Corporate Environmental Behavior and the
Effectiveness of Government Interventions
OPENING DAY
INTRODUCTORY REMARKS BY
MIKE STAHL
U.S. EPA, DIRECTOR FOR THE OFFICE OF COMPLIANCE
A WORKSHOP SPONSORED BY THE U.S. ENVIRONMENTAL PROTECTION
AGENCY'S NATIONAL CENTER FOR ENVIRONMENTAL ECONOMICS (NCEE),
NATIONAL CENTER FOR ENVIRONMENTAL RESEARCH (NCER)
April 26-27, 2004
Wyndham Washington Hotel
Washington, DC
Prepared by Alpha-Gamma Technologies, Inc.
4700 Falls of Neuse Road, Suite 350, Raleigh, NC 27609
-------
ACKNOWLEDGEMENTS
This report has been prepared by Alpha-Gamma Technologies, Inc. with funding from
the National Center for Environmental Economics (NCEE). Alpha-Gamma wishes to
thank NCEE's Cynthia Morgan and Ann Wolvertan and Project Officer Ronald Wiley for
their guidance and assistance throughout this project.
DISCLAIMER
These proceedings are being distributed in the interest of increasing public understanding
and knowledge of the issues discussed at the workshop and have been prepared
independently of the workshop. Although the proceedings have been funded in part by
the United States Environmental Protection Agency under Contract No. 68-W-01-055 to
Alpha-Gamma Technologies, Inc., the contents of this document may not necessarily
reflect the views of the Agency and no official endorsement should be inferred.
2
-------
Introductory Remarks by Mike Stahl
U.S. EPA, Director for the Office of Compliance
at the
Corporate Environmental Behavior and the
Effectiveness of Government Interventions
Washington, D.C.
April 26, 2004
3
-------
Thank you, Matt. I don't know whether it's a good thing to be given the credit for having
the vision to have started all of this—I guess that remains to be seen. I had a
management professor once who used to tell me that there's a fine line between a vision
and a hallucination, so I'm hoping that we end up with this being more of a vision than a
hallucination. I think, based on what I've read about the discussion to come today, this is
going to be a very promising and useful discussion for us.
First, I want to say that I am actually here pinch-hitting for Phyllis Harris, who is our
Deputy Assistant Administrator in the Office of Enforcement and Compliance Assurance,
who had hoped to be here but her schedule didn't permit her to come. I am only too
happy to sit in for her because my roots with this effort do go back a ways, and I'm happy
to see so many papers and so much research coming to fruition now, a little while later.
So, it's a pleasure for me to be here.
I wanted to start out by mentioning that I consider myself both a practitioner in the field
of compliance and enforcement, and something of an academician in that I am on the
adjunct faculty at George Mason University in their master's degree in public
administration. Now, that doesn't qualify me as being able to say I have a scholarly
career that I'm pursuing on the side here, but it has caused me to think and reflect from
time to time about this connection between social science research and academia
generally and people in government agencies who are trying to get certain kinds of results
and make certain things happen. I guess my observation is that having a foot sort of in
both camps, as a practitioner and as an academician, I think the connection between those
two worlds is much too faint and not nearly as strong as it should be, and I think there is a
great deal that the two groups can learn from one another. So, I'm very pleased to see
this effort, where I think we are beginning to blend folks from the social sciences and
from an interdisciplinary background with practitioners who are actually on the front
lines trying to increase compliance, protect the environment, and generally deliver
government services in a more effective way.
One of the things I wanted to lay out for you as you begin your discussions today is the
notion of "smart enforcement," and this is something that our program now at EPA is
beginning to use as an umbrella concept for several directions that we're trying to move
in simultaneously. Let me explain a little bit about smart enforcement and what it means
and what are the components of it, because I think it will bear directly on a number of the
papers that you have here today.
Essentially, smart enforcement means "using the most appropriate enforcement or
compliance tools to address the most significant problems to achieve the best outcomes
as quickly and effectively as possible." So, in order to try to carry out that notion of
smart enforcement, we have a number of components now that we have actively under
way or that we have moved toward over the last couple of years.
-------
The first of those [components] is finding and addressing significant problems. We're
doing much more to use data to try to determine what are the non-compliance patterns
that really matter to us and which ones should we be focusing federal attention on.
The second component is to use data to make strategic decisions. We've got a number of
national databases that talk about the compliance records and compliance behavior of
facilities under the major environmental statutes, but it's really only been over the last
three of four years that we have begun to use that data in much more expansive ways and
in ways that allow us to manage the program in a smarter fashion.
The next component is to use the most appropriate tool to achieve the best outcomes.
When we talk about tools in this program, we generally refer to four: The first is what
we would call "compliance assistance," which is essentially just giving information to the
regulated community to help them understand how to comply. Environmental
requirements tend to be rather complex, so I think EPA, especially over the last seven or
eight years, has put much more emphasis on trying to provide information to regulated
entities to help them understand how to get into compliance and stay there.
The second tool that we talk about in our program are incentives, and the primary
example there is our audit and self-policing policy, which provides incentives to
companies to do facility audits, find violations, disclose them to EPA, and correct them.
That policy, having been in place now for about four or five years, has led to a number of
facilities stepping forward to do their own audits and their own self-policing to try to, in
effect, get ahead of the curve and discover violations and correct them.
The next tool that I would talk about in terms of our smart enforcement approach is what
I would consider to be the more traditional compliance monitoring—this includes
inspections and investigations. Over the last several years, these have become more
sophisticated, as we have done more-in-depth investigations at particular facilities and
gone beyond just the normal onsite compliance check that our inspectors had been doing
over the many years that EPA has been in business.
The final tool is enforcement, both civil and criminal. This is when we have reached the
point where we feel that we have to take an action to correct some violation that is of
significance to us and we feel that none of the other tools can work to get the result as
effectively or as quickly as enforcement.
So, those four tools—assistance, incentives, monitoring, and enforcement—are the tools
that we like to talk about in this program as being the ones that we're now trying to mix
in the right combinations and apply to particular non-compliance patterns.
Another component of the smart enforcement approach is assessing the effectiveness of
our program. We have put a great deal of effort in the last two or three years into looking
at the performance information that we've now been collecting about EPA's enforcement
and compliance program and getting some benefit out of the analysis of that information,
in terms of recommendations about different ways to operate, adjustments that need to be
-------
made to our strategies. . . I think this notion of assessing the effectiveness is something
that you'll see EPA continuing to do in its enforcement and compliance program over the
foreseeable future.
The final element of smart enforcement is effectively communicating the outcomes of our
activities, and we have, over the last couple of years in particular, begun to talk more
about the pounds of pollution that we're reducing and less about just the number of
enforcement actions that we take in a given fiscal year. We have moved very much
toward communicating with the public in terms that they will find, I think, more valuable
and more understandable and that really speak to what it is we're trying to produce for
the environment.
So, this notion of smart enforcement, I think, is something that can serve as a bit of a
touchstone for you as you go through the day and talk about the various papers and
studies that are going to be presented here. Just based on a rather cursory review of what
you're going to be talking about today, I think many of the papers will bear directly on
our efforts to move in the direction of smart enforcement. So, I would urge that we do
more of this over time, that this link between the academic world and the world of the
practitioner be strengthened over time, and that some of the questions that practitioners
have be picked up by academia and analyzed so that we can learn more about the right
ways to operate programs. For example, I would hope that over time we can get a better
sense of what forces or what incentives motivate compliance or non-compliance in the
regulated community—what government interventions seem to be most effective in
maximizing compliance—and what tools or combination of tools are most effective
against particular patterns of non-compliance.
This is the kind of approach that we hope to be taking in EPA's Enforcement and
Compliance Assurance program over the years, and I think that the research being done
by those of you out here today who are going to be presenting papers will be very helpful
in moving us more in the direction of smart enforcement and moving us into a more
effective program over time. Let me leave you with that—I promised that I was going to
be mercifully brief so that you could get to your papers and get to the really important
discussions, and that's what I intend to do. I appreciate your participation in the
conference, and we will make use of the research that all of you have done. Thanks very
much.
-------
Corporate Environmental Behavior and the
Effectiveness of Government Interventions
PROCEEDINGS OF
SESSION I: ENFORCEMENT ISSUES
A WORKSHOP SPONSORED BY THE U.S. ENVIRONMENTAL PROTECTION
AGENCY'S NATIONAL CENTER FOR ENVIRONMENTAL ECONOMICS (NCEE),
NATIONAL CENTER FOR ENVIRONMENTAL RESEARCH (NCER)
April 26-27, 2004
Wyndham Washington Hotel
Washington, DC
Prepared by Alpha-Gamma Technologies, Inc.
4700 Falls of Neuse Road, Suite 350, Raleigh, NC 27609
-------
ACKNOWLEDGEMENTS
This report has been prepared by Alpha-Gamma Technologies, Inc. with funding from
the National Center for Environmental Economics (NCEE). Alpha-Gamma wishes to
thank NCEE's Cynthia Morgan and Ann Wolvertan and Project Officer Ronald Wiley for
their guidance and assistance throughout this project.
DISCLAIMER
These proceedings are being distributed in the interest of increasing public understanding
and knowledge of the issues discussed at the workshop and have been prepared
independently of the workshop. Although the proceedings have been funded in part by
the United States Environmental Protection Agency under Contract No. 68-W-01-055 to
Alpha-Gamma Technologies, Inc., the contents of this document may not necessarily
reflect the views of the Agency and no official endorsement should be inferred.
11
-------
TABLE OF CONTENTS
Session I: Enforcement Issues
Factors Shaping Corporate Environmental Performance: Regulatory
Pressure, Community Pressure, and Financial Status
Dietrich Earnhardt, University of Kansas 1
Deterrence and Corporate Environmental Behavior
Rovert Kagan, University of California at Berkeley, and Neil
Gunningham, Autralian National University 48
When and Why Do Plants Comply? Paper Mills in the 1980s
Wayne Gray, Clark University 70
Discussant
Nick Franco, OECA 108
Discussant
Randy Becker, U.S. Bureau of Census 114
Summary of Q&A Discussion Following Session 1 127
in
-------
Factors Shaping
Corporate Environmental Performance:
Regulatory Pressure, Community Pressure,
and Financial Status
Dietrich Earnhart *
Department of Economics
University of Kansas
May 10, 2004
Dietrich Earnhart
Department of Economics
213 Summerfield Hall
University of Kansas
Lawrence, KS 66045
785-864-2866 (phone)
Earnhart@ku. edu
Abstract: This paper analyzes the effects of external pressure - regulatory and community pressure
- on the level of environmental performance at individual polluting facilities. It considers two
dimensions of regulatory pressure: (1) specific deterrence, which is generated by actual government
interventions - namely inspections and penalties - performed at particular facilities, and (2) general
deterrence, which is generated by the threat of receiving an intervention. As important, it compares
the effects of deterrence - specific and general - based on the source of the intervention. For
inspections, it compares state and federal inspectors; for penalties, it compares EPA administrative
courts and federal civil courts. Second, the study measures community pressure indirectly using key
community characteristics (e.g., education) that proxy for actual pressure. Finally, it considers the
effects of facility- and firm-level characteristics, especially corporate financial status, on
environmental performance. For this empirical analysis, the study examines wastewater discharges
by chemical manufacturing facilities in the US for the years 1995 to 2001.
I thank Dylan Rassier for his research assistance. This manuscript was developed under a STAR Research
Assistance AgreementNo. R-82882801-0 awarded by the U.S. Environmental Protection Agency. It has not
been formally reviewed by the EPA. The views expressed in this document are solely those of Dietrich
Earnhart and the EPA does not endorse any products or commercial services mentioned in this manuscript.
1
-------
1. Introduction
Recently the Environmental Protection Agency (EPA) has been expressing a strong interest
in understanding better the factors that shape corporate environmental performance at individual
polluting facilities (hereafter "environmental performance"). In particular, the EPA wishes to assess
the effectiveness of government interventions, such as inspections and enforcement actions, for
inducing better environmental performance. This broad interest in environmental performance
echos concerns about compliance with environmental protection laws and the adequacy of
environmental enforcement previously expressed in government reports (GAO, 1983; GAO, 1995;
EPA, 1994).
To understand better these concerns and inform the EPA's more general interests, this study
analyzes two sets of external pressure factors - regulatory and community pressure - that shape the
level of environmental performance at water polluting facilities. It also considers the effects of
facility- and firm-level characteristics, especially corporate financial status, on environmental
performance. The study primarily measures environmental performance by the ratio of absolute
discharges to effluent limits - relative discharges (i.e., compliance level), which captures both
noncompliance and overcompliance. For this calculation, the study must consider specific
pollutants. To produce more generalizable results, the study focuses on two common pollutants:
biological oxygen demand (BOD) and total suspended solids (TSS).1 As a broader measure of
compliance, the study also examines the monthly count of effluent limit exceedances across all
permitted pollutants. While this latter measure is exhaustively broad, it cannot capture
overcompliance.
As the primary broad objective, this study attempts to identify the effects of certain
government interventions on environmental performance at individual facilities in the industrial
sector of chemical and allied products. The analysis considers various government interventions:
(1) state inspections, (2) EPA inspections, (3) EPA administrative penalties: fines, injunctive relief
sanctions, and supplemental environmental projects (SEPs), and (4) federal civil penalties: fines,
injunctive relief sanctions, and SEPs.2'3 Moreover, it examines the effects of these government
interventions in two dimensions. The first dimension considers specific deterrence, which captures
corporate responses to specific government interventions against particular facilities at given
1 This study also measures environmental performance by the absolute level of BOD and TSS
wastewater discharges. Analysis of these measures is available upon request.
2 For the chosen sample of facilities over the chosen sample period, no cost-recovery penalties, which
are related to remediation, are imposed by federal courts. Injunctive relief sanctions represent court-imposed
orders to perform particular a beneficial act or to stop performing a particular harmful act that relates to a
facility's operation, e.g., install a new treatment system. SEPs represent court-imposed orders to perform an
environmentally beneficial act that is not related to a facility's operation, e.g., fund an Earth Day parade.
3 Future analysis will also consider state penalties. Collection of state penalties for the entire US
would be very time consuming since no single database contains these data; instead, each state maintains its
own separate database and some states do not maintain an electronic database. The study has collected data
on penalties from the four states with the largest concentration of chemical manufacturing facilities (e.g., New
Jersey). This manuscript seeks to examine the broadest sample of facilities. Obviously, the inclusion of state
penalties would dramatically reduce the scope of the analysis.
2
-------
moments in time (Earnhart, 2004b). The second dimension considers general deterrence, which
captures the underlying "threat" of receiving an intervention (Earnhart, 2004b). To measure this
threat, this study uses indicators of interventions against other similar facilities for the relevant time
period and location (e.g., average number of federal inspections against other major chemical
facilities in each EPA region for a given year).
While the primary broad objective seeks to identify the overall effects of government
interventions on environmental performance, this study further derives six specific objectives that
either identify the main effects of government interventions or determine whether these effects differ
based on three factors: source of intervention, type of facility, and type of firm. The first specific
objective seeks to identify the effects of actual government interventions - specific deterrence - on
environmental performance. The second specific objective seeks to identify the effects of
intervention threats - general deterrence - on environmental performance. The third obj ective seeks
to compare the effects of specific and general deterrence based on the source of the intervention.
For inspections, the study compares state and federal inspectors; for penalties, it compares EPA
administrative courts and federal civil courts. EPA inspections may more greatly affect corporate
decisions than do state inspections since facilities may believe that federal involvement indicates
greater regulatory pressure. Similarly, civil penalties may more greatly affect corporate decisions
than do administrative penalties since facilities may believe that Department of Justice involvement,
which is required for civil cases, indicates greater regulatory pressure.4 The fourth objective seeks
to identify the effects of facility-level characteristics, such as type of production (based on the four-
digit SIC code) or size, on environmental performance. The fifth specific objective seeks to identify
the effects of firm-level characteristics, such as ownership structure and financial status on
environmental performance. This study examines two dimensions of financial status. The primary
dimension concerns overall financial performance, as measured by the rate of return on assets. The
secondary dimension concerns financial resources immediately available for investment in better
environmental management, as measured by annual revenues. To capture the effect of financial
status, the study must limit itself to facilities owned by publicly-held firms since financial data on
privately-held firms are not available. The sixth specific objective involving government
interventions seeks to identify the interactions between the effects of specific and general deterrence
and both facility-level and firm-level characteristics. This obj ective seeks to learn whether different
types of facilities or facilities facing different corporate conditions respond differently to
government interventions.
4 Future analysis will consider a related objective. It will seek to compare the effects of monetary
penalties (i.e., fines) and non-monetary penalties (i.e., injunctive relief, SEPs) on environmental performance.
Even though both monetary and non-monetary penalties drain corporate financial resources, they affect
corporate welfare differently. While fines provide no benefits to the firm, injunctive relief provides benefits
in the form of reduced future scrutiny, due to improved environmental management, and increased financial
payoff, whenever better environmental management is profitable. Similarly, SEPs may benefit a facility by
improving its reputation. The current manuscript does not consider this objective since few civil non-
monetary penalties were imposed on the sample of chemical manufacturing facilities during the identified
sample period, making a comparison of civil and administrative non-monetary penalties difficult to implement
properly. Future analysis will compare monetary penalties and non-monetary penalties without any
distinction between the penalties' source: administrative or civil court.
3
-------
As the secondary broad objective, this study explores the influence of local community
pressure on environmental performance. The analysis measures community pressure indirectly
using key community characteristics. These characteristics serve as proxies for pressure since they
are correlated with actual pressure (Earnhart, 2004c; Pargal and Wheeler, 1996). Specifically, this
study analyzes the influences of the following key community characteristics: (1) local labor market
condition, as measured by the unemployment rate; (2) political engagement, as measured by the
voter turnout rate; (3) political proclivity, as measured by the percent of Democratic voters; (4)
intellectual sophistication, as measured by educational attainment [proportion of residents with a
bachelor's degree]; (5) community size, as measured by the population density level; (6) community
attachment, as measured by the (6a) proportion of owner occupied households and (6b) median age;
(7) health concerns, as measured by the (7a) proportion of family households, (7b) proportion of
family households with children, and (7c) proportion of male residents; (8) wealth, as measured by
per capita income; (9) dependency on chemical manufacturing, as measured by proportion of private
earnings generated by chemical production; and (10) racial composition, as measured by proportion
of non-white residents.5 As an illustrative example, a more intellectually sophisticated (i.e., better
educated) community may be expected to mobilize its citizens more easily against and exert pressure
more effectively upon local polluters than a less sophisticated community. The study measures
community characteristics using Census data at the locale level (e.g., city) and Commerce
Department Regional Economic Information Service (REIS) data at the county level.
Since overcompliance is quite prevalent in the studied sample, the analysis is able to examine
the effects of community pressure on facilities' motivations to comply as well as to overcomply with
effluent limits. In general, each objective speaks equally to facilities' abilities and motivations to
comply with effluent limits as well as their abilities and motivations to overcomply with these same
limits. At a minimum, the objectives not related to community pressure speak to facilities' abilities
to overcomply with effluent limits. This general capacity represents a strength of the analysis.
The remainder of the paper is organized as follows. Section 2 reviews the relevant literature
and identifies the present study's contribution to this literature. Section 3 presents the empirical
application based on inspection, enforcement, and compliance data for chemical manufacturing
facilities in the US from 1995 to 2001. Section 4 presents the econometric model. Section 5
presents the estimation results. Section 6 concludes.
2. Previous Literature and Contributions of Present Study
Previous analysis on the factors shaping corporate environmental performance is limited.
Mark Cohen, the Director of the Vanderbilt Center for Environmental Management Studies, reports
that surprisingly few empirical studies of environmental enforcement have been conducted and that
they focus on a few industries: oil transport, steel mills, and pulp and paper mills (Cohen, 1999).
Jon Silberman, the Senior Attorney in the EPA Office of Enforcement and Compliance Assurance,
reaffirms the need for more empirical research (Silberman, 2000). In particular, previous economic
analysis on the effectiveness of government interventions on facility environmental performance is
5 The analysis purposively excludes the level of environmental organization membership as a
community pressure factor since it is most likely endogenously determined, especially in the case of
discharges. (Besides, local-level data are not available in any reasonably accessible form.) Instead, the
analysis relies upon more general community characteristics that might affect facility performance (Brooks
and Sethi, 1997).
4
-------
limited (Cohen, 1999). In the economics literature, few articles examine the effectiveness of
government interventions on facility environmental performance involving standard emissions (i.e.,
non-accidental discharges) and they focus exclusively on two industrial sectors - pulp/paperboard
and steel (Gray and Deily, 1996; Magat and Viscusi, 1990; Nadeau, 1997; Laplante and Rilstone,
1996; Helland, 1998a; Helland, 1998b).6 In the realm of wastewater management, previous studies
of industrial facilities examine only the former sector and consider only the effects of government
inspections. Additional studies of wastewater management examine publicly-owned wastewater
treatment plants and their responses to both inspections and penalties (Earnhart, 2004a; Earnhart,
2004b; Earnhart, 2004c). The only previous studies of penalty imposition on industrial facilities
exist in the realm of air emission management. No previous study of industrial facilities considers
specific deterrence stemming from penalties. In addition to standard emissions, a few studies
examine the effect of government interventions on oil spills (e.g., Epple and Visscher, 1984;
Anderson and Talley, 1995). Finally, two previous studies examine other dimensions of
environmental performance. Stafford (2002) examines the effect of a new EPA enforcement
protocol on facility compliance with hazardous waste regulations. May and Winter (1999) examine
compliance with agro-environmental regulations.7
This study's examination of government interventions captures deterrence in two forms:
specific and general. Previous studies on the effects of government interventions on facility
performance address the two forms of deterrence in various combinations. Some studies analyze
only specific deterrence, which stems from actual interventions at specific facilities (Magat and
Viscusi, 1990; Helland, 1998a; Helland, 1998b; Smith, 1979; Gray and Jones, 1991a; Gray and
Jones, 199 lb). Some studies analyze only general deterrence, which stems from intervention threats.
Consistent with economic theory of expected utility, this threat divides into two components: (1) the
likelihood of an intervention and (2) the size (or burden) of the intervention, conditional on its
occurrence. To capture the likelihood of an intervention, some studies use aggregate measures of
government interventions within specified locations and/or time periods (Cohen, 1987; Anderson
and Talley, 1995; Epple and Visscher, 1984; Viscusi, 1979; Bartel and Thomas, 1985). Other
studies use the predicted probability of an intervention (e.g., Gray and Shadbegian, 2000). One study
uses both likelihood measures simultaneously (Nadeau, 1997).8 No previous study directly
examines the expected conditional burden of an intervention. However, some previous studies
examine indirectly variation in the conditional burden of an intervention (e.g., Gray and Jones,
1991a). Some studies separately examine both deterrence forms by considering first actual
interventions and second predicted interventions (Laplante and Rilstone, 1996; Gray and Deily,
6 Other similar studies focus exclusively on agency behavior regarding inspections and/or
enforcement actions (e.g., Deily and Gray, 1991; Earnhart, 1997; Earnhart, 2000a; Earnhart, 2000b).
7 In addition to environmental performance, other studies explore the effects of government
interventions on performance related to worker or consumer safety regulations (e .g., Gray and Jones, 1991a;
Gray and Jones, 1991b; Olson, 1999; Viscusi, 1979; Bartel and Thomas, 1985).
8 Other studies do not focus on the likelihood of an intervention directly but instead focus on
variation in the likelihood of an intervention based on identifiable factors (e.g., Stafford, 2002; Olson, 1999;
Gray and Jones, 1991a;Viladrich-Grau and Grace, 1997).
5
-------
1996). Three studies jointly analyze the two deterrence forms: Scholz and Gray (1990), Earnhart
(2004a), and Earnhart (2004b). The first study does not consider environmental performance; the
latter two studies consider environmental performance of publicly-owned wastewater treatment
plants.
The present study is the first to examine jointly the two deterrence effects on industrial
facilities. To analyze the effects of both deterrence forms on environmental performance, this
particular empirical analysis examines a panel of data on wastewater discharges by large chemical
manufacturing facilities across the US for the years 1995 to 2001.
Drawing upon the deterrence literature, this analysis uses the noted empirical studies as a
point of departure to expand - in three other important directions - the analysis on the effects of
government interventions on corporate environmental performance. In other words, the present
study contributes to the literature in three other ways. First, it examines the distinction between
federal and state inspections and compares their effects on industrial facility performance. Second,
it examines the distinction between federal administrative and civil penalties and compares their
effects on industrial facility performance.9 Third, this study examines how different types of
facilities and firms respond differently to government interventions.
Other economic studies examine the effects of non-regulatory factors on environmental
performance and/or behavior. In particular, these studies explore the reasons for overcompliance,
which need not be explained by regulatory pressure. McClelland and Horowitz (1999) explore the
possibility of zero marginal abatement costs. Brannlund and Lofgren (1996) explore stochastic
emission patterns. Arora and Cason (1996) explore firms' desire to present a "green" image to
consumers. Downing and Kimball (1982) assess the possibility that management's concerns over
corporate image induce overcompliance.10
Community pressure may also explain overcompliance. A few economic studies explicitly
explore the effect of community pressure on environmental performance and/or behavior. Henriques
and Sadorsky (1996) explore the effect of self-reported community pressure on Canadian firms'
decisions to adopt an environmental plan. Dasgupta et al. (2000) explore the effect of self-reported
community pressure (presence versus absence) on Mexican firms' decisions to adopt certain
environmental management practices.
Other economic studies implicitly explore the effect of community pressure on environmental
performance and/or behavior by examining polluters' responses to the potential for citizen action,
which is measured by proxies for community pressure. In general, these studies rely upon
community characteristics to serve as the proxies. Maxwell et al. (2000) explore firms' desire to
9 Earnhart (2004b) examines the differential effects of government interventions on facility
performance by publicly-owned wastewater treatment plants, not industrial facilities; moreover, it does not
examine the difference between administrative and civil penalties.
10 Other studies explore the effects of non-regulatory factors on environmental performance and
behavior without addressing overcompliance. Hammit and Reuter (1988) raise the possibility of "ignorant"
compliance, while Brehm and Hamilton (1996) consider the possibility of ignorant non-compliance. Neither
study addresses overcompliance. Hamilton (1995) and Khanna et al. (1998) explore the effect of stockholder
pressure on Toxic Release Inventory (TRI) emissions. Since TRI emissions are mostly unregulated, these
two analyses address neither compliance nor overcompliance.
6
-------
preempt citizen political action for more stringent regulations at the state level; the expectation of
citizen lobbying affects facilities' decisions to reduce emissions. Hamilton (1993) examines how
hazardous waste facilities consider the potential for community action when deciding where to
locate. Pargal and Wheeler (1996) explore the effects of community characteristics on facility-level
industrial wastewater discharges in Indonesia and interpret these characteristics as capturing
community-generated "informal regulation" against facilities. Wolverton (2002) examines the
effects of community characteristics on the location decisions of Texas plants that report Toxic
Release Inventory (TRI) emissions. Becker (2002) examines whether community characteristics
help to explain the level of pollution abatement expenditures by manufacturing plants. Using a
community characteristic more tightly linked to the potential for citizen action, Konar and Cohen
(1997) explore the effect of community right-to-know laws on TRI emissions.11 Lastly, Blackman
and Bannister (1998) use a facility-specific feature — membership in a local political organization
— as a proxy for community pressure when examining the adoption of propane use by traditional
Mexican brickmakers. Similar to these previous studies, the present analysis indirectly explores the
effect of community pressure on environmental performance using proxies for actual community
pressure. In other words, while the analysis does not explicitly measure actual pressure, the effects
of community characteristics on performance should be highly suggestive of actual pressure.
By drawing upon these previous analyses, the present study contributes to the literature that
examines the effects of community pressure on corporate environmental performance in several
ways. First, it examines the effects of community pressure on compliance as measured against an
identifiable regulatory standard — permitted effluent limit — unlike all the previous studies of
corporate environmental performance. Moreover, it examines the extent of overcompliance (and
noncompliance).12 This measure of performance may better capture the effects of community
pressure since these effects may only serve to complement the effects of formal regulation, which
11 Other economic studies explore the connection between community characteristics and locally-
aggregated emissions. For example, Brooks and Sethi (1997), the most sophisticated analysis ofthese studies,
explore the relationship between zip code-level community characteristics and locally-aggregated Toxic
Release Inventory (TRI) air emissions. Brooks and Sethi (1997) catalog and describe other studies that use
simple correlations to link levels of or reductions in regionally aggregated air emissions and community
characteristics. These studies, in addition to Brooks and Sethi (1997), fail to control for other factors that may
influence emission reductions, especially regulatory factors.
12 As a matter of fact, this contribution regarding compliance levels generalizes to most studies of
environmental performance. Less than a handful of studies examine emissions relative to effluent limits
(Laplante and Rilstone, 1996; Earnhart, 2004a; Earnhart, 2004b; Earnhart, 2004c). Some studies examine
the simple distinction between compliance and noncompliance (e.g., Helland, 1998a; Nadeau, 1997; Gray
and Deily, 1996), which is too limited since it ignores the fact that many facilities overcomply with effluent
limits. [For example, McClelland and Horowitz (1999) report that aggregate emissions from pulp and paper
plants in 1992 were roughly 50 % of the permitted emissions; as another example, several firms voluntarily
reduce their emissions through participation in programs such as the EPA's 33/50 program (Aroraand Cason,
1996).] Other studies analyze absolute emission levels without reference to permitted limits (Helland, 1998b;
Magat and Viscusi, 1990), which is too limited since it ignores variation in effluent limits across facilities and
across time for a given facility. All studies using TRI emissions do not address compliance levels since these
emissions are mostly unregulated.
7
-------
may sufficiently induce compliance but not overcompliance (i.e., community pressure may mostly
affect the degree of compliance rather than the status of compliance).13 Second, this study
comprehensively incorporates government interventions and their threat. It examines separately
federal and state inspections and federal enforcement, in the realm of both specific and general
deterrence.14 Similarly, it controls for other regulatory factors, namely general permit conditions.15
Finally, the present study contributes to the environmental literature by considering financial
performance or status. In the economics literature, only one previous empirical study examines the
link from firm-level financial status to facility-level environmental performance (Gray and Deily,
1996). Other studies explore the link from firm-level financial status to firm-level environmental
performance (Konar and Cohen, 2001; Gottsman and Kessler, 1998; Earnhart and Lizal, 2002). The
present study represents only the second study of firm-level financial status to facility-level
environmental performance by linking corporate revenues and rates of return on assets to facilities'
compliance levels (i.e., relative discharges) and degree of noncompliance (i.e., monthly frequency
of effluent limit violations). In addition, this study contributes by examining how facilities facing
different corporate financial conditions respond differently to government interventions.
The results of this study generate benefits beyond these noted contributions to the literature.
First, the results should help federal and state environmental regulatory agencies to allocate
effectively their resources to achieve environmental protection. The results can provide this help
by explaining how different types of facilities or facilities in different corporate "environments"
respond differently to various influences and combinations of influences, including government
interventions and community pressure. Second, the results should help entities of the
environmentally-regulated community, chemical sector in particular, to allocate its resources
effectively to improve their compliance level and overall environmental performance in terms of
wastewater discharges. In particular, the results should help to identify which corporate
characteristics permit improvement. Since the chemical and allied products sector is a large source
of manufacturing output and wastewater discharges, the results should be strongly generalizable to
the economy as a whole and pollution control as an overall concern.
These contributions and benefits aside, this research certainly has its limitations. While the
analysis includes many influential factors on corporate environmental performance, it does not
consider several other noteworthy factors, such as criminal penalties, social norms, citizen suits,
market forces, and third-party liability claims (Cohen, 1999). Also, this research cannot claim to
identify causation, only statistically significant correlations, for the included factors. Thus, it must
qualify any claims to identifying the motivations and/or abilities behind compliant or overcompliant
13 Formal regulation may induce overcompliance when emissions are stochastic, an issue explored
by Brannlund and Lofgren (1996), as noted above.
14 The comprehensiveness of the current study stands in stark contrast to previous studies of
community pressure on corporate environmental behavior and/or performance, which do not control for
regulatory factors. As the only exception, Dasgupta et al. (2000) control for self-reported formal regulatory
presence (yes/no).
15 Earnhart (2004c) makes similar contributions for environmental performance by publicly-owned
wastewater treatment plants.
8
-------
behavior and performance.
The remaining sections use the noted literature to guide and interpret the empirical analysis
of facility-level environmental performance.
3. Empirical Application
3.1. Selection of Research Sample
To examine the effectiveness of government interventions, the influence of community
pressure, and the effect of financial status, this paper examines a specific type of environmental
performance: wastewater discharges by the 508 large ("major") chemical manufacturing facilities
across the US during the years 1995 to 2001. This selection is quite appropriate for several reasons.
First, unlike other media, regulators systematically record wastewater discharge limits, which are
critical for calculating the level of compliance (or noncompliance), and actual discharges. Second,
the EPA focuses its regulatory efforts on EPA-classified "major" facilities.16 The 508 major
facilities represented 21 % of the 2,481 chemical facilities in the National Pollutant Discharge
Elimination System (NPDES) in 2001. Moreover, they represented the bulk of wastewater
discharges from this sector. Therefore, the results from this sample of facilities are strongly
representative of the chemical industry as far as pollution control is concerned.
As the most important criterion for this sample selection, the sector of chemical and allied
products serves as an excellent vehicle for examining corporate environmental performance. [The
two-digit Standard Industrial Classification (SIC) code for this sector is 28.] Several reasons exist.
First, the EPA has demonstrated a strong interest in this sector as evidenced by its study (joint with
the Chemical Manufacturing Association [CMA]) on the root causes of noncompliance in this sector
(EPA, 1999) and its study on the compliance history for this sector [Chemical Industry National
Environmental Baseline Report 1990-1994 (EPA 305-R-96-002)]. Second, the CMA has
demonstrated a strong interest in promoting pollution reduction and prevention with its Responsible
Care initiative. Similarly, this sector is expected to display a wide variety of environmental
performance, involving noncompliance and overcompliance. Analysis of all major chemical
manufacturing facilities confirms this variety of compliance rates. For example, the mean level of
biological oxygen demand (BOD) relative discharges is 0.28, while the standard deviation is 0.34
and the range is 0 to 10.52. Similar data for TSS relative discharges confirm this assertion. The
mean level is 0.32, the standard deviation is 0.36, and the range is 0 to 9.87. Third, this sector
permits the analysis to exploit similarities and differences across the four-digit SIC sub-sectors. In
the sample used for this study, the mean level of BOD relative discharges varies dramatically across
the sub-sectors from a low of 0.09 to a high of 0.70. For TSS relative discharges, the mean level
varies from a low of 0.03 to a high of 0.57. Fourth, one of the sub-sectors, industrial organics (SIC-
code 2869), is regarded by the EPA as a priority industrial sector. Fifth, this sector is a large source
of manufacturing output and wastewater discharges. For this last reason, results should be strongly
generalizable to the economy as a whole and pollution control as an overall concern.
To retain this strong generalizability, the study focuses on two pollutants common to most
16 Major industrial facilities meet one of two criteria: (1) possess a discharge flow of 1 million
gallons per day, or (2) cause significant impact on the receiving waterbody. The EPA's Permit Compliance
System (PCS) database only systematically records wastewater discharges and effluent limits for major
facilities in the National Pollutant Discharge Elimination System (NPDES).
9
-------
regulated facilities: biological oxygen demand (BOD) and total suspended solids (TSS).17 Analysis
of both BOD and TSS appears warranted since the two measures seem to capture different
dimensions of performance based on the weak correlation - only 0.11- between the two measures.
As a broader measure of compliance, the study also examines the monthly count of effluent limit
exceedances across all permitted pollutants.
3.2. Government Regulatory Influence
This chosen sample permits analysis of government regulatory pressure. Government efforts
to control water pollution begin with the issuance of facility-specific permits. Although the EPA
possesses the authority to issue permits, this authority has been delegated to states that meet federal
criteria. Permits are issued generally on a five-year cycle. Within a five-year permit, agencies may
impose initial or interim limits, which serve as a transition to the final limits, which are generally
more stringent. In other cases, agencies may impose final limits immediately. To ensure compliance
with the permits, the EPA and state agencies periodically inspect facilities and take enforcement
actions as needed. While the EPA retains authority to monitor and sanction facilities, state agencies
are primarily responsible for monitoring and enforcement. Inspections represent the backbone of
environmental agencies' efforts to monitor compliance and collect evidence for enforcement
(Wasserman, 1984); inspections also maintain a regulatory presence (EPA, 1990).18 As for
enforcement, agencies use a mixture of informal enforcement actions (e.g., warning letters) and
formal enforcement actions (e.g., administrative orders), which include penalties.19 In particular,
EPA regional offices may initiate an administrative proceeding to impose an administrative penalty.
Alternatively, the EPA regional offices may request the Department of Justice (DOJ) to initiate a
civil court proceeding to impose a civil penalty. As likely, EPA regional offices may request the
initiation of a civil court proceeding after the imposition an administrative penalty, especially when
the administrative penalty fails to induce compliant behavior.
3.3. Data Collection
To examine the effects of regulatory pressure — inspections and enforcement, community
17 BOD and TSS are two of the five conventional pollutants (as classified by the EPA); conventional
pollutants are the focus of EPA control efforts. The EPA considers BOD the most damaging of the
conventional pollutants and the focus of their control efforts (Helland, 1998a; Magat and Viscusi, 1990).
[Conversations with federal officials confirm this point.] TSS is also damaging. All previous wastewater
studies focus exclusively on BOD. The one exception is Laplante and Rilstone (1996), who also consider
TSS. In sum, a focus on BOD and TSS discharges need not be limiting.
18 In general, federal and state inspection guidelines are minimal, according to EPA officials. As one
example, the Enforcement Management System advocates that inspections follow a systematic plan that
considers time since the last inspection and compliance history (EPA, 1990). (Further details on inspection
guidelines are available upon request.)
19 EPA policies provide only general enforcement guidelines; instead, much discretion is left to EPA
regional offices and administrative and civil courts (Lear, 1998). According to EPA officials, certain factors,
such as the economic benefit of noncompliance and compliance history, may explain the likelihood of
enforcement actions. (Further details on EPA enforcement guidelines are available upon request.) Certain
penalty types are not considered formal. The present study does not distinguish formal and non-formal
penalties.
10
-------
pressure, and financial status on the environmental performance of US chemical manufacturing
facilities, this study gathers data from various databases. The EPA Permit Compliance System
(PCS) database provides the following data elements for each chemical facility: (1) permit issuance
dates, (2) type of discharge limit [initial, interim, or final], (3) indication of changes to a permit
during the current five-year issuance period, (4) monthly wastewater flow [in millions of
gallons/day], (5) BOD and TSS monthly discharge limits, (6) BOD and TSS monthly discharges,
(7) indicator of effluent limit exceedance for each regulated pollutant, (8) four-digit SIC code, and
(9) location.
Further discussion on discharge measurements, limits, and limit exceedances is needed.
First, facilities monitor and facility-specific effluent limits restrict discharges according to two
pollution measures: monthly average and monthly maximum. Conversations with government
officials and the EPA's definition of significant noncompliance, however, suggest that regulators
especially care about the average limit (GAO, 1996). Thus, this study focuses on the average
discharge and limit. Second, facilities may monitor and facility-specific effluent limits may restrict
only quantities (e.g., kilograms of BOD), only concentrations (e.g., milligrams of BOD per liter of
water), or both. By focusing on compliance levels, the study is able to compare across all facilities
regardless of the form of their discharge measurement and effluent limit. The analysis calculates
relative discharges - the ratio of absolute discharges and effluent limits - regardless of the type of
discharge and limit. If both quantity and concentration limits apply, the analysis calculates the mean
level of compliance. Third, each facility may have several points of discharge and several sources
of wastewater generation. For each combination of discharge point and generation source, the
analysis identifies the relevant discharge level and effluent limit and then calculates the level of
relative discharges. In order to generate a single observation for each specific facility at a particular
moment in time, the analysis calculates the mean relative discharge level across all multiple
combinations of points and sources. In this way, the data on environmental performance match with
the facility-level data, especially the information on government interventions.20 Fourth, the
monthly count of effluent limit exceedances across all regulated pollutants is calculated in a similar
fashion by summing across all multiple combinations of points and sources. Fifth, a given facility
may not discharge any pollution in a specific month. If true, BOD and/or TSS discharges are
recorded as zero.
The PCS database also provides data on inspections performed by federal and state
regulators. Both the PCS database and the EPA Docket database provide data on federal penalties
imposed by EPA administrative courts. However, only the EPA Docket database provides data on
federal penalties imposed by civil courts. Penalties represent the sum of three penalty components:
monetary fines, value of injunctive relief, and value of SEP. (For the chosen sample of facilities and
study period, cost-recovery penalties, which are related to remediation, are not imposed.)
Accordingly, the study integrates the two databases, while using the Docket database to identify civil
penalties.
20 For facilities with multiple point/source combinations, the analysis also calculates the maximum
level of compliance. Similarly, the analysis also calculates and examines the maximum level if both
quantities and concentrations are measured and restricted in the same month for a particular facility.
Preliminary estimation of these maximum compliance levels generates results similar to those reported for
the average level of compliance.
11
-------
The U.S. Census Bureau provides information on natural resource-related budgets for local
and state agencies.21 Since all EPA activities are related to natural resources, this study utilizes more
specific budgetary information on the Enforcement and Compliance Assistance program within the
EPA. However, this information is available only for the EPA regional offices.22 For the central
EPA office, this study uses simply the entire agency budget, as provided by the Office of
Management and Budget. The National Council of State Governments provides data on the number
of business establishments located in a given state.
Two sources provide data on community characteristics. The U. S. Census database provides
data on certain community characteristics at the locale level for 1990 and 2000. The study translates
these decennial data into annual data by interpolating between the two endpoints, except for the year
2001, which utilizes data for the year 2000. The Commerce Department Regional Economic
Information Service (REIS) database provides data on certain community characteristics at the
county level on an annual basis.23 The specific community characteristics are as follows: (1) voter
turnout rate and Democratic voting percentages in available presidential elections, (2) proportion
of residents with a bachelor's degree, (3) income per capita, (4) proportion of owner occupied
households, (5) unemployment rate, (6) population density, (7) median age, (8) proportion of family
households, (9) conditional proportion of family households with children, (10) proportion of non-
white residents, (11) proportion of private earnings generated by chemical manufacturing, and (12)
proportion of male residents.
The EPA Toxic Release Inventory (TRI) database provides information on a facility's parent
company. The Business and Company Resource Center database provides data on a parent
company' s ownership structure: privately-held or publicly-held. The Compustat / Research Insight
database provides annual financial data on publicly-held firms.24 (Future analysis will additionally
21 While consideration of all natural resources may be too wide, data on water pollution control
expenditures is not readily available. While the National Council of State Governments provides information
on water quality-related budgets for local and state agencies, it is available only for one year - 1996 - of the
sample period. (Results generated using this alternative measure are available upon request.) Also, data on
state and local natural resource-related budgets are available only for the years 1995 to 1999. The study
extrapolates these data to cover the years 2000 and 2001.
22 EPA regional data exist only for the years 1998 to 2002; the study backward extrapolates these
data to cover the years 1995 to 1997.
23 Thus, the analysis also considers the county as a relevant scale for identifying a "community". This
scale arguably also captures an appropriate population whose utility is affected by local water quality that is
influenced by a sampled facility's discharges. A smaller scale, such as locale, is certainly useful. However,
it may omit people whose utility is affected by local water quality, especially since each facility is a major
polluter. A larger scale, such as state regulatory district, would probably include water quality unaffected by
the local facility.
24 In certain cases, the TRI database does not provide data on a facility's parent company for a
specific year. The study is still able to identify the parent company in most cases using additional data
available in either the PCS or TRI database. As the most useful method, the study uses the parent company
from the preceding and succeeding years if the name remains the same. If no parent company name is
12
-------
consider quarterly financial data.)
All dollar-denominated values are deflated to 1995 levels using the Consumer Price Index.
This study considers different sub-samples when examining different measures of
performance and different sets of explanatory factors. First, it considers three types of performance:
BOD discharges, TSS discharges, and monthly frequency of effluent limit exceedances across all
regulated pollutants. The sample for monthly effluent limit exceedances includes all maj or chemical
facilities for all months across the entire sample period. This broad sample includes 508 facilities
that were active at some point over the sample period: January, 1995, to June, 2001.25 Of these 508
facilities, 456 were active throughout the entire sample period. In contrast, 25 facilities entered the
sample at some point after January, 1995, while 27 facilities exited the sample at some point before
June, 2001.26 Although technically possible, no facility is ever temporarily inactive; instead, each
exiting facility remains permanently inactive. By including all ever active facilities, the analysis
greatly minimizes any survivor bias. Of course, the study cannot eliminate this bias since it must
select some starting point. However, any survivor bias is expected to be small since very few
facilities exit the sample: attrition represents only 5 % of the overall sample over a relatively long
6.5-year period.
The sub-samples for BOD and TSS discharges are smaller. Even though most major
reported within the TRI database, the study uses the facility name to match with the Business and Company
Resource Center database and Compustat / Research Insight database. The study assumes that a facility name
is sufficient to identify a publicly-held firm. Thus, if neither of the databases indicates publicly-held
ownership structure, the facility is assumed to be owned by a privately-held firm. In certain cases, the
Business and Company Resource Center database does not provide data on ownership structure. For these
cases, the study uses the Compustat database to identify ownership structure. By default, the company is
publicly-held if found in the Compustat database, and privately-held if not found. Finally, while the TRI and
Compustat databases provide annual data for the entire sample period, the Business and Company Resource
Center database provides data only starting in 2001. Nevertheless, the study is able to identify changes in
ownership structure based on the Compustat database, given the assumption that the Compustat database
contains all publicly-held firms. Fortunately, the Business and Company Resource Center database generally
indicates changes in ownership during the sample period (1995 to 2001). This history permits the study to
search for changes in ownership structure using the annual data reported within the Compustat database.
Without this historical information, the study would need to search the Compustat database for each firm and
for all years prior to 2001.
25 The study does apply a few other criteria for inclusion in the sample. Specifically, the study
excludes particular types of discharge and certain types of facilities. First, it excludes discharges reported
on a non-monthly basis. Without this restriction, it would be very difficult to compare across facilities. This
restriction eliminates few relevant observations since practically all major facilities facing effluent limits
report their discharges monthly. Second, the study excludes bio-solid (i.e., sludge) discharges. Third, the
study excludes industrial users, i.e., facilities that discharge into pre-treatment programs run by publicly-
owned treatment works. This restriction eliminates only three major facilities. Together, the latter two
restrictions indicate the study's focus on direct discharges into surface water bodies.
26 The PCS database does not indicate the date of activation. Instead, it indicates only the date of
inactivation. Nevertheless, the study identifies the apparent activation date based on the presence of DMR
records. Details on this identification are available upon request.
13
-------
chemical facilities discharge both BOD and TSS, several discharge only one or neither. Therefore,
this study considers two separate sub-samples: one for BOD and one for TSS. To remain in each
sub-sample, a given facility must discharge the particular pollutant at least once during the seven-
year sample period. Based on this restriction, the BOD sub-sample contains 380 facilities and the
TSS sub-sample contains 461 facilities.27 Moreover, not all facilities discharging either BOD or TSS
(or both) possess a permit that imposes effluent limits on these specific pollutants. Given the focus
on compliance level as a measure of environmental performance, to remain in each sub-sample, a
given facility must face an effluent limit for the relevant pollutant in the particular month of
discharge. This restriction eliminates 1,832 observations from the BOD sample, dropping its size
from 26,172 to 24,340. The same restriction eliminates 3,152 observations from the TSS sample,
dropping its size from 32,378 to 29,226.28
This study also considers different sub-samples when examining different sets of explanatory
factors. It considers all major facilities, when excluding financial status as an explanatory factor,
and only major facilities owned by publicly-held firms, when including financial status as an
explanatory factor. The second set of facilities represents 63 % of the overall sample.
Section 4 structures the econometric analysis of these collected data, including the creation
of measures to capture deterrence. It also interprets the statistical summary of the collected and
formatted data. Section 5 displays the analytical results.
4. Econometric Approach
4.1. Regression Framework
This paper analyzes the effectiveness of government interventions and community pressure
for inducing better environmental performance. To analyze these effects, consider the following
notation. Let YitJ represent the level of environmental performance type j for facility i in time period
t, where ye {BOD, TSS, ALL}, BOD represents BOD relative discharges, TSS represents TSS
relative discharges, and ALL represents the monthly frequency of effluent limit exceedances across
all regulated pollutants.29 This performance level depends on several explanatory variables. With
only a few exceptions, which are noted where relevant, this set does not vary across the three types
of performance: je {BOD,TSS, ALL}. Therefore, the notation for the explanatory variables does not
include the superscript j.
To estimate the effects of government interventions on environmental performance, the
27 Most facilities discharge both BOD and TSS (N=389). Some discharge only TSS (N=86). Very
few discharge only BOD (N=5). And few discharge neither (N=42). Further examination of these various
sub-samples is available upon request. Results of a comparison between facilities that rarely discharge a
specific pollutant and facilities that almost always discharge is also available upon request.
28 The PCS database does not provide a record for each month of a facility's existence. The analysis
assumes that no missing record includes an operative effluent limit. This assumption is unlikely to generate
a selection bias since the absence of a record is driven by poor recordkeeping according to EPA officials.
29 For BOD and TSS discharges, preliminary analysis also estimates absolute discharge levels and
the qualitative state of noncompliance versus compliance using a Probit model (Maddala, 1983). These
results are available upon request. The study focuses on the compliance level (i.e., relative discharges) as the
primary measure of environmental performance since it is the most comprehensive indicator and captures
overcompliance, which is very prevalent in the sample.
14
-------
analysis must first sort out deterrence. One form of deterrence - specific deterrence - stems from
actual interventions at specific facilities. Facilities may be able to respond to actual interventions
within the same month of the intervention. In this case, performance and interventions would be
simultaneously determined. However, facilities most likely need at least a few weeks, if not several
months, to respond to interventions (Magat and Viscusi, 1990; Earnhart, 2004b). Accordingly, the
analysis uses lagged, not current, values of interventions as regressors. In the case of inspections,
the analysis generates the cumulative count of inspections performed by the state at a specific
facility in the preceding 12-month period, denoted as Iit_i2ST, and generates the similar cumulative
count of inspections performed by the EPA, denoted as Iit_i2EPA In the case of enforcement, the
analysis generates the cumulative count of EPA administrative penalties and conditional mean
administrative penalty magnitude imposed against a specific facility in the preceding 12-month
period, collectively denoted as Pit.i2ADM, and generates the cumulative count of federal civil penalties
and conditional mean civil penalty magnitude, collectively denoted asPit-i2CIV30
By using lagged, not current interventions as regressors, the analysis implicitly claims that
performance and interventions are not simultaneously determined.31 To buttress this claim, the study
considers the determination of interventions. While current interventions may depend on current
performance, it is highly doubtful that agencies are cognizant of a facility' s performance in the very
month chosen for an actual intervention. Agencies more likely base their intervention decisions on
30 This construction needs elaboration. First, the study chose a period of 12 months for various
reasons: (1) major polluters should be inspected once per year, (2) previous studies, such as Laplante and
Rilstone (1996) and Earnhart (2004a,b), examine a 12-month period of lagged interventions, and (3)
preliminary analysis indicates that other time periods [e.g., 6 and 24 months] generate less significant results.
Second, the chosen approach of accumulating interventions is more consistent with reality than the alternative
approach of including multiple monthly indicators of lagged interventions (e.g., Magat and Viscusi, 1990).
According to EPA officials, regulatory agencies generally induce better performance by repeatedly inspecting
polluters. As for enforcement, penalties are sufficiently uncommon as not to warrant multiple indicators.
Nevertheless, it seems helpful to accumulate administrative penalties over a 12-month period since
administrative penalties appear to be imposed over the course of a time period longer than a month. On
average, the number of penalties over a 12-month period is 10 times greater than the number in a single
month. In contrast, civil penalties do not accumulate over a 12-month period. At the most, only a single civil
penalty is imposed over a 12-month period. Thus, the civil penalty specific deterrence variable serves more
as an indicator variable. Moreover, the conditional mean civil penalty magnitude equals the sum of civil
penalties for the same period. In this way, the analysis can explicitly interpret the mean magnitude as an
interaction between the penalty indicator and penalty sum. The chosen approach of cumulative interventions
also retains the explanatory power of potentially multiple inspections within one regressor rather than
dissipating the explanatory power across several regressors. The same dissipation of explanatory power may
apply to penalties. Nevertheless, future analysis should explore the use of multiple monthly indicators since
this approach permits the testing of whether the effects of specific deterrence are persistent (Laplante and
Rilstone, 1996).
31 Within an instrumental variables approach for resolving any potential simultaneity between
performance and interventions, lagged interventions serve as highly proper instrumental variables for current
interventions since lagged interventions are certainly exogenous with respect to current performance
(Laplante and Rilstone, 1996; Magat and Viscusi, 1990). Thus, the assumed connection between lagged
interventions and current performance need not be troubling.
15
-------
past performance since they need time to evaluate performance before responding to it (Magat and
Viscusi, 1990). In this case, again, performance and actual interventions are not simultaneously
determined. Instead, lagged performance is pre-determined relative to current interventions.
The other form of deterrence - general deterrence - stems from the threat of an intervention.
As noted above, the threat divides into its two constituent components: likelihood and conditional
burden. Similar to most previous studies of inspections, the analysis assumes that the burden of each
inspection does not vary across the facilities (e.g., Earnhart, 2004b; Laplante and Rilstone, 1996;
Gray and Deily, 1996; Nadeau, 1997). [Only Helland (1998b) differentiates according to the type
of inspection (e.g., performance audit versus compliance evaluation).] Instead, the analysis focuses
exclusively on the likelihood of an inspection. The analysis denotes the likelihood of an EPA
inspection and a state inspection as ILitEPA and ILitST, respectively. Unlike similar studies, the
analysis allows the conditional burden of each penalty to vary across the facilities (e.g., Earnhart,
2004b; Gray and Deily, 1996; Nadeau, 1997). Thus, the present study considers both components
of enforcement-based general deterrence: likelihood and conditional burden. The analysis denotes
the likelihood of an EPA administrative penalty as PLitADM and federal civil penalty as PLitCIV. To
capture inspection and penalty likelihoods, the analysis employs a pair of proxies based on the
annual aggregate measure of interventions against other similar facilities - major chemical facilities
- in the same relevant location (e.g., state) and same time period (Earnhart, 2004b; Nadeau, 1997).32
One proxy captures the inspection likelihood; the other captures the penalty likelihood. This
approach of considering other facilities keeps separate the two deterrence forms. To adjust for
differences in the number of major chemical facilities across states or EPA regions and across time,
the analysis divides each aggregate count of interventions by the number of other major chemical
facilities in each state or EPA region of the given year. When examining the threat of enforcement,
the analysis captures the conditional burden component of general deterrence using the conditional
mean penalty magnitude imposed against other major chemical facilities in the same EPA region.
These conditional mean magnitudes are denoted as PMitADM and PMitCIV for EPA administrative and
federal civil penalties, respectively. Since the mean penalty magnitude is conditional on the
imposition of a penalty, no adjustment for the number of major chemical facilities is needed.
These constructed general deterrence measures imply a particular way of understanding a
facility's expectations about future regulatory pressure. As constructed, each facility gauges its
expectation of monitoring and enforcement based on the observed experience of other similar
facilities. By considering annual aggregate measures, the analysis assumes that each facility has
fully rational, forward-looking expectations: it perfectly estimates the amount of regulatory pressure
over an entire year at the beginning of each year and retains this expectation throughout the year.
32 Conversations with EPA officials confirm that aggregate measures of interventions properly proxy
the likelihood of an intervention. (They also confirm the expectation that increased likelihoods prompt better
facility performance.) Nevertheless, this approach assumes that the likelihood is generic to all similar
facilities. Future analysis will attempt to refine the determination of "similar facilities" by expanding the
dimensions used to define "similar". Currently, the analysis considers only two-digit SIC code, EPA
classification ("major"), location (e.g., state), and time period (i.e., current year). In the case of inspections,
the dimension of EPA classification is quite important since the frequency of inspections is dramatically
greater at major facilities than at minor facilities, due to a federal guideline to inspect major facilities at least
once annually (EPA, 1990).
16
-------
Certainly, other perspectives on general deterrence expectations exist. Preliminary analysis
indicates that use of backward-looking expectations that are updated annually generate similar or
worse estimation results. Use of monthly-updated measures are probably overly sensitive to
monthly variations in monitoring and enforcement events. Future analysis will consider a 12-month
moving window of historical and/or future interventions against other similar facilities (e.g., 6
historical months and 6 future months).
These general deterrence measures should not depend on the particular facility' s performance
since the interventions are imposed against other facilities. Instead, these interventions should
depend on other facilities' performance levels. In addition, it is highly doubtful that one facility's
performance depends on other facilities' performance. (Of course, all facilities' performance may
depend upon common factors, such as seasons (e.g., treatment may be more difficult in cold
weather). As a matter of fact, the general deterrence proxies rely upon factors that are common to
all similar facilities. These common factors capture exogenous elements of regulatory pressure:
exogenous variation in regulatory pressure across regions / states and time.
In addition to these deterrence measures, other regulatory factors may affect the level of
environmental performance. First, the analysis captures variation in regulatory pressure not
reflected in the specific and general deterrence measures by including three regressors that
separately measure annual budgetary resources expended by state and local agencies (by state), EPA
regional offices (by region), and the EPA federal office (for the entire US). Each budgetary measure
is adjusted by the number of establishments in each state, region, and country, respectively, for the
relevant year (Helland, 1998a). The analysis also includes EPA regional indicators. Second, the
analysis includes facility-specific NPDES permit conditions as regressors, which collectively
capture certain dimensions of regulatory stringency:
(1) permitted effluent limit level (in pounds/day);
(2) limit type: interim versus final;
(3) magnitude of expiration (in days);
(4) indicator for any modification(s) to NPDES permit after issuance.
For comparability, the analysis converts each concentration limit to a quantity limit using the
facility' s reported flow of wastewater for the specific month.33 Limits vary across facilities and time
due to variation in effluent guidelines across sub-sectors, seasonal variation for facilities located on
certain waterways, and use of water-quality-based standards. To control for seasonal variation, the
analysis also includes a set of season indicators. Let Git collectively denote these additional
regulatory conditions.
33 Generation of this regressor demands elaboration. First, some facilities have multiple points of
discharge and/or sources of wastewater generation. For each combination of discharge point and generation
source, the analysis identifies the relevant effluent limit and wastewater flow level, converts any concentration
limit to a quantity limit using the relevant flow rate, and finally calculates the mean effluent limit across all
multiple combinations of points and sources. This approach generates a single observation for each specific
facility at a particular moment in time. Second, in certain cases, no monthly measurement of wastewater flow
is available. Rather than dropping these observations, the analysis imputes a replacement value based on the
following hierarchy depending on data availability: (1) facility-specific annual average flow, (2) facility-
specific sample average flow, and (3) sample-wide average flow. This imputation affects only 0.04 % of the
TSS sample and 0.9 % of the BOD sample.
17
-------
In addition to regulatory pressure, community pressure may also affect corporate
environmental performance. The analysis measures community pressure indirectly using the
following key community characteristics:
(1) local labor market condition, as measured by the unemployment rate,
(2) political engagement, as measured by the voter turnout rate;
(3) political proclivity, as measured by the percent of Democratic voters in Presidential
elections;
(4) intellectual sophistication (or educational attainment), as measured by the proportion of
residents with at least a bachelor's degree;
(5) community size, as measured by the population density level;
(6) community attachment, as measured by these two characteristics:
(a) the proportion of owner occupied households, and
(b) median age;
(7) health concerns, as measured by these three characteristics:
(a) proportion of family households,
(b) proportion of family households with children, and
(c) proportion of male residents;
(8) wealth, as measured by per capita income;
(9) dependency on chemical manufacturing, as measured by proportion of private earnings
generated by chemical production;34 and
(10) racial composition, as measured by proportion of non-white residents.
Let Cit collectively denote these community characteristics.
Exploration of the connection between community characteristics and wastewater discharges
may not capture properly the effect of community pressure because both facility and household
location decisions potentially generate endogeneity problems. First, a firm that wants to build a new
facility is more likely to choose a location that is more receptive to high pollution facilities (i.e.,
lower expected community pressure); this receptivity may be correlated with identifiable
socioeconomic factors. Second, people who choose to live in a neighborhood near an existing
polluter are more likely to have a higher tolerance for pollution. Similarly, once a polluter has
located at a specific site, lowered property values may prompt individual households to leave or
enter the affected community. Again, this tolerance and the re-location choices may be correlated
with identifiable socioeconomic factors. In general, it is difficult to avoid these endogeneity
34 The REIS database does not provide data on private earnings generated by chemical manufacturing
when these data would permit the identification of individual facilities. Rather than omitting these
observations lacking data, thus introducing a potentially strong bias, the analysis imputes replacement values
according to the following hierarchy based on availability: (1) facility-specific mean over the entire sample
period, (2) state-wide mean for the relevant year, i.e., state within which the facility resides, and (3) sample-
wide mean for the relevant year. This imputation affects roughly 20 % of the sample. However, the
imputation rarely draws upon the sample-wide mean (<0.1 % of the sample). Instead, 7 % of the sample uses
imputed values based on facility-specific means and 14 % of the sample uses imputed values based on state-
year specific means. The former imputation serves as a good proxy if chemical production for a given locale
varies little overtime. The latter imputation serves as good proxy if chemical production varies little across
space within a given state.
18
-------
concerns. Nevertheless, the econometric analysis attempts to mute these concerns by employing a
fixed effects model when estimating the panel data of environmental performance. In this way, the
analysis controls for inherently "dirty" or "clean" facilities, reducing any potential omitted variable
biases associated with the effects of community characteristics.35
The level of environmental performance also depends on factors besides external pressure.
In particular, it depends on firm-level characteristics: (1) financial status, as measured by annual
revenues and the rate of return on assets, which represents the ratio of net income to total assets; and
(2) ownership structure indicators: privately-held and publicly-held. When examining the link from
financial status to environmental performance, the analysis avoids using current financial status,
since contemporaneous financial status and environmental performance are most likely jointly
determined. Instead, the analysis uses lagged financial status, which is considered as predetermined
(Lizal and Svejnar, 2002a,b; Earnhart and Lizal, 2003). Thus, lagging financial status avoids any
endogeneity problem (Austin et al., 1999). Moreover, one would expect a lag between the
generation of financial resources and the ability to invest in ways of reducing wastewater discharges.
Similar to firm-level characteristics, environmental performance most likely depends on
facility-level characteristics:
(1) flow capacity, as measured by the average flow of wastewater over the preceding 12-
month period (millions of gallons / day);36
(2) marginal compliance costs, as proxied by the ratio of actual wastewater flow to flow
capacity (Helland, 1998a);
(3) stochasticity of wastewater discharges, as measured by the standard deviation of BOD
or TSS relative emissions over a current calendar year;37 and
(4) industrial sub-sector indicators (Table l.b provides a full listing).
According to Brannland and Lofgren (1996), as discharge variability rises, facilities may choose to
increase their compliance level (i.e., decrease level of relative emissions). LetFit collectively denote
35 Future analysis will attempt to avoid this endogeneity concern by estimating the effects of current
community characteristic levels on subsequent changes in performance levels, e.g., the effect of 1995
community characteristic levels on the change in performance levels between 1995 and 1996 (Brooks and
Sethi, 1997). This future analysis will consider several starting points (e.g., 1995, 1996, 1997) and several
time frames for calculating performance changes (e.g., one-year change between 1995 and 1996, two-year
change between 1995 and 1997).
36 In certain cases, no monthly measurement of wastewater flow is available. Rather than dropping
these observations, the analysis imputes a replacement value based on the following hierarchy depending on
data availability: (1) facility-specific annual average flow, (2) facility-specific sample average flow, and (3)
sample-wide average flow. This imputation affects less than 3 % of the sample. As a check for robustness,
analysis estimates only those observations with available data on wastewater flow. The estimation results
are highly similar to the reported results.
37 Preliminary analysis also uses standard deviations of absolute discharge levels to measure
stochasticity. The estimation results are roughly similar to those reported.
19
-------
these firm- and facility-level characteristics.38
Lastly, the analysis interacts the various measures of specific and general deterrence with
the firm- and facility-specific regressors. These interactions help to indicate whether different types
of facilities or facilities facing different corporate conditions respond differently to government
interventions. Let Xjt collectively denote these interactions.
The following regression equation captures the functional relationship between
environmental performance and the noted explanatory variables, especially regulatory and
community pressure:
f(Yitj)=pEPAIit.12EPA+ pSTIit.12ST + pADMp^^ADM + Pavlov + QEPA EPA + qST jj^ST +
yADMPLitADM + \|/CIVPLitCIV + y^PM^ + \|/CIVPMitCIV + r|GGit + r|cCit + r|FFit + r|xXit
+o\t + SYitj, (1)
where sYitj represents the error term and V represents the inverse Mills ratio associated with BOD
and TSS relative emissions [/e{BOD,TSS}], which is defined in the immediately following
paragraph. When estimating BOD and TSS relative emissions (i.e., y=BOD,TSS), the analysis
employs a semilog specification: f(YuJ)=ln(YitJ)39, When estimating the monthly frequency of
effluent limit exceedances (i.e.,j=ALL), the analysis employs a linear specification: f(YitJ)=YitJ,
Before estimating environmental performance, the econometric analysis must first address
the fact that facilities do not always submit discharge monitoring reports with measured discharges,
even though federal regulations require their monthly submission. This concern does not apply to
effluent limit exceedances since only a handful of observations indicate the failure to submit a
discharge monitoring report with information on limit exceedances. From the BOD sample, 225 of
the 24,340 observations lack data on measured discharges; from the TSS sample, 252 of the 29,226
observations lack data on measured discharges. Thus, any bias introduced by the failure to report
38 Previous studies of environmental performance explore two other characteristics. First, some
previous studies using panel data include the lagged dependent variable as a regressor (e.g., Earnhart, 2004b).
This regressor may capture potential inertia in the treatment process. This inertia most likely stems from the
use of fixed control equipment, whose installation generally requires time (Laplante and Rilstone, 1996).
Consequently, the regressor may provide information on the facility's stock of pollution control capital and
the general character of its abatement technology (Magat and Viscusi, 1990). However, inclusion of the
lagged dependent variable as a regressor greatly complicates the use of panel data models. Fortunately,
inclusion of facility-specific constants in the fixed effects model may more adequately control for the general
character of a facility's abatement technology if it varies little over time. Second, some previous studies
include the production price index for the identified sector, chemical manufacturing in this case, as a regressor
that attempts to control for variation in the opportunity cost of any production reductions prompted by efforts
to improve environmental performance (Shimshack and Ward, 2003; Helland, 1998a). The study has
obtained this information and generated this regressor. Future analysis will include this regressor.
39 This paper also estimates a linear specification for BOD and TSS relative discharges. Based on
a goodness-of-fit measure - adjusted R2 - and the prevalence of significant coefficients, the analysis focuses
on the semilog specification as the better model. The use of log values for the dependent variable also
minimizes the effect of outliers (Gray and Deily, 1996; Earnhart, 2004b).
20
-------
discharge measurements may be quite small.40 To address the non-reporting of discharges data, the
study uses a Heckman correction procedure to adjust for any potential sample selection bias
(Heckman, 1979; Earnhart, 2004b). As the first step in this procedure, the analysis estimates a
probit model of the facility's decisions to report monthly discharges. Let Rjt indicate the decision
of facility i to report discharges in time period t. Let Kit indicate the set of explanatory variables.
Equation (2) captures this reporting relationship:
R^SK. + s^, (2)
where eRU represents the error term for equation (2). This estimation generates useful results. For
BOD reporting, roughly two-thirds of the slope coefficients are statistically significant at the 10 %
level; in particular, flow capacity and general deterrence strongly affect the BOD reporting decision.
For TSS reporting, roughly three-fourths of the slope coefficients are statistically significant at the
10 % level; in particular, industrial sector classification and state inspection-related specific
deterrence strongly affect the TSS reporting decision. [Further details on this estimation are
available upon request.41] As the second step of this procedure, the analysis uses the estimated
probit coefficients and associated variables to generate an inverse Mills ratio, \t, for each
observation with reported emissions. This ratio serves as the correction term for sample selection
in the third step of the procedure, which involves estimation of reported relative discharges, shown
in equation (1). The inverse Mills ratios are computed for BOD and TSS discharges and included
as regressors in the environmental performance equations for BOD and TSS relative discharges.
The study estimates the three performance equations using the following three econometric
regression models: pooled ordinary least squares (OLS), fixed effects, and random effects (Hsiao,
1986). The latter two models are standard panel data models. Each specific panel data model stems
from a more general model that captures differences across the various polluters by incorporating
an individual term for each facility. If this facility-specific term is uncorrected with the other
regressors in equation (1), then the random effects model is appropriate. The random effects model
captures differences across the various polluters by including a random disturbance term that
remains constant through time and captures the effects of excluded factors specific to each facility.
If the facility-specific term is correlated with the other regressors in equation (1), then the fixed
effects model is appropriate. The fixed effects model captures differences across the various
polluters by estimating an individual constant term for each polluter. (Note that use of this model
eliminates the ability to estimate a coefficient for any time-invariant regressor; the analysis considers
two time-invariant regressors, EPA region and industrial sub-sector; nevertheless, the fixed effects
40 Self-monitoring is the most important source of information utilized by state and federal regulators
to assess environmental performance (EPA, 1990). Although facilities may have incentives to under-report
emissions, stiff sanctions for false reports, including incarceration (Shimshack and Ward, 2003) and periodic
inspections provide countervailing incentives to report honestly (Magat and Viscusi, 1990).
41 The analysis uses a two-stage estimation process for estimating the reporting decision and
performance levels. The nonlinearity of the probit model is sufficient for identifying the two related
equations (Greene, 1997). Nevertheless, to help identify these two related equations, the probit equation for
the reporting decision excludes certain variables that relate to performance and includes certain variables not
related to performance (e.g., preceding 12-month average of BOD mass loadings). A likelihood ratio test
statistic confirms that the excluded variables are jointly significant only at levels greater than 10 %.
21
-------
model indirectly captures the effect of industrial sub-sector when it is interacted with deterrence
measures.) The analysis uses an F-test of fixed effects to discern whether the fixed effects model
dominates the pooled OLS model, i.e, the F-test rejects the null hypothesis of no fixed effects. The
analysis uses the Hausman test of random effects to evaluate whether the estimation can use the
more efficient random estimates or whether these estimates are inconsistent when compared to the
fixed effects. When the Hausman test signals that the random effects estimator is consistent with
the fixed effects estimator, the random effects estimator is preferable since it is more efficient by
construction.42 Unlike the BOD- and TSS-related performance measures, the monthly frequency of
effluent limit exceedances need not represent a continuous variable; instead, this measure represents
integer or count data. Accordingly, the study also estimates this third performance equation using
a count data model, namely the Poisson model (Greene, 1997).43 Since least squares regression
generates consistent results from count data (Greene, 1997) and the use of a Poisson model generates
estimation results sufficiently similar to the reported results, the current paper does not provide the
Poisson results. Moreover, attempts to adjust for the panel data structure by incorporating fixed
effects into the Poisson model did not generate convergence. Future analysis will more strongly
focus on the estimation of these count data. Lastly, estimation of this third performance measure
omits certain regressors since they are not available or relevant for effluent limit exceedances: (1)
permitted effluent limit level and (2) stochasticity of relative wastewater discharge level. Besides
effluent limit, other permit conditions - limit type, expiration, and modifications - apply equally to
all regulated pollutants with minor exceptions.
4.2. Statistical Summary of Regression Variables
Table 1 provides statistical summaries of the formulated dependent variables and regressors.
These summaries draw upon the samples used for the regression analysis. First, Table l.a
summarizes the environmental performance measures. Facilities on average exceed 0.31 of their
limits in a given month. Consistent with this small average, facilities do not exceed a single limit
in 79 % of the months (not shown in Table 1 .a). Facilities on average generate BOD discharges that
are 82 % below their BOD monthly limit. This figure indicates a need to analyze the degree of
compliance rather than the status of compliance. At the other end, BOD discharges surge as high
as 952 % above the permitted limits. This figure indicates a need to analyze the degree of
noncompliance rather than the status of noncompliance. The comparable figures lead to the same
two conclusions: on average, TSS discharges are 78 % below permitted limits, yet they surge as high
42 Future analysis will jointly estimate BOD- and TSS-related environmental performance using a
seemingly unrelated regression (SUR) approach, which improves the efficiency of the coefficient estimates
and permits proper testing of differences between BOD- and TSS-related coefficients (Greene, 1997). The
current paper does not provide this SUR estimation since implementation requires a sub-sample restricted to
observations with both BOD and TSS relative discharges. The current paper seeks to examine the broadest
set of facilities.
43 Preliminary analysis also attempts to use a negative binomial model with limited success. Future
analysis will refine the use of this alternative model.
22
-------
as 887 % above the permitted limits.44
Second, Table l.b summarizes the regressors common to all measures of environmental
performance, while excluding financial-related regressors. This summary includes information on
inspections and penalties.45 It also includes information on community characteristics. The average
community contains about 685 people per square mile, provides at least a bachelor' degree to about
18 % of its residents, voted for the democratic presidential candidate at a 47 % rate, enjoys nearly
$ 22,600 in income per person, and endures a 5.3 % unemployment rate.
Third, Table l.c summarizes the financial-related regressors.
Fourth, Tables 1 .d and 1 .e summarize the regressors unique to BOD and TSS, respectively.
These regressors mostly relate to permit conditions. Facilities face interim limits about 2 % of the
time. Facilities possess expired permits for 197 days on average. The mean BOD discharge limit
is roughly 800 pounds per day. The mean TSS discharge limit is roughly 1,280 pounds per day.
Both BOD and TSS discharge limits vary across facilities, across years, and within years.46 This
variation confirms the need to examine relative discharges, rather than simply absolute discharges.
5. Estimation Results
5.1. Organization of Results
Finally, the analysis estimates the three environmental performance equations, one for each
type of performance. Initially, the analysis omits financial status as a regressor, in order to examine
all relevant facilities. In the second-to-last sub-section of this section, the analysis includes financial
status as a regressor, with and without its interaction with deterrence measures, while examining the
sub-sample of facilities owned by publicly-held firms. Since the regressor list includes various
measures based on a preceding 12-month period, e.g., cumulative EPA inspections, or preceding
calendar year, e.g., annual revenues, the regression sample period starts on January, 1996.
Consequently, the sample sizes drop to 20,398 for BOD discharges, to 23,228 for TSS discharges,
and to 32,109 for limit exceedances. To test the differences between pairs of intervention types
(e.g., administrative penalties vs civil penalties), the analysis considers an econometric specification
that omits the interactions between deterrence measures and facility/firm characteristics. Inclusion
of these interactions complicates this testing of differences because the effect of each intervention
type depends on facility/firm characteristics. (Future analysis will transform the facility/firm
characteristics in order to facilitate the comparison of intervention types based on a specification that
includes the noted interactions.) To test differences across facilities' responsiveness to deterrence,
44 When estimating relative discharges, the analysis deletes a handful of observations that indicate
BOD or TSS relative discharge levels greaterthan 10, i.e., discharges exceed the permitted limit by more than
900 %, since the regression analysis is sensitive to outliers.
45 A few penalties do not impose a positively-valued sanction. These zero values are incorporated
into the conditional mean associated with penalty magnitudes. Preliminary analysis attempted to discern
zero-value penalties from positive-value penalties. This effort did not seem to improve the analytical ability
to understand facilities' responses to the imposition of penalties. Future analysis will hope to refine this
effort.
46 According to government officials, BOD limits are sometimes lowered to address ambient surface
water quality concerns associated with dissolved oxygen. A similar logic applies to TSS limits.
23
-------
the analysis considers a second specification that includes the interactions between deterrence
measures and facility/firm characteristics.
As noted above, the analysis uses three econometric models - pooled OLS, random effects,
and fixed effects - and uses standard tests to assess these models. When an F-test indicates
significant facility-specific effects, the fixed effects estimator dominates pooled OLS. Since this
dominance always holds, the study only reports the pooled OLS estimates for the sake of
comparison. When the Hausman test signals that the random effects estimator is consistent with the
fixed effects estimator, the random effects estimator is preferable to the fixed effects estimator since
it is more efficient by construction.
5.2. Omit Financial Variables and Deterrence Interactions: Interpret Effects of Deterrence
Initially, the estimation omits both the financial variables and the deterrence interactions.
Estimation results for BOD relative discharges, TSS relative discharges, and monthly limit
exceedances are shown in Tables 2,3 and 4, respectively. As the first performance measure, this
section interprets the results for BOD relative discharges. Based on the F-test of facility-specific
fixed effects, the fixed effects estimator dominates the pooled OLS estimator, and based on the
Hausman test of fixed effects, the fixed effects estimator dominates the random effects estimator,
as shown in Table 2. Thus, this study focuses on the results of the fixed effects model. Also, the
insignificant coefficient associated with the inverse Mills ratio indicates that the Heckman two-step
method is not needed to correct a selection bias associated with the reporting of BOD discharges,
as shown in Table 2.
More important, the results shown in Table 2 indicate that both specific and general
deterrence affect BOD relative discharges. Consider inspection-related deterrence. The
significantly negative effect of preceding 12-month cumulative EPA inspections indicates that
greater federal presence on site at specific facilities improves performance. The effect of specific
deterrence from state inspections is insignificant. While the effect of state aggregate inspections is
significantly positive, the effect of EPA aggregate inspections is insignificant. Thus, the threat of
neither federal nor state inspections prompts better environmental performance.47 Consider also
penalty-related deterrence. While the significantly positive effect of preceding 12-month cumulative
administrative penalties indicates that an increase in the number of administrative penalties against
specific facilities undermines performance, the significantly negative effect of the preceding 12-
month average administrative penalty magnitude indicates that a larger administrative penalty
improves performance. The estimated effects for civil penalties indicate the opposite conclusion:
more civil penalties improve performance, while a larger civil penalty undermines performance.
The opposite results for the count of administrative penalties and civil penalties may be explained
by the much lesser prevalence of civil penalties. Accordingly, facilities might respond more
strongly to an increase in a less frequently-imposed sanction. As for penalty-related general
deterrence, the effect of average number of administrative penalties against other similar facilities
is significantly positive, while the effect of the average administrative penalty magnitude against
47 The economic theory of regulation may help to explain this unexpected result for KDHE aggregate
inspections. Kambhu (1989) and Kadambe and Segerson (1998) argue that increased regulatory scrutiny may
generate an indirect effect on polluters' performance by prompting them to evade scrutiny more strongly.
This indirect effect mitigates the direct effect of increased scrutiny on polluters' performance. Thus, the
overall effect of an increased inspection threat on facility performance may be negative.
24
-------
other similar facilities is insignificant. In contrast, the effect of average number of civil penalties
against other facilities is insignificant, while the effect of the average civil penalty against other
facilities is significantly negative. These last four results indicate that the threat of more
administrative penalties undermines performance, while the threat of larger civil penalties improves
performance. In sum, these results indicate a mixed degree of effectiveness for interventions and
intervention threats in terms of both inspections and penalties.
As the second performance measure, this section interprets the deterrence-related results for
TSS relative discharges, as shown in Table 3. Based on the F-test of facility-specific fixed effects,
the fixed effects estimator dominates the pooled OLS estimator, and based on the Hausman test of
fixed effects, the random effects estimator dominates the fixed effects estimator, however, this
dominance is only marginal (i.e., the Hausman test only marginally rejects the null hypothesis of
consistent estimates). Consistent with this marginal dominance, the random effects and fixed effects
estimation results are quite similar. Thus, this study considers the results of both the random and
fixed effects models even though the conclusions are identical. Also, the significant coefficient
associated with the inverse Mills ratio indicates that the Heckman two-step method is needed to
correct a selection bias associated with the reporting of TSS discharges, as shown in Table 3.
Similar to the BOD results, the TSS estimation results indicate that deterrence affects
performance. First, both specific and general deterrence stemming from state inspections
significantly improves performance. Specific deterrence stemming from neither administrative nor
civil penalties affects performance. General deterrence stemming from the number of administrative
penalties undermines performance, while general stemming from the average administrative penalty
magnitude improves performance. Lastly, general deterrence stemming from the average civil
penalty magnitude improves performance. Again, in sum, these results indicate a mixed degree of
effectiveness for specific and general deterrence stemming from both inspections and penalties.
As the third performance measures, this section interprets the deterrence-related results for
limit exceedances. Based on the F-test of facility-specific fixed effects, the fixed effects estimator
dominates the pooled OLS estimator, and based on the Hausman test of fixed effects, the random
effects estimator dominates the fixed effects estimator, however, this dominance is only marginal
(i.e., the Hausman test only marginally rejects the null hypothesis of consistent estimates).
Consistent with this marginal dominance, the random effects and fixed effects estimation results are
similar in general. Thus, this study considers the results of both the random and fixed effects models
even though the conclusions are nearly identical. Of the inspection-related effects, only general
deterrence stemming from state inspections significantly affects performance, oddly enough, it
undermines performance. Of the penalty-related effects, only two are significant. Specific
deterrence stemming the number of administrative penalties undermines performance, yet specific
deterrence stemming the average administrative penalty magnitude improves performance. Again,
in sum, the results are mixed.
The interpretation of these results fulfills the first two specific objectives of the primary
broad objective: (1) to identify the effects of actual government interventions - specific deterrence
- on environmental performance, and (2) to identify the effects of intervention threats - general
deterrence - on environmental performance.
5.3. Comparison of Interventions based on their Source
The third objective seeks to compare the effects of specific and general deterrence based on
the source of the intervention. For inspections, the study compares state and federal inspectors; for
25
-------
penalties, it compares EPA administrative courts and federal civil courts. Put differently, the
analysis tests the difference between the effect of federal inspections and the effect of state
inspections on facility performance (i.e., difference between federal and state coefficients). It also
tests the difference between the effects of EPA administrative penalties and the effects of federal
civil penalties. The analysis tests these differences using F-tests, which pose a null hypothesis of
equal effects, as shown in Table 5. The testing uses the estimation results of the fixed effects model
for BOD discharges and the random effects model for TSS discharges and limit exceedances. As
noted above, these estimation results stem from a regression model that omits the interactions
between deterrence measures and facility/firm characteristics, omits the financial regressors, and
uses the sample of all relevant facilities regardless of ownership structure. Examine first the
difference between federal inspections and state inspections. And consider first the specific
deterrence measures. Based on BOD results, the effect of actual federal inspections at a specific
facility is negative, while the effect of actual state inspections is insignificant; moreover, the
difference between the two effects is significant, as shown in Table 5. Thus, specific deterrence
stemming from EPA inspections is more effective at improving performance than specific deterrence
stemming from state inspections. However, TSS results generate the opposite conclusion: actual
state inspections against specific facilities more effectively improve performance than do EPA
inspections. Results for limit exceedances indicate no significant difference between the two
inspection-related specific deterrence effects. Consider second the general deterrence measures.
The results of all three performance measures indicate the same conclusion: no significant difference
exists between EPA inspection-related general deterrence and state inspection-related general
deterrence. These results are consistent with the expectation that actual federal inspections against
specific facilities improve facility performance more strongly than do actual state inspections.
However, these results do not support the same expectation for the threat of inspections.
Examine second the difference between EPA administrative and federal civil penalties.
Consider first specific deterrence measures. The BOD results indicate that an increase in the number
of civil penalties improves performance more greatly than does an increase in the number of
administrative penalties, while an increase in the average administrative penalty magnitude improves
performance more greatly than does an increase in the average civil penalty magnitude. The TSS
results and exceedances results indicate no significant differences. The same conclusions apply for
general deterrence measures. These results, at least the BOD results, are consistent with the
expectation that specific and general deterrence stemming from civil penalties, at least their
frequency, improves facility performance more strongly than does deterrence stemming from
administrative penalties. This difference may support the conjecture that involvement on the part
of the Department of Justice implies greater scrutiny. However, the same BOD results indicate the
completely opposite conclusion in terms of the average penalty magnitude. This study must analyze
more fully this apparent contradiction.
5.4. Effects of Facility and Firm Characteristics
The fourth and fifth specific objectives of the primary broad objective seeks to identify the
effects of facility- and firm-level characteristics on environmental performance. To capture the
effect of firm-level financial status, the study must limit itself to facilities owned by publicly-held
firms. As noted above, the final sub-section of this section estimates that effect.
This sub-section interprets the results for the other characteristics. First, facilities owned by
publicly-held firms significantly outperform facilities owned by privately-held firms, according to
26
-------
BOD results. No significant effect is indicated by the TSS or limit exceedances results. Second,
environmental performance in general does not depend on the type of production as captured by the
industrial sub-sector. Within the TSS results, only one individual sub-sector generates a statistically
significant coefficient. Within the limit exceedances results, no relevant coefficients are significant.
The BOD results cannot generate these coefficients since the industrial sub-sector indicator is time-
invariant and the fixed effects model is the dominant model, i.e., neither pooled OLS nor the random
effects model generates consistent estimates. Third, the highly positive effect of flow capacity
indicates that larger facilities underperform smaller facilities; i.e., pollution treatment involves
diseconomies of scale. (The effects on BOD and TSS relative discharges are highly significant; the
effect on limit exceedances is not significant.) Regulators should evaluate the monitoring and
enforcement pressure placed on larger facilities given these highly significant results. Fourth, the
significantly positive effect of the flow to flow capacity ratio on TSS relative discharges indicates
that facilities facing higher marginal compliance costs increase their relative discharges, given the
interpretation that the flow to flow capacity ratio proxies for marginal compliance costs. The effects
of this proxy on BOD relative discharges and limit exceedances are insignificant. Fifth, an increase
in the stochasticity of wastewater discharges significantly undermines performance in terms of BOD
and TSS discharges. According to Brannland and Lofgren (1996), as discharge variability rises,
facilities may choose to increase their mean compliance level (i.e., decrease their average level of
relative discharges). However, these results indicate that a facility may not be able to separate its
mean level of discharges and the deviation about this mean level.
5.5. Interactions between Deterrence Measures and Facility/Firm Characteristics
The sixth specific objective of the primary objective seeks to identify the interactions
between both facility-level and firm-level characteristics and the effects of both specific and general
deterrence. This objective seeks to learn whether different types of facilities or facilities facing
different corporate conditions respond differently to government interventions. In other words, the
analysis tests whether the effects of deterrence differ according to facility- or firm-level
characteristics. To test for these differences, the analysis interprets the results of the second
specification, which includes the interactions between deterrence measures and facility- and firm-
level characteristics. In particular, the analysis assesses whether the coefficients on interactive terms
are significantly different from zero.
As noted above, civil penalties are imposed much less frequently than are administrative
penalties. As a matter of fact, the imposition of civil penalties is not sufficiently frequent to permit
the interaction between civil penalty-related deterrence and facility/firm characteristics.
Consequently, the analysis is only able to estimate the interactive terms between (1) inspection-
related deterrence and administrative penalty-related deterrence and (2) facility- and firm-level
characteristics.
For each performance measure, based on the F-test of facility-specific fixed effects, the fixed
effects estimator dominates the pooled OLS estimator, and based on the Hausman test of fixed
effects, the random effects estimator dominates the fixed effects estimator. (These F-test and
Hausman test results are available upon request.) Accordingly, the analysis focus its interpretation
on estimation results from the random effects model for each performance measure. Rather than
reporting the estimation results from the numerous individual interactive terms, Table 6 reports the
results from F-tests that discern whether a particular set of interactive terms collectively differ from
zero. (Needless to say, complete regression results are available upon request.) The analysis
27
-------
considers the following sets of interactive terms:
(1) ownership structure;
(2) industrial sub-sectors; and
(3) flow capacity, flow to flow capacity ratio, and stochasticity of discharges.
Moreover, these sets are divided into four sub-sets:
(1) inspection-related specific deterrence,
(2) inspection-related general deterrence,
(3) administrative penalty-related specific deterrence, and
(4) administrative penalty-related general deterrence.
Thus, Table 6 reports twelve F-test results for each performance measure.
These F-test results generate the following conclusions. First, both specific and general
deterrence stemming from both inspections and administrative penalties depend on the ownership
structure of the firm owning the relevant facility. This conclusion applies to all three performance
measures. For example, in terms of BOD relative discharges, facilities owned by publicly-held firms
respond more strongly to specific deterrence stemming from state inspections than do facilities
owned by privately-held firms. In contrast, facilities owned by publicly-held firms respond less
strongly to specific deterrence stemming from the number of administrative penalties than do
facilities owned by privately-held firms, in terms of BOD relative discharges. Second, both specific
and general deterrence stemming from both inspections and administrative penalties depend on the
facility's industrial sub-sector. This conclusion applies to all three performance measures. For
example, in terms of BOD relative discharges, facilities producing industrial organic chemicals
respond less strongly to general deterrence stemming from federal inspections than do facilities
producing "other types of chemicals". In contrast, facilities producing industrial organic chemicals
respond more strongly to specific deterrence stemming from the number of administrative penalties
than do facilities producing "other types of chemicals", in terms of BOD relative discharges. Third,
both specific and general deterrence stemming from both inspections and administrative penalties
depend on the facility' s characteristics other than industrial sub-sector, such as flow capacity. This
conclusion applies to all three performance measures. For example, in terms of BOD relative
discharges, larger facilities, as measured by their flow capacity, respond less strongly to specific
deterrence stemming from federal inspections than do smaller facilities. In contrast, larger facilities
respond more strongly to general deterrence stemming from the average administrative penalty
magnitude than do smaller facilities, in terms of BOD relative discharges.
5.6. Sample of Facilities Owned by Publicly-Held Firms
The two preceding sub-sections examine all of the facility- and firm-level characteristics
except firm-level financial status. This sub-section examine financial status by considering the sub-
sample of facilities owned by publicly-held firms, while incorporating financial status as a regressor
in the estimation process. This sub-section considers the two noted specifications: (1) excluding and
(2) including interactions between deterrence measures and facility/firm characteristics. First, the
analysis examines the effect of financial status on the three performance types. Second, it examines
the interaction between financial status and deterrence measures based on the second specification.
Table 7 reports the estimation results from the first specification. It reports the results of the
F-test for Fixed Effects and Hausman Test for Random Effects, along with adjusted R-squared
values. Based on the results for the two aforementioned tests, the fixed effects model dominates
for BOD relative discharges, while the random effects model dominates for both TSS relative
28
-------
discharges and monthly limit exceedances. Rather than reporting the coefficient estimates for each
model, Table 7 reports only the coefficients from the dominant model. Moreover, Table 7 reports
only the coefficients related to financial status: annual revenues and return on assets. First,
regardless of performance measure, the effect of lagged annual revenues is negative. While this
negative effect is never significant, it is almost marginally significant for TSS relative discharges
and limit exceedances. If truly significant, the estimated effects would indicate that a greater flow
of cash may help to improve subsequent environmental performance, perhaps by helping to alleviate
any liquidity constraint facing firms who wish to invest in better environmental management
techniques. Second, increases in the (lagged) return on assets significantly undermine subsequent
environmental performance in terms of BOD and TSS relative discharges. (The effect on all limit
exceedances is insignificant.) This effect indicates that current financial success, as measured by
a more healthy profit stream relative to total assets, actually impedes future environmental success.
Next, the analysis examines the interaction between financial status and deterrence measures
based on the second specification noted above. Again, the imposition of civil penalties is not
sufficiently frequent to permit the interaction between civil penalty-related deterrence and firm-level
characteristics. Consequently, the analysis is only able to estimate the interactive terms between (1)
inspection-related deterrence and administrative penalty-related deterrence and (2) firm-level
financial status. Rather than reporting the estimation results from the several individual interactive
terms related to the two dimensions of financial status, Table 6 reports the results from F-tests that
discern whether a particular set of interactive terms collectively differ from zero. (Complete
regression results are available upon request.) The analysis considers the following sets:
(1) inspection-related specific deterrence,
(2) inspection-related general deterrence,
(3) administrative penalty-related specific deterrence, and
(4) administrative penalty-related general deterrence.
As shown in Table 6, both specific and general deterrence stemming from both inspections and
administrative penalties depend on the financial status of the firm owning the relevant facility. This
conclusion applies mostly to BOD and TSS relative discharges. Based on BOD discharges, all four
types of deterrence depend on financial status. For TSS discharges, only general deterrence
stemming from inspections does not depend on financial status. For all limit exceedances, only
specific deterrence stemming from inspections depend on financial status. Although not shown in
Table 6, the estimated sign and statistical significance of individual interactive coefficients reveal
the following examples. In terms of BOD relative discharges, facilities owned by firms enjoying
a larger revenue flow respond more strongly to specific deterrence stemming from federal
inspections than do facilities owned by firms suffering a smaller revenue flow. In contrast, based
on TSS relative discharges, facilities owned by firms enj oying a greater return on assets respond less
strongly to general deterrence stemming from the number of administrative penalties than do
facilities owned by firms suffering a lesser return on assets.
5.7. Effects of Community Characteristics
Finally, the analysis examines the effects of community pressure on environmental
performance by interpreting the estimated effects of key community characteristics. The study
draws upon the regression results generated without interactions between deterrence measures and
facility/firm characteristics and based on the sample of all relevant facilities regardless of ownership
structure. Moreover, as noted above, use of a fixed effects model helps to avoid endogeneity
29
-------
concerns between facility and household location decisions. Thus, the analysis interprets only the
fixed effects estimates regardless of the identified dominant model. These estimates are shown in
Tables 2, 3 and 4.
These conclusions follow. First, communities suffering higher unemployment rates
fortunately enjoy better environmental performance. This estimate effect runs counter to an
expectation that communities distracted by unemployment would not be motivated to pressure local
facilities for better environmental performance. Second, communities that vote more often suffer
worse BOD-related performance, while enjoying better TSS-related performance (the latter effect
is only marginally significant with a p-value of 0.11). The latter effect is consistent with the
expectation that more politically engaged communities will more effectively pressure facilities for
better performance. Third, communities that vote more greatly for Democratic presidential
candidates enjoy better environmental performance. This result is consistent with a potential
expectation that Democratic voters care more about environmental protection. Fourth, more
sophisticated communities, as measured by their educational attainment, suffer worse BOD-related
and exceedance-related performance. This result runs contrary to an expectation thatbetter educated
communities would more effectively pressure facilities for better performance. Fifth, more rural
communities, as measured by population density, enjoy better BOD-related performance, while
urban communities enjoy better TSS-related performance. The latter effect is consistent with a
potential expectation that more densely populated communities might be better to mobilize pressure
against local facilities. Sixth, communities consisting of more homeowners suffer worse BOD-
related performance, while enjoying better TSS-related performance. The latter effect is consistent
with the expectation that homeowners are more attached to their communities, thus, more willing
to pressure facilities for better performance. Median age also proxies for community attachment.
Older communities enjoy both better BOD- and TSS-related performance. The latter effect is
consistent with the similar effect of homeownership. However, older communities suffer worse
exceedance-related performance. Seventh, communities with more families suffer worse TSS-
related performance. This results runs counter to the expectation that families care more about the
health concerns associated with water pollution. Consistent with this contrary result, communities
with more families containing children suffer worse BOD- and exceedance-related performance.
In contrast, communities with more families containing children enjoy better TSS-related
performance. Thus, an increase in the proportion of families increases TSS relative discharges, but
a shift in this proportion towards families with children actually decreases TSS relative discharges.
The proportion of male residents has no effect on any performance measures. Eighth, more wealthy
communities enjoy better BOD- and TSS-related performance, while suffering worse exceedance-
related performance. The former two effects are consistent with the expectation that environmental
quality is a normal good; as income rises, residents demand better environmental performance from
their local facilities. Ninth, communities more dependent on chemical manufacturing for their
private earnings enjoy better TSS-related performance. This result runs counter to the expectation
that communities more beholden to local facilities would be less likely to pressure these facilities
for better performance. Tenth, less white communities sufferworse TSS-related performance, while
enjoying better exceedance-related performance. (The positive effect of non-white residents on
BOD relative discharges is insignificant with a p-value of 0.15.) Thus, concerns of environmental
justice are possibly evident only for TSS relative discharges.
6. Conclusion
30
-------
This paper analyzes the effects of external pressure - regulatory and community pressure -
on the level of environmental performance at individual polluting facilities. It considers two
dimensions of regulatory pressure: (1) specific deterrence, which is generated by actual government
interventions - namely inspections and penalties - performed at particular facilities, and (2) general
deterrence, which is generated by the threat of receiving an intervention. As important, it compares
the effects of deterrence - specific and general - based on the source of the intervention. For
inspections, it compares state and federal inspectors; for penalties, it compares EPA administrative
courts and federal civil courts. Second, the study measures community pressure indirectly using key
community characteristics (e.g., education) that proxy for actual pressure. Finally, it considers the
effects of facility- and firm-level characteristics, especially corporate financial status, on
environmental performance. As the primary broad objective, this study attempts to identify the
effects of certain government interventions on environmental performance at individual facilities
in the industrial sector of chemical and allied products. Within this primary objective, this study
derives certain specific objectives that either identify the main effects of government interventions
or determine whether these effects differ based on three factors: source of intervention, type of
facility, and type of firm. All but one specific objective is described above. The remaining specific
obj ective seeks to identify the interactions between the effects of specific and general deterrence and
both facility-level and firm-level characteristics. This objective seeks to learn whether different
types of facilities or facilities facing different corporate conditions respond differently to
government interventions. As the secondary broad objective, this study explores the influence
of local community pressure on environmental performance. The analysis measures community
pressure indirectly using key community characteristics, such as per capita income, which serve as
proxies for actual pressure. For this empirical analysis, the study examines wastewater discharges
by chemical manufacturing facilities in the US for the years 1995 to 2001.
This concluding section neither summarizes nor re-interprets the estimation results reported
above.
31
-------
REFERENCES
Anderson, Eric and Wayne Talley (1995), "The Oil Spill Size of Tanker and Barge Accidents:
Determinants and Policy Implications," Land Economics, v. 71, n. 2 (May), pg. 216-28.
Bartel, Ann and Lacy Thomas (1985), "Direct and Indirect Effects of Regulation: A New Look at
OHSA's Impact," Journal of Law and Economics, v. 28, n. 1 (April), pg. 1-25.
Berkin, Andrew (2002), "Federal Study Calls Spending on Water Systems Perilously Inadequate,"
New York Times, pg. A18, April 10, 2002.
Burby, R.J. and R.G. Patterson (1993), "Improving Compliance with State Environmental
Regulations, Journal of Policy Analysis and Management, v. 12, n. 4, pg. 753-772.
Cohen, Mark (1987), "Optimal Enforcement Strategy to Prevent Oil Spills: An Application of a
Principal-Agent Model with Moral Hazard," Journal of Law and Economics, v. 30, pg. 23-
51.
Deily, M.E. and Wayne Gray (1991), "Enforcement of Pollution Regulation in a Declining
Industry "Journal ofEnvironmental Economics and Management, v. 21, n. 2, pg. 260-274.
Dion, C., P. Lanoie, and B. Laplante (1998), "Monitoring of Pollution Regulation: Do Local
Conditions Matter?" Journal of Regulatory Economics, v. 13, n. 1, pg. 5-18.
EPA (1990), "A Primer on the Office of Water Enforcement and Permits and Its Programs",
Environmental Protection Agency, Washington, DC.
EPA (1994), "1994 Study of Monthly Averages and Related Impacts," Office of Regulatory
Enforcement, Environmental Protection Agency, unpublished report.
Earnhart, Dietrich (2004a), "Panel Data Analysis of Regulatory Factors Shaping Environmental
Performance," Review of Economics and Statistics, forthcoming.
Earnhart, Dietrich (2004b), "Regulatory Factors Shaping Environmental Performance at Publicly-
Owned Treatment Plants," Journal of Environmental Economics and Management,
forthcoming.
Earnhart, Dietrich (2004c), "The Effects of Community Characteristics on Polluter Compliance
Levels," Land Economics, forthcoming.
Earnhart, Dietrich and Lubomir Lizal (2002), "Effects of Ownership and Financial Status on
Corporate Environmental Performance," William Davidson Working Paper # 492.
Epple, Dennis and Michael Visscher (1984), "Environmental Pollution: Modeling Occurrence,
Detection, and Deterrence," Journal of Law and Economics, v. 27, n. 1 (April), pg. 29-60.
GAO (1983), "Report to the Administrator, Environmental Protection Agency: Wastewater
Dischargers are Not Complying with EPA Pollution Control Permits," General Accounting
Office.
GAO (1995), "EPA and the States: Environmental Challenges Require a Better Working
Relationship," General Accounting Office.
GAO (1996), "Water Pollution: Many Violations Have Not Received Appropriate Enforcement
Attention," General Accounting Office, Report # GAO/RCED-96-23, Washington, DC.
Gray, Wayne and M.E. Deily (1996), "Compliance and Enforcement: Air Pollution Regulation in
the U.S. Steel Industry," Journal of Environmental Economics and Management, v. 31, n.
l,pg. 96-111.
Gray, Wayne and Carol Jones (1991a), "Longitudinal Patterns of Compliance with Occupational
Safety and Health Administration Health and Safety Regulations in the Manufacturing
Sector," Journal of Human Resources, v. 26, n. 4 (Fall), pg. 623-653.
32
-------
Gray, Wayne and Carol Jones (1991b), "Are OSHA Health Inspections Effective? A Longitudinal
Study in the Manufacturing Sector," Review of Economics and Statistics, v. 73, n. 3
(August), pg. 504-408.
Gray, Wayne and Ronald Shadbegian (2000), "When is Enforcement Effective - or Necessary?"
presented at the NBER Environmental Economics Workshop, August 8, 2000.
Hall and Associates (1997), "Evaluation of the Kansas Surface Water Quality Standards," Hall and
Associates, Inc., Washington, DC, June 5, 1997.
Harrington, Winston (1988), "Enforcement Leverage When Penalties are Restricted," Journal of
Public Economics, v. 37, pg. 29-53.
Heckman, James (1979), "Sample Selection Bias as Specification Error," Econometrica, v. 47, pg.
153-161.
Helland, Eric (1998a), "The Enforcement of Pollution Control Laws: Inspections, Violations, and
Self-Reporting," Review of Economics and Statistics, v. 80, n. 1, pg. 141-153.
Helland, Eric (1998b), "The Revealed Preferences of State EPAs: Stringency, Enforcement, and
Substitution," Journal of Environmental Economics and Management, v. 35, n. 3, pg. 242-
261.
KDHE (1994), "Cost of Compliance Analysis of Proposed Ammonia Water Quality Standard,"
Kansas Department of Health and Environment, Topeka, KS.
Kadambe, S. and Kathleen Segerson (1998), "On the Role of Fines as an Environmental
Enforcement Tool "Journal of Environmental Planning and Management, v. 19, pg. 72-85.
Kambhu, J. (1989), "Regulatory Standards, Noncompliance and Enforcement," Journal of
Regulatory Economics, v. 1, pg. 103-114.
Laplante, B. and P. Rilstone (1996), "Environmental Inspections and Emissions of the Pulp and
Paper Industry in Quebec," Journal of Environmental Economics and Management, v. 31,
n. 1, pg. 19-36.
Lear, Kelly (1998), "An Empirical Examination of EPA Administrative Penalties," working paper,
Kelley School of Business, Indiana University, March 1998.
Magat, W.A. and W.K. Viscusi (1990), "Effectiveness of the EPA's Regulatory Enforcement: The
Case of Industrial Effluent Standards," Journal of Law and Economics, v. 33, pg. 331-360.
McConnell, Virginia and Gregory Schwarz (1992), "The Supply and Demand for Pollution Control:
Evidence from Wastewater Treatment," Journal of Environmental Economics and
Management, v. 23, n. 1, pg. 54-77.
Nadeau, L. W. (1997), "EPA Effectiveness at Reducing the Duration of Plant-level Noncompliance,"
Journal of Environmental Economics and Management, v. 34, pg. 54-78.
Naysnerski, Wendy and Tom Tietenberg (1992), "Private Enforcement of Federal Environmental
Law," Land Economics, v. 68, n. 1, pg. 28-48.
Olson, Mary (1999), "Agency Rulemaking, Political Influences, Regulation, and Industry
Compliance," Journal of Law, Economics, and Organization, v. 15, n. 3 (October), pg.
573-601.
Peltzman, Sam (1976), "Toward a More General Theory of Regulation," Journal of Law and
Economics, v. 19, n. 2, pg. 211-240.
Scholz, John and Wayne Gray (1990), "OSHA Enforcement and Workplace Injuries: A Behavioral
Approach to Risk Assessment," Journal of Risk andUncertainty, v. 3, n. 3, September 1990,
pg. 283-305.
33
-------
Schwarz, Gregory and Virginia McConnell (1993), "Local Choice and Wastewater Treatment: Plant
Performance," Water Resources Research, v. 29, n. 6, pg. 1589-1600.
Smith, Robert (1979), "The Impact of OSHA Inspections on Manufacturing Injury Rates," Journal
of Human Resources, v. 14, n. 2 (Spring), pg. 145-170.
Stafford, Sarah (2002), "The Effect of Punishment on Firm Compliance with Hazardous Waste
Regulations," Journal of Environmental Economics and Management, forthcoming.
Stigler, G.J. (1971), "The Theory of Economic Regulation," Bell J. ofEcon. and Mgt Science, 2: 3-
21.
Viladrich-Grau, Montserrat and Theodore Groves (1997), "The Oil Spill Process: The Effect of
Coast Guard Monitoring on Oil Spills," Environmental and Resource Economics, v. 10, n.
4 (Dec), pg. 315-339.
Viscusi, Kip (1979), "The Impact of Occupational Safety and Health Regulation," Bell Journal of
Economics, v. 10, n. 1 (Spring), pg. 117-140.
Wasserman, Cheryl (1984), "Improving the Efficiency and Effectiveness of Compliance Monitoring
and Enforcement of Environmental Policies, United States: A National Review," OECD.
34
-------
Table 1
Summary Statistics
Table l.a. Environmental Performance Measures
Variable
N
Mean
Standard
Deviation
Min
Max
Monthly Frequency of Limit Exceedances
32,019
0.31825
0.9036791
0
41.00
BOD Relative Discharges
20,398
0.316
0.355
0
9.871
BOD Relative Discharges (logs)
20,398
-1.91297
1.2300820
-10.61684
2.2533948
TSS Relative Discharges
23,228
0.282
0.341
0
10.521
TSS Relative Discharges (logs)
23,228
-1.59803
1.1030648
-11.36389
2.2060348
l.b. Regressors Common to All Dependent Variables (Except Financial-related Regressors)
Variable
N
Mean
Std Dev
Min
Max
Preceding 12-month Cumulative EPA Inspections
32019
0.0876355
0.3342007
0
4
Preceding 12-month Cumulative State Inspections
32019
1.3554452
1.8052616
0
27
Annual EPA Inspections of Others / # of Others
32019
0.0802682
0.0763154
0
0.7272727
Annual State Inspections of Others / # of Others
32019
1.2807481
1.2581372
0
9.625
Preceding 12-month Cumulative Admin Penalties
32019
0.0447859
0.2737716
0
4
Preceding 12-month Avg Admin Penalty ($/action)
32019
3227.74
111289.73
0
8225931
Preceding 12-month Cumulative Civil Penalties
32019
0.0013742
0.0460643
0
2
Preceding 12-month Avg Civil Penalty ($/action)
32019
235.559168
10187.3
0
556881
Annual Admin Penalties on Others / # of Others
32019
0.0410064
0.0530531
0
0.15
Annual Average Admin Penalty on Others ($/action)
32019
59995.07
128615.55
0
587827.64
Annual Civil Penalties on Others / # of Others
32019
0.0020406
0.0058374
0
0.037037
Annual Average Civil Penalty on Others ($/action)
32019
64255.97
202684.31
0
1266976
State and Local Budget / # of businesses ($ per)
32019
43308.0566
33.2059753
9.1575639
566.9949875
EPA Regional Budget / # of businesses ($ per)
32019
677.5628346
154.43205
473.7963493
1229.3
EPA Overall Budget / # of businesses ($ per)
32019
17489.4022
1.5932892
15.7866106
20.6166986
Region 2 (1,0)
32019
0.0800775
0.2714173
0
1
Region 3 (1,0)
32019
0.1244886
0.3301433
0
1
Region 4 (1,0)
32019
0.2600019
0.4386421
0
1
Region 5 (1,0)
32019
0.1220838
0.3273876
0
1
Region 6 (1,0)
32019
0.3265873
0.4689722
0
1
Publicly-Held Ownership (1,0)
32019
0.6299697
0.48282
0
1
SICOl: alkalies/chlorine, gases, inorganic pigments 3
32019
0.0761423
0.2652298
0
1
SIC02: organic fibers, surface agents, adhesives 3
32019
0.0534995
0.2250308
0
1
35
-------
SIC03: toilet preparations, pharmaceuticals 3
32019
0.0608389
0.2390383
0
1
SIC 19: industrial inorganics 3
32019
0.1271433
0.3331386
0
1
SIC21: plastic materials and resins 3
32019
0.1961648
0.3971008
0
1
SIC65: cyclic crudes and intermediates 3
32019
0.0518442
0.2217159
0
1
SIC69: industrial organics 3
32019
0.2427309
0.4287404
0
1
Flow Capacity (million gallons / day)
32019
2.417954
4.3171077
0.000695
65.0416667
Flow to Flow Capacity (ratio)
32019
1.077509
2.1378475
0
203.5253933
Winter Season (1,0)
32019
0.2575346
0.4372831
0
1
Spring Season (1,0)
32019
0.2729317
0.445473
0
1
Summer Season (1,0)
32019
0.2426372
0.4286842
0
1
Unemployment (rate)
32019
0.0531925
0.0198478
0.009
0.158
Voter Turnout (rate)
32019
0.3749613
0.0560933
0.2234865
0.532039
Democratic Vote (proportion)
32019
0.4702322
0.0900864
0.218163
0.8251648
Bachelor's Degree or more (proportion)
32019
0.1827148
0.0896708
0.025
0.703
Population Density (person/sq mile)
32019
684.8055623
1245.4
14.8
11412.3
Owner Occupied Housing (proportion)
32019
0.6550819
0.1182281
0.297
0.942
Median Age (years)
32019
35.4350948
3.9474492
22
49
Family Households w/ Children (proportion)
32019
0.4812143
0.0542532
0.2584856
0.6835023
Family Households (proportion)
32019
0.6842599
0.0765644
0.388
0.8778
Male Residents (per 100 females)
32019
89.5252016
10.7172648
69.7
194.78
Per Capita Income ($/person)
32019
22600.82
4841.08
12955
50002
Chemical-Related Private Earnings (proportion)
32019
0.09928
0.1112577
0
0.6067182
Non-White Residents (proportion)
32019
0.2534437
0.1987409
0
0.9109927
11 The omitted category for industrial sub-sector is "other", which contains these sub-sectors: 2822 (synthetic
rubber), 2841 (soaps), 2842 (polishes), 2861 (sanitation goods), 2879 (gum/wood chemicals), and 2892
(explosives).
I.e. Financial-Related Regressors, which are Common to All Performance Measures
Variable
N
Mean
Std Dev
Minimum
Maximum
Total Revenues [lagged] ($)
Return on Assets Raggedl (ratio)
18073
18073
18,014,833,701
0.0690892
24,348,745,654
0.0811904
18,359,000
-0.2896256
147,045,823,484
0.436318
l.d. Regressors Unique to BOD Relative Emissions
Variable
N
Mean
Std Dev
Min
Maximum
Monthly Effluent Limit (lbs/day)
20398
800.8735078
2437.18
0
31686.83
Interim Limit Type (1,0)
20398
0.0205903
0.1420116
0
1
Modification to Permit (1,0)
20398
0.1016766
0.30223
0
1
Permit Expiration (days)
20398
197.1533
544.27407
0
5145
Standard Deviation of Relative Discharges
20389
0.1813026
0.9624849
0
38.466098
36
-------
I.e. Regressors Unique to TSS Relative Emissions
Variable
N
Mean
Std Dev
Min
Max
Monthly Effluent Limit (lbs/day)
23228
1283.31
4038.01
0
50000
Interim Limit Type (1,0)
23228
0.0162304
0.1263633
0
1
Modification to Permit (1,0)
23228
0.0867488
0.2814726
0
1
Permit Expiration (days)
23228
210.33507
580.15835
0
5693
Standard Deviation of Relative Discharges
23228
0.1737951
0.3744842
0
17.0494
37
-------
Table 2
Estimation of BOD Relative Emissions
Variable
Pooled OLS
Random Effects
Fixed E:
Tects
Coefficient
P-value
Coefficient
P-value
Coefficient
P-value
Preceding 12-month Cumulative
EPA Inspections
-0.0285
0.2278
-0.03446
0.0562
-0.03553
0.0501
Preceding 12-month Cumulative
State Inspections
0.00055411
0.9205
0.007303
0.1923
0.006263
0.2672
Annual EPA Inspections of Others /
# of Others
-0.05603
0.757
0.071251
0.5817
0.067272
0.6059
Annual State Inspections of Others /
# of Others
0.03395
0.0003
0.06393
<.0001
0.061703
<.0001
Preceding 12-month Cumulative
Admin Penalties
0.20817
<.0001
0.042691
0.0878
0.044217
0.0784
Preceding 12-month Avg Admin
Penalty
-1.21e-07
0.0435
-1.18e-07
0.0073
-1.18e-07
0.0075
Preceding 12-month Cumulative
Civil Penalties
0.17198
0.4437
-0.52146
0.0021
-0.53663
0.0016
Preceding 12-month Avg Civil
Penalty
-0.00000346
0.0002
1.33e-06
0.054
1.37e-06
0.0474
Annual Admin Penalties on Others /
# of Others
-0.30756
0.3303
1.001932
<.0001
0.988102
<.0001
Annual Average Admin Penalty on
Others
5.38e-08
0.6126
2.48e-08
0.7608
2.37e-08
0.7733
Annual Civil Penalties on Others / #
of Others
-12.31464
<.0001
-8.15503
<0001
-7.92676
<.0001
Annual Average Civil Penalty on
Others
3.27e-07
<0001
2.63e-07
<0001
2.57e-07
<.0001
State and Local Budget / # of
businesses
0.00146
0.0025
0.000523
0.5452
0.000754
0.4008
EPA Regional Budget / # of
businesses
0.00082242
<.0001
0.001134
<.0001
0.00116
<.0001
EPA Overall Budget / # of businesses
-0.089
<.0001
-0.07673
<.0001
-0.07618
<.0001
Region 2
-0.33199
<0001
-0.80376
0.1012
N/A
Region 3
0.12458
0.038
-0.28801
0.5214
N/A
Region 4
0.53165
<.0001
0.211143
0.621
N/A
Region 5
0.72466
<.0001
-0.14162
0.7638
N/A
Region 6
0.23757
0.0005
-0.32289
0.4499
N/A
Monthly Effluent Limit
-0.00004459
<.0001
-0.00026
<.0001
-0.00042
<.0001
Interim Limit Type
0.49856
<.0001
-0.19814
<.0001
-0.204
<.0001
Modification to Permit
0.03449
0.2214
-0.11758
0.0012
-0.12326
0.0008
38
-------
Permit Expiration
-5.39e-08
0.0009
3.75e-09
0.8567
1.38e-08
0.5122
Publicly-Held Ownership
-0.03928
0.034
-0.06807
0.0023
-0.06709
0.0031
SICOl
-0.0409
0.4354
0.176095
0.7412
N/A
SIC02
-0.06409
0.1221
-0.00632
0.9885
N/A
SIC03
-0.30997
<.0001
0.287759
0.5033
N/A
SIC 19
-0.18737
<0001
-0.23919
0.5034
N/A
SIC21
0.22934
<0001
0.260544
0.3694
N/A
SIC65
-0.59664
<0001
-0.59721
0.1375
N/A
SIC69
0.19939
<0001
0.357438
0.2034
N/A
Flow Capacity
0.02667
<.0001
0.072786
<.0001
0.072752
<.0001
Flow to Flow Capacity Ratio
-0.02942
<.0001
0.002972
0.5159
0.003415
0.4571
Std Deviation of Relative Discharges
0.0944
<.0001
0.034071
<.0001
0.033154
<.0001
Winter Season
0.18732
<.0001
0.191222
<.0001
0.191235
<.0001
Spring Season
0.12959
<.0001
0.120142
<.0001
0.119394
<.0001
Summer Season
0.0571
0.0154
0.028412
0.0798
0.026611
0.1021
Unemployment
-4.54041
<.0001
-3.52847
<0001
-4.09001
<.0001
Voter Turnout
-0.19833
0.3959
2.963887
<0001
3.527721
<.0001
Democratic Vote
0.63181
<.0001
-0.98449
0.0327
-0.93586
0.0602
Bachelor's Degree or more
0.43304
0.0004
1.774633
0.1138
0.43304
0.0004
Population Density
-0.00014036
<.0001
0.000139
0.1347
0.001107
0.0001
Owner Occupied Housing
0.9763
<.0001
1.667313
0.0501
2.59903
0.0148
Median Age
-0.02989
<.0001
-0.0083
0.7647
-0.02989
<.0001
Family Households w/ Children
0.80605
0.0047
6.736635
<.0001
7.855912
<.0001
Family Households
0.16969
0.5484
-0.75455
0.5872
-2.01927
0.2422
Male Residents
-0.00038563
0.6836
-0.00069
0.874
-0.00127
0.8013
Per Capita Income
-0.00000748
0.0091
-0.00007
<.0001
-0.0001
<.0001
Chemical-Related Private Earnings
-0.28285
0.0048
-0.46379
0.3153
-0.3215
0.5599
Non-White Residents
0.3875
<.0001
0.730073
0.0815
0.876621
0.1543
Inverse Mills Ratio
-2.68841
<.0001
0.329728
0.244
0.364889
0.201
Adjusted R-squared
0.1326
0.0421
0.5991
Number of Observations
20388
20388
20388
Regression also includes an intercept term.
Hausman Test for Random Effects: statistic = 83.74, degrees of freedom = 39, p-value = 0.001
F-Test for Fixed Effects: statistic = 74.57, degrees of freedom = 341, p-value = 0.0001
The sample includes only those observations where a legal limit applies and emissions are reported.
The analysis uses a two-stage estimation process when estimating the reporting and performance equations.
To help identify these two equations, the analysis excludes two variables. An LRtest confirms these
variables are insignificantly different from zero when tested collectively.
39
-------
Table 3
Estimation of TSS Relative Emissions
Variable
Pooled OLS
Random Effects
Fixed E:
Tects
Coefficient
P-value
Coefficient
P-value
Coefficient
P-value
Preceding 12-month Cumulative
EPA Inspections
0.04059
0.0428
-0.01271
0.4003
-0.01163
0.4439
Preceding 12-month Cumulative
State Inspections
-0.079
<0001
-0.0411
<0001
-0.04116
<.0001
Annual EPA Inspections of Others /
# of Others
-0.1444
0.3391
-0.10332
0.3365
-0.08472
0.4344
Annual State Inspections of Others /
# of Others
0.01008
0.1964
-0.02468
0.0124
-0.02353
0.02
Preceding 12-month Cumulative
Admin Penalties
0.06517
0.0088
0.029165
0.1471
0.027687
0.1713
Preceding 12-month Avg Admin
Penalty
1.02e-07
0.0546
-3.00e-08
0.4437
-3.19e-08
0.4171
Preceding 12-month Cumulative
Civil Penalties
0.41178
0.0147
0.041392
0.7436
0.042133
0.7404
Preceding 12-month Avg Civil
Penalty
-0.00000221
0.0039
6.44e-07
0.2527
6.36e-07
0.2603
Annual Admin Penalties on Others /
# of Others
0.44785
0.0885
0.300147
0.1121
0.332448
0.0808
Annual Average Admin Penalty on
Others
-1.89e-07
0.0348
-2.09e-07
0.0023
-2.09e-07
0.0025
Annual Civil Penalties on Others / #
of Others
-2.65808
0.2595
1.666018
0.3187
2.003711
0.2354
Annual Average Civil Penalty on
Others
-1.19e-07
0.0606
-2.36e-07
<0001
-2.47e-07
<.0001
State and Local Budget / # of
businesses
0.00087426
0.029
-0.00027
0.7283
-0.00099
0.2192
EPA Regional Budget / # of
businesses
0.00073485
<.0001
0.000494
0.0012
0.000401
0.0099
EPA Overall Budget / # of businesses
0.03943
<.0001
0.026177
0.0016
0.037358
<.0001
Region 2
-0.17616
0.0003
-0.02394
0.9446
N/A
Region 3
-0.03855
0.46
0.154999
0.6333
N/A
Region 4
0.21292
0.0023
-0.14322
0.6366
N/A
Region 5
0.79363
<.0001
0.664953
0.0456
N/A
Region 6
-0.18558
0.0042
-0.31789
0.3077
N/A
Monthly Effluent Limit
-0.00001979
<.0001
-0.00001
0.0093
-8.50e-06
0.0578
Interim Limit Type
-0.10605
0.0515
0.003581
0.9375
-0.00244
0.9578
Modification to Permit
-0.0441
0.0984
-0.13013
0.0002
-0.13216
0.0002
40
-------
Permit Expiration
-1.75e-08
0.17
-4.47e-08
0.0043
-4.95e-08
0.0019
Publicly-Held Ownership
-0.0095
0.5306
0.003638
0.8478
0.007132
0.7108
SICOl
-0.54621
<0001
-0.3946
0.1669
N/A
SIC02
-0.57503
<.0001
-0.40152
0.2034
N/A
SIC03
-1.08986
<.0001
-0.89429
0.0081
N/A
SIC 19
-0.40261
<0001
-0.21289
0.392
N/A
SIC21
-0.18738
<0001
-0.04482
0.8416
N/A
SIC65
-0.32694
<0001
-0.28793
0.3682
N/A
SIC69
-0.32979
<0001
-0.15917
0.4636
N/A
Flow Capacity
-0.00617
0.0191
0.054325
<.0001
0.060198
<.0001
Flow to Flow Capacity Ratio
0.01836
<.0001
0.019456
<.0001
0.019755
<.0001
Std Deviation of Relative Discharges
0.48727
<.0001
0.214487
<0001
0.209525
<.0001
Winter Season
0.06417
0.0012
0.081536
<.0001
0.080849
<.0001
Spring Season
0.04973
0.011
0.065256
<.0001
0.064532
<.0001
Summer Season
0.01952
0.3227
0.0269
0.0463
0.026284
0.0526
Unemployment
-5.6557
<.0001
-4.00766
<0001
-3.85062
<.0001
Voter Turnout
0.41705
0.0306
-0.54908
0.3219
-0.9713
0.1106
Democratic Vote
0.10637
0.3553
-1.1986
0.0013
-1.44499
0.0004
Bachelor's Degree or more
0.05844
0.5472
0.790687
0.3247
0.05844
0.5472
Population Density
0.00002487
0.0105
-0.00006
0.3622
-0.00081
0.0002
Owner Occupied Housing
0.03115
0.8185
-1.55116
0.0188
-3.09663
0.0003
Median Age
-0.01105
0.0006
-0.00397
0.8517
-0.01105
0.0006
Family Households w/ Children
0.65589
0.0073
-0.98229
0.2064
-2.00544
0.0209
Family Households
0.94729
<.0001
5.248145
<.0001
9.451691
<.0001
Male Residents
-0.00298
0.0001
-0.00132
0.6979
-0.00629
0.1217
Per Capita Income
-0.00002447
<.0001
-0.00004
<.0001
-0.00002
0.0098
Chemical-Related Private Earnings
-0.0217
0.7907
-0.51673
0.1727
-1.06125
0.0238
Non-White Residents
-0.04976
0.3583
1.11386
0.0006
2.12361
<.0001
Inverse Mills Ratio
-4.21054
<.0001
-2.6524
<.0001
-2.68021
<.0001
Adjusted R-squared
0.1323
0.0361
0.6024
Number of Observations
23201
23201
23201
Regression also includes an intercept term.
Hausman Test for Random Effects: statistic = 53.62, degrees of freedom = 39, p-value = 0.0596
F-Test for Fixed Effects: statistic = 71.36, degrees of freedom = 405, p-value = 0.0001
The sample includes only those observations where a legal limit applies and emissions are reported.
The analysis uses a two-stage estimation process when estimating the reporting and performance equations.
To help identify these two equations, the analysis excludes two variables. An LRtest confirms these
variables are insignificantly different from zero when tested collectively.
41
-------
Table 4
Estimation of Monthly Effluent Limit Exceedances
Variable
Pooled OLS
Random Effects
Fixed E:
Tects
Coefficient
P-value
Coefficient
P-value
Coefficient
P-value
Preceding 12-month Cumulative
EPA Inspections
0.06379
<.0001
0.016833
0.256
0.015684
0.2932
Preceding 12-month Cumulative
State Inspections
0.03748
<.0001
0.004338
0.3201
0.003205
0.4698
Annual EPA Inspections of Others /
# of Others
0.29109
0.0054
-0.03774
0.6991
-0.07453
0.4508
Annual State Inspections of Others /
# of Others
-0.02889
<0001
0.02654
0.0035
0.032722
0.0006
Preceding 12-month Cumulative
Admin Penalties
0.1943
<.0001
0.037241
0.0599
0.031489
0.1146
Preceding 12-month Avg Admin
Penalty
5.25e-09
0.9088
-7.17e-08
0.1047
-7.48e-08
0.0921
Preceding 12-month Cumulative
Civil Penalties
0.16871
0.2486
-0.02975
0.8354
-0.05748
0.69
Preceding 12-month Avg Civil
Penalty
-5.89e-07
0.3747
-6.93e-09
0.9913
6.91e-08
0.9143
Annual Admin Penalties on Others /
# of Others
0.49641
0.0105
0.199677
0.2765
0.20588
0.2671
Annual Average Admin Penalty on
Others
-5.78e-08
0.3699
2.23e-08
0.7311
1.60e-08
0.8096
Annual Civil Penalties on Others / #
of Others
-0.48975
0.7682
-0.55276
0.716
-0.57523
0.7093
Annual Average Civil Penalty on
Others
1.93e-09
0.9661
7.69e-09
0.8561
2.06e-08
0.6379
State and Local Budget / # of
businesses
0.00011234
0.5757
-0.00096
0.1223
-0.00099
0.1899
EPA Regional Budget / # of
businesses
0.00017339
0.076
0.000071
0.6044
0.000123
0.3936
EPA Overall Budget / # of businesses
-0.02097
<0001
-0.01783
0.0038
-0.01723
0.0212
Region 2
-0.18581
<0001
-0.27123
0.0714
N/A
Region 3
0.12528
<.0001
0.06876
0.6357
N/A
Region 4
-0.0444
0.2911
-0.00963
0.9456
N/A
Region 5
0.04883
0.2937
-0.10296
0.4841
N/A
Region 6
-0.18752
<0001
-0.12647
0.3559
N/A
Interim Limit Type
0.01998
0.6008
0.013823
0.7415
0.011358
0.7892
Modification to Permit
-0.00185
0.9276
0.06593
0.0394
0.07751
0.0191
Permit Expiration
-4.54e-09
0.6769
-7.10e-08
<.0001
-7.52e-08
<0001
42
-------
Publicly-Held Ownership
-0.05301
<.0001
-0.01823
0.2972
-0.01552
0.3947
SICOl
0.17581
<.0001
0.143367
0.2648
N/A
SIC02
-0.04662
0.0662
-0.10345
0.4742
N/A
SIC03
-0.147
<.0001
-0.1957
0.1802
N/A
SIC 19
0.03561
0.0603
-0.0321
0.7643
N/A
SIC21
-0.02501
0.1416
-0.03111
0.7521
N/A
SIC65
0.00594
0.8177
-0.02882
0.8461
N/A
SIC69
0.07255
<.0001
0.017905
0.8489
N/A
Flow Capacity
0.00791
<.0001
0.002906
0.4311
0.000198
0.9633
Flow to Flow Capacity Ratio
-0.00193
0.4082
-0.00014
0.9493
-0.00026
0.9044
Winter Season
0.04459
0.002
0.04847
0.0002
0.048383
0.0002
Spring Season
-0.00228
0.8733
0.002323
0.8561
0.002121
0.869
Summer Season
0.01955
0.1793
0.019956
0.1248
0.019657
0.1318
Unemployment
-1.46681
<.0001
-1.15431
0.0837
-0.90661
0.2025
Voter Turnout
0.10322
0.4506
0.055088
0.9023
-0.5354
0.3504
Democratic Vote
-0.44481
<.0001
-0.31722
0.2835
-0.52036
0.1932
Bachelor's Degree or more
0.386
<.0001
0.496075
0.214
0.386
<0001
Population Density
5.87e-07
0.9231
1.87e-06
0.9507
-0.00016
0.4037
Owner Occupied Housing
-0.30829
0.0021
-0.13524
0.7628
0.383716
0.6539
Median Age
0.00781
0.0014
0.001151
0.9194
0.00781
0.0014
Family Households w/ Children
-0.19432
0.2809
0.20037
0.752
1.44521
0.0916
Family Households
1.26454
<.0001
0.976237
0.185
0.594767
0.6615
Male Residents
-0.00063886
0.252
-0.0044
0.0713
-0.00137
0.7329
Per Capita Income
-0.00000785
<.0001
4.16e-06
0.4752
0.00002
0.0211
Chemical-Related Private Earnings
-0.14771
0.0165
0.267834
0.3157
0.500985
0.2731
Non-White Residents
0.26043
<.0001
-0.17528
0.3724
-2.2445
<.0001
Adjusted R-squared
0.0286
0.0047
0.2346
Number of Observations
32019
32019
32019
Regression also includes an intercept term.
Hausman Test for Random Effects: statistic = 47.58, degrees of freedom = 35, p-value = 0.0762
F-Test for Fixed Effects: statistic = 17.49, degrees of freedom = 504, p-value = 0.0001
43
-------
Table 5
Differences between Effects of Paired Interventions
on Corporate Environmental Performance:
Results of F-Tests a
Table 5.1. Comparison of Federal and State Inspections
Comparison
BOD
TSS
All Exceedances
F-value
P-value
F-value
P-value
F-value
P-value
Specific Deterrence
4.76
0.029
3.02
0.082
0.64
0.422
General Deterrence
0.10
0.966
0.53
0.465
0.43
0.511
Table 5.2. Comparison of EPA Administrative and Federal Civil Penalties
Comparison
BOD
TSS
All Exceedances
F-value
P-value
F-value
P-value
F-value
P-value
Specific Deterrence
Number of Penalties
11.7
0.001
0.01
0.923
0.22
0.640
Average Magnitude
4.62
0.032
1.43
0.232
0.01
0.919
General Deterrence
Number of Penalties
19.73
0.000
0.67
0.415
0.24
0.621
Average Magnitude
7.64
0.006
0.15
0.701
0.05
0.828
a Based on the fixed effects model for BOD discharges and the random effects model for TSS
discharges and all limit exceedances. Both models exclude interactions between deterrence
measures and facility/firm characteristics, exclude financial-related regressors, and use the sample
of all facilities regardless of ownership structure.
44
-------
Table 6
Collective Significance of Interactive Terms involving
Deterrence Measures and Facility / Firm Characteristics:
Results of F-Tests a
Table 6.1. Federal and State Inspections
Set of Regressors
BOD
TSS
All Exceedances
F-value
P-value
F-value
P-value
F-value
P-value
Specific Deterrence
Ownership Structure
8.75
0.000
3.73
0.024
15.15
0.000
Industrial Sub-sector
4.45
0.000
5.24
0.000
3.01
0.000
Capacity, Flow ratio, Stochasticity
31.14
0.000
20.50
0.000
0.81
0.516
Financial Statusb
9.84
0.000
6.08
0.000
2.22
0.065
General Deterrence
Ownership Structure
8.21
0.000
7.88
0.000
12.55
0.000
Industrial Sub-sector
4.79
0.000
4.66
0.000
1.76
0.038
Capacity, Flow ratio, Stochasticity
2.03
0.058
11.12
0.000
3.61
0.006
Financial Statusb
7.31
0.000
0.73
0.571
0.58
0.065
a Unless otherwise noted, based on the random effects model that includes interactions between
deterrence measures and facility/firm characteristics, yet excludes financial-related regressors, and
uses the sample of all facilities regardless of ownership structure.
b Based on the random effects model that includes interactions between deterrence measures and
facility/firm characteristics, including financial-related regressors, and uses the sample of only
facilities owned by publicly-held firms.
45
-------
Table 6.2. EPA Administrative Penalties
Set of Regressors
BOD
TSS
All Exceedances
F-value
P-value
F-value
P-value
F-value
P-value
Specific Deterrence
Ownership Structure
1.33
0.265
0.57
0.564
7.35
0.001
Industrial Sub-sector
6.89
0.000
4.99
0.000
4.60
0.000
Capacity, Flow ratio, Stochasticity
2.23
0.037
6.04
0.000
7.84
0.000
Financial Statusb
6.69
0.000
3.23
0.012
0.81
0.520
General Deterrence
Ownership Structure
3.81
0.022
18.28
0.000
0.94
0.391
Industrial Sub-sector
5.25
0.000
4.44
0.000
0.99
0.461
Capacity, Flow ratio, Stochasticity
5.25
0.000
6.38
0.000
6.32
0.000
Financial Statusb
1.99
0.093
6.66
0.000
0.27
0.900
a Unless otherwise noted, based on the random effects model that includes interactions between
deterrence measures and facility/firm characteristics, yet excludes financial-related regressors, and
uses the sample of all facilities regardless of ownership structure.
b Based on the random effects model that includes interactions between deterrence measures and
facility/firm characteristics, including financial-related regressors, and uses the sample of only
facilities owned by publicly-held firms.
46
-------
Table 7
Estimation of Environmental Performance Measures:
Inclusion of Financial Status as Regressor
Based on Sample of Facilities Owned by Publicly-Held Firms
Variable / Statistic
BOD a
TSS b
All Exceedancesb
Coeff/
Statistic
P-value
Coeff/
Statistic
P-value
Coeff/
Statistic
P-value
Annual Revenues
-709E-15
0.544
-14E-13
0.155
-122E-14
0.165
Return on Assets
0.7724
0.0001
0.3653
0.003
-0.0682
0.591
F-test for Fixed Effects
58.27
0.0001
64.16
0.0001
12.51
0.0001
Adjusted R2: Fixed Effects
0.6285
0.6242
0.2162
Hausman Test for
Random Effects
59.72
0.006
24.43
0.944
33.06
0.562
Adjusted R2: Random Effects
0.0848
0.0320
0.0079
The regression for BOD, TSS, and All Exceedances includes all of the regressors listed in Tables
2, 3, and 4, respectively, with the exception of Public Ownership Structure.
a Based on a fixed effects model.
b Based on a random effects model.
47
-------
5/16/2004
Deterrence and Corporate Environmental Behavior
Dorothy Thornton (University of California, Berkeley)
Neil A. Gunningham (Australian National University)
Robert A. Kagan ((University of California, Berkeley)
This research addresses the assumption that "general deterrence" is an important key to
enhanced compliance with regulatory laws. Through a survey of 233 firms in 8
industries in the U. S., and in-depth interviews with 34 firms in the chemical and
electroplating industries, asked (1) When severe legal penalties are imposed against a
violator of environmental laws, do other companies in the same industry actually learn
about such "signal cases"? (2) Does knowing about "signal cases" or other "general
deterrence messages" change firms' compliance-related behavior? (3) How important is
the threat raised by general deterrence compared with other factors in inducing legal
compliance? We found that only 42% of respondents could identify the "signal case.
But 89% could identify some enforcement actions against other firms, and 63% of firms
reporting having taken some compliance-related actions in response to learning about
such cases. Overall, we conclude that because most firms already are in compliance (for
a variety of other reasons), this form of "explicit general deterrence" knowledge usually
serves not to enhance the threat of legal punishment but as reassurance that compliance
is not foolish and as a reminder to check on the reliability of existing compliance
routines.
Prepared for the US EPA Corporate Environmental Behavior Research Workshop,
Washington, DC, April 26-27, 2004
48
-------
5/16/2004
Deterrence and Corporate Environmental Behavior1
In most regulatory programs, officials formally prosecute and obtain legal sanctions
against violators in only a small percentage of infractions. They deal with most detected
violations at the bottom of the "pyramid of sanctions" (Ayres & Braithwaite, 1992) - that is, by
means of warnings, demands for remedial action, repeated re-inspection, and other informal
pressures. At the same time, most regulatory officials, scholars, and environmental advocacy
groups believe that governmental capacity to impose severe legal penalties, together with
relatively frequent use of that capacity, is crucial to the implementation of regulatory norms.
Underlying this belief is the theory of general deterrence, which holds that each tough legal
penalty sends a "threat message" that reverberates through the community of regulated
businesses. That threat presumably raises the perceived risk and cost of violations, and business
executives increase their investment in compliance commensurately. Yet there is surprisingly
little research that examines the extent to which general deterrence actually is important in
motivating business firms' environmental behavior. This paper summarizes the results of a
research project designed to explore that issue.
I. Explaining Regulatory Compliance: Alternative Hypotheses
A good deal of sociolegal scholarship questions the relationship between general deterrence
and corporate regulatory compliance. First, in the cacophony of news, information, and demands
of all kinds received by business firms in contemporary society, it is not clear how often business
enterprises learn about legal penalties imposed on other firms in other places. Even if they do,
business executives may not think that their firm (which may differ in many ways from the
sanctioned firm) faces an enhanced risk of being found in violation and punished (see
Braithwaite & Makkai, 1991). Thus against the general deterrence thesis, which assumes
widespread dissemination and attention to clear deterrence messages, one might counterpose a
"weak signal, weak threat" hypothesis — that is, that the message often doesn't get through or
send a meaningful threat.
Second, some research indicates that it is not general deterrence (hearing about legal
sanctions against others) but "specific deterrence" — the fear triggered by a firm's experience of
1 This work was supported by the U.S. EPA's STAR Grant Program and the Center for the Study of Law and
Society at University of California, Berkeley. The authors are also grateful to Manuel Vallee and Peter Younkin for
their research assistance.
49
-------
5/16/2004
being inspected, warned, or penalized itself- that is the chief driver of enhanced compliance
efforts (Gray & Scholz, 1991; Gray & Shadbegian, 2004; Mendel off & Gray, 2004).
Third, in economically advanced democracies, many corporate officials regard the threat of
informal social sanctions - such as the damage to corporate reputation that can flow from
negative publicity about a firm's environmental pollution - as far more salient and economically
costly than the risk of legal penalties (Gunningham et al, 2003; Mehta & Hawkins, 1998;
Prakash, 2000). In consequence, most general deterrence messages may be redundant, exerting
little impact on corporate compliance behavior.
Fourth, high levels of compliance often are observed in contexts in which the threat of legal
enforcement is relatively remote. Hence some scholars argue that for most firms, compliance
stems not from fear of legal sanctions but from a sense of social duty or legal obligation. In
democratic societies with a strong rule of law tradition, the theory suggests, most business
managers have "internalized" (or simply agree with) the norms that underlie most regulatory
rules. Or they are generally committed, as a matter of socialization and citizenship, to complying
with duly enacted laws and regulations. For these firms, too, one might hypothesize that general
deterrence signals are redundant, adding little if anything to compliance efforts.
Together, the alternative theories of corporate compliance suggest that corporate motives
vary, and hence deterrence messages have variable effects. Some firms - the "amoral
calculators" or "bad apples" (Bardach & Kagan, 1982:64-66) are responsive only to the threat of
imminent legal sanctions (general and specific deterrence), while "good apples" respond
primarily to social pressures and felt normative obligations (Malloy, 2003). For others, a
combination of "fear" and "duty" may be operative;2 they regard it as both prudent and right to
commit to a policy of full compliance with governmental regulations. The legally binding
character of regulations alone implies both a threat and an obligation. For firms responsive to this
"implicit general deterrence," learning about legal sanctions against other firms does not
motivate them to comply, but reminds them of preexisting commitments to comply, perhaps
impelling them to intensify audits of their established compliance routines.
2 Research on individual taxpayers has indicated that "fear" and "duty" tend to interact in producing compliance
with income tax law (Schwartz & Orleans, 1967; (Scholz & Pinney, 1995). However, another study found that
among taxpayers with a similar sense of duty, those who had lower fear of being caught (greater opportunity to
cheat) had lower levels of self-reported compliance - indicating that "fear' has independent effects (Scholz &
Lubell, 1998).
50
-------
5/16/2004
There is still one more way in which general deterrence messages may matter. Chester
Bowles (1971:25), reflecting on his job as head of the U.S. Office of Price Administration during
World War II, opined that 20 percent of the population would comply with any regulation, 5
percent would attempt to evade it, and the remaining 75 percent would go along with it as long
as the 5 percent were caught and punished. Officials in other regulatory agencies often echo that
theory, arguing that penalizing the "bad apples' helps keep the "contingently good apples" good
(Bardach & Kagan, 1982). This suggests that explicit general deterrence messages often matter
not because of the threat they signal but because they reassure companies that make costly
compliance-related investments that they will not be at a competitive disadvantage vis-a-vis
firms who violate the law.
II. The Research Project
In light of the complex and varied behavioral pathways suggested by the sociolegal research,
a basic empirical puzzle remains: to what extent is "explicit general deterrence" salient and
important in shaping corporate environmental behavior? Motivated by that question, this
research project was designed to seek preliminary answers to these more specific questions:
1. When a tough legal penalty is imposed against a particular violator, how loud is the
"deterrence signal" it sends? That is, how widely is it publicized, and to what extent do
other companies in the same industry actually learn about it and remember it (or other
penalty cases)?
2. To what extent does hearing about the "signal case" (or other penalty cases) change the
compliance-related behavior of other firms?
3. In stimulating compliance, how salient are the "explicit general deterrence" messages sent
by formal legal sanctions against other firms, compared to (a) the "specific deterrence"
engendered by inspections of and legal sanctions against the firm itself, and (b) the
"implicit deterrence" message sent simply by the dissemination of governmental
regulations?
4. Compared to legal deterrence, how salient are other factors - such as the threat of informal
economic and social sanctions, or normative commitments to compliance with laws and
regulations — as stimuli for compliance efforts?
5. To what extent do motivations vary across firms - depending, for example, on the type
and size of organization, or the characteristics of particular industry sectors?
51
-------
5/16/2004
To begin to answer those questions, we proceeded in several steps. First, we identified a
population of "penalty cases", based on press releases (n =112) issued by the US EPA between
January 2000 and June 2001.3 Second, after selecting a stratified random sample of 40 such
press releases,4 we searched a variety of news media databases to determine the breadth of
coverage the media accorded the cases and penalties described in the EPA press releases.5
Third, we sought to assess to what extent other firms had "heard" the deterrence signals
presumably sent by those (or other) penalty cases, and whether other firms changed their
environmentally-relevant behavior in response. To that end, we conducted an 8-industry survey,
organized by selecting 8 of the 40 "signal cases" whose media dissemination we had tracked.
The 8 cases were chosen non-randomly to include a range of industries, localities, and penalties
- such as a southern state aluminum fabricator in that repeatedly had discharged pollutants in
excess of its permit limits (and was fined $1.1 million) and a California wastewater treatment
plant official who had tampered with monitoring equipment to shield discharges that bypassed
the treatment system (and was sentenced to 5 months in prison). The 8 cases are listed on Table
1, Appendix A). For each signal case, we identified business firms in the same industry and
state.6 After selecting a random sample of such facilities, we telephoned the "person responsible
for environmental compliance" at each. Officials in 233 facilities agreed to be interviewed, a
response rate of 80%.7 Approximately 70% of the facilities whose officials we interviewed had
3 We included only press releases of completed enforcement actions (for example, we did not include those simply
announcing a prosecution) and excluded those involving "wholly illegal enterprises,' such as firms that operated
entirely outside the law (midnight dumpers, unlicensed businesses). In truncating the period, we sought to
concentrate on actions that were relatively more recent, so that respondents might have a better chance of
remembering them, but not so recent that news of them might not have had time to circulate in the industry
4 The sample was stratified to ensure we would have a mix of criminal and civil cases, and those in which the
assessed legal penalty was against individual corporate officers as well as the company
5 We searched for media coverage via Lexis-Nexis, major newspapers, local newspapers, radio and television news
transcripts, industry news outlets, newswires and regional newspaper files. Of course, industry officials can and
often do get news of penalty case from other kinds of sources as well, such as newsletters and direct
communications from legal counsel, suppliers, customers and competitors. For us, it was feasible to survey coverage
of the penalty cases only in more public, on-line news dissemination sources
6 We compiled a list of facilities by searching EPA's Envirofacts database for facilities in the same state and SIC
code as the signal facility. In addition, Switchboard.com and Yellowpages.com were searched for additional
facilities in the same industrial categories as the signal facility, as was Hoovers.com. Where available, state
databases of the relevant facilities were obtained.
7 Response rates were 100 percent for sanitary treatment facilities (n=40 in Florida, 39 in CA), 76% for aluminum
fabricators (26/34), 75% for steel fabricators (30/40); 73% for chemical manufacturers and blenders (29/40), and
69% for Colorado Electroplaters (22/32), 75% for asbestos abatement companies in New York (24/32) and 70% for
52
-------
5/16/2004
fewer than 100 employees; and only those in the chemical industry had a significant (25%)
proportion with more than 1000 employees.
After obtaining general information about respondents and their firm,8 we explored
respondents' knowledge of the signal case and of other enforcement actions against other firms.9
Other questions sought to assess respondents' perception of various legal risks associated with
regulatory enforcement. For example, based on a hypothetical situation modeled on the
regulatory violation in the signal case, respondents asked for their estimate of the likelihood of
detection, and if detected, the likelihood and severity of the resulting legal penalty.10 Finally,
respondents were asked if hearing about a fine or prison sentence at another company in their
industry ever induced them to (1) review their environmental programs; (2) change their
management plans or monitoring methods; (3) change their employee training; or (4) change
their equipment or other aspects of their physical plant. We regarded a company as having "taken
an environmental action" if they reported having taken any of the actions listed above. (For a
fuller account of this phase three research, see Thornton, Gunningham & Kagan, 2004).
The fourth phase of the project entailed longer in-depth interviews with officials at 17
chemical manufacturing facilities and 17 electroplating facilities in the states of Washington and
chemical manufacturers in Louisiana (23/33). Although these are unusually high response rates for survey research,
the possibility remains that nonresponding firms are more likely, on average, to have responded to deterrent
messages differently than those that did respond. At the time of the research, it was not possible to consistently
compare compliance records of nonresponding and responding firms, and in any case, our dependent variable in this
study is not change in compliance but whether hearing general deterrence messages impelled firms to take measures
to improve their environmental performance.
8 Data was gathered on company size and what percentage of his or her time the respondent spent on environmental
work (degree of environmental professionalism)
9 Awareness of the signal case was obtained by presenting a vignette based on the signal case and asking of
respondent had heard of such a case. Respondents were also asked (a) how many instances they could recall in
which, during the last year or two, a company or individual had been fined (or incarcerated) for environmental
violation, and (b) to describe as many particular infractions and penalties a they could.
10 For example, chemical manufacturers and blenders in Louisiana were asked: "Assume for a moment that there
was a chemical manufacturing plant that released CFCs into the air, 35% in excess of their permit limits, and then
repeatedly failed to locate or repair the leaks that led to this excess. On a scale of 0 to 100, what do you think the
chances are that the plant would be found out by law enforcement? "If they were found out, on a scale of 0 to
100, what do you think the chances are that the plant would be fined? "Can you give me a ballpark estimate of
how much they might be fined?" (The latter two questions were then asked with respect to individual fines against
plant operators/owners, the likelihood of incarceration, and the likelihood that penalties might result in plant
closure.
53
-------
5/16/2004
Ohio, exploring the role of general deterrence messages as compared with specific deterrence,
social pressures, and normative beliefs in shaping facilities' environmental behavior. (For a fuller
account, see Gunningham, Thornton, Kagan 2004). Of the eight industries surveyed in the third
phase of our project, the chemical industry had a greater proportion of large firms, and
electroplating a large proportion of small firms.
III. Findings
A. Media Coverage of the Signal Case
Despite their seriousness, the 40 "signal cases" we selected from EPA press releases did
not generally get widespread publicity in the news media. Only 10 of the press releases received
"wide" media coverage (16 to 145 stories) and 14 cases received "low' media coverage (0-6
stories).11 The apparent threshold for obtaining wide media attention was an unusually large fine
(in excess of $4 million) or an unusually long jail sentence (e.g., 17 years) which occurred in
only one case.
B. How Loud the Deterrent Message? Knowledge of Legal Penalties against Other Firms
When a specific "signal case" —was described to representatives of other firms in the
same industry, only 42% of 233 respondents recognized and remembered it.12 That is, a majority
of officials responsible for compliance either hadn't heard of a serious penalty for a serious
offense against a similar firm in their own state, less than two years earlier, or else they had not
regarded it as sufficiently relevant or important to remember. This lends support to a "weak
signal/weak threat" hypothesis.
On the other hand, general deterrence seems to have a cumulative effect on the
consciousness of regulated companies: 89% of our respondents remembered at least one instance
of some company having been penalized for an environmental violation in the past year or two,
and 71%) could describe at least one particular example of a person or business being penalized
11 We compared only the wide and low media coverage groups in our analysis to avoid classification errors. The
remaining 16 cases had intermediate coverage" (7-15 stories).
12 All other things being equal, electroplaters (71%) were significantly more likely to recognize the signal case than
were respondents from any other industry. The more professionalized the environmental staff person, the more
likely they were to remember the signal case (Logistic Regression: Recall spc. ex.= f(#employees,
professionalization, industry). No other variables were significant)
54
-------
5/16/2004
for an environmental offense. Nevertheless, their knowledge was limited and vague.
Respondents report having heard of far fewer fines than actually occur.13 In terms of particular
cases respondents described, they tended to remember only those with unusually large financial
penalties and/or cases where someone was sentenced to jail.14 At the same time, respondents
overwhelmingly underestimated the actual penalties when the signal cases were presented as
hypotheticals, and a significant minority of respondents could not recall any particular instance
of a penalty against an individual. Clearly, then, while respondents generally were conscious of
the possibility of a significant penalty, they do not make special efforts to obtain timely and
accurate information.
C. Perception of Legal Risk
Most respondents thought that serious infractions, such as those described in the signal
case, would be detected; the median perception of detection risk was 70%. However,
respondents' risk-of-detection perceptions were highly variable, ranging from close to 0 to 100%
in most industries. Respondents generally felt that if a serious infraction resembling the signal
case were to be detected, the offending company would be penalized; 92% of respondents felt
the odds of a company fine were greater than 50:50. But 7% of respondents believed there was
no possibility that an individual owner or operator would be fined personally, while only 11%
believed he would certainly be fined. The median risk-of-individual-fine perception was 40%.
Respondents were even less certain that an individual would be incarcerated: 53% of respondents
believed that the chance that an owner or operator would be incarcerated for a serious
environmental infraction was 10% or less.15
Most respondents thought it unlikely that environmental penalties would result in the
closure of an offending facility.16 But expectations of the magnitude of company fines varied
13 For example, the median number of fines against other companies (anywhere in the United States, in the last year
or two) that respondents could recall was only eight. Yet in Louisiana alone, in a 1-year period (July 2001 through
June 2002), 31 companies were fined for environmental infractions. Five of the 31 exceeded $100,000.
14 Of the 107 respondents who gave a magnitude estimate, 43% cited fines of $1 million or more, 67% cited fines of
$100,000 or more, while 26% of respondents who could describe a specific enforcement action noted that someone
at the other company had been incarcerated.
15Electroplaters perceived the risk of incarceration as higher than did all other industries (median probability is
above 50%, while for all other industries it is at or below 20%).
16 For 50% of respondents, there was no chance that environmental penalties would eventually lead to facility
closure, and 85% believed the probability of such a closure was 10% or less. But for the remaining 15% of
55
-------
5/16/2004
widely (from $0 to $20 million) as did estimates for owner/ operator fines ($0 to $2 million).17
Based on a hypothetical modeled on the signal case, many respondents (68/223) could offer no
estimate of the magnitude of the likely company fine. For those that guessed, 68% of
respondents underestimated the fine actually imposed by an order of magnitude, 28% gave an
estimate of the same order of magnitude, and 4% overestimated the fines by an order of
magnitude.18 On the other hand, after being told the actual penalty in the signal case, 85% of
respondents felt that the punishment in the case was reasonable. Of the respondents who felt that
the penalty had been unreasonable, slightly fewer than half (40%) felt that the punishments given
were unreasonably stringent while the remainder (60%) felt that the punishments were too
lenient. This support for tough legal sanctions against firms that had committed serious
violations is consistent with the notion that publicized penalties serve a "reassurance function"
for firms that regard themselves as compliant "good apples."
D. The Effect of Knowledge on Perception of Legal Risk.
There was no strong association between (a) knowledge of enforcement actions against
other firms and (b) our measures of respondents' perceptions of the risk of detection and
punishment. Five linear regression analyses were performed, each modeling a risk perception
variable (likelihood of facility closure, detection, company fine, jail, individual fine) as a
19 20 21
function of company size, degree of professionalization, knowledge (general deterrence)
and industry. All models were statistically significant but did not, in general, explain a large
portion of the variation (see adjusted R2).22
respondents, the risk of forced closure was real, and in a very few cases, substantial. Electroplaters and asbestos
abatement companies were more likely to think that fines might lead to facility closure. In fact, none of these
respondents felt that the probability of facility closure was zero. Conversely, the vast majority of sanitary treatment
facility respondents in both California and Florida deemed closure impossible, which seems a reasonable assessment
given the indispensability of their function. Some chemical manufacturing facilities viewed the probability of
facility closure as reasonably high, while most aluminum fabricators and steel fabricators viewed it as highly
unlikely.
17 Fifty percent of respondents believed that if a company official were incarcerated, the length of the sentence
served would be 6 months or less. The longest period of incarceration envisaged was ten years.
18 Those respondents that had heard of the signal case also tended to underestimate the fine, but less often (59%)
than those who had not heard of the signal case (74%).
19 Company size is divided into "large" (100 or more employees) and "small" (less than 100 employees).
20 Measured as a percent of their time the respondent spent on environmental work.
21 Three different measures used. First, the quantum of fines recalled, categorized as: none, one or two, three to 9, 10
to 15, 16 to 30, more than 30. Second, the number of particular cases recalled and described (none, one or two), and
third, whether or not the signal case was recalled.
22 Reference industry=sanitary treatment facilities in Florida.
56
-------
5/16/2004
E. Compliance-Related Behavior
A majority of companies (65%) reported that they had increased their compliance-related
activity in some way after hearing about a fine or prison sentence at another company. Thus it
appears that general deterrence messages, at least cumulatively, do matter. Our questions did not
distinguish whether or not it was knowledge of the signal case (as opposed to other penalty
cases) that triggered responsive environmental action. But employing a series of assumptions, we
can estimate that 10 to 20% did respond to the signal case.23
The most commonly-reported responsive action (57% of respondents) was to review existing
environmental control programs. But 23% changed their employee training, and 32% reported
having changed equipment, suggesting that that a substantial fraction of facilities respond in
potentially expensive ways to environmental enforcement actions taken against other firms.
F. Knowledge, Perception of Legal Risk, and Behavior
What distinguishes firms that do and do not report environmental actions in response to
deterrence messages? A logistic regression model of company environmental action as a
function of demographic, knowledge, risk perception, and other variables was developed. Table
3 (see Appendix) presents descriptive statistics for the variables employed in the model. Table 4
presents the results of the logistic regression. Company size was significantly and positively
associated with the likelihood of taking environmental action. The degree of professionalization
variable was not significantly associated with taking environmental action. Interestingly,
respondents who could describe more particular examples of enforcement actions against other
firms were more likely to report having taken an environmental action in response to deterrence
signals. On the other hand, remembering the signal case, or remembering a larger number of
instances of enforcement actions, were not significantly associated with taking environmental
Facility Closure: df=190, F=10.168, p<000, Adj R2=0.367, Sig Vars: Asbes, Elec, Steel, Chem-KY(all +ve)
Detection: df=195, F=3.679, p<000, Adj R2=.142, Sig Vars: Chem-KY (+ve)
Company Fine: df=193, F=1.851, p=.043 Adj R2=.050, Sig Vars: Particular Exs (+ve)
Individual Fine: df=185, F=2.390, p=.007, Adj R2=.083, Sig Vars: Steel (-ve)
Jail: df=186, F=3.532, p<000, Adj R2=.140 Sig Vars: Elec (+ve) SanTx-CA (+ve)
23About 65% of facilities reported they had taken an action in response to hearing about some legal penalty against
some other company. About 40% had heard of the signal case. If we assume that the action-in-response rate (65%) is
the same for that 40% of firms that were attentive to the signal case, then perhaps 24% of facilities took
environmental action in response to the signal case. Since that may overestimate the signal case response, our guess
is that 10-20% would be more realistic
57
-------
5/16/2004
actions. And firms that thought that the risk of detection and the magnitude of legal penalties
were great were, on average, no more likely to take responsive environmental action than firms
with lower estimates of legal risk.24 Associations among variables are summarized in Figure 1.
Figure 1: Summary of Results
Environmental Protection
Actions (in response to
general deterrence
messages)
Perceived Risk of
Detection and
Company Fine
Perceived
magnitude of
company fine
Perceived Risk of
Facility Closure
Knowledge of Signal
Case
Knowledge of Many
Enforcement Cases
Recall of Particular
Examples of
Enforcement Actions
Indicates statistically significant association
IV. Discussion
Classic deterrence theory predicts clear relationships between knowledge of "high
profile" enforcement actions (fines and incarcerations) and improved compliance-related
behavior. In most descriptions of the theory, (1) regulated entities are presumed to monitor their
environment for information about enforcement activity and to have heard about high profile
prosecutions and penalties; (2) knowledge of high profile cases is presumed to increase
perceived risk of non-compliance; and (3) higher perceived risk of legal sanctions is presumed
to improve overall compliance-related behavior.
Our results from the 8-industry, 223-firm survey provide only limited support for this
theory. The majority of firms (65%) report having, at some point in the past, taken an
environmental action in response to hearing about an enforcement action at another company.
But we find only a weak association between increased information about other penalty cases
24 Of the risk perception variables, only the perception that penalties might lead to facility closure was significantly
associated with taking an environmental action. However, this result appears to be driven by the electroplating
facilities in the sample, and is no longer significant (p=0.095) if electroplating cases are excluded from the dataset.
58
-------
5/16/2004
and increased perception of legal risk, and firms with higher risk perceptions of detection or fine
were not significantly more likely to have taken an environmental action than those with lower
risk perceptions. Moreover, contrary to the assumptions of general deterrence theory, firms did
not obtain or retain accurate information about the frequency or magnitude of fines and other
penalty information.
This seeming inconsistency25 may reflect the previously-mentioned theory that general
deterrence signals often serve not to enhance fear of sanctions per se but to remind reasonably
'good apples' - firms already committed to compliance as a general business strategy - that
noncompliance can occur due to slippage in their company's own self-regulatory systems. This
would explain the finding that the most common response to news of sanctions against other
firms is to review one's own compliance program. And on occasion, a deterrence signal will
inform a good apple of non-compliance (or risk of noncompliance) in their own facility
(stemming, e.g., from employee error or deviance, or with respect to a regulation they were
unaware of or had interpreted incorrectly); hence the signal will spur them into more than simple
confirmation routines. In this way, information could affect behavior without changing risk
perceptions.26 Similarly, the examples of noncompliance cited by respondents were often
couched in judgmental tones, critical of the behavior of the company punished. This supports the
notion that explicit general deterrence messages serve a "reassurance function," informing
already compliant firms that they are not foolish for doing so, since their competitors who
"cheat' are getting caught and punished.
V. What The In-Depth Two-Industry Survey Adds
25 The puzzle may reflect the possibility that the measures we constructed do not accurately reflect the underlying
constructs. For example, we asked respondents, "If a company is violating in this manner, what do you perceive the
risk of detection or punishment for that company to be?" However, we could not sensibly ask respondents how
likely they were to commit the same violations, and their response may reflect their estimate of risk of detection for
"bad apples" (not their own firm). Nor did we directly ask: Did hearing about an enforcement action at another
company ever change your perception of risk? Furthermore, we obtained only a snap shot of current risk
perceptions, but asked for an aggregate measure of behavior change, asking if companies had 'ever' taken
environmental actions in response to deterrence signals. Our measure thus does not rule out the possibility that firms
that acted in response to deterrence signals had higher risk perceptions at that prior time.
26 Such a reminder function of deterrence comes through quite dramatically in our in-depth interviews in the
electroplating and chemical industries. See Sec V, below
59
-------
5/16/2004
The two-industry survey — longer, in-depth interviews of 34 firms in 2 of the 8 industries
(chemicals and electroplating) — sought a richer contextual understanding of what motivates
management and how regulatees think about and respond to deterrence and to regulatory and
social scrutiny more generally.27
Specific deterrence. Specific deterrence in its narrowest sense - a previous sanction
against a company inclining it to make more strenuous efforts to avoid future penalties - had a
significant impact on a substantial minority of companies in our sample, particularly smaller
firms. Twenty-four per cent (4/17) of electroplaters and 11% (1/9) of chemical small-or-medium-
sized enterprises (SMEs) said that a legal penalty against their company in the past had
influenced its subsequent environmental actions. But the large chemical companies in our
sample, who reported having had only minor violations over the last decade, had experienced no
significant enforcement. For them, therefore, specific deterrence was not a salient driver of
environmental actions.
Specific deterrence in its broader sense also includes the impact of inspections (with their
implicit threat of sanctions). For electroplaters, inspections played an important role, prompting
them to undertake whatever action was required of them in the belief that further enforcement
action, with potentially profound consequences, would have followed from continuing non-
compliance. Inspections also had an important"reminder function " for firms inclined to comply
because they said it was the 'right thing to do.' Again, however, chemical companies said that
inspections did not have a significant influence on them; only one identified inspection as an
important reason for taking particular environmental actions. Most stated that they were already
substantially beyond compliance, and so inspections held no fear for them.
Explicit General Deterrence. Knowledge about legal sanctions against other companies,
according to our interviews, played only a very modest role in the case of electroplaters and an
even smaller one for chemical companies. In the case of the former, only 12% (2/17) said a fine
or prison sentence at another company had influenced specific environmental actions (less than
the average of our 8 industry survey). Only 1/17 saw general deterrence as a powerful motivator
for specific actions; 11/17 saw it as a relatively unimportant motivator. Among chemical SMEs,
27 For a fuller account, see Gunningham, Thornton & Kagan, 2004
60
-------
5/16/2004
no one identified an environmental action that occurred against another company as having
influenced particular environmental actions in their facilities. However, when prompted, many
felt that hearing about another firm being penalized might influence them if the circumstances
were sufficiently similar. Large chemical companies reported that they were not at all influenced
by such considerations.
There seem to be three reasons why the impact of explicit general deterrence was small.
First, companies had great difficulty comparing their own circumstances with those of the
company that had been penalised, and most commonly dismissed the latter as being irrelevant
(see also Braithwaite & Makkai, 1991). Second, the very large majority of our respondents
claimed to be in compliance or even "beyond compliance." In these circumstances, hearing about
punishments imposed on recalcitrants did not resonate with their own circumstances and
triggered little fear in them. Third, some respondents suggested that it was only hearing about
someone in similar circumstances going to prison, rather than merely being fined, that would
influence them.
However, as in the 8-industry survey, explicit general deterrence did have a significant
reminder function for both electroplaters and chemical companies - prompting them to review
their own operations and think about environmental risks that otherwise might not have gained
their immediate attention. Nevertheless, few reported making any significant changes as a result
of such a reassessment.
Explicit general deterrence also fulfilled a reassurance function. Many respondents
conceded that without effective enforcement, the overall performance of the industry would
decline over time, as compliant firms would lose confidence that there was a 'level playing field'
in terms of environmental standards. Many respondents placed considerable emphasis on this
function, as complaints about enforcement commonly focused on the injustice of others not
being punished, or not being punished heavily enough.
Implicit General Deterrence. For these respondents, what we have called "implicit
general deterrence" - the threat of legal sanctions implied by the mere promulgation or history of
enforcement of laws and regulations in the contemporary United States - was much more salient
than either specific deterrence or explicit general deterrence. Although many of our respondents
acted for instrumental reasons, they did not seem to engage in any careful weighing of the
61
-------
5/16/2004
benefits of non-compliance versus the probability of being discovered and punished, as predicted
by traditional deterrence theory. On the contrary, almost all our respondents gave the impression
that there was no point even debating whether to comply or not. Compliance was regarded as
mandatory. Electroplators and chemical SMEs saw legal punishment of serious violations as
virtually inevitable.28
Our interviews indicate that "implicit general deterrence" arises from the general history
of a particular regulatory regime (in this case targeted enforcement over the previous decade). In
these industries inspection and enforcement activity have generated a 'culture of compliance',
such that it becomes almost unthinkable to regulatees that they would calculatedly (as opposed to
inadvertently) break the law. Most of our respondents took a similar view to EWs-7: "It's
ludicrous to let things go and imagine you won't get into trouble... We are subject to inspection
and to fines, huge fines, for not doing it. You can't fight that. You either comply or get out of the
business." Thus it was the regulations themselves (rather than hearing about enforcement actions
against other firms) that had the most direct impact on behavior. But that occurs against a
backdrop where the common perception was that 'you go out of business if you don't comply.'29
For large chemical manufacturers, however, the mechanisms that led to compliance were
rather different. Such firms commonly described regulation as only 'the baseline,' implying that
it was a taken-for-granted minimum standard which they would usually substantially exceed for
a variety of reasons discussed below. For them, regulation was taken for granted not because of
the perceived inevitability of sanctions (that is, implicit general deterrence) but because they felt
28
Electroplaters voiced this sense most strongly, which may reflect enforcement actions these facilities had
experienced in the past: 8/17 electroplating companies mentioned previous violations, fines, jail sentences, or threats
of facility closure. Every electroplating facility was regularly inspected at least once a year: by the local sewer
district if they had a discharge to the sewer, plus by the fire department, plus by state and federal environmental
agencies. However, even smaller chemical companies (another industry subjected to substantial regulatory scrutiny
and penalties in the past) commonly voiced a similar sense of "regulatory inevitability. "
This sense of regulatory inevitability was reinforced by the widespread perception among respondents that it
was firms 'like theirs' who were most vulnerable to inspection and enforcement. Thus large firms believed that
small firms were 'getting away with it' while they themselves were not, while the converse was the perception of
small enterprises.
29
Indeed, for many interviewees, the regulations had become so embedded in their culture that they exerted an
almost unconscious influence on decision-making. Some respondents attributed legally required environmental steps
at their facilities not to regulation but to the firms' environmental ethos, seemingly oblivious to the extent to which
they operated in a thick regulatory soup which constrained many of their choices.
62
-------
5/16/2004
a failure to comply would send very undesirable signals to important stakeholders, triggering a
variety of informal sanctions. Yet the law was seen as a salient standard in the minds of their
investors, employees, customers, and local governments; hence they had to attend closely to
legal compliance.
Such instrumental considerations, even in the more complex form of implicit general
deterrence, were not the only ones that weighed upon our respondents. Almost half of our
respondents also provided a range of normative explanations for why they complied. In essence,
many of them perceived themselves as 'good guys', complying with environmental regulation
because it was the right thing to do. However, they struggled to disentangle normative from
instrumental motivations, and wrestled with the temptation to backslide when environmental
improvements proved expensive. In the absence of regulation and implicit general deterrence, it
is questionable whether their good intentions would have translated into practice.
In any event, deterrence in any form was of far greater concern to SMEs than it was to large
ones. For major reputation-sensitive firms in the environmentally sensitive chemical industry,
regulation and its enforcement played only a minor role ('as a baseline') and most chose to go
substantially beyond compliance for reasons that related to risk management considerations and
to the perceived need to protect their social license to operate. Crucial in this regard was
maintaining the trust and support of local communities, of avoiding the attention of
environmental groups and other potentially critical stakeholders, and of preserving the
company's reputation as an environmentally responsible entity (see Gunningham et al, 2003).
Large companies appeared to differ from the smaller companies in terms of how they went
about complying or over-complying. In their responses, they treated regulation and liability rules
as sources of substantial additional costs, and hence as economic signals - to which they
responded by seeking out solutions that substantially mitigated those costs and occasionally even
saved them money overall. In this regard, they were proactive and innovative in a way that
boundedly-rational small companies, particularly electroplaters, most certainly were not.
Thus there are various strands that must be taken into account in understanding what
motivates corporate environmental behavior. There is a tight coupling for example, between
normative and instrumental explanations for compliance. Even those who see themselves as
'good guys' and who comply because it is 'the right thing to do,' suggest they would be reticent
to do so if they are not confident that the 'bad guys' are being effectively regulated and
63
-------
5/16/2004
sanctioned. Similarly, there is a connection between informal social pressures and formal legal
ones. Because the law is seen by many (including local communities) as a moral barometer, any
company found in non-compliance risks not only legal sanctions but the informal stigma and
reputation damage that the community and other stakeholders may inflict.
Finally, how these various strands play out depends very much on the size and
sophistication of companies themselves and on the characteristics of the industry sector within
which they are located. Electroplaters responded very differently to various external drivers than
did chemical companies, and even within the latter, small and medium sized companies were
influenced by substantially different considerations from large companies. Overall, there was
little support for models of business firms as "amoral calculators," who carefully weigh the
certainty and severity of sanctions and who can be manipulated through a judicious mix of
specific and general deterrence.
VII. Conclusion
Our research provides only weak support, at best, for the classical "general deterrence'
hypothesis (which we would now label 'explicit general deterrence"). Many EPA-imposed legal
penalties, especially the less severe ones, do not get substantial coverage in the newsmedia.
Fewer than half (42%) of 229 respondents in our 8-industry survey recognized and remembered
the specific signal case. On the other hand, general deterrence seems to have a cumulative effect
on the consciousness of regulated companies: most respondents thought the risk of detection of
violations was high, and for many in our in-depth study, virtually inevitable. In the 8-industry
survey, 89% of our respondents remembered at least one instance of some company having been
penalized for an environmental violation in the past year or two. And some 63 percent reported
having taken some environmental protection measures after learning about penalties against
other companies. Most often, the reported reaction was to review their own compliance
programs, but a substantial minority changed equipment, monitoring practices or employee
training.
Yet many relationships predicted by the classical deterrence model did not appear in our
data. Respondents who recognized the signal case or recalled a larger number of other cases
were not more likely to report having taken environmental action in response. Officials who saw
the risk of formal detection and punishment as high were not, on average, more likely to report
64
-------
5/16/2004
taking environmental measures in response to general deterrence messages. Company managers
were not closely attentive to the penalties assessed against violators, generally underestimating
them. This suggests to us that penalties against other firms - at least in the United States near the
beginning of the 21st Century — play a somewhat different role from the one embedded in the
classical general deterrence theory, which assumes that the imminent threat of legal punishment
is the primary driver of compliance efforts.
Our survey as well as the in-depth interview evidence, rather, suggests that for most
firms, general deterrence primarily serves a reminder and a reassurance function. For the "good
apples" - firms that are generally committed to compliance for a variety of normative and
reputational reasons - learning about penalties against other firms reinforces their perception of
the need to continue compliance activities and of the potential disastrousness of non-compliance.
Sometimes, a deterrence signal prods them to check and take further action. Deterrence signals
both reassure 'good apples' that free-riders will be punished and remind them to make sure that
they are responsible corporate citizens with no need to fear the social and economic costs that
can be triggered by serious violations.
It must be remembered, however, that this research was conducted in the United States in
the early 21st Century, more than a quarter century after American states and the federal
government started serious enforcement of environmental laws. Hence the "implicit general
deterrence" mechanism has matured, so that the enforcement and normative legitimacy of
environmental regulations is taken for granted by many firms. And social and political support
for environmental norms has given many companies a substantial economic stake in avoiding a
reputation for being bad environmental citizen. Thus our research has little to say about the
importance of explicit general deterrence messages at earlier stages in regulatory programs,
when their value added may well be greater, or for firms (or industrial subsectors) that are
deliberate evaders or chronically at the edge of or out of compliance.
65
-------
5/16/2004
Appendix A
Table 1. Signal <
^ases
Industry
Infraction
Penalty
Company Fine
Jail Sentence
Individual
Fine
Electroplating, CO
The VP of a Denver plater, who, despite
56 warnings over 10 years allowed Zn,
Cd, Cu, Cr, and Ni to be continually
discharged into the Denver municipal
sewers.
$250,000
12 months +
lOOhrs
community
service
Waste Water
Treatment, CA
The district manager of a Rodeo,
California treatment plant who admitted
to allowing wastewater to bypass a
chlorine contact chamber and to
tampering with monitoring methods on
473 days between 1995 and 1997.
5 months prison
+ 5 months
home
confinement + 1
year probation
$3000
Chemical
Manufacturing or
Blending, KY
In 1995, a plant in KY stored fuming
sulfuric acid in a tank that had cast iron
piping instead of steel piping. The iron
corroded, and the company did not
inspect the piping. This resulted in about
24,000 gallons of sulfuric acid solution
being released into the air in a four-hour
period, creating a chemical cloud. A
thousand nearby residents had to be
evacuated and several were treated for
burns of their eyes, nasal passages and
lungs.
$850,000 penalty
+
$650,000 on an
emergency
notification
system
Aluminum
Fabrication -
Southern States
An aluminum fabricator in Port Allen,
LA, who discharged wastewater
contaminated with hexanol and with a
COD of 1,737 ppm (13X their permit
limit) into an intercoastal waterway
$1.1 million
5 years probation
100 hours of
community
service
$2000 to
$5000
Waste Water
Treatment, FL
South Bay Utilities of Sarasota county,
who discharged an estimated 290 gallons
of inadequately treated wastewater, along
with additional periodic discharges
amounting to 1.5 tons of nitrogen in a two
year period, into Dryman Bay.
$1.3 million
$445,000
(president
of the
company)
Steel Fabrication,
IN
A corporation that settled allegations that
it failed to control the pollution at eight
steel minimills, resulting in thousands of
tons of illegal air emissions of NOx, and
mismanaged discharges of K061 dust in
the soil and groundwater. The company
contends that it had not violated any
environmental law.
Civil penalty of
$9 million
$4 million on
environmental
projects
$85 million on
new control tech.
Asbestos
Abatement
Services, NY
While carrying out an asbestos abatement
project, between December 1997 and
March 1998, the company failed to notify
the EPA; knowingly sent workers into an
asbestos "hot zone" for more than 12
weeks, without providing them with
41 months
$59,700
restitution
66
-------
5/16/2004
Industry
Infraction
Penalty
Company Fine
Jail Sentence
Individual
Fine
protective gear, or even informing them
of the presence of asbestos; failed to have
a certified contractor perform the work, to
properly wet and bag the asbestos, to
properly label the containers filled with
asbestos, and to dispose of the asbestos at
a landfill approved for that purpose.
Chemical
Manufacturing, LA
A chemical company in Westlake, LA
was charged with releasing CFCs into the
air in excess of the 35% limit and then
repeatedly failing to locate and repair
leaks.
$4.5 million
penalty and
Fund an
"environmental
justice" project in
Westlake, LA
Table 3: Responses to General Deterrence Messages: Descriptive Statistics
Valid
Missing
look cn\ ironmcnlal
ad ion in response lo
(k'lcrivnco signal
"1 "1 -
6
Compain si/o
224
9
Large (>100
employees)
27%
I'mTiil lime spent on
cii\ironmcnlill work
228
5
0 to 25%
33%
26 to 75%
33%
(u'calcr llian "5".,
No of instances of
comp;in\ linos
recalled
5
0
11%
1
6%
2-5
25%
6-10
18%
>10
39%
\1a\imiim
Renumber a
particular example
2'2
i
u
29%
1
45%
2
26%
Heard of (ho signal
Case
42".,
22')
4
26-75%
41%
"(i-|()()"„
Probability of
Company Hue
22(>
-
0-25%
4".,
26-75%
12%
76-100%
84%
Risk--
225
s
0-2500
28%
2501-7500
42%
7501-10000
30%
Magnitude of
( ompain l ino
(dollars)
1 •;<.
37
0
1%
Thousands
9%
Tens of thousands
38%
Hundreds of
thousands
18%
Millions or more
34%
Prohahilil.t of
l-'acilil\ Closure
2I<>
14
0
50%
1 to 10
35%
11 to 25
11%
26 to 75
3%
76-100
1%
Prohahilil.t of
DcU'Clion'-
228
5
0-25%
23%
*Probability of Detection= Response to the question: "on a scale of 0 to 100, what do you think the chances are that
the plant (in hypothetical based on signal case) would be found out by law enforcement?" Estimated
67
-------
Probability of Company Fine, Magnitude of Company Fine, and Probability of Facility Closure measures
based on similar question about fate of company in hypothetical based on signal case.
**Risk= probability of detection x probability of company fine
Table 4. Logistic Regression Model of Corporate Environmental Action30
Dependent Variable: Taking environmental action in response to deterrence signals (binary).
B
S.E.
Wald
Df
Sig.
Exp(B)
Demographic Variables
Company size (large/small)
1.254
.491
6.529
1
.011
3.504
Degree of Professionalization
.008
.006
1.838
1
.175
1.008
Knowledge Variables
Number of instances of company fines
.002
.002
.840
1
.359
1.002
Recall particular examples (0.1. or2)
.980
.289
11.516
1
.001
2.665
Recognize signal case
.386
.405
.908
1
.341
1.470
Risk Perception Variables
Risk=prob of detection x prob co. fine
.000
.000
.887
1
.346
1.000
Magnitude of company fine (0,1,2,3,4)
-.013
.184
.005
1
.944
.987
Risk that penalties will lead to closure
.072
.029
6.227
1
.013
1.074
Constant
-1.775
.691
6.608
1
.010
.169
Shaded and italicized results show variables significant at or below a p=0.05 level.
30
Number of cases included in the analysis=176 (=75.5% of all cases). The model chi-square is 50.706 which is
significant at p=<000. The -2 Log likelihood value is 175.150 and the Cox and Snell R Square is .250. A second
model was also run including dummy variables for each industry. The addition of this block of variables was not
significant at a 0.05 level and so these variables were not included in the model (Chi-square=9.812, df=7, p=0.199).
A correlation matrix was calculated. No bivariate correlations exceeded .30.
68
-------
References
Aoki, Kazumasu, Lee Axelrad, & Robert A. Kagan "Industrial Effluent Control in the United States and Japan," in
Robert A. Kagan & Lee Axelrad, eds, Regulatory Encounters: Multinational Corporations and American
Adversarial Legalism. Berkeley, CA: University of California Press, 2000)
Ayres, Ian & John Braithwaite (1992) Responsive Regulation. New York: Oxford University Press
Bardach, Eugene & Robert A. Kagan (1982) Going by the Book: the Problem of Regulatory Unreasonableness.
Philadelphia: Temple University Press [reprinted by Transaction Publishers, 2002]
Bowles, Chester (1971) Promises to Keep (N.Y.: Harper & Row)
Braithwaite, John & Toni Makkai (1991) "Testing an Expected Utility Model of Corporate Deterrence," Law &
Society Review 25: 7-40
Gray, Wayne, & John T. Scholz (1991) "Analyzing the Equity and Efficiency of OSHA Enforcement," Law &
Policy 13: 185-214
Gunningham, Neil, Robert A. Kagan & Dorothy Thornton (2003) Shades of Green: Regulation, Business,
Environment. Stanford University Press
Gunningham, Neil, Dorothy Thornton and Robert A. Kagan (2004) "Motivating Management: Corporate
Compliance in Environmental Protection" forthcoming, Law and Policy.
Kagan, Robert A., Neil Gunningham & Dorothy Thornton (2003) Explaining Corporate Environmental
Performance: How Does Regulation Matter?" Law & Society Rev. 37: 51
Kagan, Robert A. & John T. Sholz (1984) "The Criminology of the Corporation and Regulatory Enforcement
Styles' in Keith Hawkins & John Thomas, eds., Enforcing Regulation (Boston: Kluwer-Nijhoff)
Malloy, Timothy (2003) "Regulation, Compliance, and the Firm," Temple Law Rev. 76: 451-531
Mehta, Alex & Keith Hawkins (1998) "Integrated Pollution Control and Its Impact: Perspectives From Industry," J.
Environmental Law 10: 61
Prakash, Asseem (2000) Greening the Firm: The Politics of Corporate Environmentalism. Cambridge, UK:
Cambridge University Press
Ruhl, J.B., James Salzman, Kai-Sheng Song & Han Yu (2002) "Environmental Compliance: Another Integrity
Crisis or Too Many Rules?' Natural Resources & Environment. 17: 24-29
Scholz, John T. & Neil Pinney (1995) "Duty, Fear and Tax Compliance: The Heuristic Basis of Citizenship
Behavior," American Journal of Political Science 39: 490
Scholtz, John T. & Mark Lubell (1998) "Trust and Taxpaying: Testing the Heuristic Hypothesis," American Journal
of Political Science 42: 398
Schwartz, Richard D. & Sonya Orleans (1967) "On Legal Sanctions," University of Chicago Law Review 35: 283
Thornton, Dorothy, Neil Gunningham, and Robert A. Kagan (2004) "General Deterrence and Corporate
Environmental Behavior," in press, Law & Policy
69
-------
When and Why do Plants Comply? Paper Mills in the 1980s
Wayne B. Gray
Clark University and NBER
Ronald J. Shadbegian
University of Massachusetts at Dartmouth
U.S. EPA National Center for Environmental Economics
For presentation at workshop on
Corporate Environmental Behavior and the Effectiveness of Government Intervention
Sponsored by
EPA's National Centers for Environmental Economics and Environmental Research
April 26, 2004
Financial support for the research from the National Science Foundation (grant # SBR-
9809204) and the Environmental Protection Agency (grants #R-826155-01-0 and #R-
828824-01-0) is gratefully acknowledged, as is access to Census data at the Boston
Research Data Center. Valuable comments were received from Alex Pfaff, Suzi Kerr,
Amanda Lee, and Maureen Cropper, as well as seminar participants in the AERE
Summer Workshop, the NBER Summer Institute, Lehigh University, Center for
Economic Studies, the University of California-Berkeley, and a 2003 AERE-ASSA
session. We are grateful to the many people in the paper industry who were willing to
share their knowledge of the industry with us. Capable research assistance was provided
by Bansari Saha, Aleksandra Simic, Nadezhda Baryshnikova and Melanie Lajoie. The
opinions and conclusions expressed are those of the authors and not the Census Bureau,
EPA, or NSF. All papers using Census data are screened to ensure that they do not
disclose confidential information. Any remaining errors or omissions are the authors'.
70
-------
Abstract
This paper examines differences in plant-level compliance with air pollution
regulation for U.S. pulp and paper mills. We test a variety of plant- and firm-specific
characteristics, to see which plants are more likely to comply with regulation. We also
test how effective regulatory enforcement is in inducing compliance, and whether plants
differ in their sensitivity to regulatory activity.
Our analysis is based on confidential, plant-level Census data from the
Longitudinal Research Database for 116 pulp and paper mills, covering the 1979-1990
period. The LRD provides us with data on shipments, investment, productivity, age, and
production technology. We also have plant-level pollution abatement expenditures from
the Pollution Abatement Costs and Expenditures (PACE) survey. Using ownership data,
we link in firm-level financial data taken from Compustat, identifying firm size and
profitability. Finally, we use several regulatory data sets. From EPA, the Compliance
Data System provides measures of air pollution enforcement activity and compliance
status during the period, while the Permit Compliance System and the Toxic Release
Inventory provide information on other pollution media. OSHA's Integrated
Management Information System provides data on OSHA enforcement and compliance.
We find significant effects of some plant characteristics on compliance rates:
plants which include a pulping process, plants which are older, and plants which are
larger are all less likely to be in compliance. Compliance also seems to be correlated
across media: plants violating water pollution or OSHA regulations are more likely to
violate air pollution regulations. Firm-level characteristics are not significant
determinants of compliance rates.
Once we control for the endogeneity of regulatory enforcement, we find the
expected positive relationship between enforcement and compliance. We also find some
differences across plants and firms in their responsiveness to enforcement. Pulp mills,
already less likely to be in compliance, are also less sensitive to inspections. Some firm
characteristics also matter here: plants owned by larger firms, whether measured in terms
of their employment or by the number of other paper mills they own, are less sensitive to
inspections and more sensitive to other enforcement actions, consistent with our
expectations and with other researcher's results.
71
-------
1. Introduction
In most economic models of government regulation, a regulatory agency establishes
standards with which regulated firms are required to comply. Compliance is usually
accomplished by having inspectors visit plants to identify violations and to impose penalties on
violators. Becker (1968) demonstrated that if both the probability of being caught and the
penalty for violations are high (relative to the costs of compliance), we would expect profit-
maximizing firms to optimally choose compliance. However, for many regulatory agencies, the
number of inspectors is small relative to the regulated population and the penalties are limited, so
there seems to be a limited incentive for compliance - yet most firms still seem to comply.
This puzzle of'excessive' compliance has led to several strands of literature. Outside
economics, researchers have emphasized the importance of social norms and a corporate culture
that encourages compliance, and have conducted interviews to identify how corporate decisions
are affected by pressures from both regulatory agencies and the general public. Within
economics, a model by Harrington (1988) shows that in a repeated game, a regulator could
substantially increase the expected long-run penalty for non-compliance by creating two classes
of regulated firms - cooperative and non-cooperative. The cooperative firms are assumed to
behave well and to be inspected only rarely. The non-cooperative firms would face much
heavier enforcement. Since facing enforcement is costly, firms would be anxious to be placed in
the cooperative group initially, and therefore would invest more in compliance at the start of the
game, than would be predicted from the expected penalty in a one-period model.
On the empirical side, there have been several studies on the effectiveness of OSHA and
72
-------
EPA enforcement, using a variety of estimation techniques. These include studies of
environmental enforcement at steel mills for air pollution (Gray and Deily 1996); at paper mills
for air pollution (Nadeau 1997) and water pollution [Magat and Viscusi (1990), Laplante and
Rilstone (1996), and Helland (1998)]; and of OSHA regulation at manufacturing plants (Gray
and Jones(1991), and Gray and Scholz(1993)). These studies generally find that enforcement
has some effect on compliance, or the goals of compliance (reduced emissions or injuries).
Since enforcement and compliance tend to be defined at the plant level, most of these studies do
not incorporate firm-level variables. However, Helland finds that more profitable firms have
fewer violations, and Gray and Deily find that compliance status is correlated across plants
owned by the same firm, though they find insignificant effects of firm size and profitability on
compliance. Gray (2000) finds little effect of corporate ownership change or restructuring on
compliance and enforcement.
In this paper we use a sample of U.S. pulp and paper mills to examine differences in
plant-level compliance with air pollution regulations. In particular, we test a variety of plant-
and firm-specific characteristics, to see which plants are more likely to comply with regulation.
We also compare the plant's air pollution compliance with its performance in other dimensions
(water pollution, toxic chemicals, and worker health and safety). Finally, we test how effective
regulatory enforcement is at inducing compliance, and whether plants differ in their sensitivity to
enforcement activity.
We use confidential, plant-level Census data from the Longitudinal Research Database
for 116 pulp and paper mills, covering the 1979-1990 period. The LRD provides us with data on
each plant's shipments, investment, productivity, age, and production technology. We also have
73
-------
plant-level pollution abatement expenditures from the Pollution Abatement Costs and
Expenditures (PACE) survey. We link in ownership information, based on the Lockwood
Directory, which allows us to identify the number of paper mills owned by the firm, and also link
in firm-level financial data taken from Compustat, identifying firm size and profitability.
Finally, we add compliance and enforcement information from several regulatory data sets,
although our focus is on the EPA's Compliance Data System, which provides measures of air
pollution enforcement activity and compliance status during the period.
We use a logit model of compliance with air pollution regulation: compliance depends on
regulatory activity directed towards the plant, as well as various plant and firm characteristics.
Regulatory activity is endogenous - regulators target enforcement activity towards plants that are
out of compliance - so a simple correlation between enforcement and compliance would be
negative, indicating (naively) that enforcement decreases compliance. To address this targeting
issue, we try two alternative ways of measuring enforcement. First, we try using lagged
enforcement as an explanatory variable, in principle purging the equations of any
contemporaneous endogeneity. Second, we try predicting enforcement from a tobit model on a
set of variables which are clearly exogenous to the plant's compliance decision (state political
support for environmental regulation and year and state dummies). We then use this predicted
value in a second-stage compliance equation. Models using lagged regulatory activity continue
to find a negative 'impact' of enforcement on compliance (which we attribute to remaining
endogeneity), while models using predicted activity yield positive coefficients, with regulatory
activity increasing compliance.
We find significant effects of plant characteristics on compliance rates: plants which
74
-------
include a pulping process, plants which are older, and plants which are larger are all less likely to
be in compliance. In contrast, firm-level characteristics are not significant determinants of plant-
level compliance rates. Plants violating other regulations (water pollution or OSHA regulations)
are more likely to violate air pollution regulations.
We also find differences across plants in their responsiveness to enforcement. Pulp mills,
already less likely to be in compliance, are also less sensitive to inspections. Finally, firm
characteristics do seem to matter for a plant's inspection sensitivity (though they did not for the
overall compliance rate). Plants owned by larger firms, whether measured in terms of firm
employment or the number of paper mills owned by the firm, are less sensitive to inspections and
more sensitive to other enforcement actions than plants owned by smaller firms.
Section 2 provides some background on environmental regulation and compliance issues
in the paper industry. Section 3 describes a simple model of the compliance decision faced by a
plant. Section 4 describes the data used in the analysis, Section 5 describes some econometric
issues with the analysis, Section 6 presents the results, and Section 7 contains the concluding
comments.
2. Paper Industry Background
Environmental regulations have grown substantially in stringency and enforcement
activity over the past 30 years. In the late 1960s the rules were primarily written at the state
level, and there was little enforcement. Since the early 1970s, the Environmental Protection
Agency has taken the lead in developing stricter regulations, and encouraging greater
enforcement (much of which is still done by state agencies, following federal guidelines). This
75
-------
expanded regulation has imposed sizable costs on traditional 'smokestack' industries, with the
pulp and paper industry being one of the most affected, given its substantial generation of air and
water pollution.
Plants within the pulp and paper industry can face very different impacts of regulation,
depending in part on the technology being used, the plant's age, and the regulatory effort directed
towards the plant. The biggest determinant of regulatory impact is whether or not the plant
contains a pulping process. Pulp mills start with raw wood (chips or entire trees) and break them
down into wood fiber, which are then used to make paper. A number of pulping techniques are
currently in use in the U.S. The most common one is kraft pulping, which separates the wood
into fibers using chemicals. Many plants also use mechanical pulping (giant grinders separating
out the fibers), while others use a combination of heat, other chemicals, and mechanical
methods. After the fibers are separated out, they may be bleached, and mixed with water to form
a slurry. After pulping, a residue remains which was historically dumped into rivers (hence
water pollution), but now must be treated. The process also takes a great deal of energy, so most
pulp mills have their own power plant, and therefore are significant sources of air pollution.
Pulping processes involve hazardous chemicals, raising issues of toxic releases.
The paper-making process is much less pollution intensive than pulping. Non-pulping
mills either buy pulp from other mills, or recycle wastepaper. During paper-making, the slurry
(more than 90% water at the start) is set on a rapidly-moving wire mesh which proceeds through
a series of dryers in order to extract the water, thereby producing a continuous sheet of paper.
Some energy is required, especially in the form of steam for the dryers, which can raise air
pollution concerns if the mill generates its own power. There is also some residual water
76
-------
pollution as the paper fibers are dried. Still, these pollution problems are much smaller than
those raised in the pulping process.
Over the past 30 years, pollution from the paper industry has been greatly reduced, with
the installation of secondary wastewater treatment, electrostatic precipitators, and scrubbers. In
addition to these end-of-pipe controls, some mills have changed their production process, more
closely tracking material flows to reduce emissions. In general, these changes have been much
easier to make at newer plants, which were designed at least in part with pollution controls in
mind (some old pulp mills were deliberately built on top of the river, so that any spills or leaks
could flow through holes in the floor for 'easy disposal'). These rigidities can be partially or
completely offset by the tendency for regulations to include grandfather clauses, exempting
existing plants from most stringent air pollution regulations.
3. Compliance and Enforcement Decisions
An individual paper mill faces costs and benefits from complying with environmental
regulation, which may depend on characteristics of the plant itself, the firm which owns the
plant, and the activity of environmental regulators. Given these constraints, the firm operating
the mill is presumed to maximize its profits, choosing to comply if the benefits (lower penalties,
better public image) outweigh the costs (investment in new pollution control equipment,
managerial attention). Regulators, in turn, allocate their activity to maximize some objective
function (political support, compliance levels, economic efficiency), taking into account the
reactions of firms to that activity.
The objective function for mill i owned by firm j at time t includes the usual revenues and
77
-------
costs of production, but these are extended to include the penalties associated with being found
in violation (Penalty), the probability of being found in violation (VProb), and the costs of
coming into compliance (CompCost):
(1) Profitijt(Comply) = Pijt*Qijt - Costyt - Penaltyijt*VProbijt(Comply) - CompCostijt(Comply)
Plants can vary their level of compliance (Comply) to maximize their profits (this assumes that
the underlying compliance decision is in fact continuous, although we only observe a 0-1
compliance status in our data. Assuming that the benefits and costs of compliance are captured
in the last two terms of equation (1), the plant will set its marginal cost of compliance equal to
the marginal benefit from compliance, measured here in terms of reductions in expected
penalties.
(2) d(-Penaltyijt*VProbijt)/dComply = d(CompCostijt)/dComply
This implicitly determines an optimal level of compliance, Comply*.
The benefits to the firm from increasing compliance come in terms of reducing the
probability of being found in violation of pollution regulations, thus reducing the expected
penalties for violations. These penalties are usually associated with regulators in terms of legal
sanctions and monetary fines, but could also be 'imposed' by customers boycotting the firm's
products in the future. In some circumstances customers might also be willing to pay more for
products that have been certified to have especially environmentally friendly production
processes, although this is currently more common in Europe than in the U.S. If we make the
usual assumption that the firm is risk-neutral, the expected benefits of compliance should be
linear in the probability of being in non-compliance, so the marginal benefit to the plant from
increasing its probability of compliance would be constant. Because of the difficulties
78
-------
associated with ensuring 100% compliance, we expect a rising marginal cost curve. Rising
marginal costs along with constant marginal benefits should lead to an interior Comply*
solution, equating the marginal costs and marginal benefits of compliance to the firm.
We focus on differences in compliance behavior across different mills, based on plant
and firm characteristics. As mentioned earlier, there are likely to be substantial differences in
pollution problems across different types of paper mills. We expect to see differences in
compliance behavior being related to the production technology at the plant (especially the use
of pulping) and related to the plant's age. There may also be economies of scale in complying
with regulations, so larger plants might find it easier to comply with a given level of stringency.
However, some of these plant characteristics on compliance could go either way: older plants
might find it harder to comply with a given standard, but they could be subject to less strict
standards due to grandfathering. Larger plants might enjoy economies of scale, but could also
have more places that something could go wrong, raising their probability of non-compliance.
Compliance behavior may also depend on characteristics of the firm which owns the mill
(e.g. the financial situation of the firm may matter). Pollution abatement can involve sizable
capital expenditures, which may be easier for profitable firms to fund - either through retained
earnings or through borrowing in capital markets. A firm in financial distress may not feel the
full threat of potential fines in an expected value sense, if they would just go bankrupt if they
happened to be caught. Firms with reputational investments in the product market may face an
additional incentive not to be caught violating environmental rules, if their customers would
react badly to the news.
Firms might also differ in the quality of the environmental support that they offer their
79
-------
plants. A large firm, or one specializing in the paper industry, is likely to have economies of
scale in learning about what regulations require, and may be in a better position to lobby
regulators on behalf of their plants. We cannot measure the strength of a company's
environmental program, but may observe a correlation in compliance behavior across plants
owned by the same firm. We may also see some effect of the firm size, either in absolute
magnitude or in terms of the number of mills they operate.
The regulatory activity faced by a plant is also expected to affect its compliance
behavior. A higher rate of inspections by regulators should increase VProb(Comply*) for any
given Comply* value, increasing the benefits from compliance. This inspection effect could be
described in terms of specific deterrence (plants who had been inspected in the past are more
careful) or general deterrence (plants with a high probability of being inspected are more
careful).1 Other enforcement actions might encourage compliance by raising the costs of being
found in violation (Penalty) without increasing the probability of being caught (VProb).
We test for differences across plants in their sensitivity to regulatory activity. Such
differences could arise for a variety of reasons. Plants owned by larger firms that sell on a
national market might be more concerned about bad publicity from environmental violations,
raising their Penalty, and hence their benefits from compliance.2 Larger plants may be used to
having regular inspections so that inspections have less of a 'shock effect' (specific deterrence)
than might be experienced by a smaller plant, reducing the benefits from compliance. Plants
may also differ in the cost of increasing their compliance, giving them different impacts from the
1 Scholz and Gray (1990) examine the impact of OSHA inspections on injury rates and find significant evidence for
both general and specific deterrence effects.
2
Conversations with people in the paper industry suggested that most large firms had strong policies encouraging
80
-------
same increase in regulatory activity.
Some of these different possibilities are shown in the three panels of Figure 1. These
panels all assume upward-sloping marginal costs and unchanging marginal benefits from
compliance. Each panel compares the impact on optimal compliance rates of an increase in the
benefits from compliance (such as might be induced by increased regulatory activity) on two
different plants. Figure la shows that even if the two plants differ in their initial level of
compliance, they could have the same change in compliance for a given increase in regulation, if
the slopes of their marginal cost curves are the same. Figure lb shows that differences in the
slopes of the marginal cost of compliance can result in very different impacts from the same
increase in regulation - here the plant with high and steep compliance costs has both lower
initial compliance and a smaller impact from the increased regulation. Finally, Figure lc shows
that plants with the same marginal cost of compliance can respond differently if the same
increase in regulation has different marginal benefits for them, as might happen if the larger firm
felt a greater desire to avoid adverse publicity (MB1').
In sum, a plant's compliance decision depends on its age and production technology, its
firm size and profitability, and the regulatory activity directed towards it, with the possibility of
some differences across plants in their sensitivity to that regulatory activity. We estimate a
model of compliance behavior as follows:
(3) Comply*ijt = f(REGS1Jt, X„ Xj, X1Jt*REGS1Jt, OComply1Jt, YEAR,).
COMPLY is the plant's observed compliance status with air pollution regulations. REGS is the
regulatory activity faced by the plant, which could be either inspections or other enforcement
100% compliance as much as possible, perhaps due to these concerns with adverse publicity.
81
-------
actions. This activity could affect either the probability of being caught in violation or the
negative consequences associated with being caught. The model includes characteristics of the
plant (X;) and firm (Xj), either of which could be interacted with enforcement activity to test for
differences in the responsiveness of plants and firms to enforcement. The plant's compliance
status with other regulatory areas is measured by OComply. Finally, year dummies (YEARt)
allow for changes in enforcement, or its definition, over time.
Now consider the regulator's decision about how to allocate its regulatory activity. If
enforcement were costless, regulators could use 'infinite' enforcement, catching all violators, in
which case setting a fine equal to the environmental damages from pollution would be optimal.
Becker (1968) notes that in a world with costly and uncertain enforcement, higher penalties
might be substituted for some of the enforcement effort, to raise the expected penalty for
violations. In fact, given limitations on the size of penalties under existing regulations, and the
high costs of controlling some pollutants, it seems puzzling why any firms would comply with
regulation. However, Harrington (1988) showed that a regulator could substantially raise the
effectiveness of enforcement, by making future enforcement conditional on past compliance. In
this model, non-compliance today not only raises expected penalties today, but the plant risks
being treated much more severely for years to come (or forever, depending on the regulator's
behavior).
If regulators are using the Harrington strategy, we would expect enforcement at a plant to
be greater in plants which violated the standards in the past. On the other hand, if most of the
differences in compliance behavior across plants are driven by fixed plant or firm characteristics,
those plants which are out of compliance may be more resistant to enforcement pressures,
82
-------
because they face higher costs of compliance. Therefore regulators might have to balance the
greater opportunity for compliance improvement against the greater enforcement effort needed to
achieve that improvement.
Regulators may also respond to differences in the potential environmental harm caused
by pollution, with plants in more rural areas facing less enforcement activity. In fact,
Shadbegian, et. al. (2000) find evidence that plants with greater benefits per unit of pollution
reduction wind up spending more on pollution abatement, suggesting that regulators are indeed
being tougher on those plants.
Observed differences in enforcement across plants and over time may also be strongly
influenced by the amount of resources allocated to regulatory enforcement in a particular state
and a particular year. During the 1980s the budgets of most regulatory agencies tended to
increase, so there were likely to be more inspections over time. There are also significant
differences in the political support for regulation across different states due to the severity of
pollution problems or to the political makeup of each state's population. On a more pragmatic
note, states may differ in the extent to which they enter all of their enforcement activity into the
regulatory databases we use.3
4. Data Description
Our research was carried out at the Census Bureau's Boston Research Data Center, using
confidential Census databases developed by the Census's Center for Economic Studies. The
primary Census data source is the Longitudinal Research Database (LRD), which contains
3 Of course the latter difference would cause problems for our estimation of the model, since seeing one 'observed'
83
-------
information on individual manufacturing plants from the Census of Manufactures and Annual
Survey of Manufacturers over time (for a more detailed description of the LRD data, see
McGuckin and Pascoe (1988)). From the LRD we extracted information for 116 pulp and paper
milla with continuous data over the 1979-1990 period. We capture differences in technology
across plants with a PULP dummy variable, indicating whether or not the plant incorporates a
pulping process. Our control for plant age, OLD, is a dummy variable, indicating whether the
plant was in operation before I9604. We control for the plant's efficiency using TFP, an index of
the total factor productivity level at the plant, which we calculated earlier when testing for the
impact of regulation on productivity in Gray and Shadbegian (1995,2003). Possible economies
of scale in compliance are captured by SIZE, the log of the plant's real value of shipments.
Finally, we include IRATE, the ratio of the plant's total new capital investment over the past
three years to its capital stock, to identify those plants with recent renovations.
In addition to these Census variables taken directly from the LRD, we use data from the
Census Bureau's annual Pollution Abatement Costs and Expenditures (PACE) survey. The
PACE survey provides us with the annual plant-level pollution abatement operating cost data
from 1979 to 1990. We divide this by a measure of the plant's size (the average of its largest two
years of real shipments over the period) to get a measure of the pollution abatement expenditure
intensity at the plant, PAOC.
To the Census data we linked firm-level information taken from the Compustat database.
enforcement action in a low-reporting state might mean the same thing as seeing several actions in a high-reporting
state.
4 We would like to thank John Haltiwanger for providing the plant age information. In our analysis we used a single
dummy to measure plant age (OLD = open before 1960) for two reasons: our sample includes some very old plants,
likely to heavily influence any linear (or non-linear) age specification, and concern with environmental issues was not
prominent before the 1960s.
84
-------
The ownership linkage was based on an annual industry directory (the Lockwood Directory),
capturing changes in plant ownership over time, which allowed us to calculate FIRMPLANT,
the log of the number of other paper mills owned by the firm. From the Compustat data we took
FIRMEMP, the log of firm employment, and FIRMPROF, the firm's profit rate (net income
divided by capital stock). We also include NONPAPER, a dummy variable indicating that the
firm's primary activity as identified by Compustat was outside SIC 26 (paper products). Since
some (not a large fraction) of our plants are privately owned and hence are excluded from
Compustat, we also include a dummy variable, MISSFIRM, to control for those observations
with missing Compustat data.
Our regulatory measures come from EPA's Compliance Data System (CDS). The CDS
provides annual measures of enforcement and compliance directed towards each plant. Our
compliance measure, COMPLY, is a dummy variable indicating whether the plant was in
compliance throughout the year (based on the CDS quarterly compliance status field - if a plant
was out of compliance in any quarter, COMPLY was zero). To measure air pollution
enforcement, we use ACTION, the log of the total number of actions directed towards the plant
during the year. We also split ACTION into INSPECT, the log of the total number of
'inspection-type' actions (e.g. inspections, emissions monitoring, stack tests), and OTHERACT,
the log of all non-inspection actions (e.g. notices of violation, penalties, phone calls). These
different types of actions may have different impacts on compliance, and may have different
degrees of endogeneity with compliance.
To supplement the air pollution data, we also use information from three other regulatory
data sets: the EPA's Permit Compliance System (PCS) and Toxic Release Inventory (TRI), and
85
-------
the Occupational Safety and Health Administration's (OSHA) Integrated Management
Information System (IMIS). The EPA's PCS provides information on water pollution regulation.
Unfortunately, this data set does not begin until the late 1980s, near the end of our period, so we
cannot include its variation over time in the model. Instead, we create WATER VIOL, the
fraction of years in which the plant had at least one reported water pollution emission that was in
violation of its permit. The EPA's TRI data set provides information on the disposal of toxic
substances from manufacturing plants. The TRI was first collected in 1987, so it also does not
provide useful time series variation for our model. Thus, we calculate the average discharge
intensity for the plant, TOXIC, as the annual pounds of environmental releases, averaged over
the 1987-1990 period, divided by the average real shipments of the plant in the same time period.
Finally, OSHA conducts inspections and imposes penalties to try to ensure safe working
conditions. We use data from OSHA's IMIS to measure the fraction of inspections during each
year that were in violation, OSHA VIOL, which is set to zero for those plants with no OSHA
inspections during the year. The OSHA data spans our entire period, so we can include the
annual values directly in our model.
5. Econometric Issues
Several econometric issues arise when we proceed to the estimation of equation (3). The
key econometric issue that any study of enforcement and compliance must face is the
endogeneity of enforcement: regulators are likely to direct more of their attention towards those
plants which they expect to find in violation. The explanation of this targeting behavior could be
as simple as a desire to avoid wasting limited regulatory resources by inspecting those plants
86
-------
which are almost certain to be in compliance (so probably no corrective action would result from
an inspection). A more complicated explanation comes from the work of Harrington (1988),
who showed that an optimal regulatory strategy could involve focusing long-run enforcement
activity on a few non-complying plants to punish them for not cooperating with regulation. In
any event, it is the case that past research has little trouble identifying a negative relationship
between enforcement activity and compliance behavior: non-complying plants get more
enforcement.
We tried two methods to overcome the endogeneity of enforcement: lagging the actual
enforcement faced by the firm and generating a predicted value of enforcement (which we also
lagged) to use in a second stage estimation (an instrumental variables method).5 The possible
problem with both of these methods is that some endogeneity may remain: for lagging, if there
is serial correlation in both the enforcement and compliance decisions, and for predicting, if the
explanatory variables used in the first stage are not completely exogenous. In addition, if the
lags are long enough or the first stage equation performs weakly enough there will be little
correlation between the instrument and the actual value of enforcement.
We use a relatively simple first-stage model to predict enforcement activity, focussing on
variables that are clearly exogenous with respect to the plant's compliance decision: year
dummies, state dummies, and VOTE. Year dummies account for changes in enforcement
activity over time, while state dummies allow for cross-state differences in enforcement activity
(or differences in reporting of that activity in the CDS). We also tested an alternative control for
state-year differences in enforcement: the overall air pollution enforcement activity rate (looking
5 Note that these two variables (lagged actual enforcement and predicted enforcement) could also be interpreted as
87
-------
at manufacturing industries, and dividing overall actions in the year by the number of plants in
the state's CDS database). The state enforcement rate was highly significant and had the
expected positive sign, but proved less powerful than the state dummies and is not used in the
final analyses shown here. Finally, we include a variable measuring the political support for
environmental regulation within the state, VOTE, which is the percent of votes in favor of
environmental legislation by the state's congressional delegation, as measured by the League of
Conservation Voters. The lagged predicted value from this first-stage model is then used in the
second-stage compliance models.
Another concern for the estimation of equation (1) is that the dependent variable in our
compliance equations (COMPLY) is discrete: a plant is either in compliance or not in
compliance. Thus we need to use an estimation method that is appropriate to a binary dependent
variable. In this case, we choose the logit model. We also estimate the model using a
(theoretically inappropriate) OLS regression model partly as a consistency check on the logit
results, but mostly so that we can easily include fixed effects into the analysis.6
A final concern for the analysis is the limited time-series variation available for key
variables. OLD and PULP never change in our data set, while other characteristics change only
slightly over time. Going to a fixed-effects model would completely eliminate OLD and PULP
and reduce the explanatory power of the other variables. If there is substantial measurement
error over time, using fixed-effects estimators could also result in a sizable bias in the estimated
corresponding to the specific and general deterrence effects mentioned earlier.
6 The fixed-effects version of the logit analysis would require estimating a conditional logit model, which in our
Census data set would probably raise disclosure concerns, making it unlikely that we could report the resulting
coefficients.
88
-------
coefficients (Griliches and Hausman (1986)). We briefly explore introducing fixed-effects into
an OLS model of compliance, but do not otherwise use fixed-effects models.
6. Results
Now we turn to the empirical analysis. Table 1 presents summary statistics and variable
definitions. Looking at the regulatory variables, compliance with air pollution regulations is
common, with about three-quarters of the observations in compliance. Enforcement activity is
also common, with plants averaging more than one enforcement action per year. Turning to
other regulatory programs, few plants show violations of either water pollution (16 percent) or
OSHA regulations (13 percent). Most of our plants (87 percent) were in operation in 1960 or
before, with slightly less than half (46 percent) including pulping facilities. The last two
columns (%CS and %TS) show the fraction of total variation in the variable accounted for by
plant and year dummies respectively. Nearly all of the variables in our data set are primarily
cross-sectional in nature, with only the productivity measure and firm profit rates showing
significant time-series variation. In any event, all of our models include year dummies, to
account for changes in overall compliance rates and definitions of compliance over the period.
In Table 2 we examine the correlations between key variables, using Spearman
correlation coefficients because they tend to be more robust to outliers. Examining plant
characteristics, we find that pulp mills are larger and spend more on pollution abatement, old
mills are less productive and are less likely to incorporate pulping, and large mills are more
productive and spend more on pollution abatement. Air pollution compliance is lower for plants
89
-------
that are large, old, incorporate pulping, and spend more on pollution abatement.7 Air pollution
enforcement activity is greater at plants which are large, incorporate pulping and spend more on
pollution abatement. Performance on other regulatory measures tends to be worse for large
plants, those incorporating pulping, and those that spend more on pollution abatement. Within
the set of regulatory measures, there is weak evidence for similar compliance behavior across
different regulatory programs: air compliance is negatively correlated with water pollution
violations, OSHA violations, and TRI discharges. Finally, air enforcement is negatively
correlated with compliance, evidence that the tendency to target enforcement towards non-
complying plants may make it difficult to observe empirically the ability of enforcement to
increase compliance.
Table 3 concentrates on the basic logit model of the compliance decision, based solely on
plant and firm characteristics. Most of the relationships are similar to those seen in the earlier
correlations. Compliance rates are significantly lower at old mills, pulp mills, and large mills,
however there is little evidence for any impact of firm characteristics on compliance. Switching
to an OLS model makes no noticeable difference in the results. However, a model incorporating
plant-specific fixed effects does give substantially different results - not surprisingly, since Table
1 showed us that most of the variables are primarily determined by cross-sectional differences,
and two of the key plant characteristics (pulping and old) are purely cross-sectional and therefore
drop out of the fixed effects model. Interpreting the magnitude of the Table 3 effects is easiest
from the OLS model (3D) — a pulp mill is 17% less likely to be in compliance, while doubling a
7 Some dummy variables in our data set (OLD, NONPAPER, and MISSFIRM) are not 'disclosable' in our analyses.
For these variables, we indicate the sign of the relationship, and double the sign (e.g. for results significant at the
10% level or better.
90
-------
plant's size reduces its compliance rate by 6% — but the transformed logit effects are nearly
identical.
Table 4 adds measures of the plant's performance on other regulatory measures. The
different regulatory measures are included separately, and then combined into a single model. In
all cases the results are similar: a plant's compliance behavior with regards to water pollution or
OSHA regulation is similar to its compliance for air pollution. The TRI results are much
weaker, and more sensitive to model specification. The weaker connection to TRI may be due to
the different regulatory structure: the TRI provides an information-driven incentive to reduce
discharges, while the other three regulatory programs follow the traditional command-and-
control model, and might therefore be more affected by a plant having a "culture of compliance"
for regulation in general. The magnitudes of the water and OSHA impacts could be substantial.
In model 4D, for example, a plant with 100% water compliance has an expected air compliance
rate 11 percentage points higher than one with 0% water compliance; a similar shift for OSHA
compliance is associated with a 14 percentage point higher expected air compliance rate.8
Table 5 provides a first look at the relationship between a plant's compliance with air
pollution regulations and a variety of measures of the enforcement effort it faces. We use both
actual enforcement and predicted enforcement measures, each lagged two years in an attempt to
reduce within-period endogeneity of enforcement.9 Based on the correlations seen in Table 2, it
is not surprising that we find evidence that plants which face greater enforcement activity, as
8 These calculations are based on the logit model's derivative of the probability of compliance with respect to the
explanatory variables equal to . 1824, evaluated at COMP's mean value of .76.
9 Predicted enforcement values come from a first stage tobit, explaining the log of each type of enforcement activity
using state and year dummies, as well as the VOTE variable. The pseudo-r-square of the tobits is .143, so we are only
explaining a relatively small part of the variation in enforcement.
91
-------
measured by lagged actual enforcement, tend to have a higher probability of being out of
compliance. We strongly believe that these results say more about the targeting of enforcement
towards violators, and do not indicate completely counterproductive enforcement. In an earlier
version of the paper, we examined the impact of enforcement on changes in compliance status.
These results indicated that enforcement activity was most effective in moving plants from
violation into compliance, rather than in preventing plants from falling out of compliance (results
available from the authors).
Once we account for the endogeneity of enforcement by using lagged predicted
enforcement we find the expected positive significant relationship between enforcement and
compliance. In particular, in model 5C, we find that increasing inspections by one raises the
probability of being in compliance by roughly 10%. However, once we include other actions
along with inspections (model 5E), the coefficient on inspections becomes a bit smaller and is no
longer significant, while the coefficient on other actions is positive and significant. The
magnitude of the two coefficients implies that increasing regulatory actions, either by one
inspection or one other action, leads to approximately a 10% increase in the probability of being
in compliance — although this increase is only statistically significant for other actions. This is
a large impact, given that only 24% of our observations are out of compliance.
In Tables 6 and 7 we consider differences in the impact of enforcement, based on plant
and firm characteristics. We focus our attention on those models which found the most positive
impacts of enforcement activity on compliance — models which use P(INSPECT)_2 and
P(OTHERACT)_2. These models include all of the plant and firm characteristics found in Table
3, which have similar signs and magnitudes to those found earlier. Table 6 considers possible
92
-------
interactive effects using the three plant characteristics that were significantly related to
compliance: plant age (OLD), plant size (SIZE), and having pulping operations (PULP). Recall
all three of these characteristics are associated with lower compliance rates. When we interact
these three variables with enforcement measures (separately), we see some differences in
response to enforcement activity by plant type: pulp mills are less sensitive to enforcement
activity. In particular, in model 6A, increasing inspections by one at a paper mill without
pulping facilities increases the likelihood of compliance by approximately 20%, whereas if the
paper mill does have a pulping facility the likelihood of compliance only rises by 5% — although
the interactive effect is not quite significant.
Table 7 presents similar results, using firm characteristics: profit rate, employment, and
number of plants (the latter two measured in log form). Although firm characteristics seemed
unrelated to compliance levels in Table 3, they appear to be strongly related to sensitivity to
enforcement, with opposite effects seen for sensitivity to inspections and to other enforcement
actions (such as notices of violation or enforcement orders). Plants owned by larger firms,
whether measured by firm employment or by the number of other paper mills owned by the firm,
are less sensitive to inspections, and more sensitive to other enforcement actions, than those
owned by smaller firms. For example, in model 7D, increasing the log of firm employment from
2.5 (its mean value) to 3.0 — only about 1/3 its standard deviation — completely eliminates any
positive effect that inspections have on the likelihood of compliance. In contrast, other actions
have a positive impact on the likelihood of being in compliance for any firm with a log of
employment greater than 1.5. Furthermore, for the same increase in log employment (2.5 to 3.0),
an additional other action raises the likelihood of being in compliance by roughly 5%. Perhaps
93
-------
larger firms have better-developed regulatory support programs and are less likely to be
'surprised' by routine inspections, but are at the same time more able to focus compliance
resources on plants with serious problems or plants in states with aggressive followup through
other enforcement actions, raising the costs of non-compliance. Smaller firms might be more
surprised by (and responsive to) routine inspections, but less able to put additional resources into
plants with serious problems and less bothered by bad publicity associated with other
enforcement actions.
7. Conclusions
We have examined plant-level data on enforcement and compliance with air pollution
regulation to: 1) test whether enforcement is effective in inducing plants to comply; 2) test
whether certain types of plants are more influenced by enforcement behavior; and 3) determine
what other firm and plant characteristics are associated with compliance. We find significant
effects of some plant characteristics on compliance: plants which include a pulping process,
plants which are older, and plants which are larger are all less likely to be in compliance. Unlike
Helland (1998), we find that firm-level characteristics are not significant determinants of
compliance at the plant level. On the other hand, plants with violations of other regulatory
requirements, either in water pollution or OSHA regulation, are significantly less likely to
comply with air pollution regulations. We do not see the same sort of effect for 'voluntary
compliance' as represented by TRI emissions. The magnitudes of the effects of plant-level
characteristics on compliance are non-trivial, at least for large changes in plant characteristics
and enforcement activity. In particular, doubling the size of a plant is associated with a 6%
94
-------
reduction in compliance; a plant with pulping has 17% lower compliance than one without
pulping; a plant in violation of water pollution regulations is 13% less likely to be in compliance
with air pollution regulations.
Measuring the impact of regulatory enforcement on compliance is complicated by the
targeting of enforcement towards plants that are out of compliance. This targeting effect
generally results in a negative relationship between enforcement and compliance. However,
when we account for the endogeneity of enforcement by using lagged predicted values of
enforcement, based on variables that are clearly exogenous to the plant's compliance decision,
we find the expected positive significant relationship between enforcement and compliance.
We also find some differences across plants in their responsiveness to enforcement, based
on plant characteristics. Pulp mills, which have difficulties in complying with regulations, are
also less likely to respond to regulatory enforcement (like Figure lb). For example, increasing
P(INSPECT)_2 by one inspection at a paper mill without pulping facilities increases the
likelihood of compliance by approximately 20%, whereas if the paper mill does have a pulping
facility the likelihood of compliance only rises by 5%. Finally, even though firm characteristics
are not found to be related to the level of compliance, we find them to be more strongly related
to a plant's sensitivity to enforcement (like Figure lc). Plants owned by larger firms, whether
measured in terms of their employment or by the number of other paper mills they own, are less
sensitive to inspections and more sensitive to other enforcement actions. For example,
increasing the log of firm employment from 2.5 (its mean value) to 3.0 completely eliminates
any positive effect P(INSPECT)_2 have on the likelihood of compliance. On the other hand, for
the same increase in log employment, one more P(OTHERACT).2 raises the likelihood of being
95
-------
in compliance by roughly 5%.
What lessons can be drawn by policy-makers from these results? First (and no surprise),
there are observable characteristics of plants which are strongly associated with their compliance
behavior. To the extent that regulators want to concentrate their enforcement activity on those
plants which are likely to be in violation, knowing which characteristics are important for a
particular industry could be useful. Second, firm characteristics seem much less important than
plant characteristics in determining a plant's compliance rate. Third, a plant's behavior in one
regulatory area appears to carry over into others, so that knowing a plant's compliance with
water pollution regulations (or even OSHA regulations) provides an indication of whether it is
likely to be in compliance with air pollution regulations. Fourth, enforcement is at least
somewhat effective in encouraging compliance.
Finally, there is evidence that plants differ in their responsiveness to enforcement
activity, and these differences are related to firm as well as to plant characteristics. In particular,
plants owned by larger firms are less responsive to inspections, and more responsive to other
enforcement actions (the effects of plant size are similar, though not statistically significant).
This is consistent with other research on regulatory impacts: Gunningham, et. al. (2003) find a
greater effect of EPA inspections for smaller firms, and Mendel off and Gray (2003) find a
greater impact of OSHA inspections on smaller workplaces.
We are planning to overcome some of the limitations of the current paper in future work.
Most importantly, we anticipate extending the data set into the 1990s. This will enable us to
include more years of data for other environmental regulatory measures, water compliance and
toxic discharges. The expanded data set will allow us to look more closely at the interactions
96
-------
between the compliance decision for one pollution medium and compliance on other media. We
also plan to expand our definition of compliance to allow us to distinguish among different
levels of compliance, ranging from paperwork violations to excess emissions, and to distinguish
between state-level enforcement activity and federal enforcement. Finally, we also plan to
examine the impact of regulation on compliance for plants in other industries including steel and
oil to see if regulatory effects differ across industries.
97
-------
REFERENCES
Bartel, Ann P. and Lacy Glenn Thomas, "Direct and Indirect Effects of Regulations: A New
Look at OSHA's Impact," Journal of Law and Economics. 28, 1-25 (1985).
Becker, Gary, "Crime and Punishment: An Economic Approach," Journal of Political Economy.
76, 169-217(1968).
Deily, Mary E. and Wayne B. Gray, "Enforcement of Pollution Regulations in a Declining
Industry," Journal of Environmental Economics and Management. 21, 260-274 (1991).
Gollop, Frank M. and Mark J. Roberts, "Environmental Regulations and Productivity Growth:
The case of Fossil-fueled Electric Power Generation," Journal of Political Economy. 91, 654-674
(1983).
Gray, Wayne B. "Environmental Compliance at Paper Mills: The Role of Regulatory
Enforcement and Corporate Restructuring," presented at AERE Winter Meetings, Boston,
January 2000.
and Mary E. Deily, "Compliance and Enforcement: Air Pollution Regulation in the U.S.
Steel Industry," Journal of Environmental Economics and Management. 31, 96-111 (1996).
and Carol A. Jones, "Longitudinal Patterns of Compliance with OSHA in the
Manufacturing Sector," Journal of Human Resources. 26 (4), 623-653 (1991).
and John T. Scholz, "Analyzing the Equity and Efficiency of OSHA Enforcement," Law
and Policy. 13, 185-214(1991).
and Ronald J. Shadbegian, "Plant Vintage, Technology, and Environmental Regulation,"
Journal of Environmental Economics and Management 46, 384-402 (2003).
and Ronald J. Shadbegian, "Pollution Abatement Costs, Regulation, and Plant-Level
Productivity," NBER Working Paper 4994 (1995).
Griliches, Zvi and Jerry A. Hausman, "Errors in Variables in Panel Data," Journal of
Econometrics. 31 (1), 93-118, 1986.
Gunningham, Neil, Dorothy Thornton, and Robert A. Kagan, "Motivating Management:
Corporate Compliance in Environmental Protection," presented at UC-Berkeley, June 2003.
Harrington, Winston, "Enforcement Leverage when Penalties are Restricted," Journal of Public
Economics. 37, 29-53 (1988).
98
-------
REFERENCES (cont.)
Helland, Eric, "The Enforcement of Pollution Control Laws: Inspections, Violations, and
Self-Reporting," Review of Economics and Statistics. 80 (1), 141-153 (1998).
Laplante, Benoit and Paul Rilstone, "Environmental Inspections and Emissions of the Pulp and
Paper Industry in Quebec," Journal of Environmental Economics and Management 31, 19-36
(1996).
Magat, Wesley A. and W. Kip Viscusi, "Effectiveness of the EPA's Regulatory Enforcement:
The Case of Industrial Effluent Standards," Journal of Law and Economics. 33, 331-360 (1990).
McGuckin, Robert H. and George A. Pascoe, "The Longitudinal Research Database: Status and
Research Possibilities," Survey of Current Business (1988).
Mendel off, John M. and Wayne B. Gray, "The Effects of Establishment and Inspection
Characteristics on the Impacts of OSHA Inspections in Manufacturing," presented at Law and
Society Association Meeting, June 6, 2003.
Nadeau, Louis W., "EPA Effectiveness at Reducing the Duration of Plant-Level
Noncompliance," Journal of Environmental Economics and Management. 34 (1), 54-78 (1997).
Scholz, John T., "Cooperation, Deterrence, and the Ecology of Regulatory Enforcement," Law &
Society Review. 18, 179-224 (1984).
and Wayne B. Gray, "OSHA Enforcement and Workplace Injuries: A Behavioral
Approach to Risk Assessment," Journal of Risk and Uncertainty. 3, 283-305 (1990).
Shadbegian, Ronald J., Wayne B. Gray, and Jonathan Levy, "Spatial Efficiency of Pollution
Abatement Expenditures," presented at NBER Environmental Economics session, April 13,
2000.
99
-------
Table 1
Summary Statistics
(N=1392)
Variable Mean Std Dev %CS %TS Description
Plant Characteristics
PULP
0.46
0.50
100
OLD
0.87
0.34
100
TFP
0.89
0.22
33
33
SIZE
10.30
0.81
93
<10
IRATE
0.13
0.17
20
<10
PAOC
0.004
0.005
77
<10
dummy, l=pulping operations
dummy, l=operating before 1960
total factor productivity (level)
real value of shipments (log)
real investment (last 3 years)/
real capital stock
pollution abatement operating
expenses / value of shipments
Firm Characteristics
FIRMEMP 2.49 1.43 70 <10 firm employment (log)
FIRMPROF 0.05 0.04 48 11 firm profit rate (net earnings/
capital stock
FIRMPLANT 2.29 0.85 80 <10 firm number of paper mills (log)
NONPAPER 0.20 0.40 . . firm's primary SIC not papermaking
MISSFIRM 0.19 0.39 . . plant not owned by Compustat firm
Air Pollution Regulation
COMPLY 0.76 0.43 31 <10 dummy, l=in compliance during year
ACTION 1.17 0.84 52 <10 total air enforcement actions (log)
(mean # actions = 3.79)
INSPECT 0.72 0.50 34 <10 air inspections (log)
(mean # inspections = 1.34)
OTHERACT 0.71 0.91 52 <10 other air enforcement actions (log)
(mean # other actions = 2.45)
Other Regulatory Measures
TOXIC
2.48
WATERVIOL 0.16
OSHAVIOL 0.13
2.86 100
29
32
100
<10
18
TRI air&water discharges/value of
shipments (1987-90 avg pounds/$000)
% water violations (1985-90 avg)
% OSHA inspections w/ penalty (79-90)
%CS = percent of variation explained by plant dummies
%TS = percent of variation explained by year dummies
100
-------
Table 2
Spearman Correlation Coefficients
(N=1392)
PULP
OLD
TFP
SIZE
IRATE
PAOC
PULP
1.000
< —)
0.036
0.538
-0.048
0.515
OLD
1.000
-0.130
-0.011
0.065
0.012
TFP
SIZE
IRATE
PAOC
1.000
0.235
0.015
0.006
1.000
0.042
0.396
1.000
-0.001
1.000
COMPLY
ACTION
TOXIC
WATERVIOL
OSHAVIOL
-0.230
0.300
0.310
-0.025
0.039
< —)
-0.071
-0.105
0.149
0.013
-0.006
0.050
0.046
-0.027
-0.090
-0.179
0.372
0.255
0.288
0.092
-0.062
0.006
0.045
0.010
0.046
-0.178
0.324
0.320
0.151
0.056
COMPLY
ACTION
TOXIC
WATERVIOL -0.075
OSHAVIOL -0.116
COMPLY
1.000
-0.295
-0.094
ACTION TOXIC WATERVIOL OSHAVIOL
1.000
0.210 1.000
0.093 0.115 1.000
0.099 0.034 0.143 1.000
Correlations exceeding about .08 are significant at the .05 level.
(--) indicates significant negative correlation.
101
-------
Table 3
model:
<3A)
Logit
Basic Compliance Models
(Dep Var = COMP; N=1160)
(3B)
Logit
(3C)
Logit
(3D)
OLS
Plant Characteristics
(3E)
F.E.
PAOC
1.064
(0.07)
0.427
(0.03)
0.072
(0.02)
0.879
(0.18)
PULP
-0.919
(-5.07)
-0.912
(-4.73)
-0.170
(-4.94)
OLD
TFP
(")
0.237
(0.59)
( —)
0.190
(0.46)
( —)
0.024
(0.35)
0.126
(1.11)
IRATE
-0.328
(-0.75)
-0.219
(-0.50)
-0.039
(-0.50)
0.019
(0.24)
SIZE
-0.303
(-2.61)
-0.365
(-2.81)
-0.055
(-2.57)
0.011
(0.12)
Firm Characteristics
FIRMEMP
-0.042
(-0.38)
0.120
(1.01)
0.018
(0.88)
-0.057
(-1.53)
FIRMPROF
2.970
(1.25)
2.468
(0.97)
0.451
(1.01)
-0.029
(-0.06)
FIRMPLANT
0.127
(1.09)
0.052
(0.42)
0.011
(0.51)
-0.073
(-2.09)
NONPAPER
(")
(")
(")
( + )
LOG-L
pseudo-R2
-609.72
0.064
-645.96
0.008
-605.97
0.070
0.075
0.341
Regressions also include a constant term and year dummies.
Firm variables include MISSFIRM.
(-) indicates negative coefficient; (--) indicates significant negative.
102
-------
Table 4
Compliance - Cross-Regulation Effects
Logit Models
(Dep Var = COMP; N=1160)
TOXIC
(4A) (4B) (4C) (4D) (4E) (4F)
Cross-Regulation Effects
-0.000 0.009 0.005 -0.031
(-0.02) (0.35) (0.17) (-1.33)
WATERVIOL
-0.713
(-2.73)
-0.618
(-2.32)
-0.670
(-2.54)
-0.601
(-2.58)
OSHAVIOL
-0.836
(-4.14)
-0.788
(-3.87)
-0.765
(-3.76)
-0.774
(-3.97)
PAOC
Plant characteristics
0.450 4.694 -1.793 1.429 2.184
(0.03) (0.30) (-0.12) (0.09) (0.14)
PULP
-0.911
(-4.68)
-1.070
(-5.30)
-0.941
(-4.82)
-1.086
(-5.26)
-1.092
(-5.62)
OLD
TFP
( —)
0.190
(0.46)
(-)
0.118
(0.28)
( —)
-0.002
(-0.01)
(-)
-0.054
(-0.13)
(-)
-0.011
(-0.03)
IRATE
-0.219
(-0.50)
-0.321
(-0.72)
-0.194
(-0.43)
-0.292
(-0.65)
-0.401
(-0.90)
SIZE
-0.366
(-2.81)
-0.245
(-1.78)
-0.324
(-2.45)
-0.220
(-1.58)
-0.154
(-1.23)
FIRMEMP
Firm Characteristics
0.120 0.099 0.108 0.095
(1.00) (0.82) (0.90) (0.78)
-0.071
(-0.63)
FIRMPROF
2.467
(0.97)
2.152
(0.83)
2.587
(1.00)
2.384
(0.90)
2.917
(1.19)
FIRMPLANT
0.052
(0.42)
0.060
(0.49)
0.073
(0.59)
0.077
(0.62)
0.103
(0.87)
NONPAPER
LOG-L
(")
(")
(")
(")
(")
-605.97 -602.26 -597.68 -594.99 -598.54 -632.17
pseudo-R2 0.070 0.075 0.082 0.086 0.081 0.029
Regressions also include year dummies, a constant term, and MISSFIRM.
(-) indicates negative coefficient; (--) indicates significant negative.
103
-------
Table 5
Compliance - Enforcement Measures
Logit Models
(Dep Var = COMP; N=1160)
(5A) (5B) (5C) (5D)
Enforcement Measures
<5E)
<5F)
P(ACTION).
-0.213
(-1.40)
ACTION.
-0.291
(-3.14)
P(INSPECT).
0.551
(1.85)
0.429
(1.40)
INSPECT.
-0.080
(-0.54)
0.045
(0.30)
P(OTHERACT).
0.483
(2.20)
OTHERACT.
-0.296
(-3.56)
LOG-L -605.01 -601.03 -604.18 -605.82 -601.75 -599.52
pseudo-R2 0.071 0.077 0.072 0.070 0.076 0.079
All models include the complete set of plant and firm characteristics from
earlier models, along with year dummies and a constant term.
104
-------
Table 6
Enforcement * Plant Characteristics
Logit Models
(Dep Var = COMP; N=1160)
(6A) (6B) (6C) (6D) (6E) (6F)
P(INSPECT).2 1.047 1.145
(2.24) (2.28)
P (OTHERACT) _2 0.123
(0.33)
-0.065 -0.033 3.827 7.051
(-0.14) (-0.07) (0.99) (1.51)
0.171 -1.314
(0.41) (-0.51)
PULP*P(INSPECT).2 -0.792 -1.124
(-1.46) (-1.89)
PULP*P(OTHERACT) .2 0.490
(1.26)
OLD*P(INSPECT).2 (++) ( + )
OLD*P (OTHERACT) _2 ( + )
SIZE*P(INSPECT) _2 -0.309 -0.628
(-0.85) (-1.42)
SIZE*P(OTHERACT) _2 0.175
(0.72)
LOG-L -603.08 -599.76 -602.89 -600.62 -603.82 -600.75
pseudo-R2 0.074 0.079 0.074 0.078 0.073 0.078
All models include the complete set of plant and firm characteristics from
earlier models, along with year dummies and a constant term.
105
-------
Table 7
Enforcement * Firm Characteristics
Logit Models
(Dep Var = COMP; N=1160)
<7A) (7B) (7C) (7D) (7E) (7F)
P (INSPECT).2
P (OTHERACT) _2
0.458 0.458
(1.18) (1.67)
0.402
(1.00)
0.685 1.311
(1.47) (2.55)
-0.713
(-1.84)
0.829 1.604
(1.32) (2.35)
-0.862
(-1.65)
PROF*P(INSPECT).2 2.464 0.529
(0.38) (0.07)
PROF*P(OTHERACT) .2 0.644
(0.14)
EMP*P(INSPECT). 2
EMP*P (OTHERACT) _2
-0.062 -0.445
(-0.37) (-2.29)
0.488
(3.89)
PLANTS*P (INSPECT) _2
PLANTS*P (OTHERACT) _2
-0.142 -0.643
(-0.50) (-2.00)
0.587
(2.94)
LOG-L -604.11 -601.73 -604.11 -593.39 -604.05 -596.80
pseudo-R2 0.072 0.076 0.072 0.089 0.072 0.084
All models include the complete set of plant and firm characteristics from
earlier models, along with year dummies and a constant term.
106
-------
Figure 1
Impact of Shift in Regulation on Optimal Compliance
MB=MB(Xp,Xf,REGS,X*REGS)
MC=MC(Xp,Xf)
$ $ $
MC1=MC2
MB1
MB 2
MB1=MB2
MC2
MB1
MB 2
MCI
MB1=MB2
MC2
MCI
MB1'=MB2
MB1=MB2
Compliance Compliance Compliance
Figure la Figure lb Figure lc
Same MB shift, Same MB shift, Different MB shifts
Different MC levels, Different MC levels, (MB 1 more
Same MC slope Different MC slopes sensitive),
Same MC
107
-------
Session I: Enforcement Issues
Discussant No. 1: Nick Franco, OECA
COMMENTS ON:
Factors Shaping Corporate Environmental Behavior and Performance:
Regulatory Pressure, Community Pressure, and Financial Status
Dietrich Earnhardt
University of Kansas
Observations
• The NPDES Permitting program has a unique regulatory structure, where regulated
facilities are required to regularly self-report their performance; and compliance status is
automatically determined in the PCS data system.
• Inspections are not the primary tool for identifying NPDES permit violations, but are a
check to ensure accurate self-reporting. Thus, there may not be a strong correlation
between inspections and specific and general deterrence. (For example, a compliance
order could be issued without an inspection.)
• Supplemental Environmental Projects (SEPs) are not an EPA penalty, but a voluntary
project undertaken by a regulated entity in conjunction with an enforcement action.
Though this represents a financial obligation, it is unclear what impact this will have with
respect to deterrence. SEPs are attractive to some regulated entities because they may be
considered as a mitigating factor when determining penalty size, they often contain a
component that may improve the regulated entities standing with the public, and so may
lessen the overall negative impact and deterrent effect of enforcement activity.
Policy Implications
• From a policy perspective it is hard to account for the mixed results, and the
inconsistency between the BOD and TSS results.
• It is doubtful that the results can be generalized to other media programs, given the
unique regulatory structure of the NPDES program.
• Where the number or average size of penalties is shown to worsen performance, this may
be accounted for the by the targeting effect, that is, penalties are given to those permitted
facilities that are out of compliance (i.e., regulatory activity is endogenous).
Other Considerations
Given the mixed results, some other areas of research that may help to clarify the impacts of
enforcement interventions, and provide clearer policy guidance are outlined below.
• What deterrent effect does the NPDES reporting system itself have on regulated entities,
does this account for the lack of consistent impact found for state and federal penalties?
How does this compare to other media programs where presence is established primarily
through inspections, investigations, and enforcement?
108
-------
• How can an analysis account for the enforcement policy of escalation? Escalation may
account for some of the difference seen between administrative and civil penalties.
Noncompliance with permitted effluent limits is normally first addressed with an
informal enforcement action (e.g., phone calls) at the state level. Then if necessary a
formal action (e.g., a compliance order, an administrative penalty order, or both). If these
administrative enforcement actions were not successful at compelling compliance then a
civil action would be initiated.
• Was the time period long enough to capture the results of injunctive relief? Capital-
intensive injunctive relief projects often take longer to implement and show results than
the time period of the study.
109
-------
Session I: Enforcement Issues
Discussant No. 1: Nick Franco, OECA
COMMENTS ON:
Deterrence and Corporate Environmental Behavior
Dorothy Thornton, University of California, Berkeley
Neil Gunningham, Australian National University
Robert Kagan, University of California, Berkeley
Observations
• General Deterrence relies upon the "threat signal" being received and understood.
• The paper is important because it shows there is a real general deterrence effect (65%
increased compliance activity base on enforcement activity against others); and the
response to the threat signal varies across groups (e.g., drive into compliance, re-enforce
compliance behavior, remind to pay attention to compliance requirements).
Policy Implications
• EPA has taken some steps to enhance and capitalize the general deterrence effect. Sought
to enhance it by issuing more press releases about concluded enforcement cases
(increased frequency of threat signal); and capitalize on it by using a signal case as an
opportunity to educate others in the industry and encourage them to take advantage of the
Audit Policy to self-disclose noncompliance by a certain time or face an inspection
(increasing personal risk perception). Though more could be done
Paper identifies a number of opportunities for EPA to Enhance General Deterrence
• Increasing Perceived Risk: given that many who responded to the survey overwhelmingly
underestimated penalty size suggests that EPA could enhance the general deterrence
effect by doing more to ensure regulated entities better understand penalty policies, and
penalties resulting from concluded cases.
• Likewise, EPA could take steps to ensure that regulated entities are better informed about
the number of penalties given out, and the broader applicability of specific enforcement
actions (e.g., helping to answer, does this apply to me?).
• More accurate knowledge about penalty size, frequency, and general applicability may
enhance the general deterrence affect.
• Reminder Function: the reminder function suggests that EPA should, at the very least,
use the occasion of a significant enforcement action to not only raise the perceived threat
in the eyes of those inclined not to comply, but to prompt those inclined to comply to
review their compliance status (e.g., an opportunity to provide self-assessment and
compliance assistance materials).
• Reassurance Function: the paper also lends support to EPA's motivation to conduct
inspections in order to maintain a presence in a particular sector in order: motivate
compliance by increasing perceived risk; and to help ensure a level playing field, which
110
-------
this paper suggests is a compliance motivator for some regulated entities (i.e., the
reassurance function).
Other Considerations
• Would the results of the study differ in an industry where there is not widespread
compliance? Would "explicit general deterrence" play more of a role?
• How can EPA foster "implicit general deterrence" (i.e., a culture of compliance) in an
industry?
• There was a lag, 1.5-2 years between the signal case and the survey. This leads one to
ask, does the general deterrence signal have a wasting impact; does it loose its affect over
time?
• How often does a general deterrence signal need to be received to have the maximum
impact in an industry? Is it more important that regulated entities remember the facts
surrounding a specific case, or that their cumulative perception of risk is maximized?
• Are companies taking into account things other than penalties when deciding to act (e.g.,
injunctive relief, impact on public image)?
• Are there industry types or structures where the threat signal is better communicated,
general deterrence has a greater impact (e.g., an industry with: a strong association,
dominated by a few large players, homogeneous operations)?
• Similarly, are there characteristics of a particular company that would make it more
inclined to head the general deterrence threat signal?
Ill
-------
Session I: Enforcement Issues
Discussant No. 1: Nick Franco, OECA
COMMENTS ON:
When and Why do Plants Comply? Paper Mills in the 1980s
Wayne Gray, Clark University
Ron Shadbegian, University of Massachusetts
Observations
• It is not generally true that firms face a rising marginal costs with regard to achieving
compliance. This seems to assume that compliance is a matter of capital outlays to
implement end-of-pipe pollution control. This does not take into account other
opportunities such process or input changes to reduce waste and pollution and
simultaneously achieve compliance and costs savings, or avoid regulation all together,
which may be available in other industries. This also seems to assume that companies
can effectively externalize pollution costs, which is likely not true.
• It is unclear what the basis is for the statement that even with limited inspection presence
and penalty size that"... most firms still seem to comply." EPA has calculated
statistically valid compliance rates for only a handful of sectors, and these have not
shown high levels of compliance.
• It appears that when populating dummy variable COMPLY, a plant was assumed to be in
compliance unless found to be out of compliance. Depending on the inspection presence
in the sector this assumption may skew the results.
Policy Implications
• The finding that where it is harder to comply, (i.e., cost of compliance is higher), plants
are less likely to be in compliance, and less likely to respond to regulatory enforcement,
seems intuitively obvious and suggests little in the way of policy prescriptions. The
agency already accounts for these factors when developing compliance assurance
strategies.
• The significance of plant-level characteristics on compliance, and the lack of significance
of firm-level characteristics, could be helpful in terms of targeting enforcement and
compliance resources. (Assuming the findings are generalizable). May be able use past
inspection data to identify common characteristics of non-compliant plants.
• The finding that firm characteristics are strongly related to a plant's sensitivity to
enforcement, but not to whether a plant is in compliance, raises a number of questions.
Does this mean that firms in this sector do not pay attention to plant level compliance
until a problem is identified? If this is the case it may suggest that compliance assistance
and general deterrence messages should be delivered at the firm as well as the plant level.
• The finding that non-compliance in one regulatory area is indicative of non-compliance
in other areas confirms the findings of other less formal analyses and anecdotal
understanding of plant level compliance. What would make this finding more useful is a
clear linkage between different types of non-compliance (e.g., if you are out of
compliance with regulation A you are likely out of compliance with regulation B);
112
-------
though this linkage is likely dependent on the industry and the mix of regulations that
they are subject to. In particular, this type of finding could be leveraged if some source
of readily available information could serve as an indicator of noncompliance that would
otherwise be difficult or costly to determine.
Other Considerations
• Additional research to determine whether the plant characteristics associated with a
higher likelihood of noncompliance in paper mills are generally applicable would
facilitate applying the finding more broadly.
General Comments
• EPA would be better served by future studies if they did not focus exclusively on
compliance, but also took into account other agency goals such as pollutant reductions.
The Office of Enforcement and Compliance Assurance has two long-term outcome goals
in the current Agency Strategic Plan, these are pounds of pollutants reduced, treated, or
eliminated, and the number of regulated entities making improvements to environmental
management practices.
• Research needs to view the suite of tools that EPA uses to ensure compliance (i.e.,
assistance, incentives, monitoring, and enforcement) in their proper context. All of these
tools are used in conjunction with one another to ensure compliance, not individually.
What would be a more fruitful line of research is looking at what combination of tools or
strategies work best to ensure compliance.
• Many economic models define deterrence as a function of the probability of being caught
in noncompliance and the cost of noncompliance. If this is the case, then inspections and
investigations speak to the probability of non-compliance being detected, and penalties
and other sanctions speaks to the cost of non-compliance. However, looking at just these
two components does not address timing and follow-through issues. With respect to
timing, is there an impact if there is a significant lag time between detection of
noncompliance and the leveling of a penalty? Likewise, what is the impact of detection
of noncompliance that results in no sanction? Is there a greater deterrent effect as the
percentage of inspections that detect noncompliance lead to a penalty increases?
113
-------
Session I: Enforcement Issues
Discussant No. 2: Randy Becker, U.S. Bureau of Census
COMMENTS ON:
Enforcement Issues:
EPA Conference on Corporate Environmental Behavior
and the Effectiveness of Government Interventions
Randy A. Becker*
Center for Economic Studies
U.S. Bureau of the Census
April 26, 2004
Introduction
The U.S. Environmental Protection Agency (EPA) has been expressing a strong interest
in understanding the factors that determine environmental performance at polluting facilities.
The three papers presented here today all examine whether regulatory actions (i.e., inspection,
penalties, etc.) result in better environmental performance at facilities.
These effects come in two basic forms: Specific deterrence measures the impact of past
regulatory actions taken directly against one's facility, while general deterrence measures the
impact of past regulatory actions taken against facilities like yours (e.g., other chemical plants in
your state).
Comments on the paper by Dietrich Earnhart
The first paper presented here examines the effects of regulatory pressure on the monthly
water pollution discharges of a panel of 508 "major" chemical (SIC 28) plants from 1995-2001.
The water pollutants he examines are (mainly) biological oxygen demand (BOD) and total
suspended solids (TSS).
* The opinions expressed here are my own and do not necessarily represent those of the U.S. Bureau of the Census.
114
-------
The focus here is mainly on the effectiveness of the various regulatory levers that the
government has at its disposal. Namely, this paper examines the effectiveness of inspections -
which come from two sources: EPA and state - as well as the effectiveness of penalties - which
also comes in two forms: EPA administrative penalties and federal civil penalties (administered
by the Department of Justice). Then there is the question of whether - within each of these 4
types of regulatory actions - general deterrence is as effective as specific deterrence.
Results suggest (to use the authors words) "a mixed degree of effectiveness." Indeed, it
seems that, currently, there are at least as many counter-intuitive effects here as there are
intuitive ones. But the paper makes clear that a lot of work is still pending, so these results
should be viewed as preliminary. Many of the 'next steps' outlined in the paper are exactly the
ones that I would suggest.
There is much to like about this paper, not the least of which is that it is a very carefully
explained study. I like the relative measure of compliance that is used here - i.e., the ratio of
absolute discharges to effluent limits - because (as the author also points out) it can capture not
only non-compliance, compliance, and over-compliance, but also the degree of non-compliance
or over-compliance. I like that the study is a joint examination of many factors: inspections vs.
penalties, EPA vs. state intervention, specific vs. general deterrence, the role of firm
characteristics and the interaction between firm characteristics and the different types of
interventions, as well as the impact of community characteristics. This is a nice broad view of
enforcement and compliance. Because there isn't much time, let me focus most of my comments
on some of the potential issues with this study.
First, I wonder whether we are studying the right facilities. (This may partly reveal my
ignorance on this subject.) I'd like to see some more context provided here — e.g., a table of
115
-------
industrial water usage (or discharge) by 2-digit SIC manufacturing industry. I wonder whether
there is a more "interesting" water-polluting industry to examine than chemicals, like pulp &
paper, or food processing, or some such. And are BOD and TSS these chemicals facilities' main
water concerns? Perhaps toxic releases, thermal pollution, etc. are as important, if not more so.
And what are the other water pollutants in the "limit exceedances" measure? I don't believe that
these are ever mentioned or discussed.
Also, it seems that air pollution is at least as problematic for chemical plants - if not more
so - than water pollution, but there is no discussion here of cross-media issues. Do the
inspection and penalty data used here also encompass air emission violations? If so, this might
explain the weak and puzzling relationships seen here (at least in part). If they do not, perhaps
there should be explanatory variables measuring how much regulatory pressure these facilities
face on other fronts, since it may affect their compliance in the water dimension. That is, if you
are constantly being inspected and fined for your air emissions, you may face some "spillover"
scrutiny of your water discharges.
I was truly struck by the magnitude of "over-compliance" here: On average, month after
month, these facilities are at 30% of their discharge limit. Since these plants are so far from
being non-compliant, I wonder whether they're even all that interesting to examine. These large
facilities have probably been regulated (and fined and inspected) for decades, which may be why
they are so compliant. Perhaps the interesting cases - the facilities closer to the margin - are the
small- and medium-sized plants that have only begun to experience more stringent regulation
more recently.
The second paper, by Thornton et al., suggests exactly this. In particular, they state that
many of the plants that they talked to were "beyond compliance" and that "hearing about
116
-------
punishments imposed on recalcitrants did not resonate with their own circumstances and
triggered little fear in them." (p. 14) Furthermore, the large chemicals companies in particular
suggested that "specific deterrence was not a salient driver of environmental actions" and that
"inspections held no fear for them." (p. 13) The authors go on to say that "in any event,
deterrence in any form was of far greater concern to [small- and medium-sized enterprises] than
it was to large ones." (p. 16)
There is also this notion that larger firms have more political clout and may be able to
negotiate more preferable emission limits. I am not sure there is strong evidence of that
necessarily, but it would be yet another reason to incorporate small- to mid-sized plants into the
analysis. I realize that this may not be possible however, because of a lack of data.
My overall concern here - which, again, may be born from my ignorance - is that we
may be focusing on facilities that may not be all that sensitive to the instruments being explored
— either because they are already super-compliant after decades of regulation and/or water
pollution is only a secondary issue for them. Focusing on a more sensitive population - e.g.,
small- or medium-sized plants in an industry that really has serious issues with BOD and TSS -
should really reveal the effectiveness of these regulatory instruments. Perhaps the author has a
sensitive population here, in which case further evidence should be presented to make that case.
Also, it seems to me that penalties are really a special case - quite a bit different from
inspections as a regulatory tool. In particular, unlike inspections, they are a tool that can only be
used in certain circumstances - namely, when there has been some sort of violation. And their
role is probably quite a bit different as well. It seems they would be used to induce compliance,
but they cannot be used to improve environmental performance in general (i.e., generate more
over-compliance) - which is really what we're talking about here in this sample of facilities!
117
-------
Penalties also seem a bit difficult to analyze empirically. In particular, penalties may
occur months after the actual violation, which would muddy any estimation of their "treatment"
effect. Furthermore, it seems to me that there cannot be a penalty without an inspection. If the
presence of a penalty always suggests the presence of an inspection, how does that affect the
interpretation of the penalty estimates (if at all)? Finally, environmental performance may
improve after an accidental discharge, with or without a penalty. Does this impact the
interpretation of results?
These issues aside, what should regulators take away from the findings of this study? It
seems to me that inspections are the only tool that they have at their disposal for any particular
plant - or at least it's the first stop. Therefore, I'm not sure that penalties - particularly in the
specific deterrence context - should receive equal and equivalent billing here in the analysis.
Finally, I would like to see more explanation of some of the counter-intuitive results.
Also, why might we expect "asymmetric" results between these two pollutants (BOD and TSS)?
And I think the author also needs to be a bit careful in interpreting his results: This study looks at
facilities that are the largest of the large and therefore the results may not necessarily generalize.
For example, at one point the claim is made that there are diseconomies of scale in water
pollution abatement. Since small- and medium-sized plants are largely absent here, the results
do not necessarily rule out a more U-shaped cost curve. Also, the compliance data here are self-
reported. Could it be that inspections and penalties induce better reporting, but no actual change
in behavior? That is, part of the inspection process may be the verification of emissions
calculations. Is there any evidence for that here? For some intuition on this subject, the author
may want to seek out the verification studies that have been done for the Toxics Release
Inventory (TRI).
118
-------
Comments on the paper by Dorothy Thornton, Neil Gunningham, and Robert Kagan
Like the first paper and the Gray & Shadbegian paper that follows, this paper explores
whether general deterrence (in particular) is important in shaping corporate environmental
behavior. The generally-held theory on general deterrence posits the following: First, firms
continually gather information on environmental inspections and penalties against others.
Second, evidence of a tough penalty against a firm reverberates throughout the community of
regulated businesses and raises their perceived risk of getting caught and facing sanctions.
Third, with this greater perceived risk, these businesses undertake measures to increase their
compliance (after some cost-benefit analyses).
Rather than infer such deterrence from volumes of data on inspections, penalties, and
plant-level pollution emissions, as do the other two studies, these authors simply ask firms
whether they are influenced by the penalization of others like them. Their survey and interviews
reveal little evidence of the sort of mechanism just outlined. I will now review some of the key
findings presented in this paper and offer some commentary along the way.
The authors begin with 112 EPA press releases on "penalty cases" from January 2000 to
June 2001 (i.e., recently but not too recent). From these, they sampled 40 cases. They then
searched many news databases to determine the extent of the media coverage received by each of
these cases. They find that most did not received "widespread" coverage. I think a bit too much
emphasis is placed here on the importance of media coverage. That is just one channel for
finding out such news. As important, if not more so, is the role of "informal" channels, such as
from workers, supplier, customers, and indeed from the regulators themselves. The authors may
have missed an important opportunity to ask firms: How do you typically hear about other
enforcement cases? Perhaps they have some anecdotal evidence that they can present, from their
119
-------
in-depth interviews with businesses.
In any event, they chose 8 of the 40 "signal" cases and drew a random sample of firms
operating in the same line-of-business and same state. Eighty percent (n=233) agreed to be
interviewed/surveyed, which is a truly exceptional response rate! Of these, 42% recalled the
signal case, which the authors think is rather low. I'm not so sure! (Is the glass 58% empty or
42%) full?) In any event, it seems like some adjustment to this statistic is warranted, based on the
"visibility" of the violating facility. That is, the responses should perhaps be "weighted"
somehow — e.g., by the (inverse of the) number of such plants in that industry-state, by the
size/prominence of the facility in question, and/or by the geographic proximity of the violator to
the surveyed business.* In an interesting result, the more "professionalized" the respondent was
(vis-a-vis the environment) the more likely the s/he recalled the signal case. I like this variable a
great deal and think that it could perhaps play a useful role in other environment-related surveys,
such as the Census Bureau's Pollution Abatement Costs and Expenditures (PACE) survey.
In what might be deemed "good news" for general deterrence, 89%> of respondents
recalled at least one recent penalty cases (if not the signal case). When told of the signal case,
however, respondents overwhelmingly under-estimated the actual penalties. The authors
conclude that, overall, the first component of the theory of general deterrence - i.e., that firms
actively seek out information on enforcement actions - is only weakly supported. Since these
particular firms appear to be super-compliant, this may not be particularly surprising.
They also find no particularly strong association between knowledge of other cases and
perceptions of the risk of detection or punishment (i.e., likelihood of being caught, of being
* Maybe we shouldn't be surprised if a chemical plant in Louisiana did not hear of the signal case because there are
in fact hundreds of chemical plants in Louisiana. Likewise, in a big state like California, it may not be surprising
that a case in Oakland or Fresno isn't known in San Diego. We should be surprised, however, if a steel mill in a
small state did not hear about the other steel mill down the road.
120
-------
fined, of being jailed, of plant being closed), implying a weak link between the first and second
components of the theory. And they find that those with a greater perception of detection and
punishment were not more likely to undertake compliance-related behavior, implying a weak
link between the second and third components of the theory.
They do find, however, that 65% of respondents report that they increased compliance-
related activity in response to hearing about another's fine or prison time, even if only meant
reviewing their existing compliance programs. This effect was a function of company size as
well as the number of other penalty cases the firm could describe.
Therefore, it seems as if general deterrence plays a role in most firms, even if it does not
follow the mechanism commonly believed. The authors argue that it serves a "reminder"
function (i.e., complying is a good thing) and a "reassurance" function (i.e., violators are
punished and there truly is a level playing field). These conclusions were supported by their in-
depth interviews.
I think it is quite right to suggest that we may be in a world that is "beyond general
deterrence" (my terminology). After decades of environmental regulation (and the EPA itself!)
there is now a "culture of compliance." Today, the very presence of regulations - rather than
who got caught - is what spurs compliance. The chemical plants in this study report that
regulations are just a "baseline" for them. Instead, protecting their reputations and avoiding
informal sanctions (by customers, investors, employees, local residents, etc.) are their much
bigger concern. The authors state that: "Overall, there was little support for models of business
firms as 'amoral calculators' who carefully weigh the certainty and severity of sanctions and who
can be manipulated through a judicious mix of specific and general deterrence." (p. 17) This is a
very optimistic conclusion - one I think we'd all like to believe.
121
-------
However, I have a few notes of caution before we dismiss general deterrence altogether.
First, there are "bad apples" out there (as evidenced by the signal cases). In particular, they may
be among the 20% who refused to respond to the survey. The authors should acknowledge that
there may be some selection bias in their statistics and (hence) their conclusions.
Second, the environmental personnel who responded to these surveys may not necessarily
be their firms' final word. I have no doubt that their hearts are in the right place — in many
ways, their career choice and livelihoods depend on regulation and environmental compliance.
But they ultimately do not decide how much resources are devoted to environmental concerns.
That decision is instead made at higher levels of the corporation and those decision-makers may
not be as pro-environment as these folks. This is, I believe, a very compelling reason to look at
actions (as in the other two papers), perhaps in addition to words.
Third, echoing my comments on the first paper, penalties are a rather special case. It is
not hard to imagine that firms do not see themselves in these particular signal cases —just as I
don't see myself in the millionaire who employs some bogus tax shelter and lands himself in a
white-collar prison. But the message that middle-class audits by the IRS are on the rise may
indeed resonate. What about inspections as general deterrence? The paper/survey is rather silent
on this possibility.
Finally, I'd like to underscore the paper's final sentence: "Our research has little to say
about the importance of explicit general deterrence messages at earlier states in regulatory
programs, when their value added may well be greater, or for firms (or industrial subsectors) that
are deliberate evaders or chronically at the edge of or out of compliance." (p. 18) I think that's
exactly right. It's very important to recognize the potential heterogeneity of firms - some will
comply no matter what, some may only respond to specific deterrence, some to general
122
-------
deterrence, some only to customers/stockholders/communities, and some to a combination of
these factors.
Comments on the paper by Wayne Gray & Ronald Shadbegian
Like the paper by Earnhart, this paper takes an empirical approach to examining specific
and general deterrence. Because of certain econometric difficulties however, these authors
(more or less) give up on estimating the former. They also focus on a different industry,
different pollution problem, and earlier time period than does Earnhart. In particular, this paper
examines the (annual) air pollution compliance of 116 pulp & paper mills from 1979-1990. This
is modeled as a function of inspections and other enforcement actions (such as notices of
violation, penalties, and phone calls), as well as plant and firm characteristics, and interactions
between these characteristics and the different types of regulatory actions. In typical Gray &
Shadbegian fashion, the paper offers a very nice discussion of the theoretical model, the previous
literature, the regulatory environment faced by these plants, their hypotheses, and so forth. (This
alone is worth the price of admission.) The paper's structure and exposition is tight.
The authors find that regulatory compliance was higher at facilities that had no pulping
activity, were younger, and/or were smaller. Firm-level characteristics - namely, size and
profitability - did not influence plant-level compliance however. "Cross-media" effects are
apparent, in that air pollution compliance was worse among facilities with violations in other
dimensions: water, toxic chemicals, and OSHA/safety.
On the key effect there is some mixed evidence. The authors find that 2-year lagged
enforcement activity (a measure of specific deterrence) actually reduced current-year
compliance, which is not what one would expect. (More on this in a bit.) On the other hand, 2-
123
-------
year lagged predicted enforcement activity (a measure of general deterrence) did in fact increase
current-year compliance, as might be expected. Here "other" enforcement actions (NOVs,
penalties, etc.) had an impact rather than inspections. And there is some evidence of differences
in sensitivities by plant- and firm-level characteristics. In particular, plants with pulping activity
are found to have been less responsive to inspections than those that didn't pulp, and larger firms
were less responsive to inspections but more responsive to other types of enforcement actions
(NOVs, penalties, etc.). The authors point out that the latter seems to suggest that smaller firms
might be more surprised by (and more responsive to) inspections and perhaps less bothered by
bad publicity associated with violations. This story seems entirely plausible, though I am not
sure that's the exact interpretation of this general deterrence measure.
In the limited time that I have, let me focus my comments on some of the potential issues
I see with this research (while making no claim to have fully thought through the various issues
I'm about to raise). First, and perhaps most importantly, I think the exclusion of plant-level
fixed effects raises the specter of omitted variable bias. I appreciate that many of the plant- and
firm-level variables included here are either time-invariant or change very little over time.
However, without such fixed effects, one will always wonder whether the variables are in fact
picking up the effects of other unobservable/unmeasured factors.
And I think there is evidence to be concerned about this: First, the Earnhart paper always
rejects OLS in favor of the fixed effect model. Second, in Table 3 of the current paper, we see
that the effect of plant size goes away with the introduction of (OLS) fixed effects. This suggests
that there is something correlated with being small - but not smallness itself - that improves
environmental performance. (At the very least, this possibility cannot be ruled out.) Third, the
perverse effect of 2-year lagged enforcement (i.e., specific deterrence) may be due to this
124
-------
variable picking up a "bad apple" effect that would otherwise be soaked up with plant fixed
effects.
The good news here is that the Chamberlain conditional logit, now available in
commonly-used statistical software, is specifically built to handle a binary outcome variable in
the context of panel data with fixed effects.* In this empirical specification, identification of
"treatment" effects comes from plants that change compliance status at least once over these 12
years. Indeed, plants that are always out-of-compliance or always in compliance fall out of this
analysis completely. Arguably, they are not the interesting population anyway (somewhat akin
to the super-compliant plants discussed above).
Another question/concern I have is with the role of pollution abatement operating costs
(PAOC) as an explanatory variable. This variable is not discussed much in the paper, perhaps
because its impact is statistically insignificant (which may say something about the quality of
these plants' PACE data). It occurred to me, however, that this variable could just as easily be
the dependent variable. That is, regulatory actions should spur PACE expenditures (abatement
activity) and then, in turn, compliance. What are we doing to our estimates by including this
variable and what happens if one were to take it out?
I also think that the authors need to be more careful when interpreting their coefficients
on the general deterrence measure. Their language suggests that they are talking about specific
deterrence when they are not (e.g., in the above example of small firms being surprised by
inspections and less bother by penalties). Finally, should the predicted probabilities used here
perhaps vary by firm characteristics?
Conclusion
* I don't fully understand the confidentiality concerns alluded to in footnote 6 (page 17).
125
-------
At this point, I think it is useful to briefly highlight a few important differences between
the Gray & Shadbegian study and the previous two papers. First, this paper explores a much
earlier time period than did the previous two. In light of the above discussion of the Thornton et
al. paper, this is exactly when one might expect to see more pronounced specific and general
deterrence effects - before compliance and over-compliance became quite commonplace (if
indeed they have). And in such a world, Gray & Shadbegian's theoretical model of (to use the
terminology of Thornton et al.) "amoral calculators" computing the optimal levels of
(non)compliance seems entirely appropriate. Finally, to the extent that the results from the Gray
& Shadbegian study differ from those of the other two, part of this may be due to the fact that
they employ HP A-reported compliance rather than .sf//-reported compliance. This again suggests
the necessity of looking at actions rather than (or in addition to) words.
126
-------
Summary of the Q&A Discussion Following Session I
Don Siege I (Rensselaer Polytechnic Institute)
Dr. Siegel directed his question to Dr. Wayne Gray "regarding the insignificance of the
firm characteristics in the model." He wondered whether it would be possible to
construct a "variable which would measure the percentage of the firm's revenue that's
represented in this industry—where you link the plant-level data to the firm-level data."
He said he thought of this because he believes there might be some diseconomies of
scope in monitoring the environmental performance of plants, and he suggested that some
sort of weighted least squares analysis might yield different results. Dr. Siegel closed by
saying he thinks "the theory predicts that some of those firm characteristics would be
important."
Dr. Wayne Gray (Clark University)
Dr. Gray responded, "Yes, we might well expect it to matter." He went on to state that in
the Compustat database they did have the SIC codes to tell them whether or not a firm's
industry affiliation was within "paper" (i.e, SIC 2600), but they "didn 7 find much of any
effect of that." Furthermore, he said, "With the census data, in principle, we could
identify all the establishments owned by that firm, but we'd only be able to do that very
well for the manufacturing part of the firm's activities—so, again, if the firm has a
substantial non-manufacturing component, I'm not sure we'd get so much out of it. It
would seem that that would be more valuable if we had found more of that sort of general
coding . . . ; it suggests that there may not be much there, but it is an interesting question
as to whether that industry focus makes you better at being in compliance or more
responsive."
Robert Kagan, (University of California at Berkeley)
Dr. Kagan commented that he was involved in a different study of the pulp and paper
industry in which he and his colleagues "looked at environmental performance at
particular facilities and at the firm level and corporate level—profitability and revenue—
the size factor." He said, "We found no correlation when we looked at cross sections;
however, we did find some relationship when we looked at corporate profitability at Time
1—say 7 or 8 years before the compliance/low-performance [problem] because the
capital expenditure at Time 1 seems more likely to have an impact at Time 2."
Dr. Gray, (directing a return comment to Dr. Kagan)
Dr. Gray stated, "You were asking people in their surveys what they predicted the penalty
would be for this sort of violation, but it seemed to me that you chose cases initially
because they were sort of big, . . . therefore, they're getting bigger penalties than the
average violation of that sort might be—you're sort of selecting on the size of the
penalty." . . . When you select them off of being really big up front, then you may be
picking ones that have an unusually high level of penalty, so maybe there are lower
estimates of what the right fine would be or how likely they were to get jail time for that,
127
-------
Dr. Kagen: So they may be right on what the average penalty was and underestimating
the serious violations—yes, I think you're right.
Pete Andrews, (UNC-Chapel Hill)
Dr. Andrews' first comment was directed to Wayne Gray. Dr. Andrews said he "was
really struck by [Dr. Gray's] comment that one of the significant variables was the
pulping facilities were just not responsive to inspections and so forth." He said he
wonders whether Dr. Gray has ever thought about "digging more deeply and whether that
was uniformly true across public facilities or whether even within that subcategory there
are better and worse performers and, if so, whether that has to do with technology
modernization and things like that. . ."
Directing his second question to Neil Gunningham, Dr. Andrews said, "You mentioned
this culture of compliance in which people have this belief that they either comply or they
get closed down, and I wonder if you've gone further to actually investigate whether that
is, in fact, objectively accurate or not." Saying that it could just be a widespread
assumption among small facilities, he noted that it shouldn't be hard to find out how
aggressive enforcement agencies are in terms of whether they actually ever close
anybody down or not.
Wayne Gray (responding to the first question)
Dr. Gray responded, "In terms of the details of the technology going on, we did do a
paper a while back that looked at different kinds of pulping . . . sulfite pulping may be
associated with more water pollution and some of the mechanical pulping might be
associated with more air pollution and such, and we did see some sense in which, in
terms of the location of these facilities, going to states where they had less stringent air or
water pollution regulations, but that may not be exactly what you're looking at. What
you were saying is that some plants may be more responsive within their [category]." He
said any time you run a regression you get the average coefficient of the group, and it
would require some sort of "observable characteristic" in terms of facilities' responses to
regulations to "differentiate the sheep from the goats," so to speak, and split them into
two groups. Dr. Gray continued, "Given that, I could then ask whether they seem to have
different coefficients and such. . . . The other problem is that with the census data we're
restricted to reporting the numerical coefficients based on the size of the number of plants
we have in each category, so I wasn't able to talk about the numbers on the old ones
because there weren't that many that were new." He closed by commenting that his
"reluctance to split things down into too small a groups" is also related to his not wanting
to reach the point where all that could be said is, "Yeah, they're different, but I can't tell
you what the numbers are."
Pete Andrews (in response)
"But it might be a useful outcome, though, in terms of targeting and figuring out what it
is in fact that's driving some businesses to do better than others, even in relatively similar
categories."
128
-------
Neil Cunningham (Australian National University), responding to Dr. Andrews' second
question
Dr. Gunningham confirmed that they did not actually check on the level of enforcement
activity and the number of closures following cited violations. He clarified that what was
"really striking" to him and his colleagues was that it was the "perceived level [of
enforcement activity that] had created this culture of ineligibility or compliance"—in
other words, "something that is perceived to be real is real in its consequences."
Jon Silberman (U.S. EPA)
Mr. Silberman offered a "couple of quick observations" regarding the term "over-
compliance," which many researchers were using in their discussions regarding Clean
Water Act permits. He clarified, "that's an economic term, not a regulatory term," and
he cited "engineering uncertainties and limitations, wet weather events and their
outcomes, and also—very important—the impact of where you are in your renewal
cycle" as factors that influence the compliance/over-compliance determination. To
clarify, he stated, "We're behind now in most of the EPA Regions and some of our
permits are being administratively continued, meaning that your permit numbers are not
ratcheted down to their new levels for up to 10 years, and a firm that is approaching the
end of that cycle is going to be desperately trying to predict where it's going to need to be
in the future relative to a firm that just had its permit renewed. So, I would just like to
suggest that when you combine that with the impact of the daily, weekly, and monthly
limits in the typical permit, what looks like over-compliance is actually the minimum the
firm really needs to do in order to avoid the types of spikes that will lead to non-
compliance on an irregular basis."
Neil Gunningham
Responding from the basis of a previous study of the pulp and paper industry that he was
involved with, Dr. Gunningham acknowledged that Mr. Silberman made a really
interesting point, but "in that study certainly some of the over-compliance we found, or
beyond-compliance activity, couldn't really be explained by this sort of permit cycle
factor." He cited the example of companies spending "millions of dollars—many
millions of dollars" to address the issue of smell, a local hot topic that's not regulated to
any great extent. He concluded that these companies' beyond compliance efforts were
obviously influenced by factors "other than just anticipating future permit laws."
Dietrich Earnhart (University of Kansas), adding to the discussion of beyond compliance
behavior
Dr. Earnhart stated, "In our particular study we've controlled for the volatility of the
discharges, which goes back to Mr. Silberman's point. It actually runs the opposite
direction of what you just proposed, if I understood it correctly—that is, if a firm's
discharges are more volatile, they should actually push their emissions or discharges
down to a lower level in order to avoid those spike." He said their study showed a "very
strong effect" of higher discharges associated with higher volatility. He added that they
had also "attempted to control for where they are in the permit cycle," not exactly the
way Mr. Silberman had captured that factor, but to the extent that they knew whether a
129
-------
facility was "working with an expired permit or not, . . . whether they're working with
final limits or interim or initial limits; we've controlled for the actual limit itself." Dr.
Earnhart closed by acknowledging that while there's surely more they can do, they've at
least attempted to capture a flavor of what he agrees are important dimensions.
Robert Kagan, asking "sort of a question back"
Dr. Kagan said that his comment "really relates to Wayne Gray's measure of compliance
as a binary variable (compliant/non-compliant)." He commented that "the notion that we
have violations that are spikes versus violations that are chronic" makes him wonder
whether enforcement people really think that a measure of compliance/non-compliance
tells them a lot. He explained, "It seems to me that it doesn't tell you much about the
seriousness of the chronic nature of compliance, given the wide variety of violations that
might be found at any moment, some of which are one time [events] and are easily
correctable—or do you think that it is a good measure because it tells you something
about. . . how much quality control a company is exercising [to achieve] compliance."
Nicholas Franco (U.S. EPA)
Mr. Franco responded, "Well certainly when we target we don't look at non-compliance
as kind of a binary thing—we pay more attention to chronic non-compliers . . . , so it's
the people that show chronic problems that indicate that it wasn't necessarily a one-time
event or spill or something like that, so it does get more attention. Maybe that's
something one could work into the analysis—the impacts of deterrence, specific or
general, on chronic non-compliers—because I would assume that for those who are in
chronic non-compliance it's going to require a much larger capital outlay for them to
come back into compliance. So, that maybe explains some of the facts that you can kind
of break those two groups out."
Magali Delmas, (UCSB)
Dr. Delmas brought a question related to the previous one about "How long does it take
for people at the plant level to actually take action?" She wondered whether, in the
efforts to determine the effectiveness of enforcement, anyone had explored "either in
your regression or during your interviews and survey . . . how long it takes for people to
take action, and does it change the result if you look at compliance 2-, 3-, 4 years after
the enforcement action or after the inspection. Also, does this time depend on the type of
enforcement?"
Wayne Gray
Dr. Gray clarified that they used a 2-year lag, and explained that the concern with trying
to do "the contemporaneous thing" was that you run into the problem of discovering that
a facility "had a really bad year—and had a lot of penalties." He explained, "You like to
have at least a little bit of a lag because of the sense that it takes a little while for things to
be corrected ... In a sense, what you want is a multi-dimensional picture of how they're
doing and the different dimension of: This problem happened because some piece broke
130
-------
and they fixed it the next week; and This problem happened because they were just
running the plant too hard and the treatment couldn't keep up with it; or something like
that. I think we don't get that clear a picture from the sort of quantitative data as you
might like to in terms of exactly what's going on."
Dietrich Earnhart, ("following up on Wayne'spoint")
Dr. Earnhart stated that working with monthly data is much easier because it helps you
avoid the "contemporaneous quagmire." He added that, "It could be possible that with a
minor amount of effort a facility could actually improve their performance even when
given a month or two, . . . and it may not be some large capital outlay—it could just be a
matter of a better way of tracking their waste stream." He stated that it could also take 2-
3 years for a company to build up the necessary capital, financial or physical, to correct a
problem and improve performance, and that's the issue—it varies on a case-by-case
basis. Consequently, Dr. Earnhart said he would be "very reluctant to say that there's one
particular lag period that would fit for all facilities." He cited the efforts of previous
researchers to assign an effect factor to each preceding month. He also cited efforts, such
as Wayne Gray's, to "slice the data various ways—1-year lag, 2-year lag, 3-year lag, and
then hope that it will be discernible across regression analysis." Dr. Earnhart added that
in his studies he has taken "3-month lags, 6-month lags, 12-month lags, 24-month lags,
and frequently it's robust across all the timeframes." In conclusion, he stated, "So it's
more to say: Something's happened reasonably recently—did that have an effect?" and
he cautioned that this only would apply to specific deterrents; general deterrents present
their own problems.
131
-------
Corporate Environmental Behavior and the
Effectiveness of Government Interventions
PROCEEDINGS OF
SESSION II: COMPLIANCE ISSUES
A WORKSHOP SPONSORED BY THE U.S. ENVIRONMENTAL PROTECTION
AGENCY'S NATIONAL CENTER FOR ENVIRONMENTAL ECONOMICS (NCEE),
NATIONAL CENTER FOR ENVIRONMENTAL RESEARCH (NCER)
April 26-27, 2004
Wyndham Washington Hotel
Washington, DC
Prepared by Alpha-Gamma Technologies, Inc.
4700 Falls of Neuse Road, Suite 350, Raleigh, NC 27609
-------
ACKNOWLEDGEMENTS
This report has been prepared by Alpha-Gamma Technologies, Inc. with funding from
the National Center for Environmental Economics (NCEE). Alpha-Gamma wishes to
thank NCEE's Cynthia Morgan and Ann Wolvertan and Project Officer Ronald Wiley for
their guidance and assistance throughout this project.
DISCLAIMER
These proceedings are being distributed in the interest of increasing public understanding
and knowledge of the issues discussed at the workshop and have been prepared
independently of the workshop. Although the proceedings have been funded in part by
the United States Environmental Protection Agency under Contract No. 68-W-01-055 to
Alpha-Gamma Technologies, Inc., the contents of this document may not necessarily
reflect the views of the Agency and no official endorsement should be inferred.
11
-------
TABLE OF CONTENTS
Session II: Compliance Issues
Corporate Self-Policing and the Environment: Factors Predicting Self-
Disclosure of Clean Air Act Violations Under the EPA's Audit Policy
Paul Stretesky, Colorado State University 1
Regulation and Compliance Motivations: Marine Facilities and Water
Quality
Peter May, University of Washington 25
Discussant
Jon Silberman, U.S. EPA, OECA/National Center for Environmental
Innovation (NCEI) 50
Discussant
John Horowitz, University of Maryland 66
Summary of Q&A Discussion Following Session II 68
in
-------
Corporate Self-Policing and the Environment: Factors Predicting
Self-Disclosure of Clean Air Act Violations under the
Environmental Protection Agency's Audit Policy*
Paul B. Stretesky
Department of Sociology
Colorado State University
Fort Collins, Colorado 80523
Telephone: 970.491.6825
Fax: 970.491.6825
E-mail: pstretes@lamar.colostate.edu.
*This research has been partially supported by a grant from the U.S. Environmental Protection
Agency's Science to Achieve Results (STAR) program. Although the research described in the
article has been funded in part by the U.S. Environmental Protection Agency's STAR program
through grant R829686, it has not been subjected to any EPA review and therefore does not
necessarily reflect the views of the Agency, and no official endorsement should be inferred.
-------
Abstract
The Environmental Protection Agency's self-policing or audit policy waives or reduces penalties
when regulated entities voluntarily discover, disclose, and correct environmental violations. This
study uses a case-control design to determine if specific deterrence, general deterrence and
compliance assistance are associated with the odds of disclosing a Clean Air Act (CAA) violation
under that policy. The event group consists of all 59 companies that disclosed a CAA violation
under the Audit Policy between October 1, 1998 and September 30, 2000. The control group
consists of a simple random sample of 59 companies that did not use the Audit Policy but were
discovered to have violated the CAA during the same time period. The results of this research
suggest that specific deterrence and compliance incentives are not related to Audit Policy use.
However, companies with facilities operating in an industry where regional inspection rates are
high are more likely to use the Audit Policy than companies with facilities operating in an industry
where regional inspection rates are low. Also important is the finding that company size is a strong
positive predictor of Audit Policy use.
2
-------
Introduction
The Environmental Protection Agency (EPA) issued a policy entitled "Incentives for Self-
Policing: Discovery, Disclosure, Correction, and Prevention of Violations" [Audit Policy] in 1995.
The Audit Policy is a compliance incentive that waives or reduces penalties for regulated entities
that voluntarily discover, promptly disclose, and correct violations of federal environmental
requirements. There are several requirements for implementing the Audit Policy. First, a company
must disclose the violation to the EPA within 21 days of the discovery of that violation. Second,
the discovery cannot be the result of a legally required monitoring, sampling or auditing procedure.
Third, the company must correct the discovered violation within 60 days of its discovery. Repeat
violations (e.g., similar violations that have occurred at the same facility within the past 3 years or
similar violations that have occurred as part of a pattern of violations at other facilities over the past
5 years) and violations that result in serious actual harm are not eligible.
Although the Audit Policy has been in existence for nearly nine years, researchers know
little about what factors may actually predict whether a company will disclose a violation under the
policy. The purpose of this study is to determine whether deterrence and compliance incentives by
the EPA encourage companies to disclose violations of the Clean Air Act (CAA) under the Audit
Policy.
Prior Research
Many regulatory personnel tend to view corporate offenders as rational actors who break
the law to maximize profit (see Kagan and Scholz 1983). This view of corporate actors can be
traced back to the materialist approaches of corporate behavior (see Edelman and Suchman 1997).
Such a view has led to the popularity of corporate deterrence research in the social science
literature (Braithwaite and Makkai 1991; Makkai and Braithwaite 1994; Russell and Gilbert 1999;
3
-------
Stafford 2003). The basic idea behind deterrence is that behavior is influenced by negative
consequences. Specific deterrence focuses on the future behavior of those actors that have already
experienced negative consequences. General deterrence focuses on the future behavior of those
actors that may or may not have experienced negative consequences, but believe that those
consequences exist because they have seen others experience them (Gibbs 1974; Reiss 1984; Tittle
and Paternoster 2000; Warr and Stafford 1993; Zimring and Hawkins 1973).
Research on specific deterrence is inconsistent (Doob and Webster 2003) and research on
corporate deterrence is no exception. For instance, some corporate specific deterrence research
shows strong positive effect while other research shows no (or even a negative) effect (May and
Winter 1999). Simpson and Koper (1992:367) found that "among a group of prior offenders,
formal-sanction severity is a strong inhibitor of re-offending." However, Block, Nold, and Sidak
(1981) found that civil penalties were more likely to have a deterrent impact than more severe
criminal penalties. Braithwaite and Makkai (1991) found evidence that sanction severity does not
appear to have a deterrent effect in the regulation of nursing homes in Australia.
Sanctions against corporations may also have a general deterrent effect when they are
administered to "make an example out of the violator and send the clear message to others who
might violate laws and regulations" (see Cohen 1998). Research on corporate general deterrence
has received much less attention in the literature than research on specific deterrence. However,
unlike specific deterrence studies, some evidence suggests that general deterrence may have an
effect on corporate behavior. For instance, Block, Nold, & Sidak (1981) studied the impact of
antitrust enforcement on price markups in the white bread industry and found that bakers in
neighboring areas reduced their prices after a price fixing case is brought in a nearby city.
4
-------
Certainty of detection has also been shown to be a strong predictor of corporate behavior. In
a study of the effectiveness of enforcement and compliance on agro-environmental policy in
Denmark, May and Winter (1999) found that the frequency of inspections were viewed as highly
effective at gaining corporate compliance of environmental regulations. May and Winter
(1999:645) suggest that "smart enforcement consists of frequent inspections targeted toward
important items." Braithwaite and Makkai (1991) also found that certainty of detection was an
important predictor of compliance in the nursing home industry. Other researchers studying
detection certainty assert that inspections are likely to be related to corporate compliance (Sholz
1991; Scholz and Gray 1997).
The EPA's Audit Policy is a compliance-oriented policy. Compliance-oriented policies are
increasing in significance and focus on education, persuasion, and cooperation rather than detection
and deterrence to improve corporate behavior (Parker 2000; Reiss 1984). According to Parker
(2000:530) the shift toward compliance-oriented policy represents the "new regulatory state, which
uses enforced self-regulation and incentives for voluntary compliance to steer corporate conduct
toward public goals without interfering too greatly with corporate autonomy and profit" (see also
Braithwaite 2000). Clarke (1987) suggests that compliance assistance efforts are more effective
than deterrence efforts (see also Stone 1975). The EPA has initiated several special programs that
provide compliance assistance to companies in order to encourage and facilitate disclosure of
environmental violations (Office of Regulatory Enforcement, EPA 1999). Moreover, regulated
entities that have received relief under the EPA's Audit Policy suggest that compliance assistance,
rather than deterrence itself, has increased their reporting of environmental violations (EPA 1999).
Friedrichs (1996) argues that regulatory agencies often confront the "choice between
emphasizing compliance or deterrence." However, it is also suggested that deterrent strategies
5
-------
combined with compliance strategies are likely to be more effective than deterrence or compliance
alone (see May and Winter 1999; Simpson and Koper 1992). Thus, it may be argued that it is
rational for regulated entities to take advantage of incentives provided in the compliance-oriented
Audit Policy if there is a perceived or real likelihood that a regulatory agency would soon discover
the violation and impose a sanction. While research on corporate compliance is pervasive, there is
still a relative lack of information on corporate self-reporting. The purpose of this research, then, is
to determine if specific deterrence, general deterrence, detection certainty, and compliance
assistance efforts are associated with the odds of disclosing a violation under the EPA's Audit
Policy. This is an important question since the Audit Policy is relatively unstudied policy and
because it falls under the type of regulatory effort that Parker (2000) describes as part of the "new
regulatory state."
Data and Methods
The unit of analysis in this study is company. The decision to study companies rather than
facilities or geographic areas such as states or counties is largely based on theoretical and practical
concerns. First, from a policy standpoint, the study of companies is desirable because they are the
entities that make the decision to disclose environmental violations under the Audit Policy
(Rebovich 1998). The Audit Policy implies a rationality that is typically associated with company-
level decisions. As Simpson (1986:860) observed, companies are "economic entities that are
rationally constructed, chartered, and owned by stockholders. Their overriding goals are economic,
i.e., profitability and market share expansion." From this point of view, the self-disclosure of
environmental violations may best be explained in terms of deterrence, incentives and factors that
are both internal and external to the company (see Kramer 1982; Shover and Bryant 1993).
6
-------
Second, many studies of organizational compliance are focused on the company. Several
interesting covariates to be included in the proposed study are based upon the findings of previous
company-level studies conducted in various industries (e.g., the nursing home industry, agricultural
industry, chemical industry, liquor industry, auto industry, pharmaceutical industry, and securities
industry). Third, many of the variables that are thought to be important in the study of corporate
behavior are readily available for companies, but not available for geographic areas or facilities
(e.g., corporate financial performance).
Sample
In order to assess whether sanction certainty, sanction severity and compliance incentives
appear to influence the decision to self-disclose a violation of the Clean Air Act (CAA) under the
Audit Policy it is necessary to compare companies that report CAA violations to companies that
violate the CAA but do not report those violations. Unfortunately, it is not possible to identify
every company that violates the CAA because some are never detected. However, investigations
uncover over one-thousand violations of the CAA each year (Scalia 1999). Companies known by
the Environmental Protection Agency to violate the CAA and not report that violation under the
Policy can be compared to companies that violate the CAA and report it under the Audit Policy. In
the proposed study, a case control design is used to make that comparison.
In general, case-control designs are used to identify factors that help differentiate cases that
experience an event from cases that do not experience an event. In the proposed case-control study,
the event group is composed of all companies that reported a CAA violation under the Audit Policy
during the 1999 and 2000 Fiscal Years (October 1, 1998 - September 30, 2000). The event group
in the proposed study is a complete enumeration and will consist of all 59 companies that were
granted relief for violating the CAA under the Audit Policy. A list of those companies was made
7
-------
available to the researcher by the EPA's Office of Regulatory Enforcement. Companies that
violated Section 211 of the CAA (mis-fueling violations) are excluded because those cases were
handled by EPA Headquarters and because the nature of the cases was so different from most EPA
regional CAA cases. The control group is the same size as the event group and also consists of 59
companies obtained by means of a simple random sample of all companies that were discovered by
the EPA to have violated the CAA during the same time period as the event group. This study is
mainly interested in making comparisons between companies that did use the Audit Policy to
companies that could have used the Audit Policy had their violation not been discovered. Thus,
companies charged with a criminal violation of the CAA are excluded from the control group
sampling frame for the simple reason that these violations are ones that are likely to be judged by
the EPA as resulting in "real harm" and would therefore are likely not be Audit Policy eligible (and
not be reported under the Audit Policy even if discovered by the company).
The sampling frame needed to select the controls was easily constructed from the EPA's
Integrated Data for Enforcement Analysis (IDEA) system. The IDEA system was developed by the
EPA in response to a need for integrated data on facilities that were potentially involved in
enforcement or compliance actions. The IDEA database has information on nearly all regulated
facilities for the purposes of facilitating enforcement case screening, enforcement/inspection
targeting, and management decision making. Because the IDEA system collects information at the
facility level, and companies are the unit of study in the proposed investigation, it is important to
make sure that facilities are linked to companies to ensure that companies with multiple facilities
are not over-represented in the control group. This was accomplished by first linking facilities to
companies and then taking the sample of companies.
Dependent Variable
8
-------
The dependent variable is dummy coded and indicates whether a company self-disclosed or
was discovered to have violated the CAA. If a company disclosed an environmental violation to the
EPA under the Audit Policy at any facility between January 1999 and December 2000 it was given
a score of "1." If a company was investigated and found to have violated an environmental law
between January 1999 and December 2000 it is assigned the score of "0."
Variables measuring specific deterrence
Prior to discussing the creation of variables that measure specific deterrence it is important
to point out a problem inherent in studying enforcement and inspection rates. It was not a problem
when a company only had one facility as enforcement data could easily be collected for that
company's facility. However, several of the companies in this dataset were linked to more than one
facility. For those companies that operated more than ten facilities it was not possible—given
limited resources—to collect enforcement data on all company facilities. Therefore, enforcement
and inspection rates had to be estimated from a simple random sample of 10 facilities for 13
companies in this sample.
In order to measure specific deterrence related to past enforcement at the facility, IDEA is
used to construct a variable that measures past enforcement as the average number of enforcement
actions by the EPA across company facilities two years prior to the discovery or disclosure of a
CAA violation. Two types of enforcement actions were counted: civil judicial and administrative
(both formal and informal). There were no criminal enforcement actions initiated or settled against
the companies in this sample during the time period under investigation. Since state regulatory
agencies often take the lead in environmental enforcement matters, those enforcement actions were
also counted as part of the total number of enforcement actions. One potential problem with
calculating enforcement actions using IDEA is that some types of enforcement actions are linked to
9
-------
more than one company facility. For example, in one case a civil judicial enforcement action
covered violations at ten facilities. In order to avoid counting one enforcement action several times
for one company, an action that appears at more than one facility is only counted once.
As noted in the literature review, previous research on the sanction severity is mixed. For
instance, Block, Nold, and Sidak (1981) found that civil penalties were more likely to have a
deterrent impact than criminal penalties. Simpson and Koper (1992:367), however, found that
"among a group of prior offenders, formal-sanction severity is a stronger inhibitor of future
compliance than are measures of sanction certainty." Sanction severity is measured by variations
in case outcomes on Audit Policy disclosure. Following Simpson and Koper's (1992:374) well
known work on antitrust violations, an effort will be made to determine whether more serious
cases are more likely to encourage self-disclosure. It should be noted, "what may be costly to one
company may be relatively insignificant to another" (Simpson and Koper 1992:355). Nevertheless,
it is reasonable to rank in order "how much of a problem particular outcomes appear to be"
(Simpson and Koper 1992:355 see also Clinard 1983; Cullen and Dubeck 1985; Frank and
Lombness 1988). IDEA is used to create two variables that indicate whether or not a civil or
administrative penalty was imposed against a company two years prior to disclosing or being
caught for an environmental violation. These variables are dummy coded (yes=l and no=0) to
indicate whether an environmental violation occurred and resulted in: (1) a civil penalty or (2) an
administrative penalty. If sanction severity is related to the disclosure of environmental violations,
companies that have experienced the most serious case initiations (e.g., civil penalties) should
have greater odds of disclosing an environmental violation under the Audit Policy than companies
with less serious outcomes (e.g., administrative penalties).
10
-------
In the proposed study it is hypothesized that the effectiveness of the Audit Policy can be
enhanced through targeted inspections. That is, the more certain it is that a violation will be
discovered by a regulatory agency the more likely it is that a company will conduct an
environmental audit and then disclose any environmental violations that are discovered as a result
(see also Scholz 1991; Scholz and Gray 1997). One measure of detection certainty is the rate of
CAA inspections at company facilities two years prior to the disclosure or discovery of a CAA
violation. The rate of company inspections by state or federal regulatory agencies is derived from
IDEA and calculated by dividing the number of inspections at company facilities by the number of
company facilities that are eligible for inspection. This variable, then, represents the average
number of inspection across company facilities 2 years prior to the discovery or disclosure of a
CAA violation. It is hypothesized that a company with a high regulatory inspection rate should
have greater odds of disclosure under the Audit Policy than a company with a low regulatory
inspection rate.
Variables measuring general deterrence
In order to determine if general deterrence is related to reporting under the Audit Policy
three different variables are created. To examine the hypothesis that the odds of disclosure under
the Audit Policy are greater for companies with facilities situated in geographic areas where
enforcement actions are more common than for companies situated in geographic areas where
enforcement actions are less common, the rate (per 1000 facilities) of criminal cases, civil cases,
and administrative cases that are initiated are computed (using IDEA) for the state or states in
which the company's facilities operate. These rates of enforcement are calculated for two years
prior to the disclosure or discovery of CAA violation. The rate of enforcement actions is calculated
by dividing the number of environmental enforcement actions in that state by the number of
11
-------
facilities in that state. In cases where company facilities are located in multiple states, those rates
of enforcement are averaged together to come up with an overall enforcement for all company
facilities.
Certainty of detection and punishment is also likely to have a "within industry" effect since
some industries are more likely to be targeted for inspections and enforcement by the EPA and
state agencies. Thus, two variables are created that indicate percent of inspections and enforcement
actions that occur within the same Standard Industrial Classification (SIC) in the EPA Region in
which the company's facilities operate for two years prior to the company's disclosure or
discovery of a CAA violation. For companies that have more than one facility operating in more
than one region, these inspection and enforcement rates are averaged across regions. The percent
of SIC regional enforcement actions can be estimated by taking the total number of SIC regional
facilities where an enforcement action occurred divided by the total number of SIC regulated
facilities in that region. The percent of SIC regional inspections can be estimated by taking the
total number of SIC regional regulated facilities that have been inspected divided by the total
number of regional SIC facilities operating in that SIC code. It is hypothesized that the odds of
disclosure under the Audit Policy are greater for companies that operate within industries that have
a high percentage of enforcement actions and/or inspections than for companies that operate in
industries that have a low rate of enforcement actions and/or inspections.
Variables measuring compliance assistance
Regulatory agencies have run several special programs that provide compliance incentives
to companies in order to facilitate disclosure of environmental violations under the Audit Policy
(Office of Regulatory Enforcement 1999). Corporate feedback on compliance assistance is largely
positive. Regulated entities that have received relief under the Policy suggest that compliance
12
-------
incentives have increased their self-reporting of environmental violations (EPA 1999). This
assertion, however, has yet to be empirically examined. To examine the extent to which disclosure
of environmental violations under the Audit Policy are related to regulatory assistance, one
variable is created that indicates whether the company is likely to have received information
concerning a compliance incentive (e.g., a letter providing the company with important incentives
related to the Audit Policy including information that the EPA is about to target that industry for
environmental violations). That variable is dummy coded with a score of "0" if the company is not
likely to have received any Audit Policy assistance in the year prior to disclosure or discovery and
"1" if the company is likely to have received assistance one year prior to disclosure or discovery.
The list of industries targeted for compliance incentives is published in the EPA's Audit Policy
Update.
Variables measuring corporate size
Size of corporations has also been identified as an important variable in corporate
performance. It has been argued that the larger and more decentralized the company, the greater
the likelihood of non-compliance (Coleman 1992; Shover and Bryant 1993). Other research has
indicated, however, that larger corporate size may actually increase environmental compliance
because large corporations have the ability and resources to minimize environmental harm (Florida
1996). To measure the size and decentralization of companies, three variables are created. The first
variable represents the number of company facilities. The second variable represents the number
of employees in the entire company. The third variable represents total company sales (in millions
of dollars). These variables are constructed from information available in the Dun and Bradstreet
database. Since data is not available in Dun and Bradstreet for all facilities, information is also
taken from Standard and Poor's Corporate Descriptions. Bivariate correlations suggest that
13
-------
branches, sales, and employees are highly correlated. Moreover, these variables are theoretically
similar in that they all, to some extent, measure some aspect of company size. Therefore, branches,
sales and employees are combined together to create a company size factor, which is the principal
component index of the three variables. This size component was highly consistent and all
component loadings exceeded .8.
Analysis
Three questions are addressed in this analysis. First, do variables related to specific
deterrence influence whether or not a company will disclose a CAA violation under the Audit
Policy? Second, do variables related to general deterrence influence whether or not a company will
disclose a CAA violation under the Audit Policy? Third, do variables that measure compliance
incentives influence whether a company will disclose a CAA violation under the Audit Policy?
Table 1 begins to address these questions.
[Table 1 about here]
Table 1 reports the mean values for the variables used in this analysis, broken down by
Audit Policy use. All specific deterrence variables have higher means for companies that disclosed
a CAA violation to the EPA than for companies that were discovered by the EPA to have violated
the CAA. Only one of those differences is statistically significant however. In the case of
administrative actions, only 1 in 20 companies discovered to have violated an environmental law
used the audit policy while nearly 1 in 5 companies who reported a violation under the Audit policy
had an administrative action against them two years prior to discovery or disclosure.
In the case of general deterrence, all means were higher for companies that disclosed a
CAA violation to the EPA than companies that were discovered by the EPA to have violated the
CAA. Except in the case of civil enforcement, companies that were located in states where civil
14
-------
enforcement rates were lower, on average, were more likely to disclose a violation of the CAA.
Two general deterrence variables were statistically significant. Both SIC inspections and SIC
enforcements within EPA region appear to be positively related to Audit Policy use. In the case of
inspections, companies that had their violations discovered operated facilities in a region in which
23.1 percent of similar industrial operations were inspected. In contrast, companies that reported
their violations to the EPA under the Audit Policy operated facilities in a region in which 34.6
percent of similar industrial operations were inspected. In the case of formal and informal
enforcement patterns, companies that had their violations discovered operated facilities in a region
in which 5.79 percent of similar industrial operations had an enforcement action. In contrast,
companies that reported their violations to the EPA under the Audit Policy operated facilities in a
region in which 8.90 percent of similar industrial operations had an enforcement action.
Again only in the case of compliance incentives, companies that used the Audit Policy, on
average, are much more likely to have been targeted by a compliance initiative. Finally, according
to the company size factor larger companies tend to use the Audit Policy.
Table 2 reports the odds ratios (OR) which are estimated from logistic regression predicting
whether or not a company used the Audit Policy. Since EPA regional variations are not controlled
for in this analysis, but may be important, fixed-effects logistic regression is also used to estimate
odds ratios (StataCorp, 2001). Fixed-effects logistic regression enables the examination of the
possibility that apparent effect of specific deterrence, general deterrence, and compliance incentive
variables reflect the effects of a network of variables correlated with EPA region. Since regional
enforcement patterns vary greatly it could be argued that what appear as effects of variables in the
model could in fact be the effects of EPA regional enforcement patterns or practices (U.S. GAO
2000: 7). One drawback with using fixed-effects logistic regression is that the procedure relies on
15
-------
variation within the matched sets (i.e., variation on Audit Policy reporting within each EPA
region). Regions with no variation on the dependent variable are then uninformative and are thus
discarded. For this reason the sample sizes for the fixed effects analysis is 107 rather than 118 as it
only represents the number of cases in informative regions, not the total number of cases.
[Table 2 about here]
As Table 2 demonstrates, when controlling for company size, none of the specific
deterrence variables are related to Audit Policy use. This null finding is consistent with the
corporate deterrence literature. The same null finding occurs for general deterrence variables,
except in the case of the regional inspection rate within the industry. In that case the odds ratio is
1.03 (p<05) and a one-standard deviation increase in the inspection rate for a particular industry
within EPA region (e.g., an increase of 19.4 inspections per 100 SIC facilities in the region)
increases the odds that a company will disclose a CAA violation under the Audit Policy by a factor
of 2.2 (95% CI; 1.09, 4.14). While it must be pointed out that disclosing a CAA violation under the
Audit Policy is a relatively rare event, it is also true that a modest one-standard deviation increase
in the EPA inspection rate within region and industry more than doubles the odds of disclosure
under the Audit Policy. In addition, the effect of SIC regional inspection rates on CAA disclosure is
replicated in the fixed-effects logistic regression model (95% CI; 1.29, 7.00) suggesting that
replication of the ordinary logistic regression results across EPA regions.
The compliance incentive variable is not statistically significant when controlling for
company size suggesting that incentives by the EPA do not increase the odds of disclosing a CAA
violation under the Audit Policy. What is clearly important in each of these models estimated is the
relationship between Audit Policy use and company size. Larger companies are much more likely
to report CAA violations under the Audit Policy than smaller companies. This provides evidence in
16
-------
support of the position that organizational size influences environmental compliance to the extent
that larger companies are more likely to report a CAA violation under the Audit Policy. This
finding is also consistent with the theoretical position that large organizations have more resources
to take advantage of self-policing environmental policies. For example, Florida (1996) found that
the number of employees in an organization was positively related to the number green practices
adopted by that organization. Moreover, the literature on organizational size suggests that larger
organizations have more resources to discover violations through self-audits and also have, on
staff, personnel able to successfully negotiate and use the Audit Policy.
Conclusion
This study is the first to examine company predictors of Audit Policy use. Specifically, this
research has examined the association between variables that measure specific deterrence, general
deterrence, compliance incentives, and company size to determine if they are associated with the
odds of disclosing a CAA violation under the Audit Policy.
The findings of this research add to the corporate behavior literature and are surprisingly
consistent with other studies that find that specific deterrence does not appear to be related to levels
of corporate compliance. Moreover, while not statistically significant, three of the four specific
deterrence variables actually appear to suggest that specific deterrence could possibly decrease the
odds that a company will disclose a violation under the Audit Policy when controlling for company
size. Under this potential scenario Audit Policy use may reflect corporate actor estimates of
sanction certainty in such a way that actors with more past environmental violations believe they
are less likely to be caught and punished for a violation of the CAA in the future. This effect is
referred to in the literature as the "positive punishment effect" (Pogarsky and Piquero 2003). While
these results cannot be used to provide evidence of the positive punishment effect since results are
17
-------
not statistically significant, they also cannot be used to provide evidence that specific deterrence
makes corporate actors more likely to use the Audit Policy. It may be that both the positive and
negative effects of specific deterrence are occurring simultaneously in this sample of companies
and that other corporate characteristics not accounted for in this study may help explain this finding
though an interaction effect. At this point, however, such conjecture is merely speculative and more
research needs to be focused on this finding.
Also consistent with previous research is the finding that industry inspection rates increase
the odds of disclosure under the audit policy. Thus, efforts aimed at increasing the number of
inspections among an industry may improve the odds that companies in that industry will disclose a
violation of the CAA under the Audit Policy. This is an important policy finding since there is an
increasing emphasis on environmental compliance though self-policing efforts.
Finally, this research suggests that company size is an important determinant of Audit
Policy. Past research clearly suggests that size matters in the case of corporate environmental
performance such that larger companies pollute at a greater rate than smaller companies (Grant,
Jones and Bergesen 2002). This research builds on those previous findings and suggests that while
corporate size may increase the rate of polluting, it may also increase the rate at which companies
are willing to disclose violations under self-policing efforts. In short, larger companies have more
resources than small companies that allow them to take advantage of the Audit Policy to mitigate
potential negative consequence that may result from being caught for a CAA violation. This
apparent large company advantage may be the result of compliance personnel who are actually on
staff and work for the company and whose job it is to know how to mitigate company liability.
Indeed, it is primarily large companies and industries that represent large companies who have
responded to a request for comments by the EPA about the Audit Policy. Thus, it appears that
18
-------
company size may be a much more important predictor in determining why companies use the
Audit Policy than variables measuring specific deterrence.
19
-------
References
Block, M., F. Nold, and J. Sidak. 1981. "The deterrent effect of antitrust enforcement." Journal of
Political Economy 89:429-445.
Braithwaite, J., and T. Makkai. 1991. "Testing an expected utility model of corporate deterrence."
Law and Society Review 25(l):7-40.
Braithwaite, J. 2000. "The new regulatory state and the transformation of criminology." British
Journal of Criminology 40(2):222-23 8.
Clarke, M. 1987. "Prosecutorial and administrative strategies to control business crime: Private and
public roles." Pp. 266-292 in Private Policing, edited by C. D. Shearing, and P. Stenning.
Sage.
Clinard, M. 1983. Corporate Ethics and Crime: The Role of Middle Management. Sage.
Cohen, M. A. 1998. "Sentencing the enviornmental criminal." Pp. 229-249 in Environmental
Crime: Enforcement, Policy, and Social Responsibility, edited by M. Clifford. Aspen.
Coleman, J. 1992. "The theory of white-collar crime: From Sutherland to the 1990s." Pp. 53-77 in
White-Collar Crime Reconsidered, edited by K. Schlegel, and D. Weisburd. Northeastern
University Press.
Cullen, F. T., and P. J. Dubeck. 1985. "The myth of corporate immunity to deterrence:Ideology and
the creation of the invincible criminal." Federal Probation 49(3):3-9.
Doob A.N. and C.M. Webster. 2003. "Sentence severity and crime: Accepting the null hypothesis."
Crime and Justice 30: 143-195.
Edelman, L. B., and M. C. Suchman. 1997. "The legal environments of organizations." Annual
Review of Sociology 23:479-515.
Environmental Protection Agency. 1995. "Incentives for Self-Policing: Discovery, Disclosure,
Corrections, and Prevention of Violations." Federal Register 60:66705-66712.
Environmental Protection Agency. 1999. "Evaluation of Incentives for Self-Policing, Discovery,
Disclosure, Correction and Prevention of Violations Policy Statement, Proposed Revisions
and Request for Public Comment." Federal Register 64 (94): 26745-26756.
Florida, Richard. 1996. "Lean and green: The move to environmentally conscious manufacturing."
California Management Review 39 (1): 80-105.
Frank, N., andM. Lombness. 1988. Controlling Corporate Illegality: The Regulatory Justice
System. Anderson.
20
-------
Friedrichs, D. O. 1996. Trusted Criminals: White Collar Crime in Contemporary Society.
Wadsworth.
Gibbs, J. P. 1974. Crime, Punishment, and Deterrence. Elsevier.
Grant, D. S., A. W. Jones, and A. J. Bergesen. 2002. "Organizational size and pollution: The case
of the U.S. chemical industry." American Sociological Review 67(3):389-407.
Kagan, R. A., and J. T. Scholz. 1983. "The criminology of the corporation and regulatory
enforcement strategies." in Enforcing Regulation, edited by K. Hawkins, and J. M. Thomas.
Kluwer-Nij hoff.
Kramer, R. C. 1982. "Corporate crime: An organizational perspective." Pp. 77-94 in White-Collar
and Economic Crime, edited by P. Wickman, and T. Dailey. Lexington.
Makkai, T., and J. Braithwaite. 1994. "The dialectics of corporate deterrence." Journal of Research
in Crime and Delinquency 31(4):347-373.
—. 1996. "Procedural justice and regulatory compliance." Law and Human Behavior 20(l):83-99.
May, P. J., and S. Winter. 1999. "Regulatory enforcement and compliance: Examining Danish
agro-environmental policy." Journal of Policy Analysis and Management 18(4):625-651.
Office of Regulatory Enforcement, EPA. 1999. "Agency launches new compliance and incentive
program for industrial organic chemical sector." Environmental Protection Agency Audit
Policy Update. EPA 300-N-99-001 4 (1):6.
Parker, C. 2000. "Reinventing regulation within the corporation: Compliance-oriented regulatory
innovation." Andministration and Society 32(5):529-565.
Rebovich, D. J. 1998. "Environmental crime research: where we have been, where we should go."
Pp. 341-354 in Environmental Crime: Enforcement, Policy, and Social Responsibility,
edited by M. Clifford. Aspen.
Reiss, A. 1984. "Selecting Strategies of Social Control of Organizational Life." Pp. 23-25 in
Enforcing Regulation, edited by K. Hawkins, and J. M. Thomas. Kluwer-Nij off.
Russell, S., and M. J. Gilbert. 1999. "Truman's revenge: Social control and corporate crime."
Crime, Law and Social Change 32(1): 59-82.
Scalia, J. 1999. "Federal Enforcement of Environmental Laws, 1997." United States Department of
Justice, Office of Justice Programs.
Scholz, J. T. 1991. "Corporate regulatory enforcement and the politics of administrative
effectiveness." American Political Science Review 85(1): 115-136.
21
-------
Scholz, J. T., and W. B. Gray. 1997. "Can government facilitate cooperation? An informational
model of OSHA enforcement." American Journal of Political Science 41.
Shover, N., and K. M. Bryant. 1993. "Theoretical explanations of corporate crime." Pp. 141-176 in
Understanding Corporate Criminality, edited by M. B. Blankenship. Garland.
Simpson, S. 1986. "The decomposition of antitrust: testing a multi-level, longitudinal model of
profit-squeeze." American Sociological Review 51(6):859-875.
Simpson, S., and C. S. Koper. 1992. "Deterring corporate crime." Criminology 30(3):347-375.
Stafford, S. L. 2003. "Does Self-Policing Help the Environment? EPA's Audit Policy and
Hazardous Waste Complaince." Pp. 30.
StataCorp. 2001. Stata Statistical Software: Release 7.0. College Station, TX: Stata Corporation.
Stone, C. 1975. Where The Law Ends. New York, NY: Harper and Row.
Tittle, C. R., and R. Paternoster. 2000. Social Deviance and Crime: An Organizational and
Theoretical Approach. Roxbury.
United States General Accounting Office. 2000. Environmental protection: More consistency
needed among EPA regions in approach to enforcement. GAO/RCED-00-108. Washington,
DC: U.S. GAO.
Warr, M., and M. C. Stafford. 1993. "A reconceptualization of general and specific deterrence."
Journal of Research in Crime and Delinquency 30(2): 123-135.
Zimring, F. E., and G. Hawkins. 1973. Deterrence: The Legal Threat in Crime Control. University
of Chicago Press.
22
-------
Table 1 Means for Variables, by Discovery or Disclosure of CAA Violation
Violation Violation
Discovered by EPA Disclosed to EPA
Specific Deterrence Variables
Past Enforcement .603 .700
Administrative Penalty .050 .200
Civil Penalty .002 .003
Past CAA Inspections .899 1.23
General Deterrence Variables
Criminal Enforcement .500 .422
Civil Enforcement .429 .423
Administrative Enforcement 10.2 14.0
SIC Regional Inspection Rate 23.1 34.6
SIC Regional Enforcement Rate 5.79 8.90
Compliance Incentive Variables
SIC Incentive .080 .340
Organization Variables
Facility Size Factor -.590 .590
Note: Bolded values indicate a statistically significant (p<05) difference between discovery and
disclosure. Results were derived from Difference of Means Tests and Mann-Whitney U
Tests. In both instances results were nearly identical for all variables.
23
-------
Table 2 Logistic Regression of Audit Policy Use
OR (95% Confidence Interval)
[Fixed Effects OR (95% Confidence Interval)]
Specific Deterrence Variables
Past Enforcement
.904 (.550, 1.49)
[.873 (.528, 1.44)]
Administrative Penalty
1.50 (.281, 8.04)
[1.11 (.192, 6.41)]
Civil Penalty
.495 (.030, 8.26)
[.437 (.017, 10.7)]
Past CAA Inspections
.793 (.539, 1.17)
[.730 (.473, 1.13)]
General Deterrence Variables
Criminal Enforcement
.986 (.146, 6.67)
[.295 (.015, 5.52)]
Civil Enforcement
1.207 (.224, 6.50)
[1.04 (.052,21.1)]
Administrative Enforcement
1.03 (.980, 1.07)
[1.07 (.992, 1.17)]
SIC Regional Inspection Rate
1.03 (1.01, 1.07)*
[1.04 (1.00, 1.09)]*
SIC Regional Enforcement Rate
.964 (.893, 1.04)
[.931 (.836, 1.04)]
Compliance Incentive Variables
SIC Incentive
2.13 (.600, 7.57)
1.20 (.211, 6.86)
Organization Variables
Facility Size Factor
5.35 (2.87, 9.96)*
4.22 (2.42, 7.37)*
4.01 (2.37. 6.81)*
[4.22 (2.18, 8.18)]*
[3.13 (1.69, 5.83)]*
[2.99(1.69,5.27)]*
Constant
1.31
.184
-.154
-2 Log Likelihood
114
107
116
[81.0]
[78.2]
[85.5]
Sample Size
118
118
118
[107]
[107]
[107]
Note: OR = odds Ratios calculated from logistic regression and fixed effects logistic regression.
* p<05
-------
Regulation and Compliance Motivations:
Marine Facilities and Water Quality
Peter J. May
University of Washington
Paper prepared for the EPA Workshop on
Corporate Environmental Behavior and the Effectiveness of Governmental Interventions
April 26 - 27, 2004
Washington, D.C.
Author Contact:
Professor Peter J. May
Center for American Politics and Public Policy
Department of Political Science
Campus Box 353530
University of Washington
Seattle, WA 98195-3530 USA
(206) 543 9842 fax (206) 685 2146
Email: pmav@u.washington.edu
25
-------
Regulation and Compliance Motivations:
Marine Facilities and Water Quality
ABSTRACT
This research examines how traditional regulatory and voluntary approaches affect motivations for
addressing potential harms to water quality. The traditional approach consists of governmental
enforcement of mandatory requirements. The voluntary approach consists of government calling
attention to potential harms and facilitating actions to address them. These approaches are best thought of
as endpoints for an array of regulatory forms, rather than as the sole choices in regulatory approach.
Three sets of findings emerge from the research. One, not surprisingly, is that traditional regulation is
more effective than is the voluntary approach alone in enhancing attention to water quality issues. A
second shows that deterrent fears and sense of duty to comply are important motivations for action in
addressing water quality. A third concerns factors that account for the variation in each motivation for
which inspections, peer reputation, and attitudes toward government are shown to be important
considerations.
These findings suggest a greater need for attention to regulatory arrangements and how those
arrangements shape affirmative considerations for the attainment of regulatory goals. Traditional
regulation establishes one set of arrangements that this research shows affect both deterrent fears and
civic duty. Requiring permits is a powerful tool for gaining attention to potential problems and for setting
forth expectations as to what constitutes the civic duty to comply. Yet, simply requiring permits is not
sufficient to motivate action. Technical and financial assistance is often necessary for facilitating and
directing actions to alleviate potential harms. These findings therefore confirm what is well known about
the importance of building commitment and capacity to take action.
The willingness of facilities that are subject to voluntary regulation, as with marinas in this research,
to take action rests largely on their good will and their civic intent. Threats of stronger regulation or
regulatory actions loom in the background making such regulation "quasi-voluntary." Showcase actions
are important in underscoring this backdrop. But, the central issue is how to enhance the commitment to
protect water quality. The challenge is to create a stronger sense of civic duty in forming what might be
considered a societal contract for protecting water quality. Actions to address water quality are stronger
when facility operators view water quality harms as a shared problem that they have a civic duty to
address. Social influences are a key to enhancing civic commitment and shared obligations to protect
water quality. These can be enhanced by strengthening associational ties (i.e., among trade groups), by
facilitating the emergence of strong industry leaders, and by providing credible evidence that action is
appropriate.
26
-------
Introduction
Despite bringing about substantial improvements in environmental quality since the early 1970s,
environmental regulations have been criticized in many quarters for having unreasonable requirements
and for engendering heavy-handed enforcement. These criticisms and the realization that the
environmental problems of today are different than those of thirty years ago have spawned a search for
viable alternatives to traditional forms of environmental regulation. Discussion of alternative regulatory
approaches has occurred in a number of books and articles (e.g., Andrews 1999; Fiorino 1995, 2001;
Gunningham and Grabosky 1998; Kettl 2002) as well as in various advisory reports (e.g., National
Academy of Public Administration 1995). The search for alternatives has brought attention to two broad
classes of regulatory approaches that have been labeled cooperative and voluntary regulation.
Cooperative regulatory approaches entail collaboration between regulatory authorities and regulated
entities either in negotiating requirements (Marcus et al. 2002; Kerwin 1999) or shared agreements for
monitoring performance and obtaining compliance (Harrison 1999; Sparrow 2000). The voluntary
approach entails a governmental role in calling attention to a potential harm and in facilitating voluntary
actions by relevant firms or industry associations to address the potential harms (Coglianese and Nash
2001; Lyon and Maxwell 2001; Potoski and Prakash 2002). While there are many variants for each of
these approaches, a basic distinction between them and traditional regulation is the degree to which
government compels actions to address environmental harms. Of particular interest for this research is
the difference between mandatory regulation that involves a central role for government and voluntary
approaches that entail a much more limited governmental role.
The notion that firms would voluntarily undertake actions to reduce pollution or go beyond minimum
requirements is contrary to the image of profit-seeking firms that ignore the external impacts of their
operations. However, experience with a diverse set of voluntary programs demonstrates that at least some
firms are motivated to act out of a combination of a sense of civic duty, good public relations, market
differentiation, and a fear of more stringent governmental regulation (Arora and Carson 1996; Prakash
2000; Segerson and Miceli 1998). At the same time, the limitations of these programs has been
demonstrated by related research that shows participation by firms is uneven, the gains in pollution
prevention are sometimes limited, and the programs are difficult to sustain and expand beyond a core
group of committed entities (Andrews et al. 2001; King and Lenox 2000; Napier and Johnson 1998;
Welch et al. 2000). Taken together, these diverse studies suggest that voluntary programs have promise
but are not a panacea.
The promise is especially important to consider because in some circumstances voluntary programs
may be the only realistic alternative. Mandatory regulation is more feasible when harms are visible and
firms concentrated so as not to overwhelm enforcement capacity. Under such circumstances, pollution
levels can be directly measured and monitored, the responsible parties are relatively easy to identify, and
enforcement can be accomplished on a regular basis if there is sufficient number of enforcement staff. In
contrast, voluntary approaches may be the only practical alternative when the sources of harms are widely
dispersed and the cost of inspection, relative to the potential harm, is high. For example, many forms of
nonpoint sources of water pollution involve widely distributed sources of harm for which it is often not
practical to enforce regulations on a case-by-case basis other than by exception (i.e., when large spills
occur). Given this, the relevant question is not whether voluntary programs are more effective than
traditional regulation, but how can voluntary programs be improved?
Understanding the motivations of firms to address potential harms is important for gauging the
promise and limitations of different regulatory approaches. If firms are motivated by a sense of duty and
concerns about potential harms, the voluntary approach has promise. If these motivations are lacking,
more coercive approaches may be necessary. As such, regulation and motivations go hand-in-hand. The
form of regulation is predicated upon assumptions about willingness to act. And, the willingness to act is
affected by the form of regulation. In getting at this interplay, the basic issues for this research are
27
-------
whether motivations and willingness to act are different under mandatory and voluntary approaches to
regulation.
This research considers these by examining the influence of different regulatory approaches upon
actions taken by marine facilities to address water quality as well as upon their motivations to take action.
A natural experiment is provided by the fact that boatyards are typically subject to traditional regulation
while marinas are typically subject to voluntary approaches. As explained in what follows, boatyards
entail point sources of water pollution. Marinas are typically addressed as non-point sources of water
pollution. By examining the role of alternative approaches and other factors in shaping actions and
motivations to address potential harms to water quality by these facilities, this research provides insights
about both the efficacy of alternative regulatory forms and potential steps to improve voluntary
approaches.
Mandatory and voluntary approaches
Mandatory regulations constitute the traditional toolkit for motivating action to address potential
harms and to protect public welfare (see May 2002). Compliance with regulatory requirements is
compelled through enforcement actions and imposition of sanctions for those found to be out of
compliance. The basic logic of this approach is a criminal law model of deterrence as examined in the
seminal work of Becker (1968; also see Ehrlich 1972). From this perspective, individuals and firms
comply because they fear the consequences of being found in violation of regulatory requirements.
Inherent in this approach is the presumption that regulated entities are unwilling to take necessary actions
to comply with regulations and therefore they must be compelled to do so. Obtaining greater compliance
is brought about by reinforcing deterrent fears.
The voluntary approach involves a very different governmental role and different assumptions about
the willingness of relevant entities to address potential harms. The governmental approach is non-
coercive for which the key policy instruments are exhortation and various forms of assistance (Grabosky
1995; Harrison 1999; Wilms 1982). Rather than mandating action, government promulgates guidelines
for best management practices and encourages adherence to them as a means of achieving desired
outcomes.1 Desired actions are encouraged through education, financial assistance, technical assistance,
and other inducements. The assumption is that those entities that contribute to environmental harms are
willing to address the harms but they fail to do so because of various complications (Brehm and Hamilton
1996; Coombs 1980; Kagan and Scholz 1984). Their failure to do so is either because they do not
recognize the existence of a problem (requiring information), do not understand what can be done to
address the problem (requiring education), or they do not have the capacity to take desired actions
(requiring financial or technical assistance).
Mandatory and voluntary approaches are perhaps best thought of as endpoints for an array of
regulatory forms, rather than as the sole choices in regulatory approach. Stated differently, variants of the
mandatory approach can involve efforts to educate regulated entities about appropriate actions and can
make available different forms of technical and financial assistance to facilitate compliance. There is
always the possibility of legal action against those who contribute to specific harms under the voluntary
approach. When pursued, these actions push the voluntary approach along the continuum toward that of
traditional regulation. In the case of water pollution, for example, harmful acts can be pursued through
the broader water-quality provisions of federal and state water quality acts. In addition, the voluntary
approach at least implicitly holds the threat of switching to traditional forms of regulation if harms are not
sufficiently addressed. For these reasons, it is perhaps better to think of this type of voluntary regulation
as "quasi-voluntary." This form of regulation is labeled voluntary in the discussion that follows for ease
of communication while keeping in mind the coercive backstop to such regulation.
HI: The mandatory approach is more effective than the voluntary approach in gaining
commitment to address potential harms by those entities that potentially contribute to
those harms.
28
-------
This hypothesis follows from the preceding discussion. Both mandatory and voluntary approaches are
expected to positively influence the actions that are undertaken to address potential environmental harms.
The mandatory approach is expected to be more effective in enhancing actions because it compels
compliance. However, the effectiveness of the approach will be undermined if regulations are not taken
seriously or if enforcement is lax. The voluntary approach is expected to engender actions by those firms
that are predisposed to act by virtue of having a higher degree of concern about potential harms, stronger
capabilities to act, or stronger fears of future regulation. However, the voluntary approach may fall short
in failing to address recalcitrant entities. Commitment to address environmental harms can be gauged by
the importance that relevant entities attach to their actions to address the harms.
Motivations to act
Consideration of the motivations of entities to take action in addressing potential harms is important
for understanding the viability of different approaches. Most of the regulatory literature on the subject is
concerned with the reasons why firms or individuals comply with regulations (see Winter and May 2001).
Less attention has been paid to motivations in the absence of formal regulations, as with the voluntary
approach, although the literature addressing the willingness of firms to voluntarily adhere to industry-
developed guidelines is relevant (see Arora and Carson 1996; Prakash 2000; Segerson and Miceli 1998;
Welch et al. 2000). Two sets of motivations that are involved are considered in what follows. One,
deterrent fears, concerns the consequences of failing to address potential harms. A second, duty to
comply, is a sense of obligation to address a potential harm.
Deterrent fears and regulatory approaches
A primary motivation for addressing potential harms is the deterrent fears associated with the
negative consequences of failing to act. As discussed by Reiss (1984), such fears are potentially activated
either through actions aimed at particular firms or individuals (specific deterrence) or through more
general warnings (general deterrence). Specific deterrence is brought about by targeting individual
entities for inspection, citations of violations of rules, and sanctions if firms fail to rectify those violations.
A number of studies (Braithwaite and Makkai 1991; Burby and Paterson 1993; Gray and Scholz 1993;
Helland 1998) reinforce the importance of inspections for compelling compliance, but vary in their
findings with respect to the influence of sanctions once violations are detected. General deterrence,
which serves to warn potential violators that negative consequences can follow from failing to act, is
brought about by more diffuse actions that include showcase actions against violators and publicity about
the consequences of failing to act. Potential shame or embarrassment of being caught violating rules has
also been shown to be an important component of deterrence (Grasmick and Bursik 1990; Grasmick et al.
1991).
H2: The mandatory approach activates a stronger sense of deterrent fears than does the
voluntary approach.
This hypothesis follows from differences in the way that the two approaches potentially activate deterrent
fears. As discussed above, the mandatory approach is based on compelling compliance by activating
deterrent fears of the negative consequences of failing to comply. This is accomplished through specific
deterrent actions involving enforcement and sanctions, and through general deterrent actions involving
showcase enforcement actions.
Specific deterrence is not relevant to the voluntary approach because there are no mandatory rules or
regulations to enforce. However, aspects of general deterrence do potentially apply. Embarrassment
might be relevant if an entity is shown to be a laggard in addressing potential harms. This is the logic of
posting "top 10 wanted lists" of environmental polluters that have been issued by environmental
organizations. A second potential fear is that a firm may be found liable for harms that, although not
specifically regulated, are subject to broader prohibitions. This fear is potentially activated by showcase
lawsuits. A third potential fear that failure to act will result in stronger governmental regulation.
29
-------
In sum, both approaches are expected to engender deterrent fears that serve as one key set of
motivations for taking action to address potential harms. The mechanisms in fostering deterrent fears
differ somewhat between the two approaches. The mandatory approach relies on both specific and
general deterrence, while only general deterrence is relevant to the voluntary approach. Because of the
greater arsenal of tools for engendering deterrent fears with specific deterrence, the mandatory approach
is expected to be stronger in activating such fears. Evidence for stronger deterrent fears consists of the
degree of concern that entities display of potential consequences for failing to act to address potential
harms.
Duty to comply and regulatory approaches
An important foundation for any form of regulation is that citizens and firms acknowledge their
responsibilities in preventing harms. Even when rules exist and are enforced, as with the mandatory
regulatory approach, there is a strong presumption that the vast majority will obey their sense of duty to
adhere to the rules. As noted by Tyler (1990), without such quasi-voluntary compliance there would be a
need for a much stronger enforcement apparatus. This duty is based on a combined sense of moral
obligation and agreement with the need and approach for addressing a given problem (McGraw and
Scholz 1991; Scholz and Pinney 1995; Winter and May 2001). As verified with findings from this
research, the sense of duty to comply can be thought of as independent of the deterrent fears discussed in
the previous section.
The more basic aspects of duty to comply relate to the moral sense of obligation to obey laws and
general ideological values. The more immediate aspects of this duty relate to the value of addressing a
given problem and acceptance of the regulatory approach to it. Such acceptance has been shown to
depend on the reasonableness of rules or guidelines, degree of trust in the agencies that promulgate them,
and the extent to which other affected entities are believed to be doing their part in addressing the
problem (Bardach and Kagan 1982; Kagan and Skolnick 1993, Levi 1988; Scholz and Lubell 1998; Tyler
1990).
H3: The voluntary approach activates a stronger sense of duty to comply than does the
mandatory approach.
This hypothesis follows from the preceding discussion. Both approaches are expected to invoke a sense
of duty to comply. However, the mandatory approach is expected to be less effective in this regard
because aspects of it potentially undermine the trust in regulatory authorities that is essential for instilling
a sense of duty to comply. The voluntary approach potentially activates a sense of duty to comply
through at least two means. One means is direct appeals to address potential harms through informational
efforts telling how actions of affected entities potentially contribute to the harms. Implicit, if not explicit,
is an appeal to the shared obligation to do one's part. A second means is more direct, one-on-one
technical assistance to help educate affected entities about harms and steps to address them. This has the
added potential benefit of gaining acceptance of the legitimacy of the recommended actions. Such
acceptance has been shown by Winter and May (2002) to depend on the extent to which those providing
advice are viewed as being competent and can be trusted. If these are lacking, the voluntary approach
loses its effectiveness and the sense of obligation to comply is not enhanced.
Because the mandatory regulatory approach also typically entails provision of information and
technical assistance, it also can be expected to activate a sense of duty to comply. Information and
assistance is provided both as part of general publicity about regulations and when facilities are inspected.
As with the voluntary approach, the trust that regulated entities place in regulatory agencies and the
perceived competence of regulatory officials are important considerations. Yet, additional considerations
also come into play. The style with which inspectors interact with regulated entities—whether inspectors
are formal or facilitative—has been shown to affect motivations for compliance and the understanding of
rules by regulated entities (Winter and May 2001). If enforcement of regulations is viewed as
unreasonable, the essential trust in regulatory authorities and sense of duty to comply will be undermined.
30
-------
Because of this potential, the mandatory approach is expected to at least partially undermine the sense of
duty to comply.
In sum, both approaches are expected to engender a sense of duty to comply that serve as a second
key set of motivations for taking action to address potential harms. The mechanisms for this are the same
for both approaches. However, the mandatory approach has the potential for a backlash against what
might be perceived as heavy-handed enforcement that in turn undermines trust in regulatory authorities
and the sense of obligation to comply. Because of this potential, the mandatory path is hypothesized as
being less effective in activating a sense of duty to comply.
The regulatory setting: Marine facilities and water quality
The setting for this research concerns marine facilities and their impacts on water quality. Marine
facilities have been shown in several studies to be sources of water pollution for estuarine areas and bays
for which there is little natural flushing (National Research Council 2000; U.S. EPA 1994, 1997). These
impacts include point sources of water pollutants—toxic metals, oils, sediments—and potential releases
of hazardous materials and toxic spills into water. Also relevant are non-point impacts from boats of raw
or untreated sewage and fuel spills. A variety of consequences can follow that include increased toxicity,
turbidity, and lower oxygen content of water. These in turn contribute to changes in the food chain in the
marine environment, altered marine life, contaminated fish, increased algae growth, and generally
unhealthy water quality.
The contamination of water is regulated by the federal government under several laws. Point sources
are addressed under provisions regarding industrial discharge under provisions of the Clean Water Act
(section 402(p)) that establish among other things the National Pollution Discharge Elimination System
(NPDES) permit system. Non-point sources of pollution are addressed in the 1987 amendments to the
Clean Water Act (section 319), and under coastal non-point sources of pollution provisions (section 6217)
of the Coastal Zone Management Act Reauthorization Amendments of 1990. These federal laws assign
key responsibilities to states for enforcement of discharge permit programs and for development of plans
for management of pollution sources.
The data for this research address behaviors of operators of boatyards and of marinas in coastal and
estuarine areas of California and Washington states. The contrast between the regulatory approaches for
each set of facilities provides a natural experiment for examining the interplay of approaches and
motivations to address harms. Boatyards are working marine facilities for repair and maintenance of
pleasure and smaller commercial vessels. These are fixed point sources of pollution from boat haul-outs,
hull pressure washing, sanding and painting, and repair operations. Boatyards are subject to mandatory
regulations and permit requirements under the NPDES program as delegated to the states.
Marinas are basins with slips for boats, usually defined as having 10 or more slips, which may or may
not provide fueling and other shore services. These have limited point sources (fueling and boat
launching areas) and more extensive non-point sources of pollution relating to potential spills from boats
of sewage, bilge, fuel, and cleaning materials. Marinas, with some exceptions, are not subject to
regulatory requirements other than the broad prohibitions of federal and state water quality acts against
impairing water quality.2 Instead, marinas are subject to voluntary approaches for which states have
encouraged, through education efforts and limited funding, adherence to best practices guidelines for
marine facilities promulgated by the U.S. Environmental Protection Agency (1993, 2001). State efforts in
this regard include information provision about best management practices, technical assistance, and
small grant programs aimed at fostering voluntary action by marinas (see Washington State Department
of Ecology 1999a).
Interviews with enforcement officials in each of the relevant regional offices of the states within
which facilities for this study are located help set the context. Neither type of facility is high priority for
enforcement actions. Because boatyards are required to have permits, they are monitored more
31
-------
extensively than are marinas. This typically consists of monitoring the filing of required, self-reported
periodic analyses of water quality samples. In most regions, field inspections of boatyards are undertaken
only every few years or less. Marinas generally fall "under the radar scope" of enforcement officials
unless there is a complaint or a spill that draws attention to problems. It is difficult to assess compliance
for either type of facility in that there is no systematic collection of reporting of field inspections other
than inspection reports filed away in regional offices. Only more serious notices of violations or fines
make it into centralized reporting. For the period January 2000 to July 2002, California state enforcement
data indicate issuance of formal sanctions for 11 boatyards and two marinas, and Washington state
enforcement data indicate formal sanctions being issued for 10 boatyards and two marinas. These
numbers correspond to the self-reporting of facilities in the survey for this study of 9 percent of the
facilities having actions taken against them for non-compliance with either reporting requirements or
other provisions.
Several aspects of marine facilities are important to consider in setting the context for this research.
Of particular concern are the comparability of the two types of facilities and the potential for systematic
motivational differences between them. The rationale for mandatory regulation of boatyards are the
greater potential harm to water quality posed by activities at boatyards and the greater ease with which
point sources of harm can be regulated. The former might suggest that boatyard operators would be more
motivated to take action but that is vitiated by the larger costs for boatyards to address potential harms.
Capital expenditures for equipment addressing environmental issues by boatyards in this study were on
average 30 percent higher than those expended by marinas. As such, it is unclear that boatyards are
inherently more motivated to act than are marinas. This potential selection bias is controlled in the
analyses that follow by introducing statistical controls concerning measures of the extent of water quality
problems and constraints upon action.
A different potential bias is that one type of facility has greater capacity to act than another. There is
no systematic evidence for this. Both types of marine businesses tend to be relatively small operations.
Among the sample for this study, the median number of full-time employees is 14 for boatyards and 3 for
marinas. Thirty percent of the marinas and 10 percent of boatyards have owners who operate more than
one facility, which is indicative of a larger scale of operations. Most boatyards, comprising 92 percent in
this sample, are privately owned while marinas have a wider mix of ownership, comprised of 30 percent
that are publicly owned facilities. Potential selection biases stemming from differences in capacity are
controlled in the analyses that follow by introducing statistical controls measuring the capacity of
facilities to address potential harms to water quality.
Operators of marine facilities are as a group conflicted in their views about addressing harmful water
quality impacts. Many are in the business because they enjoy boating and value clean water, but many
also question the need for governmental regulation. Among the study sample, 31 percent of marina
operators and 43 percent of boatyard operators moderately to strongly agree that government intrudes too
much on business and that preserving property rights is more important than protecting water quality.
Facility operators also share a perception that their efforts to address water quality pale in comparison to
what they typically view as much more harmful impacts by such sources as municipal storm water
outfalls and industrial discharges. In our interviews, boatyard operators in particular felt that their
businesses were being singled out for attention when far greater sources of potential harm were being
ignored.
Data and measures
The data were collected from a mail-out survey of marine facilities in California and Washington.
Listings of 281 marinas and 140 boatyards in coastal and estuarine areas of the two states were identified
from various publications and governmental permit listings. Questionnaires were mailed in late January
2002 with two rounds of follow-up lasting into March 2002. Valid responses were received from 144
marinas (51 percent response rate) and from 61 boatyards (44 percent response rate) for an overall
32
-------
response rate of 49 percent. Data collected from a preliminary phone contact with all facilities were used
to gauge potential response biases. No statistically significant differences were found for marinas with
respect to the number of slips or type of marina ownership, and none were found for boatyards with
respect to number of employees or type of boatyard ownership. Although this non-response assessment is
limited in scope, it suggests that the mail-out survey responses are representative of a range of marinas
and boatyards in the two states. The unit of analysis is the individual marine facility. The data are
weighted to take into account differences in response rates among the two types of facilities and among
different geographic regions.3
The measures for this research relate to testing hypotheses concerning the role of regulatory
approaches and other factors in influencing deterrent fears and sense of duty to comply among marine
facility operators. All measures are drawn from the survey results. The key distinction is between the
mandatory and voluntary approaches as reflected in the different treatment of boatyards (mandatory
regulation) and marinas (voluntary approach). The differences in approach for the two sets of facilities
are verified in analyses that follow.
The two sets of motivations are central considerations. Each is measured as a dichotomous variable
for which a contrast is drawn between those facilities that score high on the motivation and the remaining
facilities that score lower. This coding overcomes the statistical limitations of working with the ordinal
measures for each set of motivations that were contained in the surveys of facilities. Deterrent fears are
measured by respondents' rating of the fear of legal liability if harm occurred, fear of fines or other
governmental action, concern that new requirements will be imposed if actions are not taken on their own,
and avoidance of embarrassing media coverage.4 The sense of duty to comply is measured by respondent
agreement with two statements: "Regardless of the extent of the problem, marinas (boatyards) have a
civic duty to address water quality issues," and "Marinas (boatyards) have a civic duty to report accidents
or spills affecting water quality to relevant authorities."5
The commitment of marine facilities to address water quality problems is measured in three different
ways. The first is the percentage of best management practices that are undertaken by each facility.6 As
detailed in the methodological appendix, these include installation of physical devices, use of appropriate
equipment, restrictions on activities at the facility, and management practices that encourage appropriate
actions by facility operators and boat owners. Standardizing this measure by computing the percentage of
best practices makes it possible to compare marinas and boatyards with respect to actions undertaken. A
second measure is respondent self-identification of the priority for their actions along a five-point scale
that ranges from "not an issue that has really affected the facility" to "a top priority—a key consideration
for facility upgrade." A third measure is the respondent's rating of efforts to address water quality in
comparison to other facilities on a five-point scale from "less attention than most" to "we are a leader."
Several factors are potentially relevant for explaining variation in deterrent fears and sense of duty to
comply. These include various attributes of each regulatory approach, attitudes toward governmental
agencies and their handing of water quality issues, social influences, and the capacity to and constraints
upon the ability to act. The regulatory attributes include specific deterrent enforcement practices
(inspection within the past five years, whether or not sanctions were imposed, and awareness of showcase
actions) and whether different types of assistance were provided (general information or education,
technical assistance, and financial assistance).
Relevant attitudes about government include ideological perspectives, the perceived reasonableness
of water quality regulations, and the perceived competence of governmental regulators. The ideological
perspective about appropriate roles of government is measured as an index of mean agreement on a five-
point scale with two statements: "In general, governmental agencies and regulations intrude too much on
businesses," and "Preserving property rights is more important than protecting water quality."7 The
perceived reasonableness of governmental water quality regulations is gauged by agreement on a five-
point scale with the statement: "The existing governmental rules concerning boatyards (marinas) and
33
-------
water quality are reasonable." Perceptions of the competence of government agencies are gauged by
agreement on a five-point scale with the statement that facility managers "can rely on governmental
agencies for advice about dealing with potential water quality impacts."
The perceived importance of peer reputation is the primary social influence for which information
was obtained. This is measured as the mean score on a five-point scale of the extent to which respondents
agree that their reputation with each of the other facilities, boaters, and governmental agencies is "an
important consideration for how we do business."8
The final set of potentially relevant considerations are various aspects of the capacity of and
constraints upon actions that facilities take to address water quality. The size of the facility, measured as
the percentile of facility size relative to other facilities, serves as a proxy for both capacity to act and the
costs of taking action. An index of perceived constraints is measured as the mean of respondent ratings
on a five-point scale of the extent to which four factors "make it difficult to address water quality."9 Two
measures serve as proxies for the extent of water quality problems. One is respondent perceptions on a
ten-point scale of water quality at their location five years ago. The second proxy is the amount of
waterfront that is adjacent to the facility.
Findings
The findings are presented in several stages. The attributes of the mandatory regulatory and
voluntary approaches are first discussed. The comparative impacts of the two approaches upon the
willingness of marine operators to act in addressing potential harms to water quality are next considered.
This leads to empirical examination of differences in deterrent fears and sense of duty to comply for the
two approaches. Finally, multivariate analyses that explain variation in each set of motivations are
reported.
Regulation of marine facilities
Table 1 shows systematic differences in provisions for marine facilities under the mandatory and
voluntary approaches.10 The greatest differences are the permit provisions and inspection of regulated
facilities with a smaller difference in the application of sanctions. Permits are the central regulatory tools
that specify the conditions for facility operation. Fines, warnings, or other actions are more infrequent
with the typical action being a warning about boatyard practices or a fine for failure to submit a required
report on time. More extreme sanctions include orders to remove contaminated soil and mandated
upgrading of catch basins. Regulated facilities also receive information and education assistance as part
of inspections and, to a lesser extent, technical assistance is sought from third parties. Very few regulated
facilities received governmental financial assistance.
Various forms of assistance were provided marinas that are subject to the voluntary approach. A
majority of marina operators report receiving information or educational assistance while fewer operators
report receiving technical assistance or funding. Financial assistance was mainly in the form of grants for
installation of sewage pump-out stations. A small percentage of marinas were also subject to sanctions
that resulted from spills or other harms to water quality. The sanctions included small fines for spills of
oil or fuel, a $13,000 fine for allowing vessel repair in the water, and warnings for leaking sewage and
fuel systems.
Overall, the major difference between the mandatory and voluntary approaches as applied to these
facilities is the use of coercive measures to compel compliance under the mandatory approach that are not
present under the voluntary approach. Both approaches employ various forms of assistance. However,
assistance is used in different ways for the two approaches. Under the mandatory approach, assistance is
a mechanism for facilitating compliance. Under the voluntary approach, assistance is a mechanism for
both gaining attention to potential problems and for facilitating action.
34
-------
Table 1. Regulation of marine facilities
Characteristics
All
Facilities Mandatory Voluntary P-value:
(Percentagesb)
Regulatory Provisions
NPDES permitc
Inspected in past 5 years
Fine, warning, other action
Assistance Provided
Information or educational
assistance
Technical assistance
Financial assistance
38
30
11
68
32
20
100
91
19
76
43
2
0
0
6
63
25
31
<.01
<.01
.02
.08
.03
<.01
Number of respondentsc 157 53 104
Notes:
a P-value for t-test of differences for independent samples (two-tailed) for any given row between the
percentages for regulated and non-regulated facilities. Facilities subject to mandatory regulation are
those boatyards with required permits. Facilities subject to voluntary approaches are those marinas
without permits.
b Cell entries are weighted percentage of respondents for each category.
0 Un-weighted number of respondents for each category.
Actions to address water quality
Water quality programs seek to do two things whether they employ mandatory or voluntary
approaches. They seek to gain attention to potential problems so that affected entities take action in
addressing them. In addition, they seek to guide the actions that are undertaken by either mandating or
recommending specific actions. Both of these are reflected by the adoption of best management practices
as well as by the degree of effort that facilities put into addressing water quality. These issues are
addressed in Table 2.
The mandatory approach is expected to be more effective than the voluntary approach in stimulating
adoption of best practices (HI). As hypothesized, those facilities that are regulated report higher levels of
action than those facilities subject to voluntary provisions. Yet, as evidenced by the results for the
unregulated facilities, the voluntary approach appears to have had a notable impact in encouraging best
practices and to a lesser extent in encouraging stronger efforts to address potential harms to water quality.
Those facilities that are subject to the voluntary approach undertake on average 82 percent of actions
undertaken by regulated facilities. As would be expected, the adoption of best practices is greater for
marinas reporting receiving information and education than those that did not (p < .05). However, no
statistical differences are detected between these groups for the other effort measures shown in Table 2.
35
-------
Table 2. Actions to address water quality
All
Facilities Mandatory Voluntary P-valuea
(Percentagesb)
Actions / Effort
Mean percentage of Best
Management Practices undertaken c
Rate water quality as top priority d
Leader in Efforts e
Perceived Impact
Substantial impact in reducing
potential problemsf
Number of respondents 8
Notes:
a P-value for t-test of differences for independent samples (two-tailed) for any given row between the
percentages for regulated and non-regulated facilities. Facilities subject to mandatory regulation are
those boatyards with required permits. Facilities subject to voluntary approaches are those marinas
without permits.
b Weighted percentage of respondents.
0 Percentage of relevant practices for addressing water quality that are undertaken by a facility.
d Percentage of respondents that rate the priority for action as a "top priority - key consideration for
upgrade."
e Percentage of respondents that rate efforts to address water quality, relative to other facilities, with the
response that "we are a leader."
f Percentage of respondents providing a rating of 7 or greater on 10 point scale of "effect of actions in
reducing potential harmful impacts on water quality."
g Un-weighted number of respondents for each category.
The lower part of Table 2 shows the perceptions of the impacts that marine facilities have had upon
reducing harms to water quality. Consistent with the findings concerning levels of effort, a much greater
percentage of regulated facility operators report substantial impacts in reducing potential harms than do
unregulated facility operators. Taken together, these findings suggest that the mandatory approach has a
greater influence on commitment to addressing harms than does the voluntary approach alone. These
findings are expected and are not especially remarkable. The finding that the voluntary approach appears
to have had an influence for a sizeable number of unregulated facilities is notable.
Other analyses of the data from this research provide insights concerning factors that explain
variation in the extent of adoption of best practices by regulated and unregulated facilities (see May
2003).11 Two aspects of the multivariate findings are of relevant for the present discussion. One relevant
finding is that permits, and associated deterrent fears, are noteworthy influences on adoption of best
practices. On average, when controlling for a variety other factors, facilities with permits have a 17
percentage point higher score in the actions that are undertaken than those that are not regulated. A
second relevant finding is that civic duty is also shown to affect the adoption of best practices. The
average effect upon best practices of increased amounts of civic duty is a 5.4 percentage point increase,
69 78
23 31
17 25
65 79
157 53
64 <.01
18 .06
12 .05
57 <.01
104
36
-------
when calculated as moving from the lowest quartile of civic duty to the highest quartile among the data.
Note, however, that civic duty was found to have a stronger effect on best practices for regulated facilities
than for unregulated facilities. This may seem counter intuitive. But, the finding suggests that
regulations (and permits) play an important role in signaling what is desired and what constitutes civic
duty. In the absence of regulations with which to comply, the understanding of what constitutes civic
duty is more amorphous.
The findings reported in May (2003) also show that the willingness of marine facilities to address
water quality is constrained by several factors. Regardless of the approach, facility operators complain
about the complexity of rules governing water quality. As would be expected, this complexity is
perceived as more constraining by those facilities that are subject to traditional regulation. The cost of
devices to address water quality is also viewed as an impediment to taking action. An overarching
consideration is the belief on the part of nearly a majority of facility operators that there is too much
uncertainty about the impacts of marine facilities upon water quality to justify actions on their part. This
belief when coupled with the fact that more than one-third of the facility operators are skeptical of
governmental regulation constrains the acceptance of the need for actions to address water quality.
Motivations to act
One way of beginning to disentangle the effects of the mandatory and voluntary approaches is to
consider their impacts upon motivations to act. Table 3 presents a comparison of various aspects of
deterrent fears and sense of duty to comply. The findings show that deterrent fears and sense of duty to
comply are relevant considerations for both types of facilities. The data further suggest that deterrent
fears are more strongly activated by the mandatory approach, while the sense of duty to comply is
comparable between the two approaches.12
The top part of the table addresses deterrent fears. The hypothesis (H2) is that the mandatory
approach activates a stronger sense of deterrent fears than does the voluntary approach. Consistent with
this hypothesis, these findings show that most aspects of deterrent fears are higher for regulated facilities
than for those that are subject to the voluntary approach. The exception is the failure to statistically detect
a difference between facilities in avoidance of embarrassing media coverage.
The more interesting aspect of these findings about deterrent fears is evidence of a notable percentage
of unregulated facilities expressing deterrent fears. Nearly one-fifth of unregulated marinas express fears
of legal liability and of fines or governmental actions. This reflects the fact that marinas are not truly
unregulated, as they are subject to broader laws that prohibit harms to water quality. The lower
percentage of marinas reporting concerns about future regulatory actions than boatyards likely reflects
differences in the salience of water quality issues for the two types of facilities. The fear of embarrassing
media attention evidenced by some operators reflects the importance of reputation to marine facility
operators.
The bottom part of Table 3 addresses duty to comply. The hypothesis (H3) is that the voluntary
approach activates a stronger sense of obligation than does the mandatory approach. This hypothesis is
not supported by these data. The difference between the regulated and unregulated facilities with respect
to each of the measures of duty to comply is not statistically significant. Nonetheless, with the exception
of regulated facilities' sense of civic duty to report spills, majorities of marine facilities report various
aspects of duty to comply as a strong motivation for their actions to address water quality. The larger
percentage of unregulated marinas reporting a strong civic duty to report spills than that of boatyards,
while not statistically significant, likely reflects the fact that educational efforts targeted at marinas have
emphasized the importance of spill prevention and reporting to assist emergency clean up.
37
-------
Table 3. Motivations to address water quality
All
Facilities Mandatory Voluntary P-valuea
Consideration (Percent rating very strong consideration b)
Deterrent fears c
29
45
19
<.01
Fear of legal liability if harm
33
54
21
<.01
occurred
Concern new requirements will
32
46
23
<.01
be imposed if actions are not
taken on own
Fear of fines or other
27
43
17
<.01
governmental action
Avoidance of embarrassing media
25
32
21
.15
coverage
Duty to Comply d
63
63
62
.88
Civic duty to address water
51
54
50
.66
quality
Civic duty to report spills
55
49
58
.29
Number of respondents e
157
53
104
Notes:
" P-value for t-test of differences for independent samples (two-tailed) for any given row between the
percentages for regulated and non-regulated facilities. Facilities subject to mandatory regulation are
those boatyards with required permits. Facilities subject to voluntary approaches are those marinas
without permits.
b Weighted percentage of respondents rating each consideration as 4.5 or greater on a scale of 1 to 5.
0 Score of 4.5 or greater for the mean of items that follow, each rated on a scale of 1 to 5.
d Score of 4.5 or greater for the mean of items that follow, each rated on a scale of 1 to 5.
e Un-weighted number of respondents for each category.
Explaining variation in motivations
Table 4 presents multivariate results for explaining variation in deterrent fears and sense of duty to
comply. The dependent variables are whether or not a given facility scores high relative to other facilities
for each motivation. Because each variable is dichotomous, logistic regression analyses were
undertaken.13 The t-tests of the coefficients provide an assessment of whether a given variable can
statistically be detected as having an effect on motivations for these data. The sign of each coefficient can
be used to tell whether a given factor has a positive or negative effect on the likelihood of a facility
scoring high on each motivation. The magnitude of the effects of variables for logistic models is not
linear and thus cannot be directly gauged from the size of the coefficients. Assessments of the magnitude
of the impacts of relevant variables on the likelihood that facilities score high on each motivation are
reported in what follows.
38
-------
Table 4. Explaining differing motivations
Logistic Analyses a
Deterrent Fears b Duty to Comply c
Regulatory Approach
Mandatory (versus voluntary) Approach d 1.68 ** .67
(.80) (.77)
Deterrent Enforcement Practices
Inspected within past 5 years [1.58]e ** [.61]e
(.91) (.83)
Sanction imposed for water quality problems .02 -2.10***
(1.10) (.90)
Awareness of showcase enforcement actions .71 ** -.24
(.33) (.27)
Assistance Provided
General information or education .50 .52
(.64) (.66)
Technical assistance -1.81** .47
(.80) (.62)
Financial assistance -.55 -1.09
(.97) (.84)
Attitudes Concerning Government
Conservative ideology .54 * -1.27 ***
(.35) (.38)
Water quality standards are reasonable -.46 ** .27
(.26) (.25)
Perceived competence of government regulators .32 .17
(.27) (.27)
Social Influences
Importance of reputation with others -.23 2.46 ***
(.59) (.61)
Facility Capacity and Constraints
Facility size relative to industry -.01 .01
(.01) (.01)
Implementation constraints .59** -.30
(.36) (.32)
Amount of water frontage (In) .17* .04
(.11) (.10)
Perceived water quality as of five years ago .24 ** -.09
(.13) (.14)
Constant -8.42 -6.43 **
(3.75) (3.35)
Model Statistics
Number of observations 101 123
Pseudo-R2 f .26 .40
Chi-square GOF for model (p-value) 29.70 (<01) 55.21 (<01)
Percent correctly classified 84.8 82.6
39
-------
Table 4 (con't)
Notes:
*** p < .01 ** p < .05 * p < .10 (one-tailed asymptotic t-tests).
" Cell entries are logit coefficients from binary logistic modeling with the standard error in parentheses.
b Contrast between those facilities that score high (scores of 4.5 or greater) and remaining facilities for
items concerning deterrent fears shown in Table 3.
0 Contrast between those facilities that score high (scores of 4.5 or greater) and remaining facilities for
items concerning duty to comply shown in Table 3.
d Facilities subject to mandatory regulation are those boatyards with required permits. Facilities subject to
voluntary approaches are those marinas without permits.
e Results from a separate logistic modeling for this variable when substituted for the regulatory approach
variable. Because only regulated facilities were inspected, the form of regulation and inspection cannot
be entered into the same model.
f Cox and Snell Pseudo R Square.
Regulatory approach. The first row of both models concerns the influence of regulatory approach
upon the two sets of motivations. The coefficient for deterrent fears clearly shows that the mandatory
regulation has a stronger influence on deterrent fears than does the voluntary approach. The multivariate
findings show that regulated facilities, when other factors are taken into account, have a 36 percent
greater likelihood of having high deterrent fears than do unregulated facilities.14 This difference is as
hypothesized and reaffirms the zero-order comparisons of Table 3.
The coefficient for sense of duty to comply is not statistically significant and thus fails to show the
hypothesized greater effect of the voluntary approach upon that motivation. The lack of a difference in
approach for duty to comply might be explained by the fact that the two approaches do not substantially
differ in their levels of assistance as was shown in Table 1. In addition, inspections under the mandatory
approach may not have the off-putting effect that was hypothesized as the reason why that approach
would have a lesser effect on sense of duty to comply. The multivariate findings rules out the possibility
that the lack of a difference in sense of duty to comply is an artifact of other differences among marine
facilities.
Enforcement practices. The items of Table 4 concerning enforcement practices show that
inspections and showcase actions contribute to deterrent fears. Inspections are central aspects of specific
deterrence while showcase actions are important components of general deterrence. Inspected facilities,
all of which are regulated boatyards, have a 35 percent greater likelihood of having high deterrent fears.
Regulated boatyard operators with high awareness of showcase actions have a 17 percent greater
likelihood of having high deterrent fears. The corresponding value for marina operators is 10 percent
greater likelihood. The failure to find an effect of sanctions on deterrent fears is consistent with findings
in the regulatory literature that inspections are a more important predictor of compliance than are
sanctions (see Braithwaite and Makkai 1991; Burby and Paterson 1993; Gray and Scholz 1993; Helland
1998).
More surprising is the negative impact of sanctions on sense of duty to comply. Although the
imposition of sanctions is relatively rare (see Table 1), these results suggest that sanctions undermine the
sense of duty to comply. Indeed, regulated facilities subject to sanctions have 43 percent lower likelihood
of having a high sense of duty to comply and unregulated facilities subject to sanctions have a
corresponding 48 percent lower likelihood. This is perhaps because, as found by May and Winter (1999),
imposition of sanctions are viewed by some entities as overly harsh responses.15
Assistance. The findings concerning the effects of various forms of assistance upon motivations,
with the exception of technical assistance and deterrent fears, fail to show an effect on either set of
motivations. The failure to detect more substantial impacts of assistance upon either set of motivations is
40
-------
puzzling. Information and technical assistance are important ingredients of the voluntary approach for
which an effect on sense of duty to comply was expected. It may be that education efforts have been too
general and technical assistance too specific to have an impact on more basic feelings of civic obligations.
The negative effect of technical assistance upon deterrent fears can be explained by the observation that
such assistance, involving consultation by third-parties about how to comply, facilitates compliance and
thus eases fears of being deemed out of compliance. This facilitation is confirmed by the earlier analyses
of variation in adoption of best practices (see May 2003). The multivariate results showed that provision
of information and technical assistance had a substantial impact on the number of practices that were
adopted by regulated facilities. (The fact that information and assistance did not have an effect on the
number of practices adopted by unregulated facilities could reflect the generality of that information.)
Those analyses also showed that financial assistance had a positive impact on adoption of best practices
by both types of facilities.
Attitudes concerning government. The findings concerning the influence of different attitudes upon
motivations to comply are dominated by the influence of ideology. Those facility operators with more
conservative attitudes about government are more likely to have high deterrent fears and less likely to
have a strong sense of duty to comply. More conservative boatyard facility operators are on average are
13 percent more likely to have high deterrent fears and 16 percent less likely to have a strong sense of
duty to comply. The corresponding values for more conservative marina operators are 8 percent greater
likelihood of deterrent fears and 23 percent lower likelihood of sense of duty to comply. The increased
deterrent fears are likely related to the lower sense of trust in government associated with conservative
ideology (Scholz and Lubell 1998) while the negative effect on sense of duty to comply is likely related to
the lower sense of agreement of the value of regulations (Levi 1988; Tyler 1990).
Social influences. The presumption of social influence, addressed by Kagan and Skolnick (1993)
and by Winter and May (2001), is that entities take action in order to earn the approval of others. This
expectation is borne out by the findings concerning social influences and sense of duty to comply. In
particular, those boatyard operators who are concerned about reputation have a 32 percent greater
likelihood of having a strong sense of duty to comply while the corresponding value for marina operators
is 44 percent. This influence may be especially relevant in industries like marine facilities for which trade
associations are relatively strong and facility operators tend to know each other.
Conclusions
This research examines traditional regulatory and voluntary approaches for addressing potential
harms to water quality. The traditional approach consists of governmental enforcement of mandatory
requirements. The voluntary approach consists of government calling attention to potential harms and
facilitating actions to address them. These approaches are perhaps best thought of as ends of a
continuum, rather than as the sole choices in regulatory approach. In examining these, this research takes
advantage of a natural experiment in contrasting actions undertaken by boatyards subject to traditional
regulation with that of marinas subject to the voluntary approach. Hypotheses have been examined about
the role of these approaches and other factors in shaping two key motivations—deterrent fears and sense
of duty to comply—that are bases for actions to address potential harms.
Three sets of findings stand out from this research. One set concerns the impact of different
approaches on the actions that have been taken to address potential harms. The findings clearly
demonstrate that traditional regulation is more effective than is the voluntary approach alone. This is
expected and in itself is not particularly noteworthy. More striking is that the voluntary approach does
have substantial impacts for which the added gain in actions taken under mandatory regulation, when
controlling for other factors, is on average 17 percent. As such, voluntary approaches of the type
employed with marine facilities cannot be written off as wholly ineffective.
A second set of noteworthy findings concern the role of deterrent fears and sense of duty to comply as
motivations for action. Both sets of motivations are evident for the facilities in this study. As expected,
41
-------
deterrent fears are stronger for those facilities that are subject to mandatory regulation than for those
subject to voluntary approaches. The sense of duty to comply does not differ between the two types of
facilities. Taken together, the findings about motivations are consistent with the stronger impacts of
mandatory regulation upon actions to address potential harms.
A third set of findings concern factors that account for the variation in each set of motivations. The
multivariate findings that explain variation in deterrent fears are consistent with prior findings about the
importance of inspections in shaping specific deterrence (e.g. Burby and Paterson 1993; Gray and Scholz
1993; Helland 1998). But, deterrence is not limited to one-on-one enforcement. The findings that
showcase actions add to increase deterrent fears, along with the fact that three quarters of marine facility
operators report moderate to high awareness of large fines or other notable regulatory actions imposed on
other facilities, underscore the value of general deterrence brought about through showcase actions.
These put facilities on notice even when actions to address harms are not mandated.
Two other factors that differ from the traditional regulatory tools stand out as relevant considerations.
One is the role of reputation among peers as a positive influence on a sense of duty to comply. A second
factor is the role of attitudes toward government in shaping predispositions to take action. In particular,
those facility operators with more conservative attitudes about government are shown to be more likely to
have high deterrent fears and less likely to have a strong sense of duty to comply. These are important
findings because they suggest that the willingness (and reluctance) to address harms is related to more
basic predispositions.
Policy implications
Regulatory scholars have conducted fairly extensive research about regulatory instruments and, as
with this research, increasing amounts of research about compliance motivations. These findings suggest
a greater need for attention to regulatory arrangements and how those arrangements shape the relevance
of normative and social considerations for the attainment of regulatory goals. The importance of context
is reinforced by the various studies that I have conducted, often with others, regarding motivations. These
studies address compliance motivations for Danish farmers (May and Winter 1999, Winter and May
2001), for homebuilders in western Washington (May and Wood 2003), and for marine facilities with this
research. Concluding that the regulatory framework is important in explaining influences upon
compliance motivations is unsatisfying to those who seek general prescriptions about regulatory
enforcement and compliance. However, the findings across the various studies suggest that how
regulatory situations are framed affects the role of different regulatory tools and the relevance of
normative and social considerations.
The situation for boatyard operators in California and Washington mirror more typical regulatory
arrangements in the United States that do not ferment shared norms or expectations. Regulators and
regulated entities operate at arms-length with very limited interaction. As a consequence, there is little
leeway or occasion for negotiation over the terms of compliance. Instead, the rules constitute fairly
standardized conditions for permit adherence. Much of the enforcement rests on self-reporting of water-
quality analyses and only occasional on-site inspections. Enforcement matters since fines are issued and
sanctions for notable violations are publicized. However, this regulatory framework falls short of a social
contract in that the relationship is one-sided with regulatory authorities dictating the terms of the permit.
Given the lack of reciprocity for the relationship between regulators and boatyard operators, the
arrangement is more coercive than contractual.
Regulations serve a variety of functions in this context. Requiring permits is a powerful tool for
gaining attention to potential problems. Moreover, the conditions established with permits are important
in setting forth expectations as to what specifically constitutes the civic duty to comply. Yet, simply
requiring permits is not sufficient to motivate action. Technical and financial assistance is often
necessary for facilitating and directing actions to alleviate potential harms. These findings therefore
confirm what is well known about the importance of building commitment and capacity to take action.
42
-------
Deterrent fears are reinforced through publicizing of showcase actions to provide a general deterrent
effect, and through conducing inspections of facilities to provide a specific deterrent effect while also
educating regulated entities about compliance steps.
The situation for marina operators in California and Washington presents a challenge in fostering
actions to protect water quality when facilities such as boats within marinas are distributed sources of
water pollution, the costs of facility-specific inspection are relatively high, and there is much uncertainty
about the efficacy about different actions. The willingness to take action rests largely on the good will
and civic intent of marina facility operators and their tenants. Threats of stronger regulation or regulatory
actions loom in the background for which showcase actions are important in underscoring this backdrop.
But, the central issue is how to enhance the commitment to protect water quality.
The challenge is to create a stronger sense of civic duty in forming what might be considered a
societal contract for protecting water quality. Facility operators and boat owners, as with any source of
nonpoint water pollution, need to recognize that individual actions matter in the stewardship of the
environment. The findings of this study show that powerful drags on this civic commitment are distrust
of government (i.e., conservative ideology) and concerns that actions will have minimal effects on water
quality. Offsetting these somewhat is the belief among some facility operators that there is a serious
problem with water quality. Actions to address water quality are stronger when facility operators view
water quality harms as a shared problem that they have a civic duty to address.
Social influences are a key to enhancing civic commitment and shared obligations to protect water
quality. The sense of civic duty of marina facility operators is enhanced by their concern for reputation.
Facility operators are more willing to take action when they believe other facilities are doing their part.
Both of these sets of findings reflect the dynamics of a collective sense of obligation to take action.
Figuring out how to bring this about is of course a key challenge. While not directly addressed in this
research, other findings (May 2003) suggest that shared commitment is enhanced by associational ties
(i.e., among trade groups), the emergence of strong industry leaders, and credible evidence that action is
appropriate. The evidence from the research reported here suggests that public recognition of facilities
for their leadership, such as environmental stewardship awards, has a potentially useful role in this
equation.
The findings of this study more generally point to the duality of deterrent fears and civic intentions as
motivations to address potential environmental harms. Given the limitations of the voluntary approach,
deterrence must serve at least as a backstop to it. Yet, fostering deterrence need not entail an elaborate
enforcement regime since steps can be taken to instill a general deterrent effect. At the same time,
individual facilities' sense of duty to address harms can be enhanced by fostering a greater collective
sense of a need for action.
Acknowledgements
A version of this paper is to appear in a forthcoming issue of the Public Administration Review. Research
assistance with this project was provided by Robert S. Wood, Chris Koski, Justin Litvack, Stephanie
McNees, Justin Estell, and Andrew Goodrich. Comments on earlier drafts were provided by Raymond
Burby, Mark Lubell, Aseem Prakash, John Scholz, Soeren Winter, and participants in the June 2003
Berkeley Workshop on Corporate Environmental Performance and the Effectiveness of Government
Interventions. Funding for this research has been provided to the University of Washington by the U.S.
Environmental Protection Agency under grant R82882501. The findings are not necessarily endorsed by
the Environmental Protection Agency, the University of Washington, or the marine facilities and
associations who have cooperated with this study.
43
-------
Endnotes
1 Not considered here are voluntary codes developed by industry associations that serve as complements
to mandatory regulations by encouraging firms to go beyond required minimum compliance levels (see
Potoski and Prakash 2002).
2 The exceptions are those marinas that have repair or other facilities that are considered potential point
sources of pollution and thus are subject to permit requirements.
3 The weighting scheme established a stratification of boatyards and marinas in Washington, northern
California, and southern California. The data were weighted so that representation of boatyards and
marinas in each stratum reflected the appropriate population values for the stratum.
4 Each item was originally measured on a scale of 1 (low) to 5 (high) for which an index was computed as
the mean of the four items. Firms with strong deterrent fears are selected as those scoring 4.5 or greater
on the combined index.
5 Each item was originally measured on a scale of 1 (strong disagreement) to 5 (strong agreement) for
which an index was computed as the mean of the two items. Firms with strong sense of duty to comply
are selected as those scoring 4.5 or greater on the combined index.
6 Because not all practices are relevant to a particular facility, the index of best management practices for
each facility is based on the percentage of relevant best management practices for a given facility.
7 The Cronbach alpha reliability measure for this index is .63.
8 The Cronbach alpha reliability measure for this index is .67.
9 The four factors are lack of information about appropriate measures, uncertain effectiveness of water
quality measures, lack of standards or requirements, and difficulty in maintaining water quality measures.
The Cronbach alpha reliability coefficient for this index is .80.
10 In order to maximize the contrast between those facilities that are regulated and those that are not, the
data exclude 7 percent of boatyards without permits and 8 percent of marinas with permits. This excludes
boatyards that fall under the minimum thresholds for permits and marinas that have permits because they
also undertake repairs. As a consequence of these deletions, all regulated facilities are boatyards and all
non-regulated facilities are marinas.
11 The findings reported in May (2003) differ in minor ways from those reported here in Tables 1 and 2.
Those differences are explained by the fact that the present analyses exclude marinas with permits and
boatyards without permits, whereas the earlier publication included each of these. As discussed in note
10, the exclusion of these facilities provides for a stronger contrast between regulated and unregulated
facilities.
12 No statistically significant differences are found for the measures reported in Table 3 between those
unregulated marinas that received information and education and those that did not.
13 Separate OLS regression modeling that use measures for each dependent variable based on five-point
scales and the same explanatory variables as the logistic regressions were also undertaken. The findings
from the OLS models were very similar with those reported here, but because of the restricted range of
the dependent variables the logistic modeling is preferred. The only differences were for the deterrent
fear modeling for which under the OLS models sanctions had a statistically significant positive effect and
facility size had a statistically significant negative effect; neither was statistically significant in the logistic
model. A Chow F-test for pooled data undertaken as part of the OLS modeling shows that it is
appropriate to pool the data for the two types of facilities when modeling variation in each set of
motivations.
44
-------
14 This is the percentage point difference between the predicted likelihoods of high deterrent fears for
regulated and unregulated facilities. The predictions for these and other effect analyses reported in the
discussion that follows are based on the logistic models of Table 4. For each of the effect analyses a
comparison is made between predictions for changes in the variable of interest. These predictions are
based on mean values for other explanatory variables except for assistance and sanctions. The values for
those variables are set to correspond to a typical facility (i.e., values of one for general information, zero
for other forms of assistance, and zero for sanctions.) Where appropriate in what follows, effects are
separately reported for regulated boatyards and unregulated marinas.
15 Separate analyses show a stronger negative effect of sanctions upon sense of duty to comply for
unregulated marinas than for regulated boatyards. This difference is consistent with the backlash
explanation since marinas would be more likely to view sanctions as unwarranted than would boatyards,
given that marinas are not subject to mandatory regulations. An alternative explanation that could not be
examined with these data is that facilities with lower sense of duty to comply are more likely to commit
infractions and thus have sanctions imposed.
45
-------
Methodological Appendix
Table Al. Marine facilities "Best Management Practices"
Boatyards
Marinas
Use of vegetation or other natural buffers to limit
surface water runoff into marine waters.
Use of vegetation or other natural buffers to limit
surface water runoff into marine waters.
Use of catch basins, diversion or filtration systems to
control surface water runoff other than hull cleaning
areas.
Installation of catch basins, diversion or filtration
systems to control surface water runoff.
Designated hull cleaning area with concrete pad that
diverts waste water away from marine waters.
Marina fuel dock/station spill containment and recovery
equipment.
Catch basins/sediment trap for hull cleaning waste
water.
Marina fuel back-pressure/automatic shutoff on fuel
nozzles.
Recycling and filtration of hull cleaning waste water.
Posted prohibitions on discharge of used oil, antifreeze,
and paint solvents into dumpsters.
Use of tarps under boats when sanding hulls or other
surfaces.
Installation of receptacles for disposal or recycling of
waste oil.
Use of dustless sanders for sanding operations.
Installation of receptacles for disposal and treatment of
bilgewater.
Use of a sanding vacuum recovery system for recovery
of sanding dust and particles.
Installation of sewage pump-out facilities at the marina.
Dedicated, enclosed paint spraying area.
Regular maintenance of pump-out facilities by marina
staff or third parties.
Use of high pressure, low volume paint spraying
equipment.
Notification of marina tenants of third-party pumpout
and oil waste collection services.
Designated receptacles for disposal of oil, paint,
solvents, or used boat cleaning materials.
Established fish-cleaning areas with receptacles for
cleaning waste.
Third-party pickup and disposal of used oil, paint,
solvents, or used boat cleaning materials.
Restrictions on boat maintenance/cleaning to above the
waterline.
Posted prohibitions on discharge of oil, paint, solvents,
or used boat cleaning materials into general waste
dumpsters.
Designated areas for maintenance/cleaning of hull areas
with grids for collection of waste materials.
Use of environmental friendly cleaning materials.
Distribution of educational materials to marina tenants
about marine environmental practices.
Storm water management written plan.
Periodic marina-sponsored recycling or waste disposal
events for disposal of oil, solvents, batteries or other
waste.
A plan to put in place additional measures to address
potential environmental harms.
Use of environmental friendly cleaning materials.
Regular monitoring and testing of water quality at
boatyard.
Storm water management written plan.
Prohibition on do-it-yourself work at the boatyard.
An established plan to put in place additional measures
to address potential water quality impacts.
Requirements for written acknowledgment of
environmental regulations by third party vendors or
for do-it-yourself work.
Regular testing of water quality at the marina.
Use of vegetation or other natural buffers to limit
surface water runoff into marine waters.
Training or education for marina employees concerning
water quality issues and control measures.
Written prohibitions concerning waste disposal
(sewage, oil, bilge) into waters as part of marina
tenant agreements.
Source: Compiled fromUS EPA (1993, 1994) and Washington State Department of Ecology (1999a, b).
46
-------
References
Andrews, Richard N. L. 1999. Regulating the Environment, Regulating Ourselves. New Haven: Yale
University Press.
Andrews, Richard N. L., Nicole Darnall, Deborah Rigling Gallagher, Suellen Terrill Keiner, Eric
Feldman, Matthew L. Mitchell, Deborah Amaral, and Jessica D. Jacoby. 2001. Environmental
Management Systems: History, Theory, and Implementation Research. In Regulating from the Inside:
Can Environmental Management Systems Achieve Policy Goals, edited by Cary Coglianese and
Jennifer Nash, 31-60. Washington, DC: Resources for the Future.
Arora, Seema & Timothy N. Carson. 1996. Why Do Firms Volunteer to Exceed Environmental
Regulations? Understanding Participation in EPA's 33/50 Program. Land Economics 74(4): 413-432.
Bardach, Eugene and Kagan, Robert. 1982. Going by the Book: The Problem of Regulatory
Unreasonableness. Philadelphia: Temple University Press.
Becker, Gary. 1968. Crime and Punishment: An Economic Approach. Journal of Political Economy
76(2): 169-217.
Braithwaite, John and Toni Makkai. 1991. Testing an Expected Utility Model of Corporate Deterrence. Law
and Society Review 25(1): 7-3.
Brehm, John and James T. Hamilton. 1996. Noncompliance in Environmental Reporting: Are Violators
Ignorant, or Evasive, of the Law? American Journal of Political Science 40(2): 444-77.
Burby, Raymond J. and Robert G. Paterson. 1993. Improving Compliance with State Environmental
Regulations. Journal of Policy Analysis and Management 12(4): 753-772.
Coglianese, Cary and Jennifer Nash. 2001. Environmental Management Systems and the New Policy
Agenda. In Regulating from the Inside, Can Environmental Management Systems Achieve Policy
Goals? edited by Cary Coglianese and Jennifer Nash, 1-25. Washington DC: Resources for the Future.
Coombs, Fred S. 1980. The Bases of Noncompliance with a Policy. Policy Studies Journal 8(6): 885-900.
Ehrlich, Ivan. 1972. The Deterrent Effect of Criminal Law Enforcement. Journal of Legal Studies 1(4):
259-76.
Fiorino, Daniel. 1995. Making Environmental Policy. Berkeley, CA: University of California Press.
Fiorino, Daniel. 2001. Environmental Policy as Learning: A New View of an Old Landscape. Public
Administration Review 61(3): 322-334.
Grabosky, Peter N. 1995. Regulation by Reward: On the Use of Economic Incentives as Regulatory
Instruments. Law and Policy 17(3): 257-282.
Grasmick, Harold G. and R.J. Bursik. 1990. Conscience, Significant Others, and Rational Choice:
Extending the Deterrence Model. Law and Society Review 24(4): 837-861.
Grasmick, Harold G., R.J Bursik, and K.A. Kinsey. 1991. Shame and Embarrassment as Deterrents to
Noncompliance with the Law: The Case of an Antilittering Campaign. Environment and Behavior
23(2): 233-251.
Gray, Wayne B. and John Scholz. 1993. Does Regulatory Enforcement Work? A Panel Analysis of
OSHA Enforcement. Law and Society Review 27(1): 177-213.
Gunningham, Neil, and Peter Grabosky. 1998. Smart Regulation, Designing Environmental Policy. New
York: Oxford University Press.
Harrison, Kathryn. 1999. Talking with the Donkey: Cooperative Approaches to Environmental
Protection. Journal of Industrial Ecology 2(3): 51-72.
47
-------
Helland, Eric. 1998. The Enforcement of Pollution Control Laws: Inspections, Violations, and Self-
Reporting. The Review of Economics and Statistics 80(1): 141-153.
Kagan, Robert A. and John T. Scholz. 1984. The 'Criminology of the Corporation' and Regulatory
Enforcement Strategies. In Enforcing Regulation, edited by Keith Hawkins and John Thomas, 67-95.
Boston: Kluwer.
Kagan, Robert A. and Jerome H. Skolnick. 1993. Banning Smoking: Compliance Without Enforcement.
In Smoking Policy: Law, Politics, and Culture, edited by Robert L. Rabin and Stephen D. Sugarman,
69-94. New York: Oxford University Press.
Kerwin, Cornelius M. 1999. Rulemaking: How Government Agencies Write Law and Make Policy, 2d
edition. Washington DC: Congressional Quarterly Press.
Kettl, Donald F. 2002. Environmental Governance: A Report on the Next Generation of Environmental
Policy. Washington DC: Brookings.
King, Andrew A. and Michael J. Lenox. 2000. Industry Self-Regulation Without Sanctions: The
Chemical Industry's Responsible Care Programs. Academy of Management Journal 43(4): 698-716.
Levi, Margaret. 1988. Of Rule and Revenue. Berkeley: University of California Press.
Lyon, Thomas P. and John W. Maxwell. 2001. Voluntary Approaches to Environmental Regulation. In
Economic Institutions and Environmental Policy, edited by Maruizio Franzini and Antonio Nicita, 75-120.
Aldershot, UK: Ashgate Publishing.
Marcus, Alfred A., Donald A. Geffen, and Ken Sexton. 2002. Reinventing Environmental Regulation,
Lessons from Project XL. Washington DC: Resources for the Future.
May, Peter J. 2002. Social Regulation. In The Tools of Government, A Guide to the New Governance
Handbook, edited by Lester M. Salamon, 156-187. New York: Oxford University Press.
May, Peter J. 2003. Marine Facilities and Water Quality. Coastal Management Journal 31(4): 297-317.
May, Peter J. and Soeren Winter. 1999. Regulatory Enforcement and Compliance: Examining Danish
Agro-Environmental Policy. Journal of Policy Analysis and Management 18(4): 625-51.
May, Peter J. and Robert S. Wood. 2003. "At the Regulatory Frontlines: Inspectors' Enforcement Styles
and Regulatory Compliance," Journal of Public Administration Research and Theory 13 (2): 117-139.
McGraw, Kathleen and Scholz, John. 1991. Appeals to Civic Virtue versus Attention to Self-Interest:
Effects on Tax Compliance. Law and Society Review 25(3): 471-493.
Napier, Ted L. and Eric J. Johnson. 1998. Impacts of Voluntary Conservation Initiatives in the Darby
Creek Watershed of Ohio. Journal of Soil and Water Conservation 53(1): 78-84.
National Academy of Public Administration. 1995. Setting Priorities, Getting Results: A New Direction
for EPA. Washington, DC: NAPA.
National Research Council. 2000. Clean Coastal Waters, Understanding and Reducing the Effects of
Nutrient Pollution. Committee on the Causes and Management of Coastal Eutrophication, Ocean
Studies Board and Water Science and Technology Board. Washington D.C.: National Academy Press.
Potoski, Matthew and Aseem Prakash. 2002. Protecting the Environment: Voluntary Regulations in
Environmental Governance. Policy Currents 11(4): 9-14.
Prakash, Asseem. 2000. Greening the Firm. Cambridge: Cambridge University Press.
Reiss, Albert Jr. 1984. Selecting Strategies of Social Control Over Organizational Life. In Enforcing
Regulation, edited by Keith Hawkins and John Thomas, 23-35. Boston: Kluwer-Nijhoff Publishing.
48
-------
Scholz, John T. and Neil Pinney. 1995. Duty, Fear, and Tax Compliance: The Heuristic Basis of
Citizenship Behavior. American Journal of Political Science 39(2): 490-512.
Scholz, John T. and Mark Lubell. 1998. Trust and Taxpaying: Testing the Heuristic Approach to
Collective Action. American Journal of Political Science 42(2): 398-417.
Segerson, Kathleen and Thomas J. Miceli. 1998. Voluntary Environmental Agreements: Good or Bad
news for Environmental Protection? Journal of Environmental Economics and Management 36(2): 109-
130.
Sparrow, Malcolm K. 2000. The Regulatory Craft: Controlling Risks, Solving Problems, and Managing
Compliance. Washington DC: Brookings.
Tyler, Tom R. 1990. Why People Obey the Law. New Haven and London: Yale University Press.
U.S. Environmental Protection Agency, Office of Water. 1993. Guidance Specifying Management
Measures for Sources ofNonpoint Pollution in Coastal Waters. EPA-840-B-93-001c. Washington,
D C.: EPA.
U.S. Environmental Protection Agency, Office of Water. 1994. Protecting Coastal Waters from Vessel
and Marina Discharges: A Guide for State and Local Officials. EPA-842-B-94-004. Washington, D.C.:
EPA.
U.S. Environmental Protection Agency, Office of Water. 1997. State Source Water Assessment and
Protection Programs Guidance. EPA-816-R-97-007. Washington, D.C.: EPA.
U.S. Environmental Protection Agency, Office of Water. 2001. Management Measures Guidance to
Control Nonpoint Source Pollution from Marinas and Recreational Boating. EPA 841 -B-01 -005.
Washington, D.C.: EPA.
Washington State Department of Ecology. 1999a. Ship Shape, Single Industry Campaign Summary
Report. Publication 99-16. Olympia: Department of Ecology.
Washington State Department of Ecology. 1999b. Water Quality Enforcement Review. Publication 99-
18. Olympia: Department of Ecology.
Welch, Eric W., Allan Mazur, and Stuart Bretschneider. 2000. Voluntary Behavior by Electric Utilities:
Levels of Adoption and Contribution of the Climate Challenge Program to the Reduction of Carbon
Dioxide. Journal of Policy Analysis and Management 19(3): 407-425.
Wilms, Wellford W. 1982. Soft Policies for Hard Problems: Implementing Energy Conserving Building
Regulations in California. Public Administration Review 42(6): 553-561.
Winter, Soeren C. and Peter J. May. 2001. Motivation for Compliance with Environmental Regulations.
Journal ofPolicy Analysis and Management 20(4): 675-698.
Winter, Soeren C. and Peter J. May. 2002. Information, Interests and Environmental Regulation. Journal
of Comparative Policy Analysis 2(4): 115-142.
49
-------
Session II: Compliance Issues
Discussant No. 1: Jon Silberman, U.S. EPA, OECA/NCEI
COMMENTS ON:
Corporate Self-Policing and the Environment:
Factors Predicting Self-Disclosure of Clean Air Act
Violations Under the EPA's Audit Policy
Paul Stretesky
Colorado State University
April 26, 2004
Introduction:
Introduce self:
- 20 years experience with EPA, 17 of them in enforcement, including experience amending
and implementing EPA's Audit Policy.
Thank you's:
Audit Policy article: - Bert Frey, Deputy Regional Counsel, R3 - 312-886-1308
- Phil Milton, Audit Policy Coordinator, ORE-SLAP, 564-5029
Marine Facilities article: - Cheryl Hawkins, OPEI - Office of Business & Comm. Innov.
- Larry Wells, Rl-Off Env. Steward. - Assistance & P2
Opening statement: "What data does - and all it does - is tell you where you've been. It's
experience, insight, & intuition that tells you what's going to happen next."
My perspective in reviewing and commenting on the two papers is:
- this is all about enhancing environmental performance by improving how and when the
government intervenes to address pollution;
- all comments are intended constructively/"for your consideration" as you revise papers.
Key ???: Is the study protocol fair and appropriate?
- does the researcher ask the "right questions?"
- does the data appear accurate/reliable/complete?
- do we agree with the factual conclusions drawn from the data?
- do we agree with the policy recommendations derived from the factual
conclusions?
- how can the study (which is a draft, interim product) be improved?
50
-------
Emphasis on practical application of the results:
- How might EPA use the study results to actually improve our policies or programs?
- Do the data/conclusions augment, reinforce, contradict, or alter existing assumptions/
approaches?
- Should EPA do: more of the same?; less of the same; different things? or simply get (or
stay) out of the way?
WE'VE ALL HEARD THE COMMON APHORISM:
"You can lead a horse to water but can't make him drink."
Unfortunately, for some of us - including those of us who work with the Audit Policy - our
job description is to figure out how to get those horses to drink!
Assume, for the moment, leader = EPA; horses = regulated community; water = AP;,
drinking = using it, & my objective is encourage drinking. What are my options?
-1- make sure it's really "water" and it's potable. Maybe the horse isn't drinking
because "he's asking for water and you're giving him gasoline" (like a B.B. King song").
4 AP examples of "sweetening the water:
3 May 2000 AP amendments to: (1) increase disclosure period from 10 days to 21
days; (2) clarify "repeat violations" condition for company-wide disclosures; (3)
address mergers and acquisitions;
(4) end of FY '99: ORE issues guidance: "Reduced Penalties for Disclosures of
Certain CAA Violations" (addresses Title V-related disclosures to allow more of
them to qualify)."
Also in May 2000: EPA amended Small Business Compliance Policy to make more
attractive.
-2- revisit your "leadership":
- leadership * simply pouring water into a bucket and leaving it there, hoping horses will
somehow "stumble onto it";
51
-------
- ensure the water is visible and immediately apparent as such to horses (now we're talking
"outreach and communication" ... especially important for small businesses).
-3- make sure you've got the right "horse"
- Essentially a targeting issue; can also be a resource/capacity-building or timing issue;
- Maybe you should drop the horse you're struggling with now, deal with him some other
time or way/try to find others that are likely to be more cooperative.
-4- make sure your horses are "thirsty"
- Also be a targeting issue ... but equally often, it's a "conditions" issue;
- consideration 1: timing, "sequencing" issues for "integrated strategies": wait until the
timing is better from the horse's perspective;
- consideration 2: thirstiness: find a way to intervene somehow to make the horse
"thirstier!"
- This is where considerations of the issue of type and scope of government interventions
to change existing conditions in order to impact behavior and performance - the focus of this
conference - comes to the forefront.
Overview of EPA AP "lessons learned":
Summary of everything EPA has learned over the past 8 years about AP use in one
sentence:
~ "The AP works best, and is used much more often, when we employ it
in the context of an integrated compliance strategy that includes targeting
a sector with known compliance issues and then employing interventions
to give firms - usually within the same sector - the choice of "voluntary
auditing and disclosure" -or- "inspections and enforcement."
Corollary: the AP does not/hasn't produced significant results when we've simply "thrown
it out there" and waited for firms to knock at our doors.
52
-------
- This has pretty much held true no matter what the firm-specific or overall Regional
inspection and enforcement numbers have been.
Other long-held Agency views of AP use based on our 8 years of experience:
- Large firms with resources -and- auditing & EMS capacity and experience are more
likely to use the AP specifically for those reasons.
- Small businesses are unlikely (= "virtually never happens") to use the Small Business
Compliance Policy absent EPA or State interventions to both -1- build capacity, and -2-
build trust, e.g., by coupling non-inspection site visits with technical assistance ("non-
enforcement options).
• Key resource-related lesson learned:
"The best scenarios for increased AP use require the fewest possible numbers
of inspections and enforcement actions for the involved sectors, companies, or
firms."
"Ideally, our strategies should produce significant self-disclosures with just
one or even no associated inspections or enforcement because site visits and
lawsuits are resource-intensive to all concerned parties.
"That's why so many of our Compliance Incentive Programs (CIPs) rely solely
on invitation letters with just a small number of (or 1) high-profile enf.
actions. This = "leveraging" our resources and impacts."
Illustrative examples of integrated strategies including AP use:
"Carrots appear tastier when backed with a stick" - EPA Region l's ongoing compliance
work with its colleges and universities:
~ 1999: R1 identified negative compliance trends at these schools;
~ Regional compliance assistance staffs initial step was to schedule a compliance
assistance workshop for colleges and universities.
53
-------
~ WhenRl first advertised the workshop, out of3 50+ colleges and universities in New
England, fewer than 40 individuals responded, many from the same college.
~ The Region then issued an administrative penalty action against the University of
New Hampshire, at the same time contacting the president of each school by letter
to highlight the action and encourage attendance at the seminar.
~ Response to the post-enforcement action letter was overwhelming; over 330 requests
were received to attend the compliance assistance workshop, which attracted a
capacity audience; a second workshop had to be scheduled.
~ As part of that process, the AP was explained and info, provided on how to work
with EPA to develop auditing agreements -4 self-auditing and disclosures.
~ Today, nearly half of New England's colleges and universities have taken part in the
New England Colleges and Universities Audit Initiative.
"Walk before you run" AKA "build capacity" - R2, Patterson, NJ Refrigeration
Repair/Geographic Initiative:
~ Best way to promote small business use of the AP is to combine it with on-site,
compliance assistance visits ("mock inspections" that aren't inspections) together
with a credible fear of enforcement for nonparticipants.
~ Standard practice of sending "notice letter of upcoming inspections if you don't self-
disclose" didn't work with these small businesses - Region got 1 taker who turned
out to be in compliance.
~ In response, Region developed on-site compliance assistance visit program -4
auditing and disclosures.
~ #1 lesson: These small businesses wouldn't/couldn't self-audit (low knowledge and
resource levels) even after receiving the Region's compliance incentive letters. But
site visits/technical assistance obtained their involvement.
"Clean up now ... before we catch you" - National Iron & Steel Minimill Compliance
Incentive Program
54
-------
~ In August 2000, EPA sent letters to 41 minimills inviting them to participate in a
voluntary multimedia audit program where violations disclosed within six months
could be self-disclosed under the Audit Policy.
~ After the letters were sent, EPA met with the Steel Manufacturers Association and
member companies to discuss the initiative and answer questions.
~ Of the 41 minimills receiving letters, 28 minimills responded; EPA also received
letters from 10 minimills that did not even receive invitation letters but wanted to
participate.
~ 24 companies ended up disclosing violations at 38 minimills. Responses included
cleaning up spilled hazardous electric arc furnace dust, repaired cracked secondary
containment around storage tanks, etc.
"Your colleague's in violation so you probably are, too!" - Telecommunications Industry
Compliance Incentive Program
~ In January 1998, GTE voluntarily disclosed CWA and EPCRA violations to EPA.
EPA and the co. developed a company-wide settlement, including auditing, resolving
violations at 314 facilities.
~ Afterwards, EPA sent letters to the rest of the telecom industry explaining our
knowledge of their likely compliance issues and offering the opportunity to self-audit
and disclose.
~ As a result, including the GTE settlement, more than 3,500 telecommunications
facilities came into compliance with four environmental laws (CAA, CWA, RCRA,
and EPCRA) through 25 civil settlements, 24 of them attributable to the use of EPA's
Audit Policy.
Research/nomenclature tips to researcher to improve paper:
Understand what EPA and the States are actually doing with the AP today, and why:
~ Empirical studies that confirm preexisting assumptions, experiences, and programs
are OK... but the ones that raise the discussion to "the next level" of potential policy
and strategic improvements are better.
55
-------
~ Know what has worked ... and what hasn't:
E.g., 1 big, well-publicized case can have an order-of-magnitude larger deterrence
impact than 20 "pin-prick" cases when it comes to Audit Policy use (we develop our
integrated strategies and initiatives with that in mind).
~ CAA violations, especially on the part of the larger companies, often involve huge
delayed or avoided EBN and/or result from differences in regulatory interpretation,
not ignorance (i.e., AP unlikely to be a "solution").
~ Paper is devoid of recognition of historical AP "lessons learned" and uses - that's
why I used most of my presentation today talking about what EPA already knows
and how we're acting on it/integrated strategies.
Be careful to employ terms of art accurately . .. 2 examples:
~ # of enf. actions alone, or inspections alone, is a component of deterrence but is not
= "deterrence."
- Also, can't tell from the data sets whether enf. actions under other statutes are also
influential - they might be key factors in prompting AP use;
- Whenever possible, be holistic/synergistic/integrated & systems-oriented, rather
than reductionist - "look at the big picture."
~ "Whether the co. received information concerning a compliance incentive, e.g.,
a letter providing AP-related incentives, including info, that EPA may target insp's."
* "compliance assistance" ... that's specific deterrence!
~ Make sure you get your dependent variables correct, e.g.:
- Pg. 15: "The findings of this research add to the corporate behavior literature and
are surprisingly consistent with other studies that find that specific deterrence does
not appear to be related to levels of corporate compliance ." ... NO!!! The
dependent variable here is NOT "compliance" - it's Audit Policy use.
56
-------
- Also:
* Can produce compliance with or without AP use.
* Or can promote increased self-auditing and correction of violations
without self-disclosure ... also leading to increased compliance.
~ Summary/in general: say what the data actually shows (e.g., CIP program letters
work well in spurring AP use) in lieu of applying broad labels ("deterrence";
"compliance") that aren't strictly accurate.
Tell us why you chose just the CAA, & FY '99-'00 (e.g., manageable data set)?
- How might this affect the transferability of your results across sectors, programs,
and years?
- Would a breakdown of the violations by type (e.g., emissions limit violations
versus failure to install required capital equipment versus reporting/ recordkeeping)
be helpful in this regard?
Re-check your numbers:
~ ORE data pull for this paper ID'd nearly 100 CAA self-disclosures for same time
researcher found 59 (taking CAA Section 211 mis-fueling violations/HQ handled
these) into account;
~ Did you ask that your data run also include the Small Business Compliance Policy?
~ What about impact of acquisitions & mergers (not discussed at all in paper)?
~ Preexisting strong positive relationship between facility size, inspection frequency,
& likelihood of enf. ... how does this impact your statistical analysis?
Consider the possibility of behavioral and/or "AP-content-related" explanations for
results contradicting your hypotheses:
57
-------
~ Behavioral example: what researcher calls "specific deterrence" - finding that ##s
of co.-specific enforcement actions or inspections are not significantly associated
with increased audit policy use ... considerations I'd look at include:
- causality issue of whether what really happening is specific deterrence not
working to promote AP use, or that actors who attract inspections and enforcement
are simply least open to using compliance incentives?
- e.g., are the same characteristics that lead bad actors to be more likely to be sued
the ones leading them to be unlikely to use incentive programs proactively?
Such companies might be, up-front, more recalcitrant and less civic duty-
oriented and thus less likely to voluntarily audit, correct, and disclose
violations.
~ "Audit Policy content" is important too, as the Policy has been
specifically configured to promote disclosures by persons who have not
yet been inspected or enforced against. Examples:
- persons with "repeat violations" get bumped by AP Cond. 7;
-Persons inspected or notified of insp's get "bumped" by AP Cond. 4 (independent
disclosure);
- persons sued may get settlements which include self-auditing as injunctive relief;
making it harder to qualify under AP in future;
Perhaps, when you take these factors into account, the better hypothesis
is to expect the odds of self-disclosure to be lower for someone who has, in
the past, been regularly inspected and enforced against!
And finally ... general suggestion:
~ Tease behavioral observations and policy recommendations from the research (don't
"just present the data ...").
Concluding comments:
58
-------
You can "lead a horse to water" when you take the time to understand the horse, ensure it's
thirsty, and offer it attractive water."
Tell us, in your final report, based on the data & analysis, what are your recommendations
or considerations for how best to achieve these goals?
59
-------
Session II: Compliance Issues
Discussant No. 1: Jon Silberman, U.S. EPA, OECA/NCEI
COMMENTS ON:
Regulation and Compliance Motivations:
Marine Facilities and Water Quality
Peter May
University of Washington
April 26, 2004
Introduction:
Before I say anything else about this paper, I must ask Prof. May about a specific finding in
his paper:
~ In the section on "ATTITUDES TOWARDS GOVERNMENT," you write:
"Those facility operators with more conservative attitudes about
government are more likely to have high deterrent fears and less likely
to have a strong sense of duty to comply. These are important findings
because they suggest that the willingness (and reluctance) to address
harms is related to more basic predispositions."
~ My question is: could you please be specific, in your final report, as to what new
policies and programs you recommend the Bush Administration sponsor to address
this finding? :)
Prof. May has put his finger squarely on two key issues facing us all today:
~ (1) "How to enhance the commitment to protect water quality by creating a stronger
sense of civic duty - [what Prof. May calls] a 'societal contract for protecting water
quality.' "
~ (2) Not whether voluntary programs are more effective than traditional regulation
[per se], but [rather] how can voluntary programs be improved?" [and better
complement regulatory programs?]
This study is extremely timely for the C&E program - I'd like to briefly mention 2 reasons:
(1) Current strong EPA focus on environmental performance through voluntary programs:
60
-------
- Voluntary programs are designed to motivate people and organizations to take actions not
required by regulation by going beyond compliance or achieving regulatory objectives in
more effective and efficient ways.
- In general, voluntary programs use partnerships, market forces, and incentives, rather than
mandatory regulations, to achieve environmental results.
- EPA is currently home to more than seventy voluntary programs dealing with a diverse set
of issues, from climate change and waste reduction to innovative technologies.
- The Deputy Administrator has asked EPA's Innovation Action Council to determine how
to enhance the effectiveness of our voluntary programs to make them more customer-focused
and results-oriented ... this research can help.
(2) OC and others are interested in making better use of so-called "social marketing" to
enhance peoples' sense of civic duty:
- Social marketing is based on the premise that, "Things don't just happen - people have to
want to make them happen."
- In his slides for last week's presentation to the National Environmental Assistance Summit
on "performance-based compliance assistance," Office of Compliance Director Mike Stahl
described it as a " blend of two ideas: performance-based management and social
marketing."
- "Social marketing" means employing commercial marketing techniques in outreach and
assistance, not just to inform, but to actually change behavior.
(1) define problem in behavioral terms;
(2) engage stakeholders in market research;
(3) identify perceived obstacles, benefits of behavior;
(4) segment & target audience based on their specific characteristics;
(5) tailor messages to audience segments;
(6) feedback loop - monitor & adjust the message.
Tools include: -1- education to increase awareness;
-2- social marketing to increase openness to change & incentives;
-3- enf. & deterrence to address resistant to change & compliance.
Challenges to performance-based marketing include:
- compliance assistance initiative require more upfront planning & analysis;
- tailored assistance makes generalizing from results ("transferability")
more difficult.
61
-------
But the results of studies like this one, and Prof. Kagan's "Tracking Deterrent
Messages in Environmental Enforcement" study, can really help.
General reaction to May report by myself and my Regional and HO contacts:
Overall, it's a through examination with logical findings:
- Our experience agrees with Prof. May's findings in many respects (I'll cite just a few
examples):
traditional regulation is critical to continued progress in protecting and improving water
quality (and potentially more effective than the voluntary approach alone):
- Myth: EPA has already harvested the "low hanging" environmental protection fruit.
- Fact: Approx. 25% of NPDES permittees still experience significant noncompliance with
their permit limits each year.
... but, voluntary approaches can be quite effective, too;
Both deterrent fears and sense of duty to comply are important motivators;
Actions to address water quality are stronger when facility operators view water quality
harms as a shared problem that they have a civic duty to address;
Civic duty can have a stronger impact on regulated than unregulated facilities, perhaps
because regs and permits signal what is desired by society.
r:?' Mandatory-approach inspections may not have the off-putting effect on the sense of duty
to comply as some hypothesize.
vs- But imposing sanctions can be off-putting when the recipients view them as overly harsh.
r:?' Regulators should consider, when crafting regulatory goals and strategies, how
performance is affected by variable such as inspections, reputation, and attitudes toward
government.
Building commitment and capacity to take action is important to increase compliance and
performance, particularly by small businesses.
Summary of EPA reactions to draft report:
Everyone with whom I discussed this report's conclusion agreed essentially with its findings.
62
-------
The report adds to our knowledge of the impacts of regulatory and voluntary programs,
though more along the lines of buttressing what we already know (as opposed to providing
startling new insights).
It would be helpful to us, Prof. May, if you could beef up discussion of potential policy
changes, or new assistance, enforcement, and/or integrated strategies you believe your
findings suggest for improving marine facilities' awareness, compliance, and stewardship.
Do this through added holistic analysis of your factual findings, your general understanding
of the relevant compliance & motivational literature, and most importantly, your personal
in-depth interactions with the marine facilities sector, "on the ground," in the course of
preparing this report.
You may also want to look at what other Regions and States are doing in the area of
improving the marine facilities' environmental performance, e.g., Regions 1, 2, 3, and 4,
Connecticut, Florida, and Maine.
We especially would benefit from analysis and development of these types of finding:
(Pg. 17) Social influences to enhance civic commitment/shared obligations to protect water
quality include:
"Social influences are a key to enhancing civic commitment and shared obligations to protect
water quality and can be enhanced by, e.g.:
- strengthening associational ties (i.e., among trade groups);
- facilitating the emergence of strong industry leaders;
- providing stakeholders with credible evidence that action is appropriate."
- increasing facilities' concerns for their reputations;
- inspiring the dynamics of a collective sense of action;
- leveraging associational ties, e.g., among trade groups;
- supporting the emergence of strong industry leaders;
- sharing credible evidence/information that action is appropriate;
- providing public recognition of leadership facilities, e.g., env. stewardship awards.
~ Don't stop with general recommendations. Use your knowledge of this sector and
the compliance/behavioral literature generally to take us to the next level of
understanding as to how, specifically, EPA, States, and other stakeholders can
leverage these types of social influences to improve compliance and performance in
this and other sectors.
~ There are many folks here at the Agency who will consider carefully and use what
you produce.
63
-------
- SELECTED COMMENTS ON RESEARCH PROTOCOL AND TRANSFERABILITY OF
RESULTS-
Dearth of environmental indicator and/or actual performance information may affect reliability of
results:
Example 1: commitment of marine facilities to address water quality problems is measured
in 3 ways:
(1) % of BMPs undertaken;
(2) self-ID of level of priority to facility of env. protective actions;
(3) self-rating of their efforts from "less attention than most" to "we are a leader."
Example 2: Two measures serve as proxies for the extent of water quality problems:
(4) respondent perceptions, on a ten-point scale, of water quality at their location five years
ago;
(5) amount of waterfront that is adjacent to the facility.
Typical reactions to these measures from EPA personnel:
(1) "All BMPs are not created equal" - e.g., distributing env. literature versus developing
and implementation a formal Stormwater Management Plan;
(2) No actual water quality measures are developed/considered at all! Perceptions can be
wrong/some QA/QC of the opinion/perception-based survey responses (= "blunt tools")
would be helpful.
Potential Limits on the Transferability of Results:
Inherent problem exist in analyzing voluntary programs, generally:
~ Because voluntary programs are by nature so diverse, it can be difficult to
generalize/transfer results from one type of program and sector to another.
~ It would be helpful to add some discussion of this issue - with observations and
recommendations - to the paper.
Geographic factors: There may also be significant issues associated with generalizing from
Washington State marine facilities to those in other areas, e.g., New England.
~ Example: Distinction between "regulated" and "unregulated" may be less black
& white than the paper suggests :
- Rl: Boatyards differ from marinas in that their primary function is boat
maintenance, repair, and storage. Marinas, however, also conduct a significant
amount of repair, maintenance and equipment cleaning.
64
-------
- Consequently, marinas are, in fact, subject to many regulations (i.e., they're not
"unregulated").
- E.g., Region 1 has determined, through site assessment visits, that approximately
70% of the marinas visited were required by law to have storm water permits.
- Many of the activities described under the BMP table (pg. 20) appear
interchangeable between marinas and boatyards.
The paper may present a more "black and white" approach to regulatory versus voluntary programs
than is justified:
• Comment of an EPA reviewer with significant voluntary programs experience: "While the
paper asserts that mandatory and voluntary approaches are not strictly either/or
propositions, they are then pitted against each other through hypotheses such
as, 'the mandatory approach is more effective than the voluntary approach in
gaining commitment to address potential harms by those entities that potentially
contribute to these harms.' "
~ "EPA views voluntary programs less as a replacement for regulations, more as an
adjunct means to fill regulatory voids where there are no rules or where we can ask
people to voluntarily go above and beyond the regulatory requirements to be good
corporate citizens in the communities in which they reside."
Expanded discussion of "role of stakeholders":
~ Region 1 comment: "Our experiences have shown that local governments and local
organizations (including some local organizations with authority in certain areas), NGOs,
and customers play a role in affecting compliance. These stakeholder impacts do not appear
to be addressed in the report. In addition, the report does not distinguish among federal,
state, and local authorities regarding regulations."
Recommendation: expand discussion of research on how to enhance civic commitment
and shared obligations to protect water quality.
~ Comment of an EPA reviewer with significant voluntary programs experience:
"Researcher makes brief mention, only, of the value of strengthening organizational
ties (i.e. among trade groups), fostering the emergence of strong industry leaders, and
providing credible evidence that action is appropriate. More extensive discussion
of these topics would have been useful for us as policy-makers and regulators."
Role of technical and financial assistance:
65
-------
Region 1 comment: "The conclusion that technical and financial assistance is often
very important to facilitate and direct actions is a significant one to keep in
mind, especially for the NPDES individual permit requirement for pressure
wash discharging."
Question for researcher: Are there other specific areas you want to recommend as a focus
for government efforts to target technical and financial assistance to this sector?
CONCLUDING THOUGHTS:
• EPA STRONG LY AGREES: Pg. 17: "Given the limitations of the voluntary approach,
deterrence must serve at least as a backstop to it. Yet, fostering deterrence need not entail
an elaborate enforcement regime since steps can be taken to instil a general deterrent effect."
This is good news for resource-strapped regulators who are open to thinking creatively and
innovatively.
• "Everyone talks about wanting to be able to tell the federal government where to go and
what to do. Here's your chance to do so and you're being paid to do it. So be specific in
your recommendations!
66
-------
Session II: Compliance Issues
Discussant No. 2: John Horowitz, University of Maryland
COMMENTS ON:
Corporate Self-Policing and the Environment:
Factors Predicting Self-Disclosure of Clean Air Act
Violations under the EPA's Audit Policy
Paul Stretesky
Colorado State University
and
Regulation and Compliance Motivations:
Marine Facilities and Water Quality
Peter May
University of Maryland
April 26, 2004
By John K. Horowitz, University of Maryland, College Park, horowitz@arec.umd. edu
Stretesky's paper examines the characteristics of firms that disclosed a Clean Air Act
(CAA) violation during a period in which EPA had a policy that appeared to reward self-
disclosure of violations. To conduct this research, he must examine both self-disclosed
violations and violations that were uncovered by EPA but not disclosed. My comments
are most technical (although they go to the heart of the inferences that can be made from
these data.). In contrast, Jon Silberman raised issues about the nature of CAA violations
and of EPA's policy.
To study the effect of the EPA's policy, Stretesky uses a statistical technique
labeled case control design. This approach entails taking a random sample of non-
disclosers (that is, violations that were uncovered by the EPA) and comparing this to the
full sample of disclosers. The population ratio of non-disclosers is about 95 percent; only
5 percent of violations were self-disclosed. Under the case control design method, where
only a sample of non-disclosers is analyzed, the data set's ratio of non-disclosers is 50
percent. It is easy to see why he might like to do this. When the true probability of a
dichotomous (yes-or-no) variable is 50-50, the power of his statistical tests is maximized.
When the true probability is much more skewed, as it is with these data, the statistical
power is much lower. Of course, these claims apply to data that is naturally distributed
50-50, not to data that is simply selected to be distributed so.
I have serious reservations about this method. At a minimum, Stretesky should
explain this statistical technique further. I think it likely that the method is flawed,
probably because the standard errors of the estimates are biased downward. My intuition
is simple: Under this technique, the author has thrown out information. This will raise
67
-------
the standard errors. Since he does not correct for this problem in calculating his standard
errors, the standard errors reported in the paper are surely too small.
There is also, of course, a severe sample selection problem: Even if all of the
non-disclosed violations are analyzed, they will represent only those non-disclosed
violations that were detected. If the characteristics that lead to a violation being detected
are correlated with the characteristics that determine whether a violator will disclose or
not, then his coefficient estimates are again biased. This correlation surely exists. There
is probably little the author can do about it, although he knows the data and
circumstances better than I and may have some ideas. Even if he cannot solve the sample
selection problem, he needs to discuss it and its ramifications.
My comments about May's paper revolve around the nature of the question he is
addressing. This paper looks at difference in the environmental behavior and attitudes of
marinas and boatyards. For short-hand, he labels marina behavior "voluntary," because
most of them are not required to have NPDES permits, and boatyard behavior
"mandatory." May recognizes that these labels are mere short-hand for describing a
complex situation. I would argue that they have greatly obscured the issue he is
addressing, and may have led to errors in the analysis.
First, since the main difference underlying his analysis is whether a facility has an
NPDES permit, it seems odd for the author to have thrown out data points in which the
boatyards did not have NPDES permits or the marinas did have such permits.
May recognizes that marinas are not completely unregulated, and argues that the
true characterization is a continuum in which marinas are "less regulated" and boatyards
are "more regulated." But is even this characterization valid? Several pieces of evidence
suggest that at least some marinas face substantial regulation; see Table 3 on p. 12 or p.
16. This evidence suggests that a continuum may not be the right metaphor. Marinas
may simply face different regulations. The interpretation of the results is thereby called
into question.
I suspect May is correct in claiming that marinas bear a less onerous regulatory
burden than boatyards. I suggest that he document this more fully, rather than relying on
the NPDES distinction.
Finally, I want to note the problems of treating the sentiment of "duty to comply"
as a socially desirable aspect of regulation. This seems to send us into a house of mirrors
in which people comply more when they don't have to and comply less when they have
to. Perceived "fairness" - of the law, of the enforcement of the law, and of the penalties
- may be a better measure of what May is trying to understand.
68
-------
Summary of the Q&A Discussion Following Session II
Joanne Berman (U.S. EPA, Office of Enforcement and Compliance Assurance)
Ms. Berman commended both the presenters for including compliance assistance in the
presentations, but said she "was a little befuddled" by how they were using the term. She
cited the "official" description of "compliance assistance is to help the regulated
community understand and comply with the laws." Saying that she appreciated their
going on the University of Pennsylvania website, she apologized for the dearth of
material saying, "that's in part because for federal folks compliance assistance is
something new and we're trapped in the measurement of it—we're in the process of
trying to find ways to track and then measure it. This may not help for your 1990-2000
data, but from my perspective compliance assistance is not the audit policy itself but what
we did ancillary to the providing the audit policy. For example, if you were using the
audit policy and re-developed audit protocols, then what we do a lot of times in these
initiatives you spoke of is—the development of the protocol and the distribution of it is
where we would find out whether compliance assistance had a value in using the audit
policy, and I wasn't sure you knew that—I see you shaking your head that you did, but it
wasn't clear to me from your presentation."
In addressing Peter May's presentation, Ms. Berman noted that sometimes "folks kind of
clump compliance assistance with voluntary programs," and she said in her work "that's
not the case because the definition says that there has to be a connection between
compliance assistance and the work that we do." She went on to add that "in the
initiatives you spoke about. . . when we do compliance assistance, we've learned that it
needs to be an integrated approach, and I could give you a laundry list of how we do it
with inspections, how we do it with enforcement, how we do it with the audit policy."
She concluded by encouraging the researchers to "think of compliance assistance in
conjunction with all these other tools." She also said that as we move toward the concept
of social marketing, EPA would really benefit from any data concerning whether
compliance assistance actually has an impact on changing behavior.
(Note: Neither of the presenters responded to Ms. Berman's comments.)
Pete Andrews (UNC-Chapel Hill, Department of Public Policy)
Addressing Peter May, Dr. Andrews questioned how valid it is to generalize based on
data from "two segments of the marine industry," which he sees as "a somewhat unique
industry." Assuming that the focus was on smaller boat yards as opposed to big
shipyards, Dr. Andrews characterized the clientele as "recreational boaters—fairly
affluent Americans who care about the environment as well." He added that, "as a
monopoly, [smaller boat yards] can just keep raising their price because nobody can take
the business away from them—at least I don't see a lot of opportunities for that on the
East Coast." He wondered how the research would play out in looking at dairy farmers,
or school systems, or "other kinds of players who are sort of below the radar in important
segments in terms of environmental performance."
69
-------
Peter May (University of Washington, Center for American Politics and Public Policy)
Dr. May responded by first asserting that he believes the marine facilities they surveyed
have a more "heterogeneous clientele" than Dr. Andrews indicated. Addressing "the
broader question of where does this fit in?" Dr. May continued by asserting that the way
this type of research is conducted today "it's hard to really come up with clear patterns
across all these different settings," and he said he thinks "the way we're thinking about
these things is wrong. We're tending to think about it too much on a sector basis, or too
much on a size basis. There has to be some other way of thinking about these kinds of
things. . . We're thinking about it as too much of the attributes of the industry and not
enough about the attributes of the regulatory context broadly defined. We need to think
about that regulatory context, and this is why I like the work that Bob and Neil [Kagan
and Gunningham] . . .and others are doing ... is that it brings in that context more, in
terms of how reputation plays in, and a variety of these more elusive kinds of things."
He continued, saying, "To me, the generalizability isn't a question of [categorizing
industries and finding similarities]—I'd rather think about regulatory regimes and a
particular kind, and this particular kind of regulatory regime is voluntary by the firms that
aren't paid a lot of attention, but where reputational influences are important."
Conceding that he was still searching for "the answer," Dr. May re-asserted that "it's not
a question of the industry—it's a question of some of these other kinds of economic and
social influences. That's why I said whatever this typology is that's not there anymore
has to come into play." He closed by characterizing his response as "a non-response," or
"a long essay that says I don't know."
Jon Silberman (U.S. EPA, Office of Enforcement and Compliance Assurance)—a follow-
up response
"One of the reasons why I encourage Professor May to milk as much as he can out of the
work that he actually did with the marine facility sector is so that even if there are
transferability limitations on the work, we'll still get a very valuable product out of his
research."
Matthew Clark (U.S. EPA, Office of Research and Development)
"Jon, I just wanted you to define what "social marketing" means."
Jon Silberman
"I just happen to have a definition. It means: employing commercial marketing
techniques and outreach and assistance not just to inform but to actually change behavior.
If it were used in an integrated context, you would combine it with education to increase
awareness. You would use social marketing to increase openness and change and use of
incentives, and [you would use] enforcement and deterrence to address resistance to
change or non-compliance issues."
70
-------
Robert Kagan (University of California at Berkeley)
Dr. Kagan stated that he was "troubled by the notion that a violation is not a violation,"
or, to be more precise, the notion that "some are serious and some are less serious," and
he said he believed this was Jon's point as well. He wondered whether the data set would
allow the researchers to break the violations down by, for example, "those that carry big
fines versus those that carry small fines" or some other way of distinguishing "that's
more technical" as opposed to just "those that result in some significant harm."
Jon Silberman
Mr. Silberman responded that "for better or worse, every EPA program has a definition
of a term called "significant non-compliance," which actually makes value judgments in
distinguishing between compliance considered to be more significant and compliance
considered to be less significant," and he acknowledged that there are always differing
views about where that line should be drawn. He went on to clarify that the classification
is used in many different ways, including "choosing between facilities to target, choosing
the enforcement response—the response to the non-compliance, etc."
Irene Xiarchos (West Virginia University, Division of Resource Management)
Ms. Xiarchos began by commenting, "I'm very grateful that this is happening—at least
for me it's a big deal." She went on to ask whether social marketing is targeted more to
non-point source pollution or whether it is also directed toward point source pollution,
where there are already some regulatory controls set up.
Jon Silberman
Mr. Silberman deferred to Joanne Berman for an answer because she is "working more
directly with it."
Joanne Berman
Ms. Berman answered by saying, "Social marketing really is taking commercial
marketing concepts and applying them to a social environment... It's a process you go
through in order to understand your audience and then find the best approach to change
that behavior." She reiterated that the process is "designed for any audience, to
understand the early adopters and then go all the way to the laggards. So, what we would
do with social marketing is do some research on your audience, understand why there are
early adopters and why they're early adopting—focus your attention and resources on
those." She continued to explain that after identifying your partners and your audience,
you would work on "spreading the word, so eventually compliance and beyond
compliance—whatever behavior you would like to [foster]—gradually increases and
moves forward." Ms. Berman also clarified that enforcement becomes part of the process
of dealing with laggards, "because you'll never get some folks to change their behavior."
She concluded by summarizing, "So, generally, the process is to get them to change their
behavior, and then when the hammer comes in, it would be your last resort," but social
marketing "could be used in any context."
71
-------
Corporate Environmental Behavior and the
Effectiveness of Government Interventions
PROCEEDINGS OF
SESSION III: APPROACHES TO ENVIRONMENTAL PERFORMANCE
A WORKSHOP SPONSORED BY THE U.S. ENVIRONMENTAL PROTECTION
AGENCY'S NATIONAL CENTER FOR ENVIRONMENTAL ECONOMICS (NCEE),
NATIONAL CENTER FOR ENVIRONMENTAL RESEARCH (NCER)
April 26-27, 2004
Wyndham Washington Hotel
Washington, DC
Prepared by Alpha-Gamma Technologies, Inc.
4700 Falls of Neuse Road, Suite 350, Raleigh, NC 27609
-------
ACKNOWLEDGEMENTS
This report has been prepared by Alpha-Gamma Technologies, Inc. with funding from
the National Center for Environmental Economics (NCEE). Alpha-Gamma wishes to
thank NCEE's Cynthia Morgan and Ann Wolverton and the Project Officer, Ronald
Wiley, for their guidance and assistance throughout this project.
DISCLAIMER
These proceedings are being distributed in the interest of increasing public understanding
and knowledge of the issues discussed at the workshop and have been prepared
independently of the workshop. Although the proceedings have been funded in part by
the United States Environmental Protection Agency under Contract No. 68-W-01-055 to
Alpha-Gamma Technologies, Inc., the contents of this document may not necessarily
reflect the views of the Agency and no official endorsement should be inferred.
11
-------
TABLE OF CONTENTS
Session III: Approaches to Environmental Performance
A Multi-Agent Model of a Small Firm
Clinton Andrews, Rutgers University 1
Do Facilities With Distant Headquarters Pollute More?: How civic
Engagement Conditions the Environmental Performance of Absentee
Managed Plants
Don Grant, University of Arizona 24
Compliance and Beyond: Strategic Government-Industry Interactions in
Environmental Policy and Performance - The Role of Technical
Information in Reducing Automobile Emissions
Jennie Stephens, Harvard University 52
Discussant
Carl Pasurka, U.S. EPA, NCEE 67
Discussant
Don Siegel, Rensselaer Polytechnic Institute 77
Summary of Q&A Discussion Following Session III 90
in
-------
A Multi-Agent Model of a Small Firm
Clinton Andrews, Ana Baptista, Shawn Patton
Edward J. School of Planning and Public Policy
Rutgers University, New Brunswick, New Jersey
cjal @rci.rutgers.edu, abaptist@eden.rutgers.edu, shawn@shawnpatton.com
This paper presents a bottom-up view of industrial ecosystems by examining the
interpersonal dynamics that influence corporate environmental behavior. Employees of
profit-making firms don't always behave in the shareholders' best interests due to
misaligned incentives, impaired information flows, and bounded rationality. Even worse,
there are sometimes conflicts between shareholder interests and the broader public
interest, evident in the moral struggles of people over their dual roles as employees and as
citizens. Employees operate within the formal, regulative structures of the firm and
government, as well as the informal, normative or cultural structures of social networks.
The paper triangulates to identify useful insights about personal networks and corporate
environmental behavior, using interviews at firms, review of archival data, and a
computer simulation model. Interviews and archival data provide empirical grounding,
while an innovative multi-agent simulation modeling exercise supports formal theorizing.
The empirical work is based on case studies of plastics processing firms in New Jersey.
The simulation model characterizes production technologies, social and economic
structures, and interpersonal interactions under a variety of conditions. The model may
eventually prove helpful to managers interested in improving on existing organizational
practices and procedures. The model could also help regulators understand corporate
environmental behavior more fully.
Findings are relevant to both the eco-park and industrial ecosystem levels of analysis.
This work is funded by a U.S. Environmental Protection Agency STAR grant.
The next section reviews the relevant literature. Then the paper provides an introduction
to the industrial sector studied, describes case studies of four firms and findings from
those cases, introduces the multi-agent simulation model and modeling processes, and
draws conclusions.
Literature on networks and organizations
Studying organizations
It is difficult to study organizations because they are so highly complex and adaptive.
They exhibit structural complexity, having both functional and product hierarchies within
whose mesh individual employees act. Organizations are also goal-oriented systems that
survive by adapting to changing external conditions. The leaders of organizations work
full time to change that which researchers study.
The range of organizational phenomena is rich enough that distinct disciplinary
approaches to their study have developed. Often researchers talk past rather than to one
another over interstitial issues such as linking structure and agency. The normal
1
-------
progressive spiral of science from induction to deduction, pattern matching to hypothesis
testing, evidence to theory and back again, seems to operate inefficiently. Many theories
co-exist, and empirical work fails to eliminate most of these rivals. This paper attempts to
get the "wheel of science" (Wallace, 1971) spinning productively by directly
triangulating across extant theory, qualitative data, quantitative data, and simulation
modeling.
As discussed elsewhere (Andrews, 2001), there is much useful economic theory for
industrial ecologists to draw upon. It explains why firms exist (scale economies,
transaction cost reductions), how they respond to changing external conditions (internal
structural change, external influence projection), and how they relate to other firms
(contracting, mixed-motive strategizing). Some strands of the contracting literature also
assign agency to employees within a firm, typically highlighting mismatched incentives
and informational asymmetries.
Prescriptive management theories have progressed over time from the efficiency studies
of Taylorism, to classical management theory that emphasized commanding and
controlling, to the gentler and more respectful human relations approach of Mayo, to the
systems approach of Senge and others.
Classical and neoclassical economic theories provide an atomized explanation of
economic actions, whereas reformist economists view economic actions as embedded
within social structures. According to these theories social relations between individual
actors impedes competitive markets and individuals pursue a narrow utilitarian, self-
interest. This view is called "undersocialization". "Oversocialization" is when behavioral
patterns are so internalized that social relations have only a peripheral effect on behavior.
Granovetter (1985, 487) posits a middle position between over and under socialization
"Actors do no behave or decide as atoms outside a social context, nor do they adhere
slavishly to a script written for them by the particular intersection of social categories
that they happen to occupy. Their attempts at purposive action are instead embedded in
concrete, ongoing systems of social relations." He furthermore rejects the neoclassical
undersocialization theory arguing that"anonymous markets of neoclassical models are
virtually nonexistent in economic life and that transactions of all kinds are rife with the
social connection " In actuality, business relations today are mixed up with social
relations all the time. For example (p. 496), "in industrial purchasing, buying and selling
relationships rarely approximate the spot-market model of classical theory ...and
evidence consistently suggests that it takes some kind of 'shock' to jolt the organizational
buying out of a pattern ofplacing repeat orders with a favored supplierThe reasons for
this type of seemingly irrational behavior include costs associated with searching for new
suppliers and establishing new relationships. These relationships are formed through
trade associations, country clubs, and other social gatherings. The survival and success of
small firms in the market are in part due to a dense network of social relations overlaid on
top of the business relations that connects such firms and reduces pressures for
integration.
2
-------
Social Networks
There is a distinction between the 'formal' and the 'informal' organization of the firm
with the formal represented by the organizational chart and the informal represented in
the social networks within the firm. Organizational charts do not reflect the way the
work gets done inside an organization. There are many actors in employees' social
networks, all of whom have the potential to influence the employee. There are two main
types of actors that have this influence, and Shah (1998, p.250) call them "cohesive and
structurally equivalent actorsCohesive actors or referents are individuals with close,
interpersonal ties, or friends. Structurally equivalent actors are individuals who share a
similar pattern of relationships with others and thus occupy the same position in a
network. Shah (1998, p.249) has shown that"employees rely on structurally equivalent
referents for job related information and on cohesive referents for general organizational
information and as social comparison referents."
Informal practices and social networks serve distinct purposes within a firm. For
example, firms' internal information is not necessarily acted upon, particularly in the
context of promotion practices. According to Granovetter (1985, 499), "internal
promotions have affirmative incentive properties because workers can anticipate that
differential talent and degrees of cooperativeness will be rewarded." Long term
employees also have built up strong informal networks within the firm (Granovetter,
1985, 501), "when many employees have long tenures, the conditions are met for a dense
and stable network of relations, shared understandings, and political coalitions to be
constructed."
Individuals are more likely to obtain general organizational information (i.e. office
gossip, organizational culture, office politics) from cohesive actors, according to Shah
(1998, p.252). Social comparison theory suggests that similarity plays an important role
in referent selection. Demographic variables such as gender, age, tenure, and education
account for different aspects of similarity within workplaces. People often select referents
of the same gender, job category and education. Similarity in tenure and age may also
serve as relevant dimensions for career comparisons. In the cases studies to follow, there
is some evidence that long tenure and seniority on the job elicits greater influence in the
work environment than simple hierarchical positions. The studies also show that there is
greater cohesion in the workforce because of the similar ethnic backgrounds of the low
skilled workers of Hispanic origin. This demographic trait is also a link to higher levels in
the organization through internal promotions.
Workplace uncertainty, socialization practices and performance ambiguity may all lead to
different types of socialization within the firm. A routine, well-defined assembly line task
may elicit few inquiries regarding job responsibilities and performance. More complex,
loosely structured positions may generate many inquiries. The plastics manufacturing
firms included in this case study would fall under the category (Toone and Jackson) of
small batch production or "job-order manufacturing for customized products in which
production is done according to demand in small runs and lots."1
1 Toone, Roland and Jackson, Dave. 1987. The Management of Manufacturing: The competitive Edge.
Springer-Verlag; New York, p. 22
3
-------
Formal vs. Informal networks
According to Scott (2001, 153), there are two distinct features of firms today. "First,
there exists a remarkable similarity in the structural features of organizational forms
operating within the same organizational field Second, students of organizations have
long observed the presence of both a formal and informal structure, the former reflecting
officially sanctioned offices and ways of conducting business, the latter, actual patterns
of behavior and work routines. An uneasy tension exists between these structuresThe
formal and informal networks that frame inter- and intra-firm behavior are defined as
follows by Schermerhorn and colleagues (1988, 199):
• "Formal groups are created via formal authority for some purpose. They typically
have rather clear cut superior-subordinate relationships, and they often appear on
formal organizational charts." Formal groups are designated by an organizational
authority and can be seen in the production pressures and technical demands of a
company. Formal groups are specified by the organization chart (and by a task
group in a matrix management situation).
• Informal groups on the other hand are not formally recognized but typically
consist of subgroups or cliques within formal groups. These informal groups can
be people within a firm that eats together or goes on breaks together. Informal
groups emerge spontaneously. Informal groups consist of groups of individuals
that want to achieve some mutual objective (not the organization's but the
group's), sometimes they are merely friendship groups or people who have
something in common. According to Scott (2001), "This is really where/how
things get done in organizations." Informal groups can be seen in the regulative,
normative and cultural-cognitive elements of the company, including company
sponsored social activities of the sort mentioned in the case study to follow.
Informal networks exist because they help individuals do their work by "offering a
network of interpersonal relationships with the potential to 'speed up' the work flow or
gain favors in ways that formal lines of authority fail to provide" (Schermerhorm et al,
1988, 200). These informal groups also help individual employees meet needs beyond
what the formal groups can provide, including:
• Social satisfaction - friendship and social relations
• Security - "opportunities to find sympathy for one's feelings and actions,
especially as they relate to friction with the formal organization; opportunities to
find help or task assistance from persons other than one's superior"
• Identification - sense of belonging by associating with people who are similar
Organizational life cycle
As organizations increase in size, they typically become more heterogeneous in their
orientations and in the products and services they provide. This often results in movement
from a simple to a more complex structure.
4
-------
Organizational Life Cycle
Entrepreneurial Stage Collectively Stage
Formalization and
Control Stage
Elaboration of
Structure Stage
Marshalling of resources
Lots of ideas
Entrepreneurial activities
Little planning and
coordination
Formation of a "niche"
Prime mover has power
Information communication
and structure
Sense of collectivity
Sense of collectivity
Long hours spent
Sense of mission
Innovation continues
High commitment
Formalization of rules
Stable structure
Emphasis on efficiency
and maintenance
Elaboration of structure
Decentralization
Domain expansion
Conservatism Adaptation
Institutional procedures Renewal
Table 1: The Organizational Life Cycle
Source: Cameron, K. S., and Whetten, D. A. 1983. Models of organizational life cycle:
Application to higher education. Rev. Higher Educ. 6(4): 269-299.
Many firms, especially smaller enterprises, never reach the later stages in the
organizational life cycle, either because they disappear or because they don't reach a size
that requires much formalization. Nevertheless, the importance of the distinction between
formal and informal social networks grows as structures become more complex.
Industry Background
The industry sector studied in this project is plastics products. It was chosen because the
technology is relatively simple, it has eco-efficiency and pollution reduction
opportunities, there are many small and medium-sized firms available as case study
candidates, and it is undergoing a dramatic transformation due to competitive pressures
from economic globalization.
Plastics Product Manufacturing
The two basic groups of plastic materials are the thermoplastics and the thermosets.
Thermoplastic resins consist of long molecules, each of which may have side chains or
groups that are not attached to other molecules, so they are not cross linked (SPI, 1999a).
Thus, they can be repeatedly melted and solidified by heating and cooling so that any
scrap generated in processing can be reused. No chemical change generally takes place
during forming. Usually, thermoplastic polymers are supplied in the form of pellets,
which often contain additives to enhance processing or to provide necessary
characteristics in the finished product (e.g., color, conductivity). The temperature service
range of thermoplastics is limited by their loss of physical strength and eventual melting
at elevated temperatures.
5
-------
Tbermoset plastics, on the other hand, react during processing to form cross-linked
structures that cannot be remelted and reprocessed. Thermoset scrap must be either
discarded or used as low-cost filler in other products. In some cases, it may be pyrolyzed
to recover inorganic fillers such as glass reinforcements, which can be reused.
Thermosets may be supplied in liquid form or as a partially polymerized solid molding
powder. In their uncured condition, they can be formed to the finished product shape with
or without pressure and polymerized by using chemicals or heat.
plastics industry jobs in New Jersey: 51,011
Employment in New
Jersey's plastics industry
totaled 51.011 jobs in
2001- New Jersey ranks
11th in the country in
plastics employment.
1994 1996 1999 2000 2001
plastics industry shipments in New Jersey: $11.7 billion
is r
Plastics industry shipments
in New Jersey totaled
$11.7 billion in 2001. New
Jersey ranks eighth in the
country' in plastics
shipments.
1994 1996 1999 2000 2001
Figure 1: New Jersey Plastics Industry Employments and Shipments
Source: SPI (2002)
New Jersey is one of the top ten states accounting collectively for 60% of the total U.S.
plastics industry shipments (SPI, 2002). Unofficial statistics suggest that both
employment and shipments have dramatically declined in this industry since 2001.
Plastic Injection Molding Industry
Injection molding is the principal method of forming thermoplastic materials. The
production process is organized around runs of product (e.g., an order for 100,000 plastic
coffee cup lids). Large volume runs of simple items (like coffee cup lids) have low profit
6
5Q,COO
40,000
30,000
20,000
10, COO
-------
margins because there are too many competitors for this type of simple product. On the
other hand, the most profitable firms deliver high quality, complex, molded products
often in smaller runs (e.g., an order for 1000 laptop computer housings). Generally, the
injection molding business has a range of production specialties. At the smaller end, the
precision molders make very small parts and at the large end they can make larger, more
complex parts (e.g., automotive parts). There are two types of injection molders:
• Custom, contract molders make parts specific to the needs of their customers
• Proprietary, captive operations make their own products
The relevant NAICS codes that apply to this industry include:
NAICS Code
Description
326199
All Other Plastics Product Manufacturing
325991 & 3261
Plastics Product Manufacturing
32613
Laminated Plastics Plate, Sheet, and Shape Manufacturing,
32614
Polystyrene Foam Product Manufacturing
3087
Custom Compounding of Purchased Plastics Resins
325991
Custom Compounding of Purchased Resin
Technology & Innovation
Injection molding is a branch of the plastics industry that involves injection under
pressure of molten plastic into the cavity of a mold followed by cooling and removal of
the solidified part that retains a replica of the mold. The injection molding industry is
arguably in its infancy. It was only during the 1960s that reciprocating screw technology
became commercially viable. With the advent of the microprocessor, there have been
significant advances in process control during the 1980s and 1990s. There have been
equally significant advances in screw technology, multi-color molding, insert molding,
gas assisted injection molding, and other niche processes. There have also been major
advances in polymer materials, mold making, and of course, predictive analysis tools for
avoiding problems before they occur and optimizing every phase of the design-to-
manufacturing process.
However, in spite of all these advances, the injection molding industry continues to
exhibit signs that it is still a very young industry. For example, it remains common to
set up and optimize the process using time-consuming and inefficient trial-and-error
methods. While molders may be able to obtain acceptable quality parts using this
method, the process usually requires constant fine-tuning to maintain quality parts
because it was not set up using a rigorous scientific quality control method. Failure to
setup and optimize using a rigorous method normally results in a process that is not
robust and therefore, is difficult to control. Beyond the setup, optimization, and
control of the process, there are additional injection-molding manufacturing tasks that
must be performed, optimized, standardized, and integrated across the company-wide
enterprise. These additional tasks include, but are not limited to, production
scheduling, preventive maintenance, process and production monitoring, statistical
process control, statistical quality control, and production reporting. It is also
becoming increasingly common for an injection molder's customers to demand value-
7
-------
added operations such as part traceability, while simultaneously demanding per-part
price decreases. Facing these challenges, injection molders must not only implement
systems and processes to achieve the value-added demands, but also accomplish them
cost-effectively while improving the efficiency of their existing operations.
Employee Tasks
Injection molders typically work in small independent firms with relatively few
employees (5-100). Most of the employees are semiskilled workers who load plastic
pellets into the injection molding machines, mixing in some recycled plastic waste as
available. Once the plastic has cooled and re-solidified the mold opens and the plastic
product is removed. If the machine's temperature is set too high, air pollution can result in
the form of fugitive volatile organic releases. In a typical machine, every 30 seconds the
machine completes a cycle, dumping a cooled molded plastic piece onto the factory floor.
Injection molding machines require thorough maintenance, otherwise they become
unreliable. Workers take the molded plastic pieces and break off the extra bits of plastic
(little nubs and frames). The amount of plastic waste is a function of the mold design and
the amount of product made.
Workers then put the waste plastic into a grinder and store it for use as recycled
feedstock. Recyclability is a function of the type of plastic material used (some plastics
can't be recycled once heated). Un-recyclable plastic is disposed of offsite. Workers
inspect the product and reject some pieces (these get recycled) and pack the product into
boxes for shipping. These boxes are shipped to customers according to a supply
schedule. A process engineer supervises multiple injection molding machine lines and
orders raw materials. A marketing manager solicits orders for products and a plant
manager coordinates the marketing and production activities, settles employee disputes,
and seeks to maintain profitability.
Industry Outlook
The injection molding business's golden era spanned thel970s - 80s when there was less
competition at the machine and process level and firms produced very high profit
margins. Now there are abundant machinery manufacturers and processors inundating the
market. Processors range from small family operations with a handful of machines to
larger companies with hundreds of machines. Other dynamics are also lowering the
margins, including increased competition from Asian imports. Asian markets have very
low costs, particularly labor costs, relative to U.S. operations.
Plastics Injection Molding Process
In injection molding, plastic material is put into a hopper that feeds into a heated injection
unit. A reciprocating screw pushes the plastic through this long heating chamber, where
the material is softened to a fluid state. At the end of this chamber there is a nozzle that
abuts firmly against an opening into a cool, closed mold. The fluid plastic is forced at
high pressure through this nozzle into the cold mold. A system of clamps hold the mold
halves shut. As soon as the plastic cools to a solid state, the mold opens and the finished
plastic is ejected from the press SPI, 1999b).
8
-------
f(mold design)
f(material)
Molded
Product
Injection
Molder
Ship Product
to assembler
Electricity
from heat
Pack Product
Break product
from molding
Recycle
Dispose Off site
Volatile Organic
Releases
Install an injection
molder
Waste molded
connectors
Stainless steel
mold from
specialist
(expensive)
Load into
injection molder
Mix X%
f(material)
recycle
Warehouse
feedstock
Pellets from
petrochemical
company
Figure 2: Plastics Injection Molding Process Diagram
Methods
Three plastics injection-molding firms in New Jersey and one multinational chemical
corporation were selected for study. The three plastics firms were selected because they
were accessible and because they have relatively simple manufacturing processes that
could be more easily modeled. The multinational chemical company was studied less
formally, specifically for their approach to environmental management. The case studies
of the three companies include in depth interviews with the presidents or owners of each
9
-------
firm, analysis of business, financial and environmental records, and site tours of the
manufacturing floor. Each interviewee reviewed and signed an informed consent form
and their names and the names of their companies remain confidential.
l-'inn
Description
Company 1
Plastics, Injection molding, Single establishment, family run
Company 2
Plastics, Injection molding, Single establishment, family run
Company 3
Plastics, Injection molding and extrusion, Subsidiary of Multinational
company
Company 4
Multinational chemical firm
Case Study Findings
In depth interviews with the presidents or managers of the three injection molding
companies revealed a great deal of information regarding the importance of: formal and
informal networks in workplace practices such as innovation and safety measures, the
role of a family run vs. a corporate culture environment, external market dynamics, stable
workforce dynamics. The following hypotheses reveal a rich picture of this particular
industry and also highlight some important lessons more generally, regarding
organizational behavior's ties to workplace practices.
Hypothesis la: As the external environment becomes increasingly competitive, the
family-run businesses decrease the social amenities available to employees. Supported.
Hypothesis lb. As family run businesses experience a generational shift, the social
practices of the company also shift. Supported.
Interviews with all three companies revealed that there has been a shift in the business
culture towards a less social work place indicated by the decrease or elimination of
company sponsored social activities such as company sports teams or company picnics.
In the case of the first two companies, which are run by a second generation of family
members, this shift to less social activities also coincides with a shift towards a more
competitive market environment and less profitability. Interestingly, the third company
which operates under a more corporate culture (subsidiary of multinational company) and
is not run by family members, offered many more social amenities to their employees
than the first two family run companies. This difference may mean that the decrease in
social activities in the first two companies may be primarily due to scarcity of resources
to support such activities. On the other hand, the quality, scope and relevance of the
social activities in place at Company 3 cannot be measured by this case study and
therefore it is difficult to compare and contrast the activities of one company with past
activities of the others. The generational shift evident in the two family-run companies
was also accompanied by a downturn in the economy and increased competition in the
industry. Were these social activities cut because the second generation was not as
socially tied to the employees or had a different relationship with employees than their
predecessors? Or is this decline a direct outcome of scarce resources to dedicate to social
amenities?
10
-------
According to Company l's President, the decline of company sponsored social events
might be attributable to both a generational shift and a resource allocation issue. The
current owner feels that if his father were still running the business that many of the
social activities he sponsored would probably still be in place. This is because part of his
father's "style" was that he felt more comfortable interacting with the employees through
these types of activities as opposed to more informal contact on a daily basis. The
current owner thinks it was partly due to his age that he didn't relate as easily with the
employees so that the social activities were a mechanism for reaching them. The owner
today and his brother take a much more relaxed approach in which they feel comfortable
with their employees and interact with them on a daily basis in a less structured way on
the factory floor. In this case, the social activities would be nice but they are not really
necessary for them to get to know the employees. On the other hand, if the current
economic situation allowed for more profitability with a surplus of money available, a
situation more similar to when their father ran the business, then they might well consider
offering more social activities. In many ways the decision to offer company-sponsored
social activities is product both of resources and personal style.
At Company 2 on the other hand, the original owners, the fathers, were very "hands on"
interacting mostly on the factory floor as technical tool-makers, they were very close to
the factory workers. The second generation in this company was less "hands on" and
more focused on the business aspects of the company. At company 2, the owners
attribute the decline in social activities more to a lack of resources and the sense that
employees were not participating in the activities. In the past the company hosted several
company sponsored social activities such as picnics that have since faded when they
moved to a new facility a few years before, "We used to have picnics, a company picnic.
The last one was before we moved over here, but we stopped them because our business
hadfallen and the money wasn 7 really there for that or we didn 7 want to use it for that. I
think there was also some distaste on our part that we didn't feel our employees were
participating at that point."
Company 3 offered a wide range of social activities including a bowling team, company
luncheons, educational training, and other activities. Company 3's manager sees a great
deal of value in sponsoring these events. The manager states, "/ would say that happens
[interaction between positions] in the sports driven activities like bowling. It gets the full
gamut of employees. You get staff management out there bowling and the maintenance
folks and set up operators and packers, one big team." This type of socialization is also
seen as a positive contributor to company morale on the factory floor. The general
manager describes the effects of such social activities on the company as follows, "It
helps on the factory floor. I don't know if I can quite put my finger on it, but when you
have a crew that's been around, that's as senior as the one I've got, there's a wonderful
camaraderie but there's also a totem pole."
Hypothesis 2: Family run businesses have strong social ties to employees, and thus may
be less likely to streamline and cut labor. This theory is based on the human relations
theory of organizational behavior. Not supported.
11
-------
This dynamic works well for explaining promotional or recruitment/ hiring practices but
is not the driving factor in the business. Company 1, for example, relies on informal
communication networks and close ties/familiarity with employees to determine
promotions and even hiring or firing decisions. But Company l's president also
emphasized the need to continuously streamline and cut low skilled positions by
automating these jobs, thereby reducing high labor costs allowing them to stay
competitive in the marketplace. Company 2 actually detailed the difference in approach
to employees with the increased competitive market, "The biggest difference between
business today and 15 years ago is that you can't stand still. Fifteen years ago you could
stand still and just make product and move things along, add an extra employee here or
there. If they weren't contributing too much we just let it go. Today it's to the point
where you can't afford any of it. And it's hard to get business because there's so much
competition. " The second generation of owners has to consider this increased
competition when making decisions about the labor force. Company 3 on the other hand,
while not a family-run business, emphasizes the importance of family members working
together in the company, "It makes us a small, entrepreneurial, family run business with
a push of a big organization behind us. It makes us human. Christmas parties are more
fun. We don 7 have kids working here but we have hadfamilies over the years, a husband
and wife, a 45-year employee with her son who's been here 25 years. So yes, we have
families. " Increased competition in the business is driving the streamlining of the
workforce, and this raw economic factor outweighs most personal connections to
workers. However, this case study was unable to document the exact pattern of hiring
and firing practices conducted by each firm and had to rely on management's account.
Hypothesis 3: Informal communication networks will be important for a variety of
business management aspects for family run businesses. Supported.
Informal networks seem to be important for how the family firms in particular (Company
1 and 2) handle issues such as: environmental and safety procedures, supplier and
customer relations, promotional, hiring and recruitment practices (and termination).
Informal Networks are important for recruitment, hiring, and promotional practices inside
all three firms. Recruitment practices in Company 1 and 2 are also based on more
informal networks or "word of mouth" from current employees. This informal
mechanism of bringing in new employees is another reason why many family members
work together on the factory floor. By hiring in this manner, existing external social
networks are transplanted into the workplace. Seniority and tenure in the workforce
matters more, in terms of stored knowledge and experience in the workplace, than formal
credentials. This reliance on experiential knowledge is evidenced by the promotional
practices in all three companies where promotion to higher skill levels occur from within
the company as opposed to bringing in new experts from the outside. For termination,
Company 3 relies on a more formal process involving the corporate human resources
department, while the first two companies rely more on informal processes for reviewing
individual employee behavior and performance both for promotional and termination
consideration.
Companies 1 and 2 describe an incremental approach to innovation in which they try a
new idea for a little while and then determine whether it is viable to go on before making
12
-------
a large investment in a new product or process. This type of innovation also relies on
informal information networks like trade journals and trade shows, relationships with
machine manufacturers and customers and relationships with senior employees that are
familiar enough with the business to develop new ideas. In Company 1 and 2, the owners
were directly involved in the innovation process with no formal R&D staff in place.
Company 3 on the other hand relied on corporate R&D support services for larger scale
innovations in the production process. But Company 3 still developed many of its
practical innovations on the factory floor with help from long term employees.
Company 3 seemed to rely on factory floor employees to improve safety measures in
particular. The company set up a subcommittee and a suggestion box on the floor to
encourage employees to bring their interests and innovation to bear on the issue of
workplace safety. The manager of Company 3 emphasized the possible importance of
employees' previous experiences or knowledge outside the firm to bring innovations and
improvements to safety procedures in the workplace. The manager perceived this input
from employees as a driving factor in the improvement of their safety record. Informal
networks also seem to be important mechanisms for financing for the two family run
businesses, Company 1 and 2. These two companies rely on long term banking
relationships as their main source of financing and this relationship is based on trust in
the reputation of the firm. The corporate firm, Company 3, relies on more formal
mechanisms for financing through their corporate structure. Within this structure,
Company 3 had to follow a formal process for justifying any new financing.
Formal networks are important for a variety of functions in all three companies although
it is more prominent in Company 3. Company 3 is tied to a corporate parent that imposes
a more formal structure on the firm than is evident in the first two single establishment
firms. All three companies comply with federal and state environmental (EPA), Safety
(OSHA) and labor standards. All three also seem to pursue environmental (recycling
waste) and safety improvements according to an eco-efficiency principle in which the
improvements are done independent of economic activity but the impact of the
improvements are felt both in economic and environment and safety measures. While all
three companies are compliant with some type of trade standards, Companies 1 and 2 are
moving towards increased compliance with newer industry standards like ISO 9001 and
14001. Company 3 seems to have many of these certifications in place already, which
again may be a reflection of more stringent corporate standards and more available
resources to come up to compliance. All three companies describe the impacts of
increased supply chain management schemes which put pressure on them to take on more
of the risk. The three companies also have a flat organizational structure with
manufacturing jobs representing the bulk of the employee base at the bottom of the
hierarchy.
Hypothesis 4: There will be high turnover in laborers because of low skill, low wage
nature of work. Not supported.
The low level employees in the company are generally low wage earners with pay
ranging from $6.25/hour to $8.50/hour. While this wage seems relatively low, compared
with other low skill level jobs in the service sector, these manufacturing jobs represent
13
-------
better opportunities because of the accompanying benefits packages. Despite the low
wages and the repetitive nature of the work, all three companies describe an extremely
stable workforce with low turnover in all levels of workers. This low turnover may also
be due in part to the opportunities for promotion within the company. The owner of
Company 1 stated, "Everyone in our supervisory positions have been promoted from
below, but she [an employee we met on the floor] was the first one to cross the picket
line, so we have a special affection for her." Company 2's described promotions from
lower levels, "There's a lot of that particularly with us. I think most molders are probably
like that. You've got somebody who started second shift to stand by a machine, he shows
a little bit of mechanical skill and interest in the job and we say well let's try him out
here. If it works out well, it keeps on going. Right now our customer manager, which is
probably one of the most important things we do here, he started out as an assistant
foreman on second shift. He's a young guy who's going to school, he spoke good English
which is important, showed a lot of energy and a lot of interest and moved up to assistant
supervisor..." The manager for Company 3 also states, "We seek to grow people within
the organization.. .we have various folks in our business who started in the plant."
Each interviewee recounted "success" stories of employees who started out in a very low
level position like operator or packer and how they worked their way up the hierarchy
through promotions due to good work habits, positive attitude and interest in moving up.
The firms seem to reward good worker traits and reinforce this through internal
promotions. The firms also did not put much emphasis on high levels of education or
schooling but emphasized more the importance of experience and reliability. The
Hispanic low-level workers are newer to the firms and are working their way up through
the ranks. In Company 1 for example, the recent promotion of a Hispanic worker into a
supervisory position is seen as a positive impact on lower level workers' morale because
they feel closer to the upper ranks and they can aspire to also be promoted. This same
worker was promoted because the owners of the company admired her loyalty to the
company during a union strike when she crossed the picket line first. This illustrates how
promotions are based on more than just efficiency or lines of command within the
organization.
The interviewees described some overarching traits that are desirable for hourly, low
skilled workers which include; manual dexterity and proficiency on the machines,
reliability in attendance, quality of products and functions, willingness or interest to leam
business, loyalty. For salaried or higher skilled workers, interviewees emphasized the
level of commitment and interest in the business, accountability, reliability and positive
attitude in the workplace. Free rider or shirking problems in the industries arose in one
example from Company 2 when the interviewees described problems of accountability
with employees. The company does not look favorably on employees who shirk
responsibility for problems or mistakes on the factory floor. This shows that where there
is a lack of accountability, shirking will occur and the human resources process weeds
out people who tend to be unaccountable.
14
-------
Outlook on the Injection Molding Industry
Company 3's business prospects for the future seem to be more secure than Company 1
and 2 due to their relationship with a larger parent corporation which provides them with
greater flexibility, mobility and resources than the small family run companies. While
Company 3 is small in terms of the number of employees at the facility, they can afford
to be leaner (in terms of employees) because of the additional resources provided by the
parent company. The drive of the plastics industry towards Asia seems to be easily
accommodated by the parent company's relationship with other business units located in
Asia. Both Company 1 and 2 experienced both a generational shift and a large market
shift in their businesses in recent time. The plastics industry became increasingly
competitive in the late 1990s while their fathers were transitioning the companies into the
hands of their sons. This dual shift may account for a transition in both the business
strategies employed by the firms to remain profitable and the social dynamics of the
employee base. Traditionally, these family-run companies relied on long term, low
skilled factory floor employees and repeat customers with little marketing or research and
development efforts. Today, all three companies face increased competition from Asian
companies that offer the same products but have much lower labor costs than US firms.
This shift in the market has forced US firms to streamline their labor force and become
increasingly automated to increase efficiency and reduce labor costs. The companies are
also forced to find competitive advantages in their product marketing and innovation.
Company 1 in contrast to Company 2 is more optimistic about its future prospects in the
business. This optimism is primarily due to Company l's multifaceted strategy for
surviving in the increasingly competitive market through streamlining, increased
automation to reduce labor costs, horizontal integration via the acquisition of smaller
operations and cornering a niche market in fire safety equipment along with a large
multinational company contract. Company 2 has increasingly automated but is struggling
to market their business and tap into new customer bases that they can keep long term.
Company 3 is perhaps the most economically stable due to its connections to business
units worldwide and their corporate resources.
One indicator for the strength of social networks in each company can be seen in the
company sponsored social activities. Company 1 and 2 both experienced a decrease in
the number of activities sponsored by the company at the same time that the dual
generational and market shifts occurred. Interestingly, Company 3 seems to offer many
more company sponsored social activities than the family run companies - suggesting that
it's not just the familial nature of the company but rather the financial stability of the firm
that matters a great deal in terms of supporting such social activities. Despite the drive
towards automation, all three companies rely on a stable workforce characterized by low
turnover and long term employees. The importance of these long-term employees in the
workplace is reflected in the "totem pole" hierarchy or informal hierarchy that is
established within the rank file between long term employees and new hires. According
to this totem pole, long-term employees' rank overrides any professional credentials a
newcomer brings to the workplace.
15
-------
The entry-level employees in all three companies are comprised of mostly Hispanic and
other non-English speaking people. The similar ethnic background of the entry-level
employees and the internal promotion practices creates a very close knit employee base
which may also have many connections outside the workplace. It is difficult to
characterize the nature or extent of these social networks because this study did not
interview or study these employees directly. Company 3's manager emphasized the
importance of personal and social networks in improving the safety on the factory floor in
particular. Social networks, social activities outside work and the presence of family
members working on the floor were perceived as a benefit to the company in terms of
improvements in safety along with more general improvement in morale and
productivity. But unlike the other two companies, Company 3 did not have family
members running the operation, only working as lower level employees.
The competition from China is putting increased pressure on all the firms to cut back low
skilled operators. At the same time, the owners seem to value company loyalty as
evidenced in their respect and admiration for long-term employees. It will be interesting
to see how these two forces - increased drive to streamline the workforce and a close
connection with the employee base will evolve over time.
Model-building activity
Researchers can productively induct theory from case studies that, like good computer
programs, offer parsimony and logical coherence (Eisenhardt, 1989). This begs a
question: why not express theory in the form of a computer program? Parsimony was a
goal of the modeling effort, and logical coherence was a handy byproduct of the
debugging process. The dual challenges with any type of modeling are to simplify reality
appropriately and to communicate the results effectively (Andrews, 2002). Both
challenges proved significant during the modeling process.
Multi-agent simulation
Object-oriented programming languages like Java make it possible to specify and
replicate software agents relatively easily. These agents can be purposive and
autonomous, and they can interact with one another and with an external environment.
Multi-agent simulation modeling, so called, is an intellectual descendent of game theory,
artificial life, and cellular automata, and it is gaining wide use as a social science research
method (Epstein and Axtell, 1996). There is already some experience with applications to
organizational behavior research (Carley and Prietula, 1998) and industrial ecology
(Axtell et al, 2002).
Using the Brookings Institution's Ascape multi-agent simulation framework (Parker,
20000), a Java programmer created PolyModel, a simulation of operations at a plastic
injection-molding firm. Approximately 100 employees interacted with the production
technology and one another, subject to changes in the firm's external environment. The
model included technology details, organizational structures, and parameter values taken
from Company 1 in the case study. The model tested alternative theoretical constructs
16
-------
explaining the behavior of employees, to be roughly validated against the evidence from
case study Companies 2 and 3.
The current model includes 22 classes of agents, related as follows.
PolyModel contains People, the Factory, and the External Environment.
The Factory includes a Warehouse, Production Lines, and a Shipping Department.
Employee extends Person.
Owner, Plant Manager, Marketer, Engineer, Shift Supervisor, Shipping Clerk, Materials
Mixer, Maintenance Technician, Machine Operator, and Janitor all extend Job. Each
Employee has a Job.
Remaining Java classes serve as computational infrastructure.
The time step in the model is hourly, so the firm cycles through the workday and the
work week over a period of years. Each employee assesses whether to go to work every
morning, based on health, social pressures, and finances. The plant manager determines
how many production lines and associated employees are needed based on pending
orders for widgets. The marketer brings in orders and tries to keep ahead of production so
that the capacity factor of the plant is high. The janitor keeps the factory clean, and other
employees become unhappy if the factory gets dirty. The materials mixer ensures that
raw materials reach the production lines, and the shipping clerk packages completed
products and sends them out the door. The maintenance technician keeps the production
lines in working order. The machine operators perform several sequential duties (load
plastic pellets, set molder temperature, separate widgets from scrap plastic). The shift
supervisor encourages machine operators to work more carefully and reports on
employee performance to the plant manager. All employees are subject to worker error
that affects the quality of their performance, and the probability that error will occur is a
function of aptitude, experience, tiredness, and happiness.
}
public double getWorkerError(){
double aptFactor = ExternalEnvironment.getAptitudeWeight() * (100-
this.getAptitude()) / 100;
double expFactor = ExternalEnvironment.getExperienceWeight() *
(65-this.getDaysWorking()/250)/65;
double hapFactor = ExternalEnvironment.getHappinessWeight() *
(100-this.getHappiness())/100;
double tirFactor = ExternalEnvironment.getTirednessWeight() *
this.getTiredness()/100;
return aptFactor + expFactor + hapFactor + tirFactor;
}
Happiness is a weighted additive function of wealth and social embeddedness.
MoneyGrubbers like wealth (90, 10), SocialAnimals like their friends (10/90), and
TheRestofUs are more balanced (50, 50). Wealth increases by getting paid at work, social
embeddedness increases by making more friends at work and elsewhere. Friendship
depends on affinity (similar intrinsic characteristics) and frequency of interaction.
As the screenshot below shows, the dynamics of these employee interactions provide
realistic drama and aggregate up to firm-level performance measures of interest to
17
-------
management. Parameters are adjustable on the fly, and various diagnostic tools allow the
user to investigate the causes of particular dynamic behaviors.
r-: Crab r k
' ' ' Mevtoi
-#¦ «¦ £7 it * rr. HMlfM
yilpptiig Clo* ht'd wfine '£
i Mi»wi lull] jgi 11ih L"S
yiipc*ng Ulw* he-a#5 Miner htf d up lln- 19
hJ«r.cr-*t* Mmr hild . ^ limi Li
Wwreftab Mi.tr held *i> line It
«:Uw ¦ ¦ II *«\ ¦ ...• * • '
! ^ itojnttuvs Q iiihii
; nt"4l Cn 12?b Hii-rsfifl "4"
f ® aX 9
-iQduc. Jf-atLwy.. Mil lit 1.
MSCIPO
Vslv*
Parameter
Com PUuic
0.25
COH Pppcr
0 LZ
Coif WtKI
OJliS
Cvfit*". HOU»
0
'Tn—ft Monln
40
0 a* Or tort
i
F.-rnr Act sijdr Wdahi
in >)
£•1X1 r LifMrtirMfe
]0.u
Lcror Macplness Wei &r.
10.0
t-ror TiraefcUEts
Vl.u
f^ice
OS
Nuii'i tr'Fj|u?tv»
lii
"lumbf Of iwnnl"*s L>*ej 9
Prwraitrvq
su
S*2t
2
ixn
112
Tyct Aftrctnr
13
'we P^ceni
In
CwiVlf# r]
—Taui Fronts
J ?'tk OsKi
5j Wcn^i Pr-riliir^r
H 4£Y«flirtS
I liwni*!
J haflu
IMrwtt-Jv 3 rve n >,
Onrt type
J ^ M ri*ve S«H:5
OU'l Coiufi
«; FcsegitiUBdl
!U[kq:Durx!
J. «« fee ground
.*3 "aii *o!it;
opdnm
^ 5&cn» ub£4
9
1flj|
^ Zv*i
J3X23!JEf^f3Ui|H- #. J&yU
M
Developing the model required intense interaction between the programmer and the
qualitative researchers. Much conversation centered on eliciting precisely what was the
theory being formalized in the model. As the researchers played with the resulting
simulations, the theoretical framework evolved.
Illustrative Result—Bringing in Worker Error
This paper briefly shows one illustrative result. The project is ongoing and the model,
underlying theories, and empirical evidence continue evolving. The model may
eventually become robust enough to serve as a management-training simulator for the
plastic injection molding industry.
The frictionless neoclassical model of the firm typically assumes that every employee
behaves like homo economicus, a rational, omniscient, selfish maximizer. Only principal-
agent problems detract from corporate performance in that model. Our model allows us to
turn worker error on and off, and thereby compare results under contrasting assumptions
regarding that element of bounded rationality. As the following graphs show, a firm
having imperfect (aka realistic) employees is less profitable and pollutes more. Policies to
reduce worker error can now be tested in silico. More detailed theorizing about the
determinants of worker error also becomes possible.
18
-------
500000
400000
300000
200000
100000
0
H
/i
LPjlM
Pi
n
it
If
1J
J1
n
0 8000 16000 24000
DataView #2
— Total Monthly Revenues
Total Monthly Expenses
—Total Monthly Profits
Homo economicus: Profits without worker error
500000
400000
300000
200000
100000
0
J1
ru
ft n
iWW.
ij
DataView #1
Total Monthly Revenues
—Total Monthly Expenses
Total Monthly Profits
0 8000 16000 24000
Realistic employee: Profits with worker error
19
-------
40000
30000
20000
10000
0
In
(v
j
DataView #2
—Total Monthly Air Pollution
0 8000 16000 24000
Homo economicus: Air pollution without worker error
Chart View
90000
80000
70000
60000
50000
40000
30000
20000
10000
0
r-TL!
h
A
Pi n
ji
Jli
J^J
\l
j
DataView #1
—Total Monthly Air Pollution
0 8000 16000 24000
\A
Realistic employee: Air pollution with worker error
20
-------
900000
800000
700000
600000
500000
400000
300000
200000
100000
0
fjji
ny
L
nil
DataView #1
Total Monthly Waste
— Total Monthly Waste Recycled
—Total Monthly Disposed Offsite
0 10000 20000
Theorizing worker error #1: Error = f(experience)
600000
500000
400000
300000
200000
100000
0
[u
ui
|TiJ
J"
~il
DataView #1
— Total Monthly Waste
Total Monthly Waste Recycled
—Total Monthly Disposed Offsite
0 10000 20000
Theorizing worker error #2: Error = f(experience, aptitude, happiness, tiredness)
21
-------
Conclusions
Theory building
Regarding the motivating question for this research—what are the relative roles of
informal social networks and formal regulatory structures?—the modeling and case study
evidence support three insights to date. First, informal networks are very important for
hiring new employees and for helping employees to decide to take job actions like strikes
and sick day protests. Second, formal structures are hugely important for explaining
almost everything else. In this industry there also appears to be a substantial amount of
technological determinism. In other words, the type and economics of the technology
explain much of the firm's overall behavior.
Lessons learned
There are two major lessons learned for researchers interested in using multi-agent
simulation models and case studies in a grounded theory-building context. First, this
project shows that highly diverse skill sets are needed. In fact, it is unlikely that a single
individual will have the requisite range of skills, necessitating recruitment of a
multidisciplinary team consisting of an interviewer, case study developer, and Java
programmer. Second, iterative modeling and interviewing is crucial because new
questions arise, and alternative theories need to be explored and elaborated.
The benefit of developing multi-agent simulations in this inductive way is that they
appear to inform action more directly than a deductively-based model built from
principles rather than evidence might. It becomes a humbler but perhaps more valuable
type of social science.
Case studies are informative but static research products. By taking the next step and
constructing a simulation model, this research becomes more dynamic and iterative. It
becomes easier to communicate theoretical expectations and to revise them. It potentially
can help with in silico management training and strategy development so that fewer
costly mistakes get made by firms and their regulators.
Future work
There are many valuable extensions of this work that deserve future attention. First, the
establishment-level model should be extended to the case of the branch plant with a
corporate parent. Then the modeling effort should expand vertically to include the supply
chain, and horizontally to include sectoral competitors. It would also be interesting to
adapt this modeling approach to industrial clusters and eco-industrial park tenants. In
addition, much more needs to be done to explore the potential for socially responsible
behavior to affect overall corporate performance. Other extensions suggested by the case
studies include further investigation of the special characteristics of family owned
companies, and of the value and measurement of employee loyalty.
References
22
-------
Andrews, C.J. (2001) "Building a micro foundation for industrial ecology." Journal of
Industrial Ecology 4(3): 35-51.
Andrews, C.J. (2002) Humble Analysis, Praeger, Westport, CT.
Axtell, R., C.J. Andrews, and M. Small (2002) "Agent-based modeling and industrial
ecology," Journal of Industrial Ecology, Summer.
Carley, K. and M. Prietula, eds. (1998) Simulating Organizations: Computational Models
of Institutions and Groups, Cambridge, MA, MIT Press.
Eisenhardt, K.M. (1989) "Building theories from case study research," Academy of
Management Review 14(4): 532-550.
Epstein, J., and R. Axtell (1996) Growing Artificial Societies: Social Science from the
Bottom Up, Brookings Institution/MIT Press, Washington, DC.
Granovetter, M. (1985) "Economic Action and Social Structure: The Problem of
Embeddedness," The American Journal of Sociology 91(3): 499, 501.
Parker, A., Cross, R, and D. Walsh (2001) "Improving collaboration with social network
analysis," Knowledge Management Review 4(2): 25.
Parker, M. (2000) "What is Ascape and why should you care?" Journal of Artificial
Societies and Social Simulation, 3.
Schermerhorn, J.R., Jr., Hunt, J.G., and R.N. Osborn (1988) Managing Organizational
Behavior, 3rd ed., John Wiley & Sons, New York.
Scott, W.R. (2001) Institutions and Organizations, 2nd ed., Sage Publications, Thousand
Oaks, CA.
Shah, P.P. (1998) "Who are employees' social referents? Using a network perspective to
determine referent others," Academy of Management Journal 41(3): 249, 252, 255.
SPI (The Society of the Plastics Industry) (1999a) Definitions of Plastic Resins,
downloadable at http J/www, socplas. ore/
SPI (The Society of the Plastics Industry) (1999b) Plastics Processes, downloadable at
http://www.socplas.ore/industry/process.htm.
SPI (The Society of the Plastics Industry) (2002) Plastics in New Jersey, downloadable at
http://www.plasticsdatasource.ore/facts/ni.pdf
Wallace, W. (1971) The Logic of Science in Sociology, Aldine-Atherton, Chicago.
23
-------
Do Facilities With Distant Headquarters Pollute More?:
How civic Engagement Conditions the Environmental
Performance of Absentee Managed Plants
Don Grant
University of Arizona
Social scientists have long speculated that factories with distant headquarters are a threat
to communities and their physical environments. Over fifty years ago, when "war plants" were
being created outside the nation's industrial heartland, several researchers warned that absentee
managed plants are the "puppets of big business" and will exploit the social and natural
resources of their host communities. Today, as the winds of globalization and capital flight
disperse still more facilities across the landscape, researchers continue to express concerns about
the local impact of absentee managed plants. They worry that due to advances in transportation
and communication technologies, more corporations than ever before will be able to externalize
their pollution by setting up plants in far-flung, less regulated areas.
However, scholars have not analyzed the environmental performance of absentee
managed plants. Nor have they explored how their tendency to pollute varies by the types of
communities that harbor them. To begin to remedy this situation, I examine how the emissions
of absentee managed plants are conditioned by their host communities.
24
-------
This topic is an especially important one. Precisely because more companies can manage
operations from afar, absentee managed plants are rapidly becoming the modal type of industrial
organization. Hence, if they are an environmental threat, as some suggest, then a type of
organizational virus is spreading throughout the eco-system that demands analysis. There is also
a strong perception within the anti-globalization movement today that absentee managed plants
pollute more than locally managed ones. Critics of globalization assume further, like Mills, that
there is little communities themselves can do about this problem because their survival depends
on attracting and accommodating footloose plants.
I contend that critics' logic is simplistic and overlooks the potential impact of local civic
engagement on pollution outcomes. First, to suggest that communities are powerless to "outside
predators" ignores how responsibility for protecting the environment from globalization and
other forces has gradually devolved from the nation-state to the local level. A prime example of
this development in the U.S. is the Emergency Planning and Community Right-to-Know Act.
This act seeks to curb industrial pollution by requiring manufacturers to submit data on the
toxins they emit. It also makes states responsible for developing right-to-know programs that
disseminate this information to their citizens. The rationale behind this decentralized strategy is
that manufacturers are more apt to voluntarily reduce their future emissions when confronted by
an active and well-informed group of local citizens.
Critics also ignore recent studies on civil society that speak to the ability of communities
to root organizations in place and control their behavior through informal means. This research
suggests that while a community can do little to change the physical distance between itself and
an absentee managed plant's headquarters, it can reduce the social distance between itself and
25
-------
the plant by incorporating the latter in a dense network of local organizations where citizens
often meet and discuss community problems. Through their involvement in these organizations,
absentee managed plants may come to identify with their host community and work to maintain
its physical integrity.
In this paper, I seek to demonstrate that absentee managed plants emit fewer toxins when
embedded in communities that are more civically engaged. At the same time, I argue that right-
to-know proponents and civil society theorists espouse two different models of civic
engagement. Hence, I compare the effects of the local institutions that they allege facilitate the
civic engagement of pollution. Specifically, I contrast the effects of states' new right-to-know
programs with those of more traditional institutions — i.e., voluntary associations, churches, and
so-called third places (i.e., barber shops, cafes, and other informal sites of public life).
Until recently, it was impossible to determine the environmental effects of absentee
management or any other organizational form because there were no organizational data on
pollution. However, using the Environmental Protection Agency's Toxics Release Inventory -
the same data that facilities must submit under the Emergency Planning and Community Right-
to-Know Act - (using these data) I am able to test for the first time the effects of absentee
management on the emissions of chemical plants in the United States. I focus on the chemical
industry because it is responsible for a disproportionate share of toxic releases. Also, as a result
of plants continuing to migrate from the corporate centers of the Rustbelt region to Sunbelt
states, and other plants owned by foreign firms moving to the U.S. to seize new investment
opportunities, an unprecedented number of chemical plants in this country now have out-of-state
headquarters.
26
-------
To situate my study in its broader context, I next review competing perspectives on the
environmental performance of absentee managed plants offered by proponents and critics of
globalization (OVERHEAD). Proponents suggest that plants with remote headquarters often use
more efficient and cleaner technologies than locally managed ones. They argue that as
companies mature and develop standardized production processes, they decentralize their
branches to periphery regions to capture the efficiencies of their best input-saving technologies.
Proponents of globalization also contend that multi-locational firms typically have more uniform
operating procedures and greater resources to invest in environmental initiatives. They suggest
further that because environmental groups are eager to sue companies capable of paying large
settlements, the satellite plants of major corporations are under intense pressure from their
headquarters to manage their chemicals as effectively as possible and perhaps even overcomply
with regulations.
In contrast, critics of globalization argue that firms are increasing their power by
decentralizing production, a phenomenon Harrison describes as "concentration without
centralization." According to these scholars, firms often relocate plants to distant areas as a way
to avoid regulation and externalize their pollution. They suggest, therefore, that absentee
managed plants are among the dirtiest. Dependency researchers, for example, argue that the
maquiladoras created in the free trade zones of northern Mexico and other parts of Latin
America are especially poor environmental stewards. They predict that as international
competition for jobs intensifies, developing countries will feel pressure to create additional
"pollution havens" to attract plants. The same dynamic allegedly operates within the United
States, where many chemical firms have tried to flee costly regulations and fend off foreign
27
-------
competition by relocating plants in the "better business climates" of southern and southwestern
states. This strategy reinforces an already strong tendency among multi-locational businesses to
stress the exchange value of natural places over their potential use values, i.e., to treat them as
expendable commodities. Consistent with this reasoning, Davis (1992) finds that the owners of
chemical companies with multiple out-of-state plants are significantly less willing to sacrifice
production to meet environmental standards.
So, which group is right? FIGURES 1 and 2 provide a preliminary answer to this
question. They report how the emission1 levels of chemical plants with out-of-state2
headquarters compare with the emission levels of other chemical plants in the year 2000
according to the Toxics Release Inventory. Since the TRI's inception in the mid-1980s, the
number of industrial chemicals determined to be toxic and therefore tracked by the EPA has
more than doubled from 319 to 667. Figure 1 compares the amount of toxins released by
different plants using the EPA's original list of 319 or core chemicals. It shows that the average
emission level of plants with out-of-state headquarters (4.2 million toxic pounds) is
approximately 25% greater than the average emission level of other plants (3.4 million toxic
pounds). Figure 2 shows that when we use the expanded or current list of toxins, the differences
between the two plant types are even more pronounced. Absentee managed plants' average
emission level (9.9 million toxic pounds) is roughly 57% more than that of other plants (6.3
million toxic pounds). Hence, there is empirical support for critics' claim that toxic emissions
are concentrated in plants that are managed from afar.
Of course, it could be that absentee managed plants emit more toxins, on average, simply
because they use more toxic chemicals. That is, firms may be emboldened to process larger
28
-------
quantities of dangerous chemicals when they can do so from a safe distance. Hence, when one
takes into account the amount of toxins that plants have at their disposal, it may be that plants
with distant headquarters are no more prone to pollute than other plants. Even so, the total
amount of toxins emitted by plants is of paramount importance to local communities. Also,
while communities may be unable to set formal limits on how many toxins a plant processes,
communities can informally pressure a plant to minimize its toxic releases. Whether
communities can reduce the emissions of plants with the least attachment to place - absentee
managed ones - is the focus of this inquiry.
As suggested earlier, in assuming that communities are powerless to outside
organizations, critics ignore recent legislation designed to empower local citizens as regulators.
In particular, the Emergency Planning and Community Right-to-Know Act marks a significant
departure from traditional regulatory policy. Instead of specifying the pollution reduction
methods to be used by industry, which had been the practice under previous command-and-
control approaches, this act seeks to reduce industrial pollution by disclosing information on
manufacturers' pollution behavior. Specifically, it requires all states to establish a system of
Emergency Planning Committees, which are to take data on the hazardous materials used by
local manufacturers and make that information available to inquiring citizens. The assumption
underlying this "regulation through information" approach is that local residents will be able to
use pollution data to exert pressure on manufacturers to lower their emissions.
Critics also ignore recent research on civil society. This work suggests that businesses
rarely operate in a social vacuum. Rather they are subject to demands from several other kinds
of organizations, including their host communities. This research stresses the fact that
29
-------
communities have always possessed problem solving capacities and local institutions such as
churches, associations, and "third places" have long served as forums for civic engagement.
These institutions have thus helped to root actors to places and enhance the local quality of life.
While both literatures concur that civic engagement matters for the environment, they
disagree over the mechanisms involved in the civic engagement of pollution and whether civic
engagement is more relevant to some businesses than others. They also stress the importance of
different local institutions for facilitating civic engagement and reducing industrial pollution.
Indeed, I would argue that they subscribe to TWO DIFFERENT VERSIONS of the civic
engagement thesis.
Proponents of the regulation through information approach subscribe to what might be
called a strong version of the civic engagement model, which suggests that state-sponsored
institutions like right-to-know programs enable citizens to voice their grievances and organize
public protests against polluters. A strong model of civic engagement also assumes that because
manufacturers in general tend to pollute, right-to-know programs should improve the
environmental performance of plants regardless of their ownership status. Hence, it would
predict that civic engagement in the form of state-sponsored right-to-know programs lowers the
emissions of all plants.
In contrast, civil society scholars tend to subscribe to a qualified or weak version of the
civic engagement model that emphasizes how local institutions function to coopt certain types of
businesses. This model suggests that civic communities emerge out of local clusters of small,
locally owned and managed establishments. While not denying that large corporations can and
do operate in such communities, it contends that without a class of small business owners, the
30
-------
odds of establishing a civic community are considerably less. Thus, it argues there are two types
of communities or local economies, those largely organized by corporate capitalism and those by
community capitalism. Workers in the former tend to look outward to the global economy and
their allegiance lies more with the firm than the community. Workers and residents of the latter
look inward to the community since it is their primary source of support.
According to a qualified version of the civic engagement model, civic communities are
best understood as "problem solving" places and the local institutions where citizens most often
assemble and address community problems are voluntary associations, churches, and "third
places". From this perspective, civic institutions provide not so much a format for venting
grievances as they create venues for citizens to solve mutual problems like pollution. These
problems can be resolved amicably or though direct confrontation. The point is that the more
such problem solving places exist in a community, the better equipped a community will be to
solve problems that face it.
It follows that the problem solving capacity of local communities has special importance
for the environmental performance of absentee managed plants. Absentee managers have no
motive to behave in a socially and environmentally responsible fashion and therefore will pollute
if they can. Local managers would like to pollute but they do not feel they can because they
have more personal and material ties to their host community and are integrated in its structures.
Local institutions are important, then, because they smother absentee managed plants and their
managers with social pressure to behave appropriately in the absence of strong local connections.
Thus, they compensate for the lack of such ties. Hence, a weak version of the civic engagement
31
-------
model would predict that civic engagement in the form of associations, churches, and "third
place" lowers the emissions of just absentee managed plants.
While the idea that civic engagement can protect communities from all or a subset of
polluters is reassuring, SERIOUS DOUBTS nonetheless remain. First, arguments about the
environmental benefits of civic engagement stand in dire need of empirical analysis. In the few
cases where pollution outcomes have been examined at all, researchers have looked at emissions
at highly aggregated levels of analysis (e.g., industry, state, nation). Consequently, it is still
unknown whether states' right-to-know programs or the types of local institutions stressed by
civic society theorists have any real effect on the environmental performance of individual
plants. Nor has it been shown that such effects exist net of other plant characteristics that are
thought to increase pollution.
Second, some argue that while there are demonstrated benefits of civic engagement for
individuals (e.g., for finding jobs and avoiding crime), to suggest that civic engagement is also a
property of communities borders on circular reasoning. They note a common tendency among
researchers to examine positive outcomes, like low rates of crime, poverty, or pollution, and then
infer the existence of civic engagement from the same outcomes.
Finally, research on civic engagement has been criticized for ignoring how factors like
class and race may account for both civic engagement and its effects (Skocpol 1996, p. 25).
Our study seeks to address these concerns. First, we empirically model the impact of
civic engagement on chemical plants' emissions. In the process, we address Portes's complaints
about circularity by treating the factors that facilitate civic engagement as separate from their
effects. It is probably true that civic engagement cannot be exactly measured at the community
32
-------
level. However, we are able to examine the relationship between pollution and what scholars
claim are some of the institutions that facilitate civic engagement — i.e., states' right-to-know
programs, associations, churches, and "third places." Finally, we test these indicators of civic
engagement alongside measures of race and class that may explain the former's impact.
DATA AND METHODS
To determine whether the emissions of all or just absentee managed plants are
conditioned by civic engagement, a data set was constructed that incorporates measures of
chemical plants' toxic releases and predictors of those releases. The unit of analysis for this
study is the chemical plant and the data file consists of 1859 cases. Since it is at the site of
production that industrial toxins are usually emitted, and absentee management is an attribute of
individual plants, we focus on pollution outcomes at the plant level rather than the firm level.4
We conduct a cross-sectional analysis of emissions in 2000 because the remoteness of a plant's
headquarters is not likely to fluctuate much from one year to the next, nor is the civic
engagement of a plant's surrounding community.5 However, in other analyses not reported here,
we looked at emission outcomes for 1990 and found the results to be virtually the identical.
As TABLE 1 indicates, our dependent variable is taken from the EPA's Toxics Release
Inventory and is operationalized as the annual pounds of chemicals released on-site (weighted by
their toxicity). Plants with high scores on this measure are those with high emission levels. To
determine if the causes of emissions differ depending on whether one uses the EPA's original list
of chemicals or its more recent, expanded one, we conduct separate analyses of each. Because
33
-------
toxic emissions are highly skewed, we transform the dependent variable when conducting our
regression analyses by taking its natural logarithm.
One feature of the Toxics Release Inventory is that it lists for each plant a unique nine-
digit identifier assigned by the Dun and Bradstreet Company. This number, in conjunction with
the listed address of each plant, allows one to append to the TRI organizational data compiled by
Dun and Bradstreet on each plant, including whether it is absentee managed.
Absentee managed plant is coded as a dummy variable (l=yes) and defined as any
chemical facility whose headquarters is located out-of-state.6
Another key independent variable, right-to-know funding, is operationalized as the
number of years since a plant's state first funded its right-to-know program under the Emergency
Planning and Community Right-to-Know Act. Under this act, states are not provided any federal
money to create their right-to-know programs. Hence, our measure is designed to distinguish
states that have a real and long-term financial commitment to disseminating pollution
information from those that run more nominal programs. Importantly, we do not claim that this
variable measures actual citizen mobilization. Rather it is intended to capture the kind of local
regulatory environment in which a chemical plant now operates that supposedly empowers
citizens as regulators. According to the strong version of the civic engagement thesis, this
variable should vary inversely with the emissions of all chemical plants.
To test the weak version of the civic engagement thesis, we interact our measure of
absentee managed plant with three indicators of civic engagement institutions — (log) number of
associations, (log) number of churches, and (log) number of third places in a plant's county.
While none of these three indicators directly measures the mechanisms said to be involved in the
34
-------
civic engagement of pollution, they do gauge the presence of institutions said to facilitate social
connectedness and problem solving. Each of these three indicators is expected to have a
negative statistical interaction with absentee managed plant or reduce just the latter's emissions.
Our models also control for several other industrial, political, demographic, and
organizational factors that are summarized in Table 1 of your handout. We conduct analyses of
the determinants of emissions using a random effects model available in LIMDEP.
FINDINGS
TABLE 2 examines the determinants of chemical plants' emissions using the EPA's
original list of toxins or "core chemicals." Looking first at the controls in model 1, we see that
log emissions are significantly lower when plants specialize in soaps/detergents. Conversely, log
emissions are significantly higher when plants have more chemicals on-site and they and their
parent firm are large. Contrary to what one might expect, the race and class characteristics of a
plant's surrounding neighborhood are unrelated to the emission of core chemicals.9
Most importantly, we see that net of the various controls, absentee managed plant has no
significant direct effect on log emissions. Other analyses not reported here revealed that the
inclusion of log toxic chemicals on-site changed the effect of absentee managed plant from
positive and significant to non-significant. This suggests that absentee managed plants have
higher emission levels - as we saw earlier in Figure 1 - in large part because their potential for
emissions is so much greater.
Findings from model 1, therefore, suggest that critics and supporters of globalization are
both wrong — absentee management per se has neither a harmful nor a beneficial impact on
35
-------
environmental performance. Still, communities have a special stake in minimizing the emissions
of absentee managed plants precisely because the latter use such large quantities of chemicals.
This raises the question of whether certain types of communities are more successful than
others at lowering the emissions of absentee managed plants or the emissions of all plants.
Contrary to the expectations of the strong version of the civic engagement thesis, other results in
model 1 indicate that local right-to-know programs have no direct bearing on the log emissions
of all plants. Neither do log associations, log churches, and log third places. However, a
qualified version of the civic engagement model suggests that the latter three factors may still
condition the environmental performance of those plants with the weakest ties to communities —
absentee managed ones. In models 2 through 4, we explore this possibility by interacting
absentee managed plant, respectively, with log associations, log churches, and log third places.
Results indicate that the emissions of absentee managed plants are significantly lower when they
are located in counties with more associations (model 2), churches (model 3), and third places
(model 4).
In TABLE 3, we replicate our analysis of the determinants of emissions but this time
using the EPA's more comprehensive list of toxic chemicals. In model 1, we see, as before, that
plants have significantly lower emissions when they specialize in soap/detergents and higher
emissions when they process more chemicals, are large, and their parent firm is large.
Interestingly, when using the more recent, expanded list of chemicals, plants have significantly
higher emissions when located in poorer neighborhoods. While one cannot generalize from this
finding that poor neighborhoods are exposed to more absolute amounts of toxins, it does speak to
how class influences the emissions of plants and the possibility that as more chemicals are added
36
-------
to the EPA's list of toxins, the environmental dangers faced by poor communities will become
more obvious.
In model 1, we see once again that the effect of absentee managed plant is non-significant
when controlling for other relevant factors, in particular the amount of toxins that a plant uses
and stores on-site. The latter suggests that absentee managed plants release more toxins back in
Figure 2 because they have more toxins at their disposal. Indeed, on average, absentee managed
plants have on-site well over twice as many toxic chemicals than locally managed ones, 36
trillion toxic pounds compared to 14 trillion.
We also find support for the qualified, but not the strong, version of the civic engagement
thesis. Right-to-know programs exert no significant, direct effect on the emissions of chemical
plants, whereas the other three indicators of civic engagement have significant, negative
interactions with absentee managed plant, suggesting again that they lower the emissions of
plants with distant headquarters.
The negative sign of the interaction term in model 2 indicates that absentee managed
plants have significantly lower emissions when nested in counties with numerous associations.
In more substantive terms, as TABLE 4 shows, if there are no associations in a plant's county,
the absentee management effect is .186 (bx + (bxy)Z); if 10 associations, the effect is -.082; if 50
associations, the effect is -.271; and so on. This suggests that only a small number of
associations needs to be in place before absentee managed plants begin to reduce their emissions.
Table 4 provides similar statistics for the interaction effects of churches and third places.
Importantly, all of the interaction effects in Tables 2 and 3 hold after controlling for a
variety of industrial, political, socio-demographic, and organizational factors that might explain
37
-------
them. That associations, churches, and third places each reduce the emissions of absentee
managed plants speaks to how social connectedness in a variety of institutional forms benefits
communities' physical environments. In sum, findings in both Tables 2 and 3 support the
prediction of the weak version of the civic engagement model that absentee managed plants
pollute less when embedded in civically engaged communities.12
Before leaving these results, let me note that we conducted several other analyses to
determine whether states' right-to-know programs had any effect. We tested their interaction
with absentee managed plant, we controlled for prior 1990 emissions to see if they affected
changes in emissions, we looked at emissions in just 1990, we experimented with different
measures of states' right-to-know programs, and so on. In every instance, results indicated that
states' right-to-know programs have no significant direct or indirect bearing on the emissions of
chemical plants net of other factors. This non-finding is an especially important one, because, as
this FIGURE 3 indicates, if you were to just compare the average emissions of plants in states
that have and have not funded their right-to-know programs in every year, you would be misled
to believe that these programs actually make a difference.
CONCLUSION
To conclude, our findings are by no means the definitive word on absentee management
and its interaction with community structures. Our analysis, for example, says nothing about the
economic/environmental tradeoffs local communities sometimes make when deciding whether to
recruit absentee managed plants. We have only considered absentee management as it manifests
38
-------
itself within the United States and therefore cannot say how absentee managed chemical plants
might impact the environment in less developed nations.
As mentioned earlier, we also do not directly test the mechanisms involved in the civic
engagement of pollution. It could be that local institutions decrease the emissions of absentee
managed plants because they instill in them a greater sense of loyalty to their social and physical
surroundings. It may be that these institutions give citizens more opportunity to voice their
grievances. Or absentee managed plants with high emission levels may tend to exit or avoid
civically engaged communities.14 Until more detailed data become available, we have no way
of determining which of these possibilities is more true.
These caveats notwithstanding, our study makes several SIGNIFICANT
CONTRIBUTIONS.
First, in identifying which types of plants are most likely to pollute and under what
conditions, our results are of great practical value in that they should help inspectors, state
emergency response commissions, and the Chemical Manufacturers Association decide where to
allocate their resources.
Second, our findings cast doubt on the efficacy of environmental federalism and states'
right-to-know programs in particular. The fact that the latter have no effect on emissions in one
of the dirtiest industrial sectors - chemicals - is striking and raises the question "Is the more
decentralized regulatory environment in which polluters now operate real or illusory?"
Third, and on a more positive note, results suggest that although today's global economy
is dominated by mobile employers, industry rarely is all-powerful and community-based forms
of regulation are still viable. Communities possess problem solving capacities that can be
39
-------
activated to limit the destruction caused by businesses, especially those with the least attachment
to place. However, the kinds of local institutions that facilitate the civic engagement of pollution
may not be the ones that policymakers expect.
Fourth, our study suggests how organizational research might be advanced. It has
become fashionable for organizational scholars to use biomaterial metaphors like embeddedness
to describe and bound organizational properties. Yet, these terms say precious little about how
such properties, in turn, influence real biomaterial outcomes. By following our lead and
studying the environmental damage caused by nested organizational structures, researchers may
discover the ecological significance of concepts like embeddedness and what makes an
organization truly sustainable.
Finally, our empirical analysis greatly improves on past environmental studies that
merely speculate about the pollution effects of absentee management and other organizational
factors. By combining EPA data on facilities' emissions with information on their
characteristics and those of their host communities, we have pioneered, I believe, an exciting
possibility for secondary research. Our study also underscores the need to study organizations
where they most immediately impact the environment - the plant level.
There are no doubt other organizational forms besides absentee management that
influence plants' environmental performance. Indeed, the study presented here is but one part of
a larger project funded by the EPA that investigates the pollution effects of several
organizational forms. For instance, another organizational factor that we examined and
discovered increases emissions is whether a plant is a subsidiary. This is an important finding
because it speaks to the possibility that in allowing parent companies to create a "liability
40
-------
firewall" between themselves and their branches by reclassifying the latter as subsidiaries, the
1986 Tax Reform Act may have inadvertently encouraged parent firms to shift their most
environmentally dangerous production activities into subsidiaries.
So, there are other organizational forms that might be studied. Likewise, the pollution
outcomes studied here are not the only ones that can now be examined at the plant level. In
future research, I plan to examine the organizational and communal determinants of plants'
emissions using the EPA's newly released Risk-Screening Environmental Indicators or RSEI.
Unlike the pollution data used in this study that gauge simply the pounds of toxins released by
plants and their relative toxicity, RSEI data also take into account the degree to which people are
potentially exposed to chemicals and the estimated size of the exposed population. RSEI data
thus provide much more accurate measures of the potential risk-related impact of facilities on
chronic human health. Using these new data, I plan to investigate the possibility that right-to-
know programs may still reduce the most serious health-related emissions. I also plan to address
an important but underresearched question in the environmental justice literature, which is What
is it about the organization of hazardous facilities that explains why some pose a greater health
threat in poor, minority neighborhoods than others? In short, we're entering a new phase of
environmental and organizational research when several key empirical issues can finally be
addressed.
41
-------
that seeks to explain why some plants pollute more than others. For instance, another
organizational factor that we examined and discovered increases emissions is whether a plant is a
subsidiary. This is an important finding because it speaks to the possibility that in allowing
parent companies to create a "liability firewall" between themselves and their branches by
reclassifying the latter as subsidiaries, the 1986 Tax Reform Act may have inadvertently
encouraged parent firms to shift their most environmentally dangerous production activities into
subsidiaries. We have also begun examining whether chemical plants with the weakest local ties
- those owned by foreign companies - pollute more.
Finally, I hope
In future research I plan to gain further insight into the determinants of facility-level emissions
using the EPA's newly released Risk-Screening Environmental Indicators. Unlike the pollution
measures used in this study and others that gauge simply the pounds of toxins released by
facilities and their relative toxicity, RSEI data also take into account the degree to which people
are potentially exposed to chemicals and estimated size of the exposed population. Thus, RSEI
data provide much more accurate measures of the potential risk-related impact of facilities on
chronic human health. Using these new data, one can address an important but grossly
underresearched question in the environmental justice literature, which is What is it about the
organizational features of hazardous facilities in disadvantaged neighborhoods that explains why
some endanger human lives more than others. In short, we're entering a new phase of
environmental and organizational research when these and other issues can finally be examined
empirically. I'll stop on that forward looking point.
42
-------
Of course, the big issue is whether certain kinds of plants pose a greater health risk. The
measure of pollution used in this study only gauges the pounds of toxins released by a facility
and their relative toxicity. Fortunately, the EPA's newly released Risk-Screening Environmental
Indicators allows one to do this and more - i.e., it also takes into account the degree to which
people are potentially exposed to chemicals and the estimated size of the exposed population.
Using these data, I plan in future work to address a key yet underresearched question in the
environmental justice literature, which is what
What is it about the organizational features of hazardous facilities that explains why some in the
same disadvantaged neighborhood
pose a greater health risk to disadvantaged neigh
in disadvantaged neighborhoods that explains why some endanger human lives more than others.
Also, depending on how many of these institutions are present, a community may or may be able
to curb the emissions of plants with the weakest local ties.
Fourth,
First, its empirical analysis greatly improves on past studies by environmental and organizational
sociologists that merely speculate about the pollution effects of absentee management. By
43
-------
combining EPA data on facilities' emissions with information on their host communities, we
have empirically demonstrated for the first time that the spatial properties of plants have
important environmental consequences and the local conditions under which this is especially
true. Our study should also sensitize researchers to the need to study organizations where they
most immediately impact the environment - the facility level.
Second, our findings inform work on globalization and the spatialization of capital. Prior
research has noted how capital mobility can create new forms of locational concentration
(Sassen 1991) or "sticky spaces in slippery space" (Markusen 1996). Our study compliments
these studies by suggesting how local institutions help root absentee managed facilities in place
and minimize their environmental destruction. Likewise, our research resonates with recent
theorizing about the spatialization of the U.S. economy (Grant 1994; Brady and Wallace 2000)
and the "spatial decentralization" of production (Romo and Schwartz 1995). But whereas this
body of work stresses how footloose employers have severed their postwar accord with workers
and citizens, our study suggests that a new accord may be possible that is grounded in social
capital. This does not imply that a move toward a less capable and involved national
government is required for civic engagement to thrive, as conservatives have suggested. Nor
does it mean that translocal agents (e.g., NGOs, social movements, political parties) will not play
a role in creating livable places (see Evans 1997; Putnam 1993, p. 176). Rather, our results
suggest that in the present global period, viable compromises between employers and
workers/citizens might still be constructed at the local level. In light of the recent concerns
raised about the relevance and efficacy of civic engagement (Portes 1998; Skocpol 1996, p. 25),
this is promising news for communities within the U.S.
Finally, and perhaps most importantly, our study demonstrates that if scholars are to
study the impact organizations have on the environment (Perrow 1997), they must consider not
simply the characteristics of businesses but those of other organizations with which businesses
interact. As research on structural embeddedness and civil society suggests, communities are
also strong organizations and how they cultivate the problem solving capacity of their citizens
can strongly influence the behavior of external organizations like absentee managed plants.
While our study cannot say whether more amicable or contentious strategies work best with
absentee managers, it speaks to the more fundamental point that communities function as
problem solving places. Indeed, although today's global economy is dominated by mobile
employers, industry rarely is all-powerful. Communities possess organizational resources that
can be activated to limit the destruction caused by businesses, including those with the least
attachment to place.
44
-------
NOTES
1 Emissions, which are reported in pounds by the EPA, are weighted here by their toxicity (see
Grant, Jones, and Bergesen 2002 for details on toxicity weights).
2 In the context of this study, "out-of-state" is not meant as an indicator of globalization, but
absentee management.
3 Unlike many other pollutants, which are subject to strict safety standards, the Environmental
Protection Agency only requires manufacturers to report their toxic releases, leaving it up to
local communities to act on that information as they see fit.
4 Examining emissions at the firm level would also introduce several complications, since firms
may own plants in several industries with very different eco-organizational properties.
5 We explored the possibility of examining changes in emissions between 1990 and 2000, but
several factors discouraged us from doing so. In particular, because of changes in reporting
requirements and the fact that hundreds of new toxins have been added to the TRI list of
chemicals since 1990, the facilities included in the 1990 and 2000 Toxics Release Inventory are
often not the same. Indeed, a plant that processes the same chemical and in the same amount in
these two years, may be required to report information on emissions for just one of these years.
Importantly, we did replicate our 2000 analysis with 1990 data using the core list of chemicals
and found the results to be basically the same. Hence, although the chemical plants included in
the 1990 and 2000 Toxics Release Inventory may differ, the pattern of relationships between
emissions and other factors appear robust across the two time points.
6 Grant et al.'s (2002) analysis of 1990 data tested the effect of branch plants in general and
therefore did not isolate the pollution behavior of branches with out-of-state headquarters. By
distinguishing absentee managed plants from others, we are able to test the thesis advanced by
critics of globalization and capital migration that the spatial characteristics of plants have
important environmental consequences.
7 A related study examines the emission rates of foreign owned plants in the United States (Grant
and Jones forthcoming). It, however, focuses on a small subset of all absentee managed plants
and with 1990 data that excludes roughly half of the industrial toxins now tracked by the EPA.
Nor does it address the key question of this paper, which is whether the environmental
performance of absentee managed plants varies by the local civic cultures in which they are
embedded. Hence, it examines the effects of absentee management in a very preliminary
fashion.
8 The sources of these indicators are the Encyclopedia of Associations 2000 (Gale Research
Corp. 2000), Census of Churches (Association of Statisticians of American Religious Bodies
2002), and the County Business Patterns (U.S. Bureau of the Census 2002).
45
-------
9 We also considered the possibility that past environmental fines might influence emissions but
discovered that because less than .005% of plants had ever been penalized, this factor could not
be included in our models without creating severe problems of multi-collinearity.
10 In substantive terms, findings suggest that if there are no churches in a plant's county, the
absentee effect is .467 (bx + (bxy)Z); if 10 churches, the effect is . 115; if 50 churches, the effect is
-.131; if 100 churches, the effect is -.238; and if 1000 churches, the absentee effect is -.590 (the
sample range for churches is 2 to 4044). Results indicate that if there are no third places in a
plant's county, the absentee effect is .307 (bx + (bxy)Z); if 10 third places, the effect is .049; if 50
third places, the effect is -.130; if 100 third places, the effect is -.209; and if 1000 third places,
the absentee effect is -.467 (the sample range for third places is 0 to 12773).
11 Importantly, Tolbert et al. (1998) suggest that their indicators of civic engagement probably
underestimate the importance of local institutions that are older and have especially deep roots in
community.
12 We experimented with other specifications of the dependent variable such as expressing
emissions as a fraction of all chemicals on-site (log(emissions/chemicals on-site) and discovered
that the results mirrored those for log emissions.
13 For example, if one were to estimate simultaneously the determinants of emissions, the siting
of absentee-owned plants, and housing segregation (Hefland and Peyton 1999; see also Downey
2003), it might be found that race and ethnicity are significant predictors of emissions.
However, the type of longitudinal data needed for such a simultaneous equation are unavailable
or limited.
14 Although, to our knowledge, nowhere in the literature on industrial location has it been
suggested or shown that civic engagement actually influences the siting of chemical facilities.
Everyone is remarkably well preserved. You look just the way I remember you. And it's
equally great to meet others who been hired since and helped to take the dept., in many respects,
to a new level of excellence
questions: 1) Do absentee owned plants or plants with distant headquarters emit more toxins than
local managed ones? and 2)
My funded project seeks to address two basic questions: 1) how does the ownership
status of a regulated facility affects its environmental performance?, 2 ) what the implications of
this for the effectiveness of community-based forms of regulation? Today, I will be talking
about a portion of this larger project that looks at absentee owned plants and how their
environmental performance is conditioned by right-to-know programs, and other types of local
civic engagement.
46
-------
There are few human-made environmental problems that are not caused by or through
organizations (Perrow 1997; Clarke 1989). While individuals' lifestyles, consumption habits,
and so on contribute to environmental degradation (York, Rosa, and Dietz 2003), many, if not
most, pollutants are emitted at the site of production or have their source in industrial
organizations. And yet sociologists have rarely examined the impact that different
organizational forms have on pollution
Today, I will be talking about a portion of a larger project funded by the EPA that
examines how the ownership status of a facility affects its environmental performance and the
implications this has for community-based forms of regulation. Specifically, I will address two
questions: 1) Do absentee owned plants or those with distant headquarters pollute more? and 2)
What effect, if any, do right-to-know programs and other local channels of civic engagement
have on these plants' emissions?
This is true in developing countries where pollution is often unregulated by national
governments and local communities must therefore negotiate environmental standards with
manufacturers (Hartman, Huq, and Wheeler 1997). It is true as well in developed countries
where command-and-control approaches to regulating industrial toxins have been slowly
replaced by strategies that rely on the participation of local citizens (Ringquist 1995).
Actually, my talk today is essentially the same one I gave just 3 days ago in D.C. at the EPA's
Conference on .
We also show that states' new right-to-know programs have no effect on the emissions of
absentee managed plants. Rather their environmental performance depends on the presence of
other local institutions that have traditionally facilitated civic engagement, namely churches,
voluntary associations, and so-called third places.
[The study I will present today is part of a larger research project funded by the Environmental
Protection Agency on the organizational determinants of pollution and effectiveness of
community-based forms of regulations. Specifically, my talk addresses two questions: 1) Do
plants with distant headquarters pollute more? and 2) Can civically engaged communities do
anything about it? Is the environmental performance of these absentee owned plants conditioned
by the civic engagement of their host communities?
47
-------
Social scientists have long speculated that ]
The study I will present today is part of a larger research project funded by the
Environmental Protection Agency on the organizational determinants of pollution and
effectiveness of community-based forms of regulations. Specifically, my talk addresses two
questions: 1) Do plants with distant headquarters pollute more? and 2) How is the environmental
performance of these absentee owned plants conditioned by the civic engagement of their host
communities?
In addition, there is a substantial body of empirical research that suggests absentee managed
plants influence social outcomes, including poverty, infant mortality, industrial conflict, and
underdevelopment. Whether absentee managed plants also impact environmental outcomes has
yet to be determined.
***use later I report findings using the 2000 edition of the Toxic Release Inventory both because
it is the most current and it covers more than twice as many industrial toxins than earlier
editions. We focus on the U.S. for reasons of data availability and because the spatial
restructuring of production has been especially great in this country during the global era. As a
result of factories migrating from the corporate centers of the Rustbelt region to the "better
business climates" of the Sunbelt states in response to global competition and other plants
operated by foreign firms moving to the U.S. to seize new investment opportunities, an
unprecedented number of plants in the U.S. are now absentee managed.
***
We study the effects of absentee management at the facility (as opposed to firm) level because
industrial toxins are emitted at specific production sites and the environmental performance of
individual facilities is of more immediate concern to local communities.
*** In conclusion where I discuss practical relevance, mention World Bank
**Write long version for OSU, then whittle it down for EPA (*where EPA talk begins, etc.)
In conclusion, note that this study was part of a larger project funded by EPA (I also examined
other organizational forms)
** If this sounds like a talk you might give to policymakers, that is because it is.
48
-------
** in other analyses, we examined whether plants with the most distant headquarters pollute
more, but found that not to be the case.
Particularly relevant to our study, scholars at the World Bank have begun exploring how
civic engagement affects the emissions of individual facilities (Hartman, Huq, and Wheeler
1997; Pargul and Wheeler 1995; Pargal, Hettige, Singh, and Wheeler 2002). They contend that
in developing countries, where formal regulation (e.g., uniform air quality standards, mandated
pollution technologies) tends to be weak or non-existent, informal regulation exercised by
communities (e.g. public appeals, protests) may strongly influence corporate environmental
performance. They speculate that civic engagement may also influence certain types of
corporate pollution in the U.S. that are largely unregulated, such as toxins released by
manufacturers.3
In short, a growing body of research suggests that communities can improve the
environmental performance of manufacturing plants by reducing the social distance between
themselves and plants. According to this work, unless plants develop social ties to their host
communities, they are unlikely to participate in public conversations about local environmental
priorities. However, where there are numerous institutional settings that allow residents and
plant managers to meet and develop a common appreciation of place, plants are more likely to
participate in public conversations about the environment and curb their emissions.
It may also be that because pollution data are self-reported, the EPA needs to use better
quality control measures. If more intentional and unintentional mistakes made in submitting
information are caught and corrected, the effectiveness of right-to-know programs might be more
apparent. There is also the possibility that existing pollution data are basically sound but how
they are processed and interpreted by intermediaries, such as interest groups, varies widely.
Along these lines, other studies report that the goals of regional environmental groups and local
citizens often conflict.
As I explained to EPA officials and other policymakers at a conference in D.C. just three days
ago, it might still be the case that states' right-to-know programs work in other sectors of the
economy than the one studied here. Nonetheless, the fact that such programs cannot explain
emission decreases in one of the dirtiest sectors - the chemical industry - is striking. It begs the
question of what might explain recent reductions in emissions if not states' right-to-know
policies? It could be, as some industry spokespersons suggest, that most of these improvements
were the result of businesses themselves taking the initiative in devising environmental
solutions. However, empirical support for this claim is thin and limited to qualitative studies of a
small, select set of chemical companies (Baram et al. 1990). Another possibility is that changes
49
-------
in emissions are due to the efforts of national actors - e.g., the news media and organizations like
Environmental Defense (see U.S. Environmental Protection Agency 2003). Previous research
has shown, for example, that when national news media report the emissions of major
companies, it can cause the value of their stocks to drop (Hamilton 1995). Whether companies
respond to stock market declines by improving their environmental performance, though, has
still to be determined. These and other possibilities need to be examined more systematically to
determine whether the more decentralized regulatory environment in which polluters now
operate is real or illusory.
findings suggest that community-based forms of regulation may still be viable in an age of
globalization.
though the local institutions that facilitate the civic engagement of pollution may not be the ones
policymakers expect. Nor do they influence the emissions of all plants.
Fourth, our study demonstrates that if scholars are to study the impact organizations have
on the environment (Perrow 1997), they must consider not simply the characteristics of
businesses but those of other organizations with which businesses interact. As research on
structural embeddedness and civil society suggests, communities are also strong organizations
and how they cultivate the problem solving capacity of their citizens can strongly influence the
behavior of external organizations like absentee managed plants. While our study cannot say
whether more amicable or contentious strategies work best with absentee managers, it speaks to
the more fundamental point that communities function as problem solving places. Indeed,
although today's global economy is dominated by mobile employers, industry rarely is all-
powerful. Communities possess organizational resources that can be activated to limit the
destruction caused by businesses, including those with the least attachment to place.
There is evidence consistent with each of these arguments. On the one hand, several
studies document the success of states' right-to-know programs in educating their citizens and
providing them with technical know-how needed to interpret and act on complicated pollution
information. Others show that, on average, total emissions tend to be lower in states with more
aggressive right-to-know programs. And the EPA reports that, for the nation as a whole, total
pounds of on-site emissions have decreased by 56.6% since the Emergency Planning and
Community Right-to-Know Act was passed (U.S. Environmental Protection Agency 2002).
On the other hand, several studies document how residents and business leaders
become integrated in communities through their participation in volunteer associations,
churches, and third places and how such civic engagement translates into lower rates of
unemployment, poverty, and crime. Others studies suggest that these effects are especially
strong in communities with many absentee managed businesses. And still others have
50
-------
demonstrated how corporate leaders, especially managers of satellite plants, can be persuaded to
contribute to local environmental projects through their involvement in local religious and
voluntary organizations.
Our goal in this paper was to advance our understanding of the environmental
degradation caused by different organizational forms. Toward that end, we analyzed the effects
of absentee management on chemical plants' environmental performance using the EPA's 2000
Toxics Release Inventory. Findings confirm the suspicion of critics of globalization that
absentee managed plants emit greater amounts of toxins. However, results also indicate this is
largely because absentee managed plants process substantially more chemicals. In fact, when
we take into account the amount of chemicals that plants have on-site and other factors that
influence facilities' emissions, we discover that the environmental performance of absentee
managed plants is no worse than that of other plants. Whether plants with distant headquarters
emit more chemicals largely depends on the presence of local institutions that facilitate civic
engagement. Specifically, when embedded in communities with more associations, churches,
and third places, absentee managed plants emit significantly fewer toxins.
51
-------
Compliance and Beyond: Strategic Government-Industry Interactions in Environmental
Policy and Performance - The Role of Technical Information in Reducing Automobile
Emissions
Jennie C. Stephens and Edward A. Parson
Paper presented at the EPA Corporate Environmental Behavior Research Workshop,
April 26-27, 2004
1. Introduction
Technical knowledge and associated uncertainty in technical feasibility play a
critical role in government industry interactions during the development and
implementation of environmental policy and regulation. Improvements in environmental
performance are dependent on making technical changes to an industry's processes or
products. While government takes actions to promote environmentally beneficial
technological change to reduce industry's environmental impact, those targeted industries
are generally reluctant to make technical changes unless they perceive an associated
competitive advantage. Within this government-industry relationship characterized by
this conflicting basic interest, technical information and its associated uncertainty are
often integral to strategic interactive behavior. Firms identified as potential targets of
regulation, either acting individually or cooperatively through industry associations, often
use technical information as they seek to oppose, influence, or delay (and occasionally
promote) environmental regulations. Governments seeking to formulate, enact and
implement socially beneficial environmental policies must attempt to understand
technological details and feasibility of technical alternatives although they often have
limited independent information.
The design and implementation of government regulation to encourage
technological change for environmental improvement involves a dynamic process
whereby regulators and industry representatives interact and respond to each other (Yao
1988). While much of the literature examining the influence of government regulation on
technology development provides useful insights on relative effectiveness of different
regulatory mechanisms on innovative behavior (Kagan 1977; Ashford 1993; Kemp 1997;
Jaffe, Newell et al. 2000; Taylor, Rubin et al. 2003), the complexities of industry-
government interactions surrounding uncertainty associated with technological feasibility
are often omitted at this scale of analysis. Recognizing the critical role that perceptions
of technical feasibility of new technologies plays in both industry's attempts to influence
government decisions and government's attempts to influence industry's decisions, this
research focuses on the detailed interactions related to knowledge, uncertainty and
technical details.
This paper explores the role of technical information in government-industry
interactions in the fifty-year history of efforts to reduce automobile emissions. By
simultaneously focusing on the strategic behavior of both the automobile manufacturing
industry and the U.S. government, we are working toward identifying resultant
characteristic patterns of outcomes that arise from the interactions. This case is one of six
case studies, chosen to represent diversity in the targeted industry, government programs,
historical time, pollutants and geographic relevance, that will be included in the final
52
-------
product of this research effort, a forthcoming book edited by Parsons and Stephens. The
other five case studies included in this book explore strategic interactions between
government and industry with respect to technical information during efforts to reduce: 1)
dioxin in the pulp and paper industry, 2) perfluorocarbons in the aluminum
manufacturing industry, 3) chlorofluorocarbons in the chemical industry, 4) methyl
bromide in the agricultural strawberry and tomato industry, and 5) workplace exposure to
vinyl chloride in the vinyl chloride industry.
This paper will first describe the historical details associated with government-
industry interactions during each of the three time periods. A discussion of the
unproductive cycle of mistrust that has developed over the years between the government
and the industry is followed by discussions on the implications of cooperation versus
competition within an industry and the critical role of third parties, and finally some
concluding recommendations for policymakers that can be drawn from this case.
2. Three distinct Time Periods of Government-Industry Interaction
The history of the government's attempts to encourage the U.S. automobile
manufacturers' to develop and implement technologies to reduce automobile emissions
provides a particularly interesting perspective to improve understanding of government-
industry interactions regarding technological information exchange because the history
can be divided into three time periods with distinctly different industry-government
relationships defining strategies of interaction (Figure 1). During the earliest period,
before 1970, the federal government had minimal influence over the industry, and the
industry resisted technological change primarily through an industry-wide cooperative
agreement that removed competitive incentive to develop or implement pollution control
technologies. During an intermediate period, after the passage of the unprecedented
technology-forcing 1970 Clean Air Act Amendment (CAAA), competitive incentive
among individual firms was restored and the industry was forced by the government to
develop technology to meet specific emission standards in a predetermined (but
subsequently and repeatedly extended) amount of time. During the most recent time
period, from the debate preceding the 1990 CAAA until now, a more complex and less
intense industry-government relationship has developed as many more actors have
become involved in the more complicated technical and regulatory details. While
industry resistance to technological change is clearly evident in all three time-periods, the
industry strategies associated with this resistance have co-evolved with the changing
regulatory framework and the changing industry-government relationship.
This paper reviews the empirical history and then highlights the most interesting
observations about government-industry interactions within this history, while the full
chronological details of government-industry interactions during these three time periods
are described in more details elsewhere (Stephens 2004; Parson and Stephens
Forthcoming).
2.1 The Early Years: Minimal Government Involvement, 1955-1970
When the automobile was first implicated as a major contributor to the urban air
pollution problem in the early 1950s (Haagen-Smit 1952), the U.S. automobile
manufacturers responded by creating an industry-wide, cooperative agreement which
eliminated competition among individual firms to develop pollution control technology.
This cooperative approach also severely restricted third party inventions through a cross-
53
-------
licensing agreement that specified royalty-free exchange of patents and a formula for
sharing the costs of acquiring patents developed outside the industry; by removing
incentive for industry consideration of third party inventions, the industry eliminated any
potential market for third party innovators. Individual firms signed on to this cooperative
agreement because it minimized the risk to them that another firm would gain
competitive advantage by being the first to develop commercially viable pollution control
technology (DOJ 1971). The following excerpt from the minutes of an April 1955
meeting of the patent Committee of the industry trade group, the Automobile
Manufacturers Association (AMA) explains this strategy.
"No one company should be in a position to capitalize upon or obtain competitive
advantage over the other companies in the industry as a result of its solution to
this problem. " (DOJ 1971)
Despite the industry's public declaration that their cooperative program was
designed to accelerate technical developments in emissions reduction, the opposite effect,
to slow-down technical progress, has been identified as the intended result of the program
by evidence collected during a grand jury investigation assessing antitrust collusion
allegations against the industry in the late 1960s (DOJ 1971). Following this
investigation, an antitrust civil suit alleging 16 years of industry conspiracy to prevent
development of pollution control technology was issued; the case was settled by consent
decree in which the industry did not admit to any illegal activity but did agree to a series
of restrictions prohibiting the exchange of restricted technical information, prohibiting
the issuing of joint announcements, and requiring open access to existing patents and
technical reports to third parties, those outside the industry (1969).
During this early period of coordinated industry resistance, the emissions problem
was perceived by the automobile industry as a management, public relations challenge,
rather than a fundamental problem for which a technical solution had to be developed.
This perception allowed the industry to successfully resist making changes by controlling
the pace of technological development. The industry did slowly implement several
simple technical solutions in response to public pressure and regulatory threats urging
them to develop a technological response to the air pollution problem. One example of
this is the industry's installation in 1963 of a simple valve that allowed for recirculation
rather than direct release of pollutants from the crankcase; this positive crankcase
ventilation (PCV) valve had been used in military vehicles for decades so it was not a
new technology, yet the industry presented the development as a result of their diligent
efforts to find technical solutions to the automobile emissions problem (DOJ 1971).
During this early period when the federal government had minimal influence over
or interaction with the automobile manufacturers, the California state government began
addressing the industry's resistance by encouraging the development of pollution control
technology through state regulation. Recognizing the industry's slow pace of
development and implementation of technological improvements, California passed
legislation in 1960 that was designed to stimulate competition within and outside the
industry and provide a mechanism for government regulators to review the subsequent
technical developments. The Motor Vehicle Pollution Control Act (MVPCA) set strict
emission standards, a 70% reduction in HC and a 57% reduction in CO, that were to be
54
-------
enforced one year after two satisfactory emission control devices were certified by the
state to meet the standards; all new cars would have to install one of the certified devices
(California 1960). In drafting this legislation, California legislators identified and
addressed two critical mechanisms with potential to accelerate the pace of emissions
control technology development and reduce the effectiveness of the industry's attempts to
resist change: 1) the competitive pressure of third-party innovators, and 2) the asymmetry
of information between regulators and the industry. The MVPCA was designed to reduce
barriers to market entry of those developing pollution control technology external to the
automobile industry by creating incentive by ensuring a market for devices certified to
meet the standards (CAMVPCB 1965; Krier and Ursin 1977). At this time catalyst
technology was a suspected possible technology, so in response to this legislation a
period of intense catalytic research began as many catalytic chemists jumped at the
opportunity to work toward this exciting potential application of catalytic technology
(Lester 1983; Briggs 1984). By requiring a detailed state certification of all devices, the
legislation also created a pathway for information sharing; in the certification process the
state regulators gained the opportunity to evaluate the potential of different technical
approaches developed.
In 1964, when four externally developed devices (three of which were based on
catalytic technology) were certified by the state triggering enforcement of the emission
standards the following year, the automobile manufacturers revealed to the state their
own internally developed technical changes, which consisted of a series of engine
modifications rather than catalytic technology. Once the state certified these industry
developed engine modifications, the automobile companies each chose to implement
their own internally designed approaches rather than implement the externally developed
catalytic devices (Krier and Ursin 1977). So although this legislation motivated and
encouraged third party inventors, the inventions were excluded from implementation
because the potential market was removed once the state certified the industry developed
technologies. Nevertheless, the technological progress that was made in the early 1960s
by third parties was influential in demonstrating to both the industry and the government
the potential of catalytic technology; this potential was incorporated into the 1970 federal
legislation discussed in the next section.
Engineers involved in the development of the catalytic technology have suggested
that if cooperative relationships among industry, government and third parties had
existed, an effective combination of engine modifications and catalyst systems could
have resulted in an efficient pollution control technology that could have surpassed the
California standards by the mid-1960s (Briggs 1984). Instead the industry's engine
modifications approach with a limited level of reductions prevailed while the catalytic
technology with a far greater potential level of reductions was not developed for
implementation until the industry had to respond to the more stringent federal regulations
issued in 1970.
2.2 Industry Resistance within a New, Stringent Regulatory Regime: 1970 - 1988
Reacting to deteriorating urban air quality in many parts of the country and the
industry's apparent reluctance to make voluntary technical changes to reduce automobile
emissions, a frustrated federal government responded with an unprecedented, stringent
technology-forcing set of regulations in 1970. The 1970 Clean Air Act Amendments
55
-------
(CAAA) mandated emission reductions for hydrocarbons (HC) and carbon monoxide
(CO) of 90% below 1970 levels by 1975, and for nitrous oxides (NOx) 90% below 1971
levels by 1976 (U.S. 1970). These standards were more stringent than the California
standards discussed in the previous section, which required a 70 and 57% reduction
respectively for HC and CO and did not include a NOx standard. Responding to the
federal government's aggressive regulatory attempt to accelerate the development of
pollution control devices, the auto manufacturers intensified their research efforts while
simultaneously intensifying their resistance, highlighting potentially critical technical
uncertainties in their claims that the standards could not be met in the designated time.
While the auto manufacturers took every opportunity to weaken and delay the
standards throughout the 1970s, third parties, those external to both the U.S. government
and the U.S. automobile manufacturing industry, played a critical role in reducing the
effectiveness of the industry's strategies to resist change. To appease industry's concern
about the technical feasibility of meeting the strict standards set in the 1970 CAAA, the
legislation included two flexibility mechanisms: 1) automakers were allowed to apply for
a one-time, one-year extension if they could demonstrate to the Environmental Protection
Agency (EPA) administrator that the technology was not yet available, and 2) upholding
the standards was contingent on the assessment of technical feasibility to be carried out
by the National Academy of Science (NAS). These flexibility mechanisms provided two
different avenues for third parties to influence the industry-government dialogue.
During the extension hearings the testimony of independent companies
developing pollution control technology influenced both the industry and the
government. Independent suppliers of catalytic converters (the primary technology
considered capable of meeting the standards), provided manufacturer-conflicting
testimony to federal regulators about the feasibility of implementing the new technology
during the 1972 and 1973 hearings to consider whether or not the industry deserved an
extension to meet the standards (EPA 1972). Additionally, technological developments
made by several non-U.S. auto manufacturers provided regulators with a more optimistic
perspective on the technological possibilities of reducing emissions than the one
promoted by the U.S. automakers. Specifically Honda developed an alternative engine
design (a stratified charge engine) that could meet the 1975 standards without a catalytic
converter (Abernathy and Ronan 1978). Although initially the EPA denied the
industry's request for an extension in 1972, the U.S. court of Appeals ordered the EPA to
reconsider the automakers' request in 1973 and this time the one-year extension was
granted. Although arguments of technical infeasibility were used in the 1972 hearings,
the 1973 hearings focused more attention on the potential business catastrophe that could
result if insufficient time was allowed for the transition to the new catalytic technology.
The second flexibility mechanism, the stipulation that upholding the standards
was contingent on the NAS assessment of technical feasibility, incorporated another way
for an independent entity, a third party, to influence industry-government interactions and
reduce industry's resistance. The 1973 NAS report was extensively researched, and the
auto manufacturers were required by law to respond in full to any requests for
information of any kind from the NAS committee members (Lester 2003). The reports
major conclusions were that the 1975 HC and CO standards could be met in the given
time frame, but to meet the 1976 NOx standard additional time would be required (NAS
1973). The report reviewed technical obstacles to successful implementation of catalytic
56
-------
converters questioning whether the technology could be optimized by the 1975 model-
year. A major quandary noted within the report was the recent development of Honda's
stratified charge engine that could meet the 1975 standards with potential to meet the
1976 NOx standards too. This report provided regulators with an independent technical
assessment of the feasibility of meeting the standards and also created a common-ground
base of information to which the government and industry could both refer to in future
debates.
In addition to Honda's technical developments minimizing the U.S. industry's
claims of technical feasibility, other foreign automobile manufacturers were also
influential. Although technical progress was being made in the early 1970s with the
development of a catalytic converter that could successfully oxidize HC and CO, a major
unresolved technical challenge was whether an effective device that could simultaneously
reduce NOx could be developed. As it became clear in 1976 that the auto manufacturers
were not going to meet the standards scheduled to come into effect for the 1978 model
year an additional set of amendments to the CAA were debated. Initially during this
debate, the industry emphasized the uncertainties, infeasibility, and potential drawbacks
related to the development of a catalytic device that could successfully reduce all three
regulated pollutants (HC, CO, and NOx), a so-called three-way converter. In mid-1976,
however, Volvo produced a car to be sold in California with a three-way catalytic device
able to meet all three standards. In response, the U.S. manufacturers shifted their
resistant arguments away from claims of technical infeasibility focusing instead on the
economic uncertainty of implementation and the technical challenges associated with
scaling-up production; ensuring effective and safe catalytic converters on every new car,
they argued, would require more time. The industry lobbying efforts were successful in
preventing agreement in Congress on what revisions should include, so the actual
amendments to the CAA were not passed until August 1977, when the automakers were
already shipping out to the dealers their 1978 models which did not meet the current
standards. The 1977 CAAA delayed the HC standard until 1980, the CO standard until
1981, and weakened and delayed the NOx standard to come into effect in 1981 (1977).
In the years following, the three-way catalytic converter was improved upon and
became standard on most U.S. cars by 1980. Throughout the 1980s, the automobile
industry's concern about emission control regulation reduced as President Reagan's
administration demonstrated interest in weakening rather than strengthening pollution
control, and Congress was deeply divided on the issue. In 1988 this situation changed as
it became clear that additional changes to the air pollution legislation were necessary and
inevitable (Cohen 1995).
2.3 A Mature Industry-Government Relationship: 1988 to present
Despite the success in the development of the three-way catalytic converter, air
pollution continued to be a growing problem throughout the 1980s due to the increasing
number of cars on the road (Taylor 1987). Following a decade of inaction on air
pollution legislation, Congressional action began to be seriously debated in 1988 building
upon proposals developed by a few key Congressmen during the 1980s (Bailey 1998).
President George H.W. Bush, recognizing the political collateral associated with being
the President who updated and strengthened the CAA, placed passing new air pollution
legislation high on his priority list (Cohen 1995). Further reductions in automobile
57
-------
emissions was only one of three main goals of the 1990 legislation; the other two goals
were to reduce acid rain by cutting sulphur dioxide and nitrogen oxides emissions and
reduce emissions of air toxics by mandating control technology. With regard to
automobile emissions, the 1990 legislation mandated phasing in the California emission
standards at the national level in all new cars starting in 1994. These standards, known as
Tier 1 standards, required a reduction in NOx emissions from 1990 levels of 60% and a
reduction in HC of 40%. A second round of standards, known as Tier 2 standards, would
further reduce emissions by 50% from 2003 to 2006, unless an EPA review found that
these more stringent standards were infeasible or unnecessary (NESCAUM 2000). This
potential flexibility in the Tier 2 standards resulted from a compromise measure to
appease those concerned about the impacts of the legislation on the automobile industry;
by providing a future opportunity to resist the most stringent standards the legislation was
more acceptable.
During the debate surrounding the 1990 CAAA standards, the automobile
industry once again pointed out the uncertainties that the technology required to meet the
emission standards being considered could be developed and implemented (Anonymous
2004). Claims of infeasibility were made although the Tier 1 standards were already
being met successfully in California cars. The industry also predicted other negative
consequences of making the technical changes necessary to meet the new standards,
including reduced fuel economy, higher costs to consumers, reduced drivability, and
more recalls (Doyle 2000). The costs associated with the new standards, they argued,
were far greater than the associated benefits. Although claims of shutdown of the
industry if the standards were upheld like those used in the 1970s were not made, the
industry predicted job losses and an economic downturn would result from the strict
standards (NESCAUM 2000).
Unlike the 1970 and 1977 CAAA, the 1990 amendments were a high priority for
the President; President Bush, after having declared himself the environmental president,
was determined to reinforce the nation's air pollution laws (Cohen 1995). In this context,
the industry's resistant claims were not as effective in influencing the regulation as they
had been in the 1970s. Additionally, the familiarity of the industry's resistant claims to
similar claims made in the 1970s that did not materialize weakened the industry's
legitimacy and associated level of concern about how stricter standards would impact the
industry. And again, despite their claims to the contrary before the regulation was in
place, the automobile manufacturers have been able to successfully produce and sell cars
that are in compliance with Tier 1 and the subsequently determined Tier 2 standards
without any major associated negative consequences (Anonymous 2004).
During this most recent time period, a higher level of complexity compared to the
1970s in the regulatory process, the legislation, and the implementation of the legislation
compared to the 1970s has diffused the intensity of interactions between the automobile
industry and the government. A much larger number of politicians and industry lobbyists
were integrally involved in drafting the 1990 legislation, and far more government
beaurocrats and industry representatives have had to focus on the implementation of the
new legislation (Cohen 1995). This increased complexity has complicated the
relationship between the auto industry and the government and minimized the influence
of the industry's strategies of resistance on government decision-making.
58
-------
One prominent example of this complexity is the involvement of the oil industry.
In addition to setting stricter national emission standards for automobiles, the 1990
CAAA also mandated changes to fuel used in automobiles to reduce air pollution. The
debate on reformulated gasoline and alternative fuels engaged the oil industry rather than
the automobile industry, and this involvement of another large, mature industry in the
regulatory debate lessened the intensity of the automobile industry's interactions with the
government considerably.
This effect of complexity of players limiting the effectiveness of the industry's
strategies to resist making changes can be explained in the context of third parties. The
oil industry, in this context, can be viewed as a third party that has altered the intense
dynamic between the automobile industry and the government. Once again in this most
recent time period, third parties have played a critical role in reducing the effectiveness of
industry's strategies of resistance.
3 Analysis of Government-Industry Interactions in This Case
3.1 Arguments of Technical Feasibility - A Cycle of Mistrust
Throughout the fifty year history (1955-present) of efforts to reduce automobile
emissions, arguments of technical feasibility have recurred. A reinforcing cycle of
mistrust associated with technical details in the industry-government relationship has
developed encouraging industry to persistently make claims of technical infeasibility.
Following an initial 15 years of minimal industry action in reducing automobile
emissions from 1955-1969, the U.S. government became skeptical of the sincerity and
level of commitment of the industry's efforts to find technical solutions and responded in
1970 by mandating drastic emission reductions that could not be achieved with current
technology. The government's setting these stringent standards, an action that has been
described as motivated more by political considerations than technical realities (Ingram
1978; Lundqvist 1980), created an intense hostility and perpetuated a cycle of mistrust on
technical details between the government and the industry.
The industry, knowing that the government developed the emission standards
without confidence in technical feasibility, felt obliged to highlight the uncertainty in
feasibility and the strong likelihood that the standards could not be met. Because the
industry felt that the government developed the standards without understanding potential
technologies, the industry's internal processes, or the costs of implementing changes, the
industry has consistently made claims of technical infeasibility often based upon the most
extreme, pessimistic possibilities. Increasingly throughout this history, government
regulators recognized the industry's tendency to be pessimistic about future technology,
so the regulators have come to view industry's perspective on technical feasibility with
skepticism and have continued to uphold and enforce standards that the industry has
claimed cannot be met. This mutual mistrust still persists today, although due to the
familiarity resulting from the longevity of the industry-government relationship both
industry and government are now better able to interpret each other's actions and claims,
i.e. in the most recent debates the industry's claims of technical infeasibility have not
been taken seriously.
An additional factor feeding into this cycle of mistrust is the industry's apparent
perception that their public position on technical feasibility must emphasize the technical
uncertainties and potential obstacles in order to counteract the overly optimistic claims of
59
-------
technical feasibility being publicized by third parties, including environmental advocacy
groups or pollution control technology companies hoping to develop a market for their
product (i.e. catalytic converter manufacturers). The industry, predicting that regulators
would reconcile differing perceptions of technical feasibility by averaging the most
extreme views, has attempted to offset the optimistic claims of others by claiming a
position that is as far on the other end of the spectrum of potential feasibility as possible.
Despite this cycle of mistrust and the hostile government-industry relationship,
automobile emissions have been reduced immensely. Whether effective pollution control
technology could have been developed sooner or more easily with a different, less hostile
type of government-industry relationship is debatable.
3.2 Cooperation vs. Competition within the Industry
When the automobile was first implicated as a major contributor to the urban air
pollution problem in the early 1950s (Haagen-Smit 1952), the U.S. automobile
manufacturers responded by creating an industry-wide, cooperative agreement which
eliminated competition among individual firms to develop pollution control technology.
This cooperative approach also severely restricted third party inventions through a cross-
licensing agreement that specified royalty-free exchange of patents and a formula for
sharing the costs of acquiring patents developed outside the industry; by removing
incentive for industry consideration of third party inventions, the industry eliminated any
potential market for third party innovators. Individual firms signed on to this cooperative
agreement because it minimized the risk to them that another firm would gain
competitive advantage by being the first to develop commercially viable pollution control
technology (DOJ 1971). The following excerpt from the minutes of an April 1955
meeting of the patent Committee of the industry trade group, the Automobile
Manufacturers Association (AMA) explains this strategy.
"No one company should be in a position to capitalize upon or obtain competitive
advantage over the other companies in the industry as a result of its solution to
this problem. " (DOJ 1971)
Despite the industry's public declaration that their cooperative program was
designed to accelerate technical developments in emissions reduction, the opposite effect,
to slow-down technical progress, has been identified as the intended result of the program
by evidence collected during a grand jury investigation assessing antitrust collusion
allegations against the industry in the late 1960s (DOJ 1971). Following this
investigation, an antitrust civil suit alleging 16 years of industry conspiracy to prevent
development of pollution control technology was issued; the case was settled by consent
decree in which the industry did not admit to any illegal activity but did agree to a series
of restrictions prohibiting the exchange of restricted technical information, prohibiting
the issuing of joint announcements, and requiring open access to existing patents and
technical reports to third parties, those outside the industry (1969).
During this early period of coordinated industry resistance, the emissions problem
was perceived by the automobile industry as a management, public relations challenge,
rather than a fundamental problem for which a technical solution had to be developed.
This perception allowed the industry to successfully resist making changes by controlling
60
-------
the pace of technological development. The industry did slowly implement simple
technical solutions in response to public pressure and regulatory threats urging them to
develop a technological response to the air pollution problem. One example of this is the
industry's installation in 1963 of a simple valve that allowed for recirculation rather than
direct release of pollutants from the crankcase; this positive crankcase ventilation (PCV)
valve had been used in military vehicles for decades so it was not a new technology, yet
the industry presented the development as a result of their diligent efforts to find
technical solutions (DOJ 1971). Very different strategies are employed during this period
of a cooperative regime than those employed later during the more competitive regime.
3.3 Role of Third Parties
During the early period from 1955-1970 when the federal government had
minimal influence over or interaction with the automobile manufacturers, the California
state government was addressing the industry's resistance by encouraging the
development of pollution control technology through state regulation. Recognizing the
industry's slow pace of development and implementation of technological improvements,
California passed legislation in 1960 that was designed to stimulate competition within
and outside the industry and provide a mechanism for government regulators to review
technical information from within and outside the industry. The Motor Vehicle Pollution
Control Act (MVPCA) set strict emission standards, a 70% reduction in HC and a 57%
reduction in CO, that were to be enforced one year after two satisfactory emission control
devices were certified by the state to meet the standards; all new cars would have to
install one of the certified devices (California 1960).
In drafting this legislation, California legislators identified and addressed two
critical mechanisms with potential to accelerate the pace of emissions control technology
development and reduce the effectiveness of the industry's attempts to resist change: 1)
the competitive pressure of third-party innovators, and 2) the asymmetry of information
between regulators and the industry. The MVPCA was designed to reduce barriers to
market entry of those developing pollution control technology external to the automobile
industry by creating incentive by ensuring a market for devices certified to meet the
standards (CAMVPCB 1965; Krier and Ursin 1977). At this time catalyst technology
was a suspected possible technology, so in response to this legislation a period of intense
catalytic research began as many catalytic chemists jumped at the opportunity to work
toward this exciting potential application of catalytic technology (Lester 1983; Briggs
1984). By requiring a detailed state certification of all devices, the legislation also
created a pathway for information sharing; in the certification process the state regulators
gained the opportunity to evaluate the potential of different technical approaches
developed.
In 1964, when four externally developed devices (three of which were based on
catalytic technology) were certified by the state triggering enforcement of the emission
standards the following year, the automobile manufacturers revealed to the state their
own internally developed technical changes, which consisted of a series of engine
modifications rather than catalytic technology. Once the state certified these industry
developed engine modifications, the automobile companies each chose to implement
their own internally designed approaches rather than implement the externally developed
catalytic devices (Krier and Ursin 1977). So although this legislation motivated and
61
-------
encouraged third party inventors, the inventions were excluded from implementation
because the potential market was removed once the state certified the industry developed
technologies. Nevertheless, the technological progress that was made in the early 1960s
by third parties was influential in demonstrating to both the industry and the government
the potential of catalytic technology; this potential was incorporated into 1970 federal
legislation.
Engineers involved in the development of the catalytic technology have suggested
that if cooperative relationships among industry, government and third parties had
existed, an effective combination of engine modifications and catalyst systems could
have resulted in an efficient pollution control technology that could have surpassed the
California standards by the mid-1960s (Briggs 1984). Instead the industry's engine
modifications approach with a limited level of reductions prevailed while the catalytic
technology with a far greater potential level of reductions was not developed for
implementation until the industry had to respond to the more stringent federal regulations
issued in 1970.
The critical role of third parties in reducing the effectiveness of industry's
resistance to making technical changes is demonstrated in the 1970s also, when the auto
manufacturers applied for an extension to the standards set in the 1970 Clean Air Act
Amendments (CAAA). To appease industry's concern about the technical feasibility of
meeting the strict standards set in the 1970 CAAA, the legislation included two flexibility
mechanisms: 1) automakers were allowed to apply for a one-time, one-year extension if
they could demonstrate to the Environmental Protection Agency (EPA) administrator that
the technology was not yet available, and 2) upholding the standards was contingent on
the assessment of technical feasibility to be carried out by the National Academy of
Science (NAS). These flexibility mechanisms provided two different avenues for third
parties to influence the industry-government dialogue.
During the extension hearings the testimony of independent companies
developing pollution control technology influenced both the industry and the
government. Independent suppliers of catalytic converters (the primary technology
considered capable of meeting the standards), provided manufacturer-conflicting
testimony to federal regulators about the feasibility of implementing the new technology
during the 1972 and 1973 hearings to consider whether or not the industry deserved an
extension to meet the standards (EPA 1972). Additionally, technological developments
made by several non-U.S. auto manufacturers provided regulators with a more optimistic
perspective on the technological possibilities of reducing emissions than the one
promoted by the U.S. automakers. Specifically Honda developed an alternative engine
design (a stratified charge engine) that could meet the 1975 standards without a catalytic
converter (Abernathy and Ronan 1978). Although initially the EPA denied the industry's
request for an extension in 1972, the U.S. court of Appeals ordered the EPA to reconsider
the automakers' request in 1973 and this time the one-year extension was granted.
4. Conclusions
The details of this case suggest that the only time a firm has an interest in
explicitly stating that an ambitious environmental performance goal is feasible is when
they want to sell the technology required to achieve the goal e.g., Honda's CVCC engine,
and Engelhard, the catalytic converter company that provided testimony to EPA on the
62
-------
feasibility of catalytic converters in 1972. For other firms, the strategic choice is between
aggressive claims of infeasibility to oppose a proposed regulation, and passive
acceptance. This choice is probably subject to a tipping point, by which it becomes
disadvantageous to continue claiming infeasibility when either a) enough technological
information has been revealed that the claims pass from appearing reasonably cautious to
appearing dishonest and obstructionist, or; b) political forces behind a proposed
regulation have become strong enough that there is no reasonable probability of
infeasibility claims succeeding in stopping it. Better recognition by policymakers of
these incentives and disincentives associated with admitting or denying technical
feasibility of meeting a regulation could allow for improved communication between
government and industry.
This case also provides useful insight on the potential role of industry
cooperation. If the purpose of a cooperative body is principally to let firms monitor each
other's efforts and announcements, its effect is likely to be to suppress rather than
facilitate innovation even if they don't (as the automakers did) have explicit agreements
to discourage efforts. Government or outside independent expert participation in
cooperative bodies is probably a sufficient guarantee against such uses of cooperative
bodies. If the threat of a required environmental performance target is credible, the cost
of failing to meet it is high enough, the collective interest of the industry is to meet it, and
there's only potential benefit, no harm, in facilitating cooperative work toward it, a
cooperative effort may also be productive. Additionally, for some industries, more
fragmented industries with lots of smaller firms with less R&D capacity in each firm,
cooperative bodies may provide crucial increments of technical capacity to solve
environmental problems
Additionally, this case has demonstrated in several different ways the critical role
that third parties can play in facilitating more productive interactions with regard to
technical details in government-industry interactions. The asymmetry in access to
technical information between industry and government (regulators often rely on
technical information provided to them by the industry because independent technical
information is limited or non-existent) can be minimized by the active involvement of
third parties. Third parties need to be encouraged to reveal information about
capabilities without jeopardizing their commercial relationships.
Finally this research demonstrates the dynamic, interactive nature of the
relationship between industries and governments; these interactions should not be
overlooked in considerations of how best to create incentives for industry development
and implementation of environmental regulations.
63
-------
References
(1969). "United States v. Automobile Manufacturers Association, Inc., General Motors
Corporation, Ford Motor Corporation, Chrysler Corporations and American
Motors Corporation." Civil Action No. 69-75-JWC United States District Court
for the Central District of California.
(1977). The Clean Air Act Amendments 1977. 95 P.L. 95; 91 Stat. 685.
Abernathy, W. J. and L. Ronan (1978). Honda Motor Company's CVCC Engine. Harvard
Business School Working Paper. Boston, MA.
Anonymous (2004). Senior EPA Official. Washington DC.
Ashford, N. A. (1993). Understanding Technological Responses to Industrial Firms to
Environmental Problems: Implications for Government Policy. Environmental
Strategies for Industry. K. Fisher and J. Schot. Washington D.C., Island Press:
277-307.
Bailey, C. J. (1998). Congress and Air Pollution. Environmental Policies in the USA.
Manchester and NY, Manchester University Press.
Briggs, W. S. (1984). Catalysts and the Automobile - 25 Years Later. Applied Industrial
Catalysis. B. E. Leash, Academic Press. 3.
California (1960). California Motor Vehicle Pollution Control Act (MVPCA). Chapter
23. §1 Cal. Stats. 1st Ex. Sess. 346. adding to California Health and Safety Code
§24378-24398.
CAMVPCB (1965). Our Exhausted Air, California's Fight Against Autombile Smog.
The Story of the California Motor Vehicle Pollution Control Board. Los Angeles,
CA, California Motor Vehicle Pollution Control Board.
Cohen, R. E. (1995). Washington at Work. Back Rooms and Clean Air. Boston, Allyn
and Bacon.
DOJ (1971). "Smog Control Antitrust Case, Confidential Memo Regarding Grand Jury
Investigation." Congressional Record - House May 18, 1971: 15626-15637.
Doyle, J. (2000). Taken for a Ride. Detroit's Big Three and the Politics of Pollution. NY,
Four Walls Eight Windows.
EPA (1972). Transcript of Proceedings, Auto Emissions Extension Hearings. Washington
DC.
Haagen-Smit, A. (1952). "Chemistry and Physiology of Los Angeles Smog." Industrial
and Engineering Chemistry 44: 1342.
Ingram, H. (1978). The Political Rationality of Innovation: The Clean Air Act
Amendments of 1970. Approaches to Controlling Air Pollution. A. F.
Friedlaender. Cambridge, MA, MIT Press.
Jaffe, A. B., R. G. Newell, et al. (2000). Technological Change and the Environment. The
Handbook of Environmental Economics. K.-G. Maler and J. Vincent. Amsterdam,
Elsevier Science: 68.
Kagan, D. M. (1977). Regulations and Technological Change in the Automobile Industry.
Science. Technology and Policy. Cambridge, MA, Massachusetts Institute of
Technology: 133.
Kemp, R. (1997). Environmental Policy and Technical Change. A Comparison of the
Technological Impact of Policy Instruments. Cheltenham, UK, Edward Elgar.
64
-------
Krier, J. E. and E. Ursin (1977). Pollution and Policy. A Case Essay on California and
Federal Experience with Motor Vehicle Air Pollution. Berkeley, CA, University
of California Press.
Lester, G. R. (1983). The Development of Automotive Exhaust Catalysts. Heterogeneous
Catalysis: Selected American Histories. B. H. Davis and J. William P. Hettinger,
American Chemical Society.
Lester, G. R. (2003). Adjunct Professor, Northwestern University and President, George
Lester, Inc., Personal Communication.
Lundqvist, L. J. (1980). The Hare and the Tortoise: Clean Air Policies in the United
States and Sweden. Ann Arbor, MI, The University of Michigan Press.
NAS (1973). Report by the Committee of Motor Vehicle Emissions. Washington DC,
National Academy of Science.
NESCAUM (2000). Environmental Regulation and Technology Innovation. Boston, MA.
Parson, E. A. and J. C. Stephens, Eds. (Forthcoming). Innovation. Regulation, and
Environmental Performance: Fighting over Feasibility.
Stephens, J. C. (2004). "Industry Resistance to Technological Change for Sustainability,
Strategies and Motivations: Reducing Automobile Emissions in the U.S."
Technological Forecasting and Social Change submitted.
Taylor, M. R., E. S. Rubin, et al. (2003). "Effect of Government Actions on
Technological Innovation for S02 Control." Environmental Science and
Technology 37(20): 4527-4534.
Taylor, R. E. (1987). Despite Nearly 2 Decades of Federal Efforts, Many in the US still
Breathe Unhealthy Air. Wall Street Journal: 56.
U.S. (1970). Clean Air Amendments of 1970. 91 P.L. 604: 84 Stat. 1676.
Yao, D. A. (1988). "Strategic Response to Automobile Emissions Control: A Game-
Theoretic Analysis." Journal of Environmental Economics and Management 15:
419-438.
65
-------
Carbon Monoxide (CO)
Hydrocarbons (HC)
100
80
60
40
20
I
16
12
1950 1970 1990
year
2010
1950
Nitrous Oxides (NOx)
I
1970 1990
year
2010
~
~
~
» Federal Standards
" Pre-standard Emissions
1950
1970 1990
year
2010
Figure 1. These graphs demonstrating the pre-standard emissions and the decreasing federal standards
for the three primary automobile pollutants, carbon monoxide (CO), hydrocarbons (HC), and nitrous
oxides (NOx) also show how the fifty-year history can be divided into the three distinct time periods
described in the text.
66
-------
Comments on
"A Multi-Agent Model of a Small Firm"
by
Carl Pasurka*
for
EPA workshop on
"Corporate Environmental Behavior and the Effectiveness of Government Interventions"
Washington, DC
April 26, 2004
* Any errors, opinions or conclusions are those of the author and should not be attributed to the
U.S. Environmental Protection Agency.
67
-------
This study investigated interpersonal dynamics that influence corporate environmental
behavior. In its present form, most of the study is not relevant to the U.S. EPA. This study
focuses on: (1) an overview of networks and organizations, (2) a survey of the plastics industry
in New Jersey, (3) a case study of three companies, and (4) a simulation. This study claims that
its ability to analyze (1) eco-parks and (2) industrial ecosystems will make it relevant to the U.S.
EPA.
I will focus my comments on those sections with economics content. In the "Industry
Background" section, the study employs 2001 data. Currently, 2002 data are available from The
Society of the Plastics Industry (SPI). While 2002 data reveal that employment and shipments
declined by approximately 10 percent between 2001 and 2002 for New Jersey, it remains among
the top ten states in shipments of plastics (source: The Society of the Plastics Industry
http://www.plasticsdatasource.org/facts/ni.pdf)
Another data issue is the discrepancy among data sources with regard to the size of the
plastics industry in New Jersey. The following table shows the employment and value of
shipments for plants included in NAICS codes 3259 and 3261 for New Jersey according to the
2001 Annual Survey of Manufactures and the Society of the Plastics Industry (SPI):
Annual Survey of Manufactures
SPI
Employment
36,371
51,011
Shipments
$6.4 billion
$11.7 billion
Source: U.S. Census Bureau http://www.census.gov/mcd/asmdata/2001/ni34.htm
In the section "Illustrative Result - Bringing in Worker Error," the study concludes that
firms with imperfect employees are less profitable and pollutes more. My initial interpretation of
this statement is that worker error is a source of technical inefficiency (i.e., firm produces inside
its production possibilities frontier). Hence, inefficiency manifests itself in the form of reduced
good output production and increased bad output production.
However, there are several possible definitions of increased technical inefficiency. One
definition is a proportional contraction of good outputs and expansion of bad outputs. Additional
definitions involve contraction of good (bad) outputs while maintaining original level of
production of bad (good) outputs. Finally, technical inefficiency can reveal itself as a
proportional contraction of good and bad outputs. These different definitions of technical
inefficiency are illustrated in the following diagram:
68
-------
y
(good)
/fT
0 b
(bad)
Joint Production Frontier
I wish to submit several questions/recommendations to the author. First, how are the
results of this study relevant to analyses of eco-parks and industrial ecosystems? Second, the
underlying economic assumptions of simulation model should be made more transparent. Most
important, the study contained no explanation of the production technology. For example, if an
imperfect worker mistakenly turns off pollution control equipment, why would the firm be less
profitable? Finally, is worker error a justification for firm over compliance in order to avoid
violations of regulations?
69
-------
Comments on
"Compliance and Beyond: Strategic Government- Industry Interactions in Environmental
Policy and Performance"
by
Carl Pasurka*
for
EPA workshop on
"Corporate Environmental Behavior and the Effectiveness of Government Interventions"
Washington, DC
April 26, 2004
* Any errors, opinions or conclusions are those of the author and should not be attributed to the
U.S. Environmental Protection Agency.
70
-------
This study focuses on the role of knowledge, uncertainty and arguments about
perceptions of technical feasibility in the development of environmental policy. Environmental
policy seeks to reduce emissions (i.e., bad outputs) which are the undesirable by-products of a
society's production and consumption activities. The study investigates issues associated with
the design of regulations that encourage technical change which incorporates environmental
improvement. Regulations change the mix of good and bad outputs produced which influences
the direction of technical change. A key question is what is the extent of this regulatory induced
technical change?
Several definitions of technical change may be employed when discussing cases when
both good and bad outputs are produced. One definition is a proportional expansion of good and
bad outputs. A second definition of technical change involves expansion of good output
production while maintaining the original level of bad output production. The third definition of
technical change, which is most relevant for this study, involves a proportional expansion of
good output production and contraction of bad output production.
The focus of this study is the development of regulations on motor vehicle emissions.
Economists have experienced great difficulty in assessing the costs of pollution abatement
activities associated with motor vehicles. This was shown by the discrepancy between EPA and
Bureau of Economic Analysis estimates of pollution abatement costs associated with motor
vehicles. This provides some indication of the difficulty of assessing the cost of implementing
regulations and may provide an indication of the difficulty of assessing the technical feasibility
of new technologies for reducing automobile emissions.
I had several questions for the authors. Would this study reach different conclusions if it
were analyzing the interaction between the auto industry and regulators for other regulations?
For example, what has been the relationship between the auto industry and government
regulators during the implementation of regulations associated with sea belts, air bags, and
Corporate Average Fuel Economy (CAFE) standards?
Is the discussion about the role of (1) third parties and (2) cooperation in R&D efforts in
industries with smaller firms relevant to the environmental R&D efforts of other industries?
Are the conclusions of this study relevant to extant literature? This includes the literature
on asymmetric information between regulators and industry, and the role of monitoring and
enforcement activities. Several articles that appeared in the Journal of Environmental
Economics and Management may be relevant to this study:
Hackett, Steven (1995) "Pollution-Controlling Innovation in Oligopolistic Industries: Some
Comparisons between Patent Races and Research Joint Ventures," Journal of Environmental
Economics and Management, 29, No. 3 (November), 339-356.
Stafford, Sarah (2002), "The Effect of Punishment on Firm Compliance with Hazardous Waste
Regulations," Journal of Environmental Economics and Management, 44, No. 2 (September),
71
-------
290-308.
Brunnermeier, Smita and Mark Cohen (2003), " Determinants of Environmental Innovation in
US manufacturing Industries," Journal of Environmental Economics and Management, 45, No.2
(March), 278-293.
72
-------
Comments on
"Do Facilities with Distant Headquarters Pollute More?: How Civic Engagement Conditions the
Environmental Performance of Absentee Managed Plants"
by
Carl Pasurka*
for
EPA workshop on
"Corporate Environmental Behavior and the Effectiveness of Government Interventions"
Washington, DC
April 26, 2004
* Any errors, opinions or conclusions are those of the author and should not be attributed to the
U.S. Environmental Protection Agency.
73
-------
This study addresses the question of whether the extent of civic engagement affects the
environmental performance of plants operated by absentee managers? Toxic Release Inventory
(TRI) data are used as a proxy for the environmental performance of plants. The TRI data are
collected under the Emergency Planning and Community Right-to-Know Act and are self-
reported data. Mary Streitweiser (1994) found substantial variation in emission
intensities of 3-digit SIC industries within the chemical industry in 1987. These results are
shown in the table accompanying these comments.
The author employs regression analysis to determine which factors are associated with
TRI emissions of chemical plants. In this study, the quantity of TRI emissions from a plant is its
measure of environmental performance. However, two factors affect the quantity of emissions:
(1) its emission intensity (i.e., bad output production per unit of good output production) and (2)
its scale of operation (size of plant). Hence, a plant can emit relative small quantities of toxic
emissions per dollar of output but emit a relative large quantity of toxic emissions if it is a large
plant.
After reading this study, I have several questions for the author. Does civic engagement
affect a plant's emission intensity or scale of operation? The emission intensity of a plant may
be determined by technology over which local managers may have little or no control. Is a
plant's scale of operation affected by a community's attitude? Does civic engagement affect
what type of plant is located in a community?
What is the effect of plant age on emissions? Do newer plants employ technologies that
produce less pollution per unit of good output production than older plants?
Would regional/state dummy variable explain some of the differences in plant TRI
emissions among states/regions?
Different communities seem to have different views about companies with external
managers. Are absentee managers imposed on communities or are they actively pursued by
communities? While there is a recent example of a California community resisting the
construction of a new Wal-Mart, there are other examples of communities using tax incentives
to influence the plant location decisions of companies.
Is the existing economics literature relevant to this study? For example, is there a link
between findings of this study and economics literature on factors affecting plant location
decisions? In addition, this study found that race and class characteristics of a neighborhood are
unrelated to emissions. This is a topic that has been of interest to some economists
Is there a link between the findings of this study and economics literature on TRI? For
example:
Henriques and Sadorsky (1996), "The Determinants of an Environmentally Responsive Firm: An
Empirical Approach, " Journal of Environmental Economics and Management, 30 , No. 3
74
-------
(May), 381-395.
Konar, Shameek and Mark Cohen 91997), "Information As Regulation: The Effect of
Community Right to Know Laws on Toxic Emissions," Journal of Environmental Economics
and Management, 32, No. 1 (January), 109-124.
Brooks, Nancy and Rajiv Sethi (1997), " The Distribution of Pollution: Community
Characteristics and Exposure to Air Toxics," Journal of Environmental Economics and
Management, 32, No. 2 (February), 233-250.
Khanna, Madhu, Wilma Quimio, and Doa Bojilova (1998), "Toxics Release Information: A
Policy Tool for Environmental Protection," Journal of Environmental Economics and
Management, 36, No. 3 (November), 243-266.
Khanna, Madhu and Lisa Damon (1999), "EPA's Voluntary 33/50 Program: Impact on Toxic
Releases and Economic Performance of Firms," Journal of Environmental Economics and
Management, 37, No. 1 (January), 1-25.
Sum Toxic
Releases1
Mean Toxic
Intensity2
Standard
Deviation
Coefficient
of Variation
Interquartile
Range
Chemicals & Allied
Products (28)
2,794.27
19.03
89.17
4.69
8.06
Industrial Inorganic
(281)
360.09
34.07
172.30
5.06
7.21
Plastics & Resins
(282)
441.41
11.30
35.07
3.10
7.22
Drugs (283)
85.23
13.73
33.09
2.41
8.26
Soaps & Cosmetics
(284)
25.75
3.22
11.86
3.68
0.85
Paints & Allied
Products (285)
65.05
9.15
27.14
2.97
7.49
Industrial Organics
(286)
1,106.63
34.50
110.04
3.19
22.09
Agricultural
Chemicals (287)
649.19
50.45
135.99
2.70
22.58
75
-------
Misc. Chemicals
60.93
7.48
35.08
4.69
4.07
(289)
founds of toxins releases and transfers (in millions)
2pounds of toxins / $1000 value of production
Source:
Streitwieser, Mary (1994), "Cross Sectional Variation In Toxic Waste Releases From The U.S.
Chemical Industry," Center for Economic Studies, Working Paper CES-WP-94-8
http://148.129.75.160/ces.php/abstract?paper= 100230
76
-------
Discussant's Comments on Papers in the Session
"Approaches to Environmental Performance"
at the
EPA Conference on
"Corporate Environmental Behavior and
the Effectiveness of Government Interventions'
April 26, 2004
Donald S. Siegel*
Professor of Economics
Department of Economics
3502 Russell Sage
Rensselaer Polytechnic Institute
110 8th Street
Troy, NY 12180-3590
United States
Tel: (518) 276-2049
Fax: (518) 276-2235
sieged@rpi.edu
-------
The purpose of this report is to provide feedback to authors of the three papers presented
in the session on "Approaches to Environmental Performance." These studies are quite diverse,
in terms of their theoretical perspectives and empirical methods, as well as the disciplinary
backgrounds of the authors. As my comments will reveal, they are also at different stages of
development. I begin with the paper that is closest to a "final product."
Paper #1: Do Facilities with Distant Headquarters Pollute More? How Civic Engagement
Conditions the Environment Performance of Ahsentee Managed Plants
This lucid and insightful paper is an econometric analysis of the determinants of the
environmental performance of chemical plants. A major contribution of this study is its attempt
to link several indicators of "civic engagement" to the propensity of absentee managed plants to
emit chemical toxins. A key finding is that absentee managed plants tend to have better
environmental performance when they are located in communities where there is more civic
engagement.
I have two theoretical concerns. The author should reflect on several recent papers that
analyze incentives firms have to be environmental socially responsible (henceforth, ESR). For
example, McWilliams and Siegel (2001) outline a theory of the firm/supply and demand
perspective on ESR. Their model assumes that firms weigh the costs and benefits of engaging
in this activity. Some of these benefits include a greater ability to differentiate products, enhance
the firm's reputation/image, and build or sustain good relations with key stakeholders (e.g.,
employees, government, and investors). In sum, firms are responding to growing demand from
various stakeholders, including consumers, employees, and portfolio managers representing
social investors, who examine pollution measures such as the TRI in their overall assessment of
firm environmental performance.
78
-------
The McWilliams and Siegel framework suggests that ESR is an integral part of a firm's
corporate and business-level strategies. More specifically, the authors conjecture that the
propensity of firms to engage in ESR is positively correlated with firm size (a variable Professor
Grant includes in his empirical analysis), scope of activities, R&D, and consumer
income/wealth. That is, larger, more diverse, and more technologically advanced firms derive
greater benefits from engaging in ESR. Information asymmetry between firms and consumers
(and other stakeholders) regarding the social desirability of managerial practices also appears to
play a critical role in determining the incidence of CSR. In a previous paper (McWilliams and
Siegel (2000)), the authors report empirical evidence that is consistent with a theory of firm
perspective. That is, they find a strong positive correlation between the social performance of
firms and the rate at which they invest in R&D and advertising. Another key paper in this area is
a study by Russo and Fouts (1997), who outline a "resource-based view" of ESR. The authors
hypothesize that ESR can constitute a resource that generates a competitive advantage. They
confirm this hypothesis using extensive data on environmental and firm performance.
The agency theory perspective also bears mentioning in this context. Agency theory is
based on the principal agent framework. An example of a principal/agent relationship concerns
shareholders, who own the firm (and thus, are the principals) and the CEO and senior
management, who are the agents of shareholders. It is well known that agents (managers) often
pursue policies that are not in the best interest of principals (shareholders). This leads to what is
referred to in the literature as "agency costs," or costs associated with insufficient effort by
employees and administrative costs associated with policies to deter such slack effort. In this
context, agency costs may be relevant when senior management at corporate headquarters is
overseeing the environmental performance of numerous manufacturing plants
79
-------
My point is that there may be diseconomies of scope in monitoring the environmental
performance of many establishments, perhaps because managerial resources are spread too thin.
The literature (see Jensen (1993)) also suggests that agency problems may be more severe for
large, diverse, publicly-traded firms, where is there typically greater separation of ownership and
control. Thus, at minimum, it might be useful to include a dummy as a right-hand-side variable
in the econometric analysis denoting whether shares of the corporate parent of the plant are
publicly traded. Note also that our previous discussion in this section on the strategic use of
ESR strongly suggests that the "returns" to ESR may also be higher for publicly-traded firms.
I also have several comments relating to measurement issues and the econometric
analysis. The author estimates the following equation:
I J
(1) ENVPERF= f (ABSENT, £ Pi CIVENG, £ 5jDEMO,
i=l j=l
PLANT, FIRM, IND) + Ui
where the environmental performance (ENVPERF) of the plant is presumed to be a function of a
dummy variable denoting whether the plant has an "absentee owner" (ABSENT), a vector of
indicators of "civic engagement" (CIVENG), several demographic factors (DEMO), including
race and class, plant and firm characteristics (PLANT, FIRM), and a few sub-industry dummies
(IND).
It is important to note that there are several econometric concerns regarding OLS
or simple random effects estimation of equation (1). These concerns are measurement error and
80
-------
specification error, which could result in biased, inconsistent, and inefficient parameter
estimates. The first issue is measurement error in the dependent variable. Note that we don't
observe the plant's "true" environmental performance (ENVPERF*), but rather an imperfect,
self-reported indicator based on the toxic release inventory (TRI) data:
ENVPERF = ENVPERF* + e
where s is the measurement error. As we know from basic econometric theory, errors of
measurement in a dependent variable yield unbiased, although inefficient estimates if these
errors are uncorrected with the independent variables. Thus, the following assumptions must
hold. Cov (s, ABSENT)=0; Cov(s, CIVENG)=0;
However, I conjecture these assumptions could be invalid, since plants with absentee
owners and those that are located in communities where there is more "civically engagement"
may have an incentive to overstate their environmental performance. In the former case,
overstatement could arise because of the monitoring problems noted earlier. In the latter case,
overstatement might result from managers being aware of the fact that communities with greater
civic engagement will expect to see superior environmental performance in local manufacturing
facilities. Alternatively, managers may believe that tightly knit communities will actively
oppose the facility if they perceive that the plant is inflicting environmental damage on the
community.
Some additional measurement problems should also be addressed. The measure of
absent ownership is a dummy variable denoting whether the plant's corporate headquarters is
located in the same state as the facility. A better measure would be based on the distance
between the plant and its corporate headquarters. I am also a bit concerned about the fact that
"civic engagement" is a rather fuzzy construct. Thus, further justification of these measures is
81
-------
needed or the author may choose to employ an econometric method that attempts to control for
such measurement error.
Another econometric concern is specification error, in this case a key omitted variable-
R&D. McWilliams & Siegel (2000) have shown that R&D is positively correlated with
corporate social and environmental performance, as well as firm size (an included regressorin
Professor Grant's model). I am also concerned about possible "Schumpeterian" effects. Joseph
Schumpeter, the eminent Harvard economist, asserted that larger and more diverse firms have a
greater propensity to engage in innovative activity than small firms. He also argued that such
companies reap higher returns to R&D than small firms. There is some empirical evidence of
Schumpeterian effects in the chemical industry (see Link (1980) and Mansfield (1980)). In his
excellent book, Scott (2003) presents an economic analysis of new primary data on
environmental research in the chemical industry and reports evidence that is consistent with
previous Schumpeterian findings.
There are several possible "solutions" to the measurement/econometric problems I have
identified. The first is instrumental variables estimation (e.g., 2SLS, 3SLS) or some form of
systems equations estimation (Griliches (1986)). Another approach is multivariate reverse
regressions, which can be used to derive bounds on the extent of the impact of the measurement
error on the parameter estimates. I believe that the best approach would be a "multi-indicators,
multiple-causes" (MIMIC) model, a type of LISREL model, in which you would attempt to
"explain" the measurement error (see Siegel (1997)). A MIMIC model is essentially a full-
information version of instrumental variables. These models have been used in numerous
sociological studies.
Paper #2 "Compliance anH BeyonH: Strategic Government-Industry Interactions in
82
-------
Environmental Pnliry anH Performance"
This paper seeks to advance our understanding of how firms react to the threat of
regulation. According to the authors, they have several options. They can attempt to oppose,
influence, delay, or support such initiatives. The authors wish to analyze these strategies. On
the other side of the equation, they also wish to examine how regulatory agencies react to such
efforts. I presume that the ultimate objective is to design better policy initiatives that take
account of strategic interaction s between firms and regulatory agencies.
I have several concerns with the current version of the paper. The first is that there is a
major disconnect between the title and the text. The papers reports findings from a single case
study. I am also a bit unclear about whether this is supposed to be an exercise in grounded
theory development. If this is true, the authors need to explain why this approach is warranted in
this context. After reading the introduction, I thought that the authors would propose to develop
a taxonomy of strategies employed by firms and public agencies. Alas, such a taxonomy was not
considered in the remainder of the manuscript. I would like to encourage the authors to move in
this direction, since this would be a really useful outcome.
The paper would also greatly benefit from additional discussion/consideration of
economic theories of regulation. "Capture" theories of regulation are of course highly relevant
in this context. These refer to instances when regulatory agencies are "captured" by the firms
they are supposed to control. In his seminal paper, Stigler (1971) argued that firms will actually
lobby for additional regulation when such legislative initiatives yield either direct monetary
subsidies to these corporations, impose constraints on substitute products or subsidies on
complementary products, create conditions that make it easier for incumbent firms to fix prices
or collude along some other dimension of competition, and when it enhances the ability of
83
-------
incumbent firms to control entry.
The authors should also consider the tension between corporate social responsibility
(CSR) and regulation (Siegel (2001), McWilliams and Siegel (2002)). My point is that there is a
possibility that CSR could be used to forestall regulation. Pre-emptive CSR strategies should
also be considered. That is, it is conceivable that CSR (and regulation) can be used a means of
raising rivals' costs. Marvel (1977) demonstrated the strategic use of CSR in the British textile
industry in the early 1800s. The bottom line is that a polluting firm might actively seek
additional environmental regulation if this would raise rivals' costs more than its own.
Another important trend needs to be considered: the growth of "strategic research
partnerships (SRPs). An SRP is defined as any cooperative relationship involving organizations
that conduct or sponsor R&D. The end result is that R&D is increasingly a collaborative
activity. The increase in SRPs can be attributed to the following policy changes: an expansion of
public -private partnerships, explicit relaxation of antitrust enforcement to promote collaborative
research (e.g., the National Cooperative Research Act (NCRA) of 1984), and policies promoting
more rapid technological diffusion from universities to firms (e.g., the Bayh-Dole Act of 1980).
Examples of partnerships involving private firms only are strategic alliances/networks, licensing
agreements, research joint ventures (RJVs), and industry consortia (SEMATECH). Examples of
public-private SRPs include co-operative R&D Agreements (CRADAs) between federal
laboratories and firms, NSF Industry-University Co-operative Research Centers (IUCRCs) and
Engineering Research Centers (ERCs), university licensing, sponsored research agreements, and
entrepreneurial startups, and publicly-funded R&D programs such as the U.S. Commerce
Department's Advanced Technology Program (ATP).
The authors mention the role of "third parties." I would like them to consider two
84
-------
additional players: activists/NGOs and social investors. Baron (2001) and Fedderson and
Gilligan (2001) assert that activists and NGOs may play a vital role in reducing information
asymmetry between firms and consumers regarding the "social desirability" of a firm's
managerial practices. Social investors and ethical funds could also be key financial stakeholders
in certain industries. This highlights the importance of firms that attempt to measure the social
performance of companies for portfolio managers of "screened" mutual funds and other social
investors (e.g., Kinder, Lydenberg, and Domini).
Paper #3 ""A Multi-Agent MnHel of A Small Firm"
I would like to begin with some words of praise. The authors deserve a great deal of
credit for mixing quantitative and qualitative methods. That is highly unusual in the social
sciences and in my view, an approach that is ideal for an interdisciplinary topic such as corporate
environmental.
This paper addresses a wide range of organizational issues, including questions that are
unrelated to environmental activity. I encourage the authors to ask a smaller set of clearly-
defined research questions that are more targeted to environmental issues. In the next version of
the paper, the authors need to provide much more information on the qualitative methods and
techniques employed. As far as I can tell, you interviewed only one person from each firm. I
could be wrong about this, but this is not clear in the text. You should add a separate section on
this issue and have a chart or table presenting the set of questions you asked these managers. I
am also concerned about your small and possibly biased sample: four firms in a single industry
(plastics) in a single state (New Jersey). You need to include more material to convince the
reader that this industry is somehow representative of manufacturing industries and worthy of
85
-------
attention.
I would also like to see a more structured approach to the qualitative work. As outlined
in Miles and Huberman (1994), there are techniques that can be employed to quantify the
qualitative data from the structured interviews. This involves a detailed analysis of the interview
transcripts, often using a software package such as Nudist. The first stage of analysis is data
reduction or the identification of themes that emerge from the interviews. The second stage is
data display or the generation of frequency counts of the number of times a particular theme is
mentioned. The quantitative results can be used to draw conclusions from the interview
transcripts (See Siegel, Waldman, and Link (2003) for an example of this approach).
As far as I can tell, the method used to identify social networks is quite crude. Nahapiet
and Ghoshal (1998) and (Gant, Ichniowski, and Shaw (2002) present more sophisticated ways of
conceptualizing and measuring the incidence and effects of social networks. For example,
Nahapiet and Ghoshal (AMR-1998) identify three dimensions of social capital. The first is the
structural aspect, which refers to the configuration of ties/relationships that emerge in the
network. Another dimension is the relational aspect, which describes the nature of the
relationships that people have developed with each other. Finally, there is the cognitive aspect,
which refers to resources that provide shared representations, interpretations, and systems of
meaning among the various actors in the network.
My final concern relates to the simulation model. It is well known that such models are
highly sensitive to assumptions researchers make regarding key parameter values. Thus, it is
incumbent upon the authors to present a clear statement of the assumptions of the simulation
model. Given the importance of this concern, I believe it would be best to include this
information on a separate chart or table.
86
-------
87
-------
References
Baron, David P. 2001. Private Politics, Corporate Social Responsibility and Integrated Strategy.
Journal of Economics and Management Strategy 10: 7-45.
Fedderson, Timothy J. & Thomas W. Gilligan. 2001. Saints and Markets: Activists and the
Supply of Credence Goods. Journal of Economics and Management Strategy 10: 149-171.
Griliches, Zvi (1986), "Economic Data Issues," in Z. Griliches and Michael D. Intriligator, eds.
Handhook of Econometrics Amsterdam: North Holland, pp. 1466-1514.
Jensen, Michael, C. (1993) "The Modern Industrial Revolution: Exit and the Failure of Internal
Control Systems," .Tonrnal of Finance 48, pp. 831-880.
Link, Albert N. and John T. Scott. 2001. "Public/Private Partnerships: Stimulating Competition
in a Dynamic Market," International Journal of Industrial Organization,
Vol. 19, No. 5, pp. 763-794.
McWilliams, Abagail and Donald Siegel. 2000. "Corporate Social Responsibility and Financial
Performance: Correlation or Misspecification?," Strategic Management Journal, vol. 21, pp.
603-609.
McWilliams, Abagail and Donald Siegel. 2001. "Corporate Social Responsibility: A Theory of
the Firm Perspective," Academy of Management Review, 26(1): 117-127.
Miles, M. B., and A. M. Huberman. 1994. Qualitative Data Analysis (2nd ed.). Thousand Oaks,
CA: Sage Publications.
Scott, John T., Environmental Research and Development: US Industrial Research, the Clean
Air Act and Environmental Damage (Cheltenham, UK; Northampton, MA, USA: Edward Elgar
Publishing, 2003).
Siegel, Donald (1997). "The Impact of Investments in Computers on Manufacturing Productivity
Growth: A Multiple-Indicators, Multiple-Causes Approach," Review of Economics and
Statistics, Vol. 79, No. 1, February 1997, pp. 68-78.
Siegel, Donald S. (2001). "Do British Companies Really Need a Minister to Make Them
Socially Responsible?," Parliamentary Brief Vol. 7, No. 5, April 2001 (special supplement on
"Business and the Community"), pp. 7-8.
Siegel, Donald S. (2003). "Data Requirements for Assessing the Impact of Strategic Research
Partnerships on Economic Performance: Analysis and Recommendations." Technology Analysis
and Strategic Management Vol. 15, No. 2, pp. 207-225.
Siegel, Donald, David Waldman, and Albert N. Link, 2003a, "Assessing the Impact of
88
-------
Organizational Practices on the Productivity of University Technology Transfer Offices: An
Exploratory Study," Research Policy, Vol. 32, No. 1, pp. 27-48 (A previous version of this paper
appeared as NBER Working Paper #7256, July 1999).
Stigler, George J. (1971) "The Economic Theory of Regulation," Bell Journal of Economics and
Management Science, Vol. 5, No. 1, pp. 71-89.
89
-------
Summary of the Q&A Discussion Following Session III
Magali Delmas (U. C. Santa Barbara)
Directing her comments to Dr. Grant, Dr. Delmas stated that she "really likes the idea of
actually looking at the firm characteristics and the plant characteristics and looking at
how community pressure can impact environmental performance and how these firm
characteristics mitigate this." However, she said she wondered whether headquarter
location is the right measure and went on to provide some feedback from her current
research into "how environmental measures at the city level respond to stakeholder
pressure." She said she has received conflicting opinions from those she has interviewed,
with some saying that the main action stakeholder is the corporate headquarters and
others saying, "No, it's the community, and it doesn't relate to the location of the
headquarters." So, this has led her to wonder "what type of other measures [could be
used] to tease out whether this is kind of a centralized type of management, with
everything kind of done at a corporate level, or whether it's more decentralized, with the
city or the plant actually having some decision-making power.
Dr. Delmas added, however, that they found that "headquarters location matters in terms
of adopting environmental management practices," and she said cities where
headquarters are located more commonly implement environmental management
practices.
Don Grant (University of Arizona)
Dr. Grant responded, "This kind of goes back to Carl's [Carl Pasurka, one of the
discussants] first attack [laughter] about how you go about measuring absentee
management plants," and he agreed with Dr. Pasurka that in a lot of the current economic
geography literature the standard policy is to measure in terms of miles between a plant
and its headquarters. He went on to explain that the reason he chose to measure in terms
of whether a plant is headquartered in the same state was "because of all the covenants
beginning in the early 1980's . . . when states were competing for businesses and they
were trying to lure them by lowering their environmental standards." Acknowledging
that the other method might work as well, he concluded by saying, "Those are the only
two alternatives I've ever come across. If there's a better one, let me know.
Andrew Hutson (University of North Carolina, Chapel Hill)
Addressing Dr. Grant also, Dr. Hutson said, "You've embedded your argument in the
larger globalization debate, and I'm wondering if the data you use aren't more
appropriate simply just for looking at different regional variations within the United
States. Citing the different dynamics that come into play on a global scale—e.g.,
regulatory dynamics, administrative capacity, community pressures (with people being
better organized)—and the fact that multi-national firms "may have different and better
incentives [than smaller, local firms] to have structured, formal environmental
management in place," Dr. Hutson questioned the "generalizability" of Dr. Grant's
findings for the larger, global debate in which he embedded his work.
90
-------
Don Grant
Dr. Grant responded that he had been "very careful not to suggest" that his findings could
be generalized and applied to other situations in the world. He stated that the reason he
couched his argument in terms of the globalization debate is "that's how it's typically
addressed today," as opposed to 50 years ago when it was more of a domestic issue. He
added that he believes that the most relevant literature is the literature that has to do with
domestic plant growth and re-location.
Going on, Dr. Grant stated, "By the same token, I wanted to suggest in the paper that
there are future avenues of research" and once more studies have produced more data, he
will then be able to test the generalizability of his study. He agreed with Dr. Hutson that
at the present time it's important to stress that studies that have uncovered the absentee
management effect are limited to the U.S. He concluded by adding that they had looked
at the effects of foreign-owned firms on environmental performance—an issue
particularly relevant to the chemical industry because so many of those firms are foreign-
owned—and they found no evidence that foreign ownership "has a bad effect on
emissions."
Wayne Gray (Clark University)
Dr. Gray addressed his comments to Dr. Grant, also, and said he was "just a bit
concerned" that Dr. Grant was getting his comparisons between locally managed and
absentee-managed plants from within the chemical industry, which isn't necessarily a
very homogeneous group. As Dr. Gray put it, "There are a lot of different sub-industries
within chemicals and such" and, depending on the product being produced, some of those
might be inherently more likely to be absentee-managed. He gave the example of
products that have high transportation costs. A firm producing such a product would be
more likely to have production facilities "spread all over the country" and necessarily
absentee-managed, whereas other products that are more easily shipped might be more
centrally produced. Dr. Gray wondered whether that factor was "correlated with any sort
of sensitivity in terms of how easy it is for them to reduce their pollution." He wasn't
sure how much or what kinds of controls had been used in the study "in terms of the
particular kind of products, either . . . controlling for the 4-digit industry or something
like that or the interaction effect with the local community."
Don Grant
Dr. Grant admitted that they haven't explored that in great detail, but agreed that there is
much heterogeneity within the industry. He added the example of continuous-processing
plants and batch-processing plants (e.g., soaps and detergents) and said that issues such as
these might have implications for how far a facility can be from the corporate
headquarters. He closed by saying, "We really haven't delved into these sub-industry
differences, but I think there could be something there."
Irene Xiarchos (West Virginia University)
91
-------
Returning to the issue of the distance factor, Ms. Xiarchos asked Dr. Siegel why one
would want to measure "the exact distance of the headquarters from the location where
the product is produced."
Don Siegel (Rensselaer Polytechnic Institute)
Dr. Siegel responded by clarifying that the argument in question is that "the farther away
the senior managers and corporate headquarters were, the more likely they were to be
disengaged from what was going on at the plant." He presented the example of himself,
living in Albany, New York, ten or fifteen miles from the Massachusetts border, yet 500
miles away from Buffalo, which is in the same state. He believes categorizing by
distance provides a much more precise picture of an area of influence.
Irene Xiarchos
Ms. Xiarchos countered that "rather than looking at distance specifically," maybe going
by geographic/political levels or boundaries—county, state, region, country, continent—
would yield a more meaningful stratification. She reiterated that she didn't "think the
distance measured in kilometers would necessarily make a difference."
Don Grant (University of Arizona)
"It's a great empirical question. Again, the reason why I studied it as we did was because
in the context of environmental regulation, states matter, and so the state boundaries are
factored into our absentee management."
Pete Andrews (University of North Carolina, Chapel Hill)
"I'd really like to hear Clinton Andrews respond to some of the comments that were
made about his paper, because having not read the paper but listening to the presentation,
I'm really intrigued by the potential of this area, but the comment is also correct that it is
very unspecified in terms of the actual presentation."
Clinton Andrews (Rutgers University)
Dr. Andrews acknowledged that in a half-hour presentation that includes "empirical stuff
plus some modeling results, there's not much time to go into the details," but said he
would gladly present more details to anyone with the time and inclination to gather
around his laptop computer. He went on to say that he did feel compelled to clear up the
discussion of worker error, and he explained: "The way we did it, which is just one way
among many of conceptualizing it, was that workers had to interact with their technology
and basically adjust it. The technology had optimal set points, and the workers would
basically err in hitting those set points. . . . When it's structured that way, then you
inevitably find that profits are going to go down and pollution is going to go up because
of the form of that particular technology and the production function."
Rob Axtell (Brookings Institute)
Dr. Axtell followed up on Professor Siegel's comments on Dr. Grant's paper, his concern
being primarily the re-writing of the model as a locational choice model—in other words,
92
-------
taking a free-choice model and adding local structures to that. He said this can
potentially present a problem because "when you think about it, normally the emissions
level would depend on the civic engagement variable. But it could also be the case that
the civic engagement level is in fact dependent on the level of emissions in some
important way." He just wanted to reiterate that this is a difficult econometric and
specification problem.
Don Siegel
Dr. Siegel responded by saying, "I don't think you can address that problem with cross-
sectional data. If you had panel data . . . and you could add some data over time, you
could probably do that."
END OF SESSION III Q&A
93
-------
Corporate Environmental Behavior and the
Effectiveness of Government Interventions
SECOND DAY
INTRODUCTORY REMARKS BY
JIM GULLIFORD
REGION 7 REGIONAL ADMINISTRATOR
A WORKSHOP SPONSORED BY THE U.S. ENVIRONMENTAL PROTECTION
AGENCY'S NATIONAL CENTER FOR ENVIRONMENTAL ECONOMICS (NCEE),
NATIONAL CENTER FOR ENVIRONMENTAL RESEARCH (NCER)
April 26-27, 2004
Wyndham Washington Hotel
Washington, DC
Prepared by Alpha-Gamma Technologies, Inc.
4700 Falls of Neuse Road, Suite 350, Raleigh, NC 27609
-------
ACKNOWLEDGEMENTS
This report has been prepared by Alpha-Gamma Technologies, Inc. with funding from
the National Center for Environmental Economics (NCEE). Alpha-Gamma wishes to
thank NCEE's Cynthia Morgan and Ann Wolvertan and Project Officer Ronald Wiley for
their guidance and assistance throughout this project.
DISCLAIMER
These proceedings are being distributed in the interest of increasing public understanding
and knowledge of the issues discussed at the workshop and have been prepared
independently of the workshop. Although the proceedings have been funded in part by
the United States Environmental Protection Agency under Contract No. 68-W-01-055 to
Alpha-Gamma Technologies, Inc., the contents of this document may not necessarily
reflect the views of the Agency and no official endorsement should be inferred.
2
-------
Introductory Remarks by Jim Gulliford
Region 7 Regional Administrator
at the
Corporate Environmental Behavior and the
Effectiveness of Government Interventions
Washington, D.C.
April 27, 2004
Good morning. Thank you, Matt, for the kind introduction.
It's indeed a pleasure to be here with you today. I hope that when
I've finished, you will take with you - a message of change -
change in the way we look at new environmental regulation or
rule roll-out.
I want to cover a couple of topics -
First, I would like to talk a little about Administrator Mike
Leavitt's vision of collaboration and a "Better Way";
3
-------
And then, I'll go over how Region 7 managed the roll-out of the
proposed and final CAFO rule. This is important as we identify
opportunities to change our behaviors toward the environment.
When Administrator Mike Leavitt was confirmed, the message
he brought was one of collaboration ... meeting in the middle. This
message talks of a "Better Way" by:
facilitating collaboration;
harnessing technology; and
creating market incentives.
We need to assure compliance in order to enforce the law. Let
me explain what that means. In his 500 Day Plan, Administrator
Leavitt makes the point that we need to make compliance our
enforcement objective. In other words, we will use the intent of the
4
-------
law to support our outreach and inspections to achieve compliance
with rules and regulations.
When it came to new regulations and rules in the past, there
wasn't much collaboration. Regulations were rolled out, with little or
no outreach, training, or communication, and the regulated
community was required to comply with the regulation. I know we
can do a better job. It's just going to take some effort to change the
process or behavior.
We're here to talk about ways to change environmental
behaviors. The biggest environmental improvement comes from
behavior change, not from command and control enforcement. We
don't really have all the command and control tools to enforce in the
ag sector, for example non-point source pollution.
5
-------
Now, let's talk about the process Region 7 used in rolling out
the proposed and final CAFO rules.
When the CAFO rule was introduced, the Secretary of
Agriculture and former EPA Administrator Christie Todd Whitman
made a joint announcement. I believe it is critical to build
meaningful relationships among EPA, federal, state and local
agencies, and other partners to realize environmental gains.
The rule covers some 11,000 large animal feeding operations
nationally, 4,300 of which are in Region 7. The rule, as it stands
today, will cut the annual nutrient runoff from CAFOs by about 61
million pounds. We expect it will also cut sediment loads by 1.1
billion pounds. That's quite an environmental benefit.
I'm very proud of the way the roll-out was handled in my
6
-------
Region. Since agriculture is the largest industry in Region 7, the
new CAFO rule affected an incredibly large part of our regulated
community. We knew that if we handled the roll-out the way new
regulations had been managed in the past, we were going to have
a huge problem. So, we decided to be proactive.
Early on, we put together a strategy that involved outreach,
holding workshops, training events, and meetings with the ag
community. We heard what was said - worked through the
problems - and developed a stronger partnership with the ag
community.
In the summer of 2002, we listened to the ag community about
their concerns. Early in December, we met with industry
associations informally on a state-by-state basis, just before the rule
came out.
7
-------
On December 19, 2002, right after the rule was signed, we met
with regulated producer associations, USDA, and the state
environmental agencies. The revised rule for Concentrated Animal
Feeding Operations became effective on April 14, 2003.
We held the meeting in Kansas City, Kansas, and introduced
our partners to the details of the Rule. At the meeting, 150 individual
state attendees met to determine how to work together to get the
word out and to help the regulated community best understand the
new regulation.
We knew we needed their networks to help us get the message
out, and they knew that having our expertise regarding the Rule
would help with clarifying the Rule for their organization members.
We were involved in these outreach activities in order to
8
-------
remove ambiguity about who is covered, and ensure the effective
management of manure at the largest CAFOs. We also reminded
everyone that as we transitioned to the new rule, we still had an
existing rule that we would continue to enforce!
We held approximately 40 outreach meetings throughout
Region 7 states.
As EPA, state regulatory agencies, and CAFO operators began
the transition into implementing this revised Rule, there was a great
deal of outreach and support from EPA and our partners, including
USDA, and many farm service organizations. Let me give you a
couple of specifics.
In Missouri, we participated in roughly ten meetings related to
the revised CAFO regulations. The first was held in early January
9
-------
2003, to build consensus on an agenda designed to educate
producers on the requirements of the CAFO regulations.
Participants included representatives from the Missouri
Departments of Agriculture and Natural Resources, producer
groups, NRCS, the University of Missouri, and the Missouri Farm
Bureau. Outreach meetings were held in six locations across the
state. Over 500 producers attended the meetings.
In Nebraska, Region 7 participated in six outreach meetings
designed to educate producers on the requirements of the CAFO
regulations. Participants included representatives from the
Nebraska Department of Environmental Quality (NDEQ), NRCS, the
University of Nebraska at Lincoln, and several stakeholder groups.
Similar to Missouri, these outreach meetings were at locations
10
-------
across the state and reached more than 500 producers. You get the
pattern.
We also joined the state in a meeting with the Sierra Club to
provide an overview of the revised CAFO regulations and
Nebraska's plans for permit program revisions.
We sponsored and participated in CAFO Regulations
Implementation Workshop in Nebraska City, Nebraska. The focus
of this event was to:
Share current scientific knowledge and resources relevant to
implementation of the CAFO regulations;
to facilitate regional discussions for implementation of CAFO
regulations and consistency between state regulatory
agencies; and
11
-------
to promote communication and possible collaborative activities
among technical resource community and regulatory agencies
-discussing opportunities for innovation and new technologies
for compliance.
Let me switch gears for a minute to discuss Alternative
Technologies. When the proposed regulations, which were based
on the use of total containment and subsequent land application.
appeared to be inflexible regarding the type of technology that could
be used, Region 7 opened the lines of communication regarding
alternative technologies.
The Rule gives limited opportunity for the use of alternative
technology. If it can be shown that another type of system can
achieve equivalent results, then the permitting authority may use this
alternative system as the basis for writing the NPDES permit. This
12
-------
is not a relaxing of the standard but, on a site-by-site basis, using an
alternative system that will achieve the same or better environmental
results.
In September 2003, a meeting was held in Des Moines to
discuss how alternative technologies could meet the new CAFO
Rule requirements. Several groups are working with Iowa State
University and other land grant university experts to demonstrate
and monitor proposed alternative technologies, and develop models
that would support the use of those technologies to satisfy the
requirements.
One example of alternative technologies is the application of
anaerobic digester systems - These systems:
cover lagoons
recover methane for producer energy needs or sale to local
13
-------
energy grids.
The benefits to the environment include reduced greenhouse
gases, odor reduction, possible hydrogen sulfide reduction, and
reduced opportunities for the release of pathogens to surface and
ground water. (Hosted St. Louis meeting - technical experts ~ ISO.
As you can see, the Region was very involved in rolling out the
revised CAFO rule. We took the responsibility to make sure that the
regulated community knew what was in the CAFO rule, and what
was going to be required to be in compliance with the Rule.
John Silberman talked yesterday about the keys to making the
Agency's "Audit Policy" successful - the obligation of leadership,
proactive outreach, communication, and availability.
14
-------
We didn't stop at the signing of the rule. We continue to hold
training opportunities, workshops, and meetings as needed. In fact,
next month we will be meeting with the state of Iowa regarding their
re-written regulations. We need to do more of this kind of
collaborative work.
However, all this does not change EPA and the states'
responsibilities as enforcers. We know that a key element of
compliance is good enforcement. We haven't missed an opportunity
to remind operators that we are continuing to enforce our current
rule, and we will enforce the new rule as well.
Timely and appropriate enforcement actions result in
compliance, not only at a specific facility where the action is taken,
but they also act as a deterrent to others in the regulated
community. The regulated community wants to stay in compliance
15
-------
with the Rule, thereby avoiding an enforcement action against them.
By its nature, the livestock and poultry industry operates on a
pretty narrow margin. Compliance, and avoiding penalties, is an
important objective for producers.
All of our efforts resulted in less confusion about the revised
CAFO Rule. We also saw better implementation of the Rule. And
as a whole, the environment will benefit from the realization of
cleaner water. As Administrator Leavitt says - compliance is where
we find environmental benefits.
We are taking important steps in Kansas City, but we at EPA
know we cannot address these issues effectively on our own.
Developing new and enhanced partnerships is becoming
increasingly important.
16
-------
In retrospect, Region 7 gained the benefit of many partnerships
being built through this effort. We were able to leverage our
outreach resources more efficiently. And, the producers received
timely and accurate information regarding the proposed and final
CAFO rule requirements.
We intend to take advantage of this as we work on additional
ag initiatives - offroad diesel ext; then others.
I hope as I close that you have a better sense of Region 7's
priority of developing a better approach to regulation roll-out,
fostering mutually beneficial working relationships with the ag sector,
and of our sincere desire to produce significant environmental
improvements.
Thanks for sharing a portion of your agenda with me-
17
-------
continued success for the rest of the day.
18
-------
Corporate Environmental Behavior and the
Effectiveness of Government Interventions
PROCEEDINGS OF
SESSION IV: EVALUATION OF VOLUNTARY PROGRAMS
A WORKSHOP SPONSORED BY THE U.S. ENVIRONMENTAL PROTECTION
AGENCY'S NATIONAL CENTER FOR ENVIRONMENTAL ECONOMICS (NCEE),
NATIONAL CENTER FOR ENVIRONMENTAL RESEARCH (NCER)
April 26-27, 2004
Wyndham Washington Hotel
Washington, DC
Prepared by Alpha-Gamma Technologies, Inc.
4700 Falls of Neuse Road, Suite 350, Raleigh, NC 27609
-------
ACKNOWLEDGEMENTS
This report has been prepared by Alpha-Gamma Technologies, Inc. with funding from
the National Center for Environmental Economics (NCEE). Alpha-Gamma wishes to
thank NCEE's Cynthia Morgan and Ann Wolverton and the Project Officer, Ronald
Wiley, for their guidance and assistance throughout this project.
DISCLAIMER
These proceedings are being distributed in the interest of increasing public understanding
and knowledge of the issues discussed at the workshop and have been prepared
independently of the workshop. Although the proceedings have been funded in part by
the United States Environmental Protection Agency under Contract No. 68-W-01-055 to
Alpha-Gamma Technologies, Inc., the contents of this document may not necessarily
reflect the views of the Agency and no official endorsement should be inferred.
11
-------
TABLE OF CONTENTS
Session IV: Evaluation of Voluntary Programs
The ISO 14001 Management Standard: Exploring the Drivers of
Certification
Andrew King, Darmouth College and Michael Lennox, Duke University 1
Participation in Voluntary Programs, Corporate Reputation, and Intangible
Value: Estimating the Value of Participating in EPA's ENERGY STAR®
Program
Lou Nadeau, ERG, Inc 43
Discussant
Charles Griffiths, U.S. EPA, NCEE 88
Discussant
Jorge Rivera, George Mason University 91
Summary of Q&A Discussion Following Session IV 93
in
-------
I lll ISO 14001 MANAGEMENT STANDARD:
EXPLORING THE DRIVERS OF CERTIFICATION
ANDREW A. KING
Tuck School of Business
Dartmouth College
100 Tuck Hall
Hanover, NH 03755
Tel: (603) 646-9185
Fax: (603) 646-1308
andrew.a.king@dartmouth.edu
MICHAEL J. LENOX
Fuqua School of Business
Duke University
PO Box 90210
Durham, NC 27708
Tel: (919) 660-8025
Fax: (919) 681-6244
mlenox@duke.edu
ANN TERLAAK
School of Business
University of Wisconsin-Madison
4261 Grainger, 975 University Ave
Madison, WI 53706
Tel: (608) 262-5227
Fax: (608) 262-8773
aterlaak@bus. wise .edu
Draft date: May 1, 2004
Working Paper. Please do not quote or cite without authors'permission
King, Lenox & Terlaak ©2004. This research was partially funded by NSF/EPA grant #R827819. We would like to
thank Pratima Bansal, William Greene, Constance Helfat, Paul Ingram, Jackson Nickerson, Michael Russo, Brian
Silverman, and Michael Toffel, for their advice and assistance with this paper. We would also like to thank
Marshall Schminke and the three anonymous reviewers for their thoughtful direction and encouragement.
1
-------
I lll ISO 14001 MANAGEMENT STANDARD:
EXPLORING THE DRIVERS OF CERTIFICATION
Abstract: In this paper, we explore the drivers of certification with the ISO 14001
environmental management standard. Scholars and practitioners debate whether ISO 14001
certification signals superior underlying environmental performance or signifies merely the
adoption of specific environmental management practices. Our paper helps to resolve this debate
by developing a theory of why and when organizations will choose to certify with ISO 14001.
We propose that certification with a management standard will be more likely when it is difficult
to communicate credibly environmental practices to supply chain partners. We develop
hypotheses concerning the type of information that will be communicated through certification.
We empirically investigate our hypotheses using a rich longitudinal database. We find evidence
that organizations certify with ISO 14001 to reduce information asymmetries with supply chain
partners. In particular, we find that geographically and culturally remote suppliers are more
likely to seek certification. We do not find evidence that certification serves as a signal of
superior environmental performance. Rather, our findings suggest that suppliers use certification
to communicate about environmental improvement efforts. In our conclusion, we discuss the
implications of our findings for public policy makers and firm managers. (190 words)
Keywords: institutions, management standards, industry self-regulation, ISO 14001
2
-------
The use of certified management standards to regulate business activity is of growing
interest to academics, business managers, and policy makers. These standards stipulate neither
product specifications nor process attributes. Instead, they require and certify the existence of a
set of internal organizational practices and routines. For business managers, these standards may
increase efficiency or solve inter-firm coordination problems. For policy makers, these standards
may provide an alternative to costly government regulation. Prominent examples of certified
management standards include the OHSAS 18000 standard (for occupational health and safety),
the International Organization of Standardization's ISO 9000 and ISO 14001 management
standards (for quality and environmental management), and the Eco-Management and Audit
Scheme (EMAS).
In this paper, we explore certification with the ISO 14001 environmental management
standard. Sponsored by the International Organization for Standardization (ISO) and designed
by an international technical committee (TC 207) comprising more than 500 members, ISO
14001 specifies a set of environmental management guidelines and practices. It creates a system
for third-party auditors to certify compliance with the standard. From the outset, the role of this
standard has been a source of considerable debate. For example, in testimony before the U.S.
Congress, members of the standard setting committee expressed differing expectations. Some
suggested that certification would help "to distinguish companies that are doing the bare
minimum from those that are committed to environmental excellence" (Morella, 1996). Others
noted that "ISO 14001 compliance may become a standard of due care in assessing whether a
company was [acting] in good faith" (Mazza, 1996). Still others suggested that the program
might provide direct operational advantages (Collins, 1996).
3
-------
Our paper helps to resolve this debate by developing a theory of why and when
organizations will choose to certify with ISO 14001. We propose that certification with a
management standard will be more likely when it is difficult to communicate credibly
environmental practices to supply chain partners. We develop hypotheses concerning the type of
information that will be communicated through certification. We hypothesize that certification
will either help buyers choose high performing suppliers or help them monitor performance
improvement among existing suppliers. We empirically investigate our hypotheses using a rich
longitudinal database. Finally, we discuss the implications of our findings for public policy
makers and firm managers.
THEORY & HYPOTHESES
Certified management standards like ISO 14001 include two fundamental elements.
First, they codify a set of standard practices and behaviors. Second, they provide a certification
system that allows organizations to demonstrate their compliance with these practices and
behaviors. Most of the pioneering work on ISO 14001 and similar standards has emphasized the
importance of the former element and made use of certification only as a mechanism for
measuring the adoption of the specified practices (Corbett & Kirsch, 2001; Delmas, 2002; Guler,
Guillen, & Macpherson, 2002). This research has tended to model adoption as a process of
institutional pressure or information based contagion.
A handful of recent studies have proposed that certification represents a distinct and
important element of these standards and fundamentally changes the way the standards are used
(Anderson, Daly, & Johnson, 1999; Jiang & Bansal, 2003). These studies suggest that more
consideration should be given to how certification could help resolve problems of credible
4
-------
communication. This evidence suggests that managers in organizations may choose to certify
with standards such as ISO 14001 when they recognize that asymmetric information could cause
inefficient exchange with skeptical supply chain partners (Anderson et al, 1999; Jiang & Bansal,
2003).
Asymmetric information causes two main problems among exchange partners. First, it
makes it harder to assess the quality of potential partners (the 'selection problem'). Second, it
makes it more difficult to evaluate improvement efforts among existing partners (the 'monitoring
problem'). Akerlof (1970) used the pre-owned car market to illustrate how asymmetric
information could result in a selection problem. He postulated a market in which sellers have
some information about the quality of used autos (maintenance, history, improvements, etc.) that
buyers do not. He pointed out that if buyers recognize the possibility that sellers could make
false claims of superior quality, buyers would be unwilling to pay a higher price for cars with
allegedly higher quality. In response, sellers would withdraw their high quality vehicles, leaving
only lemons in the market. As a result, even if both suppliers and buyers would prefer to deal in
high quality used cars, this selection problem will cause a market in which only low quality cars
are bought and sold.
The second type of asymmetric information problem, the monitoring problem, occurs
when asymmetric information makes it difficult to know if agreements have been met.
Asymmetric information between suppliers and buyers may make it difficult to observe fully the
actions of the supplier (Silverman, Nickerson & Freeman, 1997; Williamson, 1985). For
example, a customer that pays a mechanic to do repair work on a vehicle may be unable to
determine if the work has been done properly. As with the selection problem, this monitoring
5
-------
problem can cause an inefficient market for goods and services, and thereby harm both suppliers
and buyers.
We theorize that certified management standards such as ISO 14001 reduce asymmetric
information problems by allowing suppliers to credibly communicate information to buyers. To
begin testing our theory, we first seek to uncover whether certification with ISO 14001 occurs
more frequently when organizations are likely to have less information about supply chain
partners. We then develop hypotheses to distinguish the use of certification to solve the selection
or the monitoring problem. We focus our theory on intermediary supply relationships.
Certification and Information Asymmetries
Numerous factors influence the transfer of credible information and thus its distribution
among parties. A common finding across many literatures is that the physical distance between
two parties is a critical factor (Allen, Lee et al., 1980; Hamilton, Godfrey & Linge, 1979; Katz &
Tushman, 1979). Distance reduces information transfer through its direct effect on transfer costs
and by its association with other restricting factors (Mariotti & Piscitello, 1995). For example,
distance may reduce the number of shared information links and so prevent receiving parties
from checking the veracity of information through redundant sources (Lane & Bachman, 1996).
Distance may also reduce the frequency of interaction and so reduce the propensity of parties to
develop a reputation as a credible source (King, 1999). Empirically, numerous studies in various
social settings have documented that information transfer decreases rapidly with increasing
physical distance between two parties (Adams, 2002; Allen, Lee, & Tushman, 1980; Hamilton,
Godfrey & Linge, 1979).
6
-------
Given the propensity for physical distance to reduce information transfer and increase
asymmetric information, we should expect suppliers to use certification to communication
credibly with buyers when buyers are more physically distant.
Hla: The more distant an organization is from its buyers, the greater the propensity for
the organization to certify with the ISO 14001 management standard.
Aside from physical distance, social, cultural, and institutional distance can reduce
information transfer and increase information asymmetries (Caves, 1982). One explanation is
that a shared culture or belief system facilitates the processing of transferred information
(Hofstede, 1980). Numerous studies have shown that cultural and physical distance increases the
difficulty and cost of selecting and monitoring foreign suppliers (Buckley & Casson, 1979;
Hamilton et al., 1979; Kogut & Singh, 1988). Such "liability of foreignness" is one of the
central tenants of international business theory (Zaheer, McEvily & Perrone, 1998). Following
this tradition, we argue that information asymmetries should be especially high in international
supply relationships. As a result, organizations that are more likely to supply foreign buyers will
be more likely to certify with a management standard.
Hlb: The more an organization sells to foreign buyers, the greater the propensity for the
organization to certify with the ISO 14001 management standard.
Certification and Selection
In the above section, we hypothesize that certified management standards like ISO 14001
help resolve asymmetric information problems. We did not explore whether they resolve
selection or monitoring problems. In this section, we develop hypotheses consistent with the use
of certified management standards to reduce the selection problem, i.e., we consider the potential
that buyers use certification to determine which organizations to use as suppliers. Recall that a
7
-------
selection problem occurs when buyers cannot observe underlying product or firm quality.
Akerlof s market for lemons exemplifies how this situation can result in inefficient markets. To
overcome this inefficiency, sellers (suppliers) may employ signals that allow buyers to
differentiate high from low quality suppliers.
Because a certified management standard does not stipulate any outcome requirements
(instead, it specifies a set of management practices), it can only provide a signal of superior
organizational performance by causing organizations to act in ways that reveal what they know
about themselves. It must entice high performing organizations to certify, while dissuading low
performing ones from doing so. In his seminal contribution, Spence (1973) used the job market
to provide an explanation for how such signaling might function. The idea is that high
productivity workers (e.g., more motivated or intelligent people) choose to get a college degree
not because they seek to learn something, but because they seek to differentiate themselves from
those that choose not to get a degree. Because the cost of getting a degree (studying, writing
papers, etc.) is less for high productivity workers than it is for low productivity ones, and
assuming that employers are willing to pay a premium to the high productivity workers (once
they can identify them), only high productivity workers will obtain a degree. In contrast, a low
productivity worker will refrain from seeking a degree since the premium that could be obtained
from having a diploma will be less than the cost of acquiring it.
To serve as a signal of superior performance, certification with a management standard
like ISO 14001 must follow a similar logic. Specifically, the cost of certifying must be lower for
high performing suppliers and buyers must be willing to pay a premium to high performing
suppliers. With respect to the first requirement, there are many reasons for buyers to pay a
premium to suppliers with higher environmental performance. Environmental problems at the
8
-------
supplier can cause supply disruptions. For example, "in 2001, a refinery fire in Illinois caused
shortages lasting for weeks and forced the EPA to temporarily rescind reformulated gasoline
requirements in the Chicago area (Slawsky, 2004)." Environmental problems at the supplying
organization can damage the reputation of its supply chain partners. For example, it was concern
about the practices of suppliers that damaged the reputation of Nike, Starkist, and Unilever
(among many others) and caused these organizations to create management practices that their
suppliers must follow. Finally, under U.S. CERCLA statutes, supply chain partners can be held
responsible for improperly disposed toxic waste (Snir, 2001).
With respect to the second requirement, there are several reasons to believe that
environmentally responsible organizations can certify at lower cost. Technical committees (like
TC 207 for ISO 14001) seek to make it easier for high performing organizations to certify by
designing the management standard so that it includes practices that have been found to improve
organizational performance (Collins, 1996). The logic of such a design is that high performing
organizations should have implemented some of the required practices and thus need to adopt
fewer additional practices to obtain certification (Collins, 1996). Empirical research provides
further evidence that certification cost are inversely related to performance (Naveh, Marcus,
Allen, et al, 1999). For the ISO 9000 quality management standard (which is the older brother of
ISO 14001 and served as a model for its design), Marquardt (1992) observes that certification
costs depend on where you start. "If you've just won a Baldridge Award, registration of a plant
or business may take you a few days. But if your quality system needs to be improved or created
from the ground up the process can take as long as a year and cost $100,000 or more"
(Marquardt, 1992: 51). In the case of ISO 14001, a survey found that a majority of respondents
felt that leading organizations could certify with ISO 14001 more cheaply than environmental
9
-------
laggards (Ferrer et al, 2003).
The literature on business and environment adds a specific reason to believe that the cost
of certification to ISO 14001 should be lower for high performing organizations. Scholars
suggest that organizations are responding to environmental pressures in stages (OTA, 1986).
They first ignore environmental problems, then perceive them to be a regulatory issue, and only
later understand them as a source of strategic advantage (Hoffman, 1997; OTA, 1986). As an
organization moves through these stages, its environmental (and potentially economic)
performance improves because its response shifts from one emphasizing technical buffers to one
emphasizing proactive environmental management (Russo & Fouts, 1997). These theories
would suggest that organizations in the later stages of evolution (and thus with higher
performance) will be more able to certify with environmental management standards. Bansal and
Hunter (2003) indeed find that organizations with better environmental reputation (and
presumably performance) were quicker to certify with ISO 14001 than those with lower
performance, possibly due to relatively lower certification costs.
The above discussion suggests that some of the conditions are present that would allow
certification with ISO 14001 to act as a signal of underlying organizational performance.
Whether or not organizations use it in this way is an empirical question. Evidence to support or
disconfirm such a signaling theory can best be found by evaluating whether high performing
organizations tend to certify. No signaling equilibrium can exist in which low quality suppliers
(or all suppliers) certify. If low performance suppliers certified, ISO 14001 would no longer
convey superiority, and no supplier would certify because doing so would entail a cost with no
benefit (recall that buyers are only willing to pay a premium to certified suppliers if the signal
provides credible evidence of superior performance). Thus, if certified management standards
10
-------
act as a market signal, and if we observe any certification at all, we should expect higher
performing organizations to have a greater tendency to certify.
H2: The higher the environmental performance of an organization, the greater the
propensity for the organization to certify with the ISO 14001 management standard.
Certification and Monitoring
In the above discussion, we assumed that suppliers use certification with ISO 14001 to
signal about underlying performance attributes, and that buyers use certification to identify better
performing suppliers. The functioning of ISO 14001 as such a signal assumes that the attributes
about which certification communicates are stable. It also assumes that buyers can easily switch
suppliers and thus have a continuous need to identify high performing suppliers. In this section,
we discuss how suppliers may use certification to provide buyers with credible information about
performance improvements (rather than performance levels) in existing supply relationships.
Stakeholder-agency theory suggests that institutions for monitoring organizational
behavior are needed when stakeholder interests are not aligned with the interests of agents
(organizations), and when information asymmetries between stakeholders and agents prohibit
direct observation of agents' activities (Hill & Jones, 1992). In the arena of environmental
performance, incentives between suppliers and buyers may be misaligned, because some of the
cost of poor supplier performance is borne by the buyer. As discussed in the previous section,
environmental problems at the supplier can impose costs on buyers through supply disruptions
and spill-over reputation damage (Reinhardt, 1997).
As a result of these conditions, buyers have an incentive to encourage suppliers to
improve or maintain their environmental performance. Unfortunately, asymmetric information
problems may prevent the buyer from providing effective incentives to suppliers. Because
11
-------
environmental improvement efforts usually involve internal processes and management
procedures, buyers may be unable to observe actions at supplying organizations. Certification
with ISO 14001 may partially resolve this monitoring problem by providing a mechanism for
gaining credible evidence of a supplier's due diligence or performance improvements.
The need for monitoring a supplier increases the more the relationship between buyers
and suppliers is ongoing. When buyers can easily switch to new suppliers, the selection problem
(i.e., the problem of selecting high quality suppliers) is paramount and market pressures provide
incentives to suppliers to improve and signal performance. When buyers cannot easily switch to
new suppliers, however, buyers seek to motivate and monitor improvement efforts among
ongoing suppliers.
Joskow (1988) demonstrated that partner specific specialized assets cause switching costs
that determine the degree to which buyers and suppliers have an ongoing vertical relationship
(Joskow, 1988; Williamson, 1985). Idiosyncratic firm and facility level differences may
determine the extent of these costs. In many cases, however, industry level differences influence
the degree organizations tend to have partner specific assets and thus the tendency for these
organizations to have an ongoing relationship with supply chain partners (Maddigan, 1981).
These industry specific effects have been shown to be both wide-ranging and tractable to
measurement (Balakrishnan & Wernerfelt, 1986; Maddigan, 1981). When organizations in an
industry tend to have ongoing relationships with their suppliers, and if buyers use ISO 14001 to
solve monitoring problems among long term supply partners, we should expect:
H3: The more an organization is engaged in ongoing vertical relationships with its
buyers, the greater the propensity for the organization to certify with the ISO 14001
management standard.
12
-------
Suggesting that a certified management standard can help solve a monitoring problem
implies that certification will be associated with some desired organizational activity. In the case
of ISO 14001, it seems likely that stakeholders are seeking to monitor the existence of
environmental management systems (EMS) and that these systems improve environmental
performance. Note that implementation and certification of an EMS are distinct concepts and
may occur at different points in time. An EMS represents a set of procedures and guidelines that
systemize and control an organization's environmental management process. ISO 14001
stipulates a particular form of an EMS. To be certified with ISO 14001, an organization must
have a stated environmental policy, must determine and monitor the environmental impacts of its
activities, must set environmental objectives and measurable targets, must monitor actions and
take corrective actions where appropriate, and must continuously review this process. An
organization could have all of these elements and choose not to certify, it could have all of these
elements and choose to communicate their existence by certifying, or it might choose to adopt
the elements it lacks in order to certify.
We theorize that certification with ISO 14001 provides a means of credibly
communicating about the existence of a performance improving EMS. We are agnostic about
whether certified organizations are informing supply chain partners about the existence of a
previously adopted EMS, or whether they are adopting additional EMS activities in order to
certify. To the extent that certification follows the former logic we should expect that the
existence of an EMS is associated with performance improvement and that certification with ISO
14001 will simply reveal this. To the extent that certification follows the latter logic, we should
expect to see that ISO 14001 certification itself is associated with performance improvement.
H4a: Adoption of an environmental management system will improve an organization's
environmental performance.
13
-------
H4b: Certification with the ISO 14001 management standard will improve an
organization's environmental performance.
DATA & METHOD
We test our hypotheses by examining a sample of 8358 facilities (49413 observations)
drawn from the population of U.S. manufacturing facilities from the year 1995 to 2001. Facility
data were derived primarily from the U.S. EPA's Toxic Release Inventory (TRI) and Dun &
Bradstreet's (D&B) directory of facilities. We also gathered industry-level data from the Bureau
of Economic Analysis (BEA) and the Census Bureau of Foreign Trade. We gathered
demographic information from the Internal Revenue Service (IRS) and the Census Department.
Our sample is limited by the reporting requirements of the TRI. Facilities must report to the TRI
if their manufacturing processes generate scrap above certain levels and if they have more than
nine employees.
The most recent TRI data extends only to 2001, but data on ISO 14001 certification is
available through 2002. Because certification with ISO 14001 did not begin in earnest until 1996,
we limit our sample to the years from 1996 to 2002 for the dependent variables (1995 to 2001 for
the independent variables) in evaluating the propensity of facilities to certify. In analyzing the
effect of management practices and ISO certification on improvement, we extend the panel back
to 1994 to allow at least a two-year pretest window.
Measures
Dependent variable. The primary dependent variable for our analysis is certification to
the ISO 14001 environmental management standard. We gathered certification data from the
QSU database of ISO 14001 certified facilities (QSU, 2002a). Certification occurs at the facility
level. We coded ISO 14001 Certification as simply whether a facility is ISO 14001 certified
14
-------
during a particular annual period. ISO 14001 Certification takes a value of" 1" for all certified
facilities in a given year and "0" otherwise.
Independent variables. To test Hypothesis la, we measured the geographic distance
from a facility to the nearest major buyer (Distance to Buyers). To calculate this distance, we
first used TRI data to gather longitude and latitude information for each facility. We then used
the BEA input-output tables to determine the major (largest percentage) buying industry for each
selling industry. For each supplying facility (identified by its 4 digit SIC code), we then
calculated the great circle distance (in miles) to the nearest member of this buying industry. We
take the natural log of this measure to reduce its skew.1 To test Hypothesis lb, we created
Foreign Buyers. This variable measures the degree to which facilities in an industry sell to
buyers outside of the United States. It captures the percentage of all goods produced by
members of an industry that are shipped to buyers outside of the U.S. We used Input-Output
data from the BEA to create this variable.
To test Hypothesis 2, we calculated a facility's environmental performance using the
King & Lenox (2000) method of estimated relative pollution among facilities in an industry. The
method estimates the relationship between facility size and facility toxic waste generation in
each 4-digit Standard Industry Classification (SIC) code and year.2 We measured the
1 To ensure the robustness of this measure, we also calculated an alternative variable that measured the number of
such buyers within a 50 mile radius of the facility. Analysis of using the natural log of this count variable
confirmed the sign and significance of our results.
2 For any four-digit SIC Code level, if there was an insufficient number of facilities to estimate the production
function, we aggregated to the three-digit code. We were able to estimate production functions at the four-digit
level for 99% of the facilities.
15
-------
standardized residual, or deviation, between observed and predicted waste generation given the
facility's size and industry sector.
ln(Wit)= ap + PIjt ln(.v„) + (52]t ln(.vf, f + s]t (1)
Environmental Performanceit = -fy/ojt (2)
where Wit is the toxicity weighted sum3 of all Toxic Release Inventory waste generated by
facility i in year /, sit is facility size, and ccjt ,/?//(, and /??„ are the estimated coefficients for sector j
in year t. Size is measured using the number of employees working at facility i in year t. We
reversed the sign of the residual to reflect the fact that more waste than predicted for a facility
represents lower environmental performance.
To test Hypothesis 3, we measured Ongoing Vertical Relationship as the likelihood that a
facility is in a long-term relationship with its buyers. To create this variable, we adopted a
method similar to that developed by Maddigan (1981) and Balakrishnan & Wernerfelt (1986).
First, we used data from the BEA to identify pairs of supplying and buying industries. For each
supplying industry in each pair, we then used the entire 1997 D&B database (500,000 facilities)
to calculate the percentage of suppliers that was owned by a corporation that also owned a
facility in the buying industry.4 We then used shipment data from the BEA input-output tables to
3 To account for toxicity differences in facility waste generation, we weight the 246 toxic chemicals that have been
consistently reported in the TRI by their toxicity using the threshold "reportable quantity" (RQ) for an accidental
spill as required in the CERCLA statute (See King & Lenox, 2000). We then sum all of the toxicity-weighted was
created by a facility to calculate the total waste generation for the facility.
4 This ownership structure was updated for other years by tracking changes in ownership reported in the TRI.
16
-------
weight this percentage.5 We take the natural log of this weighted percentage value to reduce the
skew of our final variable. Thus, the final value estimates the log percentage of any dollar
produced by each industry (SIC code) that is shipped to a vertically integrated buyer. Previous
research suggests that this industry level variable approximates well an industry's propensity to
employ long-term contracts or have ongoing vertical relationships with buyers (Balakrishnan &
Wernerfelt, 1986; Maddigan, 1981).
To test Hypothesis 4a, we measured the existence of an operating environmental
management system by analyzing the reports of pollution reduction activity in the TRI (EMS).
As part of their annual TRI submission, facilities report changes they have made to the
production processes that could reduce waste or control pollution. The types of changes can be
broken into two main categories: 1) technical modifications and 2) changes in the environmental
management process. Facilities also report the sources of these technical changes. We coded
EMS as a binary variable indicating whether or not these sources provided evidence of
systematized environmental management practices. Sources of change that indicated evidence of
an operating EMS are: (1) internal pollution prevention opportunity audits, (2) materials balance
audits, (3) participative team management, (4) employee recommendations under a formal
company program.
Control variables. Experience with related management standards has previously been
shown to influence the tendency for an organization to certify with the ISO 14001 environmental
management standard (King & Lenox, 2001). Previous experience may increase a facility's
absorptive capacity with respect to management standards. This would allow adoption at lower
5 Some supplying-buying industry pairs have greater interaction (according to dollar values shipped) then other
pairs, and these differences must be captured to account more accurately for the supply chain relationships.
17
-------
costs and thereby increase adoption propensities. To account for this tendency, we measured
whether a facility participates in the Responsible Care Program. The Responsible Care Program
is sponsored by the American Chemistry Council and, like ISO 14001, requires the establishment
of environmental management practices. We captured program participation using a binary
variable (RC Member) that indicates if that facility was owned by a firm that participated that
year in the Responsible Care Program. We also created a binary variable (ISO 9000 Certified)
that is coded such that a "1" indicates any year in which the facility is certified with the ISO
9000 quality management standard. ISO 14001 was modeled after ISO 9000, and the structural
resemblance of the two standards may facilitate certification with ISO 14001 subsequent to
certification with ISO 9000. We gathered ISO 9000 certification data from the ISO 9000
Registered Company Directory of North America (QSU, 2002b).
Supply chain pressures could influence the tendency of facilities to adopt environmental
management practices and to certify. These supply pressures could emanate from both waste
and product streams. More specifically, 'buyers' of waste may request their suppliers to adopt
environmental practices and certify with ISO 14001 with the expectation that this would make
the supplying facility's waste more predictable and less toxic, thereby facilitating waste
treatment. To capture the pressures from waste stream partners, we created two binary variables.
Offsite Waste Transfer indicates whether or not the facility transfers waste to an offsite waste
processor that either recycles or treats the waste. POTW Waste Transfer measures the potential
for regulatory pressure from Publicly Owned Treatment Works (POTW). To create the measure,
we determined if the facility sent any waste material to a POTW in each year. A value of "1"
indicates evidence of a physical connection to the POTW. To capture the pressures emanating
from product supply streams, we created Auto Supplier, which is a binary variable that indicates
18
-------
whether or not the facility sells products to automobile assemblers. Ford, GM, and Toyota have
all announced that they will give preference to ISO 14001 certified facilities.
Regulatory and stakeholder pressures could also influence the propensity to certify with
ISO 14001. To account for these, we created several other control variables. Industry Waste
Generated measures the degree to which an industry generates toxic waste (and thus is likely to
be the target of regulation and stakeholder pressure). It is measured as the mean of the natural
log of the toxicity weighted waste generation for all facilities within each 4-digit SIC code.
Regulatory Pressure measures the stringency of state-level environmental regulation. It is
constructed using a measure devised by Meyer (1995) based on the logged aggregate emissions
per state over the sum of the Gross State Product in four polluting sectors (chemicals, pulp &
paper, textiles, and petroleum products). Research has also shown that local stakeholder pressure
is related to the affluence of the surrounding community (Walsh, Rex, & Smith, 1993). To
measures the Affluence of citizens in the area surrounding a facility, we calculated the annual
average local income using IRS data on the 5-digit zip code area. Scholars have argued that the
Responsible Care initiative could reduce stakeholder pressure on an industry by reducing the
likelihood of regulatory action. To control for this potential effect we also measured the annual
percentage of the facilities in the industry (RC Industry) that participate in the Responsible Care
initiative.
Finally, a number of firm and facility attributes could influence a facility's decision to
certify. A facility's size could influence the availability of resources and thus its propensity to
adopt an environmental management system or certify with ISO 14001. We measure Facility
Size as the normalized (by industry and year) log of the number of employees at that facility.
Foreign ownership could also influence the propensity for certification. Foreign parents may use
19
-------
certification as a means to monitor their overseas facilities. Foreign parents may also require
certification of their international facilities in an attempt to standardized practices across
facilities. We created a binary variable that measures whether a U.S. facility is owned by a
foreign parent (Foreign Owned). Foreign ownership was determined using D&B's Who-Owns-
Whom dataset. In some cases, the database did not list a nationality. For these, we individually
verified the nationality of the ultimate parent. We coded the variable Foreign Owned to be "1" if
the ultimate parent firm is non-U.S. owned, "0" if it is U.S. owned. Common corporate
ownership of buyers and suppliers could influence the propensity for certification since vertical
integration can reduce market incentives and thereby increase the need for monitoring.
Alternatively, common ownership may facilitate information transfer between supplier and
buyer, thereby reducing the need for certification. We created a binary variable, Vertically-
Integrated Buyer, to capture these potential effects. The variable takes on a value of "1" if at
least one potential buyer of the facility's output (as determined by the BEA input-output tables)
has the same corporate parent as the facility. Finally, Firm Size measures the annual count of the
number of facilities owned by the target facility's parent. The count is logged to reduce the skew
of the distribution.
Table 1 summarizes our measures and provides the pair wise correlation between
variables.
Insert Table 1 about here
Method
Our analysis requires evaluation of a facility's propensity to certify with the ISO 14001
standard. It also requires that we evaluate the effect of environmental management practices.
20
-------
For the first analysis, we use a discrete time random effect logistic model. For each facility, we
predict certification with ISO 14001. As soon as a facility is certified, we no longer consider it
in our sample, as it is no longer at risk to certify. The model is specified as:
Plt+i = F(Z) = F(a; + bXlt) = e(ZltV(l + e(Zlt})
where P is the probability that facility i will certify with ISO 14001 in the next year (t+1). The
vector X;t represents the characteristics of the ith facility in year t. The facility random effects are
measured as a;. We use a random, rather than a fixed effect specification because the fixed effect
model would disregard all observations that do not certify with ISO 14001 within our panel.
Furthermore, a fixed effect specification would prohibit the interpretation of any variables with
values that do not vary across groups (or time). To investigate the robustness of our model
specification, we also employed a maximum likelihood proportional hazard model (with an
exponential base-line hazard) and a Cox's non-parametric partial-likelihood estimation
procedure. The Cox estimation is inefficient, but does not require specification of a particular
functional form of the base line hazard. These robustness checks generate results that confirm the
reported ones in coefficient sign and significance.
Our model includes a potential selection problem. It is possible that some unobserved
disturbance causes both the decision to adopt an EMS and to certify with ISO 14001. For
example, organizations with a particular culture or leadership might tend to adopt both. Even if
EMS is included in a second stage regression, this disturbance term will tend to bias coefficient
estimation. Unfortunately, solving this problem in a logistical regression analysis of panel data
is on the frontier of statistical knowledge. For all but a few cases of simple attrition, correcting
for selection in panels longer than two periods remains impractical (Honore & Kyriazidou, 2000;
Kyriazidou, 2001).
21
-------
To address the selection problem, we therefore chose to shrink the panel to a cross
section to eliminate the panel analysis problem and to allow use of estimation techniques with
normally distributed disturbance terms. We use the approach developed by van de Ven and van
Praag (1981) which specifies a selection model (adoption of an EMS) and a probit model
(certification with ISO 14001).
Prob(ISO=l) = prob(Bx; + un > 0) (4)
Prob(EMS=l)=prob(Zxi + u2; >0) (5)
where B & Z are separate coefficient vectors and x; is our set of explanatory variables. The two
disturbance terms uh and U2i are assumed to be bivariate normally distributed but correlated p.
Using methods developed by Heckman (1979) and van de Ven and van Praag (1981), both the
coefficients and this correlation can be calculated either through a two stage procedure or
through a single maximum likelihood estimation. We employ the second approach.
Finally, we use a differences-in-differences approach to analyze the effect of
environmental management systems on environmental performance. This approach vastly
reduces the propensity for unobserved organizational attributes to bias estimates and cause
spurious findings. Specifically, we estimate:
y; (t+1) = B[yi(t),Xi(t)] + 5; + 8; (6)
where i index the facilities, y, is the facility's environmental performance, B is a vector of
estimated coefficients, x, is a vector of measured facility level attributes, 8; is dummy variable
capturing unmeasured facility fixed attributes, and s, is the error term. Because of the lagged
independent variable, this formulation is prone to autocorrelation. We use a method developed
by Anderson and Hsiao (1982) to correct for this potential problem.
22
-------
Our sample is a large one and this can cause an inflated tendency to reject the Null
hypothesis. Previous research using large samples has tended to correct for this by reporting
significance only for p < 0.01 and p < 0.001.
ANALYSIS & RESULTS
Analysis of Certification
Table 2 presents the first part of our analysis of the causes of certification with the ISO
14001 environmental management standard. Model 1 presents estimates for a baseline
specification that includes only our control variables. The estimates suggest that the propensity
to certify with ISO 14001 is greater in the presence of related practices (ISO 9000 Certified),
supply chain pressures (POTW Waste Transfer, Auto Supplier), and larger facility and firm size.
Interestingly, RC Industry is associated with a lower propensity to certify, suggesting that a high
degree of participation in Responsible Care among firms in an industry may reduce the need for
ISO 14001 certification. Furthermore, facilities that have foreign parents (Foreign Owned) are
more likely to certify with ISO 14001. We also find that vertical integration between a facility
and its buyers (Vertically-IntegratedBuyer) increases the propensity to certify. Finally, facilities
with existing environmental management systems (EMS) are more likely to certify, presumably
to take credit for previously pursued activities. (Note that we include EMS as a control variable
in Table 2. EMS will become an independent variable as we test Hypothesis 4a).
In Model 2, we add our measures capturing the likelihood of information asymmetries
and the need for signaling and monitoring. The addition of these independent variables in Model
2 provides a significant increase in the explanatory power over the base case (as indicated by a
significant incremental %2 test). Coefficients estimated in Model 2 support Hypotheses la, lb,
and 3. Consistent with Hla, we find that the propensity for a facility to certify with ISO 14001
23
-------
increases with greater distance between the facility and its buyers. Consistent with Hlb, we find
that a facility's propensity for certification increases with the tendency of the industry to export
to foreign buyers6. Taken together, these results suggest that certification with ISO 14001 is
more likely if information asymmetries in the supply chain are high, thereby supporting our
proposition that facilities use certification to reduce asymmetric information with buyers.
Insert Table 2 about here
Turning now to whether facilities use ISO 14001 to help resolve problems of asymmetric
information in selection or monitoring, we find support only for the monitoring hypothesis. We
find no significant evidence that superior environmental performance (i.e., relative facility waste
generation) positively influences certification propensities. Thus, we have no evidence that ISO
14001 is operating as a signaling mechanism. We do find consistent evidence that ISO 14001
certification may act as a monitoring device. With respect to Hypothesis 3, we find that the
coefficient for Ongoing Vertical Relationship is positive and strongly significant, suggesting that
the greater the likelihood that a facility is in an ongoing vertical relationship with its buyers, the
higher the propensity for ISO 14001 certification.
Interestingly, our analysis provides hints that asymmetric information between corporate
parents and local facilities may also be an important driver of certification. Throughout our
analysis, we find evidence that foreign owned facilities are more likely to certify with ISO
6 To ensure that this effect was not caused by exports to particularly environmentally sensitive regions, we
investigated the effect of exports to different regions. We could find no evidence that exports to Europe,
Australia, Asia, or Central America had a different effect on certification than exports to North America (Canada
and Mexico).
24
-------
14001. Information asymmetries between foreign parents and domestic facilities may have
caused facility managers to seek certification in order to signal to corporate parents about their
management abilities. We also find that Vertically-Integrated Buyer - one of our control
variables - consistently and significantly increase certification propensities. This suggests that
suppliers that are vertically integrated (i.e., commonly owned) with their buyers are more likely
to certify with ISO 14001. This finding lends further support to Hypothesis 3. It suggests that
the move towards integrated governance structures (like long term contracts as hypothesized in
H3 or hierarchy as captured by this control variable) is associated with greater monitoring needs
(due to higher switching costs and lack of market incentives), thereby triggering certification.
To explore the potential for confounding unobserved industry effects, we include two-
digit SIC code fixed-effects in Model 3. Our results are consistent with Model 2. With the
exception of Foreign Buyer, the coefficients for the variables of concern remain significant at a
minimum of p<0.01. The inclusion of industry fixed-effects does reduce the significance of
Foreign Buyer but does not change the coefficient estimate. Since Foreign Buyer is an industry
level variable (calculated on the 4-digit SIC code level), co-linearity with industry fixed-effects
may cause its significance to drop if the variable varies little across two-digit SIC codes. Note
that an incremental %2 test suggests that the inclusion of industry fixed-effects does not improve
model fit, suggesting that we have no evidence of other unobserved industry effects - at least as
captured at the two digit SIC code.
Separating Adoption and Certification
Similar industry and organizational attributes might determine the propensity both to
adopt environmental management practices (systems) and to certify with ISO 14001. If we fail
25
-------
to capture in our right hand side variables some factors that explain both adoption and
certification, our analysis of the causes of certification could be biased. To account for these
missing factors and correct this problem, we perform a two-stage estimation. In the first stage,
we estimate the facility and industry attributes that are associated with a higher propensity to
have an environmental management system (EMS) in place.7 In the second stage, we estimate
which facilities are likely to certify with ISO 14001. This method also has the appealing
property that it allows us to analyze the factors that cause adoption of EMS practices and
compare these factors with those that cause certification.
As explained in the methods section above, this two-stage estimate requires us to collapse
our data into a cross-section so that we can use a probit specification. To check that such
conversion from a panel to a cross-section does not change our results, we first estimate a model
similarly to Model 3 (in Table 2) using our collapsed panel. We report the results of this Model
(Model 4) in Table 3. Note that because some facilities enter the panel after 1996, collapsing the
panel reduces our data set to 7899 facilities. The estimates from our probit model in Model 4 are
similar in direction and significance to those in Model 3 in Table 2. We continue to find a
significant positive effect for Distance to Buyer, Foreign Buyer, Ongoing Vertical Relationship,
Vertically-Integrate Buyer, and EMS. Our estimate for Environmental Performance remains
negative and is now significant at the p < 0.01 level.
Insert Table 3 about here
We present the results of our two-stage analysis in Model 5 in Table 3. The first column
7 We use the idea of stages for expository convenience. In the actual analysis the two stages are calculated
simultaneously using a maximum likelihood estimator.
26
-------
reports the likelihood that a facility will have an EMS. The estimates from this "selection
model" are then used to correct for potential unobserved attributes that might bias our estimates
in the second stage probit model. Due to the particular structure of this technique, the second
stage estimates reported in Model 5 for ISO 14001 certification are based only on data from
those facilities that had an EMS prior to 1996 (reducing our sample to 3300 facilities). An
alternative specification using only those that did not have such an EMS delivered similar
results.
Estimates from this model (see the far right column in Table 3) confirm the findings
presented in Table 2. We again find support for both HI a and Hlb - facilities with more distant
and more foreign buyers are more likely to certify. We again find support for our hypothesis that
certification will be higher for facilities in industries with ongoing vertical relationships with
buyers (H3). Finally, we find strong evidence to disconfirm Hypothesis 2 - facilities with lower
(not higher) environmental performance are more likely to certify.
Our two-stage approach allows us to differentiate the sources of EMS adoption from the
causes of certification. Comparing the two columns of Model 5, some important differences
appear. In general, we find that technical, regulatory, and experience differences explained
differences in the tendency to have an EMS. In contrast, factors influencing the need for
communication with buyers about improvement efforts influenced certification.
Specifically, as shown in Table 3 column 2, technical and regulatory differences strongly
influence EMS adoption (but less so certification with ISO 14001). Regulatory Pressure and
Industry Waste Generation have a significant effect on implementing an EMS. Likewise,
facilities that transfer waste to public or private outside processors (Offsite Waste Transfer,
POTW Waste Transfer) tend to have an EMS. This pattern of results seems to suggest that
27
-------
facilities facing greater demand from regulators or waste handling stakeholders tend to adopt an
EMS. This may suggest greater ability among these stakeholders to directly monitor EMS
activities and environmental performance. In contrast, only Offsite Waste Transfer has a
significant effect on ISO 14001 certification. Unlike in Model 3, POTW Waste Transfer is no
longer significant, suggesting that the results found in Table 2 may have been confounded by an
unobserved EMS selection problem.
Consistent with previous studies (King & Lenox, 2001), we find that ISO 9000
certification is associated with the adoption of environmental practices - in this case a
functioning EMS. We also find evidence that Responsible Care Participants are more likely to
have an operating EMS, but can find no evidence that these facilities have a higher propensity to
certify with ISO 14001. This seems to suggest that participants in the Responsible Care program
are indeed implementing some of the associated environmental management practices.
However, we find evidence that facilities in industries with many RC participants have a lower
propensity to certify with ISO 14001 — suggesting that conflicts or substitution exist between the
two programs.
Our results suggest that organizations with lower relative environmental performance are
more likely to implement an EMS. The implementation of a formal management system may
reflect a desire to catch up among industry laggards. Interestingly, we find that suppliers that
tend to have ongoing vertical relationships with their buyers are less likely to adopt an EMS.
This is consistent with our hypothesis that without a credible means of monitoring actions at
suppliers, buyers have difficulty encouraging real investments in performance improvement.
Our other independent variables (.Distance to Buyers, Foreign Buyer, and Vertically Integrated
Buyer) have no significant effect on adopting an EMS.
28
-------
Performance Improvement Analysis
We can now turn to testing our final hypotheses in support of a theory that ISO 14001
allows buyers to monitor improvement efforts at suppliers. To analyze whether EMS adoption
or ISO certification is associated with improvement (H4a and H4b), we specify a cross sectional
time series regression predicting waste generation in the next year (see Model 6 in Table 4). We
include facility and year fixed effects to control for underlying facility heterogeneity and time
effects. Because we also control for this year's waste generation, this model represents a form of
the highly conservative and robust differences-in-differences approach. We also include a log
count of the number of non-management source reduction changes reported for each facility
{Pollution Reduction Activity). This variable uses TRI information.8 To ensure robustness, it was
coded in three different ways: 1) as a binary variable, 2) as a count variable, and 3) as a log count
variable. The reported results use the last formulation but the sign and significance of the
reported results are robust to all formulations. Including this variable helps to ensure that our
measure of EMS does not simply reflect change activities.
We find that the existence of an EMS in year t is associated with significant increases in
environmental performance in year t+1 (H4a). With respect to Hypothesis 4b, we do not find
significant evidence that certification is associated with such improvement. The coefficient for
ISO 14001 certification is significant at only the p<0.05 level. In a sample of this size, such a
finding must be viewed with great caution. Thus, we have strong support that adoption of an
8 Source of change that indicated technological change activity included external pollution prevention opportunity
audits, employee recommendations independent of a formal company program, state and federal government
technical assistance programs, trade association/industry technical assistance programs, and vendor assistance.
29
-------
EMS provides evidence of improvement, but we do not find evidence that certification itself is
associated with improvement.
Insert Table 4 about here
This lack of evidence may suggest that ISO certification does not itself cause
improvement but merely provides credible evidence of an underlying and possibly pre-existing
EMS that causes improvement. We should be careful, however, not to commit a type-II error
and confuse a lack of a finding with disconfirming evidence. The short timeframe over which
most organizations have been certified makes it very difficult to estimate ISO 14001 generated
improvements. For now, all we can say is that a facility's certification with ISO 14001 is
associated with having an EMS (both logically and statistically), and having an EMS is itself
related to improvement.
DISCUSSION
In summary, we find evidence that organizations certify with ISO 14001 to reduce
information asymmetries with supply chain partners. In particular, we find that geographically
and culturally remote suppliers are more likely to seek certification. We do not find evidence
that certification serves as a signal of superior environmental performance. Rather, our findings
suggest that suppliers use certification to communicate about environmental improvement
efforts. Specifically, we find that in the face of high switching costs and in the absence of
market incentives - aspects that are inherent in more integrated governance structures like long
term contracts or vertical integration - certification may serve to fulfill the greater need for
monitoring supplier behavior.
30
-------
Our results are robust to a large number of controls and specifications. We attempt to
control for unobserved heterogeneity by including industry and year fixed-effects. We utilize a
two-stage selection model to address potential concerns about self-selection. While we only
investigate one certified management standard, our sample covers a wide number of industries,
which gives us greater confidence in the external validity of our findings.
Despite our conservative approach to our analysis and our robustness tests there are
reasons to interpret our findings cautiously. Scale and chemical emission thresholds for
reporting to the Toxic Release Inventory could cause a potential sample selection problem. Our
sample may fail to pick up facilities with superior environmental performance that are not
required to report to the TRI. We have investigated this problem statistically and believe our
results to be robust. Nevertheless, we believe care should be exercised in extrapolating our
findings to predict the behavior of firms of all sizes and industries.
Another potential confound is that we measure the existence of an EMS through a
facility's report on pollution reduction activities. This could cause a measurement error for
facilities that have an environmental management system in place but do not routinely make
changes to production processes or that have made a number of pollution reducing improvements
in the past and no longer have the need to further reduce pollution levels. Fortunately, the effect
of this bias should be conservative, because it should make it harder to find a relationship
between adoption of an EMS and improvements in environmental performance.
Finally, ISO 14001 is still in its relatively early stages of diffusion. As the standard
diffuses and more facilities seek certification, the profile of those seeking certification may
change. In particular, as the number of ISO 14001 certifications rises, the pressure on non-
certifiers to certify will likely increase. As these pressures increase, the marginal costs and
31
-------
benefits will shift such that organizations that face relatively low information asymmetries may
seek certification. While this does not contradict our fundamental thesis that the desire to
monitor and communicate about behavior is driving certification decisions, it suggests caution in
extrapolating our analysis to all temporal periods of the adoption process.
CONCLUSION
In this article, we explore the drivers of ISO 14001 certification and we answer specific
questions about how this certified management standards may function. We theorize that
certification with ISO 14001 provides one way of reducing asymmetric information. We
develop hypotheses for when organizations will use certification to communicate credible
information to supply chain partners. We then explore what kind of information is conveyed by
certification.
Our findings suggest that organizations certify with ISO 14001 to overcome problems of
asymmetric information. We find that certification does not provide evidence of superior
organizational performance (as expected by many of its creators). Instead, we find evidence that
suppliers use certification to communicate improvement efforts to long-term supply chain
partners (buyers).
Our research should not be interpreted to support an overly simplified functionalist notion
of the ISO 14001 standard. Evidence suggests that many of the framers of this new standard
expected it to serve a different social purpose than it came to have. Many expected ISO 14001 to
provide a means of credibly differentiating organizations with better environmental performance
(Mazza, 1996). Our analysis suggests that this expectation went unfulfilled. Yet it also suggests
that the standard came to play an alternative functional role.
32
-------
Our research emphasizes a fundamental paradox in the design of certified management
standards like ISO 14001. It suggests that standards that include beneficial practices may seldom
act as a market signal. For a certified management standard to be useful as a market signal, high
performing organizations must benefit from certification, while low performers must not. If low
performers gain significant operational benefits from certifying, this condition will not hold.
Moreover, if supply chain partners target their incentives to organizations where improvement
can most easily be achieved, they may tend to encourage the worst performers to adopt and
certify. Thus, our research suggests that the more an environmental management standard
provides direct operational benefits, the less likely it will provide a means of signaling superior
performance.
Our findings suggest avenues for future research. First, our analysis suggests the need for
additional consideration of the design of management standards. In future research, we hope to
analyze other attributes that cause standards to function either as a tool for improvement or as a
means to signal. Second, future research could extend our analysis to other stakeholder relations.
Our study focused on the potential of management standards to address problems of asymmetric
information within supply chains but provides hints that management standards may also play a
critical role in reducing information asymmetries within firms. We found that facilities in larger
corporations and those that were foreign owned tended to certify with ISO 14001. This may
suggest that managers use certification as a signal in the market for corporate resources, or that
corporate parents employ certification as a means to monitor their facilities. Future research
should assess how certified management standards are used within firms, and whether internal
use alters their function.
33
-------
REFERENCES
Adams, J.D. 2002 Comparative localization of academic and industrial spillovers. Journal of
Economic Geography 2(3): 253-278
Akerlof, G. 1970. The markets for Lemons: Quality uncertainty and the market mechanism.
Quarterly Journal of Economics. 488-500.
Allen, T. J., Lee, D. M. S., & Tushman, M. L. 1980. R&D Performance as a Function of Internal
Communication, Project-Management, and the Nature of the Work. IEEE Transactions
on Engineering Management, 27(1): 2-12.
Anderson, S. W., Daly, J. D., & Johnson, M. F. 1999. Why firms seek ISO 9000 certification:
Regulatory compliance or competitive advantage? Production & Operations Management,
8(10): 28-43
Anderson, T. and C. Hsiao 1982. Formulation and Estimation of Dynamic Models Using Panel Data.
Journal of Econometrics 18: 47-82.
Balakrishnan, S. & Wernerfelt, B. 1986. Technical Change, Competition and Vertical Integration;
Summary. Strategic Management Journal, 7: 347-350.
Bansal, P. & Hunter, T. 2003. Strategic explanations for the early adoption of ISO 14001. Journal of
Business Ethics. 46(3): 289 - 299.
Buckley, P. J. & Casson, M. 1979. A theory of international operations. Amsterdam-London:
North-Holland.
Caves, R. E. 1982. Multinational Enterprise and Economic Analysis. London, U.K.:
Cambridge University Press.
Collins, B. L. 1996. Prepared statement before the house science committee subcommittee on
technology, Federal News Service, June 4, Section: In The News.
Corbett, C. J. & Kirsch, D. A. 2001. International Diffusion of ISO 14000 Certification.
Production & Operations Management, 10(3): 327-342.
Delmas, M. A. 2002. The diffusion of environmental management standards in Europe and in the
United States: An institutional perspective. Policy Sciences, 35(1): 91-119.
DiMaggio, Paul J. 1988. Interest and agency in institutional theory. In L.G. Zucker (Ed.),
Institutional Patterns and Organizations: Culture and Environment. Cambridge,
MA: Ballinger: pp 3-21.
Ferrer, G. Gavronski, I. and de Laureano Pavia, P. 2003. ISO 14001 Certification in Brasil:
Motivation and Benefits. Working Paper, Kenan-Flagler Business School
34
-------
Granovetter, M.S. (1985). Economic action and social structure: the problem of embeddedness.
American Journal of Sociology, 91: 481-510.
Guler, I., Guillen, M. F., & Macpherson, J. M. 2002. Global Competition, Institutions and the
Diffusion of Organizational Practices: The International Spread of ISO 9000 Quality
Certificates. Administrative Science Quarterly, 47: 207-232.
Hamilton, I., Godfrey, F. E., & Linge, J. R. 1979. Industrial System. In I. Hamilton & J. R. Linge
(Eds ), Spatial Analysis, Industry and the Industrial Environment. New York: John
Wiley.
Heckman, J. 1979. Sample selection bias as a specification error. Econometrica, 47: 153-161.
Hill C. & Jones T. 1992. Stakeholder-Agency Theory. Journal Management Studiesx29(2):
131-154
Hoffman, A. 1997. From Heresy to Dogma: An Institutional History of Corporate
Environmentalism. San Francisco: The New Lexington Press.
Hofstede, G. 1980. Culture's Consequences: International Differences in Work-related
Values. Beverly Hills: Sage Publications.
Honore, B.E. and Kyriazidou, E. 2000. Panel Data Discrete Choice Models with Lagged Dependent
Variables. Econometrica, 68(4): 839-874.
Ingram, P. & Silverman, B. S. 2002. The New Institutionalism in Strategic Management. In P.
Ingram & B. S. Silverman (Eds.), Advances in Strategic Management, Vol. 17.
Jiang, R. J. & Bansal, P. 2003. Seeing the Need for ISO 14001. Journal of Management
Studies, 40(4): 1047-1067.
Joskow, P. L. 1988. Asset specificity and the structure of vertical relationships: empirical
evidence. Journal of Law Economics and Organization, 4(1): 95-117.
Katz, R. & Tushman, M. 1979. Communication Patterns, Project Performance, and Task
Characteristics: An Empirical Evaluation and Integration in an R&D Setting.
Organizational Behavior and Human Performance, 23: 139-162.
King, A. 1999. Retrieving and transferring embodied data: Implications for the management of
interdependence within organizations. Management Science, 45(7): 918-935.
King, A. & Lenox, M. 2000. Industry Self-Regulation without Sanctions: The Chemical
Industries Responsible Care Program. Academy of Management Journal, 43(4): 698-
716
King, A. and Michael Lenox. 2001. Lean and Green: Exploring the Spillovers from Lean Production
to Environmental Performance. Production and Operations Management, 10(3): 1-13.
35
-------
Kogut, B. & Singh, H. 1988. The effects of national culture on the choice of entry mode.
Journal of International Business Studies, 18(3): 411-432.
Kyriazidou, E. 2001 Estimation of Panel Data Sample Selection Models. Review of Economic
Studies, 68: 543-572.
Lane, C. & Bachman, R. 1996. The Social Constitution of Trust: supplier Relations in Britain
and Germany. Organization Studies, 17(3): 365-395.
Maddigan, R. J. 1981. The Measurement Of Vertical Integration. Review of Economics &
Statistics, 63(3): 328-336.
Mariotti, S. & Piscitello, L. 1995. Information costs and Location of FDIs within the Host
Country: Empirical Evidence of Italy. Journal of International Business Studies, 26(4):
815-841.
Marquardt, D. 1992. ISO 9000: A universal standard of Quality. Management Review, 81: 50-
52.
Mazza, S. 1996. International Standards and US Industry, Hearing Of The Technology
Subcommittee Of The House Science Committee, Sergio Mazza ed. Washington, DC:
Federal News Service.
Meyer, S. 1995. The Economic Impact of Environmental Regulation. Journal of Environmental
Law and Practice, 3(2): 4-15.
Morella, C. 1996. International Standards and US Industry, Hearing Of The Technology
Subcommittee Of The House Science Committee, Sergio Mazza ed. Washington, DC:
Federal News Service.
Naveh, E., Marcus, A., Allen, G., & Koo Moon, H. 1999. ISO 9000 Survey '99: An Analytical
Tool to Assess the Costs, Benefits and Savings of ISO 9000 Registration. New York:
MacGraw-Hill, Inc.
OTA. 1986. Serious Reduction of Hazardous Waste: Office of Technology Assessment.
Government Printing Office, Washington, DC
QSU 2002a. ISO 14001 Registered Company Directory North America. Fairfax, VA, QSU
Publishing Company.
QSU 2002b. ISO 9000 Registered Company Directory North America. Fairfax, VA, QSU
Publishing Company.
Reinhardt, F. 1997. Market Failure and the Environmental Policies of Firms. Journal of Industrial
Ecology, 3(1): 9-21.
Russo, M.V. and Fouts, P. 1997. A resource-based perspective on corporate environmental
performance and profitability. Academy of Management Journal 40(3): 534-560.
36
-------
Silverman, B. S., Nickerson, J. A., & Freeman, J. 1997. Profitability, Transactional Alignment
and Organizational Mortality in the U. S. Trucking Industry. Strategic Management
Journal, 18(Summer): 31-52.
Slawsky, R. 2004. High crude oil prices pinch drivers at pump. New Orleans City Business New
Orleans, LA, April 12.
Snir, E. M. 2001. Liability as a Catalyst for Product Stewardship. Production and Operations
Management, 10(2): 190-206
Spence, A. M. 1973. Job Market Signaling. Quarterly Journal of Economics, 87(3): 355-375.
Staw, B. M. & Epstein, L. D. 2000. What Bandwagons Bring: Effects of Popular Management
Techniques on Corporate Performance, Reputation, and CEO Pay. Administrative
Science Quarterly, 45: 523-556.
van De Ven, W. P. & B. M. Van Praag 1981. The Demand for Deductibles in Private Health
Insurance: AProbitModel with Sample Selection. Econometrica, 17: 229-252.
Walsh, E., Rex W., and D. Clayton Smith 1993. Backyards, NIMBYs, and Incinerator Sitings:
Implications for Social Movement Theory. Social Problems, 40:25-38.
Westphal, J.D., Gulati, R., & Shortell, S. 1997. Customization or conformity? An institutional and
network perspective on the content and consequences of TQM adoption. Administrative
Science Quarterly, 42(2): 366-394.
Williamson, O. 1985. The Economic Institutions of Capitalism: Firms, Markets, Relational
Contracting. The Free Press: New York.
Zaheer, A., McEvily, B., & Perrone, V. A. 1998. Does Trust Matter? Exploring the Effects of Inter
organizational and Interpersonal Trust on Performance. Organization Science, 9: 141-159.
37
-------
TABLE la
Descriptive Statistics
Variable
Description
Mean Std Dev Min Max
Distance to Buyers
Foreign Buyers
Environmental Performance
Ongoing Vertical Relationship
EMS(t-i)
Responsible Care Participant
ISO 9000 Certified
Offsite Waste Transfer
POTW Waste Transfer
Auto Supplier
Industry Waste Generation
Regulatory Pressure
Affluence
RC Industry
Facility Size
Foreign Owned
Vertically-Integrated Buyer
Firm Size
Log of geographic distance to nearest 2.23
buyer
Log percent of industry production 1 42
shipped to buyers outside the U.S.
King & Lenox (2000) measure of -0.06
environmental performance
Log of industry percentage of suppliers 0.37
with vertically integrated buyers.
Binary variable indicating existence of 0.46
an environmental management system
Binary variable indicating facility 0.09
owned by a member of Resp. Care
Binary variable indicating ISO 9001 0.25
certification
Binary variable indicating the facility 0.83
transfers waste offsite (not to POTW).
Binary variable indicating a facility is 0.34
connected to public water treatment
Binary variable indicating facility 0.06
supplies the automotive industry
Log average total waste generation for 4.84
sector in which the facility operates
The regulatory stringency of the 0.13
facility's state.
Average family income within the 10.2
facility's zip code.
Percentage of facilities in industry that 0.09
are owned by RC members.
Natural log of facility employees. 0.07
(normalized by industry and year)
Binary variable indicating that a facility 0.04
is foreign owned
Binary variable indicating a potential 0.58
buyer shares the same corporate parent
Count of firm facilities 1 -49
1.29 0 6.22
0.75 0 4.06
0.98 -4.34 4.08
0.34
0.50
0.29
0.43
0.37
0.47
0.24
0.12
0.49
1.43
0 2.39
1.46 1.15 11.89
0.02 0.11 0.2
0.28 6.16 13.1
0 0.67
0.95 -4.79 5.73
0.20
0 5.32
n= 49413
38
-------
TABLE lb
Descriptive Statistics: Correlations
Variable
1
10 11 12 13 14 15 16 17 18
1. Distance to Buyers
2. Foreign Buyers
3. Environmental Performance
4. Ongoing Vertical Relationship
5. EMS(t_i)
6. Responsible Care Participant
7. ISO 9000 Certified
8. Offsite Waste Transfer
9. POTW Waste Transfer
10. Auto Supplier
11. Industry Waste Generation
12. Regulatory Pressure
13. Affluence
14. RC Industry
15. Facility Size
16 .Foreign Owned
17. Vertically-Integrated Buyer
18. Firm Size
1.00
0.00 1.00
-0.01 0.00 1.00
-0.05 0.26 0.01 1.00
0.01 0.11 -0.11 0.02 1.00
-0.04 0.13 -0.04 0.15 0.14 1.00
0.01 0.17 -0.04 0.09 0.11 0.12 1.00
-0.02 0.01 -0.17 -0.04 0.13 0.03 0.10 1.00
-0.03 0.06 -0.05 0.02 0.12 0.02 0.09 0.32 1.00
0.09 0.12 -0.01 -0.03 0.02 -0.06 0.03 0.05 0.06 1.00
0.00 0.21 0.02 0.46 0.13 0.24 0.03 0.04 0.04 0.00 1.00
-0.16 -0.07 0.05 -0.05 0.01 -0.05 -0.01 0.06 0.05 -0.04 -0.09 1.00
-0.12 0.04 0.01 -0.03 0.04 0.02 0.04 0.05 0.04 0.00 -0.02 0.15 1.00
-0.07 0.29 0.00 0.33 0.13 0.44 0.07 -0.02 0.01 -0.13 0.55 -0.09 0.02 1.00
0.03 0.01 0.00 0.00 0.16 0.12 0.16 0.14 0.13 0.00 0.00 -0.04 0.04 0.00 1.00
-0.02 0.04 -0.02 0.03 0.06 0.05 0.04 0.05 0.01 -0.02 0.06 -0.02 0.05 0.20 0.04 1.00
0.07 0.20 -0.07 0.15 0.12 0.23 0.16 0.10 0.06 0.08 0.14 -0.11 0.00 0.15 0.21 0.12 1.00
0.06 0.17 -0.06 0.15 0.14 0.34 0.13 0.09 0.07 0.10 0.18 -0.13 -0.01 0.21 0.22 0.14 0.58 1.00
n= 49413
39
-------
TABLE 2. Predicting Certification with ISO 14001,1996-2002
(Discrete Time Random Effect Logistic Model)
Model 1 Model 2 Model 3
Distance to Buyer
Foreign Buyer
Environmental Performance
Ongoing Vert. Relationship
0.09
(0.03)
0.19
(0.06)
-0.09
(0.04)
0.51
(0.13)
*
*
**
0.08
(0.03)
0.15
(0.07)
-0.09
(0.04)
0.66
(0.14)
*
**
EMS(t-i)
0.33
**
0.32
**
0.31
**
(0.09)
(0.09)
(0.09)
Responsible Care Participant
-0.29
-0.26
-0.27
(0.16)
(0.17)
(0.17)
ISO 9000 Certified
0.80
**
0.72
**
0.72
**
(0.08)
(0.09)
(0.09)
Offsite Waste Transfer
0.29
0.30
0.36
(0.17)
(0.17)
(0.18)
POTW Waste Transfer
0.28
*
0.27
*
0.27
*
(0.09)
(0.09)
(0.09)
Auto Supplier
1.45
**
1.42
**
1.41
**
(0.10)
(0.11)
(0.22)
Industry Waste Generation
0.03
-0.02
0.01
(0.03)
(0.04)
(0.04)
Regulatory Pressure
-4.11
-2.33
-2.27
(2.99)
(3.03)
(3.07)
Affluence
0.09
0.16
0.15
(0.15)
(0.15)
(0.16)
RC Industry
-2.57
**
-2.87
**
-3.09
**
(0.53)
(0.54)
(0.61)
Facility Size
0.36
**
0.38
**
0.38
**
(0.05)
(0.05)
(0.05)
Foreign Owned
0.65
**
0.72
**
0.66
**
(0.17)
(0.17)
(0.17)
Vertically-Integrated Buyer
0.48
**
0.42
**
0.41
**
(0.12)
(0.12)
(0.12)
Firm Size
0.19
**
0.18
**
0.18
**
(0.03)
(0.03)
(0.03)
Constant
-5.50
**
-6.74
**
-6.89
**
(1.54)
(1.59)
(1.62)
Year Dummies
included
included
included
Industry Dummies
included
Log Likelihood
A Chi square
(Nested Comparison Model)
-2756.15
(constant)
-2736.62
39.07 (4)°°
(Model 1)
-2727.86
17.51 (13)
(Model 2)
Number of facilities = 8358; Number of observations = 49413
**p< 0.001; *p<0.01; 00 Change in Chi square significant at p<0.01
40
-------
TABLE 3. Predicting Certification with ISO 14001,1995 cross section
(Probit and Heckman Corrected Probit Models)
Model 4
Model 5
ISO 14001 Certification
EMS (Selection)
ISO 14001 Certification
Distance to Buyer
0.06
**
0.01
0.07 *
(0.02)
(0.06)
(0.03)
Foreign Buyer
0.12
*
0.06
0.21 **
(0.04)
(0.07)
(0.07)
Environmental Performance
-0.10
**
-0.15
**
-0.17 **
(0.02)
(0.02)
(0.04)
Ongoing Vertical Relationship
0.34
**
-0.14
*
0.34 *
(0.08)
(0.06)
(0.12)
EMS(t-i)
0.15
*
(0.05)
Responsible Care Participant
-0.12
0.21
**
-0.07
(0.09)
(0.06)
(0.13)
ISO 9000 Certified
0.11
0.17
**
0.17
(0.07)
(0.05)
(0.09)
Offsite Waste Transfer
0.24
**
0.24
**
0.43 *
(0.08)
(0.04)
(0.14)
POTW Waste Transfer
0.13
*
0.16
**
-0.07
(0.05)
(0.03)
(0.07)
Auto Supplier
0.85
**
-0.22
0.99 **
(0.14)
(0.11)
(0.18)
Industry Waste Generation
0.03
0.14
**
0.04
(0.02)
(0.02)
(0.04)
Regulatory Pressure
-0.49
3.97
**
0.23
(1.63)
(1.00)
(2.25)
Affluence
0.10
0.06
0.09
(0.08)
(0.06)
(0.12)
RC Industry
-1.40
*
-0.52
-1.62 *
(0.32)
(0.20)
(0.44)
Facility Size
0.19
**
0.17
**
0.20 *
(0.03)
(0.02)
(0.04)
Foreign Owned
0.31
*
0.06
0.25
(0.10)
(0.07)
(0.14)
Vertically-Integrated Buyer
0.20
*
0.03
0.26 *
(0.07)
(0.04)
(0.10)
Firm Size
0.08
**
0.05
**
0.06 *
(0.02)
(0.01)
(0.03)
Industry Dummies
included
included
included
Observations
7899
7899
3300
Rho9
0.08
Chi Square
629.85
**
297.50 **
** p<0.001, * p< 0.01, + SIC 22 and SIC 39 removed because perfectly predicts no ISO certification
9 Rho is the correlation between the disturbance terms in the selection and certification models. That this
correlation is not statistically significant does not preclude the potential for biased estimates in uncorrected
analyses.
41
-------
TABLE 4. Predicting Environmental Performance, 1995-2002
(Fixed-Effects Specification)
Model 6
EMS(t-i)
0.04 *
(0.01)
ISO 14000 Certification^)
0.03
(0.02)
Environmental Performance (t_i)
-0.35 **
(0.00)
Pollution Reduction Activity (t.i)
-0.02
(0.01)
Year Dummies
included
Facility Dummies
included
N
54138
Facilities
10080
F-stat
662.05 **
**p<0.001, * p<0.01
-------
Participation in Voluntary Programs,
Corporate Reputation, and Intangible Value:
Estimating the Value of Participating in EPA's EnergyStar® Program
Lou Nadeau,
Eastern Research Group
Jeff Cantin,
Eastern Research Group
and
Richard Wells,
The Lexington Group
June 24, 2003
We would like to thank Andy King, Frank Dixon, Billy Pizer, Madhu Khanna, and Nishkam Agarwal for
insightful comments on an earlier draft of this paper. We are extremely grateful to Sol Salinas of EPA
ENERGYSTAR® for providing guidance in asking the right questions and for providing support for this
work. We would also like to thank Stuart Brodsky and Bill Vonneida of EPA EnergyStar® for
providing valuable data and information. Sophie Delano of ERG provided superb research assistance.
This research was supported under EPA contract 68-W-01-021.
43
-------
Abstract
This paper estimates the market value of the EnergyStar® Buildings program to companies that
participate in the program and the market value associated with energy efficiency in general. To do this,
we specify econometric models of the relationship between intangible value, as measured by Tobin's q,
and participation in the EnergyStar® Buildings program. As part of our modeling effort, we also
attempt to control the influence of two unobservable factors: corporate reputation and a company's
"inherent" energy efficiency. We use data on 124 Real Estate Investment Trust (REITs) measured
quarterly from 1999 to 2001. Our results indicate EnergyStar® partnership results in a return of $16,026
per million dollars in assets owned, compared to not being a partner. This return represented 3.66 percent
of the market value of these companies. Additionally, our models indicate that energy efficient REITs
earn a return of $45,564 per million dollars of assets owned above less energy efficient REITs. The return
for energy efficiency represented 10.4 percent of the market value of these companies. Finally, we
estimate the "lost opportunity" of not joining the EnergyStar® Buildings program (i.e., the value non-
participants would have earned if they had joined). In the sample of REITs, 121 were not a partner for at
least one quarter. Of these 121, 50 would have been better off as an EnergyStar® partner at some point
during the sample period. Of these 50, 20 of them would have been better off for the entire sample period.
Furthermore, the lost value from not joining represented close to 10 percent of the asset value of these
companies.
44
-------
1. Introduction
In the past decade voluntary environmental programs have been increasingly utilized as a public
policy tool because they provide greater flexibility than traditional command-and-control regulations and
they enable public agencies such as the U.S. EPA to influence corporate decision-making in areas where
they do not have a statutory mandate. One of the key challenges facing voluntary environmental programs
is how to quantify the program's value to participants. In this paper, we develop a model that relates
participation in the EPA EnergyStar® Buildings Program to Tobin's q, a measure of corporate
intangible value. We develop statistical models to explore this relationship and then estimate the value of
the ENERGYSTAR® program to participants and the market value of energy efficiency in general.
The lack of evidence that a link in fact exists between participation and firm financial
performance affects program recruitment efforts. By definition, voluntary programs rely on firms to make
commitments to join the program. Although some companies may join voluntary environmental programs
out of altruistic reasons, reliance on such altruism is not a viable strategy to increase and maintain
program membership. Most companies will require some benefit to participation, whether it be tangible or
intangible. Without some benefit, managers of public companies may be violating their fiduciary
responsibility to shareholders if the programs result in significant cost for the companies. Prior
evaluations of voluntary programs have cited "reputational value" as one of the key benefits reported by
program participants (e.g., Wells, 2000; Reed, 2001). Still other studies have noted that public perception,
a key component of reputation, has a significant influence on the decision to join voluntary programs
(Khanna and Damon, 1999; Videras and Alberini, 2000; Arora and Cason, 1995, 1996). Companies report
that they are able to translate their participation in such programs—and any associated public
recognition—into an enhanced corporate reputation, which has a real (albeit intangible) value to them
(Wells, 2000).
Public programs such as EnergyStar® now operate in a period of increased emphasis on
accountability. Voluntary programs can no longer rely on numbers of participants as a measure of
success, but must now demonstrate positive results. For EnergyStar®, key results would include
reductions in greenhouse gases associated with the improved energy efficiency of participating
companies. Another important result, however, would be to provide a mechanism that increases the
market value of the participating companies. Voluntary programs that increase the value of participating
45
-------
companies and that reduce environmental impacts are successful on two scales. Additionally, there is an
important connection between the two goals: programs that create value for participants will attract more
companies leading to additional environmental improvements.
Finally, participation in a voluntary environmental program reflects a company's commitment to
reducing its impact on the environment. Quantitative information on the relationship between voluntary
program participation and financial performance would provide financial analysts and investors with a
ready measure of how participation in these programs affects market value. That is, participation in
ENERGYSTAR® may indicate which firms are energy efficient and thus provide a better investment
opportunity. This would be the case if participation in the program is associated with higher levels of
intangible value, after controlling for other relevant factors.
To investigate the relationship between financial performance and participation in the
ENERGYSTAR® program, we construct econometric models that relate participation in the program to
Tobin's q, a financial measure of intangible value. We use a sample of 124 Real Estate Investment Trusts
(REITs) measured quarterly from 1999 through 2001. Our sample represents 75 percent of the equity
REITs and more than 80 percent of market capitalization among equity REITs. In addition to controlling
for a number of factors that influence financial performance, we also control for self-selection into the
EnergyStar® program. Based on our statistical results, we derive estimates of the market value of (1)
participating in the ENERGYSTAR® program, (2) building energy assessments (i.e., benchmarking), and (3)
energy efficiency in general.
ENERGYSTAR® is set of voluntary programs that are designed to protect the environment by
promoting energy efficiency. The program covers products, construction of new homes, improvement of
existing homes, commercial real estate, and corporate real estate. We focus on the EnergyStar®
Buildings Program, which is primarily concerned with improvements to commercial and corporate real
estate. The program provides a comprehensive set of technical resources on ways to make buildings more
energy efficient. The Buildings Upgrade Manual, for example, provides information on upgrades that can
be made to lighting, fan systems, heating and cooling systems, and other building systems. ENERGYSTAR®
also provides guidance on developing energy efficient purchasing policies and on developing energy
efficient operations and maintenance policies. The program also supplies information and tools on how to
assess the financial viability of upgrades. Finally, the program provides an online tool called Portfolio
46
-------
Manager that can be used to benchmark a building's energy usage against similar buildings. Companies
enter building-specific information into Portfolio Manager which then generates a score between zero and
100 that rates the building's energy efficiency compared to similar buildings.1 Buildings that have a score
exceeding 75 and that meet other requirements can then apply for a "label" from the EnergyStar®
program that certifies the building as one the most energy-efficient in the country. Combined, these
resources provide a set of valuable resources for companies looking to improve the energy efficiency of
their buildings.
Companies can become EnergyStar® partners by submitting a "partnership letter" to the
program. The letter must be signed by the Chief Executive Office or Chief Financial Officer of the
company (or equivalent). In the letter, the company commits to:
¦ Measuring, tracking, and benchmarking its energy performance using the tools provided
by EnergyStar®;
¦ Developing a plan to improve energy performance; and
¦ Educating its staff and the public about being an EnergyStar® partner.
Commitments are voluntary and the EnergyStar® program does not monitor implementation of the
commitments.2 Additionally, partnership with EnergyStar® is not a prerequisite for access to the
energy-related tools discussed above. Those tools and resources are freely available to the public. Rather,
joining EnergyStar® is an outward statement by companies of their commitment to improving energy
efficiency.
1 The Department of Energy's Commercial Buildings Energy Consumption Survey (CBECS)
provides a set of buildings against which the Portfolio Manager database makes comparisons.
2 Additionally, companies are not removed from the program for failure to meet commitments.
47
-------
2. Previous Literature
There is growing body of literature that looks at the relationship between environmental
performance and financial performance.3 Although our study is related to and draws from this literature,
the studies are substantially different from what we are doing in this paper. First, we are looking at how
good energy performance, rather than good environmental performance, relates to financial performance.
Energy performance has significant financial implications, especially for REITs, which are the subject of
this paper. Certainly, environmental performance can also have financial impacts, but these effects are
different than those of energy performance. Firms pay directly for energy usage resulting in a direct
impact on costs. Improving environmental performance, on the other hand, usually has a more indirect
effect on financial performance such as reducing penalties or future regulatory requirements. Second, we
are looking at the financial benefits of joining a public voluntary program. To our knowledge, only one
other study (Khanna and Damon, 1999) has addressed this issue. Most studies in this area look at some
measure or measures of environmental performance and relate those measures to financial performance.
Part of our study follows this path: we use participation in the EnergyStar® program as a measure of
energy performance and relate that to financial performance. Our analysis, however, goes beyond that and
provides an estimate of the value of the program to participants.
Despite the differences between studies of the environmental-financial performance link and our
study, these other studies still inform our work. In short, these studies tend to find a significant positive
relationship between good environmental performance and good financial performance, although the
validity of the results is often questionable. An overriding concern in these studies is whether or not the
observed relationship reflects some other factor such as reverse causation (good financial performance
leads to good environmental performance), good managerial practices, or other firm- or industry-specific
considerations. This paper attempts to address these issues using statistical techniques and detailed data.
In contrast to most of these studies, we focus on only one industry (REITs). Although this focus limits the
applicability of the results to other sectors, we expect that is also represents an advantage over some of
the other studies that used samples from a wide-ranging set of industries. By focusing on one sector, we
can more easily control for some of the factors which influence firm value.
3 Koehler and Cram (2001) provide an extensive review of this literature. We provide a review of
only those studies that are directly relevant for our work in this paper.
48
-------
Three papers are particularly relevant for our work: Konar and Cohen (2001), King and Lenox
(2001), and Khanna and Damon (1999). Konar and Cohen (2001) look at the relationship between
Tobin's q and environmental performance for 321 firms in 1989. After controlling for a number of firm-
specific factors, they find that firms with better environmental records have higher values of Tobin's q.
To measure environmental performance, they use Toxic Release Inventory (TRI) emissions and the
number of environment-related lawsuits pending against each firm. A distinctive feature of the paper is
that the authors translate their estimates into the monetary value associated with environmental
performance. They find that the average firm suffered a $380 million loss in intangible market value
associated with environmental performance. That is, if firms had zero TRI emissions and no pending
lawsuits, the market value of the average firm would have been $380 million higher. This number
represented nine percent of the asset value of these firms. The amount also varied by industry, with
chemical firms suffering a loss worth 31.2 percent of their asset value. Thus, environmental performance
can account for a substantial portion of a firm's value.
King and Lenox (2001) attempt to answer a number of unresolved questions in the debate over
the environmental-financial performance link by using a long panel of data. They use a sample of 652
firms measured annually from 1987 to 1996. In using a panel data set, they can control for firm-specific
effects through panel data methods (e.g., fixed and random effects models). They attempt to address
whether or not it "pays to be green" or whether it "pays to be in a green industry." They find that a firm's
environmental performance, relative to other firms in its industry, is positively related to Tobin's q, after
controlling for industry-level environmental factors and other firm-specific factors. Another statistical
specification, however, contradicts this relationship to some degree. Additionally, they were unable to
conclude that causation runs from environmental performance to financial performance. Thus, King and
Lenox (2001) demonstrate that the relationship between environmental and financial performance is not a
simple one. They conclude that the statistical relationship between the two is influenced heavily by model
specification and choice of sample.
Khanna and Damon (1999) is the most closely related to this paper. They look at the relationship
between participation in EPA's 33/50 program, reduction in the releases of toxic chemicals covered by
the program, and financial performance. They focus on firms in the chemical industry and control for self-
selection into the program. They begin by modeling participation in the 33/50 program. This model is
later used to formulate controls for self-selection into the program. Next, they relate participation in the
49
-------
program to reductions in releases of chemicals covered under the program. They find that 27.9 percent of
these reductions from 1991 to 1993 are attributable to the participation in 33/50. Khanna and Damon then
relate participation in the program to two measures of financial performance: return on investment (ROI)
and excess value as a percentage of sales (EV/S). ROI is measured as the ratio of income to total invested
capital and is a good measure of the current financial performance of a firm. EV/S is measured as the
difference between the market and book value of a firm divided by sales and is a measure of future
prospects for a firm. They find that participation in the program is significantly and negatively related to
ROI, but significantly and positively related to EV/S. Thus, Khanna and Damon conclude that firms
suffer losses in the short-term from program participation, but gain over the long term. The estimates
translate into a 1.2 percent decline in ROI and a 2.2 percent increase in EV/S.
In addition to the literature on the environmental-financial performance link, we also draw from
the literature on participation in public programs. This literature is relevant for our model of decisions to
participate in the ENERGYSTAR® program. Khanna and Damon (1999) make a contribution to this
literature, but we also draw from Videras and Alberini (2000), Arora and Cason (1995, 1996), Henriques
and Sadorsky (1996), and DeCanio and Watkins (1998). In brief, these studies relate participation in
voluntary programs to a number of firm-related factors such as financial health, threat of future
regulation, and past environmental performance. In developing our model of participation decisions by
REITs we provide reference to these studies to justify our model.
3. REITS
Real Estate Investment Trusts (REITs) are companies that own and in most cases operate income-
producing real estate. For the most part, they are public companies that are openly traded on the major
stock exchanges. The REIT corporate structure was authorized in 1960 by an act of Congress which
intended to make large-scale, income-producing real estate holdings available to small investors. REITs
played a relatively minor role in the real estate sector until the Tax Reform Act of 1986 made REITs a
more attractive investment option. Following the Tax Reform Act, the number of REITs began to grow
significantly.
50
-------
To be classified as a REIT, companies must meet three main requirements. First, a majority of the
company's assets must be in real estate held for the long term. Second, the company must earn most of its
income from real estate. Finally, the company must return 90 percent of its taxable income to
shareholders. The primary advantage of organizing as a REIT is the reduced tax burden. The primary
disadvantage to companies organizing as a REIT is the restriction on retained earnings. REITs can be
divided into three general categories: equity, mortgage, and hybrid. Equity REITs primarily own and
manage real estate. Mortgage REITs, on the other hand, primarily own loans or other obligations that are
backed by real estate. A hybrid REIT is a company that engages in both equity and mortgage investment
strategies. In our analysis, we focus on equity REITs because these are the companies that own property
and thus have an interest in reducing energy costs.
The REIT market has grown substantially since the early 1970s. In 1971 there were 34 REITs
with a total market capitalization of $6.5 billion ($2001) (NAREIT, 2003).4 By 2001, there were 182
REITs worth atotal of $161.9 billion ($2001) (NAREIT, 2003). Based on these numbers, the average
market capitalization per REIT increased almost five-fold between 1971 and 2001, from $191.1 million
($2001) in 1971 to $889.8 million in 2001 ($2001).
Energy is an important component of a REIT's operation. Innovest (2002) notes that energy
expenditures are a major part of a REIT's operating cost. Parker and Chao (1999) note that energy costs
for REITs can represent between 20 and 40 percent of the company's total operating cost. This substantial
reliance on energy makes REITs a good case study for looking at the relationship between energy
efficiency and intangible value. Presumably, being an ENERGYSTAR® Buildings partner should provide
some benefits to REITs given their substantial energy needs.
4. Conceptual Considerations
In this section we develop a conceptual framework to characterize the relationship between
participation in EnergyStar® and intangible value. To accurately model this relationship, however, we
must also account for the influence of two other factors on intangible value: corporate reputation and a
4 The market capitalization value was converted from 1971 dollars to 2001 dollars using the
Consumer Price Index.
51
-------
REIT's propensity to be energy efficient. We begin with a general discussion of intangible value and its
relationship to energy efficiency. We then develop a more formal model of the relationship between
intangible value and participation in EnergyStar®. We complete this section by developing a model of a
REIT's decision to join EnergyStar®, which we use to control for self-selection into the program.
Intangible Value
The market value of any firm can be viewed as the sum of two components: value derived from
tangible assets and value derived from intangible assets (Konar and Cohen, 2001). Energy performance
should influence both sources of value. Tangible value is derived from tangible assets such as buildings
and other capital items, as well as a company's earnings. Intangible value is derived from intangible
assets, which include anything that can be linked to future earnings but which does not appear on a
standard corporate balance sheet. Generally, such assets are neither accounted for internally nor
externally, hence it is difficult to assess their magnitude using standard accounting practices (Lev, 2001).
By some estimates, however, intangible assets can make up as much as 80 percent of a company's value
(Reed, 2001).
Lev (2001) groups intangible assets into three categories, related to:
¦ Discovery—The ability to innovate to capture market share.
¦ Organizational practices—Better and smarter ways of doing business.
¦ Human resources—Having better-trained and better-qualified people than competitors.
We expect that ENERGYSTAR® participation, or good energy performance in general, is most closely
related to the second of these, i.e., it adds value because it represents a better way of doing business.5
5 Certainly, energy efficiency could manifest itself in other ways, including those related to
discovery and human resources. For example, a REIT could develop a new way of operating buildings
that is more energy-efficient, representing a business innovation that feeds into its intangible value
(discovery).
52
-------
For REITs, adapting better business practices as a result of EnergyStar® participation leads to a
number of intangible benefits. First, energy efficient buildings cost less to operate, which should result in
bottom-line benefits for REITs. Furthermore, lower operating costs increase the net operating income
(NOI) which in turn increases the property value for a building.
Second, energy efficient buildings provide a more attractive space and work environment. Romm
and Browning (1998) found that energy efficient lighting upgrades led to productivity improvements in a
series of case studies. This can lead to better tenant retention and the ability to charge a premium for the
improved space. Third, through its EnergyStar® participation the company starts to gain recognition as
an environmentally conscious business. This leads to improved public perception, which has been
identified as an important source of value for companies joining voluntary programs (Wells, 2000;
Khanna and Damon, 1999; Videras and Alberini, 2000; Arora and Cason, 1995, 1996).
Third, efficient energy management may improve the long-term sustainability of the company.
That is, alliance with ENERGYSTAR® may help build intellectual capital in the company with regard to
energy management techniques. Those techniques can provide insulation against energy shocks, supply
interruptions, or price spikes. Finally, competency in energy management may be interpreted as an
indicator of superior overall management (Innovest, 2003). Energy is a complex management issue,
requiring an ability to recognize and address emerging issues. Companies that manage energy issues
effectively may find that their overall management reputation is enhanced.
To formalize our notion of intangibles, we write a firm's market value (MV) as the sum of
tangible and intangible value:6
MV = VT + V, (1)
If we divide through by VT in equation (1), we get
MV _ Vi
VT + VT
(2)
The left-hand side of equation (2) can be interpreted as Tobin's q, the ratio of market value to
replacement value of tangible assets. Thus, based on (2), a firm's q value is determined by the ratio of
intangible to tangible values. A firm with no intangible value (Vj=0) will have a q value equal to one.
6 The following treatment of q draws from Konar and Cohen (2001).
53
-------
Following Hirsch and Seaks (1993) and Konar and Cohen (2001) we specify V,/Vr as a linear
function of a set of explanatory variables:
where X is a matrix of explanatory or control variables, (3 is a vector of regression coefficients on the X
variables (including a constant term), and 8 is an error term. Thus, we get the following expression for a
regression model for Tobin's q:
This model can be estimated using either q or q-1 as the dependent variable.
The Relationship Between Intangible Value and EnergyStar® Participation
The goal of our analysis is to examine the relationship between intangible value and participation
in ENERGYSTAR®. Our central hypothesis then is that participation in the program will lead to higher
levels of intangible value. We expect that participation in the EnergyStar® program represents a smarter
way of doing business, resulting in higher intangible value. Decisions to join EnergyStar®, however,
are not random, but reflect a decision process on the part of companies. To accurately capture the impact
of participation in the program on intangible value we must account for this self-selection into the
program.
There are two other considerations that we expect to be important in looking at the relationship
between participation in EnergyStar® and intangible value: energy efficiency and corporate reputation.
In regards to the first, it is clear that companies can pursue energy efficiency goals in the absence of the
EnergyStar® program. There may be some level of management commitment to energy efficiency for
business reasons (e.g., lowering cost) or for social reasons (e.g., reducing impacts on the environment).
Regardless of the reasons, a company's "inherent" propensity for energy efficiency will influence its
intangible value independent of the ENERGYSTAR® program. Additionally, a company's energy efficiency
will both influence and be influenced by its decision to join EnergyStar®.
Vr
— = XP + 8
(3)
q = 1 + X$ + 8
(4)
54
-------
Corporate reputation is a more encompassing concept than energy efficiency, but will also exert
an influence on the relationship between intangible value and EnergyStar® participation. Companies
with better reputations will have higher levels of intangible value as the market is willing to pay more to
hold the stock of companies with better reputations. Thus, reputation will exert an influence on q. We
might also expect reputation to have a positive influence on decisions to join EnergyStar® and on
energy efficiency: i.e., companies with better reputations may be more likely to join EnergyStar® or be
more energy efficient. Although these tend to be correlational rather than causal relationships, we need to
address them in our modeling in some form.7
Although we expect both energy efficiency and reputation to be important factors influencing the
relationship between intangible value and EnergyStar® participation, we do not have data to objectively
measure either energy efficiency or reputation. Nevertheless, we expect that both reputation and a
propensity to be energy efficient will influence intangible value and decisions to join EnergyStar®. Our
solution is to assume that joining EnergyStar® and not joining EnergyStar® represent completely
different scenarios for companies which must be represented by completely different regression models.
The two different scenarios stem from the fact that corporate reputation and energy efficiency make
companies that join EnergyStar® significantly different than those that do not. Empirically, this is a
switching regression model and we can write this as
where D is a binary variable equal to one if the company is an EnergyStar® partner and zero if the
company is not a partner, Xl and X are matrices of explanatory variables, f), and (32 are vectors of
regression coefficients, and ul and u2 are regression error terms. The "program effect" in this model can
be calculated by comparing the predicted q value from (5a) to the predicted q value from (5b) for
companies that join the program. This is discussed in more detail in the econometric methods section.
7 The converse might also be true: joining EnergyStar® or being energy efficient will lead to an
improved corporate reputation. That is, participation in ENERGYSTAR® or being energy efficient leads
society to place a higher reputational value on those companies. This is partly what we are measuring in
our statistical models.
q = X1$l + ul for D = 1
(5 a)
q = X2$2 + u2 for D = 0
(5b)
55
-------
This structure allows the set of explanatory variables (Xt) to differ between partners and non-
partners, although we expect that a number of common factors are relevant for both groups. As discussed
below, however, two program-related activities (benchmarking and labeling) may influence partners'
intangible value, but they are irrelevant for explaining variation in non-partner q values. Additionally, this
structure allows the coefficient values ((3j) to differ between partners and non-partners for factors that are
common to both regression models. Thus, we let the relationship between q and the explanatory factors
differ completely between the two groups. We expect that the relationship between q and its determinants
(i.e., the variables in the Xmatrices) will be substantially different for partners than for non-partners and
that the difference between the two stems from energy efficiency and corporate reputation considerations.
In other words, companies that join EnergyStar® differ from those that do not join EnergyStar® with
respect to energy efficiency and corporate reputation. Without objective data to measure these
differences, we need to model them separately.
An alternative to the switching regression model in (5a) and (5b) is to pool the partners and non-
partners and use a binary indicator to capture the effect of being an EnergyStar® partner. This is a
standard program evaluation model and can written as:8
q = X$ + a D + u (6)
where a reflects the effect of participating in the EnergyStar® program. Although this model appears to
be more tractable, the estimated value of a will be biased if we exclude corporate reputation and a REIT's
propensity for being energy efficient because each will have a positive influence on both intangible value
and decisions to join EnergyStar®. As discussed in more detail in the econometrics methods section,
however, the biased estimate will incorporate both a "reputation effect" and a "propensity for energy
efficient effect" into the estimated coefficient. Based on this, the biased estimate of a in an estimation (6)
provides a measure of how energy efficiency in general affects q values, with "energy efficiency" being
the sum of a program effect, reputation effect, and propensity for being energy efficient effect. In essence,
a biased estimate will still be valuable because we have an idea about where the bias is coming from and
we can therefore give the estimated coefficient a broader interpretation.
8 The '1' and '2' subscripts are removed to reflect that we have pooled the data.
56
-------
We develop empirical models based on both (5a) and (5b) and (6). The estimates from (5a) and
(5b) are used to develop an estimate of the program effect: i.e., the intangible value of the EnergyStar®
program to REITs. The estimated value of a from (6) is interpreted more broadly as the intangible value
of energy efficiency, as proxied by EnergyStar® partnership. In both sets of models, we control for self-
selection of REITs into the program. This is discussed in more detail in the econometric methods section.
As noted above, there are two program-related activities that may affect a REIT's intangible
value. First, companies that join EnergyStar® are encouraged by the program to benchmark their
buildings. Benchmarking involves entering building-specific information into EPA's Portfolio Manager
database which then generates a score for the building.9 Benchmarking is voluntary for the partners, but
reflects a desire on the part of REITs to understand the energy efficiency of their buildings relative to
similar buildings. Companies that own low-scoring buildings should be motivated to improve the
buildings' energy efficiency because the benchmark score indicates that similar buildings are more
efficient and thus operating at lower cost. The number of buildings that a company benchmarks, however,
is not public knowledge. Thus, benchmarking acts more as an indicator of energy efficiency concern
(among partners) than as an outward sign of energy efficiency. We measure benchmarking by calculating
the percentage of total square footage owned that each REIT benchmarked in the previous year
(BENCH)}0
Buildings that achieve a score of 75 or higher can receive a label from the EnergyStar®
program, provided they meet other requirements, entitling the owner to display a bronze plaque bearing
the EnergyStar® logo.11 A label is a certification by the EnergyStar® program that the building meets
high energy efficiency standards. Thus, labeled buildings represent the best performing buildings of their
9 Portfolio Manager ranks the building's energy efficiency (kwH per square foot) against a
national database of buildings, developed by the Department of Energy, adjusted for building type,
occupancy patterns, and climate (location).
10 We have quarterly observations, so this measure can change from quarter to quarter. For
example, for the third quarter of 1999, we calculated the percentage of total square footage that had been
benchmarked between fourth quarter 1998 and third quarter 1999. Our presumption is that a benchmark
score is valid for one year.
11 To receive the label, the building owner must complete an application letter and a Statement of
Energy Performance (provided by EPA) and then have a Professional Engineer certify that the data
entered into the Portfolio Manager database is accurate and that the building conforms to industry
standards for indoor environment.
57
-------
type and are among the most energy-efficient in the country. Furthermore, EPA publishes lists of labeled
buildings, making a label a form of public recognition for energy efficient operation. To the extent that
the market acknowledges this public recognition, a label acts as an indicator to the market that the REIT
owns and operates energy-efficient buildings. Thus, there may be some intangible value associated with
building labeling. Similar to benchmarking, we measure labeling as a percentage of total square feet
owned by the REIT (LABEL).
In addition to these energy-related factors, there are a number of other factors that will influence a
company's intangible value. These other factors make up the Xmatrices in the equations above and can
be divided into three groups: market conditions, firm-specific financial factors, and firm characteristics.
We expect that better market conditions for REITs will result in higher intangible values, all else equal.
To measure market conditions, we use the quarterly return to the National Association of Real Estate
Investment Trust (NAREIT) index for equity REITs. The NAREIT Equity Index tracks the market
performance of all equity REITs. We expect that this will act as a good proxy for general market
conditions for our sample of REITs. For each quarter in the data, we calculated the percentage change in
the NAREIT index from the previous quarter and used that as a measure of market growth or decline. We
label this variable NRET in our empirical model.
There are a number of firm-specific financial characteristics that can have an influence on a
firm's intangible value. First, we use the REIT's return on assets (ROA) as a measure of REIT
profitability. We measure ROA both concurrently to q and lagged by one quarter. Second, the riskiness of
a REIT's stock may also influence the value the market places on its intangible assets. To measure this,
we calculated each REIT's stock market beta (SBETA) for the 1996-2001 time period and used that as a
measure of riskiness.12 Finally, we control for each REIT's baseline level of intangible value. We expect
that each REIT's intangible value during the sample period (1999-2001) will be positively influenced by
its intangible value in the period leading up to the sample period. If REITs with higher intangible values
prior to the sample period are concentrated among partners, then we may find that partnership is
significantly associated with higher q values than non-partners when such a relationship actually does not
exist. In other words, good performance in terms of intangible value may carry over from the pre-sample
12 Thus, there is one calculated value of beta for each firm which is used to measure riskiness in
each quarter.
58
-------
period to the sample period. To measure this, we calculated the average value of Tobin's q for 1996-1998
for each REIT and refer to this variable as qb.u
Intangible value can also be influenced by a number of firm-specific non-financial characteristics.
Studies of intangible value often use the size of the firm, measured as the number of employees, as a
control variable. We use the total square footage owned by the REIT as a measure of firm size rather than
the number of employees. Previous studies of Tobin's q values for REITs have also accounted for the
concentration of a REIT's holdings in one sector (e.g., office, retail) (Capozza and Seguin, 1999). Our
meaure of the concentration (CONC) is a Herfindahl index and is defined as:
CONC =
y=i
i ( 2
7 5.
where .v. is the square footage owned by the REIT in sector /. and S is the total square footage owned by
the REIT.14 Finally, we also expect that the sector where a REIT has concentrated its holdings may
influence q values. In equation (6), we control for the two major sectors in the data: office (OFFICE) and
retail (RET) using binary control variables set equal to one if the REIT had more than 75 percent of its
square footage in either the office or retail sector, respectively.15
Based on these considerations, we can write equation (6) as
q = f(D, BENCH, LABEL, NRET, TSQFT, ROA,
ROAt_l, qb, SB ETA, CONC, OFFICE, RET)
(7)
where D is the binary variable equal to one if the REIT was a partner and all variables are as defined
above. Equation (5a) excludes D and equation (5b) excludes D, BENCH, and LABEL. We measure total
square feet (TSQFT) in natural logarithm form to reduce the influence of any large square footage values.
We discuss our implementation of this model more fully in the econometric methods section below.
13 Similar to the stock market beta, there is only one value, the pre-1999 q value, for each REIT.
14 The seven sectors covered in our analysis are: office, retail, industrial and warehouse, health
care, lodging, residential, and self-storage.
15 We were unable to include the sector-specific controls in the switching regression model due to
multicollinearity issues in the estimations.
59
-------
Decisions to Join EnergyStar®
We also expect that the decision to join EnergyStar® is not random and can be represented by
the following equation:
D* = Zy + e (8a)
where D* is the net benefit from joining EnergyStar®, Z is a set of explanatory variables that influence
the net benefits, and y is a vector of coefficients. We expect that companies will join EnergyStar® if the
net benefits of joining exceed zero and will not join if the net benefits are negative. Thus, assignment into
the two regression equations for q (i.e., (5a) and (5b)) can be written as
D = 1 ifD* > 0 (8b)
D = 0 ifD* < 0 (8c)
This specification assumes that the decision of whether or not to join EnergyStar® reflects some
assessment of the net benefits of joining the program by each REIT. If the REIT determines that the net
benefits are positive, then it joins the program. We do not expect that each REIT performs a detailed net
benefit calculation for this, but rather more of a subjective calculation. The goal of characterizing the
decision to join is to include a set of factors in a statistical model of equation (8a) that reflects the net
benefits of joining the EnergyStar® program. We expect that four sets of factors are relevant: market
conditions, company size, the firm's financial health, and the company's primary sector.
Improvements in the market conditions affecting REITs should have a positive influence on the
decision to join EnergyStar®. As overall conditions improve, REITs as a group will have more
resources to allocate to voluntary programs such as EnergyStar®. As conditions worsen, on the other
hand, REITs would be expected to focus their energies on weathering the down market rather than joining
voluntary programs. Videras and Alberini (2000) found that increases in industry-level sales were
positively associated with an increased probability of joining EPA's Green Lights program, but were not
associated with joining EPA's WasteWise or 33/50 programs. DeCanio and Watkins (1998) found that an
increase in industry earnings per share was positively related to joining Green Lights. In our statistical
60
-------
model of the decision to join ENERGYSTAR®, we control for market conditions by using the quarterly
return in the NAREIT equity REIT index discussed above.
Larger REITs may also have an incentive to join compared to smaller ones. The larger the REIT,
measured either in terms of square footage or total assets, the more opportunity there is to reduce costs
through energy efficiency programs. Previous studies that looked at voluntary program participation
decisions all used some measure of company size (Khanna and Damon, 1999; Videras and Alberini,
2000; Arora and Cason, 1995, 1996; Henriques and Sadorsky, 1996; and DeCanio and Watkins, 1998).
We control for REIT size in the model of participation decisions by including the total square footage
owned by the REIT and the total assets owned by the REIT. In our statistical models, we measure each in
logarithmic form to reduce the influence of large values.
REITs that are doing better financially relative to their sector peers should also be more likely to
join ENERGYSTAR®. A number of other studies have examined this issue, but each found no relationship
between financial health and participation (DeCanio and Watkins, 1998; Videras and Alberini, 2000; and
Arora and Cason, 1995). We measure firm profitability by ROA concurrently and lagged by one quarter.
Additionally, we also include the average q value for 1996-1998, the period preceding our sample period.
We expect that firms that had a higher levels of intangible value at the beginning of the sample period
will be more likely to join the program. We also include the firm's stock market beta as a measure of risk
faced by owning the firm's stock. We expect that the riskiness of a firm's stock should affect decisions to
join EnergyStar®, but the direction of that effect is left to the empirical model.
Finally, decisions to join ENERGYSTAR® will be influenced by the types of buildings that are
owned by the REITs. For example, in interviews with 33 REITs, Parker, Chao, and Gamburg (1999)
found that in the retail sector the tenants tend to pay for energy costs while in the office sector the REIT
tends to pay for energy costs. Additionally, the energy requirements and opportunities for energy
efficiency upgrades will differ markedly across the sectors we use in our analysis (office, retail, industrial
and warehouses, health care, lodging, residential, and self-storage). These sector differences may
influence decisions to join ENERGYSTAR®. Finally, early recruitment efforts of the ENERGYSTAR®
Buildings Program focused on office properties. Both Videras and Alberini (2000) and DeCanio and
Watkins (1998) use industry sector controls in their analyses. Therefore, we include a set of binary
variables that reflect the sector that each REIT operates in.
61
-------
Based on these considerations, our model for joining EnergyStar® can be written as
I) = f(NRET, TSQFT; TA, ROA, R()At ,, qb, SB ETA, SECTOR) (9)
where TA is total assets, SECTOR is a set of control variables for the sectors in our data, and all of the
variables are defined as above. We use a probit model to estimate this equation. The following section
discusses our econometric method, which involves using the model of decisions to join EnergyStar® in
equation (9) to adjust for self-selection into the program.
5. Econometric Methods
In this section, we review the econometric issues involved in estimating the models discussed in
the previous section. There are two models that we discuss: the switching regression model of (5a) and
(5b) and the pooled model of (6). The switching regression model is used to derive estimates of the
program effect of EnergyStar® for REITs while the pooled model is used to derived the intangible
value associated with energy efficiency overall. In this section, we also provide reasoning for this
interpretation of the pooled model.
62
-------
Switching Regression Model With Endogenous Switching
The model in (5a) and (5b), along with the self-selection into EnergyStar® described by
equations (8a)-(8c), is a switching regression model with endogenous switching. To begin, we specify the
model formally as:
hu = P1X7 + uiu f°r Du = 1 dOa)
hit = P2xiu + u2,t f°r Du = 0 00b)
Du = l'zit + % (10c)
D f = 1 if D* > 0
(lOd)
Dn = 0 i/D,; <0
where all variables and parameters are as defined above and / indexes REITs and t indexes time.
Estimation of (10a) and (10b) in their present form will result in biased estimates because the error terms
in both equations will have non-zero expectations due to the selection mechanism defined in (10c) and
(lOd) (Maddala, 1983).
To estimate this system, we follow the two-step procedure outlined in Maddala (1983). First, we
define the following two ratios:
Wllt = Hj'ZItVH j'Zit) (lla)
W2lt = (P(y^f)/[1 - Hj'zu)] (lib)
where cp and
-------
EnergyStar® (Z) and the coefficients associated with those variables (y). These ratios can be computed
for each observation in the sample. We can now redefine equations (10a) and (10b) as16
lit = Pi*iu " °uWiit + viit f°r D,t = 1
(12a)
% = $2X2it + G2eW2U + V2it fOT D it = 0
(12b)
where ols is the covariance between u, and s, o2s is the covariance between u2 and 8, and vjit {/' = 1, 2} are
new residuals with zero conditional means.
The new equations can now be estimated using a two-stage process. First, we estimate a probit
model for decisions to join EnergyStar® ((10c) and (10d)), using D values in place of the unobserved
D* values. This provides a consistent estimate of y which we call y. Next, we generate estimates of Wm
and W2jt by substituting y for y in (1 la) and (1 lb). In the second stage, we estimate our equations for
Tobin's q, (12a) and (12b), using Wljt and W2it in place of Wlit and W2it, respectively. The estimated
coefficients for Wm and W2it will be consistent estimates of ols and o2e, respectively (Maddala, 1983).
As we noted in Section 4, we expect both corporate reputation and a REIT's propensity for being
energy efficient to affect both partnership decisions and intangible value. The switching regression model,
however, mitigates the effects of these omitted variables by separating the partners from the non-partners.
To see this, we write our two q equations as
where R* is an unobserved measure of corporate reputation and 0] and 02 measure the impact of
reputation on q. We will focus this discussion on reputation because the propensity for energy efficiency
can be handled in an analogous manner. Next, we can assume that corporate reputation and partnership
have a simple relationship such as (ignoring /' and t subscripts and any error term):
16 The detailed derivation of these equations, along with a more detailed account of this method,
can be found in Maddala (1983, pp. 223-228).
?i/r = P^iit + QiR * + uiit f°r D,t = 1
(13 a)
V2U = $2X2it + M ' + U2rt fOT D,t = 0
(13b)
64
-------
R* = b0 + bxD
(13c)
where b0 and h] are regression coefficients. To see how having this unobserved relationship affects (10a)
and (10b), we substitute (13c) into (13a) and (13b) and use the condition for Dit for each relationship:
lut = P+ Wo + bi) + uilt f°r D,t = 1 (14a)
hit = P2X2U + Q2b0 + U2H f°r DU = 0 (14b)
Thus, the terms + &i) and 02^o are unmeasured effects of corporate reputation on q}1 Corporate
reputation, however, is likely to have a significant firm-specific component, especially when measured
over time in a panel. If we assume that b0 can be written as b: instead (i.e., as a firm-specific effect) then
our unmeasured effects of corporate reputation can be handled with a panel data method such as random
effects and by assuming that the remaining effect is subsumed into the error term. Thus, in the partner
model, we are assuming that some of the reputation effect (i.e., is randomly distributed among
partners after controlling for other factors included in X. This same assumption is not valid for a pooled
model because reputation affects both decisions to join and q values.
Pooled Model For Valuing Energy Efficiency
The pooled model from equation (6) can be written more formally as
ht = Xi$ + aDu + Uu (15)
where a is once again the program effect of EnergyStar® on intangible value. To account for self-
selection into the program, we combine (15) with the REITs' decisions to join EnergyStar® modeled in
(10c) and (lOd). Estimation of (15) without controlling for self-selection into EnergyStar® will result in
biased estimates of a due to the relationships in (10c) and (lOd) (Maddala, 1983). To avoid this, we
17 The other unmeasured factor that we consider important, a REIT's propensity for being energy
efficient, can be handled in a similar manner, resulting in additional terms similar to the ones for
corporate reputation in (14a) and (14b).
65
-------
follow a path similar to the switching regression modes above and add the inverse mills ratio as a new
regressor to our model.18 This can be written as
% = Xit P + aDu + KW,t + Vu
(16)
where Wit = (p(y'Zlt)/
-------
coefficient on D, say a, which will equal (a + 0 />,). In other words, the estimated coefficient on D in (18)
will be the sum of the program effect (a), a corporate reputation effect (0 bx), and the "propensity for
being energy efficient" effect. Given that we cannot separate out the pure program effect, we interpret our
estimated coefficient in (16) as the intangible value of energy efficiency, operating through participation
in ENERGYSTAR®, corporate reputation, and the efforts of REITs to operate in an energy efficient manner.
A final point is that even if we assume that b0 can be written as a firm-specific random effect br we will
still be unable to separate out the program effect from the other two unobserved influences. Nevertheless,
we assume that there are firm-specific factors that are unmeasured in this model and use the random effect
procedure in the estimations.
In both the switching regression model and the pooled model we use predicted probabilities from
a first stage probit model. The first stage probit model contains many of the variables also in the second
stage switching regression model and the pooled model. In order for our predicted probabilities to have
some independent variation, we need to include a variable in the first stage probit model that acts as an
instrument for joining ENERGYSTAR®. We use total assets as our instrument since it satisfies the criteria
for a good instrument: it is related to the probability of joining ENERGYSTAR®, but is not correlated with
the outcome variable (i.e., Tobin's q).
6. Data
Data for our analysis come from a variety of sources. We began by obtaining data from Standard
and Poor's Compustat database for all companies in Standard Industrial Code 6798 (Real Estate
Investment Trusts). This included 198 REITs active between 1996 and 2001. We then restricted the
sample to equity REITs that had sufficient financial data over the sample period.21 Next, we collected
information on square footage by property type owned by the REITs from the companies' Security and
Exchange Commission (SEC) filings and from corporate annual reports. This also resulted in some loss of
companies from the sample due to missing or inadequate data. Additionally, some REITs were considered
out of scope for this analysis and were also dropped.22 Our final sample includes 124 equity REITs.
21 Information on which REITs were classified as equity REITs was taken from NAREIT's web
site (http://www.nareit.com).
22 These were REITs primarily involved in manufacturing and development of homes.
67
-------
Information on EnergyStar® partnership and building benchmarking and labeling were
obtained from EPA's ENERGYSTAR® Buildings Program. This included a list of partners in the
Commercial Real Estate sector and the dates on which each partner joined the program. Information on
benchmarking and labeling were taken from the program's Portfolio Manager database which tracks those
activities. This information was matched to our sample of 124 REITs.
We followed the method proposed by Chung and Pruitt (1994) to calculate Tobin's q. Their
measure of q can be written as
(MVE + PS + DEBT)
v = Ya (19)
where MVE is the market value of common stock (share price multiplied by shares outstanding), PS is the
liquidating value of the firm's outstanding preferred stock, DEBT is the value of a firm's short term
liabilities net of its short term assets plus the book value of the firm's long-term debt, and TA is the book
value of the firm's total assets. This is a common formulation of q in the literature.
To calculate the stock market betas (SBETA), we used end-of-month stock prices for the period
1996 through 2001. For each REIT, we regressed the monthly stock price return on the monthly return to
the Standard and Poors 500 Index. The estimated slope coefficient from these regressions are, by
definition, stock market betas.
Our final data set is comprised of 124 REITs with quarterly observations spanning 1996 through
2001. In our statistical models, however, we restrict the sample to 1999 through 2001, since recruiting for
ENERGYSTAR® did not begin until 1999. This results in a potential panel size of 1,488 observations (124
REITs x 12 quarterly observations). Due to missing data, however, our workable sample size was 1,434
observations. Of the 1,434 observations, 202 were for the partners and the remaining 1,232 were for non-
partners.
There were 23 REITs (19 percent) that became partners over the course of the sample period.
Companies could be in both the partner and non-partner samples since partnership was defined as being a
partner at a specific time. Thus, a company that joined after the beginning of the sample period, but
68
-------
before the end of the period, would be in both samples. Of the 23 partners in the sample, only three were
partners for the entire sample period.
Table 1 provides definitions and summary statistics for our data. The average value of Tobin's q
for the 124 REITs during the sample period was 0.923. Partners had an average q value of 0.945 and non-
partners had an average q value of 0.917, a difference that is not statistically significant. Over the sample
period, 202 observations (14.1 percent) corresponded to REITs being partners. The average REIT in our
sample owned 32.6 million square feet and had $1.6 billion in assets. Additionally, on average the REITs
property type holdings tended to be very concentrated with an average Herfindahl concentration index of
0.89.23 Finally, more than half of all the observations corresponded to retail REITs. The office and retail
sectors combined for a total of 64 percent of all observations.
7. Econometric Results
In this section we present the results of estimating our equations for Tobin's q and the decision to
join EnergyStar®. We begin by discussing the probit model for REIT's decisions to join EnergyStar®
since this model is used to generate our control for self-selection into the ENERGYSTAR® program for
both the switching regression model and the pooled estimation. We then review the results for our
regression models for Tobin's q. In the section that follows, we translate the regression results into dollar
values associated with the EnergyStar® program.
23 The construction of the index implies that values can range from 0.14 (equal diversification
across property types) to one (complete concentration in one sector).
69
-------
Decisions to Join EnergyStar®
Table 2 presents the results of the probit model for decisions to join EnergyStar®. We present
both the estimated coefficient and the marginal effects of each variable on the probability of joining
EnergyStar®. The probit model was run for the sample of 124 REITs measured quarterly from 1999 to
2001 for a total of 1,434 observations. In the probit model we include a set of dummy control variables
for the sectors using the retail sector as the base.
The estimated coefficient for the return to NAREIT index indicates that general market
conditions are significantly and positively related to joining EnergyStar®. Although significant, the
actual impact on the probability of companies joining EnergyStar® is small. The marginal impact for
the return to the NAREIT index indicates that a one percentage point increase in the index increases the
probability of the average company joining EnergyStar® by 0.35 percentage points.
In terms of REIT size, the amount of assets owned by REITs is significantly and positively
related to joining EnergyStar®, but the total square footage is not related to the probability of
companies joining the program. The marginal effect for total assets indicates that a one percent increase in
the average company's total assets increases the probability of joining EnergyStar® by 5.6 percentage
points. The positive relationship between participation and size is consistent with previous studies that
look at voluntary program participation (Khanna and Damon, 1999; Videras and Alberini, 2000; Arora
and Cason, 1995, 1996; and DeCanio and Watkins, 1998).
The four REIT-specific financial characteristics that we used in the analysis are all significantly
related to joining EnergyStar®. Return on assets, both concurrently and lagged by one quarter, has a
positive impact on joining EnergyStar®. That is, firms that are doing better financially are more likely
to join the program. This differs from previous studies which found little effect of financial health on
decisions to join voluntary programs (Videras and Alberini, 2000; Khanna and Damon, 1999).
Additionally, if we also take ROA as a measure of management effectiveness, firms that are better
managed are more likely to join. Firms that had higher intangible values in the period leading up to our
sample period, as measured by pre-1999 Tobin's q values, were also more likely to join EnergyStar®.
Finally, joining EnergyStar® was positively related to firm-specific risk associated with the REITs'
stock prices. REIT stocks, however, tend to be low-risk relative to the stock market. Our estimated betas
70
-------
ranged from 0.127 to 0.363. Thus, the finding that riskiness increases the probability of joining
EnergyStar® may be limited to our sample given the small range of betas.
Finally, the sector in which the REIT owned most of its property played a significant role in
determining the probability of joining EnergyStar®. We included controls for six of the seven sectors,
making retail, the largest of our sectors, the base sector in the analysis. REITs in the residential, industrial
and warehousing, lodging, health care, and self-storage sector were all less likely to join EnergyStar®
than retail REITs. Office REITs, on the other hand, were much more likely to join EnergyStar® than
retail REITs.
The estimated model in Table 2 was used to generate a values for WUt and W2it24 For each
observation, we calculated the value of Zity using each observation's values for Zit and the estimated
coefficients in Table 2 for y. We then calculated cp(Zity), the standard normal density function, and (•). These two ratios are added to
the models for Tobin's q to control for endogenous selection into the EnergyStar® program.
Intangible Value
Table 3 presents the results for the switching regression model and the pooled regression model.
The switching regression equation for partners had 202 observations and covered 23 REITs, while the
non-partner equation had 1,232 observations and included 121 REITs. The pooled model contained all
1,434 observations and all 124 REITs. All three models were run using quarterly observations running
from 1999:1 to 2001:4. In this section we discuss the statistical results. In the next section we derive
quantitative measures of the value of the EnergyStar® program and the value of energy efficiency based
on these models.
We will start by discussing the non-program variables. Market conditions for REITs, as measured
by the return to the NAREIT index, are positively related to q values. The coefficient is significant in the
24 Note that Wut in the switching regression model and Wit in the pooled model are defined
identically and are even calculated from the same probit estimation (Table 2). They are defined over
different samples, however.
71
-------
pooled sample and for non-partners, but the coefficient is not significant in the partners' model.25 The
magnitude of the effect, however, is small. A one percentage point increase in the return to the index
increases q values by slightly more than 0.003 points, or 0.32 percent of the average q value for the
sample period.
The riskiness of a REIT's stock has a significant negative effect on q values. That is, REITs with
more volatile stock prices tend to have lower q values. The effect of volatility is more than four times
larger among partners than non-partners. REITs, however, tend to be lower risk stocks and thus it is not
possible to extrapolate this result beyond REITs.
Pre-1999 Tobin's q values are highly significant and exert a positive influence on q values in the
sample. Larger values of the coefficient indicate a stronger influence of pre-1999 q on sample period q
values. The relationship is strongest for partners with a coefficient of 0.83. Thus, partners that had high
intangible values in the pre-sample period would also have high values in during the sample period. The
use of pre-1999 Tobin's q controls for this carry-on effect of good performance.
Return on assets, both contemporaneous and lagged, is negatively associated with Tobin's q in
our sample. The estimated coefficients, however, are only significant in two of six cases. In those two
cases, they are significant only at the 10 percent level. Nevertheless, the prevalence of the negative signs
does seem to indicate a negative relation. We expected that REITs with better profitability would have
higher q values, so this result runs counter to our expectations.
The size of the REIT is significantly and positively related to q values. In the pooled model, a one
percent increase in total square footage owned by a REIT increases q values by 0.024 points. The
relationship is larger among partners compared to non-partners, however. A one percent increase in total
square footage increase q values by 0.059 points for partners, but among non-partners the increase n q is
only 0.027.
25 The lack of significance in the partner model may be caused by too few observations for
partners. Specifically, the coefficient estimate on NRET for partners is (numerically) close to the
estimated values in the non-partners and pooled models, but the standard error in the partners model is too
large to generate a significant coefficient.
72
-------
The concentration of a REIT's holdings across sectors is also positively related to q values. That
is, REITs that are less diversified across property types tend to have higher q values. This result confirms
previous research on the relationship between REIT property type concentration and firm value (Capozza
and Seguin, 1999). This relationship, however, is not significant for partners. The pooled model also
contains two sector controls: one for the office sector and one for the retail sector. Both control variables
are negative and significant, indicating that non-office, non-retail REITs fared better over the sample
period than REITs concentrated in the office and retail sectors.
The set of program-related variables (participation, benchmarking, and labeling) differs between
the three equations. The non-partner model contains none of the program-related variables and the pooled
model contains all three. In the pooled model we find that the partnership variable is significantly and
positively associated with higher q values. As noted in the econometric methods section, however, this
estimated coefficient reflects the sum of three influences: the direct effect of partnership on q, a
reputational effect, and the effect of a company's propensity for being energy efficient. Combined, we
interpret this coefficient as the effect of energy efficiency, as proxied by partnership in EnergyStar®, on
q. Thus, after controlling for market conditions, firm-specific financial factors, firm characteristics, and
primary sector, we find a significant effect of energy efficiency on q. In the next section we translate the
estimated coefficient in this model to a measure of the value of energy efficiency.
Our measure of benchmarking activity is included in the partner model and the pooled model. In
both models benchmarking is associated with significantly larger q values. The estimated coefficient is
larger in the partner-only model than in the pooled model because the pooled model contains non-partners
that perform no benchmarking. The result says that REITs that benchmark a larger proportion of their
buildings have higher q values. Benchmarking, however, is not something that is observed by the market.
Thus, the market must be reacting to an outcome associated with benchmarking. Given that we have
controlled for several other factors (firm-specific financial factors and pre-1999 q). we expect that the
significant coefficient on BENCH reflects the market's valuation of energy efficient building operation.
Energy efficient building operation will manifest itself in increased profits for these companies. Thus,
companies that benchmark larger proportions of their buildings operate those buildings efficiently and
this leads to higher intangible values by increasing profitability. Another possibility is that the market
recognizes which REITs are more energy efficient and places some value on that energy efficiency. If the
more energy-efficient REITs also tend to benchmark more of their buildings, then our significant
73
-------
coefficient on BENCH reflects the value of energy efficiency. In either case, the market is placing some
intangible value on energy efficient operation.
Our measure of labeling, included in both the pooled model and the partner model, is not
significant. This would seem to indicate that the market does not recognize an outward, objective sign of
energy-efficient building operation. We expect that a statistical reason may also be leading to
insignificance of the labeling variable. First, if we remove BENCH from either equation, LABEL is both
positive and significant.26 Second, all labeled buildings are also benchmarked ones. This implies that
when we include both measures in the same model, we are measuring almost identical events.27 The
benchmarking variable has a larger variation and therefore consumes almost all of the explanatory power
of the combined measure.
Before we turn to estimating the value of the program and energy efficiency in general, we note
that the results associated with self-selection are mixed between the two formulations. In the pooled
model, the self-selection term is significant—indicating that self-selection is a significant factor in
explaining q values. In the switching regression models, the self-selection controls provide no explanatory
power—indicating that self-selection plays almost no role in explaining q values. Nevertheless, we retain
the self-selection framework because of the pooled model result and because there are good theoretical
reasons to believe that self-selection will be important component of this evaluation.
8. Estimates of Market Value
In this section we translate the regression estimates from Section 6 into estimates of the market
value of the ENERGYSTAR® program and of energy efficiency in general. There are four measures of
participation in the program that we are interested in:
26 These results are not reported here, but are available from the authors upon request.
27 The correlation between the two measures is 0.78.
74
-------
¦ The value of being an EnergyStar® partner;
¦ The value of benchmarking for EnergyStar® partners;
¦ The value of energy efficiency, as proxied by EnergyStar® partnership; and
¦ The value that non-partners would have earned had they joined EnergyStar®.
The first of these, the value of EnergyStar® partnership, is calculated using the ratio of the two
equations in the switching regression model. The value of benchmarking is calculated using the partner's
model in the switching regression model. The value of energy efficiency, is calculated using the pooled
model. The last measure, which we refer to as the lost opportunity of not joining EnergyStar®, is
calculated using the switching regression model in a manner analogous to the first measure.
To calculate these values we begin by calculating the "premium" associated with each measure of
participation. The premium is the defined as the q value associated with each participation measure
relative to the q value associated with not participating. For example, for valuing the EnergyStar®
program, the premium measures the value of q for being a partner relative to not being a partner. A
general expression for the premium can be written as:
Premium =
E\q | Participation]
E\q | No Participation\
- 1
x $1 million (20)
where £[•] is the expected value operator. Thus, the premium is the ratio of two predicted q values and is
converted into millions of dollars of a REIT's assets. To understand our formulation of this premium
consider the following. Before subtracting off one and multiplying by $1 million, the premium reflects
how much larger, in percentage terms, Tobin's q is for those participating in the program compared to
those who do not participate in the program. Tobin's q reflects the value the market places on a firm's
assets. Thus, before we multiply by $1 million, the premium reflects the percentage markup the market
places on the assets of those firms performing the activity. Subtracting one and multiplying by $1 million
converts the percentage to the value per million in assets.
75
-------
We then multiply the premium by the average level of assets for the REITs that are participants to
estimate the market value associated with the measure of participation. This follows from the fact that q
values provide the market's valuation of a firm's assets. Thus, multiplying by the level of assets converts
the premium into a market value. We begin by discussing the estimated premiums and market values for
joining EnergyStar®, benchmarking, and energy efficiency in general. We follow that with our estimate
and discussion of the lost opportunities of not joining EnergyStar®.
Value of .Joining EnergyStar®. Benchmarking, and Energy Efficiency
We estimate the value of the EnergyStar® program by using the estimated switching regression
model. The structure of this model allows us to estimate q values for partners if they had not joined
ENERGYSTAR®. Thus, we can compare the q associated with partnership to the q associated with not being
a partner for each partner in the sample at each time period. The q value associated with partnership is
calculated by generating predictions from the partner model for each partner observation. The q value of
partners had they not joined the program is calculated by generating predicted values for the partners
using the non-partner model. Maddala (1983) suggests looking at the difference between these two
numbers to gauge a program's effect. We use the ratio instead because relative q values are more easily
interpreted as ratios. That is, for partner the benefit of participating in the program at time t can be
calculated as
eujd,, -1] = *,„Pi - °i,y„,
EYi,„\Dn = 1] jrw.p2 _ vjrn
where X* is the Xl matrix with BENCH and LABEL removed.28 The average of equation (21) over all
observations where Dit = 1 provides an estimates of the average benefit of participating in the
EnergyStar® program. We use this average ratio in calculating the EnergyStar® premium.
28 In calculating (21) it is necessary to adjust the matrix Xl in the denominator or the vector (32 so
that their dimensions make multiplication possible. Specifically, BENCH and LABEL do not have
corresponding coefficients in (32. To handle this, we remove those variables from X] and re-label the new
matrix X{;'\
76
-------
To value benchmarking, we use the estimated model for partners in Table 3 to predict a q value
associated with some level of benchmarking (say B) and compare that to a predicted q value associated
with not benchmarking any buildings. This can be written as
E[qut | BENCH = B]
E[qut | BENCH = 0]
We use the median level of benchmarking among partners to compute the numerator of (22) and evaluate
all other variables also at their median levels. Equation (22) provides an estimate of the increase in q
values for partners that perform a median level of benchmarking, compared to performing no
benchmarking.29
As discussed in the econometric methods section, the estimated coefficient for the partnership
variable in the pooled model will reflect the sum of three separate effects: the direct effect of
EnergyStar® on q, the reputational effect associated with partnership, and the effect of a company's
propensity for being energy efficient on q. We interpret this combined effect as the effect of energy
efficiency, as proxied by EnergyStar® partnership, on intangible value. To calculate this value, we use
the pooled model and predict the q value associated with partnership (/) = 1) and compare that to the q
value associated with not being a partner (D = 0). The ratio of these predicted q values can be written as
E[qit | D = 1]
E[qit 1-0 = 0]
To calculate these predicted values, we set each variable in the equation equal to its mean value except for
D, BENCH, and LABEL. We set BENCH and LABEL both equal to zero to remove the influence of these
other program-related variables on our measure.
Table 4 provides our estimates for each of the three measures discussed above. We estimate that
the ENERGYSTAR® premium is $16,026 per million dollars of assets for REITs that are partners. That is,
for every million in assets, EnergyStar® partners earned an average return of $16,026 in intangible
29 The median amount of benchmarked space among the partners in our sample is 1.9 percent of
total floorspace per quarter per REIT. In constructing the benchmarking variable, a building was
considered to be "benchmarked" in a certain quarter if a benchmark score had been generated for it within
the last year. The average amount of benchmarked space in our sample of partners was 9.2 percent of total
floorspace per quarter.
77
(22)
(23)
-------
value above what they would have earned had they not joined the program. Based on our formulation and
our use of the switching regression model, we attribute this premium directly to a REIT's partnership in
EnergyStar®. Multiplying by the average level of assets for REITs that are partners, this premium
translates into an average market value of $51.67 million, representing 3.66 percent of the average market
value of these companies.
We estimate that benchmarking has earned partners a premium of $6,437 per million in assets.
Thus, partners performing a median level of benchmarking (i.e., 1.9 percent of total square footage)
earned an average return of $6,437 for every million in assets beyond what they would have earned if
they performed no benchmarking. This translates to an average market value of $20.75 million for
partners that benchmark, representing 1.47 percent of the average market value for these companies.
The estimated premium associated with benchmarking also indicates that active participation in
the EnergyStar® program has benefits. Not all EnergyStar® partners benchmark their buildings. Our
estimates indicate that a small amount of benchmarking (1.9 percent of total square footage) earns a return
of 0.64 percent on assets. This return is in addition to the return earned for being a partner. Thus, a
partner benchmarking about two percent of its floorspace earns a premium of $22,463 per million in
assets, or a return of 2.2 percent. This translate to $72.42 million in market value, representing 5.13
percent of the market value of REITs that are partners.
Finally, we estimate that energy efficiency earns REITs a premium of $45,564 per million in
assets. That is, taking participation in EnergyStar® as a measure of energy efficiency, REITs that are
partners earned a return of $45,564 for every million in assets above that of REITs that are not partners,
and presumably less energy efficient. This translates into a market value of $146.89 million, representing
10.4 percent of the market value of these companies.
Lost Opportunities from Not Joining ENERGYSTAR®
We now look at the extent to which REITs that did not join the program lost intangible value
from not joining.30 Above, we estimated the intangible value that partners would have had if they had not
30 The switching regression model assumes that partners and non-partners self-select into their
respective groups, with self-selection based on an evaluation of the (unobserved) net benefits of joining.
78
-------
joined ENERGYSTAR®, which provided a means of assessing the program benefits. In a similar manner, we
can use our estimated models to predict the q value for non-partners if they had joined the program. We
can calculate this in a manner analogous to equation (21). This can be written after making the
appropriate substitutions as:
E\-1u,\D« - "I = Kf>, + °,A,
ElhJD,: = 0] X2,fi2 + o2,#2„
In calculating (24), we need to make an assumption on how much benchmarking and labeling each non-
partner would have done under the program. Given that both are program-related activities, we think it
reasonable to assume that non-partners would have done no benchmarking or labeling. Thus, we add two
columns of zeros to X that correspond to these variables and call the new matrix^*. The average over all
/' and t where Dit = 0 will provide information on whether or not non-partners were better off not
participating in the program. A more interesting measure, however, is the average of all observations
where (24) is greater than one. These observations represent cases where the non-partners would have
been better off as a partner and can be called "lost opportunity."
We calculated equation (24) for each of 1,232 non-partner observations. The average value for
that calculation was 0.901, implying that, on average, non-partners had higher intangible value from not
joining ENERGYSTAR®. There are, however, 377 observations (30.5 percent of the 1,232) in which the
ratio exceeded one, implying that in those cases non-partners would have been better off by joining
EnergyStar®.31 These 377 observations were distributed across 50 different REITs (41 percent of the
non-partners). Thus, there were 50 REITs that would have been better off at some point between 1999
and 2001 had they joined EnergyStar®. Of these 50, 20 of them would have been better off as a partner
for the whole sample period. The average value of the ratio for observations that exceeded one was 1.099,
which translated to an average lost premium of $98,925 per million in assets for those observations.32
Thus, not joining EnergyStar® resulted in a significant level of lost value for non-partners.
Finding cases where non-partners would have had higher intangible values had they joined
EnergyStar® does not invalidate that assumption because intangible value is only one component of the
net benefit calculation.
31 Once again, an observation corresponds to a REIT in a specific quarter. Thus, there were 377
quarters in which REITs could have had higher intangible values by joining ENERGYSTAR®.
32 This estimate is large, but reflects only observations that exceed one.
79
-------
9. Summary and Conclusions
This paper looks at the relationship between participation in the EPA ENERGYSTAR® program and
a firm's intangible value. We used a sample of REITs measured quarterly from 1999 to 2001. We
constructed models of the relationship between Tobin's q, a measure of intangible value, and participation
in the ENERGYSTAR® program. Our models controlled for a number of factors, including self-selection
into the ENERGYSTAR® program by companies, market conditions, firm characteristics, and firm-level
financial factors. We found that the REITs involved in the EnergyStar® program received a return of
$16,026 for every million in assets above the amount they would have earned had they not joined the
program. Based on the modeling procedure that we used, we attribute this return to the EnergyStar®
program. We also found that ENERGYSTAR® partners that benchmark a small number of buildings (1.9
percent of their total floorspace in a quarter) earn a return of $6,437 per million in assets. We attribute this
benefit to activities that are associated with building benchmarking, such as efficient building operation.
Finally, we found that energy efficiency, as proxied by participation in the EnergyStar® program,
earned partners a return of $45,564 per million in assets. This return translated into 10.4 percent of the
market share of these companies. Thus, for REITs, where energy is a substantial concern, energy
efficiency represents 10.4 percent of the market share of the energy efficient companies.
Our results are applicable to three distinct areas. First, we have provided estimates of the value
created by a public voluntary program for program participants. The 1993 Government Performance and
Results Act (GPRA) requires federal agencies to assess their performance. Our results are directly
relevant to assessments of program performance under GPRA. Our results indicate that REIT partners
earn a return of $16,026 per million in assets for being a partner. Furthermore, these results are directly
attributable to the EnergyStar® program. Although the key results for the EnergyStar® program are
related to reducing energy-related environmental effects, our results provide valuable information on the
market value created by the program.
The results of this paper are also directly relevant for program recruitment efforts. The key for
voluntary environmental programs to improve environmental conditions is recruit participants that take an
active role in the program. Our results indicate that both joining the program and being an active
participant create value for REITs. As noted above, REITs that join EnergyStar® earned an average
80
-------
return of $16,026 per million in assets. Thus, joining EnergyStar® has value. Additionally, REITs that
took an active role in the program, as measured by benchmarking buildings, earned an additional return
of $6,437 per million in assets for only a modest level of benchmarking (1.9 percent of total floorspace in
any given quarter). Thus, a partner benchmarking about two percent of its floorspace earns a premium of
$22,463 per million in assets. Both of these results should provide valuable information for the program's
recruiting efforts.
Finally, our results should be directly relevant for investors and analysts that follow the REIT
industry stocks. The results from our pooled model indicates that energy efficiency earned REITs a return
of $45,564 per million in assets. This translated to a market value that represented 10.4 percent of the
value of these companies. To measure energy efficiency in our model, we used whether or not the REIT
was an ENERGYSTAR® partner. Thus, partnership in EnergyStar® offers a convenient measure of which
companies earn superior returns for their energy efficiency.
References
Arora, Seema and Timothy N. Cason, 1995. "An Experiment in Voluntary Environmental Regulation:
Participation in EPA's 33/50 Program." Journal of Environmental Economics and Management,
vol. 28, pp. 271-286.
Arora, Seema and Timothy N. Cason, 1996. "Why Do Firms Volunteer to Exceed Environmental
Regulations? Understanding Participation in EPA's 33/50 Program," Land Economics, vol. 72
no. 4, pp. 413-32.
Capozza, Dennis and Paul Seguin, 1999. "Focus, Transparency and Value: The REIT Evidence," Real
Estate Economics, vol. 27, no. 4, pp. 587-619.
Chung, Kee H. and Stephen W. Pruitt, 1994. "A Simple Approximation of Tobin's q." Financial
Management, vol. 23, no. 3, Autumn, pp. 70-74.
DeCanio, Stephen J. and William E. Watkins, 1998. "Investment in Energy Efficiency: Do the
Characteristics of Firms Matter?" Review of Economics and Statistics, vol. 80 no. 1, pp. 95-107.
81
-------
Greene, William H., 1993. Econometric Analysis, Macmillan Publishing Company, New York.
Henriques, Irene and Perry Sadorsky, 1996. "The Determinants of an Environmentally Responsive Firm,"
Journal of Environmental Economics and Management, vol. 30, pp. 381-395.
Hirsch, Barry T. and Terry G. Seaks, 1993. "Functional Form in Regression Models of Tobin's g,"
Review of Economics and Statistics, vol. 75, no. 2, May, pp. 381-385.
Innovest Strategic Value Advisors, 2002. Energy Efficiency and Investor Returns: The Real Estate Sector.
Khanna, Madhu, and Lisa A. Damon, 1999. "EPA's Voluntary 33/50 Program: Impact on Toxic Releases
and Economic Performance of Firms.'' Journal of Environmental Economics and Management,
vol. 37, pp. 1-25.
King, Andrew A. and Michael J. Lenox, 2001, "Does it Really Pay to Be Green V Journal of Industrial
Ecology, vol. 5, no. 1, pp. 105-116.
Koehler, Dinah and Don Cram, 2001. "The Financial Impact of Corporate Environmental Performance,"
Harvard School of Public Health, unpublished manuscript.
Konar, Shameek and Mark A. Cohen, 2001. "Does the Market Value Environmental Performance?," The
Review of Economics and Statistics, vol. 83, no. 2, pp. 281-289.
Lev, Baruch, 2001. Intangibles: Management, Measurement, and Reporting, Washington, DC: Brookings
Institute.
Maddala, G., 1983. Limited Dependent and Qualitative Variables in Econometrics, New York:
Cambridge University Press.
National Association of Real Estate Investment Trusts (NAREIT), 2003. "Annual Market Capitalization,"
Available at: http://www.nareit.com/researchandstatistics/marketcap.cfm.
82
-------
Parker, Gretchen, and Mark Chao, 1999. "REITs and Energy Management," Energy and Environmental
Management, August.
Parker, Gretchen, Mark Chao, and Victoria Gamburg, 1999. "Market Opportunities For Energy Service
Companies Among Real Estate Investment Trusts," Institute for Market Transformation, paper
presented at the 10th National Energy Services Conference, December.
Reed, Donald J., 2001. Pursuing the Elusive Business Case for Sustainable Development, World
Resources Institute, Sustainable Enterprise Series.
Romm, Joseph J., and William D. Browning, 1998. Greening the Building and the Bottom Line:
Increasing Productivity Through Energy-Efficient Design, Rocky Mountain Institute.
Videras, Julio and Anna Alberini, 2000. "The Appeal of Voluntary Environmental Programs: Which
Firms Participate and Why?", Contemporary Economic Policy, vol. 18, no. 4, pp. 449-461.
Wells, Richard, 2000. "Corporate Environmental Governance: Benchmarks Toward World Class
Systems," The Conference Board.
83
-------
Table 1. Definitions and Summary Statistics for Variables Used in the Analysis
Standard
Deviation
Variable
Definition
Mean
Tobin's q
Partners
Non-Partners
Ratio of market value to replacement value of assets.
0.9231
0.9453
0.9174
0.2084
0.2188
0.2062
PART
BMRATIO
LBRATIO
NRET
SBETA
TSQFT
Ln (TSQFT)
TA
Ln(TA)
CONC
%
ROA
OFFICE (office)
RET (retail)
1ND (industrial/
warehousing)
HC (health care)
LODGE (lodging)
APT (residential)
STOP (self-
storage)
Equal to one if REIT i was a partner in quarter t and zero otherwise
Percentage of buildings that are benchmarked for REIT i in quarter t.
Percentage of buildings that are labeled for REIT i in quarter t.
Quarterly return for NAREIT's Equity REIT Index, multiplied by 100.
Stock market beta calculated over the 1996-2001 time period using end-
of-month stock prices.
Total square feet owned by the REIT (millions).
The natural logarithm of total square feet.
Total assets in millions of dollars.
The natural logarithm of total assets.
Herfmdahl index of property-type concetration.
Average value of Tobin's q between 1996 and 1998.
Return on assets
Equal to one if more than 75 percent of the REIT's total square footage
was in the office sector and zero otherwise.
Equal to one if more than 75 percent of the REIT's total square footage
was in the retail sector and zero otherwise.
Equal to one if more than 75 percent of the REIT's total square footage
was in the industrial and warehousing sector and zero otherwise.
Equal to one if more than 75 percent of the REIT's total square footage
was in the health care sector and zero otherwise.
Equal to one if more than 75 percent of the REIT's total square footage
was in the lodging sector and zero otherwise.
Equal to one if more than 75 percent of the REIT's total square footage
was in the residential sector and zero otherwise.
Equal to one if more than 75 percent of the REIT's total square footage
was in the self-storage sector and zero otherwise.
0.1409
0.0151
0.0041
2.887
0.1271
32.59
16.523
1,558.8
6.933
0.8935
0.9076
0.0104
0.1262
0.5160
0.0753
0.0635
0.0962
0.1632
0.0251
0.3480
0.0748
0.0376
6.001
0.0969
70.57
1.301
2,255.3
1.175
0.1790
0.3106
0.0323
0.3322
0.4999
0.2439
0.2439
0.2950
0.3670
0.1564
84
-------
Table 2. Results for Probit Model For Joining ENERGYSTAR® Program (N = 1,434)
Variable
Estimated Coefficient
Marginal Effect On
Probability of Joining
EnergyStar® [a]
Constant
-4.402***
(-4.02)
-
Return to the NAREIT Equity Index (NRET)
0.026***
(3.01)
0.0035
Log of total square feet (ln(7:\6>/<7))
-0.038
(-0.352)
-0.0051
Log of millions of total assets (ln(7M))
0.418***
(3.14)
0.0557
Return on assets (ROA)
1.364*
(1.85)
0.1817
Return on assets, lagged one quarter (R()Ah])
1.551***
(2.02)
0.2065
Pre-1999 Tobin's q fq,j
0 749***
(4.573)
0.0997
Stock market beta {SBETA)
1 294***
(2.32)
0.1723
Residential sector (APT)
-7.883***
(-55.16)
-0.0961
Industrial and warehousing sector (IXD)
-0.386*
(-1.86)
-0.0506
Lodging sector (LODGE)
-7.343***
(-47.64)
-0.0961
Health care sector (!!(')
-7.343***
(-47.64)
-0.0961
Office sector (OFFICE)
1.202***
(9.94)
0.3634
Self-storage sector (STOR)
-7.664***
(-55.54)
-0.0961
Log-likelihood
-366.70
-
Likelihood Ratio Statistic
432.54***
-
Note: The base sector in this table is retail.
[a] This is the effect of each variable on the probability of joining the ENERGYSTAR® program. For the
continuous variables (NRET, 1 n(TSQFT), ln(E4), RO.l. ROA,_h qb. and SBETA), this is measured at the mean
value for each variable. For the remaining binary variables, this reflect the change in probability associated with
setting the binary variable equal to one.
85
-------
Table 3. Estimated Regression Models for Tobin's q
Switching Regression Model Pooled Regression
Variable
Partners
(N = 202)
Non-Partners
(N = 1,232)
Model
(N = 1,434)
Constant
-0.8087**
-0.0340
0.0709*
(-2.10)
(-0.52)
(1.82)
ENERGYSTAR® partnership {!))
-
-
0.0417***
(3.82)
Benchmarking {BENCH)
0.2966**
(2.10)
-
0.1445***
(2.19)
Labeling {LABEL)
-0.0460
(-0.20)
-
0.0950
(0.78)
Return to the NAREIT Index {NRET)
0.0034
(1.48)
0.0031***
(4.04)
0.0030***
(7.34)
Stock Market Beta {SBETA)
-0.4166*
(-1.96)
-0.0995*
(-1.96)
-0.1939***
(-7.15)
Log of Total Square Feet (lnC/:,V6>/<7:))
0.0593***
(2.83)
0.0269***
(7.41)
0.0243***
(11.83)
Concentration {CONC)
0.0327
(0.33)
0.1907***
(6.04)
0.1789***
(9.35)
Pre-1999 Tobin's Q (7/,,)
0.8300***
(9.98)
0.3810***
(25.64)
0.3974***
(49.8)
Return on Assets (RO.i,)
-2.7518*
(-1.78)
-0.1291
(-0.95)
-0.1522*
(-1.81)
Return on Assets, previous quarter
-1.508
-0.081
-0.0966
(ROAJ
(-0.97)
(-0.59)
(-1.14)
Office Properties Sector
-
-
-0.0623***
(-5.00)
Retail Sector
-
-
-0.0469***
(-5.668)
Self-Selection Correction
-0.0111
0.0015
-0.0066***
(Wlit, w2u, WJ
(-0.29)
(0.06)
(-5.62)
R2
0.8382
0.8184
0.8046
Adjusted R2
0.8298
0.8172
0.8028
Asymptotic t-ratios appear in parentheses.
*** Significant at the one percent level.
** Significant at the five percent level.
* Significant at the ten percent level.
86
-------
Table 4. Estimated Premiums and Market Value for EnergyStar® Program, Benchmarking, and Energy
Efficiency.
Measure
Estimated Premium
(Value Per Million
Dollars of Assets)
Market Value
(Millions)
Percentage of
Total Market
Value
Value of Being an EnergyStar®
Partner
$16,026
$51.67
3.66%
Value of Benchmarking Buildings for
Energy Performance
$6,437
$20.75
1.47%
Value of Being an EnergyStar®
Partner and Benchmarking Buildings
$22,463
$72.42
5.13%
(Active partnership)
Value of Energy Efficiency $45,564 $146.89 10.40%
87
-------
Session IV: Evaluation of Voluntary Programs
Discussant No. 1: Charles Griffiths, U.S. EPA, NCEE
COMMENTS ON:
The ISO 14001 Management Standard:
Exploring the Drivers of Certification
Andrew King
Darmouth College
and
Participation in Voluntary Programs, Corporate
Reputation, and Intangible Value: Estimating the Value of
Participating in EPA's ENERGY STAR® Program
Lou Nadeau
ERG, Inc
April 27, 2004
Recently, two of my colleagues, Ann Wolverton and Keith Brouhle, and I were asked to
write a book chapter on U.S. voluntary programs. We have been working on ways to evaluate the
effectiveness of voluntary programs, so this was a good way to review the literature. Our conclusion
was that the current economic literature has not produced strong evidence of improved
environmental performance due to voluntary programs. Some programs have been recognized as
improving performance, but, in general, voluntary programs have not yet been shown to produce
dramatic improvements. We do recognize, however, that there are a number of obj ective, other than
improved environmental performance, that might justify a voluntary program. Objectives such as
improved economic efficiency (that is, the program might produce greater net benefits to society);
savings in administrative, monitoring, and enforcement costs for the same environmental impact;
the inducement of innovation; or increased environmental awareness.
As you might guess, the response from individuals in the voluntary program offices was
mixed. Two responses, however, are of interest. First, one individual said that economists are
obsessed with economic efficiency. If it doesn't look and smell like cap-and-trade, then they don't
like it. His concern seemed to be that there are other, non-economic and non-quantifiable obj ectives
for these programs that economists miss. The second comment was that the text was very negative.
After recognizing that the economic literature may not have found much environmental impact of
these programs, I was asked if I could include a sentence along the lines of "that said, well designed
voluntary approaches can be a highly effective tool for environmental protection." The review
offered no additional basis for the inclusion of this sentences, but I understood that it reflected a
88
-------
deep belief that voluntary programs are an important component of the EPA mission, even if the
economists can't quantify it.
Certainly there has been a growth in voluntary programs. Our chapter identifies 55 voluntary
programs administered by the EPA and established since 1991. There is also anecdotal evidence
that they have had some impact. As one reviewer of our chapter informed me, Robert J. Eaton,
when he was chairman of Daimler Chrysler and chair of the National Academy of Engineering said,
"Life Cycle Management had convinced us that 'pollution prevention pays.' Not only doe it pay,
but it can be a competitive advantage."
In this session, we have heard two excellent papers that attempt to get at the advantages of
voluntary programs. Both are econometrically sophisticated. King, Lennox, and Terlaak (2004)
explicitly take into account the correlation between establishing an EMS and ISO 14001
certification. Personally, I expected to see a two stage approach, particularly if we believe that ISO
certification follows the establishment of an EMS. One approach would be to use an inverse Mills
ratio as in the next paper. Nadeau, Cantlin, and Wells (2004) use this approach to explicitly account
for self-selection into a program. In this paper, I would suggest allowing the participation variable
to affect the slope term as well as the constant.
If we look at the left hand side variables in these papers, we see a very common construction,
and one which I would like to talk about. Both studies looked at some outward mark of participation
in a voluntary program, but not necessarily at the environmental improvements that this participation
produces. King, et al. examine the establishment of an EMS or ISO 14001 certification in the
majority of their paper, and only look at the effect on environmental performance at the end.
Nadeau, et al. looked at the market value of participation in Energy star and not the environmental
benefit of the program. Both sets of authors recognize this point. Nadeau, et al. state that they "...
are looking at how good energy performance, rather than good environmental performance, relates
to financial performance." The decision to look at this relationship, rather than environmental
performance, may be related to the difficulty in finding measurable environmental gains due to
participation in these programs. As King, et al. note regarding their program, "many expected ISO
14001 to produce a means of credibly differentiating organizations with better environmental
performance. Our analysis suggests that this expectation went unfulfilled."
A more careful look at these analyses, however, may suggest a more complicated story. In
King, et al., regulatory pressure affects the establishment of an EMS and an EMS is a statistically
significant determinant of environmental improvements. EMSs, then, are important but are a less
transparent measure of corporate environmental actions. In contrast, the outward mark of good
environmental management, ISO 14001 certification, is not statistically affected by regulatory
pressure and does not have a significant impact on environmental performance. Similarly, in
Nadeau, et al., the measure of good environmental activity which is not witnessed by the market,
benchmarking buildings, is a significant factor in the Tobin's q premium. The outward label of
Energystar, on the other hand, is not statistically significant. So, while the outward mark of
environmental performance, ISO 14001 certification and the Energystar building label, is not a
significant driver in environmental or financial performance, the less evident measure is significant.
Why is this important? Because the outward mark is measurable and is often the only thing
available for evaluation. It is, however, confounded by the fact that it could simply be a measure
of EPA's efforts in recruiting partners. This has been the recent concern of OMB and others, that
voluntary programs shouldn't measure success by the number of partners they recruit, since it is
89
-------
measuring the desired output, environmental gains, by measuring inputs, the success in recruiting
partners. This may be why some researchers have not found big environmental gains by measuring
participation in voluntary programs - it is confounded by the effort in recruiting those partners. The
participation variable may be a noisy indication of the actual, unobservable measure of
environmental performance. To the degree that participation requires or increases the unobservable
measure of good corporate environmental activity, an EMS or benchmarking, then the voluntary
program is responsible for improved environmental performance. It may be difficult to evaluate,
however, because of a lack of appropriate data.
In this case, the authors were lucky to the appropriate data. King, et al. had TRI data, which
included a measure of EMS activity. Nadeau et al. had internal Energystar benchmarking data.
Many times, however, this type of data is not available, and it is precisely this lack of data that my
coauthors and I have found as a limiting factor for evaluating voluntary programs. It is hard to avoid
the conclusion that researchers have, so far, not found big environmental gains from voluntary
programs. The point is that this may be due to the fact that the outward mark of environmental
performance may only be a noisy indicator of actual underlying corporate activity. In other words,
and as my reviewers were trying to emphasize, there may be unquantified factors that economists
sometimes miss and perhaps we shouldn't be so negative about voluntary programs. Only more
accurate, underlying data and additional research will tell.
90
-------
Session IV: Evaluation of Voluntary Programs
Discussant No. 2: Jorge Rivera, George Mason University
COMMENTS ON:
The ISO 14001 Management Standard:
Exploring the Drivers of Certification
Andrew King
Darmouth College
and
Participation in Voluntary Programs, Corporate
Reputation, and Intangible Value: Estimating the Value of
Participating in EPA's ENERGY STAR® Program
Lou Nadeau
ERG, Inc
1. Strategic enactment of a new institution: Exploring the causes of certification
with the iso-14001 management standard. By King, Lenox, and Terlaak
This is an excellent paper from both the theoretical and empirical perspectives. Using a
strategic analysis approach the manuscript develops a conceptual framework and
hypotheses to explain under what circumstances corporate facilities are more likely to
certify with ISO-14000. The authors argue that supply chain's information asymmetries,
which make the exchange of credible environmental reputation difficult, are one of main
reasons why facilities decide to obtain ISO-14000 certification. These hypotheses are
then tested using a proportional hazard model and 1995-2002 panel data for a sample US-
based facilities.
Their findings suggest that ISO-14000 certification is used as a signal of environmental
improvement efforts rather than an indication of superior environmental performance.
Certification is more likely for facilities that are more distant in terms of geography and
culture and for those with long term or vertically integrated associations to downstream
buyers. My main criticism to King, Lenox, and Terlaak's approach is that they use US-
based facilities to assess ISO-14000, an international standard whose participants are
mostly outside the US.
2. Participation in voluntary programs, corporate reputation, and intangible value:
Estimating the value of participating in EPA's Energy Star Program. By Nedeau,
Canting, and Wells.
91
-------
This paper relies on quarterly 1999-2001 data of 124 real investments trusts to estimate
the market value (measured as Tobin's q) of the Energy Star Building program. The
authors suggest that the Energy Star provides its partners with benefits of about sixteen
thousand dollars per each million dollars of assets.
I believe that the manuscript uses a very good approach and valuable data to try to answer
a critical question regarding the use of voluntary environmental programs as an
alternative policy tool to promote environmental protection. Yet, I would encourage the
authors to improve the paper in the following areas:
1. The review of the literature needs to incorporate the research on voluntary programs
published in the management and public policy journals.
2. Given its focus on intangible assets and capabilities, the authors may find it valuable to
incorporate arguments from the resource view of firm to support their hypotheses,
discussion, and conclusions.
3. The authors put a big emphasis on the importance of reputation but only use an
instrumental variable that accounts for this and other constructs. Thus, it is necessary to
incorporate a direct measure of reputation in the analysis.
4. The issue of reverse causality needs to be addressed in the discussion and conclusions.
5. Finally, the authors need to include a section that explicitly highlights the limitations of
their approach and findings.
92
-------
Summary of the Q&A Discussion Following Session IV
Matt Clark (U.S. EPA, Office of Research and Development)
For people who are looking for money for data, I understand the Japanese and European
economies are doing pretty well right now. [laughter]
I'm relaying a question from William D'Alessandro from Crosswinds Bulletin—one of
the people who have joined in on the phone and the internet: Mike, could you re-state,
for Bill's benefit, why companies certify ISO 14001?
Michael Lenox (Duke University)
In responding to this question, Dr. Lenox clarified that their research indicates that
certifying "seems to be signaling simply that you have an EMS in place and not
signifying some secure, underlying environmental quality."
Eric Otis (University of Pennsylvania)
Addressing the first of two questions to Dr. Lenox, Dr. Otis referred to the literature that
TRI commonly uses and asked "whether a footnote isn't appropriate there as to what it
really is measuring, because [he thinks] it's at least possible to say that EMS's and ISO
14000 programs actually are providing some environmental performance benefits that are
not captured by the TRI measure." Characterizing the TRI measure as "very high level,
gross information," Dr. Otis said he doubted whether this was the best measure for
gauging firm-level environmental performance, and he advised Dr. Lenox to "at least
qualify your result on that point."
Dr. Otis directed his second comment to Louis Nadeau regarding the correlation between
"better firms, in terms of monitoring conditions, etc." and participation in the Energy Star
Program. Dr. Otis questioned the assumed causation direction of the correlation and said
he believes the causation can go both ways. In other words, since it is as likely that
"better firms can afford to be doing Energy Star" as it is that "Energy Star [participation]
is worth more money, . . . you really wouldn't want to conclude that there's a huge
amount of value in Energy Star. It may be going the reverse direction—those firms that
are better managed already can afford to do Energy Star Programs, which may be
providing environmental benefits, but it's not then as clear what follows from what." Dr.
Otis wondered whether Dr. Nadeau had accounted for "that potential reverse causation,"
which, he believes, "shows up in a lot of other studies as well."
Michael Lenox
Dr. Lenox responded that because they were dealing with manufacturing firms ... at the
facility level, "at some level, emissions is a good measure of facility-level environmental
liabilities." He went on to acknowledge that what Dr. Otis said is "absolutely correct," in
that the researchers used this measure "as a proxy for some kind of unobserved
environmental quality, and there could be a number of attributes and elements in that."
93
-------
Dr. Lenox went on to say that he and his colleagues, and other researchers as well, really
should have been and really need "to be curious about the correlation between TRI
measures and things like NOx-SOx emissions, accidents, violations and the like." He
closed by saying that it was his hope and speculation "that there is some significant
correlation between these various metrics," but he isn't aware of anyone who has actually
performed an analysis of that as yet.
Lou Nadeau (ERG, Inc.)
Dr. Nadeau responded to the second question from Dr. Otis pertaining to the probable bi-
directional nature of causation between a firm's participation in the Energy Star Program
and that firm's financial health by saying, "It's a valid point—clearly, better companies
tend to join Energy Star, and so any sort of correlation in compliance needs to be
interpreted in that light." He said that they used the Heckman self-selection tool in the
first stage of the study to capture the first level of causation, from corporate value to
participation in Energy Star, and are "hopefully getting at the causation running from
participation to value" in the second stage of the study. He closed by affirming that the
team is attempting to control for as many things as possible.
Jon Silberman (U.S. EPA, Office of Enforcement and Compliance Assurance)
Mr. Silberman opened by stating, "First, I'd just like to point out that there is, in fact, an
extensive amount of research out of Europe that also looks at the question of ISO 14000
certification and its relationship to performance, but since we have one of the leading
European researchers, Chris Howes from the United Kingdom Environment Agency,
waiting to speak at the other microphone, so I'll just stop there."
Mr. Silberman continued, directing this comment to Dr. Lenox: "I'm wondering if ISO
14000 coupled with certification is not functioning very similar to how a rule might
function as a mandatory requirement out of government followed by inspections." He
clarified this idea by adding, "if people who adopt EMS's without certification are doing
better than people who certify, is the certification process potentially dumbing down
people's EMS's by making them managed towards achieving a piece of paper that, based
on my ISO 14001 auditor training and years of experience with ISO 14000, is quite easy
to get and totally dissociated from actual performance?"
Michael Lenox
In response to the availability of data from Europe, Dr. Lenox commented that these data
are often difficult to use from a researcher's perspective due to the lack of comparability
across the data sets. As an example, he suggested that finding a European measure "that
maps very nicely to TRI. . . would require some coordination that. . . would be very
difficult to achieve."
Addressing Mr. Silberman's second issue, Dr. Lenox stated that "suggesting that the
EMS is great and then the certification dumbs it down . . . would be an incorrect
interpretation of our findings. It's simply the fact that there are forces that are driving the
94
-------
adoption of EMS's independent of certification, and that often the existence of an EMS
makes it much easier, then, for you to get certification."
Chris Howes (Environment Agency, England and Wales)
Mr. Howes cited two European studies of environmental management systems—one that
looked at performance data and compliance history from over 2,500 regulated sites, and
another pan-European industry study that looked at data from 450 sites. He said these
two studies had fairly common findings: basically, "there is no correlation between good
environmental performance or compliance and certification with ISO 14001, or indeed
registration to EMAS," (a European program that Mr. Howes characterized as "ISO
14000 plus"). Mr. Howes went on to say that "the very clear message with regard to
legal compliance is: If we as a regulator wanted to target sites based on whether or not
they had EMS's, which seems to be sensible thing to do, . . . we should target those sites
with ISO 14001 or EMAS because they are more likely to have noncompliances and they
are more likely to have poor environmental performance."
Citing a current 3-year project being managed by the Environment Agency (the REMAS
Project—more information available at www.remas.info). Mr. Howes advocated looking
at broad "benchmark performance and the existence, or otherwise, of EMS's at a much
greater level of sensitivity" than is typical with ISO or REMAS. "It's looking at the
elements of a management system that are in place and comparing those to compliance
and to the normally regulated issues in terms of emissions to air, land, and water." Mr.
Howes also mentioned the more recent regulatory categories for major industry in Europe
of energy efficiency and resource efficiency. Getting to his main point, Mr. Howes
asked, "if certification and registration in the U.S. and the rest of the world doesn't add
value, shouldn't this be the issue for all stakeholders in ISO 14001—the public,
regulators, and industry? . . . Shouldn't we be pushing for outcomes, not process, from
ISO—from the accreditation bodies, from registrars?"
According to Mr. Howes, UKAS, the accreditation service in the U.K. that is somewhat
equivalent to RAB [Registrar Accreditation Board, established in 1989 by the American
Society for Quality], has recently come out and said that, essentially, "ISO 14001 is not
driving improvement; ... the qualifications of offices aren't good enough; the
accreditation bodies aren't good enough." Ultimately, Mr. Howes wonders, "What are
we [including the U.S. EPA and others] going to do about this?"
Jay Benfor ado (U.S. EPA, National Center for Environmental Innovation)
Before turning to the paper writers for their responses, Mr. Benforado paraphrased the
question as: "Could you foresee some utility in certification of performance rather than
certification of process?"
Michael Lenox
Dr. Lenox responded, "First of all, I take a little issue with the idea that certification
doesn't add value. The question is: Who does it add value to and to what ends? I think it
95
-------
does probably add value for those in the supply chain who want to try to have some,
perhaps, management over those facilities. From a public policy standpoint, does this
lead perhaps to a reduction in environmental emissions and the like? Once again, we're
finding that might not be working the way we had hoped. So, should public policy
perhaps get involved and step in and try to put more teeth in something like ISO 14001?
Perhaps, but to the extent you're interested in self regulation, it begins not to look like
self regulation much anymore, obviously, with the EPA stepping in and mandating and
dictating."
Allison Christie Sajan (Natural Resources Canada)
Ms. Christie Sajan said that Natural Resources Canada has been looking at many of these
same questions and that the companies they have heard from who have applied various
types of environmental management tools, such as EMS's, believe they are realizing real
benefits from these efforts. She commented that they are in the first phase of a 3-year
study of companies that have not employed an EMS and would welcome any dialogue or
suggestions.
Madhu Khanna (University of Illinois)
Dr. Khanna raised a question "related to the result. . . that firms who have an EMS did
show some improvement in environmental performance but not the ones that actually got
certified." She went on to say, "If I understand correctly what you're doing, you're
looking at firms that just started that have an EMS, and the second group is firms that
have an EMS but also got certified." Dr. Khanna concluded by saying she was "really
intrigued by why it is that firms that actually went ahead and got the certification, which
presumably verified that their EMS had all the right elements and so on, did not achieve
the environmental improvements that the other firms did."
Michael Lenox
Dr. Lenox replied, "I apologize for that—that is not correct—the interpretation is that
that is the pool of all who get EMS, not just the ones who don 7 get certified. So, we're
looking at two pools here—those who get EMS and that effect on improvement, and then
certification, which is a sub-sample of those who have an EMS. So, the argument, which
is not surprising actually, is that certification in and of itself does not lead to any
improvement, and I'm not sure why it would be expected to. If you have a functioning
EMS, that's what should lead to improvement—not the certification per se. So, to be
clear on that, in that pool of [firms with an] EMS is both those who get certified and
those who don't get certified."
Madhu Khanna
Dr. Khanna then suggested investigating whether having an EMS and certification
together is better than just having an EMS.
Michael Lenox
96
-------
Dr. Lenox responded that "supposedly the certification variable tries to pick that up, and
we don't find that. . . . Again, to be clear, there are incentives, perhaps, to improve
environmental performance that drive you to adopt an EMS. There are other kinds of
incentives that are dangled giving you incentives to certify that are not necessarily
commensurate with that."
97
-------
Corporate Environmental Behavior and the
Effectiveness of Government Interventions
PROCEEDINGS OF
NEW GRANTEES IN CORPORATE ENVIRONMENTAL BEHAVIOR
A WORKSHOP SPONSORED BY THE U.S. ENVIRONMENTAL PROTECTION
AGENCY'S NATIONAL CENTER FOR ENVIRONMENTAL ECONOMICS (NCEE),
NATIONAL CENTER FOR ENVIRONMENTAL RESEARCH (NCER)
April 26-27, 2004
Wyndham Washington Hotel
Washington, DC
Prepared by Alpha-Gamma Technologies, Inc.
4700 Falls of Neuse Road, Suite 350, Raleigh, NC 27609
-------
ACKNOWLEDGEMENTS
This report has been prepared by Alpha-Gamma Technologies, Inc. with funding from
the National Center for Environmental Economics (NCEE). Alpha-Gamma wishes to
thank NCEE's Cynthia Morgan and Ann Wolverton and the Project Officer, Ronald
Wiley, for their guidance and assistance throughout this project.
DISCLAIMER
These proceedings are being distributed in the interest of increasing public understanding
and knowledge of the issues discussed at the workshop and have been prepared
independently of the workshop. Although the proceedings have been funded in part by
the United States Environmental Protection Agency under Contract No. 68-W-01-055 to
Alpha-Gamma Technologies, Inc., the contents of this document may not necessarily
reflect the views of the Agency and no official endorsement should be inferred.
11
-------
TABLE OF CONTENTS
New Grantees in Corporate Environmental Behavior
Pollution Prevention: The Role of Environmental Management and
Information
Madhu Khanna, University of Illinois 1
The Effect of Self-Policing on Hazardous Waste Compliance
Sarah Stafford, College of William and Mary 10
Comparative Plant-Level Analysis of Three Voluntary Environmental
Programs
Richard Morgenstern, Resources for the Future 15
Oregon Business Decision for Environmental Performance
David Ervin, Portland State University 21
in
-------
Pollution Prevention: The Role of
Environmental Management and
Information
Madhu Khanna, George Deltas, Satish Joshi
Donna Ramirez
University of Illinois and Michigan State University
Pollution Prevention (P2)
• Promises several advantages over end-of-pipe controls
- Focuses on multi-media pollution control
- Prevents trace emissions/bio-accumulative pollutants
- Requires greater integration between environmental and
business decisions; encourages innovation and cost-
effectiveness
• Waste reduction at source implies increased efficiency
in production
- Potentially higher profits and a win-win strategy
- Differentiated products that respond to environmentally
conscious consumers
- Reduced environmental risks to shareholders
- Improvement in corporate reputation
1
-------
Approaches for Pollution Prevention
• Regulatory agencies encouraging P2
- Voluntary programs, technical assistance and training
- Environmental Leadership Programs/Adoption of Enviromnental
Management Systems
- Information collection and disclosure through the Toxics Release Inventory
• Requires reporting on toxic releases and adoption of 8 types of P2 activities
Firms are
- Participating in stewardship programs
- Adopting Enviromnental Management Systems (EMSs): Total Quality
Enviromnental Management (TQEM)
• Seeking continuous progress in reducing waste and improving product quality
• Undertaking internal environmental audits, employee training and involvement
• Making process and product modifications to increase efficiency and reduce
Motivation for this Research
• What is motivating some firms to adopt EMSs/TQEM and P2?
- Which types of firms are more likely to adopt?
• Do EMSs encourage P2 and which types of P2
- Visibility of EMSs may provide stakeholder benefits to firms even in
absence of P2
- Some P2 is costly; less observable by public
- Adoption rates of TQEM high (50%) but of P2 low (25-33%)
• Is P2 effective in improving environmental performance of firms?
• Does pollution prevention really pay?
- Which types of P2 in particular and for what types of firms?
2
-------
Theoretical Issues Addressed
Can market pressures (consumer preference for green
products) motivate P2 as a strategy to differentiate
products and achieve social optimality?
Are supplementary regulations needed (minimum
quality standards, taxes/subsidies) and their
implications for social welfare, firm profits and prices
Incentives for P2, EMS adoption and social
optimality of market based pressures when all
consumers cannot observe P2 but can observe a
firm's EMS
Empirical Analysis
Motivations for P2: Role of TQEM and information
provision about toxic releases
Impact of P2 adoption on environmental
performance
Impact of P2 on economic performance of firms
- Event study analysis of impact of P2 and EMS
adoption on stock market reactions to toxicity
weighted TRI
- Impact of P2 on expected future profitability of
firm, price earnings ratios and market shares
-------
Theoretical Framework
Assumptions
- Product attributes:
• Greenness
• Reliability
• Others: style, design, convenience
- All consumers care about greenness to same degree; differ in
preferences for other attributes
- Consumers willing to internalize the externality to some extent
- Greenness measured by emissions intensity (P2)
- Consumers can observe extent of P2
- Rival firms in an industry seek to differentiate their products
- Increasing greenness of product by a firm imposes fixed costs
that increase with greenness
• can lead rivals to match greenness or lower prices
Theoretical Framework
Q.
Location of consumers by preference
Market share of A
Market share of B.
Other
Attributes of
Product by
Firm A J
Other \
Attributes of
Product by j
Firm B /
Reliability and other attributes of brands pre-defined
Firms choose greenness and product prices to maximize profits
Consumers choose the product that maximizes their benefits net of prices
4
-------
Specific Questions Investigated
• Whether firms with a higher intrinsic quality are
more/less likely to choose more P2
• Incentives for P2 due to
- Impact of increased consumer awareness about environmental
attributes of products
- Cost sharing of P2 by regulators
• Impact of P2 on market shares, prices and profits
• Whether consumer preferences are sufficient to achieve
socially optimal level of P2 by all firms
• Implications of minimum quality standards, taxes/
subsidies for P2, firm profits and social welfare.
Initial Findings
When consumers observe and care about product greenness
- Firm with higher intrinsic quality does more P2, charges a higher
price and has a greater market share than rival firm
- Even if consumers fully internalize the environmental externalities,
market pressures will not lead to an optimal provision of the
environmental attribute
• Need to supplement market pressures with regulatory intervention
- Impact of a minimum quality standard on social welfare is
ambiguous
• Higher quality firm may overcomply with standard but would do less
P2 than in absence of standard
Work in Progress
Implications of product quality taxes/subsidies and cost sharing
policies for P2, firm profits and social welfare
Implications for P2 and social welfare when only some consumers
observe product greenness but all care about it and firms adopt an
EMS to indicate product quality
-------
Sources of Data for Empirical Analysis
• Adoption of TQEM: IRRC surveys 1994-96
- 228 parent company level observations each year
- 3500 observations at the facility level each year
Toxic Releases and P2 activities: TRI
- Types of P2 activities: Good operating practices;
Spill and Leak Prevention; Process and Product
Modification
- On-site Releases, Off-site Transfers, Hazardous Air
Pollutants
Superfund sites, inspections and civil penalties: IDEAS
data
Financial Performance: Research Insight Data
Environmental Pressure Indicators: Census and other
sources
Sample of Firms: S&P 500 firms that report to TRI and
completed IRRC survey 1994-96
% of TQEM Adopters
m\ mi
Toxic Releases
it,
l=ffltJ
^TT1
Nil
19
4 1995
Nil
1996
~ ONSITE
¦ OFFSITE
~ TOTAL TRI
¦ TRI/SALES
30
25
20
15
10
5
0
Adoption of P2 Activities
~ Good operating
practices
¦ Process and Product
Modifications
Spill and Leak
Prevention
~ " Total P2 activities"
~ "% of P2
opportunities"
12
6
-------
Motivations for TQEM
Probit Analysis using Panel Data Methods
Explanatory Variables
Effect
Market Pressures: Final Good
Market Share
Asset/Sales
+
+
+
Regulatory Pressure: NPL sites
Volume of HAP
Civil Penalties
Frequency of Inspections
+
+
Information Provision: Off-site Transfers
On-site Transfers
+
Finn Characteristics: Size
R&D expenditures
Older Assets
+
+
+
+: Significant positive effect; Others Insignificant effect
Motivations for P2
Explanatory Variables
Spill and
Leak
Prevention
Process/
Product
Modification
Off-site
Transfers/
Sales
TQEM
+
-
Market Pressures: Final Good
Market Share
:
;
Regulatory Pressure: NPL sites
Volume of HAP
Civil Penalties
Frequency of Inspections
:
:
Info. Provision: Off-site Transfers past
On-site Transfers past
Number of TRI Records
+
;
+
Firm Characteristics: Size
R&D expenditures
Older Assets
Sales/Asset
;
Facilities in Non attainment Counties
Facilities in States with high compliance
expenditure
+
7
-------
Impact of TQEM on Process/Product Modification
Activities Varies Across Finn Types
• Firms in the Top Quartile of R&D expenditures:
- Larger R&D expenditure more likely to lead to more P2
- TQEM has an insignificant impact on P2
• Firms in the lower 3 quartiles of R&D expenditures
- Larger R&D expenditures less likely to lead to more P2
- TQEM has a positive impact on P2
• Firms in the top 3 quartiles of market share
- Firms with larger market share more likely to do P2
- TQEM has a positive significant impact on P2
Event Study Analysis: Stock Market Reactions
to Toxic Release Information
Hypothesis to be Tested:
• HI: There is a significant negative association between the
quantity of pollutants released and a firm's abnormal stock
returns.
• H2: The toxicity level of the releases is negatively associated
with the stock market returns.
• H3: Higher P2 activities exhibit a positive association with
stock market returns
• H4: A firm's degree of readiness to improve its environmental
performance, signaled by its adoption of an EMS, and it's
stock market returns are positively associated.
8
-------
Other Work in Progress
• Analysis of effects of TQEM and other
practices (such as corporate reporting) on P2
• Facility level analysis of impact of P2 and
source of information/assistance on P2 on
Toxic Release performance and on criteria
pollutants
• Impact of P2 on financial performance of firms
9
-------
i The Effect of Self-Policing on
Hazardous Waste Compliance
Sarah L. Stafford
The College of the William and Mary
EPA's Corporate Environmental Behavior Workshop April 27, 2004 Research Supported by STAR Grant R831036
Objectives
¦ Determine whether self-policing policies have
affected compliance with hazardous waste
regulations.
Understand the extent to which companies use
self-policing.
. Develop compliance model that incorporates self-
policing.
« Provide feedback on the effectiveness of self-
policing policies.
EPA's Corporate Environmental Behavior Workshop April 27, 2004
10
-------
Self-Policing
¦ Self-Policing: a situation in which a regulated
entity notifies authorities that it has violated a
regulation or law.
.. Not necessarily the same as self-reporting.
Federal "Audit Policy" encourages self-
policing by reducing or eliminating penalties
for self-disclosed violations.
« State self-policing policies and environmental
audit privilege and immunity laws also
encourage self-policing.
EPA's Corporate Environmental Behavior Workshop April 27, 2004
Theoretical Framework
¦ A self-policing policy without a change in
inspection targeting or fines cannot increase
compliance.
: Can increase environmental protection by
requiring remediation.
.. Should only effect inadvertent, not willful,
violations.
* If a self-policing policy is combined with a
redistribution of enforcement it can increase
compliance.
: Can affect willful violations as well as inadvertent.
-------
Empirical Analysis
¦ Ideal analysis would consider effect of self-
policing policies on auditing, self-policing, and
compliance.
.. Comprehensive data not available.
* First, look for changes in compliance
behavior after imposition of federal and state
policies.
» Based on results, conduct more focused
analysis.
EPA's Corporate Environmental Behavior Workshop April 27, 2004
Initial Analysis
¦ Use panel data on inspections and detected
violations before and after imposition of federal and
state policies.
.. Probability of inspection, and thus probability of detection,
is not exogenous.
.. Use censored bivariate probit with errors clustered by
facility.
¦ Data for 9,500 hazardous waste facilities from 1992
to 2001.
. No newly regulated facilities, one-time generators, small
quantity generators, or federal facilities.
-------
Initial Results
Federal policy accompanied by change in targeting,
but no significant change in overall level of
violations.
State policies appear to have had a more significant
effect:
... States with audit privilege only: lower probability of
inspections and violations.
.. States with audit privilege and immunity: higher probability
of inspection, lower probability of violation.
. States with self-policing: lower probability of inspections
and violations.
Questions Still to be Answered
¦ Is the change in targeting due to self-policing
policies or merely coincident?
b# Can the change in violations be attributed to
self-policing or are there other causes?
iPA's Corporate Environmental Behavior Workshop April 27.
-------
Follow-up Analysis
Use data on 2001 self-disclosures to examine
impact of self-disclosure on probability of an
inspection.
.. Challenges:
¦ How accurate is the data on self-disclosures?
¦ Are there enough self-disclosures to make estimates?
Find data on audit adoption to determine
whether increased auditing could be
responsible for decreased violations.
.: Possible sources?
-------
Comparative Plant-level Analysis of
Three Voluntary Programs
Richard Morgenstern
William Pizer
Jhih-Shyang Shih
April 27, 2004
I RESOURCES
ft row THE FUTUW*
Status of Voluntary Programs for
Environmental Protection
• 00's in Germany, Netherlands: national
government, industry associations
• 000's in Japan: local agencies, firms
• in U.S. 'public voluntary programs" or
'government lead challenges' popular
• 54 EPA programs in 1999, up from 28 in
1996
• U.S. climate policy dominated by
voluntary efforts: EPA, DOE, DOA
fa*"***88'* RESOURCES
-------
Potential Advantages of Voluntary
Programs
• Increased flexibility for government
and industry
• Reduced confrontation
• Reduced transaction costs,
litigation, etc.
• Pilot test new approaches,
especially in absence of legal basis
for mandatory program
RESOURCES
uni-Till'" row THE FUTUW*
Are Voluntary Programs Really
Effective?
• Concerns expressed that programs
do not push firms beyond baseline
performance
• Without regulatory or price signals
few incentives to develop/use new
technologies
• Shifts emphasis from 'worst'
polluters to those willing to act
voluntarily
O RESOURCES
16
-------
Two Types of Voluntary Programs
• Focusing on particular technologies,
e.g., Green Lights
• Focusing on environmental
performance, e.g., 33/50, Climate
Wise, 1605b
I RESOURCES
ft row THE FUTUW*
Goal of Research
Expand understanding of
environmental effectiveness as well
as efficiency of voluntary programs
- Current information is often too
aggregate, without clear baseline
- Pollution prevention and GHG
reduction are growing areas of policy
interest
fa*"***88'* RESOURCES
-------
Principal Contributions of
Research
• Shift focus from firm-level to plant-level
analysis, thereby controlling for changes in
output, other key factors
• Improve modeling of participation,
emission reductions: focus on differences
between participants and non-participants
• Expand breadth of academic-style studies
beyond 33/50 to include GHG reduction
programs
• Validate/improve data quality
Plant-level Data
• Unlike most previous studies which rely
on firm-level information, focus is on
plant-level data
• Available on confidential basis from US
Census (LRD, QFR)
• Need to link Census data with public
information: 33/50, Climate Wise, 1605b
• Builds on researchers" previous
experience with Census Bureau data
RESOURCES
-------
Methodology
Problem: firms self-select to join
programs. Thus participation is not
random
Method 1: Ignore problem
Method 2: Condition selection on
observable data, e.g., size, profits, etc
Method 3: Condition selection on
unobservable data (analyze residuals)
(Heckman & Hotz)
55= RESOURCES
Early Progress
• STAR grant awarded Fall, 2003
• Initial focus on literature review,
assembling publicly available data,
formal approval from Census
(Predominant Purpose Statement)
• Currently on second round of PPS
reviews
• Optimistic about near-term approval
resources
-------
Expected Research Results
• Key characteristics of program
participants vs non-participants
• Environmental performance of
participants vs non-participants
• Factors influencing performance
including size, profitability, industry, firm
type, early/late joiner, etc
• Inter-program comparisons
• Effect of program participation on
performance in other areas
saws****58 RESOURCES
-------
Oregon Business Decisions for
Environmental Performance
U.S. EPA Funded Project on Corporate
Environmental Behavior and Effectiveness
of Government Intervention
Portland State University, University
5/17/2004 of Illinois and Oregon State 1
University
Project Team
• David Ervin, PI/PD, Portland State II.
• Madhu Khanna, PI, U. Illinois at
Champaign-Urbana
• Patricia Koss, PI, Portland State U.
• Junjie Wu, PI, Oregon State U.
• Cody Jones, GRA, Portland State U.
5/17/2004
Portland State University, University
of Illinois and Oregon State
University
21
-------
Project Objectives
1. Identify and measure the major elements
of environmental performance, e.g., toxic
waste compliance, solid waste recycling
and water use efficiency, for Oregon firms.
2. Collect primary data on the set of
environmental practices used by a random
sample of Oregon firms.
Portland State University, University
of Illinois and Oregon State
University
Project Objectives cont'd
3. Collect data on firm, industry, regulatory,
and 'voluntary' environmental program
factors hypothesized to influence the
environmental performance.
4. Test the influences of firm, industry,
regulatory conditions, simultaneously with
voluntary program factors, on the
adoption of environmental practices.
5/17/2004
Portland State University, University
of Illinois and Oregon State
University
22
-------
Project Objectives cont'd
5. Test the influences of firm, industry,
regulatory, and voluntary program factors
on firms' environmental performance.
6. Infertile 'voluntary' program features
(e.g., practices and incentives) and other
conditions that significantly improve firm
environmental performance.
Portland State University, University
of Illinois and Oregon State
University
Hypotheses
1. The decision to adopt a particular
environmental practice is related to
characteristics of the firm, industry, and
regulatory environment, as well as
voluntary program incentives.
2. The environmental performance induced
by a particular environmental practice is
also related to specific firm, industry, and
regulatory characteristics and program
incentives.
Portland State University, University
5/17/2004 of Illinois and Oregon State 6
University
23
-------
Hypotheses cont'd
• The effects of the firm, industry, and
regulatory characteristics and program
incentives vary across environmental
performance elements, e.g., toxic releases
and solid waste recycling.
• The effects of the firm, industry, and
regulatory characteristics and program
incentives on environmental performance
vary across sectors, e.g., building
construction, agriculture.
Portland State University, University
5/17/2004 of Illinois and Oregon State 7
University
Major Activities
• Review potential environmental programs
available to Oregon industries
• Conduct industry focus groups to identify
practices and performance measures
• Select stratified sample of firms
• Implement survey with Washington State
University survey research center
• Conduct multivariate analyses to test
hypotheses
Portland State University, University
5/17/2004 of Illinois and Oregon State 8
University
24
-------
Approach/Methods
• 2-stage model to analyze, simultaneously,
the determinants of program participation
and environmental performance.
• 1st stage -- firm's choice of environmental
plan (or combination of practices)
• 2nd stage -- explanation of environmental
performance as influenced by firm,
industry, regulatory and program factors
Portland State University, University
5/17/2004 of Illinois and Oregon State 9
University
Approach/Methods
• Polychotomous-choice selectivity
model to address self selection bias
and interaction between practices
• Stratified random sample to assure
sufficient number of participating
and non-participating firms
Portland State University, University
of Illinois and Oregon State
University
25
-------
Planned Schedule
Environmental program review 1-5/04
• Industry Focus groups 6-9/04
• Survey nstrument design 6-9/04
• Sample selection 8-9/04
• Survey enumeration 10/04- 3/05
• Data cleaning 4/05-6/05
• Analysis 7/05-12/05
• Writing and outreach 1/06-9/06
Portland State University, University
5/17/2004 of Illinois and Oregon State 11
University
Progress ^Environmental
Program Review
• Many environmental programs are
available to Oregon firms.
• Participation may be affected by
business composition - 97.6% of
Oregon firms are classified as small.
• Most programs allow firms to choose
best environmental practices.
Portland State University, University
5/17/2004 of Illinois and Oregon State 12
University
26
-------
Progress sEnv iron mental
• A preliminary finding is that certain
practices appear to be common
across programs and industries
-Supply chain management
-Employee behavior modification
-Environmental personnel
-Training - employees, contractors,
vendors
Portland State University, University
5/17/2004 of Illinois and Oregon State 1|
University
Progress - Environmental
Program Review
Selected EPA Cross-Sector Programs
Incentives
Statistics |
"Green"
Label/Public
Recognition
Enforcement
Discretion/
Regulatory Relief
Tech. Assistance
Oregon
Participation
Climate Leaders
~
2
Energy STAR/Climate Wise/Green Lights
~
109
National Environmental Performance Track
~
~
~
4
Waste Wi$e
~
~
10
Portland State University, University
of Illinois and Oregon State
University
27
-------
Progress ^Environmental
Program Review
Selected International Programs
Incentives
Statistics
"Green"
Label/Public
Recognition
Enforcement
Discretion/
Regulatory Relief
Tech. Assistance
Oregon
Participation
CERES Endorser Program
~
2
Forest Stewardship Council Certification
~
62
ISO 14001 Certification
~
NA
Responsible Care/RC 14001 Certification
~
3
Portland State University, University
of Illinois and Oregon State
University
Progress gEnv iron mental
Program Review
Selected National/State/Local Programs
Incentives
Statistics |
"Green"
Label/Public
Recognition
Enforcement
Discretion/
Regulatory Relief
Tech. Assistance
Oregon
Participation
Food Alliance (National)
~
32
LEED Green Building Certification (National)
~
"7
73
Eco-Logical Business Program for Auto Shops (Oregon)
~
~
45
The Oregon Natural Step Network (Oregon)
~
~
127
Eco-Logical Business Program for Landscapers (Regional)
~
~
NA
G/Rated Green Building Program (Regional)
~
50
Portland State University, University
of Illinois and Oregon State
University
28
-------
Corporate Environmental Behavior and the
Effectiveness of Government Interventions
PROCEEDINGS OF
LUNCH PANEL DISCUSSION
A WORKSHOP SPONSORED BY THE U.S. ENVIRONMENTAL PROTECTION
AGENCY'S NATIONAL CENTER FOR ENVIRONMENTAL ECONOMICS (NCEE),
NATIONAL CENTER FOR ENVIRONMENTAL RESEARCH (NCER)
April 26-27, 2004
Wyndham Washington Hotel
Washington, DC
Prepared by Alpha-Gamma Technologies, Inc.
4700 Falls of Neuse Road, Suite 350, Raleigh, NC 27609
i
-------
ACKNOWLEDGEMENTS
This report has been prepared by Alpha-Gamma Technologies, Inc. with funding from
the National Center for Environmental Economics (NCEE). Alpha-Gamma wishes to
thank NCEE's Cynthia Morgan and Ann Wolverton and the Project Officer, Ronald
Wiley, for their guidance and assistance throughout this project.
DISCLAIMER
These proceedings are being distributed in the interest of increasing public understanding
and knowledge of the issues discussed at the workshop and have been prepared
independently of the workshop. Although the proceedings have been funded in part by
the United States Environmental Protection Agency under Contract No. 68-W-01-055 to
Alpha-Gamma Technologies, Inc., the contents of this document may not necessarily
reflect the views of the Agency and no official endorsement should be inferred.
11
-------
TABLE OF CONTENTS
Lunch Panel Discussion: Progress Towards an Environmental Facility Research
Database
Panelists: Dietrich Earnhardt, University of Kansas; Richard Andrews,
University of North Carolina at Chapel Hill; Mike Barrette, OECA; Pat
Garvey, Office of Environmental Innovation 1
111
-------
Transcript of Tuesday's Lunch Panel/Discussion
Matt Clark (U.S. EPA, Office of Research and Development)
I'll give you a very brief background. As you can see, we have a fairly significant
number of researchers who are using facility-specific data from EPA, whether it's TRI or
what we call PCS, which is our water pollution system, or some of the enforcement stuff,
through ECHO and IDEA, or the air system, or the biennial reporting system for people
who are working with hazardous wastes—there's a whole raft of these things that we
want to try to align because every time I have a researcher who is trying to pull these
things together, it's taking $100,000 off a grant. So, I don't want to keep spending that
$100,000 over and over again, and what we're winding up with is—I mean, it would be
nice if we could just bring all this stuff in, but you can't, because the researchers are
looking at a specific industry, or like Dave Ervin, at a state, and things like that. We're
hoping that there's some economy of scale here, some efficiency to be gained, by trying
to address sort of holistically what the difficulties are associated with databases and to
resolve some of those difficulties.
We had sort of an open meeting on September 3rd, I believe, of last year, and I'm just
going to recap it. We had some of the same people who are up here talking, both the
EPA database managers, who are describing their plans—a great deal of improvement in
EPA databases has been made over the last few years—and some of the researchers
expressing some of the difficulties that they've had in trying to pull stuff together. In a
lot of ways, it works both ways, because the researchers need to know a little bit more
than just what the data are—they have to know what the data mean and how they were
collected.
So, just to recap it, there's a lot of interest in using EPA facilities-specific databases for
research; the researchers are cumulatively expending a lot of time and resources matching
and cleaning data. There are difficulties in combining information from different
databases, most particularly for panel data sets, which is what people are most interested
in using. And, there's difficulty on the researchers' end. To people who work in the field
at EPA it becomes readily apparent that the researchers don't always understand the
data's purpose—how and why it was collected, its limitations, and what it actually
means. There has been substantial progress in creating single identifiers for facilities and
linkages in the Facility Reporting System (FRS) and Envirofacts, which is another sort of
a window to all our data world. And, there were problems in linking with Census that
Randy Becker talked about, and Randy is from Census.
What we're trying to do today is sort of move on. We've been talking, and we seem to
have had, perhaps not agreement, but a lot of nods in the audience in the September 3rd
meeting that it might be a good idea to develop a combined database, a facility research
database, and our thought was that we would use the LRD (longitudinal research
database) model and develop some sort of sample of the roughly 1.8 million facilities or
"things that are regulated out there" in the EPA data sets. When you start talking about
panel data, that is just too much to deal with, so we have to figure out a way to limit it
and make the kind of matches and corrections that we need. The idea is to match the
1
-------
data, on a facility basis, from a variety of databases, including enforcement and all the
others I mentioned, and some I didn't mention. It would be nice if we could link it to
Census, particularly the longitudinal research database, in some fashion. There are
problems with the IRS, and privacy rights, and things like that, so we have to try to figure
out how to address those, but also, somehow, to have ways to identify what kind of
company it is and link to Dun and Bradstreet.
Clearly, we want to build a panel data set, which is going to be very hard. We're going to
need things like dedicated servers—it's a lot of information and a lot of different
facilities. I don't know how big it has to be, . . . and one of the things that we discussed
in the meeting is that this is something that would have to be regularly updated—it has to
be a living thing that we have to keep up-to-date over time—on a quarterly, monthly,
whatever the right basis is—to keep it going. Ideally, it would be easily aggregated to
state, industry sector, etc.
Now, we really have some questions here—it seems to be the way to go, but we don't
know. We really want our panelists—and we'll give you a few minutes each—to talk
about this. Is this a desirable thing to do—certainly from the researchers' perspective,
but also from the database managers'? Is it something that would be useful internally?
Secondly, is it feasible? And at what cost—half a million dollars—$1 million—$5
million—$10 million? What would it take to make this happen? What are the desirable
components? Start talking about sampling protocol—do you want just a random sample
of all the industries out there? Do we want to over-sample some of the bigger industries,
the more highly regulated industries, and have a random sample of dry cleaners, auto part
shops, and things like that? How do we stratify the samples, if that's the choice? What
should be the elements in the sampling protocol? How would we stratify them—based
on the NAIC codes—the size of the firm—the location—the state? Over sampling—
what do we want to do, census the big facilities and polluting industries? What kind of
documentation [do we want]? One of the things that the people at EPA and the
researchers both pointed out is that they need some documentation to explain what the
data mean and what the purpose was. Finally, what kind of accessibility can be granted?
-and any other ideas.
Okay, so I'll just move on down the line of panelists here and get their input.
Pat Garvey (U.S. EPA, Office of Environmental Information)
I'm the National Program Manager for the Facility Registry database that Matt referenced
that includes about 1.8 million unique places regulated or monitored by either the Federal
government or state government. It has about 2.2 million IDs associated with that—so
those large companies, like Alcoa, that might come under the Toxic Release Inventory
might have hazardous wastes, might have a discharge pipe, might have an air stack—
those are all IDs, programmatic areas that we would try to link together at a given place.
The whole research community has raised five related issues, and I'll review them
quickly here. One is the issue of definition of terms. At EPA we have created this large
database, and for lack of a better word, we've just used the word "facility." But, the
2
-------
database also has monitoring stations—the air quality subsystem monitoring stations (and
it has the ability to have water monitoring stations)—brownfields properties, Superfund
sites, all these things that are technically not facilities (i.e., there's not a company, an
owner, or an operator there). But, the definition of terms, I think, skews the aspect of a
researcher coming into a very large data set and saying, "None of these things meet the
criteria that I want."
I know researchers have a strong desire for a historical perspective, but as Matt just said,
the database is a living database—it needs to be refreshed. Dun and Bradstreet tells us
that about 11 percent of all companies in the United States go into business or out of
business per year. As soon as you go 3 or 4 years into a historical perspective, you've
got potentially 44 percent of your universe either not part of the initial sample or not in
the final sample because of that historical perspective.
We do have data gaps in key fields, and we don 7 have key fields defined as you
[researchers] would want, such as a primary SIC code for a North American Industrial
Code. We in the Facility Registry System collect all SIC codes ever reported by that
company, because, as I'm sure all companies do their absolute best job in defining
themselves because they have to report to TRI and other regulatory programs and have a
high community right to know, they might be creative in the designation of their SIC
code or a change of SIC code process, which would then throw them out of, maybe,
historical perspective or trend analysis. The Agency has never dealt with trying to
establish a definition of "primary SIC code," so many of our databases collect all the SIC
codes, and thus you've got a big melting pot.
You've also got the other key fields that might not be populated because of burden
reduction issues, so that the Office of Management and Budget or the Agency or the
Administration has decided that, "Gee, for this individual data area, be it small-quantity
generators for hazardous wastes or something else, we don't want to collect all the
different kinds of data elements that complete a large database like the FRS."
Then, we also have the last issue—many of our smaller, targeted programs, such as
Green Light or Energy Star and those kinds of things, deal more at the company level and
not specifically at the facility level. So, if in your research problem you want to deal with
small-quantity generators, which obviously constitute a large number, we've got
databases that satisfy that need, but if you're looking at some of the boutique kinds of
programs, such as Energy Star of brownfields or P2—you know them all better than I—
their definition of "facility" or "site" versus "company" or "corporation" doesn't match
up cleanly.
So, I think these are all very strong challenges that the research community has—Good
luck\
Matt Clark
Randy will be talking about the length between Census and all the problems dealing with
that. He's familiar also with being a researcher, not just being a bureaucrat.
3
-------
Randy Becker (U.S. Bureau of the Census)
One of the reasons why you'd want to link EPA data to Census data is that we can fill in
some of the missing information about plants, such as their size, which isn't generally
collected in the EPA databases—that's the critical one. But, you can also fill in the
information about industrial detail—we have our own industrial code—as well as
production information, so you can look at how productive they are, how many
production workers they employ, their input usage, their capital investment—and that's
basic information that's collected annually on the Annual Survey of Manufacturers and
the Census of Manufacturers. Plus, there are other databases as well that can be matched
into those databases that I just mentioned, and most commonly we're talking about the
Pollution Abatement Cost and Expenditures Survey, which was collected for a number of
years and was collected again in 1999. It is now being redeveloped—we hope to have
another one in the field in a couple of years.
So, there is interest in bringing the regulatory data to the census data. A couple of earlier
speakers talked about the challenge of the census process, and part of that is that we get
our information on what businesses are out there—we call it our business register—from
the IRS. You may recall a number of years ago the IRS was raked over the coals by
Congress because of its employees browsing the data, so they [Congress] took it out on
the IRS and they took it out on agencies that the IRS provides data to, and they wanted to
know who's looking at the data provided to the agencies. The Census Bureau is a very
large user of IRS data, ... so the IRS makes the claim that because they provide the
underlying frame information, and some of that data makes its way into the final census
data—that is, it's co-mingled—they think that they have some say on who looks at it and
for what purposes, and so forth. So, this is, as of 1999-2000, very subject to a lot of
regulatory oversight. Now, most of the Census Bureau is immune from that since there is
legislation in place to actually use IRS data to draw samples . . . and so forth. Using the
data to do research and handing it over to egghead academics doesn't necessarily sit that
well with the IRS. (Of course I don't speak for the Census Bureau when I say all these
things.)
But, essentially, that's the conflict that we're faced with, and we are now through the
woods, as it were. There aren't much more reporting requirements and more processes in
place to get things approved, as Dick Morgenstern was talking about earlier. But, there is
a process in place and we're hoping that it's speeding along. So, there is a fixed cost to
using Census data, but there is a great return, as well.
So, I think what we envision, and we talked about this at the September 3rd meeting, is
that once EPA has their data together and all the facilities in a common identifier, one of
the things that could be done is some sort of name-and-address matching with the Census
data. Essentially, there would be a "zipper" file—that is, the Census ID and the EPA
ID—and because our data is confidential (and I think some of the EPA's might be as
well), but because our data is confidential, that zipper would reside at Census. However,
any researchers who would want to link any EPA data could tap into that zipper that's
been constructed and bring in whatever EPA data there are.
4
-------
Some of the issues that Matt was talking about—we have 1.8 million on the EPA site and
we have many more on the Census site. You know, if we're talking about sampling and
restricting scope, . . . why not just do it all? As soon as you start restricting the scope,
there's going to be someone out there who wants to look at something that's outside of
that scope. The only real constraint is just computing power—and brain damage, to some
extent—but I think it is certainly possible—the Census has experience with linking many,
many more records and doing name and address matching. There's sophisticated
statistical software out there to do name and address matching and so forth, so I'd say
let's think big and if it doesn't work, we can always contract our motivation at that point,
but let's not start out that way.
Ron Shadbegian (University of Massachusetts, Dartmouth)
I do have a lot of experience using the Census data and linking in EPA data to that, and as
a number of people have already said, it's not an easy process to go through. In the old
days we didn't have these wonderful computer matching programs, so we did a lot of
linking by hand. The other problem is that name and address matching is fine, but not all
facilities give their particular address—they give you the address of their headquarters or
some post office box—so, there's a lot of hand wringing, and you look at those with the
same zip code . . . there are a lot of tricks to sort of looking and trying to identify the
matches that you don't get through the computer programs, so it's not an easy thing,
necessarily, to do, but as a first crack, maybe that is what we would do.
We've also linked in data from other sources like the Lockwood Directory, which is a
paper industry directory that gives information such as the size of the firms in terms of
their capacity, whether they use pulping or not—do they by pulp?—do they make their
own pulp?—what techniques do they use? So, we've used that sort of data in the past as
well, and weaved that in with census data, and we've used Pollution Abatement Cost and
Expenditure Survey to collect information on how much these plants spend on pollution
abatement capital, on the air side, on the water side—how much they spend on operating
costs on the air side, water side, solid waste side.
Wayne Gray and I have used these data sets that we've put together, and continually put
together over time, to answer lots of interesting questions. We started off looking at the
effect of EPA regulation on productivity, and our EPA regulatory measure was the
pollution abatement expenditures—and so that quickly became: well, do pollution
abatement expenditures overstate actual expenditures on pollution abatement or do they
understate actual expenditures on pollution abatement? ... So, we've looked at things
like that. We've looked at the effect of regulation on pollution abatement investment, on
production investment, and on compliance, as Wayne talked about yesterday.
So, there are lots of interesting things that you can do once you make that link with
Census data, so I agree with Randy here in saying that we should go for the whole thing,
and if we have to start paring it down, then we can do that as well.
5
-------
Michael Barrette (U.S. EPA, Office of Enforcement and Compliance Assurance)
Thanks for inviting me today—I'll just give a few remarks about facility data. Let me
provide just a little bit of background: First of all, I'm actually involved in two pieces of
the puzzle, which is actually using the data as somebody that does targeting and analysis
with our Regional offices in the enforcement program, and in system design and web
help tools, including our integrated data for enforcement analysis (IDEA) system, which
is basically a mainframe system that's extremely powerful, but only a handful of people
really know how to use it, because it's so complicated. And then two spin-off projects
from that—one is what we call OTIS, which is our Online Tracking Information System,
an internal web-based tool that basically gets the IDEA data out using a web browser, and
then ECHO, which launched about a year or so ago, which provides data . . . and we
integrate information on RCRA, the Clean Air Act, the Clean Water Act, TRI,
census/demographic data, things of that nature.
So, we're very familiar with the Facility Registry System—we've been working with
Pat—we've worked with him basically to put together an error tracking system, so if
somebody is using the databases and they see an error, they can submit that and it goes
through a fairly elaborate process of getting it to the right data steward. The thing to keep
in mind when using the facility data is there are, as Pat mentioned, somewhere in the
range of 2 million records. Now, those records come from EPA in many strange ways—
we have different reporting cycles; we have different regulations; we have some
programs where you have to notify when you shut down and others where you don't have
to notify. We have some states that are maintaining their own database and then they
decide to just send us uploads, say, once a month, so they don't really view that data set
as what they use to manage—they may pay less attention to it. We have other states that
are entering [data] indirectly. We also have some systems that are modernized, some that
we call legacy, and some that are kind of in between, being modernized now. The
bottom line is that facility data that winds up at EPA come in many shapes and forms,
and basically Pat's program has to figure out what matches up out of all those things that
come in. Over the last 5 or 6 years we've seen a big improvement in the ability to match
those things, and we've also added a data steward network on top of the computer
programs. But, still, if you really want to work with a very tight data set and ensure there
are no mistakes, you're never going to get a hundred percent with that system because
things are constantly changing.
At headquarters we don't, obviously, know what's going on in the field, and so—I don't
know if I'm stating this the way Pat would, but—what we have with FRS now is pretty
much what you're going to have in the future with the exception that maybe more
programs will be added. I think his budget's pretty much locked in—he's got a staff
that's able to do data quality, to respond to errors, to respond to people who might send in
batch files and ask them to clean them up and they always clean them up very quickly—
but there's a lot of information out there that people maybe have not scrubbed down or
looked at, and there's always going to be some type of mistakes. That doesn't mean that
6
-------
you should discount the system—you just have to keep that in mind when you're using it.
One example is the fact that we have somewhere around a million facilities in the RCRA
database. Well, there's no flag in RCRA that really tells you that they're not in existence
anymore, so that's something we're working on—we're trying to get that in place by
December 2005—to have an "inactive" flag in that database. But, you have to keep in
mind that if you're using that database, as many as 30-, 40- 50-percent may be companies
that are not in existence.
The other piece is that there has been some progress internally to start looking at
corporations and how to match those together, but at this point in time EPA contracts
with Dun and Bradstreet and it's not clear exactly what data will somebody outside the
Agency use? That's kind of an ongoing thing, but it's something the enforcement office
is very interested in—figuring out a good way to profile a company, because every week
we get requests, whether from the Administrator or a voluntary program or the White
House—somebody needs to know about a company—and there's no way that you could
just press a button and say, "Okay, these are the ones that had enforcement actions, this is
their TRI release, and this is how many had open violations." You'd have to do a lot of
custom work to get that information. Hopefully, if we make some progress with Dun and
Bradstreet, we'll be able to automate some of that.
The issue of developing a small subset, or a panel, of data that can be used by
researchers—we actually have some experience in that because in 1995 the White House
Reinvention Committee basically required EPA to publish data for five industries on the
web, which we ended up doing in 1998 under the Sector Facility Indexing Project. Let
me relate a couple of experiences with that: First of all, if you tap into the EPA systems
and ask how many refineries there are, which is SIC 2911, when we ran that back in 1995
it was something like 850. That doesn't necessarily mean that the linkage data, or FRS,
is wrong, but what it means is there's a lot of extra SIC code data in all these systems that
it's pulling from, and anything that has any relationship to a petroleum refinery,
somebody might have put in a 2911, and then all of a sudden what should be 180
facilities looks like 850. We couldn't put out the list of 850 because it was wrong, and
we ended up having to—basically through brute force and a lot of grunt work—figure out
which facilities were actually refineries, what were the EPA numbers across all the
programs, which ones were operating and not operating, and we used some of the same
sources, like Lockwood Post, the Department of Energy for the refineries, the auto
industry has trade association data, and we actually got that nailed down and published
the data in 1998.
But, it's a pretty intensive effort to make sure that that information stays correct, and we
have to have contractor support every year to look at that information. There are all sorts
of questions on definition—if a refinery is totally located with a chemical plant and you
want to look at emission trends or you want to look at the release amounts versus the
production, they may have reported both their chemical and their petroleum together, and
how do you piece that apart? So, it's not really that easy to do—it's a very intensive
effort that, I guess I would say, unless there's a dedicated staff that is spending money
and doing that every year and making sure that there's data surveillance, it's very hard to
7
-------
accomplish and I think that, as some of the other folks have said, you're going to get a lot
of people coming in that want to do other studies.
So, I don't know what the solution is, but one of the things (since I come from the
enforcement compliance program) that we've tried to do is to focus in on a subset, which
we call the "majors." The "minors" in most of the programs have fewer reporting
requirements—the states don't have to tell us all that information. So, unless you're
looking at compliance assistance to small facilities or something, we think that the data
on the majors, which would be Clean Air Title V permitees, water major dischargers,
RCRA large-quantity generators—we spend a lot of time scrubbing that data. The
reporting requirements are more frequent, so we know the facility data are better. Once
you start entering into the world of Clean Water Act minors or things like that, you never
know what you're going to get, and those of you that have used the data probably know
that. So, that's one way that you could develop a subset, but I don't know if that's
necessarily the best way.
The only other thing I'll say is that for those of you that are interested in Clean Water Act
data, we are developing a PCS modernization project. Our hope at this point, and I think
we have a pretty solid commitment from the states, is that when that modernization is
completed, we will continue to get all the majors' discharge reports, from which we'll
calculate a compliance determination. But, we're hoping that we're going to get all the
minors in as well, and the reason why we think we can do that is because we're moving
toward electronic reporting of that data. The biggest barrier to getting that data in is state
staff having to keypunch every little thing every month. So, if we can get electronic
reporting of those DMRs right from the facility into a central receiving database, then
there's a good chance that we might go from the current situation of about 7,000 facilities
with good data up to close to 100,000 in the water program. It's not going to happen
overnight, but hopefully we'll get there soon.
Matt Clark
Pete—any suggestions on this—what you would like to see?
Pete Andrews (University of North Carolina at Chapel Hill)
I'm the first of the academic eggheads up here, so I'll try to be as hard-boiled as
everybody else.
As a researcher trying to use these data, let me first comment on some of the things that
have been really helpful. I want to really thank the gentleman at my left here, Mike
Barrette, who has been tremendously helpful with our project, both in getting the data we
needed and learning to use it and so forth, and working through the portion of our grant
that it took to get these data into usable shape for what we were trying to do with them.
EPA has been working on making all the data more user friendly, and this is a great help.
Simple things, like single identifiers, keeping the address and contact information up to
date. This comment that was made about which ones have gone out of business—one of
8
-------
the greatest unexpected learning experiences we've dealt with just in the process of our
project is discovering how much of our own sample turns over just in the period of the
project. So, that is very helpful.
The other thing that I think would be extremely helpful would be just relatively simple
things, like more-user-friendly guides to the data sources—what were they collected
for?—so that on the web, if you're using this for the first time, they're just easy buttons
to click for so you understand how these data were collected, for what purposes, what
their updating cycles are, if that's something that can be put up there. That will save
some costs at the user end; it will certainly save a lot of costs at the EPA end of hand
holding all of us individually on those kinds of questions which we've had to ask in the
past. Those things are really helpful.
I'm more of a skeptic also—I wasn't expecting, after your comment about the nodding
heads in 2003 that everybody was going to sound skeptical of the panel data project, but I
am a skeptic of it. I think it could be a very useful thing to do for some purposes, but I
don't think it would solve the problem you're trying to solve with it. There are obviously
some areas of research that could come from it and that could be strengthened by it, but
one of the basic problems I think it would run into is "bucket size" for doing statistical
analysis—how many facilities you've got in each cell. There are so many different kinds
of research that cry out to be done in this area, and when we start figuring how many
facilities there are in a given sector, in a given state, of a given size, I worry about the
bias of a panel like that towards: once you've created the panel, you want researchers to
use it, rather than them looking at some of the new and emerging sectors. You know, we
ought to have more people looking at agriculture today, but this is largely a
manufacturing facilities and utilities and POTWs database. So, I worry about the bias—I
worry about your putting so much money into building and maintaining and feeding a
database like this that it would be less productive of the kind of research that, even with
the rough kinds of issues we're having to deal with, could be improved by putting the
money into some things that you're already starting to do and really helping us a lot by
doing. And then, let researchers keep having their own head about the kind of questions
that look important and how we work with that.
Dietrich Earnhart (University of Kansas)
I will echo most of these points. Going back to the presentation I made remotely
September 3rd of last year, what I would want would be a nice, clean, unique identifier
across all the databases. If that is already in place, as I understand what Mr. Garvey said,
then I have what I want, in general. If 99 percent of the time it's a nice clean identifier,
then I'm done—you don't need to integrate anything more for me—I'll gladly run
through 1.8 million observations to match up what I need. Now, if that's the case, that's
great, and I think it would be helpful to advertise that more, because actually it wasn't
until I talked to Sarah Stafford in January of this year that I realized that that had been
done.
9
-------
I, like Ron, have hired research assistants who painstakingly tried to match facility name,
facility zip code, etc., so I don't want to go back to that. I would echo the point that Pete
just brought up in terms of going back to natural underlying databases—it would be
helpful to have more accurate information on contact information. In our survey we
distributed to every single last chemical manufacturing facility across the entire U.S., and
we found that 35 percent of the data that EPA listed was wrong, way wrong, not even
close to being right. We searched numerous databases to try to find the people who
supposedly were active discharging facilities—we could not find them after months and
months—we called state regulators—we called the EPA Regional offices—they were no
where to be found. In addition to that, the active status was way wrong, even for some of
the major polluters. So, in some ways, possibly, it would be helpful to clean up the data
before you actually integrate it. (Sorry about that.)
Documentation would be really helpful. I will add that the people at EPA have been very
helpful—I'll put my plug in for Steven Rubin, a fantastic man who has provided me with
the Permit and Compliance System database on an ongoing basis. It would be nice if
there actually might be a workshop of this sort to teach us, or teach us in combination
with our research assistants—I know Madhu Khanna has an army of RAs working with
her. If you could come and say this is how we use databases, especially if we had a
database that is really powerful but a bugger to use, then I know I'll never touch it.
All of these things would be helpful as part of this integration/modernization process.
One minor point about the Dun and Bradstreet: I'm wondering how or why that was
chosen as the way of connecting things—most people who are looking for financially
related data use the Compustat Research Insight Database, which I know definitely does
not have the Dun and Bradstreet number. So, it would be really nice to find somebody
that is not proprietary. Once again, maybe there's some great logic behind how it was
chosen, and maybe there's a different database that people can match to—I know
Michael Lennox made a reference to that—so maybe we all can learn from his previous
research.
Michael Lennox (Duke University)
As a doctoral student, I actually did that process of matching facilities to Dun and
Bradstreet data, then matching to Compustat—the Compustat data, though, doesn't give
you the structural trees of ownership, and Dun and Bradstreet does—and Dun and
Bradstreet has facility-level data, which Compustat does not, so that's why you have to
use Dun and Bradstreet.
Michael (Barrette), to make your job even more difficult, I just want to throw out another
recommendation: It would be great to have these over time—affiliations-
longitudinally—which I know is incredibly difficult but something that we've worked
really hard on trying to capture—but it gives you some incredibly powerful statistical
powers there by looking at facilities that have changed ownership and how that might
affect behavioral performance. [Matt Clark interjection: You're speaking temporally
10
-------
rather than geographically.] Yes. So, in other words, as Pete was suggesting, corporate
affiliation and ownership change more rapidly than you would think across these
samples, so you would almost have to have a corporate affiliation per year and have it
reflecting how that changes over time. I know that's a very tall order—we've talked to
Dun and Bradstreet, and they're not thrilled about dealing with past data—they just want
to deal with it here and now. But, that would be incredibly valuable for a researcher.
Pat Garvey
We've looked at that a half dozen times and there's no regulatory statute that EPA can
sort of "hang its hat on" to do that information collection. We've heard that over and
over again—it's just that there's an issue of reporting burden by the regulated community
to a regulatory agency, and there's always a tension there, as you can imagine.
Michael Lennox
I know that the TRI fields—that's what people were saying—are notoriously ill-reported:
The Dun and Bradstreet Number is the wrong one or they get the wrong firm or its
subsidiaries—I know it's incredibly difficult to get.
Matt Clark
I have a question: Is there a business service like Lexus-Nexus? I would think that a lot
of the sales and changes of ownership would be recorded in the Wall Street Journal or
things like that, so that would not be an information collection problem. I was just
wondering: Is there a service out there, that particularly you business professors have
used, that would allow us to go back in time and see what these changes have done?
Pat Garvey
We deal with the number of firms and stuff, and D & B sort of raises its hand highest
among those service agents.
Dinah Koehler (University of Pennsylvania)
Wharton has this WRDS database—I don't know what WRDS stands for, but I'll find
out—and it's a pretty detailed database related to business-type issues. I think the way it
works, and, again, I'm talking sort of "off-line" here, but something along the lines that
you can request a data set to be prepared—it's for a fee—and they then prepare that data
set, and they will probably match across various different individual data sets. So, we
can check up on that and give you guys more information if you're interested as one
potential model. Now, it is run by Wharton and it is for a fee. It's actually a pretty good
business for Wharton.
I do have a question: We are in the process of trying to match the RADP Accident
Database with TRI. What I'm trying to use is the Dun and Bradstreet facility ID. Now,
Mike just said that that's not great with TRI. Can you give us any advice, or is this in the
FRS already so I can go use that—or do we have to go this D & B facility ID path, which
seems to be the only link that we can find between these two databases?
11
-------
Pat Garvey
The Department of Homeland Security does not allow me to provide Risk Management
(Audit) Program, or RMP, linkages to any FRS data. That was specifically locked down
on September 22, 2001, and that's not publicly available. It is available to EPA staff.
Dinah Koehler
So, just to qualify, Wharton is in a cooperative agreement with the EPA to work on it,
given all the confidentiality issues. So, we do have that database and we want to work
with it and try to link it with TRI, so I guess we're going to have to do that on our own.
Pat Garvey
The TRI and RMP were our very first two databases that we linked together in 1999. On
the Dun and Bradstreet side, we just finished a matching with D & B, and they did match,
through an automated process, 65 percent of the 1.8 million facilities to a D & B number.
As Mike said, the percentage of the 80,000-84,000 "majors" that we've designated in the
database even had a higher matching. One could, of course, extrapolate that RMP
facilities normally are pretty large, or a certain threshold probably qualified as the
"majors" category, but it's not available—we're not allowed to provide that information.
Pete Andrews
Let me just add another comment—maybe this is something else that ought to go into
your notes, Matt, as something to work on. Certainly with the Census data, dealing with
the IRS, we've had to deal with confidentiality problems and so forth. Particularly with
the Department of Homeland Security taking the position it's taking in some places, it
may be important to start working on the question of having a validated method for
researcher access to this information. It may not be available on the internet to the
general public, but certainly EPA researchers and others have got to be able to have
access to these data if EPA and the country are going to get the information they need.
Maybe they won't get the individual facility coordinates that a terrorist might find useful,
but they certainly need to be able to use these data to do valid and useful studies.
Michael Barrette
It's been probably four years since I've even really looked at the RMP database, but from
what I remember, the reporting form requires an EPA ID to be provided, and I think it
isn't really specified exactly which ID needs to be provided, but I think we recommended
in the instructions that you would start with the TRI, so I'm not sure in the RMP program
itself what they provide out of their own database. But, if they do provide any of that
data and you can get that one field that has the EPA identifier that was self reported, you
may be able to construct some of that—I'm guessing.
Pat Garvey
I was in conversation with Homeland just three months ago, and they just made another
absolute comment about lockdown on RMP.
Michael Barrette
12
-------
I believe my comment is related more generally, I think—it might be worthwhile thinking
about this in terms of putting this database together, because it's not only researchers who
are going to be concerned about this, but there's, I think, some sort of a partnership
directly with DHS on this problem of information provision. Letting RMP [data out] is
not the only kind of issue—in most of the very important environmental risks that many
researchers are concerned about, whether it's nuclear or chemical facilities, oil refineries,
water utilities, a lot of these are really potential targets from a Homeland Security point
of view. Therefore, it makes sense at the beginning, as you're putting this together, to
maybe find some common ground or groups.
I think, also, maybe broadening it to provide incentives rather than only looking at it as
an obstacle to research—because I really think there could be some connections on the
informational regulation theme, where there have been some potential benefits of broader
public information from the environmental protection point of view. The same might be
the case from a domestic protection point of view. Some of this information could be put
together rather than being at cross purposes. I think otherwise you're really going to run
into very serious problems in other fields when you start to put a lot of detailed
information together, you start running off facilities, specific locations, and things like
that—my guess is that that's when our friends in the anti-terrorism world are, for good
reasons, going to sort of become concerned about it.
Matt Clark
I know from working with DHS that our STAR grant researchers actually do have to get
clearance on some of these things. RFS has also done some research looking at issues
about the risk of having too much information available, so that might be able to be
continued.
Eric Orts (University of Pennsylvania)
A short story: Just three weeks ago, we had a number of letters come into EPA from
local government and county government that said your public website had the words
"water tower, water intake, water processing, water association" and they asked that you
please strike off facilities that have those combinations. There's a tremendous pressure
out there—from governors, mayors, county commissioners—to take a lot of information
down from public access.
Matt Clark
Maybe research access should be different than public access, and having gone through a
screening process, maybe that would work. The Census has a process.
Joel Garner (Joint Centers for Justice Studies, Inc.)
I'm working with Sally Simpson at the University of Maryland on a Justice Department
study using the PCS data. I want to re-emphasize that a big problem is connecting
facilities to firms. Any help we can get on that would be greatly appreciated.
I'd like to suggest a small solution as opposed to a "big guns" solution: I think it would
be useful if all the projects funded under this program were required to place their data
13
-------
set at the end of their project in the public archives. Wayne Gray's particular project—
those data—publicly funded data—should be sent to the University of Michigan or some
other public archive so that others can use that. You can get more research per dollar
spent—other researchers can use it. Researchers don't like giving up "their children" like
this. It's just good science—it's good public policy.
Matt Clark
We require it. There are going to be limits to that, of course, with the Census data.
Joel Garner
And there are other public archives—Michigan also has mechanisms whereby
confidential information is resided there, and it's available only when the original
investigator approves researchers using it. The point is there are other places and
mechanisms to do that, and it's a . . .
Randy Becker
But there are such mechanisms here as well. These data are archived and there is a
process for getting at it.
Joel Garner
If I wanted to come and use the previous data, I could call you and I could use those data?
Randy Becker
Well, you have to submit a proposal to actually use those data.
Joel Garner
Right. I just think that's a very important thing to do with existing projects. The second
thing I would suggest is bringing professional data archivists to the task. There are
people who know the data and work generating it—it's a very different profession, and
they like doing this! And they're good at it. And the staff pays them to do it. They do
good documentation—we don't have to do that work—they can do it for us.
And then the last suggestion is: Maybe working with the states, generate the data
originally, and you talked about doing some of that. If that work could be done
originally, then you don't have to do it in the archives.
Pat Garvey
Matt, I'd be remiss not to tell you that at the facility level there's another large program at
EPA called the exchange network. It's got about $20 million a year behind it that we
give to States and Tribes. We've made arrangements already with 20 states to, on a
regular basis—as Mike was saying, every two weeks to a month—exchange their master
facility record with EPA's master. In that exchange effort, we have a goal of 35 states by
the end of this calendar year, so that exchange network ... is aggressively moving
forward.
Sarah Stafford (College of William and Mary)
14
-------
I heard this comment from a couple people—this is straight to the EPA folks: I find
mistakes all the time and I don't report them. Do you want me to report them, and if so,
how? If I'm trying to do the linkage, and I find a matching that's not there, should I
report that? Should I just use the report function in Envirofacts? Is there a better way—
to do it as a batch? I'm also concerned, because, you know, it's a 95 percent hunch,
because I've done a lot of connections, but I could be wrong too. So, what type of
information do you want from us to help get these databases better? I think probably
everyone in this room has found a number of mistakes—has anyone ever reported them?
I find a lot, every day, so any information you have to help me help you would be great.
Pat Garvey
I'll let Mike answer, but I think we had 17,000 reported error notifications on the first
year of release of ECHO, and Envirofacts gets about 310-325 per month. We don't see a
lot, so anybody who finds especially issues of linkage IDs that are poorly done or wrong
lat-longs or wrong names of facilities because they've gone out of business, please,
please, please report. As Mike said, I run Error Tracker and I run Facility Registry
System, and I'm also the staff director of the exchange network, so I'm in a pretty much
unique position.
Michael Barrette
Just quickly on that, in terms of the practicality of reporting errors, first of all, the web
site makes it very easy one at a time. So, if you're actually just playing around on the
web site—ECHO, Envirofacts, or whatever—and you see an error, hit the button, and if
it's a linkage issue it's going to go to Pat's staff, and they're going to figure it out. What
I found is when I do a detailed targeting project and I'm looking at, let'say, comparing air
releases in one system to TRI air releases or permit data, and I'll look at almost 700
facilities and maybe 200 of them or 100 of them or whatever aren't lining up right. I'll
kind of scrub those down, and I'll send a batch file over to Pat and say, "These things
look suspect—they're reporting TRI releases to water but they don't have a water permit
attached to it, or whatever it is. Normally, he fixes it within 2-3 weeks. If you're dealing
with a large data set—I don't know if Pat's willing to do this—but it's easier to send him
the file than to key punch them [errors] in one at a time to the web site.
Pat Garvey
I'll take it any way—whenever you have a large data set at this level, information and
feedback is the most important thing.
Lori Snyder (Harvard University)
This is another solution—I almost think that this already exists—a list-serve, where
researchers who are using these databases could subscribe, and then we could post when
we discover something—perhaps not every single error we find, but general things that
we find that might be of interest to the broader research community could be shared that
way, because I know a lot of us are in different fields—we don't always go to the same
conferences—we're not always talking to one another—so that might be a way to
facilitate communication.
15
-------
Deanna Matthews (Carnegie Mellon University)
Something that would be helpful—I now know Michael Barrette's name, so I know
where to go to for help with these data sets. Perhaps as a Project Officer, something that
could be done is assigning somebody who does have control within the database area to
each of the new projects, so they have a contact to go to rather than just the link on the
bottom of the data page—so that they can say, "I'm looking for this specific data. Who
should I go to to find help on that?"
Irene Xiarchos (West Virginia University)
One comment I wanted to make just because it's related to the idea that maybe we should
have different access for the research community and for the public. The research
community is very broad, and sometimes it includes students that may not be able to be
in direct communication with data links, so they are pretty much between the public and
the research community, but a lot of research is going to come out from them. I wanted
to come out and say that because I am a graduate student.
Another question I have, a personal question if anybody can answer—I don't know who
could answer, so that's why I'm posing it here—is there any data that you know of on
recycling, but for the industry—for facilities—not municipal?
[There was no response from the panelists or the audience, and Matt Clark closed the
session.]
16
-------
Corporate Environmental Behavior and the
Effectiveness of Government Interventions
PROCEEDINGS OF
SESSION V: ENVIRONMENTAL MANAGEMENT SYSTEMS
A WORKSHOP SPONSORED BY THE U.S. ENVIRONMENTAL PROTECTION
AGENCY'S NATIONAL CENTER FOR ENVIRONMENTAL ECONOMICS (NCEE),
NATIONAL CENTER FOR ENVIRONMENTAL RESEARCH (NCER)
April 26-27, 2004
Wyndham Washington Hotel
Washington, DC
Prepared by Alpha-Gamma Technologies, Inc.
4700 Falls of Neuse Road, Suite 350, Raleigh, NC 27609
i
-------
ACKNOWLEDGEMENTS
This report has been prepared by Alpha-Gamma Technologies, Inc. with funding from
the National Center for Environmental Economics (NCEE). Alpha-Gamma wishes to
thank NCEE's Cynthia Morgan and Ann Wolverton and the Project Officer, Ronald
Wiley, for their guidance and assistance throughout this project.
DISCLAIMER
These proceedings are being distributed in the interest of increasing public understanding
and knowledge of the issues discussed at the workshop and have been prepared
independently of the workshop. Although the proceedings have been funded in part by
the United States Environmental Protection Agency under Contract No. 68-W-01-055 to
Alpha-Gamma Technologies, Inc., the contents of this document may not necessarily
reflect the views of the Agency and no official endorsement should be inferred.
11
-------
TABLE OF CONTENTS
Session V: Environmental Management Systems
Institutional Pressure and Environmental Management Practices: An
Empirical Analysis
Magali Delmas, University of California at Santa Barbara 1
Environmental Management Systems: Informing Organizational
Decisions
Deanne Matthews, Carnegie-Mellon University 47
Formalized Environmental Management Procedures: What Drives
Performance Improvements? Evidence From Four U.S. Industries
Richard Andrews, University of North Carolina 71
Discussant
Chuck Kent, U.S. EPA, OPEI 100
Discussant
Pat Atkins, Alcoa 104
Summary of Q&A Discussion Following Session V 108
in
-------
Institutional Pressure and Environmental Management Practices:
An Empirical Analysis
Magali A. Delmas
University of California, Santa Barbara
and Michael W. Toffel
University of California, Berkeley
May 2004
Draft: Please do not quote or circulate without the consent of the authors
ABSTRACT
Despite burgeoning research on companies' environmental strategies and environmental
management practices, it remains unclear why some firms adopt environmental management
practices beyond regulatory compliance. This paper leverages institutional theory by proposing
that stakeholders—including governments, customers, activists, local communities,
environmental interest groups, and industry associations—impose coercive and normative
pressures on firms. However, the way in which managers perceive and act upon these pressures
at the facility level depends upon facility- and parent company-specific factors, including their
track record of environmental performance, the competitive position of the parent company and
the organizational structure of the facility. Beyond providing a framework of how institutional
pressures influence facility's environmental management practices, we provide preliminary
results based on the empirical analysis of a survey of 3160 environmental managers in the United
States.
Research funded by the US EPA Star Grant program.
1
-------
INTRODUCTION
Why do some firms adopt environmental management practices that go beyond regulatory
compliance? Is the adoption of these practices driven by potential performance outcomes or by
institutional pressures? Some research has analyzed specific factors external to the firm that
drive the adoption of environmental strategies such as regulation and competitive forces
(Aragon-Correa, 1998; Christmann, 2000; Dean & Brown, 1995; Hart, 1995; Nehrt, 1996; Nehrt,
1998; Russo & Fouts, 1997; Sharma & Vredenburg, 1998), and pressure from non-governmental
organizations (Lawrence & Morell, 1995). Other research has looked at the role of the
characteristics of the firm to explain the adoption of "beyond compliance" strategies. This
includes the influence of organizational context and design (Ramus & Steger, 2000; Sharma,
2000; Sharma, Pablo, & Vredenburg, 1999) and organizational learning (Marcus & Nichols,
1999). Other analyses have focused on the individual or managerial level, examining the role of
leadership values (Egri & Herman, 2000), and managerial attitudes (Cordano & Frieze, 2000;
Sharma, 2000; Sharma et al., 1999). While each has provided a piece of the puzzle, there is still a
lack of understanding of the conditions under which these various rationales matter to explain the
adoption of practices beyond regulatory compliance at the facility level. In a rare exception,
Gunningham, Kagan, and Thornton (2003) examined the external and internal pressures that
drive firms to improve their environmental performance beyond regulatory compliance in the
pulp and paper industry. As others recently pointed out, 'our understanding of factors that foster
strong environmental management practices within a firm, particularly with operations at the
facility level, still remains limited' (Klassen, 2001, p. 257). This paper offers a perspective that
not only evaluates the relative influences of external stakeholders exerting institutional pressures
on firms, but also depicts how firm characteristics and organizational structure moderate these
pressures. Beyond providing a framework of how institutional pressures influence facility's
environmental management practices, we provide preliminary results based on the empirical
analysis of a survey of 3160 environmental managers in the United States.
The institutional sociology framework emphasizes the importance of regulatory, normative and
cognitive factors that affect firms' decisions to adopt a specific organizational practice, above
and beyond the practice's technical efficiency. Institutional theory emphasizes legitimation
processes and the tendency for institutionalized organizational structures and procedures to be
2
-------
taken for granted, regardless of their efficiency implications (Hoffman & Ventresca, 2002).
However, the institutional perspective does not address the fundamental issue of business
strategy: why do organizations subject to the same level of institutional pressure pursue different
strategies? Building on the institutional framework, we argue that firms adopt heterogeneous sets
of environmental management practices because they interpret these pressures differently due to
facility and parent company characteristics. In our model, managers of different facilities are
subject to the same level of institutional pressures but they are expected to perceive these
pressures differently due to disparities in their parent companies' organizational structure,
strategic position, and financial and environmental performance. This difference between
'objective' and 'perceived' pressure leads to different calculations and responses. The adoption
of environmental management practices by firms varies therefore not only due to different levels
of institutional pressures but also because of the organizational process that transforms objective
pressures into perceived pressures.
To be tested empirically, this comprehensive framework of the drivers of the adoption of
environmental management practices necessitates an empirical approach that combines both
existing publicly available databases, as well as original data from a survey questionnaire at the
facility level. Publicly available databases can provide information on "objective pressures"
while the survey questionnaire can give information about the perception of pressure and the
actions taken in response. The combination of these sources of information allows the evaluation
of the difference between objective and perceived pressures and the resulting adoption of
environmental management practices.
INSTITUTIONAL THEORY
Institutional theory emphasizes the role of social and cultural pressures imposed on organizations
that influence organizational practices and structures (Scott, 1992). DiMaggio and Powell (1983)
argue that managerial decisions are strongly influenced by three institutional mechanisms—
coercive, mimetic and normative isomorphism—that create and diffuse a common set of values,
norms and rules to produce similar practices and structures across organizations that share a
common organizational field. An organizational field is defined as "those organizations
that...constitute a recognized area of institutional life: key suppliers, resource and product
3
-------
consumers, regulatory agencies, and other organizations that produce similar services or products
(DiMaggio & Powell, 1983: 148).
Jennings & Zandbergen (1995) were amongst the first to apply institutional theory to explain
firms' adoption of environmental management practices. They argue that because coercive
forces—primarily in the form of regulations and regulatory enforcement—have been the main
impetus of environmental management practices, firms throughout each industry have
implemented similar practices. Consistent with most institutional theorists, Jennings &
Zandbergen claim that firms that share the same organizational field are affected in similar ways
by institutional forces that emanate from them. They cite the examples of how the Three Mile
Island crisis undermined the legitimacy of all firms in the US nuclear power industry, and how
the discovery that chlorofluorocarbons (CFCs) depleted stratospheric ozone undermined the
legitimacy of manufacturing and using those products, and quickly led to institutional coercive
forces via the establishment of the Montreal Protocol to phase out the manufacture of CFCs.
Delmas (2002) proposed an institutional perspective to analyze the drivers of the adoption of the
ISO 14001 Environmental Management System (EMS) international standard in Europe and in
the United States. She described how the regulatory, normative and cognitive aspects of the
institutional environment within a specific country affect the costs and potential benefits of ISO
14001 adoption, and therefore explain differences in adoption rates across countries. Other
researchers have explored how companies operating in different organizational fields are subject
to different institutional pressures. As a result, different practices become commonplace. For
example, distinct levels of coercive pressures are exerted upon different industries, which may
lead to different environmental strategies (Milstein, Hart, & York, 2002).
While such studies examine dynamic and cross-industry institutional forces, they avoid the
question more fundamental to strategic management: why do organizations within the same
organizational field pursue different strategies, despite experiencing isomorphic institutional
pressures? In other words, how might institutional forces lead to heterogeneity, rather than
homogeneity, within an industry? Hoffman (2001) argues that while organizations do not simply
react to the pressures dictated by the organizational field, they also do not act completely
autonomously without the influence of external bounds. Institutional and organizational
dynamics are tightly linked. A few researchers have begun to investigate this question
empirically (D'Aunno, Succi, & Alexander, 2000; Levy & Rothenberg, 2002).
4
-------
Levy & Rothenberg (2002) describe several mechanisms by which institutionalism can
encourage heterogeneity. First, they argue that institutional forces are transformed as they
permeate an organization's boundaries because they are filtered and interpreted by managers
according to the firm's unique history and culture. Second, they describe how an institutional
field may contain conflicting institutional pressures that require prioritization by managers.
Third, they describe how multinational and diversified organizations operate within several
institutional fields—both at the societal and organizational levels—which expose them to
different sets of institutionalized practices and norms.
D'Aunno et al. (2000) explore the circumstances under which organizations are more likely to
abandon institutionalized structures or practices in favor of new ones, such as by diversifying
into new services. They find that market forces (proximity to competitors), institutional forces
(poor compliance with government regulations, being a member of a multidivisional firm), and
mimicry of changes observed in other organizational fields each encourage strategic change that
diverges from institutional norms.
We hypothesize that organizational structure, strategic positioning, and performance will affect
how firms perceive institutional pressures and how they decide to respond. Individuals in
organizations focus on different aspects of the firm's external and internal environments,
depending on the cognitive frame through which they view the world (Hoffman, 2001).
Cognitive frames are mental representations individuals use to interpret and make sense of their
world. Frames can come to be collectively held within organizations, especially through the
influence of the organizational leader (Barr, Stimpert, & Huff, 1992; Weick & Roberts, 1993).
Institutional pressures
In this section, we describe a model that links institutional pressures to organizational
characteristics to explain the adoption of environmental management practices at the facility
level. Figure 1 illustrates our model.
***
[Insert Figure 1. about here]
***
5
-------
This figure shows that facility-level managers' perceptions of institutional pressures are a
function of stakeholders' actions but are moderated by the organizational characteristics of the
facility and the parent company as well as the strategic positioning of the parent company. We
describe how these coercive and normative pressures can affect the adoption of environmental
management practices by facilities. We focus on a subset of the institutional actors identified by
Hoffman (2001) who we believe are most likely to directly influence environmental practices at
the facility level: governments, customers, competitors, community and environmental interest
groups, and industry associations. The actors we focus upon are important to consider when
assessing a firm's environmental performance (Lober, 1996).
Government pressure
Perhaps the most obvious stakeholders that influence firms' adoption of environmental practices
are various government bodies. Legislation authorizes agencies to promulgate and enforce
regulations, a form of coercive power. Many researchers have focused on the influence of
enforced legislation and regulations on firms' environmental practices (Carraro, Katsoulacos, &
Xepapadeas, 1996; Delmas, 2002; Majumdar & Marcus, 2001; Rugman & Verbeke, 1998). In
particular, Delmas (2002) found that governments play an important role in firms' decision to
adopt ISO 14001. First, governments can act as a coercive force by sending a clear signal of their
endorsement of ISO 14001 by, for example, enhancing the reputation of adopters. Second,
government can facilitate adoption by reducing information and search costs linked to the
adoption of the standard by providing technical assistance to potential adopters. In this paper, we
refer to political pressure as the level of political support for broader or more stringent
regulations. Regulatory pressure represents the extent to which regulators threaten to or actually
impede a company's operations based on their environmental performance.
Customerand competitive pressures
In addition to government actors, firms may facilitate coercive and mimetic isomorphism. For
example, multinationals are widely recognized as key agents in the diffusion of practices across
national borders by transmitting organizational techniques to subsidiaries and other organizations
in the host country (Arias & Guillen, 1998). Firms may also mimic practices that successful
leading firms have adopted. In addition, firms respond to customer requirements. The customer-
6
-------
supplier relationship is perhaps the primary mechanism through which quality management
standards have diffused (Anderson, Daly, & Johnson, 1999). Several studies have found that
firms that adopted environmental management practices were motivated by customer concerns.
A survey of the largest Canadian firms showed that customer pressure was the second most cited
source of pressure to adopt an environmental management plan, after government pressure
(Henriques & Sadorsky, 1996). Khanna and Anton (2002) found that U.S. companies that sell
final goods adopt more comprehensive EMSs than companies that sell intermediate goods. This
suggests that retail consumers exert more pressure on companies to adopt environmental
management practices than commercial and industrial customers. Christmann and Taylor (2001)
showed that customers in developed countries have influenced companies in China to improve
their environmental compliance and adopt the ISO 14001 EMS standard.
Community and environmental interest group pressures
Local communities can also impose coercive pressure on companies through their vote in local
and national elections, via environmental activism within environmental non-government
organizations (NGOs), and by filing citizen lawsuits. Several studies have found that company
decisions to adopt environmental management practices are influenced by the desire to improve
or maintain relations with their communities. Henriques and Sadorsky (1996) surveyed 700 firms
in 1992. These firms indicated that community group pressure influenced them to adopt an
environmental plan. Florida and Davison (2001) investigated why facilities had adopted EMS's
and instituted pollution prevention programs. They found that the adoption of these programs
was positively correlated with firms' active engagement with community stakeholders (Florida
& Davison, 2001). Another study based on a survey of ISO 14001 certified companies across 15
countries found that one of the strongest motivating factors to pursue certification was the desire
to be a good neighbor (Raines, 2002).
Some communities may be better able than others to encourage facilities to adopt environmental
practices. Communities with larger minority populations, lower incomes and less education have
greater exposure to toxic emissions (Arora & Cason, 1999; Brooks & Sethi, 1997; Khanna &
Vidovic, 2001). Some researchers have begun examining whether socioeconomic community
characteristics are associated with facilities' decisions to adopt environmental management
practices. One study examined facility-level adoption of a United States Environmental
-------
Protection Agency (US EPA) voluntary program, and found that adoption was more likely in
communities with higher median household income (Khanna & Vidovic, 2001).
Greater declines in toxic emissions have been observed among facilities located in communities
with higher voting rates (Hamilton, 1999) and in states with higher membership in environmental
interest groups (Maxwell, Lyon, & Hackett, 2000). Maxwell et al. (2000) assert that higher
environmental interest group membership levels indicate a community's pro-environmental
stance and greater propensity to use these organizations to lobby for more stringent regulation.
As such, the authors conclude that higher membership rates provide a credible threat of increased
regulation, which in turn drives firms to self-regulate.
Many of the firms studied by Lawrence & Morell (1995), especially the larger ones, were
motivated to improve their environmental performance by their concern over 'environmental
organizations that had aggressively publicized firms' lapses in environmental responsibility'
(Lawrence & Morell, 1995, p. 111). There are many examples where companies have amended
their environmental practices in response to environmental group pressures (Baron, 2003). For
instance, after Mitsubishi Corporation was subject to a protracted consumer boycott led by the
Rainforest Action Network (RAN), Mitsubishi announced it would no longer use old-growth
forest products (World Rainforest Movement, 1998).
Industry pressure
Institutional researchers have argued that organizations are more likely to mimic the behavior of
other organizations that are tied to them through networks (Guler, Guillen, & MacPherson,
2002). Several studies have found that industry associations have motivated firms to adopt
environmental management practices. Kollman & Prakash (2002) examined why the United
Kingdom, Germany and the United States have such different rates of EMS certification. They
found that the decision of whether to pursue certification, and which standard to certify against
(ISO 14001 or the European Union's Eco-Audit and Management Scheme) was strongly
influenced by stakeholder pressures from industry associations in addition to regional chambers
of commerce, suppliers and regulators.
Market concentration within an industry may also affect the rate of diffusion of environmental
management practices. If an industry is dominated by a few big players that require their
8
-------
suppliers to adopt particular environmental management practices, this is likely to lead to a
greater diffusion of these practices than if the industry were more fragmented. This is a major
reason why automotive suppliers in the United States have adopted similar quality and
environmental practices.
Interactions
The interaction between these institutional pressures is likely to moderate their individual
influence on company practices. For example, the pressure from environmental groups may
encourage the formulation of more stringent regulations. This, in turn, can induce industry
leaders to encourage laggard firms to adopt environmental practices. Similarly, following its
chemical disaster in Bhopal in 1984, Union Carbide along with other large chemical companies
faced mounting public pressure for more stringent safety and environmental regulations. In
response, the chemical industry developed and promoted a set of environment, health and safety
(EHS) management practices—the Responsible Care program—to chemical industry
associations in Canada and the United States (King & Lenox, 2000; Prakash, 2000).
The moderating effects of firm characteristics
Within the same industry, firms may be subjected to different levels of institutional pressures.
For example, multinational corporations are often held to higher standards for social and
environmental responsibility than national companies because they are subject to the additional
pressure of stakeholders from foreign countries (Zyglidopoulos, 2002). Furthermore, the
visibility of leading firms often subjects them to more pressure. For example, social and
environmental activists have targeted Nike, McDonald's, Starbucks and Home Depot in part
because of their market leadership position (Roberts, 2003; Rowley & Berman, 2000).
Furthermore, firms with historically poor environmental records are often subjected to more
scrutiny by their local communities and regulators. Thus, multinational companies, market
leaders, and firms with poor environmental records may have more to gain by developing
sophisticated mechanisms to anticipate and manage external pressures.
9
-------
PERCEPTION OF PRESSURE
Firm and facility characteristics can affect not only the level of institutional pressure exerted on a
facility, but also how facility managers perceive institutional pressures. This is important
because, even if institutional pressures were exerted at the same level on two facilities, these two
facilities may well perceive and respond differently.
First, institutional pressures are exerted at various levels of a firm. For example, community
pressures are often directly targeted at a particular facility, while shareholder pressures target the
corporate level. Second, organizations channel these institutional pressures to different subunits,
each of which frames these pressures according to their typical functional routines (Hoffman,
2001). For example, legal departments interpret pressures in terms of risk and liability, public
affairs does so in terms of company reputation, environmental affairs in terms of ecosystem
damage and regulatory compliance, and sales departments in terms of potential lost revenues.
Consequently, the pressure is managed according to the cultural frame of the unit that receives it:
either as an issue of regulatory compliance, human resource management, operational efficiency,
risk management, market demand, or social responsibility (Hoffman, 2001). One implication of
this process is that the internal organization of the firm matters because it influences how
institutional pressures are perceived. Facility managers may perceive these external pressures
more intensively (and respond to them accordingly) in firms where they have more open
channels of communications with the immediate receptor of pressures (corporate functional areas
responsib le for finance, law, strategy, communication, and the environment).
Information sources may also play a role in cultural framing. Environmental managers may learn
about management practices from a variety of sources. For example, a facility may learn in an
industry association meeting about a pending boycott of a competitor because of its
environmental performance. The source from which managers get their information on existing
environmental management practices can also influence their decision to adopt environmental
management practices.
A firm's historical environmental performance may also influence both how managers perceive
stakeholder pressures and how they respond to them. Managers in firms whose reputations have
suffered from pollution accidents may be more sensitive to environmental issues than those in
10
-------
other companies (Prakash, 2000). After major accidents, firms may rearrange their organizational
structure to prevent recurrences and to facilitate more rapid responses. Such reorganizations may
also begin actively engaging with those stakeholders from whom the firm expects more scrutiny
(e.g., regulators, environmental activities). These reorganizations may also occur within
competing firms if heightened institutional pressures spill beyond the firm that experienced the
accident. For example, the disclosure of environmental information in the annual reports of oil
companies increased significantly in the years following the Exxon Valdez oil spill (Patten,
1992).
FIRM RESPONSES TO INSTITUTIONAL PRESSURES
Firms can adopt various types of environmental management practices in response to
institutional pressures. These can be based on (1) environmental strategies of conformance that
focus on complying with regulations and adopting standard industry practices, or (2) voluntary
environmental strategies that seek to reduce the environmental impacts of operations beyond
regulatory requirements (Sharma, 2000). Voluntary strategies involve creative problem solving
and collaborative interactions with stakeholders (Sharma & Vredenburg, 1998). For example,
firms adopting voluntary approaches can implement EMS elements by creating an environmental
policy, developing a formal training program, or instigating routine environmental auditing. In
addition, management can choose to have the comprehensiveness of their EMS validated by a
third party by pursuing ISO 14001 certification. Management can also convey the importance of
environmental management by including it as a criterion in employee performance evaluations
(Nelson, 2002).
Companies can also seek to improve relations with regulators and signal a proactive
environmental stance by participating in government or industry sponsored voluntary programs.
Indeed, the US EPA, some industry associations, and several NGOs have recently created
voluntary standards to provide incentives for firms to go beyond minimal regulatory
requirements. For example, the US EPA has developed several voluntary agreements between
governmental agencies and firms to encourage technological innovation or reduce pollution
while providing relief from particular procedural requirements (Delmas & Terlaak, 2001).
Industry programs include Responsible Care and Sustainable Slopes (King & Lenox, 2000;
11
-------
Rivera & de Leon, 2003). NGO programs include the Natural Step and the Global Reporting
Initiative Guidelines (Bradbury & Clair, 1999; Hedberg & von Malmborg, 2003).
Companies can also work directly with customers and suppliers to improve their environmental
performance. Furthermore, they may engage in "systematic communication, consultation and
collaboration with their key stakeholders...(and) host stakeholder forums and establish permanent
stakeholder advisory panels at either the corporate level, the facility level, or to address a specific
issue" (Nelson, 2002, p. 18).
Methodology
Data for this study are derived from two main sources: (i) a survey questionnaire sent to 3160
facilities in the fall of 2003; and (ii) publicly available databases. The survey provided
information about the management practices each facility has adopted (our dependent variable)
as well as the number of environmental staff, the types of environmental auditing conducted, and
perceptions of institutional pressures. Various "objective" institutional pressures as well as firm
and facility level characteristics were obtained from existing databases.
Sample
Our sample focuses on heavily polluting industrial sectors, which we identified based on their
share of toxic chemical emissions reported to the US EPA's Toxic Release Inventory (TRI)
program. The following sectors were selected: electric utilities (SIC 49), electrical/electronics
(SIC 36), petroleum refining (SIC 29), chemical and allied products (SIC 28), automotive (SIC
37), machinery manufacturing (SIC 35), primary metals manufacturing (SIC 33), and pulp, paper
and paperboard mills (SIC 26). In 2001, the 11,622 facilities from these industries represent 47%
of the total number of facilities that reported data to TRI and 78% of the total TRI toxic air
emissions reported.1 To ensure we would have access to recent environmental performance data,
we restricted our sample to facilities that reported air emissions to the TRI program in at least 3
years within 1996 - 2000. These facilities must report TRI data annually when they employ 10
or more individuals and manufacture, import, process, or use more than designated minimum
EPA 260-S-03-001. July, www.eoa.gov/tri/tridata/triO 1
12
-------
thresholds (typically 10,000-25,000 pounds) of any of 650 toxic chemicals.2 To ensure the
availability of financial data, we further restricted our sample to facilities owned by publicly
traded companies. These restrictions reduced our sample to 3160 facilities.
Survey questionnaire
Little detailed information about environmental management practices (EMPs) is publicly
available at the facility level. Consequently, we conducted a survey to gather this information as
well as managers' perceptions of the factors that influenced their facility to adopting EMPs. Our
survey also asked about how the facility's environmental management was structured
organizationally. The survey questionnaire instrument is included in the Appendix.
Pre-testing. To ensure that our survey questions were clearly understood and easily answerable
by our respondents, we pre-tested our survey instrument by having a panel of experts complete
the survey. These experts included environmental managers and environmental, health and safety
(EHS) managers from twelve large companies in our sample's industries, a few environmental
management consultants, and several faculty members whose research interests include
environmental management. We then interviewed these individuals to probe their interpretation
of each question and to solicit suggestions to clarify them. This process resulted in refinements to
several survey questions and response anchors.
Survey respondents. The ideal survey respondent must be knowledgeable about the facility's
EMPs and have informed perceptions about the drivers of its environmental management efforts.
As such, we targeted the survey toward the facility's environmental manager or EHS manager.
The Survey Research Center (SRC) at the University of California at Santa Barbara called each
facility to obtain the name of this individual.
Survey administration. The survey was sent to the respondents in three waves. The survey was
sent to the entire sample twice, on October 13 and November 4, 2003. The cover letter that
accompanied the survey provided a unique login identification number to enable respondents to
complete the survey online via a secure website if they preferred that to the enclosed paper
2 US Environmental Protection Agency. 2001. The Emergency Planning and Community Right-to-Know Act: Section 313
Release and Other Waste Management Reporting Requirements. EPA 260/K-01-001. February.
http://www.epa.gov/tri/guide docs/index.htm
-------
version. To encourage responses, the SRC placed calls to 2312 facilities (73% of the sample)
between October 23 and November 12. In addition, postcards were sent in January 2004 to those
who had not yet replied.
Response rate. In total, we received 303 responses by mail and 233 by web for a total of 536. Of
our total sample of 3160, this represents a response rate of 17%, which is considered as an
acceptable response rate for a sample of that size. The response rate among the 2312 facilities
that received a follow-up call from SRC was 20.3%, significantly higher than the 7.2% response
rate of the other 930 facilities. Sample representativeness was tested in three ways. First, we
compared the size of respondent and non-respondent facilities using facility-level employment
data obtained from D&B. Respondent and non-respondent facilities employed, on average, 479
and 422 employees respectively, but this distinction is not statistically significant (p=0.19). We
then examined the response rates across industries and found they were quite similar, ranging
from 13%) (Refining and Electric utilities) to 17%> (Machinery, Electrical/electronics) to 19%>
(Automotive, Primary metals). We also compared the pollution levels of the respondents to the
non-respondents. The two groups' total annual toxic emissions released to air, logged and then
averaged over 2000 and 2001, were statistically indistinguishable (p=.41). In addition, we
compared the two groups' average environmental harm by weighting each chemical by the US
EPA's TRACI scheme and then aggregating annual totals (Toffel & Marshall, 2004) and logging
the result. By this measure, the two groups were also statistically indistinguishable (p=0.80).
This provides further assurance that respondents are representative of the entire sample.
Dependent variable
The dependent variable represents the comprehensiveness of environmental management
practices at the facility level. We proxy this unobserved quality by aggregating the observed
environmental practices adopted by a facility. Our measure includes the extent to which: (1) the
facility adopts and communicates an environmental policy; (2) employees receive environmental
training; (3) employee performance reviews incorporate environmental performance; (4)
procurement decisions incorporate environmental concerns; and (5) the facility participates in
government and industry-initiated voluntary environmental programs. In addition, internal and
external audit frequency and whether ISO 14001 certification were included.
14
-------
There are several methods to aggregate these management practices and create the dependent
variable. One method is simply to sum up the number of practices that each facility has adopted
and the level of implementation of each adopted practice (Khanna & Anton, 2002). This method
implicitly weights each practice equally.
A second method is to run a factor analysis on the original variables within each category (e.g.,
training) and then add the category totals to generate the final dependent variable. The main
applications of factor analytic techniques are: (1) to reduce the number of variables and (2) to
detect structure in the relationships between variables. Klassen used this method in his study of
environmental management practices in the furniture industry (Klassen, 2001).
EMS comprehensiveness via summing variables
Table la describes the methodology to derive each category of environmental management
practice. We focus on the following categories: environmental policy promotion (POLICYD);
audit frequency (AUDITSD); training comprehensiveness (TRAIND); environmental
performance review (REVIEWD); environmental procurement policy (PROCURD);
participation in voluntary programs (VOLPRGD); and ISO certification (ISO D). Each
category is the sum of the variables composing it. For example, the category representing
environmental policy is the sum of four variables, each coded 0 ("no") or 1 ("yes"), constructed
from the following survey questions: Is your environmental policy distributed to employees?
(POLICYD) Is your environmental policy posted on the Internet? (POLICYI) Is your
environmental policy discussed with managers or supervisors? (POLICYM) Is your
environmental policy posted at the facility? (POLICYP) We then normalized each category sum
to a maximum score of 1, and he added these normalized category totals to create the dependent
variable (EMP_SUM).
Comprehensiveness of environmental management practices (EMP_SUM)= POLICY D +
AUDITS D + TRAIN D + REVIEW D + PROCUR D + VOLPRG D+ ISO D
As depicted in Figure 2, this variable has a normal distribution, a condition that facilitates the use
of ordinary least squares (OLS) regression.
***
15
-------
[Insert figure 2. about here]
***
EMS comprehensiveness via factor analysis
A specific category of factor analysis is called Principal Component Analysis (PC A). We used
PCA to investigate whether we could consolidate some of our variables within a specific
category. PCA is an eigenanalysis technique that extracts a set of eigenvectors and their
associated Eigenvalues by a step-wise procedure. The first eigenvector is extracted in a manner
that causes it to account for a maximum amount of variance in the data. After each eigenvector is
extracted a residual data matrix is calculated and the procedure is repeated until there are no
significant eigenvectors left. The variance accounted for by each eigenvector is measured by its
Eigenvalue. The variance is equal to the square of the Eigenvalue. Examination of the
Eigenvalues and their relative magnitudes allows an estimation of the number of significant
'factors' or components in the matrix. We retain only factors with Eigenvalues greater than 1, a
very common criterion (Kaiser, 1960).
Not all of the environmental management practices variables could be subject to PCA due to the
coding of some variables.3 We use PCA with the following categories: environmental training;
employee performance reviews; procurement; audit frequency; and voluntary environmental
program participation. Table lb describes the PCA analyses and results, including the original
variables, the number of factors with Eigenvalues above 1 and their Eigenvalues of these factors,
the percentage of variance explained by the factors, and the name of the new variable(s) created.
***
[Insert Table lb about here]
***
3 For example, our variable assessing the existence and communication of an environmental policy consist of dummy variables
on which it is not possible to run a PCA.
16
-------
We add the categories created with PCA to the variable representing environmental policy
(POLICYD) to create the second measure of the comprehensiveness of environmental
management practices (EMPPCA).
Comprehensiveness of environmental management practices using PCA (EMP PCA) =
POLICY D + AUDITS C + TRAIN C + REVIEW_C + PROCUR_C + VOLPRG C +
ISOC
EMP PCA is normally distributed, as illustrated in Figure 3.
***
[Insert figure 3. about here]
***
The two dependent variables that we created (EMPSUM and EMP PCA) are highly correlated
(0.985) but EMP SUM (541) has more observations than EMP PCA (480) due to the difference
in methodology used to create the variables.
Independent variables
Many stakeholder pressures can be measured through publicly available data sources, though in a
few cases internal company information may be significantly more accurate (e.g., customer
pressure). The perception of stakeholder pressure can also be assessed through a survey
questionnaire addressed to managers (Henriques & Sadorsky, 1996). Relying on both publicly
available databases and a survey enable us to assess differences between "objective pressures"
(measured by the former) and "perceived pressures" (measured by the latter).
Political and regulatory pressure
Political and regulatory pressure is measured in several ways. First, we include Congressional
members' "National Environmental Scorecard" values published annually by the League of
Conservation Voters, a measure that has been widely used for this purpose (Hamilton, 1997;
Kassinis & Vafeas, 2002; Viscusi & Hamilton, 1999; Welch, Mazur, & Bretschneider, 2000).
The average of the League of Conservation Voters' 1996 scores for the state's US Senate and
House delegations to Congress was calculated (LCV96CON). Second, we include the number of
17
-------
state-level environmental policy initiatives (toxic waste, air pollution and recycling programs)
each state has implemented (GINDEX50) (Hall & Kerr, 1991: 142), a measure recently been
used by Welch et al. (2000). Third, we incorporate Renew America's assessment of how
comprehensively each state's policies have addressed 17 environmental domains (e.g., air
pollution, groundwater, soil conservation) (GINDEX17) (Hall & Kerr, 1991: 146). Fourth, we
include a variable representing the state's average inverse of pollution intensity over the years
1995-2001 (PII9501). Each year's figure is the ratio of the state's gross product (million current
US $) to the unweighted sum of the state's Toxic Release Inventory emissions (lbs).
Competitive pressure
Pressure toward mimetic isomorphism exerted on a facility is measured by the extent to which
the facility perceives that its competitors have adopted an EMS (COMPEMS). Survey
respondents could choose from five categories representing percentage ranges from 0-20% to 80-
100%. We also included an additional "don't know" category that was subsequently recoded as
the average response of facilities within in the same industry.
Community and environmental interest group pressure
Community pressure is measured using several indicators, including propensity for collective
action, environmental attitudes and demographics. Because communities with a higher
propensity for collective action are likely to be capable of exerting greater institutional pressure
on local facilities, various proxies for a community's propensity for collective action are
employed. First, community environmental activism is measured using the proportion of the
population within the facility's state that are members of major environmental and conservation
organizations, an approach used in several other studies (Maxwell et al., 2000; Welch et al.,
2000; Wikle, 1995). These data were collected through a survey of 80 main environmental and
conservation NGOs in 2003 (Delmas, 2004). The number of environmental NGO members per
state was normalized by dividing it by thousands of state residents in 2000 (NGOPCS00).
Second, a community's propensity to file lawsuits against facilities based on environmental
issues is estimated based on the proportion of a facility's state's population who are
environmental lawyers (Delmas, 2003). The number of environmental lawyers per state, obtained
18
-------
from the Martindale Law Directory, was normalized by dividing it by each state's population
(LAWYERST).
Third, community demographics may also matter. The fact that communities with lower income,
less education, and greater proportions of minorities are often exposed to more pollution may be
due to facilities' perceiving such communities as possessing less institutional power. Community
demographics data including income, race, education, and population density in the United
States are available from US Census Bureau and have been used in several studies to examine
the influence of communities on organizations' environmental practices (e.g., Arora & Cason,
1999; Hamilton, 1993). We include the following variables from the US Census: median per
capita income (MDINCT); percentage of the population over 25 years that attended college
(EDUCCOL); percentage of population whose race was reported "white" (WHITEPOP);
percentage of housing units that are owner-occupied (OWNED); and percentage of urban
population (URBAN). Each of these measures represents the average US Census values
pertaining to the zip codes within a five-mile radius of each facility.
Perception of pressures
In addition to using objective measures of these stakeholder pressures, we also assess how
managers perceived these pressures. Survey respondents reported the extent to which they
perceived various stakeholders influencing their facility to improve environmental performance.
The list of stakeholders included customers, suppliers, competitors, trade associations, local
community, environmental organizations, regulators/legislators, the media, corporate
management, employees, other facilities within the company, socially responsible investment
funds, and shareholders. Respondents ranked each stakeholder influence on a five-point scale
from "no influence" to a "very strong influence." Many of these variables are highly correlated.
As with environmental management practices, we conducted a principle components factor
analysis with "varimax" rotation to combine these variables into a few factors. Missing
observations were excluded listwise. The underlying variables loaded onto three factors:
COMMERCIAL PRESSURE (pressure from customers, suppliers, competitors, trade
associations), NON-MARKET PRESSURE (pressure from local community, environmental
organizations, regulators, media), and FIRM INTERNAL PRESSURE (pressure from corporate
management, employees, other facilities within the company). The loading of these underlying
-------
variables on the three factors is reported in Table 5. These three factors explained 58% of the
variance. Only two underlying variables (socially responsible investment funds and
shareholders) loaded evenly across factors and had lower coefficients (< .50). We ran additional
PCA without these two variables. The resulting factors of this subsequent analysis were strongly
correlated to the three initial factors (correlations > 0.980 and significant at the 0.01 level).
The moderating effects of firm andfacility characteristics
We identify each facility's main business strategy by asking survey respondents to rate, on a 7-
point scale, the extent to which their company "provides low cost products or services" versus
seeks to "differentiate [their] products" (BIZSTRAT). Facility size is measured through its
employee headcount, in thousands (EMPLHERE), which we obtained from Dun and Bradstreet.
A firm's historical environmental record could be measured using the sum of environmental
compliance violations and resulting penalties accrued over the preceding years at all of its
facilities (Kassinis & Vafeas, 2002; Khanna & Anton, 2002; Russo & Fouts, 1997). We use the
plant's logged toxic releases averaged over 2000 to 2001 as a proxy of the attention that the
media and community are paying to the performance of the facility (UNW0001L).
Results
Tables 2 and 3 provide descriptive statistics. We run two sets of models. Environmental
management practice comprehensiveness, constructed using the two approaches described above
(sum and PCA), were regressed on various measures of institutional pressures as well as facility
characteristics and industry dummy variables.
The first set of models includes the "objective" measures of institutional pressure. Results
are reported in Table 4.
The second set of models includes the "subjective" measures of institutional pressures.
Results are reported in Table 6.
In the first set of models, few "objective" measures of institutional pressure are significant. None
of the regulatory or legislative pressure variables appears to significantly influence the adoption
of environmental management practices. Two variables that measure community activism—
20
-------
median 1999 household income and percentage of white population—are significant and
negative. Their significance confirms prior research that finds correlations between community
demographic indicators and pollution, though the negative relationship identified here is in sharp
contrast to prior findings that poorer communities with more racial minorities are subjected to
more pollution. Our findings suggest that companies located in such communities adopt more
environmental management practices. We also find that competitor pressure, measured by the
proportion of competitors a firm believes has implemented an EMS, is positively and
significantly related to the number of environmental management practices the firm adopts. This
direct relationship suggests that firms may mimic their peer groups in terms of how many
environmental management practices to implement, whether this be a lot or few. The variable
representing pollution spotlight (average unweighed TRI air releases, 2000-01) is also significant
and positive, suggesting that facilities that may attract more media and regulatory attention
because of their relatively high mass of emissions have implemented more environmental
management practices. Finally, more environmental management practices have been adopted by
facilities that are part of an organization with operations across more continents and whose
headquarters is outside the US. All of these findings are robust to two alternative ways we
measure EMS comprehensiveness (EMPSUM or EMPPCA).
In the second set of models, we initially individually included each original "perceived
stakeholder pressure" survey variable and found all of them to be significant and positive except
the variable representing the perceived influence of regulators/legislators (not shown). Next, we
instead included the three-factor solution to the principle component analysis to evaluate the
impact of commercial, non-market, and internal pressures. Perceived commercial and internal
pressures to improve environmental performance are both positive and significant determinants
of adopting environmental management practices. Non-market pressures to improve
environmental performance appear to have no influence on the adoption of environmental
management practices. Consistent with the previous objective pressure models, firms with their
headquarters based in the US adopted significantly fewer environmental management practices,
and firms that had a wider geographic scope and more competitor pressure adopted more
environmental management practices. In contrast to the objective pressure models, neither
community demographic measure is significant here, nor is the variable representing pollution
21
-------
spotlight (average unweighted sum of TRI air releases, 2000-01). All of these findings are robust
to two alternative ways we measure EMS comprehensiveness (EMP_SUM or EMP_PCA).
In view of these results it is interesting to note that 'perceived' pressures have a better
explanatory power than 'objective' pressures to predict the adoption of environmental
management practices at the facility level. Some of these findings may result from the limitation
of the proxy of our 'objective' pressures. It would be interesting to get a better understanding of
the regulatory pressure exerted at the facility level by assessing the number of times regulatory
agencies have actually contacted, visited or fined the facility. In addition, we still have yet to
investigate the relationship between objective and perceived pressures, including when their
levels are aligned and, perhaps more importantly, when their levels substantially differ. These
findings have potentially important policy implications. Indeed understanding how facilities
perceive pressures will help policy makers focus their efforts on the appropriate channels.
Our finding that regulators and other non-market actors apparently do not significantly impact
the adoption of comprehensive environmental management practices suggests that other
stakeholders are more effective in exerting direct influence. Policy makers can take advantage of
these stakeholders to enhance the level of adoption of environmental management practices at
the facility level.
Conclusion
This paper provides a model that describes how stakeholders including regulators, customers,
activists, local communities and industry associations impose institutional pressures on facilities
and their parent companies. We also suggest how a variety of facility- and parent company
factors moderate how managers perceive and act upon these pressures. Moderating factors
include historical environmental performance, the competitive position of the parent company
and the organizational structure of the facility.
Our approach complements institutional theory as it suggests that both institutional pressures and
organizational characteristics influence organizations to adopt environmental management
practices. Firm and facility characteristics are viewed as moderating factors because they are
expected to magnify or diminish the influence of institutional pressures.
22
-------
Our results show that the most important factors explaining the adoption of environmental
management practices are how environmental managers perceive the institutional pressures. In
particular, we find that stakeholder pressures from the private sector influence facilities'
adoption of environmental management practices. We also find that the corporation has a strong
influence on facilities' decisions to adopt environmental management practices. The
characteristics of the facility and the firm also matter. In particular, large facilities, facilities that
are part of firms that operate internationally, and facilities whose headquarters are outside the US
tend to adopt more environmental management practices. We plan to further investigate the
relationship between objective and perceived institutional pressures, and how various
institutional pressures influence the adoption of specific environmental management practices.
23
-------
References
Anderson, S. W., Daly, J. D., & Johnson, M. F. 1999. Why firms seek ISO 9000 certification:
Regulatory compliance or competitive advantage. Production and Operations
Management, 8(1): 28-43.
Aragon-Correa, J. A. 1998. Strategic proactivity and firm approach to the natural environment.
Academy of Management Journal, 41: 556-567.
Arias, M. E. & Guillen, M. F. 1998. The transfer of organizational management techniques. In J.
L. Alvarez (Ed ), The Diffusion and Consumption of Business Knowledge. 110-137.
London: Macmillan.
Arora, S. & Cason, T. N. 1999. Do community characteristics influence environmental
outcomes? Evidence from the toxics release inventory. Southern Economic Journal,
65(4): 691-716.
Baron, D. P. 2003. Private politics. Journal of Economics & Management Strategy, 12(1): 31-
66.
Barr, P. S., Stimpert, J. L., & Huff, A. S. 1992. Cognitive change, strategic action, and
organizational renewal. Strategic Management Journal, 13: 15-36.
Bradbury, H. & Clair, J. A. 1999. Promoting sustainable organizations with Sweden's natural
step. Academy of Management Executive, 13(4): 63-74.
Brooks, N. & Sethi, R. 1997. The distribution of pollution: Community characteristics and
exposure to air toxics. Journal of Environmental Economics and Management, 32(2):
233-250.
Carraro, C., Katsoulacos, Y., & Xepapadeas, A. (Eds.). 1996. Environmental policy and market
structure. Boston: Kluwer Academic Publishers.
Christmann, P. 2000. Effects of'best practices' of environmental management on cost advantage:
the role of complementary assets. Academy of Management Journal, 43: 663-680.
Christmann, P. & Taylor, G. 2001. Globalization and the environment: Determinants of firm
self-regulation in China. Journal of International Business Studies, 32(3): 439-458.
Cordano, M. & Frieze, I. H. 2000. Pollution reduction preferences of US environ-mental
managers: applying Ajzen's theory of planned behavior. Academy of Management
Journal, 43: 627-641.
D'Aunno, T., Succi, M., & Alexander, J. A. 2000. The role of institutional and market forces in
divergent organizational change. Administrative Science Quarterly, 45(4): 679-703.
24
-------
Dean, T. J. & Brown, R. L. 1995. Pollution regulation as a barrier to new firm entry: initial
evidence and implications for future research. Academy of Management Journal, 38:
288-303.
Delmas, M. 2004. Survey of Environmental NGO membership in the United States. Donald Bren
School of Environmental Science and Management Mimeo.
Delmas, M. 2003. In Search of ISO: An institutional perspective on the adoption of international
management standards. Stanford, CA: Stanford Graduate School of Business Working
Paper 1784.
Delmas, M. & Terlaak, A. 2001. A framework for analyzing environmental voluntary
agreements. California Management Review, 43(3): 44-63.
Delmas, M. 2002. The diffusion of environmental management standards in Europe and the
United States: An institutional perspective. Policy Sciences, 35: 91-119.
DiMaggio, P. J. & Powell, W. W. 1983. The iron cage revisited: Institutional isomorphism and
collective rationality in organizational fields. American Sociological Review, 48(2): 147-
160.
Egri, C. & Herman, S. 2000. Leadership in the North American environmental sector: Values,
leadership styles and contexts of environmental leaders and their organizations. Academy
of Management Journal, 43: 571-604.
Florida, R. & Davison, D. 2001. Gaining from green management: Environmental management
systems inside and outside the factory. California Management Review, 43(3): 64.
Guler, I., Guillen, M. F., & MacPherson, J. M. 2002. Global competition, institutions, and the
diffusion of organizational practices: The international spread of the ISO 9000 quality
certificates. Administrative Science Quarterly, 47(3): 507-531.
Gunningham N., Kagan R.A., &Thornton, D. 2003. Shades of green: business, regulation, and
environment. Stanford University Press, Stanford, Calif
Hall, B. & Kerr, M. L. 1991. 1991-1992 green index : a state-by-state guide to the nation's
environmental health. Washington D C.
Hamilton, J. T. 1993. Politics and Social Costs: Estimating the Impact of Collective Action on
Hazardous Waste Facilities. RAND Journal of Economics, 24(1): 101-125.
Hamilton, J. T. 1997. Taxes, torts, and the toxics release inventory: Congressional voting on
instruments to control pollution. Economic Inquiry, 35(4): 745-762.
Hamilton, J. T. 1999. Exercising property rights to pollute: Do cancer risks and politics affect
plant emission reductions? Journal of Risk and Uncertainty, 18(2): 105-124.
25
-------
Hart, S. L. 1995. A natural-resource-based view of the firm. Academy of Management Review.,
20(4): 986-1014.
Hedberg, C.-J. & von Malmborg, F. 2003. The Global Reporting Initiative and corporate
sustainability reporting in Swedish companies. Corporate Social Responsibility and
Environmental Management, 10(3): 153-164.
Henriques, I. & Sadorsky, P. 1996. The determinants of an environmentally responsive firm: An
empirical approach. Journal of Environmental Economics & Management, 30(3): 381-
395.
Hoffman, A. J. 2001. Linking organizational and field-level analyses - The diffusion of corporate
environmental practice. Organization & Environment, 14(2): 133-156.
Hoffman, A. J. & Ventresca, M. J. (Eds ). 2002. Organizations, policy and the natural
environment: hstitutional and strategic perspectives. Stanford: Stanford University
Press.
Jennings, P. D. & Zandbergen, P. A. 1995. Ecologically sustainable organizations: An
institutional approach. Academy of Management Review, 20(4): 1015-1052.
Kaiser, H. F. 1960. The application of electronic computers to factor analysis. Educational and
Psychological Measurement, 20: 141-151.
Kassinis, G. & Vafeas, N. 2002. Corporate boards and outside stakeholders as determinants of
environmental litigation. Strategic Management Journal, 23(5): 399-415.
Khanna, M. & Anton, W. Q. 2002. Corporate environmental management: Regulatory and
market-based pressures. Land Economics, 78(4).
Khanna, N. & Vidovic, M. 2001. Facility participation in voluntary pollution prevention
programs and the role of community characteristics: Evidence from the 33/50 program:
Binghamton University Economics Department Working Paper.
King, A. & Lenox, M. 2000. Industry self-regulation without sanctions: The chemical industry's
responsible care program. Academy of Management Journal, 43(4): 698-716.
Klassen, R. D. 2001. Plant-level environmental management orientation: The influence of
management views and plant characteristics. Production and Operations Management,
10(3): 257-275.
Kollman, K. & Prakash, A. 2002. EMS-based environmental regimes as club goods: Examining
variations in firm-level adoption of ISO 14001 and EMAS in U.K., U.S. and Germany.
Policy Sciences, 35(1): 43-67.
Lawrence, A. T. & Morell, D. 1995. Leading-edge environmental management: Motivation,
opportunity, resources and processes. In D. Collins, & M. Starik (Eds.), Special Research
26
-------
Volume of Research in Corporate Social Performance and Policy, Sustaining the
Natural Environment: Empirical Studies on the Interface Between Nature and
Organizations. 99-126. Greenwich, CT: JAI Press.
Levy, D. L. & Rothenberg, S. 2002. Heterogeneity and change in environmental strategy:
Technological and political responses to climate change in the global automobile
industry. In A. J. Hoffman, & M. J. Ventresca (Eds ), Organizations, Policy and the
Natural Environment: Institutional and Strategic Perspectives. Stanford: Stanford
University Press.
Lober, D. J. 1996. Evaluating the environmental performance of corporations. Journal of
Managerial Issues, 8(2): 184-205.
Majumdar, S. K. & Marcus, A. A. 2001. Rules versus discretion: the productivity consequences
of flexible regulation. Academy of Management Journal, 44(1): 170-179.
Marcus, A. A. & Nichols, M. L. 1999. On the edge: heeding the warnings of unusual events.
Organization Science, 10: 482-499.
Maxwell, J. W., Lyon, T. P., & Hackett, S. C. 2000. Self-regulation and social welfare: The
political economy of corporate environmentalism. The Journal of Law & Economics,
43(2): 583-619.
Milstein, M. B., Hart, S. L., & York, A. S. 2002. Coercion breeds variation: The differential
impact of isomorphic pressures on environmental strategies. In A. J. Hoffman, & M. J.
Ventresca (Eds ), Organizations, Policy and the Natural Environment: Institutional
and Strategic Perspectives. Stanford: Stanford University Press.
Nehrt, C. 1996. Timing and intensity effects of environmental investments. Strategic
Management Journal, 17: 535-547.
Nehrt, C. 1998. Maintainability of first mover advantages when environmental regulations differ
between countries. Academy of Management Review, 23: 77-97.
Nelson, J. 2002. From the margins to the mainstream: Corporate social responsibility in the
global economy. In N. Hajensgard, & A. Wahlberg (Eds.), Campaign report on
European CSR excellence 2002-2003: It simply works better!. 14-19. Copenhagen: The
Copenhagen Centre, CSR Europe and the International Business Leaders' Forum.
Patten, D. M. 1992. Intra-industry environmental disclosures in response to the Alaskan oil spill:
a note on legitimacy theory. Accounting, organization and society, 17(5): 471-475.
Prakash, A. 2000. Responsible Care: An assessment. Business & Society, 39(2): 183-209.
Raines, S. S. 2002. Implementing ISO 14001—An international survey assessing the benefits of
certification. Corporate Environmental Strategy, 9(4): 418-426.
27
-------
Ramus, C. A. & Steger, U. 2000. The roles of supervisory support behaviors and environmental
policy in employee "ecoinitiatives" at leading-edge European companies. Academy of
Management Journal, 43(4): 605-626.
Rivera, J. & de Leon, P. 2003. Voluntary environmental performance of western ski areas: Are
participants of the Sustainable Slopes Program greener? Annual Research Conference
of the Association for Public Policy Analysis and Management, Washington DC.
Roberts, S. 2003. Supply chain specific? Understanding the patchy success of ethical sourcing
initiatives. Journal of Business Ethics, 44(2/3): 159-170.
Rowley, T. & Berman, S. 2000. A brand new brand of corporate social performance. Business
and Society, 39(4): 397-418.
Rugman, A. M. & Verbeke, A. 1998. Corporate strategies and environmental regulations: An
organizing framework. Strategic Management Journal, 19(4): 363-375.
Russo, M. V. & Fouts, P. A. 1997. A resource-based perspective on corporate environmental
performance and profitability. Academy of Management Journal, 40: 534-559.
Scott, W. R. 1992. Organizations: rational, natural, and open systems (3rd ed ). Englewood
Cliffs N.J.: Prentice Hall.
Sharma, S. 2000. Managerial interpretations and organizational context as predictors of corporate
choice of environmental strategy. Academy of Management Journal, 43: 681-697.
Sharma, S., Pablo, A. L., & Vredenburg, H. 1999. Corporate environmental responsiveness
strategies: The importance of issue interpretation and organizational context. Journal of
Applied Behavioral Science, 35(1): 87-108.
Sharma, S. & Vredenburg, H. 1998. Proactive corporate environmental strategy and the
development of competitively valuable organizational capabilities. Strategic
Management Journal, 19(8): 729-753.
Toffel. M. W & Julian D Marshall, J.D. 2004. Comparative analysis of weighting methods used
to evaluate chemical release inventories. Journal of Industrial Ecology 8(2).
Viscusi, W. K. & Hamilton, J. T. 1999. Are Risk Regulators Rational? Evidence from Hazardous
Waste Cleanup Decisions. The American Economic Review, 89(4): 1010-1027.
Weick, K. E. & Roberts, K. H. 1993. Collective mind in organizations: Heedful interrelating on
flight decks. Administrative Science Quarterly, 38(3): 357-381.
Welch, E. W., Mazur, A., & Bretschneider, S. 2000. Voluntary behavior by electric utilities:
Levels of adoption and contribution of the climate challenge program to the reduction of
carbon dioxide. Journal of Policy Analysis and Management, 19(3): 407-425.
28
-------
Wikle, T. 1995. Geographical patterns of membership in US environmental organizations. The
Professional Geographer, 47: 41-48.
World Rainforest Movement. 1998. End of boycott: "Eco-Agreement" between RAN and
Mitsubishi. World Rainforest Movement Bulletin, (9) February.
Zyglidopoulos, S. C. 2002. The social and environmental responsibilities of multinationals:
Evidence from the Brent Spar case. Journal of Business Ethics, 36(1/2): 141-151.
29
-------
Figure 1. A model of institutional pressures moderated by parent company and facility
characteristics
Affect
Institutional fields
Pressure
Organization characteristics
Influence
Parent company
• Shareholders
• Competitors
• Industry associations
• Government
• Consumers
• Activists
Parent company
• Firm size
• Firm competitive position
• Level of internationalisation
• Corporate Environmental Health and Safety
organisation
I
Pressure
Plant's adoption of
environmental
management practices
• Environmental management
system (EMS)
comprehensiveness
• Management of stakeholder
relations
Plant
• Government
• Consumers
• Community
Plant
• Plant size
• Sources of information on environmental mgt
practices (industry associations, regulators,
non-governmental organisations, customers,
suppliers)
• Historical environmental performance
(regulatory compliance, pollution levels
relative to neighbours and competitors
t
Influence
Affect
30
-------
Table la. Construction of environmental management practices categories (sum of variables)
Variable name
Construction
Environmental policy
promotion (POLICY D)
S (POLICYD Env policy distributed to employees (0/1); POLICYI; Env
policy posted on internet (0/1); POLICYM Env policy discussed with
mgr/supervisor (0/1); POLICYP Env policy posted at facility (0/1 ))/4.
Annual audits
(AUDIT S_D)
S (AUDITEXT External audits last 3 yrs; AUDITINT Internal audits last 3
yrs)/3
Training
comprehensiveness
(TRAIN_D)
S (TRAINDES Proportion of engineering/r&d/design dept receive env.
Training; TRAINEHS Proportion of EHS dept receive env. Training;
TRAINMGT Proportion of management receive env. Training;
TRAINMNT Proportion of maintenance dept receive env. Training;
TRAINOPS Proportion of operations dept receive env. Training;
TRAINPUR Proportion of purchasing dept receive env. Training;
TRAINSAL Proportion of sales dept receive env. Training) /number of
departments for which responses were given.
Environmental
performance review
(REVIEW_D)
S (REVDES Env Review of engineering/r&d/design staff; REVEHS Env
Review of EHS staff; REVMGT Review of management include env.
Performance; REVMNT Review of maintenance staff include env.
Performance; REVOPS Review of operations staff include env.
Performance; REVPUR Review of purchasing staff include env.
Performance; REVSAL Review of sales staff include env. Performance)
/number of departments for which responses were given
Environmental
procurement policy
(PROCUR_D)
S (PURPOL Extent to which purchasing uses green policy; PURISO Extent
to which purchasing requests ISO 14001 of suppliers; PURINFO Extent to
which purchasing requests env. info of suppliers)/3
Voluntary programs.
Extent to which the facility
adopts government and
industry voluntary
(VOLPRG_D)
S (GOVVOL Status of implementing govt voluntary programs; INDVOL
Status of implementing industry voluntary programs)/8
ISO certification (ISO D)
S (IS09 ISO 9000 status + Isol4 ISO 14001 status)
31
-------
Table lb. PCA analyses of environmental management practice categories
Category
Variables included
Eigenvalues
>1
Variance
explained
Variable
created
Annual audits
External and internal audit frequency
1.99
99%
GEN
AUDITS C
Training
comprehensiveness
Comprehensiveness of environmental training
across seven departments
3.21
54%
TRAIN_C
Environmental
performance review
Extent to which performance reviews incorporate
environmental management tasks across seven
departments
4.36
62%
REVIEW_D
Environmental
procurement
Extent to which purchasing decisions incorporate
environmental criteria
1.95
65%
PROCUR_C
Participation in
voluntary programs
Participation in government- and industry-
initiated voluntary environmental programs
1.37
69%
VOLPRG_C
ISO certification
Status of adopting ISO 9000 and ISO 14001
1.47
74%
ISO C
32
-------
Figure 2. EMS comprehensiveness based on sum of original variables (EMP SUM)
60
50
40
30
20
10
Std. Dev= 1.06
Mean = 3.99
N = 541.00
0
1.25
2.25
3.25
4.25
5.25
6.25
1.75 2.75 3.75 4.75 5.75
EMS comprehensiveness as sum of 0-1 normalized category sum
Figure 3. EMS comprehensiveness based on PC A of original variables (EMP PC A)
Std. Dev = 4.16
Mean = .2
N = 480.00
EMS: pea each category, then sum results
33
-------
Table 2. Descriptive statistics: objective pressures
Variable Description
Obs Mean SD Min Max 1
3
1 STGRN_C5 PCA (lcv96con gindex50 gindexl7 pii9501 ngopcs03)
2LAWPCS03 Environmental lawyers
3LCV96CON League of Conservation Voters
4GINDEX50 Number of state env policies of 50 by Hall & Kerr
5GINDEX17 Rating of 17 state policies by Renew America
6PII9501 State's average inverse of pollution intensity, 1995-2001
7NGOPCS03 Environmental NGO members per 1000 state residents, 2003
8COMPEMS Proportion of competitors with EMS
9BIZSTRAT Company's main business strategy (low cost vs differentiation)
10MDINCT Median 1999 household income within 5 miles, in thousands
11WHITEPOP Percentage population race within 5 miles
12 EDUC C OL Pet pop 25+ years within 5 miles
13EMPLE1RT Plant employees
14UNW0001L Log of Average unweighted TRI air releases, 2000-01
15 TRC0001L Log of Average TRACI-weighted LRI air releases, 2000-01
16 INTL_D Geographic breadth of operations
17HQUS HQ located in US
18INDSLRY1 Automotive industry
19INDSLRY2 Electrical/Electronics industry
20INDSLRY3 Machinery industry
21INDSLRY4 Primary Metals industiy
22INDSLRY5 Refining industry
23INDSLRY6 Utilities industry
526 -0.02 0.97
533 6.39 2.60
526 43.61 21.05
526 22.17 7.36
-1 2 1.00
2 19 0.45 1.00
0 91 0.75 0.25 1.00
5 38 0.92 0.41 0.57 1.00
526 90.27 24.23 46 134 0.92 0.26 0.67 0.90 1.00
526 0.01 0.01 0 0.1 0.72 0.51 0.40 0.54 0.49 1.00
533 2.91 1.71 0.02 8 0.88 0.52 0.52 0.77 0.74 0.70 1.00
571 2.97 1.15 1 5 0.03 0.04 -0.02 0.06 0.06 -0.01 0.00 1.00
535 4.42 1.87 1 7 0.10 0.00 0.10 0.12 0.16 -0.04 0.06 0.07 1.00
524 41.45 12.37 9 94 0.41 0.25 0.35 0.35 0.40 0.24 0.38 0.06 0.03
524 0.79 0.18 0.2 1 -0.04 -0.10 0.03 -0.05 0.02 -0.17 -0.06 0.00 -0.09
524 0.48 0.12 0.2 0.8 0.23 0.17 0.09 0.17 0.24 0.18 0.29 -0.04 -0.01
421 0.48 0.70 0 6 0.01 -0.13 0.04 0.06 0.10 -0.12 -0.06 -0.01 0.13
501 8.76 3.47 -4 16 -0.08 -0.19 0.01 -0.07 -0.07 -0.12 -0.09 -0.07 -0.06
325 5.00 5.76 -9 19 -0.02 -0.11 0.05 0.01 -0.01 -0.08 -0.06 -0.05 -0.24
575 0.59 0.41 0 1 0.11 -0.02 0.11 0.16 0.16 -0.07 0.04 0.24 0.21
575 0.86 0.34 0 1 0.00 0.00 0.02 -0.02 -0.01 0.02 -0.01 -0.13 0.02
575 0.10 0.29 0 1 -0.09 -0.10 -0.08 -0.07 -0.06 -0.09 -0.09 0.21 0.01
575 0.18 0.39 0 1 0.10 0.04 0.06 0.09 0.10 0.09 0.07 0.12 0.05
575 0.10 0.30 0 1 -0.01 -0.12 -0.02 0.03 0.06 -0.07 -0.05 -0.13 0.11
575 0.14 0.35 0 1 0.12 0.04 0.06 0.14 0.13 0.03 0.11 -0.17 0.08
575 0.24 0.43 0 1 -0.05 0.20 -0.09 -0.04 -0.09 0.05 -0.02 0.16 0.06
575 0.08 0.27 0 1 -0.03 -0.12 0.13 -0.13 -0.11 0.04 -0.03 -0.30 -0.43
Table 2. Descriptive statistics: objective pressures (continued)
Variable Description
10
11
12
13 14
15
16
17 18 19 20 21
10MDINCT Median 1999 household income within 5 miles
1.00
11 WHITEPOP Percentage population within 5 miles
0.16
1.00
12EDUCCOL Pet pop 25+ years within 5 miles
0.65
0.13
1.00
13EMPLHRT Plant employees
0.00
-0.02
0.01
1.00
14UNW0001L Log of Average unweighted TRI air releases, 2000-01
-0.08
-0.11
-0.09
0.11 1.00
15TRC0001L Log of Average TRACI-weighted TRI air releases, 2000-01
-0.08
-0.01
-0.11
-0.11 -0.01
1.00
16INTL D Geographic breadth of operations
0.11
0.01
0.03
0.16 -0.15
-0.03
1.00
17HQUS HQ located in US (dummy)
-0.09
-0.06
-0.03
-0.06 -0.07
0.08
-0.05
1.00
18INDSTRY1 Automotive industry
-0.02
0.12
-0.02
0.22 -0.07
-0.06
0.15
0.02 1.00
19INDSTRY2 Electrical/Electronics industry
0.19
0.10
0.09
0.04 -0.20
-0.06
0.17
-0.16 -0.14 1.00
20INDSTRY3 Machinery industry
-0.01
0.08
0.11
0.21 -0.22
0.02
0.10
0.07 -0.12 -0.13 1.00
21INDSTRY4 Primary Metals industry
-0.07
-0.02
-0.05
-0.10 -0.27
0.24
-0.08
0.10 -0.17 -0.18 -0.16 1.00
22INDSTRY5 Refining industry
-0.01
-0.21
0.00
-0.17 0.21
-0.32
0.03
-0.03 -0.24 -0.25 -0.22 -0.31
23INDSTRY6 Utilities industry
0.03
0.07
-0.01
-0.10 0.28
0.22
-0.32
0.07 -0.12 -0.12 -0.11 -0.15
22 23
1.00
-0.21
1.00
34
-------
Table 3a. Descriptive statistics: perceived pressures
Variable Description
Obs
Mean
SD
Min
Max
EMP_SUM Comprehensiveness of environmental management
319
3.95
1.02
1.50
6.10
practices (sum of normalized category sums)
EMP_PCA Comprehensiveness of environmental management
283
0.10
4.00
-9.95
9.37
practices (sum of category-level principle-component
factors)
INFLUPRI perceived influence of commercial pressures (comp, cust,
319
-0.02
1.01
-2.46
2.48
suppl, trade assoc)
INFLUCOM perceived influence of non-market pressures (community,
319
0.04
1.00
-2.88
2.65
ngos, regulators, media)
INFLUFIR perceived influence of internal pressures (corp mgt,
319
-0.04
1.01
-2.79
2.47
employee, other facilities)
COMPEMS Competitors with EMS
319
2.92
1.14
1
5
MDINCT Median household income
319
40.28
11.61
9.4
94.2
WHITEPOP Percentage population white
319
0.79
0.20
0.2
1.0
EMPLE1RT plant employees
319
0.48
0.67
0.001
5.5
UNW0001L TRI air releases
319
8.72
3.33
0
15.8
INTL_D Geographic breadth of operations
319
0.60
0.40
0
1
HQUS HQ located in US
319
0.93
0.26
0
1
BIZSTRAT Company's main business strategy
319
4.33
1.88
1
INDSTRY1 Automotive industry
319
0.12
0.33
0
1
INDSTRY2 Electrical/Electronics industry
319
0.18
0.38
0
1
INDSTRY3 Machinery industry
319
0.11
0.31
0
1
INDSTRY4 Primary Metals industry
319
0.17
0.37
0
1
INDSTRY5 Refining industry
319
0.28
0.45
0
1
INDSTRY6 Utilities industry
319
0.09
0.29
0
1
Table 3b. Correlations: perceived pressures
Variable
Description
1 2
3
4
5 6
7 8
9
10 11
1
INFLUPRI
perceived influence of commercial pressures
(comp, cust, suppl, trade assoc)
1.00
2
INFLUCOM
perceived influence of non-market pressures
(community, ngos, regulators, media)
-0.03 1.00
3
INFLUFIR
perceived influence of internal pressures
(corp mgt, employee, other facilities)
0.04 0.03
1.00
4
COMPEMS
Competitors with EMS
0.23 -0.12
0.04
1.00
5
MDINCT
Median household income
0.01 -0.04
-0.13
-0.03
1.00
6
WHITEPOP
Percentage population white
0.01 -0.16
-0.18
0.03
0.09 1.00
7
EMPLHRT
Plant employees
0.19 -0.06
0.00
0.04
-0.01 0.03
1.00
8
UNW0001L
TRI air releases, 2000-01
0.03 0.17
0.10
-0.10
-0.06 -0.07
0.12 1.00
9
INTL_D
Geographic breadth of operations
0.09 -0.21
0.13
0.24
0.12 0.03
0.17 -0.11
1.00
10
HQUS
HQ located in US
-0.14 -0.04
0.02
-0.15
-0.09 -0.02
-0.07 -0.05
-0.03
1.00
11
BIZSTRAT
Company's main business strategy
0.07 -0.06
0.23
0.02
0.04 -0.06
0.14 -0.05
0.25
-0.05 1.00
35
-------
Table 4 Regression results: Objective institutional pressure. Dependent variable comprehensiveness of Env. Mgt Practices4
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)
Sum
Sum
Sum
Sum
Sum
Sum
Sum
Sum
Sum
Sum
Sum
Sum
Sum
PCF
state greenness=PCA
0.07
[0.06]
Environmental lawyers
0.00
[0.02]
League of Conservation Voters
0.00
[0.00]
Number of state env policies
0.01
[0.01]+
Rating of 17 state policies
0.00
[0.00]
State's inverse pollution intensity
1.57
[3.76]
Environmental NGO members
0.05
[0.03]
Competitors with EMS
0.25
0.25
0.25
0.25
0.25
0.25
0.25
0.25
0.25
0.25
0.25
0.25
0.24
0.93
[0.05]**
[0.05]**
[0.05]**
[0.05]**
[0.05]**
[0.05]**
[0.05]**
[0.05]**
[0.05]**
[0.05]**
[0.05]**
[0.05]**
[0.06]**
[0.21]**
Company's business strategy (
0.04
0.04
0.04
0.04
0.04
0.04
0.04
0.04
0.03
0.04
0.04
0.03
0.05
0.17
[0.03]
[0.03]
[0.03]
[0.03]
[0.03]
[0.03]
[0.03]
[0.03]
[0.03]
[0.03]
[0.03]
[0.03]
[0.04]
[0.13]
Median 1999 household income
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.05
[0.00]*
[0.00]*
[0.00]+
[0.00]*
[0.00]*
[0.00]+
[0.00]*
[0.00]*
[0.00]+
[0.01]+
[0.02]*
Percentage population White
-0.60
[0.27]*
-0.55
[0.27]*
-0.57
[0.27]*
-0.21
[0.34]
-2.35
[1.11]*
Pet pop 25+ years college
-0.66
[0.43]
-0.60
[0.43]
Unweighted TRI air releases
0.05
0.04
0.05
0.05
0.05
0.05
0.04
0.04
0.04
0.04
0.04
0.04
0.15
[0.02]**
[0.02]*
[0.02]**
[0.02]**
[0.02]**
[0.02]**
[0.02]*
[0.02]*
[0.02]*
[0.02]*
[0.02]*
[0.02]*
[0.07]*
TRACI-weighted TRI air
0.00
[0.01]
plant employees,
0.16
0.18
0.17
0.16
0.16
0.17
0.18
0.18
0.18
0.18
0.17
0.18
0.21
0.59
[0.08]*
[0.08]*
[0.08]*
[0.08]*
[0.08]*
[0.08]*
[0.08]*
[0.08]*
[0.08]*
[0.08]*
[0.08]*
[0.08]*
[0.09]*
[0.34]+
Breadth of oper
0.41
0.40
0.41
0.40
0.41
0.42
0.40
0.40
0.38
0.37
0.40
0.38
0.47
1.32
[0.14]**
[0.14]**
[0.14]**
[0.14]**
[0.14]**
[0.14]**
[0.14]**
[0.14]**
[0.14]**
[0.14]**
[0.14]**
[0.14]**
[0.17]**
[0.57]*
HQ located in US
-0.50
-0.40
-0.50
-0.50
-0.50
-0.49
-0.41
-0.41
-0.42
-0.41
-0.42
-0.43
-0.15
-2.45
[0.20]*
[0.20]*
[0.21]*
[0.20]*
[0.21]*
[0.21]*
[0.20]*
[0.20]*
[0.20]*
[0.20]*
[0.20]*
[0.20]*
[0.32]
[0.84]**
Industry dummies
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Constant
3.12
2.98
2.98
2.87
2.93
2.99
2.96
3.00
3.14
2.97
3.41
3.40
3.45
-0.85
[0.47]**
[0.48]**
[0.46]**
[0.46]**
[0.47]**
[0.46]**
[0.46]**
[0.46]**
[0.48]**
[0.47]**
[0.50]**
[0.51]**
[0.59]**
[2.09]
Observations
354
360
354
354
354
354
360
360
360
360
360
360
236
314
R-squared
0.22
0.21
0.21
0.22
0.21
0.21
0.21
0.21
0.21
0.2
0.22
0.21
0.2
0.22
Standard errors in brackets + significant at 10%; * significant at 5%; ** significant at 1%
4 Dependent variables: Sum = EMP_SUM; PCA= EMP_PCA
36
-------
Table 5. PC A Institutional Pressures (Rotated Component Matrix)
Component
1
2
3
COMMERCIAL PRESSURE
Influence of competitors
.805
.154
.129
Influence of customers
.784
.165
Influence of suppliers
.696
.285
.166
Influence of trade associations
.506
.383
.238
Influence of SRI funds
.488
.470
.102
NON-MARKET PRESSURE
Influence of media
.278
.758
.133
Influence of environmental organizations
.296
.699
.161
Influence of local community
.201
.669
.286
Influence of regulators/legislators
-.146
.660
.173
FIRM INTERNAL PRESSURE
Influence of corp mgmt
.117
.759
Influence of other facilities in firm
.206
.172
.721
Influence of employees
.319
.683
Influence of shareholders
.420
.298
.509
Extraction Method: Principal Component Analysis. Rotation Method: Varimaxwith Kaiser Normalization.
37
-------
Table 6. Regression results of perceived institutional pressures
Dependent variable: Comprehensiveness of environmental management practices5
(1)
(2)
Sum
PCF
Perceived influence of commercial pressure
0.31
1.25
[0.05]**
[0.20]**
Perceived influence of non-market pressure
-0.01
0.06
[0.05]
[0.20]
Perceived influence of firm internal pressure
0.37
1.37
[0.05]**
[0.20]**
Competitors with EMS
0.20
0.70
[0.04]**
[0.19]**
Median 1999 household income
0.00
-0.02
[0.00]
[0.02]
Percentage population White
-0.31
-1.63
[0.24]
[0.99]
TRI air releases, 2000-01
0.01
0.05
[0.02]
[0.07]
Plant employees
0.13
0.42
[0.07]+
[0.31]
Geographic breadth of operations
0.33
1.03
[0.13]**
[0.53]+
HQ located in US
-0.44
-2.27
[0.18]*
[0.74]**
Company's business strategy
-0.01
0.05
[0.03]
[0.12]
Industry dummies
Y
Y
Constant
3.86
0.2
[0.46]**
[1.93]
Observations
319
283
R-squared
0.43
0.42
Standard errors in brackets + significant at 10%; * significant at 5%; ** significant at 1%
' Dependent variables: Sum = EMP_SUM; PCA= EMP_PCA
38
-------
UNIVERSITY OF CALIFORNIA, SANTA BARBARA
BERKELEY • DAVIS • IRVINE • LOS ANGELES • MERCED • RIVERSIDE • SAN DIEGO • SAN FRANCISCO
SANTA BARBARA • SANTA CRUZ
1 > -v .« A* . -7 * ' _
Survey on Environmental Management Practices
Thank you for agreeing to take part of this benchmarking exercise on environmental management
practices.
• All individual responses will be kept strictly confidential
• Please try to answer every question, even though you may not be 100% sure of your answer.
If you ha-ve any questions, please feel free to contact Paolo Gardinali, Associate Director of the
Social Science Survey Center (SSSC) at (805) 893-3887 or paolo@survev.ucsb.edu
Project Principal Investigators:
Professor Magali Delmas
Professor Dennis Aigner, Dean
Donald Bren School of Environmental Science and Management
University of California, Santa Barbara
Graduate Research Assistant: Mike Toffel, Haas School of Business.
University of California, Berkeley
Please return this questionnaire to:
Social Science Survey Center
ISBER, 2201 North Hall
University of California
Santa Barbara, CA 93106
Fax: (805) 893-7995
You can also fill the questionnaire online at http://www.survey.ucsb.edu/env
with the code:
CONFIDENTIAL
39
-------
Section I. General information
1. Parent company information
Parent company name:
Location of parent headquarters (country):
2. Does your company operate facilities outside of the United States? ~ Yes ~ No
If YES, where are they located? (check all that apply): ~ Europe ~ Asia ~ Elsewhere
In this questionnaire, we are asking questions about your facility: a facility includes buildings
that are on a contiguous site and under common control by a company.
Section 2. 1 ji\ ironmental management organization
3. Which of the following most closely reflects your position? Please check one:
~ EHS manager or specialist
~ Environmental manager or specialist
~ Plant manager
~ Other, please describe
4. Approximately how many full time equivalent employees (FTEs) are working on environment,
health and safety issues for your facility?
EHS / Environmental department
Other departments
Total
Section 3. 1 ji\ iroiinicntal management practices
5. If your facility or company has an environmental policy, how is it communicated?
Please check all that apply
~ We do not have an environmental policy We have an environmental policy and:
~ post the policy around our facility
~ post it on the internet
~ distribute it to all facility employees
~ most employees have discussed the policy
with a manager/supervisor
CONFIDENTIAL
40
-------
6. Over the past 3 years, how many times has your facility had an internal environmental audit
conducted by your facility staff and/or corporate staff? (If none, please enter zero)
7. Over the past 3 years, how many times has your facility had an external environmental audit
conducted by third-parties such as consultants, not including regulators or corporate staff?
8. Approximately, what proportion of your employees at your facility have received environmental
training over the past 12 months in the following departments?
Environmental training includes coursework or team meetings where environmental policies,
procedures and impacts are discussed or disseminated.
Please check one for each department.
No such
0-20%
21-40%
41-60%
61-80%
81-100%
department
at my facility
Management
u
u
u
u
u
u
()|vmlions
J
J
J
_l
_l
_l
Maintenance
u
u
u
u
u
u
I jiL'iiKvniiL' R&D. Design
J
J
J
_l
_l
_l
Sales
u
u
u
u
u
u
Purchasing hocuivmenl
J
J
J
_l
_l
_l
9. In job performance reviews for employees at your facility, how important do you consider the
contribution of your employees to environmental performance?
Please check one for each department. Use your best estimate if you are unsure.
Not part of
Low
Moderate
Important
Very
No such
review
importance
importance
important
department
at my facility
General
U
u
U
u
u
u
management
()|vmlions
J
J
J
J
J
J
\l;iinlon;inco
J
J
J
J
J
J
Iji'jiiKvriii'j
J
J
J
J
J
J
R&D. Design
Environment,
u
u
u
u
u
u
health & safety
Snl os
J
J
J
J
J
J
Purchasing
Procurement
/ U
u
u
u
u
u
CONFIDENTIAL
41
-------
Section 4. Relations with stakeholders
Never Occasionally Frequently All the
time
10. To what extent does your purchasing department ~ ~ ~ ~
use a green purchasing policy?
11. To what extent does your purchasing department ~ ~ ~ ~
request your suppliers to be ISO 14001 certified?
12. To what extent does your purchasing department ~ ~ ~
ask suppliers to provide information about their
environmental management practices?
~
13. Approximately what proportion of 0_20o/o 21.40o/o 41.60o/o 61_80% 81.100o/o D°n'1
your competitors have adopted an know
environmental management system ~~~~~~
(EMS)? (Certified or non-certified).
14. What is the status of your participation in
voluntary US EPA or state programs such
as Energy Star, Wastewise, Environmental
Performance Track, etc.
,, ,, . t-i , Planning Currently
Not being Future ^ . . :
• j , • , to participating
considered consideration . . A r 1 &
participate
~ ~ ~ ~
15. What is the status of your participation in
industry-led environmental programs such ~ ~ ~ ~
as Responsible Care, industry climate
challenge programs, etc.
16. How often does your facility solicit opinions
from environmental non-profit
organizations, such as involving them in site
planning or in identifying environmental
impacts?
Never Rarely Sometimes Often Very All
often the
time
~ ~ ~ ~ ~ ~
CONFIDENTIAL
42
-------
17. What is the status of the following certifications at your facility?
Not being Future Planning to Currently Successfully
considered consideration implement implementing implemented
ISO 9000 certification ~ ~ ~ ~ ~
ISO 14001 certification ~ ~ ~ ~ ~
18. Approximately how many complaints has your facility received from the surrounding
community about odors, noise, smoke, dust, effluents, water pollution, or aesthetic appearance
in the last three years?
19. To what extent have each of the following groups influenced your facility to improve its
environmental performance?
No influence
Little
influence
Some
influence
Strong
influence
Very strong
influence
Customers
U
U
U
U
U
Suppliers
U
u
u
u
U
Competitors
u
u
u
u
u
Trade associations
u
u
u
u
u
Local community
u
u
u
u
u
Socially responsible investment funds
u
u
u
u
u
Environmental organizations
u
u
u
u
u
Media
u
u
u
u
u
Shareholders
u
u
u
u
u
Corporate management
u
u
u
u
u
Employees
u
u
u
u
u
Other facilities within company
u
u
u
u
u
Regulators/legislators
u
u
u
u
u
CONFIDENTIAL
43
-------
20. To what extent have the following corporate departments influenced your facility to improve its
environmental performance? Please check one for each department:
Our
No Little Some Strong Very strong corPoral'on
• r, r, r, f! ii does not have
influence influence influence influence influence
such
department
Corporate environmental
management
~
~
~
~
~
~
Corporate legal & regulatory affairs
U
u
u
u
u
u
Corporate public relations
U
u
u
u
u
u
Corporate strategy
u
u
u
u
u
u
Corporate marketing
u
u
u
u
u
u
Corporate product design
u
u
u
u
u
u
Other (please specify):
~
~
~
~
~
~
Section 5. Measuring and reporting environmental performance
21. To what extent are your facility's environmental Not at To a To some To a To a
costs identified in cost accounting? aii limited extent large great
extent ' extent extent
~ ~ ~ ~ ~
22. Which of the following issues are significant environmental issues at your facility?
~ Air emissions
~ Water pollution
~ Solid Waste
~ Hazardous Waste
~ Noise
~ Other, please describe
23. For which of these environmental issues do you have objectives and targets?
~ Air emissions
~ Water pollution
~ Solid Waste
~ Hazardous Waste
~ Noise
~ Other, please describe
24. Do you disseminate your facility Toxic Release Inventory (TRI) data to the public in an easily
accessible format (beyond reporting this data to the US EPA)? ~ Yes ~ No
CONFIDENTIAL
44
-------
Section ft. Molixnlions lor en\ iionmenlal mnnngemenl practices
25. In addition to improving environmental performance, how important are the factors listed below
in motivating your facility to implement environmental management practices?
Not Somewhat T Very
important important important
Increase customer loyalty
u
u
u
u
Roach now cuslomors
J
J
J
J
Improve eni|ilo\oo moiiwilion or moralo
J
J
J
J
1 lolp L'onomlo now producls or son icos
J
J
J
J
lni|no\o royukilorx complianco
J
J
J
J
Inlluonco ivikIiiil' loL'iskilion or lobulations
J
J
J
J
Improve relations with environmental non-profit
organizations
~
~
~
~
lni|no\o rokilions Willi our local communil\
J
J
J
J
26. Please describe your company's main business strategy using the following scale, from "provide
low cost products or services" (1) to "differentiate our products" (7)?
1 2 3 4 5 6 7
Differentiate
Provide low cost ~~~~~~~ our Pr°ducts
products or services on the
market
27. Please check the following box if you would like to receive a copy of the final report in the
future.
~ Yes, I would like to receive a copy of the report.
Name:
E-mail:
Facility name:
Street address:
City:
State and Zip code:
~ No, I am not interested.
CONFIDENTIAL
45
-------
28. Please provide any comments and suggestions here:
Thank you!
CONFIDENTIAL
-------
Environmental Management Systems: Informing Organizational Decisions
Chris T. Hendrickson, Deanna H. Matthews, and Lester B. Lave
Carnegie Mellon University
Pittsburgh, PA
Abstract
Approaches to improve environmental performance have expanded to include voluntary programs
that encourage organizations to go beyond regulatory compliance. Environmental management
systems (EMS) are a recent type of voluntary initiative expected to produce general reductions in
pollution discharges. Unlike voluntary programs such as Green Lights or 33/50, the specific goals
and benefits of EMS implementation to participating organizations and government regulators has
not been defined. Participating companies expect to lower environmental costs or to improve
performance. Society should benefit from reduced energy consumption, pollution, and waste
generation. However, not all EMS will be equally effective. For example, a process-oriented EMS,
such as ISO 14000, may not provide a comprehensive view of environmental issues across an
organization or develop the data needed to assess environmental improvements and cost savings.
We examined existing EMS at various levels. First, at a macro-level, we assessed the change in
environmental performance as a result of adopting an EMS. Second, at a micro-level, we examined
the existing EMS in a number of organizations and the extent to which the system provides relevant
data and analysis to inform company decisions. The data provide a basis for identifying the EMS
attributes that are useful and necessary for decision-making. Organizational leaders can use the
results to improve the effectiveness of existing or new EMS. Policy makers can use the results to
determine requirements for a potential voluntary program for implementation of environmental
management systems.
47
-------
Introduction
U. S. Environmental Protection Agency actions to improve companies' environmental performance
have expanded to include voluntary programs that encourage organizations to go beyond regulatory
compliance. Environmental management systems (EMS) are a recent type of voluntary initiative
expected to produce general reductions in pollution discharges. At the same time, environmental
issues are becoming more strategic to business, so firms are implementing EMSs as a means to
capture and assess environmental issues across operations. Participating companies expect to lower
environmental costs or to improve performance. Society should benefit from reduced energy
consumption, pollution, and waste generation. However, not all EMS will be equally effective at
achieving these goals. The research presented here examines existing EMSs to help address the
future role of environmental management systems in environmental policy and organizational
decision-making. The research recognizes the need for evaluating individual components of
environmental management systems, as well as whole systems. If EMSs are effective, then fewer
problems from regulatory compliance may occur and the burden of regulation may be reduced. It is
essential for both government agencies and firms to know what components and ways of operating
the system are most effective in improving environmental performance. This research investigates
how EMSs and environmental information in general are used by organizations and, in turn,
proposes a design for an EMS that provides the information needed for better decisions.
The research examines EMSs at two distinct levels. First, at the macro-level, we assess facility
performance in relation to ISO 14001 EMS adoption and certification. The macro-level analysis
uses publicly available information, most of which is collected by government agencies via
regulatory requirements, in order to assess overall patterns in environmental performance. The
analysis provides a picture of what can be learned from existing data about environmental
performance. Specifically, the analysis assesses the ISO 14001 environmental management system
standard in relation to these data providing policy makers with information on how the standard
currently integrates with other environmental initiatives. The results show little difference between
the environmental performance of facilities. These results have serious implications for policy
makers in how certified environmental management systems should be used to evaluate facility
environmental performance.
Once we know generally how firms with a formal EMS compare to those without, the micro-level
analysis allows us to find out why. By investigating the environmental information available
internally to an organization, we learn what data are used merely for regulatory reporting and which
are used to guide decision makers. An effective EMS is expected to include information collection
and dissemination of useful, relevant, and timely data to inform company decisions. Results show
that most data are regulatory based, reported outside environmental groups infrequently, and thus
limit their use in decision-making toward improvements in environmental performance. This
detailed inquiry into data availability and use can help businesses and policy makers choose
measures that reflect environmental performance, can be understood by various stakeholders, and
lead to improvements across industry.
The overall results of the research lead to suggestions for improving environmental management
systems for organizational decision-making and policy development. First, EMSs must stretch
beyond the current regulatory issues to be effective in long-term improvement. Unlike other
voluntary programs, EMSs do not require organizations to operate outside the boundary of current
regulatory issues. Second, EMS goals, targets, and resulting performance must be made more
transparent. The link between efforts to reduce environmental problems via an EMS and reported
48
-------
environmental performance metrics is tenuous. Finally, as various environmental issues continue to
shift in importance, EMSs must adapt to changing organizational focus and monitor potential future
regulatory issues. For example, corporate social responsibility and calls for sustainability
(incorporating social and economic factors, as well as environmental factors) and the global
importance of carbon emissions (currently unregulated in the U.S.) are influencing corporate
strategy. An EMS should support decision-making on these pressing issues, especially for
multinational companies operating under different regulatory schemes.
Macro-Level Analysis - Facility Environmental Performance In Relation to ISO 14001
Certification
The ISO 14000 series for environmental management systems continues to grow in popularity as a
means for organizations to address environmental issues in their facilities across operations.
Approximately 62,000 organizations worldwide have been certified as following the ISO 14001
EMS standard as of December 2003, but only about 3,500 in the U.S. (1). Since participation is
voluntary, certification to ISO 14001 is regarded as an indication of a firm's interest in
environmental improvement and going "beyond compliance" to address environmental problems.
This research investigates the level of environmental performance in facilities in relation to
certification to the ISO 14001 EMS standard. The analysis focuses on automobile assembly
facilities in the U.S. consisting of approximately 50 facilities in 20 states producing a variety of
cars, vans, trucks, and SUVs. The sector has a history of dealing with environmental issues. Both
the emissions and wastes from manufacturing operations as well as from the use-phase of the final
product have resulted in public and regulatory attention to the industry. In addition, the automotive
industry has been a leading sector in ISO 14001 EMS certification. Ford Motor Company initiated
a commitment to establishing EMSs and certifying its facilities to the ISO 14001 EMS in the early
stages of the standard. All of Ford's U.S. facilities were certified by early 1999. The other two
major U.S. firms as well as foreign-owned firms followed this lead and began to implement the
standard in their own facilities. To further reinforce their commitment to environmental
management systems, the major U.S. firms announced that suppliers would be required to
implement and certify to the standard as well, with deadlines for certifying in late-2002 or mid-2003
(2, 3). Approximately one-fifth of all U.S. certifications are held by facilities with business related
to the automotive sector (4).
The study examines four different measures of environmental performance over the period from
1993 to 2003. The measures include toxic chemical releases, criteria air pollutant emissions,
hazardous waste generation, and compliance to regulatory requirements. Most performance
measures have been normalized to production, allowing a comparison of facilities on a per-vehicle
basis. During the time period, all facilities, regardless of EMS status, made steady progress in
reducing their environmental burdens. The analysis shows that in later years, certified facilities are
not performing better than facilities that chose to certify their EMS later. In addition, the results
give no indication that the facilities are achieving an increase in the rate of improvement that was
seen prior to adopting and certifying to the standard. In some cases, certified facilities are more
likely to have worse performance once the EMS is operating.
The ISO 14001 EMS standard defines an EMS as "that part of the overall management system
which includes organizational structure, planning, activities, responsibilities, practices, procedures,
processes and resources for developing, implementing, achieving, reviewing, and maintaining the
environmental policy" (5). The intent of the standard is to provide a structured framework for an
EMS that is based on a commitment to continual improvement in overall environmental
49
-------
performance. By periodically reviewing and evaluating the environmental management system, the
organization will identify opportunities for improvement.
These implications of the standard lead to two main hypotheses. First, facilities that have
established and implemented an ISO 14001 EMS and now have a system for and commitment to
environmental performance improvement should have better environmental performance than those
facilities without such a system in place. A structured EMS gives a facility the ability to
systematically identify and address environmental problems, which presumably leads to reduced
environmental discharges.
Second, facilities that have established and implemented an ISO 14001 EMS have undergone a
change in operations. In theory, the system provides them with a tool to reduce environmental
impacts that was previously unavailable. Once the standard is in place and operations are regularly
being assessed and evaluated, the facilities should continually improve on the past performance
levels and likely at a greater rate of improvement. From year to year after certification, one would
expect a facility with a certified ISO 14001 EMS to make additional reductions in overall
environmental impact than had been achieved previously.
Research Method
The sample includes approximately 50 automobile assembly facilities in the U.S. across the time
period from 1993-2003. The number of facilities in each analysis varies depending on data
availability. The Automotive News Market Databook provides parent company, facility location,
vehicle descriptions, and production information. Company websites or correspondence supplied
some data and helped to validate existing data. The World Preferred Registry provides a list of
entities holding registrations to the ISO 14001 EMS standard and the date of certification. Three
variables were constructed from the information on ISO certification. First, we separate the
facilities into two groups, defined as "early adopters" if certification occurred in 1998 or 1999 and
"late adopters" if certification occurred in 2000 or after (adoption status). Second, the date of
certification establishes the stage of implementation of the ISO 14001 EMS in a facility (ISO
Stage). The variable is a counter variable that increases by one as a facility moved through different
stages - from no system (0), implementing (1), certifying (2), and operating under the EMS (3 and
up). A third variable represents the status of the EMS as either certified or not certified in a given
year (ISO status).
Environmental performance data were collected from the U.S. Environmental Protection Agency
public databases. The Toxics Release Inventory (TRI) (1993-2001), AirData on criteria air
pollutant emissions (1996 and 1999), National Biennial RCRA Hazardous Waste Report (1993-
2001, odd-numbered years), and the Enforcement and Compliance History On-line (ECHO) (1996-
2003) provide data. We construct the final sample for each metric by matching entries in the
various databases by the name of the facility, address information, and EPA facility information
which includes identification numbers for the facilities under various programs.
Results
Table 1 shows the average total toxic chemical releases and average total toxic chemical waste
managed per vehicle for the facilities from 1993 to 2001. The number of facilities in the Certified
Facilities group increases each year as facilities certify their EMSs. Overall, the facilities show a
decreasing level of chemical releases and wastes per vehicle as expected. From 1997 to 2001, the
non-certified facilities show a much larger improvement in total toxic chemical releases and amount
50
-------
of toxic chemical waste managed - almost twice that of the certified firms. It should also be noted
that the variance in releases per vehicle is large for the group of facilities, ranging from a low of
-0.0 pounds/vehicle to a high of 8.0 pounds/vehicle in 2001.
Table 1: Trend in Average Toxic Chemical Releases and Waste Management.
Total Releases
Total Waste Management
(pounds per vehicle)
(pounds per vehicle)
All Facilities
All Facilities
1993
6.19
13.18
1994
5.10
10.74
1995
4.80
11.26
1996
4.63
10.96
1997
4.01
9.85
Certified
Non-certified
Certified
Non-certified
Facilities
Facilities
Facilities
Facilities
1998
3.64
3.87
9.24
8.11
1999
3.64
3.66
8.66
7.92
2000
3.36
3.30
8.28
8.11
2001
3.32
2.77
8.40
6.03
% Change 1997-2001
-17%
-31%
-15%
-39%
Using standard OLS multiple regression analysis to determine if any difference exists between
facilities once certification has been achieved produced no significant statistical support for the two
proposed hypotheses.
51
-------
Table 2 shows the results of three different regression models. For each regression model, the
dependent variable is pounds of toxic releases per vehicle. As expected, coefficients for time and
production are negative, suggesting improvement over time and some scale efficiency. Facilities
that assemble trucks produce only about 0.5 pound more toxic chemicals per vehicle than those
facilities that assemble cars. Facilities in the DaimlerChrysler company have lower emissions than
average, compared to both Ford and General Motors facilities which have higher emissions of the
same magnitude. Foreign or multiple-owner facilities are the base case. Each of these seven
parameter coefficients are significant (p-value < 0.05). Adoption status, differentiating between
early and late adopters, has a negative coefficient indicating that early adopters generally have
lower toxic wastes per vehicle than later adopters. Likewise, the longer the ISO 14001 EMS is in
place the lower toxic wastes per vehicle, as noted by the negative coefficient of the ISO stage
variable. The ISO Status variable, indicating whether a facility has a certified ISO 14001 EMS in
place in a given year, has a positive coefficient indicating increased toxic wastes for these facilities.
However, none of these three coefficients is statistically significant (p-values >0.4) and the
estimated magnitude of the effect is small. Similar results are achieved for the other TRI variables
investigated. No difference is seen across facilities in each group for total waste managed, waste
that is recycled, recovered, or treated, or when weighting the chemicals for toxicity prior to analysis.
In addition, no statistical significance is found for other model specifications considering such
variables as change in TRI pounds and change in production from the previous year
52
-------
Table 2: Results of OLS Regression for Toxic Chemical Releases.
Model 1
Model 2
Model 3
Constant
4.43
(0.737)
4.470
(0.769)
4.457
(0.758)
Time (years since 1993)
-0.35
(0.043)
-0.357
(0.064)
-0.354
(0.057)
Production (1000's of vehicles)
-0.0083
(0.0011)
-0.0083
(0.0011)
-0.0083
(0.0011)
Car
2.50
(0.443)
2.498
(0.444)
2.498
(0.444)
Truck
3.03
(0.417)
3.032
(0.418)
3.031
(0.418)
DaimlerChrysler
-1.08
(0.525)
-1.075
(0.526)
-1.076
(0.525)
Ford
1.12
(0.396)
1.116
(0.398)
1.118
(0.397)
GM
0.94
(0.466)
0.935
(0.467)
0.935
(0.467)
Adoption Status (early vs. late)
-0.33
(0.469)
-0.353
(0.492)
-0.345
(0.485)
ISO Stage (years with certified EMS in place)
-0.024
(0.128)
ISO Status (certified EMS in place)
0.060
(0.371)
p-values all less than 0.05 except for adoption status (Model 1
Model 3: p-value=0.47), ISO stage (Model 2: p-value = 0.85),
p-value = 0.48, Model 2: p-value = 0.47,
and ISO Status (Model 3: p-value= 0.87).
53
-------
Table 3Table 3 and Table 4 show the average tons of criteria air pollutant emissions per 1000
vehicles from 1996 and 1999. Nineteen facilities are in the group of early adopters certified by
1999, while 27 facilities are in the group of late adopters certified after 1999. First, consider the
comparison of the "average" facility in 1996 - when no facilities were certified - to the two groups
of facilities in 1999 - certified and non-certified (Table 3). Both groups of facilities achieved
reductions in criteria air emissions over the average 1996 levels despite increases in production,
with facilities with a certified EMS having an overall lower level of emissions per vehicle in 1999
than facilities without a certified EMS. In fact, facilities without certified EMS had an increase in
particulate matter emissions.
54
-------
Table 3: Average emissions of criteria air pollutants (tons/1000 vehicles).
1996
999
Average Facility
ISO-Certified
Not ISO-Certified
Production
207,000
246,000
217,000
CO
0.371
0.231
0.309
NOx
0.882
0.341
0.679
VOC
6.311
4.348
4.890
S02
0.618
0.043
0.513
PM25
0.088
0.025
0.120
PM10
0.117
0.082
0.163
Total
8.301
5.047
6.556
Table 4Table 4T indicates, however, that the facilities which certified an EMS by 1999 (the early
adopters), did not attain reductions in criteria air pollutant emissions to the extent that facilities
which did not certify an EMS (late adopters) were able to attain. Overall, late adopters show a
greater reduction in air emissions - a 37% decrease, versus a 5% decrease for the early adopters
who were implementing and certifying their EMSs in the interim years. During this time, the late
adopters also increased production considerably. The largest absolute improvements came in VOC
emissions which account for 75% of all emissions. The early adopters made large relative
improvements in emissions of sulfur dioxide compared to the late adopters; however, the total
reduction of sulfur dioxide by late adopters was more than three times the amount. Regression
analysis using OLS methods again did not indicate any statistical significance in the ISO
certification variables. Statistical analysis to determine if a difference exists between the mean
values for the different groups of facilities did not show significance.
Table 4: Average emissions of criteria air pollutants (tons/1000 vehicles).
1996
1999
ISO-Certified
ISO-Certified
ISO-Certified
ISO-Certified
by 1999
after 1999
by 1999
after 1999
Production
240,000
183,000
246,000
217,000
CO
0.126
0.543
0.231
0.309
NOx
0.489
1.158
0.341
0.679
VOC
4.468
7.608
4.348
4.890
S02
0.112
0.974
0.043
0.513
PM25
0.080
0.094
0.025
0.120
PM10
0.103
0.127
0.082
0.163
Total
5.299
10.413
5.047
6.556
For hazardous wastes, again a general trend of improvement is seen across all facilities, although no
difference is apparent between facilities with certification and those without certification of an
EMS. Table 5 shows the trend in hazardous waste generation from 1993 to 2001. Both groups of
facilities experienced similar generation rates in the latter years as some facilities were undergoing
implementation and certification of an EMS. In comparing the initial 1993 figures to the final 2001
figures, facilities that had certified the EMS to ISO 14001 had a higher percentage of reductions - a
27%) decrease over the 8 years, versus 19% decrease for those without certification. In a statistical
comparison of the mean tons per 1000 vehicles for each year, no significant difference exists
between the two groups of facilities, indicating that time of adoption does not have a greater
55
-------
influence on hazardous waste generation. Again, OLS regression shows no statistical significance
in the ISO-certification variables in relation to waste generation per vehicle.
Table 5: Trend in Hazardous Waste Generation.
Hazardous Waste Generation
(tons per vehicle)
All Facilities
1993
7.15
1995
6.03
1997
5.78
Certified
Facilities
Non-certified
Facilities
1999
4.82
4.33
2001
4.22
4.67
% Change 1997-2001
-27%
-19%
The data for regulatory compliance appear in Table 6. Note that the data represent information for a
two-year period inclusive; inspections, violations, penalties or corrective action could have occurred
at any point in the two-year period. The first column represents a period when no facilities had an
EMS certified to the ISO 14001 standard, although several facilities were in the implementation
phase. The second column represents 1998-2000 and the facilities are separated into two groups -
those which certified an EMS during the time and those which did not (effectively early adopters
and late adopters). The third column represents 2001-2003 and again the facilities are separated
into two groups - those with EMSs which had been certified and in operation during the time period
and those which were implementing an EMS for certification (again, effectively early adopters and
late adopters). Inspection rates across the first two time periods are consistent, as are rates between
the groups in the final year. This provides a baseline that compliance issues are not biased by a
change in inspection activity from regulators.
The data reveal several differences between the groups of facilities. First, facilities that are in the
implementation phase prior to certification show higher occurrences of violations or noncompliance
events. This is true for the early adopters in the 1998-2000 period, and the late adopters in 2001-
2003, although it is more pronounced for the early adopters. Perhaps more importantly, 77% of
facilities with an established EMS had a regulatory compliance issue, indicating that the EMS is not
entirely successful in allowing facilities to maintain compliance. The overall rate of violations or
noncompliance events for all facilities in 2001-2003 is 78%, similar to the rate in 1996-1998
without any certification. On average, facilities were out of compliance for 6 of the 8 quarterly
periods over each time frame. It is possible that violations and non-compliance events are likely
getting resolved more quickly over time. The number of continuing events (those with 8 quarters of
continuous noncompliance) decreased from 66% in the first period to about 40% in the last period.
56
-------
Table 7: Regulatory compliance of firms from 1996-2003 in relation to ISO 14001 EMS certification.
1996-1998
1998-2000
2001-2003
No
Facilities with
Facilities with
Facilities with
Facilities with
facilities
EMS certified
no EMS
EMS certified
EMS certified
with EMS
during time
certified during
prior to time
during time
certified
period
the time period
period
period
Total number of facilities
50
22
28
30
20
Facilities that have been inspected
46
20
26
26
17
Total number of inspections
186
71
99
75
42
Percent of facilities inspected
92%
91%
93%
87%
85%
Average number of inspections per facility
4.0
3.6
3.8
2.9
2.5
Facilities with violation or noncompliance event
38
20
19
23
16
Percent of facilities with violation or noncompliance event
76%
91%
68%
77%
80%
Quarterly periods with 1 or more violation or noncompliance
event
240
115
112
159
96
Average periods with violation or noncompliance event per
facility
6
6
6
7
6
Continuing noncompliance/violations
25
11
10
9
7
Percent of facilities with continuing noncompliance or
violations
66%
55%
53%
39%
44%
Facilities with significant noncompliance
13
4
1
12
10
Percent of facilities with significant noncompliance event
34%
20%
5%
52%
63%
Facilities with pollutant release exceedances
9
4
3
6
4
Number of parameters over limit
18
6
4
8
5
Average number of parameters over limit per facility
2
1.5
1.3
1.3
1.3
Percent of facilities with pollutant exceedances
18%
18%
11%
20%
20%
Number of reports over limit
92
28
21
26
14
Number of reports submitted
9843
6116
3246
5495
1429
Facilities with pollutant spills
10
3
0
10
9
Total number of pollutant spills
13
5
0
15
19
Average number of pollutant spills per facility
1.3
1.7
0
1.5
2.1
Percent of facilities with pollutant spills
20%
14%
0%
33%
45%
Facilities where enforcement actions were taken
9
9
3
18
5
Number of Facilities assessed penalties
4
5
1
17
5
Total penalties assessed
$144,000
$490,000
$56,000
$1,650,000
$176,000
Total number of enforcement actions taken
15
14
5
24
6
Average number of enforcement actions taken per facility
1.7
1.6
1.7
1.3
1.2
Percent of facilities where enforcement actions taken
18%
41%
11%
60%
25%
57
-------
In the latter time period, of the 39 facilities which had a violation or noncompliance event, 22 had
events with significant noncompliance ratings corresponding to 52% of facilities with EMS certified
prior to the time period, and 63% of facilities in the implementation stage. These percentages are a
large increase from the baseline year (34%) or the middle period (20% and 5%). Facilities with no
certified EMS had fewer pollutant release exceedances. More pollutant spills are occurring in the
facilities over time. During the last period, almost 40 spills occurred at 20 facilities, about twice
that of the initial period. Facilities with an operating EMS received far more enforcement actions
(18 out of 30 facilities) and faced more fines ($1.7 million) than the other groups of companies
despite having a comparable number of enforcement actions per facility.
Discussion of Results
The results do not support either hypothesis formulated about the relationship between facility
environmental performance and ISO 14001 certification. The first hypothesis proposed that
facilities that have established and implemented an ISO 14001 EMS and now have a system for and
commitment to environmental performance improvement should have better environmental
performance than those facilities without such a system in place. Overall, the U.S. automobile
assembly facilities exhibit no substantial difference in environmental performance in relation to the
implementation and operation of an EMS certified to ISO 14001. The total emissions and wastes
from all facilities decreased over the 1993-2001 period which is to be expected. Emissions and
wastes for individual facilities trend downward, although fluctuations both up and down are
widespread. Similarly, the wide variance in performance for any metric in any year indicates that
some facilities have large volumes of waste or emissions that have been eliminated by other
facilities with low volumes. Regulatory compliance has not changed considerably over the time
period, for better or for worse.
The second hypothesis proposed that facilities with a certified ISO 14001 EMS should continually
improve on past performance levels and likely at a greater rate of improvement. However, once a
facility had certified an EMS to the ISO 14001 standard, improvements in environmental
performance did not accelerate from past performance. In some cases, facility performance was
actually worse after an EMS was implemented and certified to the ISO 14001 standard.
One important consideration is whether the performance measures chosen for the analysis match
areas where the facilities put efforts for improving performance. Within the ISO 14001 EMS
framework, each facility may choose specific impacts to target for reduction or improvement.
These goals and targets are not made public and may not correspond to the environmental
performance measures represented by the public data used in the analysis. However, as the ISO
14001 EMS standard claims to provide a framework for overall improvement in environmental
performance, some relationship would be expected. As the facilities move to organized methods of
documenting impacts, assigning responsibilities, and incorporating a general awareness of
environmental issues, all environmental impact areas should see some improvement.
The firms in the automobile assembly industry publicize their performance in environmental areas
using a variety of metrics. Most common for facility performance are measures of energy use and
carbon dioxide emissions, measures of materials consumption and waste generation, and water
consumption. None of these data are public at the facility level, except hazardous waste data which
is only a small portion of waste generation (on the order of 1% of total waste). So, a relationship
would be expected, and this is the area where we see the strongest support of performance in
relation to ISO 14001 certification. Early adopters decreased hazardous waste generation by 27%
while late adopters only decreased generation by 19% from 1993 to 2001. Still, no statistically
58
-------
significant difference in the groups exists. In addition, the early adopters experience a 24% increase
in hazardous waste generation from 1999 to 2001 when the facilities would have had an ISO 14001
EMSs in operation approximately 3 years. Clearly not continued improvement in environmental
performance.
One aspect outside the scope of this research is how cost influences changes in environmental
performance. It is possible that through implementing and certifying an EMS to the ISO 14001
standard that facilities achieved reduction in waste or emissions, which may be equal to reductions
at facilities which did not initiate ISO 14001 implementation, at a lower cost. The standard may
allow facilities to identify more cost-effective methods of improving performance. Organizations
would then have to consider if these cost savings match the costs of implementing and maintaining
the EMS to the ISO 14001 standard.
Two factors of how the ISO 14001 EMS standard is structured may explain to the results and, after
a longer time period, may lead to better performance. First, given the work related to implementing
the standard, facilities undergoing certification may have a better grasp of environmental impacts
existing at the facility. From a thorough audit and investigation of facilities operations for creation
of an EMS, facilities may identify sources of waste and emissions, do a better inventory of wastes
and emissions, or become aware of or more diligent of compliance issues. This would increase the
figures used to measure performance in comparison to earlier years without the standard. Yet, these
figures would be a more accurate depiction of actual performance. If this is true, longer time
periods of analysis would identify improvements once the baseline for emissions and waste
generation had been shifted to these new levels. However, the data for toxic releases and hazardous
wastes, which provide measures for about 20 facilities after 3 years of implementation do not show
significant improvements with time. Similarly, the number of violations and noncompliance events
for these facilities with well-established ISO 14001 EMSs are similar to those prior to
implementation.
A second factor of the ISO 14001 EMS standard that may not be reflected in the performance data
is the potential impact of better general management of environmental issues once a facility
implements ISO 14001. The standard requires facilities to identify responsibilities and initiate a
cycle of audit and review of operations of the EMS. These requirements, while not necessarily
having a direct impact on day-to-day performance of operations, may in the long run assist in
reducing environmental impacts, especially one-time problematic events. For example, if an
employee with environmental oversight leaves the position, the duties and concerns of the position
are documented and more easily transferred to other personnel. This ensures continuity of
operations and reduces the chance that certain activities, such as monitoring an effluent or annual
training, will go unchecked. Better overall management of environmental issues would not be
directly translated to environmental performance improvements. This then identifies a shortcoming
in the ISO standard for achieving improvements.
Micro-Level Analysis - EMS Structure and Information Systems
An EMS is intended to address all activities related to environmental issues, including such
activities as monitoring wastes and emissions, complying with regulatory requirements, developing
new products, and providing service to customers. In many cases, organizations gather existing
environmental activities under a single framework to establish a formal EMS. The ISO 14001 EMS
is becoming the de facto model for EMS. The standard provides a structure to EMS by outlining
59
-------
specific elements that must be included, but is non-prescriptive in how the elements might be
fulfilled, allowing for flexibility among users.
The basic components of an EMS include: a mission statement, documented environmental policy,
goals, timelines, data collection and organization, information systems, identification of
environmental impacts, regulatory requirements, personnel responsibility and task list, training and
awareness, management review, organizational decision process, audits, annual reports, security
measures, and emergency plans (6-9). The general process for developing an EMS is to document
how tasks with an environmental impact are to be done, complete the tasks as documented, and
check periodically to verify that the tasks are being done as intended and, if not, correct the
problem. Implementation includes obtaining commitment from top management, communicating
the importance of environmental efforts, establishing environmental policy and objectives,
assessing current impacts, developing a plan for improvement, assigning responsibilities, recording
achievements, auditing results, and reviewing the system.
Given this wide consensus in content and implementation, one might expect that all EMS would be
similar, regardless of the type of the organization. To test this claim, we investigated the EMSs of
nine companies via structured and open-question interviews, a likert-scale survey, facility visits, and
a review of publicly available information. The case studies involved both EHS and non-EHS
personnel and inquired about both the EMS and environmental management information systems
(EMIS) used to support EMS. In the area of EMS generally, the research considered the
components and structure of the EMS and the value the EMS had provided to the organization. The
intent of the research was to determine if a consensus EMS existed, and to identify components, if
any, that were unique. One particular interest was the integration of environmental, health, and
safety (EHS) issues at some companies, and the influence it has on EMSs. The research on EMIS
considered the data use and availability within the organization for decision-making. The intent
was to determine if common data were available, if unique data existed, and how the data were
utilized across the firm.
Case Study Methods
Nine companies agreed to participate in the case study research. The companies' names have been
withheld to maintain anonymity. The companies that participated are from a variety of industry
sectors in the United States including leaders in the fields of electronics, transportation, chemicals,
and construction. Eight of the nine companies fall within the North American Industry
Classification System (NAICS) manufacturing sectors (32 and 33). The other company is from the
construction industry (NAICS 23). All companies are multinational corporations with operating
sites around the world, resulting in a complex array of environmental requirements that must be
followed. With a range of employees from 10,000 to 100,000, the range of environmental issues is
wide.
For each company, we conducted the investigation at one representative location, and in some cases
interviewed corporate environmental staff as well. The facilities were typical sites of the firms.
Corporate staff provided a broader picture of the EMS function across the firm. The companies
represent an opportunity sample selected from contacts of the researchers. Selection resulted in
many of the participants being known environmental leaders in their fields, although among the
nine companies environmental performance varies widely. The status of EMSs at each company
varied as well, with some having mature EMSs to some only beginning to implement formal EMSs.
All of the companies that participated had at least a few facilities (ranging from 5% of facilities to
60
-------
100% of facilities) that were ISO 14001-certified. Six of the nine facilities visited were ISO 14001-
certified. Further certification of facilities will depend on the demands of the customers as well as
corporate policy decisions. Seven of the nine companies have integrated the functions of
environment, health, and safety issues. This integration of function influences the make up of the
EMS. Safety and health information was identified by most participants as part of the EMS
information.
Each case study consisted of an initial conversation detailing the project and verifying a company's
willingness and interest in participating, an exchange of information on the company's EMS or
environmental programs, a site visit with a tour of the facility, a structured survey, a set of open-
ended questions, additional questions that arose during the visit, and then follow-up contacts to fill
in data gaps and verify information. The information exchanged by the company with the
researchers on their EMS or environmental programs ranged from entire EMS manuals to corporate
presentations. The information for each case study was augmented with publicly available
environmental information for both the company and the facility available through company
websites, press releases, and government databases. Due to the small sample size, complex
statistical analyses of the data do not produce reliable conclusions. However, qualitative
information and simple statistics were used to summarize the data collected. A summary of the
responses from some questions results in statements on general trends and the beginning of a
contingent typology of EMS and EMIS within corporations.
Comparison of EMS Components
Based on the nine companies that participated, the major components of the EMSs were the same.
The reliance on common components suggests that a "consensus" EMS has been established across
many industries. Each company had a corporate-wide environmental policy recognizing
responsibility in environmental matters (and in health and safety matters in some cases). Policies
pledged various activities, usually compliance with regulations, communication with stakeholders,
and continuous improvement of performance. Often, companies included establishing an EMS as
part of its policy or pledge.
Each company has established goals for its environmental performance to be monitored and
measured by its EMS. Four of the nine companies use the strategic plan of the corporation to
develop strategic environmental goals. Once corporate goals are established, the goals are handed
down to divisions and business units, which in turn pass them down to individual facilities for
achieving them. Seven companies reevaluate goals annually, one semi-annually, and one company
reevaluates goals "as needed" to maintain improvements. Various tools and techniques are used to
identify environmental issues that the EMS monitors. Four of the companies use cross-functional
teams to assess operations. These teams involve environment, health, and safety staff, business
division staff, and research and development personnel. Three of these teams use formal tools - an
assessment matrix, a checklist, and a product characterization process - in their evaluation. Other
companies look only toward their EHS staff to identify issues. Past audit results are the main
source of information, while brainstorming exercises or offering employee incentives are also used
to identify environmental issues.
All of the companies conduct internal environmental audits at each site with at a frequency ranging
from 2 to 5 years. Audits require a few days to two weeks. Often these audits consist of
employees, particularly environmental professionals, from many different facilities who are able to
share their expertise. Some of the companies have developed self-auditing tools which can be used
61
-------
to prepare for audits or do internal audits more frequently. These tools include checklists, electronic
tracking systems, email reminders, and guidance documents. External audits occur as specified by
any certifications.
Each company did have unique characteristics of its EMS. Some companies have developed
extensive process maps or mass balances of their facilities in order to identify all inputs and outputs
as well as environmental impacts that should be covered by the EMS. Some companies have
incorporated a means to track, identify, and prioritize future risks to their business and facilities into
the EMS. Many have produced EMS skeletons to help ease the implementation burden as the EMS
is initiated at different facilities. A few companies have seen a benefit from industry-wide groups
that bring environmental professionals together to share successes and challenges in EMS activities.
Within some industry sectors, e.g., the chemicals sector, sharing of EHS strategy occurs. In other
sectors, each company independently deals with these issues. The EMSs are similar across all
sectors, however. Finally, while most companies have a procedure to determine if capital
expenditures have any environmental consequences as part of the EMS, few have procedures to
involve EHS staff early in product development
EMS Structure
While the organization's EMS had similar components, the management structure of the various
EMSs was different across firms. By structure, we mean how the EMS is arranged and operated
within the organizational layout of the firm. For example, some systems are centralized at the
corporate level while others operate independently at the business unit or facility level. One
company has a corporate certification to ISO 14001 where all manufacturing facilities are covered
and audited for conformance on a rotating basis. Two other companies have a corporate-wide EMS
protocol (not ISO 14001) which covers (or will cover once fully implemented) all facilities. Some
facilities within one of those organizations are certified to ISO 14001, however. The other
companies, while maintaining a central, corporate EHS function, have facility-level environmental
management systems. Many of these facility EMSs conform to the ISO 14001 standard. The
structure of an EMS depends on whether a company repeatedly produces the same product year-to-
year, changes the product continuously, provides a service rather than a product, or has a few large
manufacturing facilities versus many, small manufacturing facilities. These variations affect the
complexity as well as the core of the EMS in terms of personnel, documentation, responsibilities,
and information systems.
Six of the companies shared their EMS manuals with the researchers of this study. Although two of
these were not ISO 14001 certified, all of these manuals aligned with the ISO 14001 suggested
structure and contained detailed information about procedures and responsibilities. Most
environmental personnel commented on the large amount of time and work required to compile
these manuals, but they also acknowledged the benefits in employee training, knowledge transfer,
and overall organization when complete. All of the companies identified customers or suppliers as
influences on their EMS. Several facilities within the organizations in the case study had
implemented EMSs according to the ISO 14001 standard based on customer demands. But this
customer demand had not influenced corporate mandates for EMS development. Individual
facilities took the initiative to implement the standard. Two companies had incorporated product
information (not simply process or operating information) into their EMSs based on requests for
product standards from customers.
62
-------
Value of the EMS
We asked participants to rate the value of their environmental management system in different
areas. Most respondents consider the EMS to provide the most value to the EHS department itself,
rather than outside the department. There could be some bias here as environmental staff answered
the question. Thus, this is how the environmental staff perceives non-environmental staff to view
the EMS. Each facility was asked to evaluate on a scale of 0 to 5 ("not valuable" to "extremely
valuable") how the EMS contributed to the following eight business opportunities as shown in
Table 7. Companies may not perceive the valuation scales exactly the same.
Table 7: Valuation of environmental management systems
EMS Characteristics
Company
A
B
C
D
E
F
G
H
J
average
Improved Environmental
Impact
5
5
5
3
*
5
3
*
5
4.43
Performance Enhancement
of Products or Processes
4
4
5
4
4
4
3
4.00
Managing Regulatory
Requirements
4
5
2
3
4
3
4
3.57
Communicating with
Management
5
3
4
4
2
1
4
3.29
Communicating with
Employees
3
4
4
4
2
4
3
3.43
Financial Savings
3
3
5
2
3
3
3
3.14
Time Savings
4
5
1
1
2
4
3
2.86
Communicating with
External Stakeholders
2
1
3
2
2
2
2.00
Overall value average
3.75
3.75
3.63
2.88
3.00
3.14
3.38
Respondents were asked to rate the value of the EMS with regard to the activities on a scale of 0 to 5 from not
valuable to extremely valuable. * Companies E and H not included because they either do not have formal
EMSs or their EMS is still in the implementation phase and are unable to rate their EMS.
An EMS provides the least value to communicating with external stakeholders. All of the
participants had information regarding their environmental programs on the corporate website, yet
the information did not correspond with components of the EMS. All companies publicized the
environmental policy and commitments to monitoring environmental issues. Only one company
stated their actual EMS goals. Some companies either on the website or within a published report
on environmental or corporate responsibility communicate performance metrics, but these are often
data already public. These data include safety accident data, toxic release inventory releases, and
total waste generation. Other information included status of ISO 14001 certification and
environmental awards presented or earned by the company or individual facilities and personnel.
While this information is beneficial to community stakeholders, it does not reflect the efforts and
activities of the companies' environmental management systems.
Comparison of Environmental Management Information Systems
Information systems are used to manage regulatory requirements, compare facilities to one another,
and monitor time trends. Table 8 lists the types of data available in environmental, health, and
safety information systems at the facilities. Note that the categories do not reflect individual
software tools, only that information of a given type is available via electronic form. Most data are
collected for regulatory purposes; some other data that have internal value are also collected. The
63
-------
most common data in information systems collected by at least two-thirds of the companies include:
air emissions management, injury and illness incident statistics, key performance indicators, non-
conformance statistics, chemical inventory and management, EHS auditing, and notice of violation
tracking. Examples of data collected for non-regulatory use include greenhouse gas emissions,
waste minimization efforts such as recycling, energy consumption, and non-reportable injury and
illness statistics. Most of the companies in the study have adopted corporate-wide environmental
information systems within the past three years.
Data Category
Company
A
B
c
D
E
F
G
H
J
Injury & Illness Incident Statistics
X
X
X
X
X
X
X
X
Air Emissions Management
X
X
X
X
X
X
X
Key Performance Indicators
X
X
X
X
X
X
X
Non-Conformance Statistics
X
X
X
X
X
X
X
Chemical Inventory/Management
X
X
X
X
X
X
EHS Auditing
X
X
X
X
X
X
Notice of Violation Tracking
X
X
X
X
X
X
Waste Management
X
X
X
X
X
X
Computer Based Training for Environmental
X
X
X
X
X
MSDS's - Incoming from vendors
X
X
X
X
X
Spill Tracking and Notification
X
X
X
X
X
Computer Based Training for Health & Safety
X
X
X
X
EHS Documents/Knowledge Base
X
X
X
X
Energy Consumption/ Energy Management
X
X
X
X
ISO 14001 Management System
X
X
X
X
MSDS's - On the Web for Customers
X
X
X
X
Pollution Prevention
X
X
X
X
Toxic Release Inventory
X
X
X
X
MSDS Creation - Outbound for Customers
X
X
X
Other
X
X
X
Regulatory Tracking Calendar
X
X
X
Wastewater Management
X
X
X
EHS Project Management
X
X
Product Liability/Product Stewardship
X
X
Regulatory Interpretation Library
X
X
Stormwater Management
X
X
Toxicology Information
X
X
EHS Cost Analysis
X
Voluntary Program Participant Requirements
able in Company Information Systems
As information systems have become more common-place, companies have begun to migrate to
formal databases or web-based systems. Companies are beginning to establish intranet systems to
accommodate multiple site data requests, such as waste generation from all facilities within a single
business unit. Often these systems utilize workers outside of EHS personnel to input and track data.
However, some data remains segregated in non-networked systems, limiting availability of the data
for decision-making.
64
-------
Companies were asked about the level of integration of their EMIS. The level of integration reflects
the linkage of the environmental management information system with other business information
systems. The level of integration can be determined by the number of other business information
systems that are linked with the environmental information systems. The survey options for
systems that may be integrated included: accounting/ financial, business development, human
resources, inventory control, maintenance, manufacturing, and purchasing. Each of these systems
can be considered a level of integration. Few of the case study companies had integrated the EMS
data with more than 2 other business information systems. Most commonly integrated was
inventory control, reflecting a need to monitor chemicals.
Each company was asked how frequently environmental data are requested with the following
seven choices: daily, weekly, monthly, quarterly, annually, infrequently, never as well as the option
of "all that apply." The question was answered for three parts of the company: management outside
the EHS function at the facility, EHS management at the corporate level, and management outside
the EHS function at the corporate level. Table 9 shows the frequency of data requested by each
company. Management within the EHS divisions at the corporate level usually requests
information more frequently than management within the non-EHS divisions of the facility and
always requests information more frequently than management within the non-EHS divisions of
management at the corporate level.
Table 9: Data Requests by Management
A
B
c
E
F
G
H
J
Management within the EHS Function of the
Company
Daily
X
*
Weekly
X
X
X
Monthly
X
X
X
Quarterly
X
X
Annually
X
X
Infrequently
Management outside the EHS function of the
Company
Daily
X
Weekly
Monthly
X
X
X
Quarterly
X
X
X
X
X
Annually
X
X
Infrequently
Management outside the EHS function of the Facility
X
Daily
X
Weekly
X
X
X
Monthly
X
X
X
X
X
X
Quarterly
X
X
Annually
X
X
X
Infrequently
X
* Company D did not answer this question because they said that using the data is an ongoing process and
that the frequency of data requests varies. Company H did not answer the question for EHS staff because
they continually have access to this data - so they do not ever have to request it
65
-------
An intensity value was assigned to the frequency from infrequently (1) to daily (6). Table 10 shows
the average data request frequency across companies. Management within the EHS divisions at the
corporate level requests data weekly (average of 4.75), management within the non-EHS divisions
of the facility company requests data more than once a month (average of 4.1), and non-EHS
divisions of management at the corporate level requests data more than once a quarter (average of
3.1). The total intensity index averages the intensity value for each level of management per
company. Table 10 also shows the total intensity index by company. Across levels of management,
Company A requests data most frequently, and Company G requests data the least frequently.
Table 10: Data Request Frequency and Intensity
A
B
C
E
F
G
H
J
average
Management within the EHS
Function of the Company
6
5
3
4
5
5
*
4
4.75
Management outside the EHS
function of the Company
4
3
3
3
4
1
4
3
3.1
Management outside the EHS
function of the Facility
5
4
4
5
4
2
4
5
4.1
Total Intensity Index
5
4
3.33
4
4.33
2.67
4
4
The total intensity index averages the intensity value for each level of management per company.
In addition to the frequency of data requested, there are differences in which types of data are
requested and which parts of the company request the data. In at least 6 of the 9 companies,
management within the EHS divisions of the company request data from: waste management, injury
and illness incident statistics, air emissions management, chemical inventory/ management, EHS
auditing, and MSDS's - incoming from vendors. The only information requested by more than half
of the management with the non-EHS divisions of the company was the injury and illness incident
statistics, which technically is not even an environmental item. The next most requested items (by 4
of the 9 companies) included waste management and EHS auditing. Management within non-EHS
divisions of the facility most commonly (by 5 of 9 companies) requests information on waste
management and injury and illness incident statistics. The interest in these two items probably
occurs because both are easy to define, easy to measure, and linked directly to costs.
Value of EMIS
The EHS information systems are considered to be valuable in many areas. Each facility was
asked to evaluate (on a scale of 0 to 5 from "not valuable" to "extremely valuable") how the EHS
information system contributed to eight business opportunities. The valuations by each company
are listed in Table 11. Companies may not perceive the valuation scales exactly the same. An
average was taken across the nine companies to prioritize the areas. From the averages, these
companies receive the most value from information systems in managing regulatory requirements
and improving environmental impact. Note that the range of value for the categories is lower than
the valuation given previously for EMSs. More value is perceived from the overall management
system than from the information system used in conjunction with the EMS. Communicating with
external stakeholders remains the lowest category for gaining value.
66
-------
Table 11: Valuation of environmental information systems
A
B
C
D
E
F
G
H
J
average
Improved Environmental
Impact
5
3
5
3
3
3
3
5
3
3.67
Managing Regulatory
Requirements
5
2
4
4
3
2
3
5
5
3.67
Communicating with
Management
4
2
5
3
3
3
3
5
3
3.44
Communicating with
Employees
3
2
5
3
3
4
2
4
3
3.22
Time Savings
4
3
1
1
1
5
4
5
5
3.22
Performance Enhancement
of Products or Processes
4
3
3
3
0
4
3
3
2
2.78
Financial Savings
3
3
2
1
1
3
4
3
2
2.44
Communicating with
External Stakeholders
2
1
5
2
0
1
3
4
1
2.11
Overall value average
3.75
2.38
3.75
2.5
1.75
3.13
3.13
4.25
3.0
Respondents were asked to rate the value of the EHS information system with regard to the activities on a scale of 0
to 5 from not valuable to extremely valuable.
Discussion
The micro-level analysis demonstrates the commonalities among EMS and information systems in
various organizations. The components within an EMS typically relate to existing regulatory
requirements. Thus, EMSs are likely to be helpful in identifying compliance issues as these issues
are targeted for action and personnel can access data readily. Each of the companies participating in
the case studies had violations against environmental permits and had received enforcement action
for environmental compliance issues in the past two years (10).
Given the commonalities in environmental management systems across corporations of various
industries, sizes, and locations, managers can take advantage of the experience from individual
facilities to build up new EMS or strengthen existing ones. The unique collaboration among
environmental professionals that some of the participants had initiated is an aspect that facilitates
learning.
In this study, management within the EHS divisions at the corporate level usually requests
information more frequently than management outside the EHS function of the facility and always
requests information more frequently than the management outside the EHS function at the
corporate level. These results indicate that environmental data may not be utilized throughout all
areas of an organization. Business managers should realize that decision-makers in all areas should
be able to access environmental data. One potential aid to exchange is integration of the EMS
information into other management information systems. Few companies had links between the
environmental data and the data in the traditional business systems, although data from these
systems is crucial to decisions. Without this integration, the EMSs are limited in how they might
assist decision makers.
Respondents in the study evaluated where environmental information systems contributed to
business opportunities. Environmental information systems provide the most value in improving
environmental impact and managing regulatory requirements. The environmental information
systems provide the least value in communicating with external stakeholders. Similarly,
67
-------
environmental management systems provided the most value in improved environmental impact
and performance enhancement of products or processes. Like the environmental information
systems, the environmental management systems provided the least value in communicating with
external stakeholders.
Implications for Policy Makers, Organizations, and the Public
The results of the analyses have important implications for policy makers, organizational managers,
and the public. From the macro-level analysis, certification of an EMS to the ISO 14001 standard
cannot be used as a proxy for improved environmental performance, and possibly more importantly,
cannot be used as a proxy for regulatory compliance. As the data indicate, facilities with a certified
EMS continue to struggle with violations and noncompliance of environmental regulations.
Regulators cannot assume that presence of a certified EMS guarantees against infringement and
resulting environmental impacts. Since no difference in operating performance is evident, policy
makers must not consider EMS certification to be a means to an end. Continued reductions in
currently regulated wastes and emissions are not assured. Any policy which gives regulatory
preference to a facility on the basis of implementing and certifying an EMS to ISO 14001 should be
considered only with additional transparency of the underlying goals and targets of the EMS.
Additional transparency allows for a more robust analysis of efforts and accomplishments made by
facilities in relation to the EMS.
Likewise, from the micro-level analysis, both EMS and EMIS revolve around regulatory
compliance issues, the systems may be less useful in allowing companies to go beyond compliance.
Unlike other voluntary programs intended to improve environmental performance by targeting
issues outside the regulations, EMSs may only assure that firms are aware and improving areas
already under scrutiny. Without additional data reporting (either internally to a firm, or externally)
on non-regulated issues or integration with non-environmental data, using the systems continues to
address only impacts with compliance aspects. At the same time, use of the data outside EHS
management is also essential in having EMS become part of the decision-making processes in an
organization.
One must consider, however, if how a facility is dealing with environmental issues is more
important that what the performance is. The ISO 14001 EMS structure emphasizes continuity and
consistency in addressing environmental impacts. As noted above, do the management aspects of
the standard have benefits outside of general environmental performance over time which serve to
improve operations? Benefits such as codifying responsibilities, establishing protocols, increasing
awareness of environmental issues, and documenting these items, albeit cumbersome, may be
shown to improve management of environmental issues overall. These factors may benefit
organizations and encourage continued use and implementation of EMSs.
But since improving environmental performance in ways that are clearly identifiable to policy
makers, local citizens, and the general public is the main goal of regulation, then the existing ISO
14001 standard is not sufficient. The fact that a consensus EMS exists can support regulators as
policies and programs develop around the concept of EMSs. Regulators can determine which
elements are essential for improving performance and concentrate on strengthening those
components within facilities, and focusing on those elements during audits. Assistance programs
can be generalized to promote further adoptions of EMS that allows for the company to have
flexibility in establishing a corporate-level or facility-level EMS.
68
-------
The results of the two analyses, along with discussion from a workshop on environmental
management for multinational corporations attended by corporate-level environmental, health, and
safety executives, led to the development of five essential elements of an EMS to aid environmental
decision-making. The five elements, shown in Table 12, provide decision makers with key
information on how environmental issues influence day-to-day and long-term operations of the
firm. The scope of issues confronted on a daily basis in multinational firms requires a broad EMS
that captures more than just regulatory requirements.
Table
2: Five elements for environmental management systems to aid organizational decision-making.
1. Process diagrams identifying material and energy inputs and outputs
2. Quantifiable goals for both short- and long-term performance consistent with the
organization's strategic plan
3. Reliable methods of collecting and disseminating environmental data
4. Risk assessment tools for current and emerging environmental issues for operations and
products
5. Collaboration and education of environmental personnel both within the organization and
outside
These EMS elements are not universal, even among ISO 14001 certified EMS. Other than
establishing goals and targets (element 2), these elements are not required for ISO 14001
certification, although at various levels a certified system may have these attributes. The five
elements do not focus on regulatory issues or compliance. Most organizations (with or without an
ISO-certified EMS) have existing systems to address regulatory requirements. Instead, the elements
center on the goal of the EMS to provide timely, relevant information for decision-making on
environmental issues that may occur across the organization.
The overall results of the research lead to suggestions for improving environmental management
systems for organizational decision-making and policy development. First, EMS must stretch
beyond the current regulatory issues to be effective in long-term improvement. Unlike other
voluntary programs, EMS do not push organizations to concentrate efforts on addressing impacts
outside the regulatory issues. Second, EMS goals, targets, and resulting performance must be made
more transparent. The link between efforts to reduce environmental problems via an EMS and
reported environmental performance metrics is tenuous. Finally, as environmental issues continue
to shift in importance, EMS and must adapt to changing organizational focus and monitor potential
future regulatory issues. For example, corporate social responsibility and calls for sustainability
(incorporating social and economic factors, as well as environmental factors) and the global
importance of carbon emissions (currently unregulated) are influencing corporate strategy. An
EMS should support decision-making on these pressing issues.
69
-------
References
(1) ISO World. ISO 14001 Registered Companies.
http://www.ecology.or.jp/isoworld/english/analyl4k.htm (Mar 24, 2004).
(2) Ford Becomes First U.S. Automaker to Require Suppliers To Achieve ISO 14001
Certification http://www.p2pays.org/ref/ll/10987.htm (3/15/03).
(3) General Motors Sets New Level of Environmental Performance for Suppliers.
http://www.p2pays.org/ref/ll/10988.htm (3/15/03).
(4) WorldPreferred.com Inc. World Preferred Registry, http://
www.worldpreferred.com (Mar 3, 2004).
(5) International Organization for Standardization. Environmental Management Systems
- Specification with Guidance for Use; Reference number ISO 14001:1996(E) ISO Central
Secretariat: Geneva, 1996.
(6) Ilinitch, A. Y.; Soderstrom, N. S.; Thomas, T. E. J. Accounting Public Policy. 1998,
17, 383-408.
(7) Cascio, J. The ISO 14000 Handbook, Port City Press: Baltimore, Maryland, 1996.
(8) Matthews, D. H. Ph.D. Dissertation, Carnegie Mellon University, Pittsburgh,
Pennsylvania, 2001.
(9) Stapleton, P. J.; Glover, M. A.; Davis S. P. Environmental Management Systems: An
Implementation Guide for Small and Medium-Sized Organizations. Technical Report, NSF
International: Ann Arbor, Michigan, 2001.
(10) U.S. EPA. Enforcement and Compliance History Online (ECHO).
http://www.epa.gov/echo/ (December 17, 2003.
70
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
Formalized Environmental Management Procedures: What Drives Performance
Improvements? Evidence from Four U.S. Industries
Andrew M. Hutson
Daniel Edwards, Jr.
Richard N. L. Andrews1
ABSTRACT
This paper examines changes in environmental performance and management benefits associated
with the introduction of environmental management systems (EMSs), and factors influencing
these outcomes. Specifically, we sought to determine whether there are systematic differences, in
EMSs themselves and in resulting environmental performance, between organizations that adopt
EMSs for their own organizational reasons ("self-initiated"), or under coercion from corporate or
customer mandates, and those that have not adopted such systems at all. Data included a survey
of 3,200 plant managers in four sectors that include many suppliers to the automotive industry,
which has mandated EMS adoption by its subsidiaries and suppliers, plus data from EPA's
IDEAS database. The results suggest important findings concerning the roles of specific
objectives for performance improvement, as opposed to adoption of an EMS per se; the limited
effects of business-to-business EMS mandates; the perceived benefits of environmental
performance improvement to business objectives; and the continued importance of governmental
regulation and inspection to environmental performance on some key indicators.
Keywords: EMS, environmental management systems, ISO 14000, compliance, pollution
prevention, eco-efficiency, product stewardship, environmental performance, voluntary
standards
Introduction
The adoption of formalized sets of environmental management practices by manufacturing
facilities has proliferated over the past decade, but the efficacy of such practices in promoting
environmental performance improvements remains uncertain. Over 50,000 organizations
worldwide, including approximately 3,000 in the United States, have certified to the ISO 14001
environmental management system (EMS) standard, and more are currently in the process of
registration.2 Many other businesses have adopted formalized sets of environmental management
practices, but have not officially registered to the ISO 14001 standard. While some may have
adopted EMSs equivalent to the ISO 14001 standard and chosen not to formally register, others
have adopted systems which fall short of ISO requirements. Nonetheless, efforts to install
1 Corresponding author. Address: Department of Public Policy, CB#3435, Abernethy Hall, University of North
Carolina at Chapel Hill, Chapel Hill, NC 27599-3435. Tel. 919-843-5011, Fax 919-962-5824. Emails:
hutson@unc.edu. deg@email.unc.edu. pete andrews@unc.edu
2 See http://www.ecolouv.or.ip/isoworld/cimlish/analv 14k.htm. Note that some ISO 14001 certificates may cover
multiple facilities of the same parent organization, while others may cover only a specified subset of functions even
at a single facility site.
71
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
systematic procedures for environmental management are becoming increasingly common, and
in some cases a de facto condition for operating in certain industries and markets.
It is plausible to expect that the actual efficacy of an EMS in promoting environmental
performance improvements may vary depending on the motivation of the business in introducing
it. An EMS might be introduced, for instance, to improve compliance with environmental
regulations, or to improve pollution prevention and increase "eco-efficiency" in the use of
materials and energy in production processes (Coglianese and Nash 2001), or to improve the
environmental performance of a company's products all the way from raw materials through
recycling or waste management ("product stewardship") (Gallagher 2002). Alternatively, it
might be introduced with an eye to improving overall plant management, whether or not it
achieved significant improvements in environmental performance per se (Florida and Davison
2002). Or it might be introduced simply as a paper process, either to promote the business's
"green" image (Darnall 2002) or merely to satisfy a mandate from corporate headquarters or a
major customer.
Many facilities are increasingly under pressure from corporate parents and major customers to
adopt formal environmental management practices. A number of major businesses have recently
mandated introduction of environmental management systems (EMSs) by their subsidiaries and
suppliers, particularly in the automotive and electronics industries: prominent among these are
recent mandates from Ford and General Motors that all company facilities, as well as facilities of
their first tier suppliers, must adopt and register an EMS in conformance with the ISO 14001
standard (Hutson 2001). Government agencies also have begun to promote such systems with
public recognition and incentives, such as EPA's National Performance Track and similar state-
level initiatives.3
A key unanswered question is what differences in actual environmental performance are
associated with the introduction of such systems, and particularly, whether such systems produce
positive changes in performance and other benefits when they are mandated or encouraged by
external incentives.
The objective of this research project was to determine what changes in environmental
performance result from the implementation of environmental management systems (EMSs), and
what differences in organizational characteristics, motivations, and decision making are
associated with these changes. Specifically, we sought to determine whether there are systematic
differences, in EMSs themselves and in resulting environmental performance, between
organizations that adopt EMSs for their own organizational reasons ("self-initiated"), or under
coercion from corporate or customer mandates, and those that have not adopted such systems at
all. Both public policymakers and businesses themselves will benefit from better information on
the consequences of EMSs for environmental performance, and on their associated benefits and
costs.
The research addressed a series of more specific questions concerning the impact of EMS
adoption:
3 http://www.eoa.gov/perforniancettack/. and state initiatives e.g. in Colorado, Illinois, Indiana, Maine, North
Carolina, Oregon, Texas, Virginia, Wisconsin and others.
72
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
• First, is the adoption of a formal environmental management system a good predictor of
environmental performance improvement? Do facilities that have introduced ISO-
certified EMSs - or comparably formalized EMSs - improve their environmental
performance more than other facilities in the same industrial sector?
• Second, if so, are some environmental performance indicators (EPIs) more likely to
improve than others? Do they improve across the board, or merely in some more limited
set of performance indicators - and if the latter, what lessons might be drawn for
understanding the strengths and limitations of EMS impacts on environmental
performance?
• Third, are such improvements associated with EMS adoption per se, or with more
specific characteristics of the EMS such as the particular types of goals and performance
objectives adopted? One required characteristic of an ISO-equivalent EMS is the
adoption of explicit objectives for performance improvement, but the nature and
stringency of these objectives is left to the discretion of the adopter. Do facilities
experience greater environmental performance improvement for outcomes that are
targeted by their EMS performance objectives?
• Fourth, do facilities that adopt a formalized set of environmental management practices
reap business benefits as well? That is, do they experience improved management
efficiencies, improved positioning in the market place, or other business benefits in
addition to (or even independent of) their actual environmental performance changes?
• Finally, do facilities that are subject to explicit requirements from corporate parents or
customers to adopt environmental management practices perform differently than
facilities that are not subject to such pressures, and that presumably therefore implement
environmental management practices - whatever ones they do implement - under their
own initiative?
Data sources included a survey of plant managers from a stratified random sample of facilities in
four industrial sectors, as well as publicly available data for those facilities from EPA's IDEAS
regulatory compliance database and its Toxics Release Inventory.4
FORMALIZED ENVIRONMENTAL MANAGEMENT SYSTEMS AND BENEFITS
To answer these questions, one must consider three bodies of theory and evidence: what impacts
does an EMS have on environmental performance, what impacts might one expect it to have on
business outcomes such as management and market benefits, and what expectations of business
benefits might explain the imposition of business-to-business mandates for EMS adoption on
subsidiaries and suppliers?
Environmental Performance
To date, studies that have attempted to determine the effects of EMSs on environmental
performance have shed some light on the issue, but have not produced systematic or consistent
answers. Case studies of facilities that have adopted EMSs point to a myriad of environmental
improvements associated with the management systems, but have not produced systematic or
reliable results (Berry and Rondinelli 2000; Rondinelli and Vastag 2000; Ammenberg 2001).
4 Analysis of the TRI data is still in progress.
73
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
Several survey-based studies also have reported the impacts of EMSs on environmental
performance, some too early after ISO 14001 adoption (in 1996) to produce compelling results
(Melnyk et al. 1999; Hamschmidt 2000), and others with improved reliability but limited
generalizability (Florida and Davison 2001; Mohammed 2000; Anton 2002; Andrews et al.
2003). Two studies have sought to measure performance outcomes using EPA's Toxics Release
Inventory (TRI) database, with conflicting results: Matthews (2001) found that facilities in the
auto industry with EMSs did not perform better than those without them, while Russo (2001)
found that ISO 14001 was a significant predictor of reduced toxic emissions in the electronics
industry. The findings of these latter studies may have differed due to differing methodological
approaches and/or sectoral variance. In any case, the results focused only on toxic emissions
rather than on a broader array of environmental performance indicators.
Additionally, most existing studies do not address the great range of discretionary variation that
exists among the EMSs that facilities adopt. ISO 14001 provides a standard template for the
process elements of an EMS - identifying environmental aspects and impacts, setting goals and
objectives, assigning responsibilities, training, corrective and preventive actions, periodic review,
and so forth - and it specifies three overarching goals for that process (compliance, pollution
prevention, and continual improvement). But all the specific content - what environmental
aspects and impacts will be considered, what environmental performance objectives will be
priorities and how rapidly they are to be achieved, and others - are left entirely to the choices of
the adopter, and businesses that do not seek ISO 14001 certification are not bound even to the
ISO template. One previous study found great variation in practice both among aspects and
impacts considered, and in the determination of which of these impacts were considered
significant and targeted as priorities for improvement (Andrews et al. 2003). An important
question for further inquiry is what impact the facility's choice of objectives - not just its
decision to adopt a formal EMS - has on environmental performance and other outcomes.
Management and Market Benefits
The implementation of a formal EMS can be an expensive and time consuming endeavor
(Darnall and Edwards forthcoming). Several theoretical justifications have been offered as to
why an EMS might produce benefits both to environmental performance and to business
outcomes. First, systematic management of functions with negative environmental consequences,
rather than haphazard and inconsistent methods for addressing them, is more likely to produce
outcomes that have both environmental and business benefits, such as minimizing costs,
environmental liabilities, regulatory penalties, and risks to the firm's image and associated brand
value (Coglianese and Nash 2001). Second, managers who address environmental problems with
formal management systems may also reap improvements in product quality and process
efficiencies that lead to positive financial outcomes (see also Porter and van der Linde 1995;
Hart and Ahuja 1996; Klassen and McLaughlin 1996; Sharma and Vredenburg 1998; Dowell,
Hart and Yeung 2000; Christmann 2000).
Third, an EMS can contribute to broader patterns of beneficial management and cultural changes
within a business organization, such as integrating environmental management with other
primary management functions and with organization-wide quality management procedures
(Florida and Davison 2001, Coglianese and Nash 2001). Additionally, management-based
approaches may be less costly and more effective than government imposed regulation, may lead
74
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
to greater buy-in from management due to a greater sense of legitimacy, ownership and control,
and may promote innovation and social learning (Coglianese and Lazar 2003). However, some
suggest that claims about the ability of EMS to lead to performance improvements may be
overstated, and must be empirically tested (Walley and Whitehead 1994).
Business-to-Business Mandates
It is becoming increasingly common for business decision makers to include comprehensive
environmental management plans as part of a broader corporate strategy. Such strategies may
include strong encouragement, as in the case of IBM, or explicit requirements, such as those
mandated by Ford, General Motors and others, for subsidiaries and suppliers to adopt formal
environmental management systems.
A corporation may require that its own facilities adopt EMSs for at least three reasons (Andrews,
Hutson, and Edwards, forthcoming). One is to minimize legal and financial liabilities associated
with poor environmental performance by its subsidiaries: through more explicit procedures for
environmental management and associated accountability, firms may prevent accidents and
better understand the potential legal risks they face. A firm may be harmed by poor
environmental behavior of its subsidiaries and may therefore seek a unified strategy to reduce
current and potential risks. Second, a company may mandate EMSs by its subsidiaries in order to
protect or improve its image, reputation, and brand value. Adoption and certification of EMSs
may be a means for presenting an image to the external world of the company's commitment to
good environmental management practices, whether or not this image in fact represents better
performance than that of other comparable firms (Darnall 2002). Third, the increasingly
widespread geographic distribution of manufacturing sites, which has coincided with the
globalization of manufacturing, has increased the need for greater standardization of practices.
Firms that use standard operating procedures throughout their global operations - including
standardization of environmental management practices - may improve both communication and
overall efficiency, both of which may improve financial performance by reducing costs and
minimizing waste.
Firms may extend such requirements to suppliers and business partners for similar reasons. In an
increasingly global economy, where firms often subcontract manufacturing functions to a
geographically disperse network of suppliers, firms face a host of challenges. First, corporations
in some sectors face shared threats to their reputations, which are addressed by creating sector-
wide standards to which members must adhere (Kollman and Prakash 2002), or "lead industry
regulation" to influence the practices of suppliers or customers whose behavior may affect their
own reputations or liabilities (Nash 2002). Second, large brand-visible firms may choose to have
their more anonymous suppliers adopt formal sets of practices and/or codes of conduct in an
effort to protect their brand image and reputation from harm caused by potential environmental
or human rights abuses down the value chain (Gereffi et al. 2001, Klein 1999). Finally, the
adoption of standardized practices by suppliers may result in increased efficiencies, cost
reductions, and even innovations whose benefits which can then be shared with lead firms
(Geffen and Rothenberg 2000, Corbett 2002). In essence, firms seek to reap the same benefits
from their suppliers as they expect from their own subsidiaries.
75
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
How effectively those who are subject to requirements from corporate parents and customers
respond to mandates is still untested. While some evidence suggests that mandates are
motivating suppliers to adopt formal EMSs, many for the first time (Hutson 2001), no systematic
studies have been conducted on mandates to adopt EMS, nor on the resulting effectiveness of
such mandates in inducing environmental performance improvements. The possibility exists that
EMSs, which are touted as efforts to "regulate from the inside," may in practice be perceived
more like traditional regulatory mandates by governments (which "regulate from the outside")
when they are imposed on suppliers and business partners. Even though facilities can choose
how to adopt the systems, and are not constrained by strict performance targets, they may choose
to adopt the system simply as a paperwork burden, or do so in the most limited form necessary to
meet the requirements. Alternatively, it is also possible that facilities which adopt environmental
management practices under pressure may do a more thorough job of implementation, as they
perceive that doing so is a precondition of their contractual relationships; or that having initially
adopted the EMS only because of a mandate, they may subsequently discover that it has
unanticipated benefits to them.
Data and Measures
Data
The data used in this analysis were collected through a survey of plant managers from a random
sample of manufacturing facilities in four U.S. industrial sectors: Motor Vehicles Parts and
Accessories (SIC 3714), Chemicals and Chemical Preparations (SIC 2899), Plastic Products
(SIC 3089) and Coating, Engraving, and Allied Services (SIC 3479). These sectors were chosen
to include a high number of certified environmental management systems, strong supplier
relationships to the automotive sector (which has mandated supplier EMSs), and significant
environmental impacts based on EPA's Toxics Release Inventory (TRI) data. The sample
included facilities that had adopted EMSs with and without external pressures or incentives to do
so, as well as controls that had not adopted formal EMSs at all. From our original sample we
discarded approximately 500 due to facility closures, re-location and incorrect mailing addresses,
and sent the survey to plant managers of over 3,200 facilities. From these we received 617
responses, a response rate of 20%, well distributed among the industries sampled.6 For each
facility, the survey requested information on current environmental management objectives and
activities, specific motivations or requirements to develop an EMS, and changes in
environmental performance indicators (EPIs) and other benefits that the respondents had
observed as a result of environmental management activities. Plant managers were the target
respondents.
Dependent Variables
The study used two primary groups of dependent variables, as reported by the respondents:
changes in seventeen environmental performance indicators, and ten categories of management
and market benefits (figure 1). Facility responses indicated whether environmental indicators
increased, decreased or were unchanged during the past three years, corresponding roughly to
6 A response rate of 19.56% was achieved after accounting for facility closures, re-location and incorrect mailing
addresses.
76
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
2000 thai 2002.7 The respondents similarly indicated the significance of management and market
benefits realized due to the adoption of environmental management activities, on a scale ranging
from 'no benefits' to 'high benefits.' 8
Figure 1: EPI and Management Benefits Examined
EPIs
Management Benefits
• Water use
• Cost savings in terms of inputs or taxes
• Energy use
• Avoidance of non-compliance penalties
• Recycled inputs
• Increase in productivity
• Recycling of waste
• Increase in market share
• Chemical inputs per unit output
• Ability to reach new markets
• Total material inputs
• Product differentiation
• Hazardous waste generation
• Improved company/plant image
• Non-hazardous solid waste generation
• Improved access to capital markets
• Wastewater effluent
• Improved competitive advantage
• Air pollution emissions
• Improved management efficiencies
• Greenhouse gas emissions
• Noise generation
• Smell generation
• Disruption of the natural landscape
• Soil contamination
• Severe leaks or spills
• Legal violations or potential violations
Independent Variables
Three primary independent variables were used to explain environmental performance outcomes
and management benefits: the degree to which environmental practices and activities were
formalized at the facility level, the relative priorities each facility placed on particular objectives
for their environmental practices, and the existence of external market pressures to adopt an
environmental management system. (For a detailed description of measures see Appendix I).
Control Variables
The study also included two types of control variables in the analysis to account for other sources
of variability in performance outcomes. The first included endogenous resources and capabilities
that might influence environmental performance outcomes, such as prior management systems or
cultural norms of the organization. The second category included exogenous factors, such as
industrial sector and regulatory pressures, which have the potential to alter environmental results
at the facility level. (For detail on control variables see Appendix II).
7 Facilities were given the option to indicate that the listed indicator either was not tracked by the facility or was not
applicable to site operations.
8 Additional tests using TRI data as dependent variables measuring environmental performance change are still in
progress.
77
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
Methods and Results
Models
This study used a binomial logistic regression model to investigate the effects of the variables of
interest on resulting environmental performance and management benefits. This model applies a
maximum likelihood estimation technique to estimate the likelihood of a certain outcome.
The model takes the linear form:
z = a + p1X1 + p2X2+....+ pkXk
While the parameter estimates produced by this method are not necessarily intuitively
meaningful, the regression coefficients can be transformed to show either the percent change in
the odds of the outcome's occurrence per a one unit change in the independent variable or the
percent change in probability in the outcome's occurrence at the margin for the mean probability
given a one unit change in the independent variable. Both transformations are useful in
understanding the effects of our independent variables on environmental performance and
management benefit outcomes. However, the former transformation allows us to generalize the
impact of the independent variables on the likelihood that improved environmental performance
is reported without assumptions about mean probability that such improvements will occur. In
contrast, the latter transformation is dependent on the sample mean for each dependent variable
of interest.
For the seventeen EPIs and ten management benefits as well as a comparison of overall benefit
or cost of environmental practices, the models take the following general form:9
Log p(cpi/)= a + Pi (EMS formalization) + P2(system objectives)! + P3(business
mandate)! + P^attitude) + P5(involvement) + Ps(time) + Pe(sizc) + P7(industry)i +
P8(ownership) + P^resources) + P^(capabilities), + Pi [(location), + Pi2(regulatory
pressure);
Log p(bcncfit/,)= a + Pi(EMS formalization) + P2(system objectives)! + P3(business
mandate)i + P^attitude) + P5(involvement) + Ps(time) + Pe(size) + PyOndustrv), +
P8(ownership) + P^resources) + Pio(capabilities) + Pi ^environmental performance)!
The model results for each of the seventeen (17) self-reported EPIs are shown in Tables 1 and
2.10 The model results for each of the ten management benefits are shown in Tables 3 and 4. The
Wald chi-square statistic is presented for each model along with a Max-rescaled R-Square
statistic. The parameter coefficient is reported along with a point estimate of its effect on the
odds that improved environmental performance was reported along with the standard error of the
parameter coefficient.
9 The model presented below is a generalized model. Slightly different specifications were used for some classes of
management benefits. See Tables 3 and 4 as well as Figure 5 in Appendix II for more detail.
10 The models for four EPIs - greenhouse gas emissions, disruption of the natural landscape, soil contamination and
noise generation - did not produce statistically significant results and are not included in the referenced tables.
78
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
Table 1: Logistic Regression Results for Self-Reported Environmental Performance Changes: Regulated and Quasi-
Regulated Indicators
EPI
Model
Hazardous Waste
Air Pollution
Wastewtr Effluent
Spills
V io latio n s
Chi Square
50.84**
52.65**
48.37**
47.66**
46.70**
R 2
0.19
0.27
0.19
0.25
0.16
n
429
348
393
344
376
Param ete r
E stim ate
Point
E stim ate
S.E.
Pa ram ete r
E stim ate
Point
E stim ate
S.E.
Param eter
E stim ate
Point
E stim ate
S.E.
Param eter
E stim ate
Point
E stim ate
S.E.
Pa ra m ete r
E stim ate
Point
E stim ate
S.E.
Variables of Interest
EMS Practices Formalization
Customer Mandate
Corporate M andate
Both B2B M a ndates
Product Stewardship Goals
Endogenous Controls
Quality Management Systems
Time with Formal EMS
Exogenous Controls
Auto Supply Sector
Plastics Sector
Coatings Sector
Private Ownership
Recent lnspections +
Recent Non-Com pliance +
Recent Fines +
Region 1
Region 2
Region 3
R egion 4
Region 5
Region 7
Region 8
Region 9
Region 1 0
Facility Size +
*p < 0.10, ** p < 0.05, *** p < 0.01, + see Appendix II for discussion of variable construction
79
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
Table 2: Regression Results for Self-Reported Environmental Performance Changes: Non-Regulated Indicators
EPI
Model
Water
Energy
Recycled Inputs
Recycled Waste
Chemical Use
Material Use
Smell
Non-Hazardous Waste
Chi Square
58.85***
52.13**
61.56***
57.97***
44.17*
47.78*
46.25**
59.12***
R2
0.22
0.17
0.27
0.20
0.17
0.16
0.28
0.20
n
445
462
382
449
399
415
279
465
Parameter
Estimate
Point
Estimate
S.E.
Parameter
Estimate
Point
Estimate
S.E.
Parameter
Estimate
Point
Estimate
S.E.
Parameter
Estimate
Point
Estimate
S.E.
Parameter
Estimate
Point
Estimate
S.E.
Paramete
r Estimate
Point
Estimate
S.E.
Paramete
r Estimate
Point
Estimate
S.E.
Parameter
Estimate
Point
Estimate
S.E.
Variables of Interest
EMS Practices Formalization
0.36"
1.43
0.16
0.34"
1.40
0.15
Customer Mandate
Corporate Mandate
-0.61*
0.54
0.36
Both B2B Mandates
-0.35"
0.43
0.35
Compliance Goals
-0.37"
0.69
0.17
-0.31''
0.73
0.15
0.44*
1.55
0.26
-0.36-
0.70
0.14
Pollution Prevention Goals
Eco-Efficiency Goals
0.34"
1.41
0.11
0.40***
1.49
0.13
0.32'"
1.38
0.12
0.22'
1.25
0.13
0.29"""
1.34
0.11
Product Stewardship Goals
0.31**
1.37
0.16
Bidogenous Controls
Other Hans
0.78*
2.17
0.37
I.I7-
3.22
0.45
1.12
3.07
0.36
0.70*
2.02
0.37
Quality Management Systems
0.63**
1.88
0.33
0.57'
1.77
0.31
0.60'
1.82
0.32
Attitude Toward Env. Mgmt
Employee Involvement
0.10*
1.11
0.06
0.I6"
1.17
0.06
Time with Formal EMS
0.17*
1.19
0.10
Exogenous Controls
Auto Supply Sector
0.81"
p pc
0.36
0.81 "
2.25
0.35
-0.73"
0.46
0.46
l. 12"""
3.06
0.37
I.34"""
3.82
0.39
-I.08"
0.34
0.52
Plastics Sector
-0.57*
0.56
0.35
-0.79"
0.46
0.37
Coatings Sector
-I.00-"
0.36
0.36
0.92'"
2.51
0.36
-0.71"
0.49
0.32
Private Ownership
Recent lnspections+
0.26"
1.01
0.01
0.04*
1.04
0.02
Recent Non-Compliance+
0.04***
1.04
0.01
Recent Fines+
0.00*
1.00
0.00
Region 1
I.22'
3.39
0.66
Region 2
1.44**
4.21
0.63
Region 3
Region 4
I.03"
2.81
0.48
0.98""
2.67
0.43
-0.82-
0.44
0.47
Region 5
0.97**
2.56
0.43
0.67'
1.96
0.40
Region 7
Region 8
Region 9
0.68"
2.40
0.52
Region 10
1.62"
5.07
0.77
I.96"
7.08
1.16
1.34""
6.30
0.77
Facility Size+
0.39'"
1.48 ,
0.11
-0.26"
0.77
0.11
0.33"
1.39
0.14
*p < 0.10, ** p < 0.05, *** p < 0.01, + see Appendix II for discussion of variable construction
80
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
Table 3: Regression Results for Self-Reported Internal Management Benefits
li
Management Benefits
Model
Cost-Benefit
Inputs
Penalities
Productivity
Mgmt Efficiency
Comp Advantage
Chi Square
35.20**
49.67***
31.09*
52.71***
59.09***
88.75***
R2
0.10
0.18
0.27
0.17
0.21
0.32
n
491
434
293
485
465
480
Paramete
r Estimate
Point
Estimate
S.E.
Paramete
r Estimate
Point
Estimate
S.E.
Parameter
Estimate
Point
Estimate
S.E.
Parameter
Estimate
Point
Estimate
S.E.
Parameter
Estimate
Point
Estimate
S.E.
Paramete
r Estimate
Point
Estimate
S.E.
Variables of Interest
EMS Practices Forma 1 ization
0.29**
1.33
0.14
0.37**
1.45
0.18
1.05***
2.85
0.27
Customer Mandate
1.23"" "
3.44
0.42
0.65'
1.92
0.40
0.86**
2.37
0.42
Corporate Mandate
Both B2B Mandates
0.70"'
2.02
0.33
Compliance Goals
Pollution Prevention Goals
0.i¦
1.24
0.11
Eco-Effjeieney Goa Is
0.27"'
1.30
0.11
0.23"
1.26
0.12
0.25"'
1.28
0.11
Product Stewardship Goals
0.27***
1.31
0.11
0.62***
1.85
0.12
Endogenous Controls
Other Plans
Quality Management Systems
0.65**
1.91
0.27
Attitude Toward Env. Mgmt
Employee Involvement
0.17***
1.19
0.05
Existence of Parent Org.
0.58**
1.79
0.27
0.58*
1.79
0.34
Time with Formal EMS
3.57
0.37
1.37"'1
3.92
Exogenous Controls
Auto Supply Sector
-0.59-
0.55
0.33
1.03*
2.80
0.62
Plastics Sector
2.35
10.45
0.87
-0.53-
0.59
0.31
Coatings Sector
1.43
4.17
0.69
Private Ownership
0.60**
1.82
0.27
Recent
Non-Compliance+
Recent
Fines+
Region
Region
Region
Reaion
Region
Region
Reaion
Region
Reaion
Facilitv
*p < 0.10, ** p < 0.05, *** p < 0.01, + see Appendix II for discussion of variable construction
Individual variables that were not included within each respective model are marked in solid black.
81
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
Table 4: Regression Results for Self-Reported External Management Benefits12
Management Benefits
Model
M arket Share
N ew M arkets
Product Differentation
Plant Image
Capital Access
Chi Square
79.58***
89.43***
70.74***
39.14**
44.78***
R 2
0.28
0.32
0.26
0.38
0.23
n
463
468
439
177
306
Param eter
Estim ate
Point
Estim ate
S.E.
Param eter
Estim ate
P o int
Estim ate
S.E.
Param eter
Estim ate
Point
Estim ate
S.E.
Param eter
Estim ate
P o int
Estim ate
S.E.
Param eter
Estim ate
P o int
Estim ate
S.E.
Variables of Interest
Endogenous Controls
Quality Management Systems
-0.16"
0.85
0.07
0.87
0.07
Exogenous Controls
Region 1
Reg ion 2
Region 3
Region 4
Region 5
Region 7
Region 8
Region 9
Region 1 0
Facility Size.t
¦:p _ 0.10,p < 0.05, *** p < 0.01, + see Appendix II for discussion of variable construction
12 Individual variables that were not included within each respective model are marked in solid black.
82
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
Results
Environmental Performance
The formalization of EMS practices was a significant predictor of improved environmental
performance for only three of thirteen environmental indicators. Facilities that had adopted a
formal EMS or ISO 14001 equivalent EMS were significantly more likely to report increases in
recycling of waste (86 percent), in use of recycled inputs (80 percent), and in reductions in
regulatory violations (70 percent), than were those with less formalized environmental
management practices. However, the formalization of environmental management practices had
no significant associations with other important environmental performance indicators, such as
air and water quality, hazardous and non-hazardous waste generation, material and energy
inputs, or even control of spills. The proportions of facilities reporting improvement in each EPI
are presented in Table 5.
The presence or absence of a formal mandate to adopt an EMS, either from a facility's corporate
headquarters or from a customer, did not appear to have any significant influence on its
environmental performance. The only indicator for which a mandate variable appeared to predict
a change in environmental performance was chemical use, for which facilities with both a
customer and corporate mandate were less likely (57 percent) to report improvement. Without
any plausible rationale or additional evidence to explain this result, we are inclined to dismiss it
as spurious.
However, the priority each facility placed on particular goals of their environmental management
activities (compliance, eco-efficiency, or product stewardship) was a broadly significant
indicator of performance improvements. Facilities that placed emphasis on compliance-centered
goals were significantly more likely to report improvements in reducing violations (38 percent),
spills (36 percent), and smells (55 percent), but were significantly less likely to report
improvements in hazardous and non-hazardous waste management (27 and 30 percent,
respectively), recycling of wastes (27 percent) and use of recycled inputs (31 percent).
Conversely, facilities that placed emphasis on eco-efficiency goals were significantly more likely
to report improvements in recycling and use of recycled inputs (38 and 49 percent, respectively),
in energy and materials use (41 and 25 percent, respectively), and in hazardous and non-
hazardous waste generation (22 and 34 percent, respectively). However, similar improvements
were not reported for water or chemicals use. Finally, facilities that placed emphasis on product
stewardship were significantly more likely than others to report improvements in hazardous
waste generation (31 percent), in air quality (25 percent), and in smells (37 percent), though not
in other indicators.
A second variable that was broadly and significantly associated with patterns of environmental
performance improvement was the presence of a formal pollution-prevention or waste-
minimization plan. Facilities that had such plans were significantly more likely to report
improvements in recycling (207 percent), in use of recycled inputs (222 percent), in water use
(117 percent), and in non-hazardous waste generation (102 percent).
83
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
Table 5: Reporting of Improved EPI Performance
Not
Not
No
Environmental Performance Indicator
Improved
Improved
Applicable
Response
Total
n
» Water use
0.38
0.49
0.12
0.01
100%
617
» Energy use
0.37
0.53
0.08
0.02
100%
617
» Recycled inputs
0.37
0.35
0.24
0.03
100%
617
» Recycling of waste
0.35
0.51
0.12
0.02
100%
617
» Chemical inputs per unit output
0.36
0.41
0.20
0.03
100%
617
» Total material inputs
0.28
0.52
0.15
0.05
100%
617
» Hazardous waste generation
0.58
0.33
0.08
0.02
100%
617
» Non-hazardous solid waste generation
0.43
0.48
0.07
0.02
100%
617
» Wastewater effluent
0.35
0.41
0.21
0.02
100%
617
» Air pollution emissions
0.49
0.42
0.07
0.02
100%
617
» Smell generation
0.20
0.35
0.43
0.02
100%
617
» Severe leaks or spills
0.32
0.33
0.33
0.01
100%
617
» Legal violations or potential violations
0.30
0.41
0.27
0.02
100%
617
Recent inspections and non-compliances appeared to have a positive impact on environmental
performance for the most heavily regulated indicators, though not for others. For example,
facilities that had experienced recent inspections were significantly more likely to report
improvements in air pollutant emissions (36 percent), energy use (1 percent), reductions in
materials use (4 percent), and wastewater effluent (2 percent). Facilities that had recently
identified non-compliances were significantly more likely to report improvements in non-
hazardous waste generation (4 percent), in reducing violations (4 percent), and in hazardous
waste generation (3 percent).
Several other factors appeared to be associated with performance changes as well. The greater
the involvement of a broad range of employees in environmental management activities, the
more likely it was that changes would be reported in some of the outcomes that were harder to
achieve, such as reduction of material use, reduction of chemical use, and air quality. For
example, we could predict that if a facility involved all site employees and other interested
parties the likelihood of reducing materials used in production might increase by 270 percent.
However, because these findings lacked consistent statistical support across other logically
similar input variables, we are inclined not to place too much emphasis on this finding.
Interestingly, the attitudes of the responding managers toward environmental issues appeared to
have no significant effects on any of the environmental performance indicators. This finding
stands in contrast to other studies that have found positive attitudes about environmental
management to be significant drivers of environmental performance improvements (see e.g.
Kagan, 2004forthcoming).
The presence of formal quality management systems was significantly associated with an
increased likelihood of reducing wastewater effluent (82 percent), as well as water use (88
percent), chemical use (82 percent) and energy use (77 percent). However, in most of these cases
(with water use as the exception) such findings were present only at the significance level of
84
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
p=0.10. While these effects are not statistically striking, the convergence is worthy of comment.
Input variables such as chemical, water and material use may be affected by quality management
initiatives that seek to limit waste and increase efficiency.
The single most systematically influential predictor seen in the data was the greater improvement
of firms in the auto supply sector compared to the chemicals sector. Facilities in the auto supply
industry were nearly 150 percent more likely to report improvements in air pollutant emissions
than were chemicals firms and 100-200 percent more likely to report improvements in use of
most production inputs including material inputs, chemical, water, and energy use. Only in the
use of recycled inputs and in smell generation were auto supply firms significantly less likely
than chemical plants to have improved (54 and 66 percent, respectively). Facilities in the plastics
industry were also more likely than the chemical industry during the period of this study to
improve their air pollution emissions (177 percent), although they were less likely than chemical
firms to reduce spills (64 percent) and to increase the use of recycled inputs (54 percent).
Finally, we note several strong regional associations that deserve further investigation and
explanation. Facilities in EPA Region 10 (Pacific Northwest) were more than 500 percent more
likely to reduce non-hazardous waste generation and over 400 percent more likely to reduce
chemical use than did firms in the base region (Region 6, the south central states). Facilities in
Region 5 (Great Lakes) and those in Region 2 (New York/New Jersey) showed disproportional
improvement in water use (156 and 321 percent, respectively), and those in Region 4 (Southeast)
showed disproportional improvement in recycling (167 percent) and use of recycled inputs (181
percent). Several other regions showed significantly less improvement in some outcomes than
the base region. Whether these differences reflect differences in regional policies and priorities
of EPA itself, or in the industrial mix or other exogenous influences in those regions, they
suggest regional patterns or influences that deserve further investigation.
Management Benefits
The formalization of environmental management activities by facilities also appeared to be a
significant predictor of some management benefits. Overall, facilities with formalized
environmental management practices in place were more likely to report the benefits of those
activities as greater than the costs (33 percent) than those with less formalized activities. These
facilities also were more likely to report increased management efficiencies (45 percent) and
improved access to new markets (72 percent), and were much more likely to report improved
competitive advantage (285 percent) associated with their environmental management activities.
Table 6 shows the proportion of facilities reporting some benefits for each investigated activity.
The presence of business-to-business mandates also appeared to be a significant predictor of
some management outcomes. However, some important distinctions must be made regarding the
type of business-to-business mandate to which a facility is subject. Facilities subject to an EMS
mandate from their corporate parent were not more likely to report any management benefits,
and in fact were less likely (62 percent) to report improved access to new markets. Conversely,
facilities subject to a customer requirement to adopt a formal EMS were not only more likely to
report several management benefits, but were much more likely to do so for some such benefits.
For example, facilities subject to a customer mandate were more likely to report increased
85
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
savings from inputs and taxes (244 percent), increased productivity (92 percent), increased
market share (237 percent), improved competitive advantage (137 percent), and improved access
to new markets (116 percent).
Table 6: Reporting of Management Benefits
No
Not
No
Management Activity
Benefit
Benefit
Applicable
Response
Total
n
»
Cost savings in terms of inputs or taxes
0.40
0.40
0.16
0.04
100%
617
»
Avoidance of non-compliance penalties
0.77
0.12
0.08
0.03
100%
617
»
Increase in productivity
0.51
0.37
0.09
0.03
100%
617
»
Increase in market share
0.32
0.51
0.13
0.03
100%
617
»
Ability to reach new markets
0.37
0.47
0.12
0.04
100%
617
»
Product differentiation
0.28
0.51
0.18
0.03
100%
617
»
Improved company/plant image
0.79
0.13
0.05
0.03
100%
617
»
Improved access to capital markets
0.26
0.51
0.19
0.03
100%
617
»
Improved competitive advantage
0.47
0.40
0.10
0.03
100%
617
»
Improved management efficiencies
0.68
0.22
0.07
0.03
100%
617
As with environmental performance, the priority each facility placed on specific goals of their
environmental management activities (compliance, pollution prevention, eco-efficiency or
product stewardship) was a significant predictor of the management benefits facilities reported.
Facilities whose activities focused on pollution prevention activities were more likely to report
cost savings in terms of inputs and taxes (24 percent), increased market share (49 percent),
improved access to new markets (36 percent), and product differentiation (35 percent). Facilities
whose activities centered on eco-efficiency goals were more likely to report cost savings from
inputs and taxes (30 percent), improved management efficiency (26 percent), and improved
competitive advantage (28 percent) stemming from their environmental management activities.
Facilities that focused on product stewardship goals reported the widest range of benefits, with
such facilities more likely to report benefits from increased productivity (31 percent), increased
market share (53 percent), access to new markets (55 percent), product differentiation (80
percent), access to capital (88 percent), and competitive advantage (85 percent). The only
management benefits that such facilities were not more likely to report were cost savings from
inputs and taxes, avoidance of non-compliance penalties, and improved image.
While the goal of a facility's EMS was an important predictor of an increased likelihood of
reporting management benefits, the value of focusing specifically on a regulatory compliance
goal was less impressive. Facilities that emphasized regulatory compliance as their EMS goal
were no more likely than others to report any management benefits from their environmental
management activities. In fact, such facilities were less likely to report improved access to new
markets (25 percent). However, facilities that had a recent legal violation were more likely to
report benefits from avoidance of non-compliance penalties (292 percent) and improved
company or plant image (775 percent). Additionally, facilities with a parent organization, and
those with a waste minimization or pollution prevention plan in place, were much more likely to
report improved image benefits from their environmental management activities (658 percent
and 616 percent, respectively).
86
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
Findings and Implications
The results of this study offer new insights into the roles that environmental management
systems play in improving environmental performance, and into their potential for management
benefits to facilities that adopt them.
Finding: One should not expect to see substantial changes in environmental performance
simply because a facility adopts a formal EMS.
The most evident changes in environmental performance associated with EMS adoption were
changes in relatively marginal practices at the day-to-day operating level. These changes did not
extend to other impacts, such as air and water pollution, that might require more significant
changes in technologies and capital investments. In short, it appears that the kinds of
improvements that were most often reported in association with formalized EMSs were those
that were easiest and cheapest to improve at the margin in day-to-day operations at the facility
level ("better housekeeping," for instance), whereas those less often reported as improved were
those that would require more significant changes in actual production technologies and
processes (and perhaps, therefore, more significant capital investments, and approval from
decision makers located outside the individual manufacturing facility).
Finding: Specific goals for improvement, rather than simple EMS adoption, appear to be a
better predictor of the success of environmental management activities.
The specific goals of environmental management practices appear to be much better predictors
of performance improvements than mere presence of a formalized environmental management
system. Our results suggest strongly that improvements in environmental performance indicators
stem not from adoption of environmental management systems per se, but instead from adoption
of specific objectives aimed at correcting impacts of particular manufacturing processes or
products. Facilities that focused their environmental management efforts on areas such as eco-
efficiency and pollution prevention tended to report improvements in EPIs which reflected those
goals. Improvements reported by facilities with formal pollution prevention and waste
minimization plans in place (in addition to EMSs) add further support for this conclusion, as
such facilities were much more likely to report improvements in EPIs that these plans targeted.
For environmental regulatory agencies and the interested public as well as other potential
adopters, these findings suggest strongly that attention and any rewards or recognition should
focus on the specific environmental performance improvements targeted as priority objectives in
the EMS, and on the facility's success in achieving them, not merely on the adoption of an EMS
per se.
Finding: Environmental regulation plays an important role in promoting improved
environmental performance.
Whatever the benefits of voluntary initiatives such as adoption of an environmental management
system, government regulation continues to play an important role in environmental performance
improvements. Indicators that were regulated and inspected regularly showed greater
87
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
improvements than those that were not. For instance, recent inspections were strongly associated
with improvements in air pollution; this association was expectable and even reassuring, since air
pollutant emissions are arguably the most systematically monitored and inspected indicator of
environmental performance (water pollution is monitored more variably, and primarily by the
states rather than EPA). Facilities with recent non-compliances, or potential non-compliances,
were also more likely to report improvements in hazardous waste generation and reduced
violations: these too are understandable and reassuring results, since hazardous waste
management is an area of high potential economic liability to the facility if non-compliances
should be discovered on inspection and cited as formal violations.
Reductions in regulatory violations and non-compliances are not by themselves a surprising
result, given prior research on the impact of formalized environmental management techniques
(Andrews et al. 2003), and they reinforce the importance of regulation in promoting
environmental performance improvements when used in conjunction with voluntary "beyond
compliance" measures such as EMS adoption. But perhaps paradoxically, facilities that
emphasized compliance as a primary goal of their environmental management activities were
less likely to report either the additional environmental performance improvements or the
management benefits that facilities with other goals reported. While a compliance-focused EMS
may well have helped to reduce regulatory violations, it did not appear to promote improvements
in beyond-compliance environmental performance measures such as eco-efficiency or product
stewardship. Further, facilities that designed their EMSs with multiple environmental
performance objectives that spanned all four areas (compliance, pollution prevention, eco-
efficiency and product stewardship) had greater potential to accrue a suite of separate benefits
associated with each category.
Finding: The accumulation of management-related benefits also is associated more with
specific management goals than with EMS adoption.
Facilities with formal EMSs in place were more likely to report a number of management
benefits, but the specific goal of a facility's environmental activities was a much better predictor
of what kinds of management benefits a facility could expect to achieve than was the mere
presence of an EMS. Those facilities that focused their activities on pollution prevention goals
were more likely to report more external benefits (such as market share, access to new markets,
and product differentiation) than internal benefits (such as cost savings from inputs and taxes).
Conversely, facilities with a focus on eco-efficiency were more likely to report more internal
benefits (including cost savings and improved management efficiencies) than external benefits
(though such facilities were more likely to report increased competitive advantage, which can be
thought of as both an internal and external benefit). Facilities with a focus on product
stewardship were more likely to report a broad array of both internal and external benefits, and
more likely in general to report a greater number of management benefits, than facilities with
other priorities for their environmental activities. Such findings suggest that facilities with
"higher-order" environmental objectives, such as product stewardship, may achieve greater
management benefits as well (see Appendix I for discussion of "higher-order" environmental
objectives).
88
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
The finding that facilities with a compliance focus were less likely to report management
benefits reveals an important additional point. Facilities that limit themselves to achieving legal
compliance, instead of broadening the scope of their activities to include more advanced
practices and to integrate environmental management with other management practices and
objectives, may be missing fundamental opportunities to extract additional value from their
investments in such activities.
Sectoral differences, where automotive parts manufacturers were more likely to report
environmental performance improvements compared with the chemical industry, may reflect
technological and historical differences that deserve further investigation and comparative study.
In the chemical industry, end products are highly resource dependent: nearly all inputs (chemical
and otherwise) are potential product. Over the past two decades, the chemical industry has been
much more proactive in the adoption of systematic environmental management practices. Since
the Bhopal disaster of the mid-1980s and the subsequent introduction of the Responsible Care©
initiative a decade ago, many within the chemical sector may already have gone through a period
of intensive environmental management improvement, a process which auto supply facilities are
just now beginning to experience.
Finding: Business-to-business mandates offer limited potential to promote environmental
protection.
Facilities that were required to adopt environmental management activities by a customer,
corporate parent, or both, were no more likely to report performance changes than those who
were not subject to such requirements. On its face, this result appears to lend little support to
advocates of systematic efforts by the private sector to issue mandates in order to promote
environmental protection. However, if these mandates serve as an impetus for EMS adoption by
facilities that otherwise would have no such formalized practices in place, they may nonetheless
spur performance improvements if the resulting systems are designed correctly. That is, if the
EMSs that emerge in these facilities place emphasis on specific performance-related goals such
as eco-efficiency, pollution prevention, and product stewardship, and not simply on adoption of
EMS procedures, then explicit requirements may nonetheless lead to eventual performance
improvements.
Finding: Private sector mandates may lead to management benefits, depending on who
issues the mandate.
While mandates did not appear to affect environmental performance, customer mandates, at
least, did appear to affect management benefits. Facilities subject to corporate mandates were not
more likely to report management benefits in any category, but those subject to customer
mandates were more likely to report benefits, of multiple types.
Two possible reasons might account for the absence of reported management benefits associated
with corporate mandates. First, facilities subject to corporate mandates might have been subject
to stricter environmental management regimes prior to the mandate, as larger corporations have
been ahead of the curve in this regard, and may thus have been less likely to experience
additional improvements in internal management benefits. In checking this possibility, however,
89
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
we determined that while facilities with corporate mandates were more likely to have a pollution
prevention plan already in place - thus suggesting the existence of a previously-developed
environmental management regime - they were no more likely to have a pre-existing waste
management plan, total quality management plan, or ISO 9000 certification, and when these
proxies were combined there was no difference. Second, it is possible that corporate subsidiaries
market their products primarily through their parent corporations (or through higher-level
business units within them) rather than directly to outside markets, and thus may be less likely
than non-subsidiaries to experience changes in external benefits such as increased market share
or access to new markets, which are primarily the concern of those at the corporate level. The
corporation as a whole may well receive benefits from company-wide adoption, but these may
not be as directly salient at the facility level. This possibility requires further investigation of the
extent of subsidiary independence in marketing and other decision-making.
Facilities subject to customer mandates, conversely, might perceive internal benefits as being
greater because for many these efforts might be their first foray into formal environmental
management, or might at least represent a greater step forward. Many auto suppliers subject to
automotive mandates to adopt ISO 14001, for instance, have had no formal system in place prior
to those mandates (Hutson 2001). However, our data do not appear to show that facilities with
customer mandates were less likely to have had pre-existing pollution-prevention or waste
management plans, which could provide proxies for the presence of prior environmental
management activities. Customer-mandated facilities might also report external benefits more
frequently because unlike their counterparts who are corporate subsidiaries, they may be more
likely to market their products to a variety of customers. In such cases, concerns about market
share, competitive advantage, and plant/company image may be more salient at the facility level.
This possibility also deserves further investigation.
Conclusion
Taken together, the findings we have presented in this paper point to three main conclusions for
policy makers and business leaders.
First, broad environmental improvements should not be expected from the simple adoption of an
EMS, either voluntarily or as a result of a business-to-business mandate (nor probably, by
extension, in response to government rewards or other incentives). While adoption may be
valuable as a tool for helping facilities to reduce regulatory violations, to better manage day-to-
day activities such as recycling, and more generally to instill more explicit environmental
management procedures and accountability, EMS adoption does not by itself lead to
environmental performance improvements across a broader spectrum of performance indicators.
Those interested in EMS as a tool for substantive improvement in environmental performance,
such as reduced natural resource use or pollutant emissions, should concentrate instead on
promoting specific performance improvements as EMS goals, and ensuring that these targets are
achieved.
Second, the management benefits that facilities may gain from EMS adoption appear
commensurate with their efforts to move beyond compliance and focus on "higher order"
90
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
environmental goals such as eco-efficiency and product stewardship. For the business
community, such results suggest that a serious commitment to environmental performance
improvements may have economic rewards. Moreover, facilities whose environmental
objectives represent a broad array of performance goals may achieve a wider array of benefits,
both environmental and management, than facilities whose objectives are more narrowly
tailored.
Finally, voluntary efforts at self-regulation should be complements to, not substitutes for, more
traditional environmental regulation and enforcement by state and federal agencies. While
environmental management practices may in fact deliver environmental performance
improvements, particularly toward goals that are specifically targeted for improvement, facilities
that were subject to recent inspections were more likely to report such improvements in
performance indicators that are subject to inspection. The persistence of traditional regulation
may therefore facilitate the effectiveness of self-regulatory efforts.
ACKNOWLEDGEMENTS AND DISCLAIMER
We acknowledge with appreciation the financial support provided to this project by the Science
To Achieve Results (STAR) Program of the National Center for Environmental Research
(NCER), U.S. Environmental Protection Agency, under its Grant No. R829440. We also are
grateful for advice and assistance from Matthew Clark (EPA project officer); from Michael
Lennox; and from the staff of EPA's IDEAS database. An earlier version of some of our
conceptual framework, literature review, and preliminary analyses is forthcoming in a peer-
reviewed volume from Resources for the Future Press (Andrews, Hutson and Edwards
forthcoming). All conclusions and any errors are solely the responsibility of the authors.
REFERENCES
Ammenberg, Jonas. 2001. How Do Standardized Environmental Management Systems Affect
Environmental Performance and Business? Licentiate Thesis No. 907, Department of Physics
and Measurement Technology, Linkoping University, Sweden.
Andrews, Richard N. L. et al. 2003. Environmental Management Systems: Do They Improve
Performance? Final Report of the National Database on Environmental Management Systems
Pilot Project. Chapel Hill, NC: Department of Public Policy, UNC-Chapel Hill. Also on line at
http://ndems.cas.unc.edu/
Andrews, R. N. L.; Hutson, Andrew M.; and Daniel Edwards, Jr.(forthcoming). Environmental
Management Under Pressure: How Do Mandates Affect Performance? Chapter in Leveraging
the Private Sector: Management Strategies for Environmental Performance, edited by Cary
Coglianese and Jennifer Nash. Washington, DC: Resources for the Future Press.
Berry, Michael, and Dennis Rondinelli.. 2000. Environmental Management in the
Pharmaceutical Industry: Integrating Corporate Responsibility and Business Strategy.
Environmental Quality Management 9(3):21-33.
91
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
Christmann, Petra. 2000. Effects of 'Best Practices' of Environmental Management on Cost
Competitiveness: The Role of Complementary Assets. Academy of Management Journal 43(4):
663-880.
Coglianese, Cary, and David Lazar. 2003. Management-Based Regulation: Prescribing Private
Management to Achieve Public Goals. Law and Society Review 37(4):691-730.
Coglianese, Cary, and Jennifer Nash. 2001. Environmental Management Systems and the New
Policy Agenda. Chapter 1 in their Regulating From the Inside: Can Environmental Management
Systems Achieve Policy Goals? Washington. DC: Resources for the Future Press, pp. 1-25.
Corbett, Charles J., and David A. Kirsch. 2000. ISO 14000: An Agnostic's Report From the
Front Line. ISO 9000 + ISO 14000 News (February), pp. 4-17.
Darnall, Nicole. 2002. Why Firms Signal "Green": Environmental Management System
Certification in the United States. Ph.D. Dissertation, University of North Carolina at Chapel
Hill.
Darnall, Nicole, and Daniel Edwards, Jr. (forthcoming). Predicting the Costs of Environmental
Management Adoption: A Resource-based View. Paper to be presented at Academy of
Management Annual Conference, New Orleans, LA., August _, 2004.
Dowell, Glen; Hart, Stuart; and Bernard Yeung. 2000. Do Corporate Global Environmental
Standards Create or Destroy Market Value? Management Science 46(8): 1059-1074.
Florida, Richard, and Derek Davison. 2001. Why Do Firms Adopt Environmental Practices (And
Do they Make a Difference)? Chapter 4 in Regulating From the Inside: Can Environmental
Management Systems Achieve Policy Goals?, edited by Cary Coglianese and Jennifer Nash.
Washington, DC: Resources for the Future Press, pp. 82-104.
Corbett, Charles J. 2002 (unpublished paper). Diffusion of ISO 9000 and ISO 14000
Certification through Global Supply Chains. Los Angeles, CA: UCLA, Anderson School, 33 pp.
Gallagher, Deborah R. 2002. From Coercion to Cooperation: Influences on Environmental
Management System Design. Ph.D. Dissertation. Chapel Hill, NC: University of North Carolina
at Chapel Hill.
Geffen, C., and S. Rothenberg. 2000. Suppliers and Environmental Innovation. International
Journal of Operations and Production Management 20 (2): 166-186.
Gereffi, Gary; Garcia-Johnson, Ronie; and Erika Sasser. 2001. The NGO-Industrial Complex.
Foreign Policy, July-August 2001, p. 57-65.
92
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
Hamschmidt, J. 2000. Economic and Ecological Impacts of Environmental Management
Systems in Companies: Experiences from Switzerland. Proceedings, EURO Environment,
Aalborg, Denmark, October 18-20, 2000 (cited in Ammenberg 2001).
Hart, Stuart L., and Gautam Ahuja. 1996. Does It Pay to be Green? Business Strategy and the
Environment 5(l):30-37.
Hutson, Andrew M. 2001. ISO 14001 and the Automobile Industry in Mexico. Master's Thesis,
Nicholas School of the Environment and Earth Sciences, Duke University, Durham, NC.
Kagan, Robert, (forthcoming). Environmental Management Style and Corporate Environmental
Performance. Chapter in Leveraging the Private Sector: Management Strategies for
Environmental Performance, edited by Cary Coglianese and Jennifer Nash. Washington, DC:
Resources for the Future Press.
King, Andrew A., and Michael J. Lennox. 2000. Industry Self-Regulation Without Sanctions:
The Chemical Industry's Responsible Care Program. Academy of Management Journal 43(4):
698-717.
King, Andrew A., and Michael J. Lennox. 2001. Who Adopts Management Systems Early? An
Examination of ISO 14001 Certifications. Best Paper Proceedings of the Academy of
Management Conference.
Klassen, Robert D., and Curtis P. McLaughlin. 1996. The Impact of Environmental Management
on Firm Performance. Management Science 42(8): 1199-1214.
Klein, Naomi. 1999. No Logo. New York: Picador.
Kollman, Kelly, and Aseem Prakash. 2002. EMS-Based Environmental Regimes as Club Goods:
Examining Variations in Firms-level Adoption of ISO 14001 and EMAS in UK, US, and
Germany. Policy Sciences 35(l):43-67.
Matthews, Deanna Hart. 2001. Assessment and Design of Industrial Environmental Management
Systems. Ph.D. dissertation. Pittsburgh, PA: Carnegie-Mellon University.
Melnyk, Steven A.; Calatone, Roger; Handfield, Rob; Tummala, R. L.; Vastag, Gyula; Hinds,
Timothy; Sroufe, Robert; Montabon, Frank; and Sime Curkovic. 1999. ISO 14001: Assessing its
Impact on Corporate Effectiveness and Efficiency. Tempe, AZ: Center for Advanced Purchasing
Studies, Arizona State University.
Mohammed, M. 2000. The ISO 14001 EMS Implementation Process and Its Implications: A
Case Study of Central Japan. Environmental Management 25(2): 177-188.
Nash, Jennifer. 2002. Industry Codes of Practice: Emergence and Evolution. Chapter 14 in New
Tools for Environmental Protection: Education. Information, and Voluntary Measures, edited by
Thomas Dietz and Paul C. Stern. Washington, DC: National Academy Press, pp. 235-52.
93
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
Porter, Michael E., and Claas van der Linde. 1995. Toward a New Conception of the
Environment-Competitiveness Relationship. J. Econ. Perspectives 9(4): 97-118.
Rondinelli, Dennis A., and Gyula Vastag. 2000. Panacea, Common Sense, or Just a Label? The
Value of ISO 14001 Environmental Management Systems. European Management Journal
18(5): 499-510.
Russo, Michael V. 2001. Institutional Change and Theories of Organizational Strategy: ISO
14001 and Toxic Emissions in the Electronics Industry. Working Paper. Eugene, OP\R:
Department of Management, University of Oregon.
Sharma, Sanjay, and H. Vredenburg. 1998. Proactive Corporate Environmental Strategy and the
Development of Competitively Valuable Organizational Capabilities. Strategic Management
Journal 19(8): 729-753.
Walley, Noah, and Bradley Whitehead. 1994. It's Not Easy Being Green. Harvard Business
Review 72(3): 2-7. 1994.
94
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
Appendix I: Primary Variables of Interest
EMS Formalization
In order to determine the degree to which a formalized environmental management system
(EMS) was in place, respondents were asked to identify specific practices in place at their
facility. Responses were evaluated and coded on a three point scale to measure the degree of
environmental activity formalization. Environmental management activity formalization is
summarized in figure 2.
Figure 2: Environmental Management Activity Formalization
Environmental Activity :
adopted a written statement of environmental poliey goals,
set speeilie environmental perlbrmanee objectives.
planned speeilie. measurable steps to meet environmental perlbrmanee objeetives.
has a single manager who has primary responsibility lor environmental management
activities.
trains employees in speeilie aetivities related to environmental aspeels of their jobs.
has a proeedure in plaee for identifying legal requirements.
regularly tracks and manages environmental compliance indieators.
regularly traeks and manages environmental indieators other than compliance.
makes some environmental perlbrmanee data available to the public.
makes results of environmental performance available to employees.
has a formal proeedure for documenting environmental management practices.
has procedures in place for responding to environmental spills or accidents.
periodically conducts top management reviews of environmental performance.
conducts regular internal audits of environmental procedures or conducts regular
external (3rd party) audits of environmental procedures
System Objectives
Facilities were asked to describe the priority placed on written objectives at the site in order to
gain insight into what impact specific goals of environmental activities might have on
subsequent environmental performance. Facilities rated twelve statements on a scale of 0 (no
objective) to 4 (high priority), and were grouped into four categories of potential EMS focus
based on their responses (Gallagher, 2002).13 Figure 3 illustrates the relationships between these
constructs and the survey questions.
13 Gallagher (2002) argues that a facility's EMS goals move along a continuum from regulatory compliance toward
environmental sustainability. Within this paradigm, facilities with less aggressive goals place a narrow focus on
achieving compliance goals while more enterprising facilities place an emphasis on goals that are progressively
centered on pollution prevention, conservation of raw materials (eco-efficiency), product design (product
stewardship), and ultimately consideration of the facility's impact on environmental quality for future generations
(sustainability). While our data cannot confirm or refute a nested, directional, and cumulative relationship among the
concepts in this typology, this categorization of facility EMS goals appears consistent with our observations.
95
Formal ISO
EMS Equivalent
EMS
0 0
&
Ta
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
Figure 3: Environmental Management System Objective Constructs
Monitoring the discharge of
various pollutants
Minimizing accidents such as
leaks, spills and overflows
Meeting regulatory
requirements
Pollution ProM'iilioii (,o;ils
Improving existing
technology that will help
minimize the discharge of
pollutants
Replacing old technology
with newer equipment that
will help minimize the
discharge of pollutants
Redesigning the production
process to eliminate potential
environmental impacts
I'li'o-llNick-no (ioiils
Reducing water, energy, raw
materials used per unit output
Recycling raw materials used
in production
Reducing materials used in
production
Product Slow ;inl (.oiih
Redesigning
products to reduce
environmental
impact
Extending the
useful life of
products
Evaluating the
environmental
impact of materials
used
Conducting life
cycle analyses of
products
Business to Business Mandates
Because many manufacturing facilities have recently made the decision to adopt a formal EMS
in the context of industry pressures, respondents were asked to rate the impact of customer
and/or corporate pressures on the decision of whether or not to adopt an EMS. Facilities were
coded into four mutually exclusive categories based on the presence of specific pressures from
customers or corporate parents: no mandate, customer mandate, corporate mandate, and
customer and corporate mandate. Dummy variables were created for each mandate category.
96
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
Appendix II: Control Variables
Endogenous Controls
Facilities that have developed a broad array of management competencies often extend these
competencies to include the adoption of environmental management practices (Christmann 2000;
Florida and Davison 2001; Darnall 2002; Andrews et al 2003). Specifically, firms that have
previously acquired systematic knowledge in quality management and pollution prevention are
better able to leverage these internal capabilities toward development of environmental
management activities (Melnyk et al 1999; Corbett and Kirsch 2000; King and Lenox 2001).
Accordingly, this study has included several control variables that take such management
capabilities into consideration.
With respect to other specific management plans or systems, the survey asked respondents
whether or not the facility had formal pollution or waste minimization plans in place. Facilities
with either a formal pollution prevention or waste minimization plan were code 1, facilities with
neither plan were coded 0. Similarly, facilities with a Total Quality Management (TQM) system
in place were coded 1, facilities with no TQM system were coded 0. Whether or not the facility
belonged to a larger organization was also considered. Facilities with a parent organization were
coded 1, independent facilities were coded 0.
The length of time a facility has engaged in environmental activities may be an important
consideration in environmental performance changes at the facility level. For instance, the "80-
20 rule" argues that as a rule of thumb, 80% of environmental benefits can be reaped with 20%
of the costs: generally the first environmental problems to be solved are the easiest and most
visible. Such activities include recycling or pollution prevention practices, where facilities may
be able to garner impressive improvements over a short period of time. Over time, however, the
additional marginal benefits gained by reducing environmental impacts may be less than the
costs of more fundamental changes that sustain prior environmental improvements. To control
for the potential effect of time on environmental performance, facilities were asked how long the
environmental management system had been in operation. Facility responses were evaluated and
coded on a 0 to 4 scale where 0=no management system, l=less than one year, 2=one to two
years, 3=two to three years and 4=more than three years.
At the pre-testing phase of research, environmental professionals reported (through anecdotal
observation) two additional factors that deserved attention. Both the general attitude a facility
manager has towards environmental practices in place, and the degree to which employees of all
levels are committed to them, may influence the effect of environmental management activities
on performance outcomes. Accordingly, we included two additional control variables to account
for the influence of managerial attitudes about, and employee involvement in the facility's
environmental activities. Respondents were asked to rate five statements related to experience
with environmental activities on a 0 (strongly disagree) to 6 (strongly agree) scale (Figure 4).
The median response was used to measure the facility attitude toward its environmental
management activities. The survey instrument also asked respondents to describe the degree to
which five categories of employees and stakeholders were involved with environmental
management activities. Respondents rated the involvement of the environmental
manager/engineer, plant manager, non-management employees, contract service providers and
97
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
interested parties/citizen groups on a 0 (not involved) to 3 (highly involved) scale. Facility
responses were evaluated and the rating for each category of interested party was summed. The
total score for each facility (0-15) measured the overall level of involvement.
Figure 4: Statements of General Attitude Toward Environmental Activities
Overall, environmental management activities have resulted in better risk management, improved
company image, cost savings, and other benefits.
()\ erall. eu\ irounieuial management acli\ Hies lia\ e added e\tra cosis and consumed scarce resources
u ithoui addinu much \ aluc to our operations
Overall, the more experience we have with our environmental management activities, the more
opportunities we find (oreos|-e(Teeli\e impivncmcnls
()\ crall. I lie more experience we ha\ e w illi our eu\ iroiinicnlal management acti\ Hies, ihc more il
hcconics |iisi a paperwork routine
Overall, the more experience we have with our environmental management activities, the more it
becomes a strategic driver in our overall business management decisions.
Exogenous Controls
Factors outside the direct control of facility management may also affect environmental
performance. The industry within which a facility operates is perhaps one of the most important
such factors. Facilities operating in different sectors often have dissimilar impacts on the natural
environment, face distinctive regulatory pressures, and have unique opportunities for
environmental improvement. Industry dummies were constructed for each of the four sectors
examined in this study.14 Facilities in the chemical industry were used as the base category in
this analysis.
Facilities that are owned by large, publicly traded corporations often consider environmental
performance measures in corporate reports and assume pro-active environmental strategies due
to pressure from shareholders and environmental groups. The analyses used dummy variables
(ownership as publicly traded or privately held) to control for the potential effect of ownership
status on environmental performance.
Backed by the power of state and federal governments, regulators have the power to command
changes in environmental performance at the facility level. To control for the effect of regulatory
pressure on facility environmental performance, we constructed three variables using data from
EPA's IDEAS database. The number of inspections, number of non-compliances and the amount
of fines at each facility during the three years prior to the study period were modeled. Two
separate variables were used in the analyses depending on the indicator being modeled; for air
pollution emissions only data from the AIRS database were considered, for hazardous waste
generation only data from the RCRA database were considered. For all other indicators, each
variable was summed across the AIRS, RCRA and NPDES databases.
14 Facilities were grouped by four-digit SIC codes as reported in the EPA TRI database.
98
-------
DRAFT for EPA conference discussion only - 4/13/2004 - do not cite or quote
Within the United States, the degree of regulatory pressure to which manufacturing facilities are
subject may vary due to differing state laws, rigidity of enforcement, industrial composition, and
public attitudes towards environmental protection. To control for potential regional differences,
we constructed dummy variables based on the EPA region (1-10) in which these facilities are
located. EPA Region 6 (south central) was used as the base category in this analysis.
Finally, the complexity of facility operations may impact its environmental performance. Large,
multi-faceted facilities may experience greater challenges in controlling the number and
magnitude of their environmental impacts due to complex or diverse production processes. As a
proxy for this aspect of facility operations, respondents were asked to report the total number of
full-time employees at their site. Facility responses were evaluated and coded on a seven-point
scale consistent with the U.S. Census Bureau's classifications for the 1997 U.S. Economic
Census. The following scale was used for number of full-time employees reported by the facility:
Mess than 20, 2=20-99, 3=100-249, 4=250-499, 5=500-999, 6=1000-2499, 7=more than 2500.
Facility environmental performance was also considered within the management models since
many of the hypothesized efficiencies are expected to accrue to the facility from an increased
management of their impacts on the natural environment. More specifically, self-reported
environmental data were included in four of the eleven benefit models. Figure 5 shows which
variables were included within the specific models.
Figure 5: Inclusion of Environmental Performance Outcomes in Management Benefit
Models
Model
Self-Reported EPI Included
(\isi s;i\ iims iii icriiis of mpiiis or ia\cs
Water Use
Energy Use
Recycled Inputs
Chemical Inputs
Material Inputs
¦\\ oiclaiicc ill \i»n-( omphaiicc IVikiIhcs
1 .cual Violations or Potential Violations
Se\ ere Leaks or Spills
Impi'in oil (\iinp;iii> 1 Jl;iill Iniauc
1 .eual Violations or Potential Violations
So ere Leaks or Spills
Noise Generation
Smell Generation
Impan oil \ccess in ( apilal \1arkcls
Legal Violations or Potential Violations
99
-------
Transcription of Session V Discussant Comments by Chuck Kent (U.S. EPA, Office
of Policy, Economics, and Innovation)
Thank you very much. First of all let me say that all three of these papers I find
fascinating and of great interest to my work. Let me describe a little bit about what that
work is and why I'm so interested.
As part of the Office of Policy, Economics, and Innovation we attempt to—we try new
ideas, we pilot projects, we pilot all kinds of proposals with companies, with states, and
at the federal level. For example, we have a National Performance Track Program that
incorporates some of the ideas that you are hearing here today in terms of focusing on
performance, trying to get companies to articulate what they would try to improve over a
period of time, and in what categories are they willing to measure their accomplishments
against those items and share that information with the public and with the government,
and so on and so forth.
We also decided to make that program partly dependent on having an EMS in place—an
EMS not being the most important element of that program but a necessary condition to
being a member, and the idea there was that it might provide a better sense of
sustainability, of dependability, whether that company would continue to perform at a
higher level. But, the kinds of things that we hear from the literature continue to raise
questions for us in terms of: Is that program design the best that we can make it, and
what does it have to say for future programs of this type?
Let me talk for a few minutes about each of the papers and then come back and talk about
some broader policy initiatives that are under way and how this research might help us
even further.
With regard to the first paper, I am fascinated by the look into the difference between
perception and objective pressures on corporate decision making. Since my business is
trying to influence the behavior of corporate decision makers, it's very interesting to me
to understand better how they think and what causes them to do what they do. It looks
like an enormously complicated enterprise, and I'm looking forward to the future
iterations of this work. I am fascinated that there's apparently stronger pressure from
private sector sources and corporate sources than even the community. That's a bit
counter-intuitive to me, but in one respect consistent with the notion that what we're
seeing, for example in Performance Track, is that where you have a corporate executive
who chooses to go a certain direction with the firm, chooses to change the reputation of
the firm and to do it in a visible way, amazing things can happen. In fact, it's so
prominent a feature of the behavior that we're seeing that we are adding a component to
Performance Track this year that will recognize the corporate level of commitment.
Heretofore, the program has been designed at a facility level, just looking at performance
and commitments at a facility level, and now we're actually going to recognize
corporations that have chosen to do something at a corporate level and are pushing hard.
100
-------
A couple of sort of minor comments, and I'm by no means the right person to comment
on the methodologies and the elaborate statistical manipulations that are going on here.
I've been out of graduate school since 1975, and I don't know most of these terms, but in
terms of your Figure 1, Magali, on the right-hand side where you're talking about sort of
the outcome that you're trying to monitor, your paper talks a lot about the emphasis on
performance, but that figure only refers to whether there is an EMS in place. So, you
might try to reflect the performance element on the right-hand side as the desired
outcome. Also, I think it's generally acknowledged that TRI data don't always match up
very well with some of the things that we're actually trying to—both what companies are
trying to do with an EMS as well as the kinds of things we're trying to measure overall,
but I don't think I need to dwell on that.
One other minor point is that as we look at what you call the "proxies " for something
like community pressure, it seems to me that the ones that you ended up choosing are a
bit remote. I know it's very difficult to find good measures of things like that, but
counting the number of environmental lawyers in a community almost sounds like the
beginning of a good joke, no offense to the study. I would hope that maybe there are
more-direct measures of pressure that we could find ultimately—perhaps counting legal
actions, complaints, hearings, —who knows? I'm also reminded that Bob Kagan's work
has talked about the concept of social licensing, and I'm intrigued by that in relationship
to this paper, and I'm actually surprised not to see a reference, since you're both from the
University of California. So, you might want to think about whether that's relevant.
As the proud co-founder of the Environmental Studies Program at Santa Cruz in 1969,
I'm pleased to see Santa Barbara and Berkeley and others all working together on these
things.
So much for attempting to focus on data and methods, because I recognize that this is a
very difficult task. I noticed in your histogram on stakeholders, Magali, "corporate
management" seemed to be the highest influence; "regulators" was the second highest—
you didn't talk about it, but the third highest was "employees." That caught my eye,
because one of things I'm seeing—at this point, it's more anecdotal, but I've heard it
referenced in certain literature and from corporate executives who use EMS's—and what
they're finding is that one of the most interesting impacts of environmental management
systems is the impact they have on employees' motivation, and thinking, and acceptance,
sort of the legitimacy of the environmental agenda within a corporation. That's an
interesting dynamic that I think deserves more attention, because it can be sort of a
hidden force within a corporation, or even a small business, in terms of how the work gets
done and the sense of ownership of environmental values.
But clearly your work at this point is showing strong business motivation as drivers for
behavior, and, as I said before, the corporate pressure is something that I'm beginning to
see and we're recognizing it in the design of Performance Track. So, I look forward to
seeing future iterations of this work, and I commend you for handling such a large data
set. I wish you the best of luck.
101
-------
With regard to Deanna's work, I recall the first iteration of this a couple of years ago. In
fact, as I recall, when you were presenting it, you were about to give birth as well, right?
I'm wondering if Pete has something equally productive to show for his . . . [Pete
Andrews interjected: "My first grandchild next month!"] There we go—there's another
pattern we should take into consideration here.
I must say that the overall conclusion that there's no substantial difference in
performance in this particular sector in response to an industry mandate is not a huge
surprise to me. I've been personally skeptical about Ford Motors telling their suppliers,
"you've gotta have this piece of paper by a certain date; otherwise, we won't buy your
stuff," and what kind of behavioral modification that might bring about. Similarly, I have
concerns about the role of government in terms of requiring that firms have that piece of
paper and what kind of behavior that would induce. As you may know, EPA has been
very cautious about requiring EMS's as part of our regulatory policy, and I'll talk more
about that in a moment.
I am interested, as you pursue your work, Deanna, whether you intend to gather more
information on the motivations of the individual companies as they apply the EMS in
their own context—motivation other than the customer requirement—because I think that
does have a lot to do with what gets done with the EMS. In fact, I kind of wish the
microanalysis that you did, which I guess was of other companies, could have been
focused on the same sector, because it would have helped us learn a great deal more
about how they think. But, obviously, there were probably methodological reasons for
that.
Minor points: I would ask you all to be careful about the language you use when you talk
about EPA's voluntary programs. I think it's probably an exaggeration to suggest that
voluntary programs are somehow taking the place of regulatory programs at EPA. They
are certainly prominent and there are bunches of them. They're a little bit sloppy—as Jay
may have talked about earlier today, we're launching a major initiative to try to add some
considerable discipline to the design and management of voluntary programs at EPA.
But, try to avoid the suggestion that voluntary programs are sort of taking over the world,
because I don't see it that way.
In summary, I would say that we have a great interest in the motivational factors that
drive a sector like the auto assembly sector. Particularly, we're interested in the effect of
mandates. We, as an agency, are certainly not comfortable in focusing on the how over
the what, because ultimately we think that, as tax payers, you're more interested in the
what.
With regard to Pete's work, this is the second major piece of work from this study.
We've followed Pete's work with great interest, and it has informed our thinking, as well,
as we've designed programs at the Agency. We're particularly interested in his probing
into the difference between a self-initiated EMS and a required EMS. I'm a little
disappointed that so far the data really don't show much difference—that's just me
wishing for something more useful out of that particular probe. But, I would hope that
102
-------
each of you would think about that and see whether there's more that you can draw out of
your research and what you can tell us at EPA.
I did notice that the presence of a formal mandate, although it didn't appear to influence
performance, generally, where specific goals were set, appeared to make a difference.
This is a really important theme that I think we'll be hearing more and more about, and
that is—and you've said this before, Pete—an EMS maybe serves better as a window into
an organization, how they think and how they operate, than almost anything else. If you
have a clear vision about what you want to accomplish, an EMS may, in fact, be a very
useful tool for getting you there, but in and of itself, there's no magic in an EMS that's
going to get you what you want. That ties into the major point that Pete made that goals
matter. When you set out to do a compliance-focused EMS, what you get is better
compliance, usually. When you set out to do eco-efficiency, you tend to get some of that
a little better, and so on and so forth. And when you tackle something big, like
stewardship, surprise—you get something bigger than you otherwise would. I find that
very encouraging. To me, it should be part of the message to all of us, in government
and in the private sector, who try to use these systems: let's not be bashful—let's go for
something more dramatic.
Let me shift then to just a quick description of recent policy initiatives at EPA that relate
to this work that you would want to know about and that you, in turn, can inform. On
April 12th the Deputy Administrator signed a major new EMS strategy. It doesn't change
our policy, but what it does is articulate first of all the set of principles that are pretty well
stated already in EPA's position statements. But then we lay out, I think, six major
policy ideas to test, and what we're responding to is a great deal of interest, particularly
in several states around the nation, to try to incorporate EMS ideas and concepts into
permits and regulations—and even at the federal level there's been quite a bit of
discussion of this. So, what we're doing is trying to channel the thinking and the analysis
along the lines of at least these six broad area policy ideas to test. I would have you at
least be aware of those in your work and see whether you can help inform that debate as
well. I won't read them all; Jon Silberman is here in the audience, and he was one of the
major authors of this document, as well as George Wyeth and many other people at EPA.
We will have that document on the web within days, I'm told. We do have one copy
here, and Jon is willing to be the contact point for copies of this document. I would have
you take a look at that—it helps clarify both the Agency's position and the things that we
want to learn.
Finally, I just want to say that I'm the chair of an Agency-wide policy group on EMS,
and we're trying to establish a learning kind of climate within the organization so that it's
clear what questions we're trying to answer, it's clear what we would do with those
answers, and try to incorporate those into the regulatory design of our programs.
Thank you.
103
-------
Session V: Discussant Comments
Patrick R. Atkins, Alcoa
Introduction
Alcoa is a large company with over 385 manufacturing sites in 40 countries throughout
the world. We have determined that an effective management system is a requirement
for driving the appropriate systems and behaviors throughout our company to achieve our
goals of excellent environmental performance and continual improvement. We have 160
companies certified to the ISO 14001 Standard and have a Global Certificate from BSI
that recognizes that our corporate management systems conforms to ISO 14001.
Locations can be certified to ISO 14001 under that global certificate.
We find that environmental performance does improve through the application of the
principles of the 14001 EMS, and that our locations demonstrate continual improvement
when the management system is utilized. The ISO process requires the locations to
identify the activities at the facility that can impact the environment, prioritize those
impacts and establish a plan and process to address the critical issues on a priority basis.
Legal and other (corporate, community, customer, etc.) requirements also have to be
identified and used in the risk matrix that establishes the priorities for the issues.
I believe it boils down this, (perhaps too simplistic a view): Businesses with good
Business Management Systems and practices are able to leverage the ISO 14001 structure
and get additional value from the process. Businesses that are weak in other
Management areas have difficult with ISO 14001. GE and Toyota have found value from
ISO 14001 because the whole organization understands the value of a proper executed
management system.
With this background in mind, I offer the following comments:
Formalized Environmental Management Procedures: What Drives Performance
Improvements? Andrews, et al.
Andrews concludes:
First, broad environmental improvements should not be expected from the simple
adoption of an EMS, either voluntarily or as a result of a business-to-business
mandate (nor probably, by extension, in response to government rewards or other
incentives). While adoption may be valuable as a tool for helping facilities to
reduce regulatory violations, to better manage day-today activities such as
recycling, and more generally to instill more explicit environmental management
procedures and accountability, EMS adoption does not by itself lead to
environmental performance improvements across a broader spectrum of
performance indicators. Those interested in EMS as a tool for substantive
improvement in environmental performance, such as reduced natural resource use
104
-------
or pollutant emissions, should concentrate instead on promoting specific
performance improvements as EMS goals, and ensuring that these targets are
achieved.
This seems to be a misunderstanding of the intent of ISO. The system itself is not the end
product, the end product and value is the execution of the Objectives and Targets defined
by the system. An organization should not put all their energy in the system but in the
making the management activity work. This conclusion should be restated to emphasize
that value is obtained when the intent of the standard is met. (This is more clearly
defined in the ISO 14001:2004 revisions.)
He then concludes:
Second, the management benefits that facilities may gain from EMS adoption
appear commensurate with their efforts to move beyond compliance and focus on
"higher order" environmental goals such as eco-efficiency and product stewardship.
For the business community, such results suggest that a serious commitment to
environmental performance improvements may have economic rewards. Moreover,
facilities whose environmental objectives represent a broad array of performance
goals may achieve a wider array of benefits, both environmental and management,
than facilities whose objectives are more narrowly tailored.
I tend to agree. The goals and objectives have to have a long-range strategic component
or the organization will lack direction and will make not make leaps of progress.
However, this means that all stakeholders' definition of environmental performance
improvement cannot be satisfied at once.
His third conclusion:
Finally, voluntary efforts at self-regulation should be complements to, not
substitutes for, more traditional environmental regulation and enforcement by state
and federal agencies. While environmental management practices may in fact
deliver environmental performance improvements, particularly toward goals that
are specifically targeted for improvement, facilities that were subject to recent
inspections were more likely to report such improvements in performance
indicators that are subject to inspection. The persistence of traditional regulation
may therefore facilitate the effectiveness of self-regulatory efforts.
I think an important point that is missing in this conclusion is the following:
Significant aspects are controlled in the management system because they have a higher
risk. A business focuses resources and attention voluntarily on ISO 14001 system
objectives, which frequently supports compliance. State and Federal regulation of a site
is not always risk based. Regulatory agencies and Business should align their efforts to
ensure compliance. This can be address from both sides: State and Federal regulation
should become more risk based, and Business should be able to obtain regulatory relief
for well functioning management system.
105
-------
Environmental Management Systems: Informing Organizational Decisions:
Matthews and Lave
This is an ambitious effort to use several sets of data to measure environmental
performance changes in facilities that have ISO certification as compared to the
performance of facilities that have not been certified to ISO 14001. The researcher then
attempt to use a survey technique to better understand why environmental performance
improvement is not being seen
A major problem with this work in my opinion is it attempts to use nationally published
goals as measures of EMS success.. .is this appropriate? Can it even be done, given the
wide range of variables in such data, the recognized poor quality of the data in such data
sets, and the many other variables that impact plant performances such as product mix
changes, investment strategies by the corporation, process changes within the plant, etc?
Another problem appears to be a lack of understanding about how an ISO EMS is
supposed to operate. The paper notes that each facility may choose the areas for reduction
or improvement, but then claims that since EMS should provide for overall improvement
in environmental performance, some relationship should be expected, and the timing of
the work suggests to me that they expect this improvement in any and all environmental
improvement parameters to be immediate (within a year of certification). This is
certainly a high (and rapid) hurdle for measuring success, especially when the ISO
system is not designed to perform this way. The results seem inconclusive, as I would
suspect from the flawed study design.
The case study approach resulted in the conclusion that an ISO EMS may not be focused
on the areas that are important to all the stakeholders.. .especially regulators. I agree with
this conclusion. The authors then proposed that there are five elements that should be
used to all environmental decision making:
o Process diagrams and material flows
o Quantifiable goals and targets
o Reliable data
o Risk assessment tools
o Environmental personnel collaboration and education
I contend that a good EMS will include these elements and much much more.
EMS must recognize how an organization operates and how best to achieve the goals and
targets within the organizational systems that exist.
All stakeholders must be included in the management process.. .including communities,
customers, suppliers, regulators, investors, employees and even the public at large.
106
-------
Perhaps a more useful approach for the next effort would be to work directly with the
auto company that were the subject of the first part of the paper to determine the true
impacts of EMS on the business and the company's environmental performance
Institutional Pressure and Environmental Management Practice: Delmas and
Toffel
This research relies on a questionnaire approach to gather information on the drivers
behind environmental management practices. I am pleased to see that a great deal of
effort was invested in the form of the questionnaire and the processes to extract and
analyze the data. Such steps are critical in studies like this. I was disappointed in the
return rate, given the degree of effort expended by the researchers. I think a higher
response rate would make the data set more robust.
The study indicates that corporate and private pressures are more important than
community/regulator pressures or customer supplier pressures. I am not in full
agreement with this conclusion, and suspect that the data may have been biased by the
level of people in the organizations that responded to the questionnaires. Often a person
"in the trenches" at an operating location will have a view of the world that is a bit too
narrow, and will provide responses that can place the focus on the clear signals from a
corporate directive, when actually there are strong community and regulatory pressures
that are also influencing the entire management structure of the location and hence the
management system.
I suggest there be more focus on case studies as this work goes forward. Issues with the
complexity of the question of performance drivers cannot be adequately addressed via
questionnaires.
107
-------
Summary of the Q&A Discussion Following Session V
Matt Clark (U.S. EPA, Office of Research and Development)
(relaying a question from William D'Alessandro, one of the remote participants, who
addresses this question to the entire panel, but specifically to Dr. Andrews)
Stipulate for a moment that the following statement is a fact: ISO 14001 was battled over
ferociously and ultimately approved to require companies headquartered in the U.S. to do
absolutely nothing they were not already required to do by law and nothing they chose
not to address. How well would this explain the findings of your work?
Pete Andrews (University of North Carolina, Chapel Hill)
Dr. Andrews said he did not dispute the point, but he also did not think it affects the
usefulness of what companies are doing with EMS's or the work he and his colleagues
are doing to try to clarify that. He stated, "A lot of the firms are choosing to do this, and
so it becomes relevant and interesting and quite important for us to examine what, in fact,
they do when they're doing it. Particularly where public agencies are now offering
benefits for it, or encouraging it, our point is to point out that, depending on what goals
companies choose to adopt, they can really get some good results—so they can get
something that just, maybe, achieves better compliance but gets them no other business
benefits, or if they focus on compliance, it may not get them better performance in other
ways, or vice versa."
Jon Silberman (U.S. EPA, Office of Enforcement and Compliance Assurance)
Directing his comments to Dr. Andrews, Mr. Silberman said, "I couldn't help but think of
Maslow's hierarchy of needs and his behavioral research for individuals when we were
talking about compliance-based EMS's, and product stewardship, and e-Gov efficiencies,
etc. I guess my question for you is: Do you think there's some kind of similar effect like
that for EMS's, for example for companies that are unlikely to be able to reach higher
levels of higher order goals until they first satisfy their lowest order goals, and maybe
that's why some of these companies are getting compliance-focused EMS's? Would it be
a natural progression, do you think, that they would move up along the continuum more
towards the higher order goals over time?"
Mr. Silberman closed by stating that EPA believes "it's much better for EMS's to focus
on higher order goals," and as evidence he cited the fact that the Office of Enforcement
and Compliance Assurance has required only "26 total compliance-focused EMS's as
adjunctive relief in settlements since 1993 in literally thousands of cases. We really try
to save them for the companies that really are not able at all to manage their compliance."
Pete Andrews
Dr. Andrews responded, "In answer to your question, we've puzzled a lot about this and
you may know the work of one of my former colleagues who is now teaching at Duke,
Deb Gallagher, who has done some of the most detailed efforts to try to figure out
whether these are, in effect, nested goals—if you'd start with compliance and move on."
108
-------
He also said he'd be interested in hearing Pat's [Pat Atkins, Alcoa] reaction to the
question, as "a voice from the industry sector itself." He added, "Certainly for the
existing facilities we've looked at, it looks like some of that may be going on, but it may
be historical," and he pointed out that claiming that an initial focus on compliance is a
"necessity" for progressing to higher-order goals "would leave out the whole category of
unregulated industries."
Pat Atkins (Alcoa)
Mr. Atkins commented, "I think one of the big issues is metrics, and compliance is a
pretty good yes/no kind of metric and everyone can follow it, and I would equate it with
safety and public-reportable injuries, a pretty good metric that everybody can understand,
and it's easy to track and to report on. But, once you get that metric in place and begin to
move beyond it, and start looking at healthy workforce or off-site accidents or the
education of people on how to be safe in their own lives, it takes on a much broader
impact on the company. I think that's what will happen with the environmental
management systems—perhaps your first steps and objectives will be those that you can
easily measure, and compliance may be one of those, particularly if you're in a highly
regulated environment like the United States. And once you show progress in those
areas, you will begin to grow your list of aspects and things will migrate upward in terms
of impacts and risks of your business. So I think you're right—it will grow."
Michael Lenox (Duke University)
Dr. Lenox voiced his concern about the causality issue in Dr. Andrews' work, saying,
"I'm a little worried—could it be that firms are in ex-post, justifying their objectives
based on the benefits that they received or the behavior that they have created? But more
importantly, are there some other kind of underlying factors that are driving both the
objectives and then ultimately what happens at the end of the day, or is it a
recommendation that we can give to other firms to simply adopt these types of objectives
and then the outcomes will follow? Those could be very different prescriptions."
Dr. Lenox also commented on Dr. Andrews' suggestion that it would be great if all the
firms with EMS's would disclose the data they are collecting to the general public. Dr.
Lenox said that at first the idea sounded reasonable to him, also, but then he realized that
companies required to disclose the internal data they collect for their EMS's might
simply choose not to write EMS's, thereby defeating the whole purpose.
Pete Andrews
Dr. Andrews responded to the causality question by saying, "Remember, we didn't just
ask them what their goal was—we constructed those from what they told us about actual
changes they've made. So, I don't think that we're just getting an artifact in that sense.
There may well be underlying factors of some sort. Clearly, there are a lot of other
relationships we didn't talk about, such as, like others, we do see it affect prior practices
like ISO 9000 and so forth. Clearly, companies that already have that management
framework in place find it easier to piggyback and to model this into that larger
integrative framework. Cary Coglianese and Jennifer Nash have done a lot of thinking
109
-------
about the root underlying factors behind all of this, and I think there probably are some.
But I do think that what we're trying to do—finding out "What effect does it make what
people choose to focus on?"—in a sense, shouldn't be rocket science. The things they
focus on are the things that improve—what gets measured gets managed—what gets
chosen as objectives gets managed. That, to me, is simply reinforcing the point that it is
what companies choose to do—what they prioritize and what they choose to really make
the management system work for—that matters rather than just that they have this labeled
management system in place. That's reassuring in a sense, but it also says, "Okay, so
let's talk about performance and not just about using the EMS or the certification for
public purposes as a proxy for good performance."
Regarding the second point that was raised, about disclosure, Dr. Andrews pointed out,
"I'm trying to make a point that I wanted to really make a difference between what
companies choose to do for their own purposes and what they choose to do for public
benefits . . . There's a larger interesting conversation going about: many companies now
do produce environmental reports, but there's no comparability, there are no standards for
them, and so forth, so they become, basically, PR exercises. There is an attempt, the
Global Reporting Initiative—there may be some other attempts as well—to create a more
standardized basis, partly on the grounds that it may be in the company's own interests to
do that. As you get more folks—you know, social-screening funds—barraging you with
questionnaires every year, all different, and so forth, it might even be beneficial to them.
But it certainly would be beneficial to the leaders to have a common set of data. It may
not do everything, but let's have common, comparable kinds of data that would be worth
having. In that sense, disclosure is a larger conversation than EMS. I certainly wouldn't
say that we ought to drive them away from doing EMS by requiring them to disclose
everything in the EMS. But, if they're trying to come to EPA or to a state agency and
say, "I'm a good environmental steward, so I want flexibility or I want a prize for it, such
as a governor's recognition award," I'm saying let's not just talk about whether you have
an EMS. If you say you have an EMS, it means that you've thought about this—you're
achieving these performance changes—let's talk about what you're achieving and focus
on the performance."
Dinah Koehler (University of Pennsylvania)
Dr. Koehler commented that she thought Dr. Andrews had concluded "somewhat
wistfully" that "Gee, wouldn't it be great if they were to make serious capital investments
in product changes and process changes." She said that led her to "think that perhaps
we're expecting too much of an ISO standard." She challenged the entire panel to
consider what exactly can be expected from ISO 14000 and she wondered what has been
learned from ISO 9000 that might help. She closed by asking, "Are we to expect these
leaps or will we just see incremental, tiny changes in whatever outcome measure we
think we're looking at, as a function of some quality management system?"
Magali Delmas
Citing her experience, Dr. Delmas said, "When you talk to environmental managers, they
will tell you that it takes several years before they see any improvement, and that actually
110
-------
was the case with 9000—it would sometimes take 4 years. At first you would get more
information, actually, about how much emissions . . . and then they make progress 4 to 5
years down the road, so maybe what we are observing right now is that it's just too early
to say anything. Facilities started adopting ISO 14000 here in the U.S. in 1999, and the
majority—the big wave—was in 2000. So, it's really early to say anything. We should
just be more positive and wait a little bit more before drawing negative conclusions."
Deanna Matthews (Carnegie-Mellon University)
Dr. Matthews stated, "I think your comment that maybe we're expecting too much from
ISO 14000 might be a good way to phrase it, but perhaps some on the management side
have looked at this as—going back to the signaling issues—this is something that we just
need to do for business, and, thus, we're not really going to see any change in
performance. If that's so, then policy makers and business managers do need to look at it
from a different perspective and figure out what they need to do to make the big changes.
As Pat says, ISO 14000 is not the end product—it is trying to get it into that strategic
element of business strategy that's going to give the capital projects that are going to
change the processes that are going to make the big steps in how facilities improve. I
could reflect on some of Alcoa's goals for their 20/20 Vision that are technology
dependent, that are built into the long-term strategy and not simply based on an
individual plant having ISO 14000 or an EMS."
Pete Andrews
Dr. Andrews gave "a couple of quick responses," saying, "I think that's right. First of all,
ISO 14001 was one of the largest suite of ISO 14000 guidelines, guidance documents,
and so forth for business. I don't know if anybody's ever even looked at the question:
Do companies that have ISO 14000 care at all about the other ISO documents. Do they
use the other documents to inform this toward a larger strategic process? I haven't seen
it, and I'd be interested but surprised if it were happening. I think so much attention has
been focused on 14001, because it's certifiable and so forth, that that's happening. It
might be that it just takes longer, but more likely, I think, is that we've all been looking at
the facility level, and it may be that the real changes don't happen [at that level]. What
happens is one facility, at some point, gets out-competed or outmoded and gets closed
down and the production goes to some more-modern facility somewhere else. That's not
something that the facility manager is going to tell you about at that facility, and it's
either happened or not."
He concluded by saying, "So, I think ISO 14001 is valuable for mainstreaming
environmental considerations—again, you may be right, Deanna, that it's not happening
at the facilities you looked at, but I think it is happening at some others, and to the extent
that it does mainstream these considerations . . . into the job description of other
managers, it can help at the margin, with managing a particular facility, but I do think it
is process-based and we need to look to some larger processes as public policy experts or
people who care about actual environmental performance outcomes . . . how do we really
reduce large-scale environmental impacts?"
Ill
-------
Monica Araya (Yale University)
Dr. Araya directed her comments to Dr. Delmas, saying, "We all agree that it's difficult
to capture that action [i.e., the role of activists] in an empirical picture. My impression is
that monitors are saying that those activists in communities are not the reason they are
doing this. . . . My question is: Do you think that their perceived reasons for engaging in
environmental management systems should be taken as the reason why they do it? The
reason why I'm asking this is because in the area I work in, which is corporate
environmental reporting, companies sometimes say they are reporting because of internal
reasons, but, in practice, they are doing it because of actions that are coming against
them. So, the fact that they don't mention that in a survey doesn't mean that it is not
happening. So, could you in your study come up with a way of capturing this by taking
companies that have a corporate mandate and companies that do not and see how they
react to external pressures?
Magali Delmas
Dr. Delmas asserted that she thinks "it's really important to have objective measures and
survey measures together, so here what we are trying to do is to see how can you assess
community pressure objectively and how do environmental managers perceive this
pressure, so we will be able to look at the difference with this. ... So we will be able
actually to compare both to be able to see the distance between perceived and objective
pressure.
Glenn Farber (U.S. EPA, Office of Solid Waste and Emergency Response)
Dr. Farber stated, "My office often maintains that minimizing waste or other kinds of
environmental compliance or performance improvements provide business benefits, even
if it's only reducing the burden that we ourselves have imposed in the first place, but I've
never heard that associated with product stewardship, and I was surprised by your result,
Pete, that showed business benefits deriving from EMS's that had product stewardship as
an objective. I was wondering if you found that surprising as well."
Pete Andrews
Dr. Andrews responded, "We're interested in it, too. We'll try to figure it out."
112
-------
Corporate Environmental Behavior and the
Effectiveness of Government Interventions
PROCEEDINGS OF
SESSION VI: INFORMATION DISCLOSURE
A WORKSHOP SPONSORED BY THE U.S. ENVIRONMENTAL PROTECTION
AGENCY'S NATIONAL CENTER FOR ENVIRONMENTAL ECONOMICS (NCEE),
NATIONAL CENTER FOR ENVIRONMENTAL RESEARCH (NCER)
April 26-27, 2004
Wyndham Washington Hotel
Washington, DC
Prepared by Alpha-Gamma Technologies, Inc.
4700 Falls of Neuse Road, Suite 350, Raleigh, NC 27609
i
-------
ACKNOWLEDGEMENTS
This report has been prepared by Alpha-Gamma Technologies, Inc. with funding from
the National Center for Environmental Economics (NCEE). Alpha-Gamma wishes to
thank NCEE's Cynthia Morgan and Ann Wolverton and the Project Officer, Ronald
Wiley, for their guidance and assistance throughout this project.
DISCLAIMER
These proceedings are being distributed in the interest of increasing public understanding
and knowledge of the issues discussed at the workshop and have been prepared
independently of the workshop. Although the proceedings have been funded in part by
the United States Environmental Protection Agency under Contract No. 68-W-01-055 to
Alpha-Gamma Technologies, Inc., the contents of this document may not necessarily
reflect the views of the Agency and no official endorsement should be inferred.
11
-------
TABLE OF CONTENTS
Session VI: Information Disclosure
Information Disclosure and Risk Reduction: The Sources of Varying
State Performance in Control of Toxic Chemical Emissions
Michael Kraft, University Wisconsin at Green Bay 1
The Effect of Reporting Thresholds on the Validity of TRI Data as
Measures of Environmental Performance: Evidence from Massachusetts
Lori Snyder, Harvard University 29
Discussant
John Dombrowski, U.S. EPA, OEI 85
Discussant
Tom Beierle, Resources for the Future 89
Summary of Q&A Discussion Following Session VI 93
in
-------
Information Disclosure and Risk Reduction: The Sources of
Varying State Performance in Control of Toxic Chemical Emissions
Michael E. Kraft
kraftm@uwgb. edu
Troy D. Abel
abelt@uwgb.edu
University of Wisconsin-Green Bay
Green Bay, WI 54311
and
Mark Stephan
stephanm@vancouver.vvsu.edu
Washington State University, Vancouver
Vancouver, WA 98686
Prepared for presentation at a conference on Corporate Environmental Behavior and the
Effectiveness of Government Interventions, sponsored by the U.S. EPA National Center for
Environmental Economics, April 26-27, 2004, Washington, D.C.
1
-------
Abstract
This paper reports on initial findings of a research project that examines the effects of
information disclosure policies on environmental decisionmaking, specifically, actions related to
control of toxic chemical emissions in the United States. The project seeks to determine why
some companies do more to reduce toxic chemical pollution than others and why some
communities encourage such pollution reduction more than others. Ultimately, we will try to
identity the variables that most directly affect pollution reduction and by implication
improvements in public health.
Theory: We examine state trends in reduction of toxic chemical emissions through two
theoretical frameworks: the lens of comparative state environmental policy and a perspective
derived from the politics of information disclosure. We hypothesize that state environmental
release and waste reductions are a function of: (1) population size and economic prosperity; (2)
state policy resources; (3) the structure of environmental and industrial interests; and (4) state
political liberalism. Method: Ordinary Least Squares regression is used on data representing
trends in reported releases and production-related waste of 11,353 facilities between 1991 and
1997. Results: Consistent with theories of information disclosure politics, the level of
conservation group membership is the most influential factor influencing a state's ratio of firms
reducing toxic releases to firms increasing them. States with less ideologically polarized politics
also tend to host more release reducers than increasers. However, multiple regression models
could only weakly account for trends in production-related waste. These findings reinforce our
longer-term goal of incorporating sub-state level analysis (quantitative and qualitative) in an
effort to explain the patterns of toxic chemical releases and the effect of information disclosure
policies.
2
-------
Since the 1980s, information disclosure policies have emerged as one of the more
promising alternatives in the public policy repertoire. Considerable confidence has been
expressed in their utility and potential impact in achieving diverse societal goals: corporate
financial responsibility, food safety and nutrition awareness, drug safety, auto safety and fuel
economy, accountable campaign financing, and environmental protection, among others
(Graham 2002; Weiss and Tschirhart 1994). As a policy approach, information disclosure
represents what Schneider and Ingram (1990, 1997) refer to as capacity building tools, that is,
policies or programs that aim to inform or enlighten and thus to empower people to act on their
concerns. These tools are attractive in part because they may complement or replace government
regulation, thereby reducing the costs and burdens often imposed by enforcement of regulatory
standards. They also are consistent with widely held values promoting citizen access to
information in a democracy, captured in the phrase "right-to-know" (Hadden 1989). In addition,
they reflect conviction on the part of many analysts that new policy strategies may improve both
public and private sector performance, particularly in areas such as environmental protection,
where emphasis is placed increasingly on community-level action to promote public and
environmental health (Mazmanian and Kraft 1999; National Academy of Public Administration
2000; Portney 2003; Sexton et al. 1999; Tietenberg and Wheeler 1998).
Despite their widespread use, there has been little systematic inquiry into how information
disclosure policies actually affect corporate or community decisionmaking, and how they might
be designed to maximize their effectiveness—for example, in the way information is
communicated to the public and actions that could improve the public's capacity to understand
and use the information. The development of training programs in use of such data is one
example.1 As one prominent illustration, the federal Toxics Release Inventory (TRI) program
authorized by a provision of the Superfund Amendments and Reauthorization Act (SARA) of
1986 often has been cited as a success story in dissemination of information about releases of
toxic chemicals by industrial facilities. Title III of SARA created the Emergency Planning and
Community Right-to-Know Act (EPCRA), section 313 of which mandates that manufacturing
facilities report their annual releases of listed toxic chemicals to the EPA; the agency in turn
makes the information public.2 The information is available in an online database that can be
accessed by the public and stakeholders, and summary statistics are provided in a TRI Public
Data Release report each year. In addition, some environmental groups, most notably
Environmental Defense, make the data available online in a variety of graphic formats that allow
community residents to assess what each industrial facility in their communities is emitting
(www, scorecard. org).
The TRI program has been evaluated positively because of its appearance of effectiveness.
Indeed, the EPA itself calls it a "tremendously successful program," the results of which "speak
loudly for themselves" (U.S. EPA 2002a).3 The agency comments refer to the dramatic reduction
since the late 1980s in the volume of toxic chemicals released by manufacturing facilities
reporting under the program. Positive conclusions about the program's utility are reached even
when observers acknowledge that most of the reductions in chemical releases occurred during
the first five years of the program and that current volumes of releases remain quite large and
continue to pose a significant risk to public and environmental health (Press and Mazmanian
2003). Still, it is easier to document reductions in the release of such chemicals than it is to
explain the mechanism by which these reductions have occurred since the late 1980s, or to
anticipate how future reductions in emissions might occur if such disclosure policies remain a
major component of federal and state environmental protection efforts (Stephan 2002).
3
-------
As Graham and Miller (2001), among others, argue, the overall reductions in release of
toxic chemicals reported in the TRI require careful interpretation in light of the complexity of the
reporting system, major changes made to it over time, and the multiplicity of variables that can
affect corporate environmental decisions. They note, for example, that reported decreases in
chemical releases "mask widely varying trends in major manufacturing industries" (15). It is
apparent that economic factors affecting particular industries, new regulations or enforcement
actions by federal and state regulators, and decisions made by managers of particular facilities
with large releases can significantly affect the national trends on which analysts usually focus
(see also Natan and Miller 1998).4 The EPA itself regularly includes comparable warnings in its
annual TRI report on the "limitations that must be considered when using the data"; these
include the widespread use of estimated rather than actual data on chemical releases and
significant variation among companies in the way they estimate such releases (U.S. EPA 2002a,
ES-13).
One conclusion is that to understand the way such information disclosure works, why it is
successful (or not), and its potential for the future requires analysis that is directed at state-level
trends and especially at decisions made in specific communities that are located near facilities
producing large quantities of toxic emissions or emissions with high risk levels. Even here,
however, interpretations of the data differ. The EPA has been quite optimistic about the way the
TRI data are likely to be used by governments and citizens at these levels, as indicated in the
following statement from the press overview that accompanied its 2002 TRI report:
Governments—federal, state, and local—have used the TRI to set priorities,
measure progress, and target areas of special and immediate concern. The public,
our most important customer, has used the TRI data to understand their local
environment, to participate in local and national debates about the choices being
made that may affect their health and the health of their children and, ultimately,
to exert their influence on the outcomes of these debates (U.S. EPA 2002a, 1).
Many analysts seem to agree with the EPA's assessments. They argue that use of the TRI
information has "contributed significantly to community organizing efforts to change facility
emission behavior" (Bouwes, Hassur, and Shapiro 2001, 2). Similarly, a survey of corporate
leaders found that over half acknowledged that "pressure from community activists" has affected
their companies' behavior, sometimes leading to a reduction in chemical pollution (cited in
Bouwes, Hassur, and Shapiro 2001).
A common argument, captured in a recent report by EPA analysts, is that public access to
information "can drive change more effectively than regulations alone," and that the release of
such information "can help to empower community residents, heighten industry accountability to
the citizenry, and support efforts to ensure environmental justice" (Bouwes, Hassur, and Shapiro
2001, 1). The authors do state clearly that availability of data is a necessary but not sufficient
condition to achieve such goals. Beyond use by such activists, state and local agencies use TRI
data in developing emergency planning procedures, formulating legislation, and monitoring toxic
waste. Many states also have supplemented EPCRA with their own right-to-know legislation and
regulations, and some mandate reduction in a facility's toxic emissions.
Even with such qualifications, we wonder whether the confidence in the use of TRI data
expressed by the EPA and others is fully warranted, particularly as such use applies to citizens
and stakeholder groups at the community level. For most of the life of the TRI program, these
4
-------
citizens and groups worked with data that reflected only the total pounds of chemical releases,
not a more meaningful indicator of public or environmental health risks (that is, information
about exposure and toxicity). Some scholars have expressed somewhat similar reservations about
the effectiveness of voluntary pollution reduction measures adopted as an alternative to
conventional regulation, where companies may lack the incentives to do as much as they would
under a regulatory regime with consistent enforcement of the law (e.g., Harrison and Antweiler
2003). Only empirical research can adequately address these concerns.
These observations lead us to some basic questions that need to be asked about information
disclosure policies of this kind. We focus on environmental policies, but the questions are
equally applicable to other forms of information disclosure. How does the collection and
dissemination of such information bring about a change in corporate behavior that leads to a
reduction in the amount of toxic chemicals released to the local environment? What specific
changes in incentives occur as a result of such policies that might lead corporate officials to alter
manufacturing processes or take other actions to reduce releases of toxic chemicals?5 What are
the effects of information disclosure on communities themselves, specifically on the level of
attention paid to and concern over chemical risks in a local area? What efforts do public officials,
community leaders, environmental and health groups, and others make to try to reduce emissions
by local industry, and how effective are those efforts?
Our larger research agenda includes the building of a framework for understanding the
effects of environmental information disclosure on corporations and communities, using the case
of the federal TRI program. The research seeks to answer two major questions: What effect does
the release of information about pollution output have on decisionmaking by corporate officials,
community leaders, and representatives of local and regional nongovernmental organizations,
and on environmental quality outcomes? What factors mediate the use of such information and
thus condition behavioral changes and the environmental outcomes they produce?6 If those
questions can be answered, we hope to be able to determine which components of the TRI are
most likely to affect behavior, in what ways, and why. That knowledge should add to our
understanding of how the processes of environmental information disclosure work within both a
corporate and community setting, and how they result in improved environmental outcomes.
There are important implications for the design and implementation of information disclosure
policies, and for the TRI program in particular.7
The Analytic Framework
A rich body of work exists on the emergence of non-regulatory measures for businesses
and communities that are aimed at increasing public involvement as well as the dissemination of
information (Dietz and Stern 2003). However, these studies give little attention to factors
normally considered from a political science perspective that can help to explain the relationship
between environmental performance in industrial facilities and decisionmaking within
surrounding communities (Stephan 2002). We hope the project will help to fill that void by
integrating a number of different theoretical perspectives that bear on non-regulatory
environmental decisionmaking into a useful analytic framework.
These perspectives are drawn from studies of (1) risk perception and communication
(where variables such as trust in the source of information, a capacity to understand it, and
surprise or shock upon release are important); (2) organizational behavior (for example,
corporate embarrassment or shame when emissions data are released, and internal firm
5
-------
characteristics that affect action on chemical risks); (3) transaction costs in acquiring and using
information (disclosure policies may reduce the cost of information for citizens and thereby
encourage a higher level of participation); (4) environmental justice (inequities in the distribution
of chemical risks and economic benefits, the likelihood of variable levels of citizen action in
communities with different racial and economic profiles); (5) political agenda setting (new
information may alter the way problems and potential solutions are viewed as well as their
saliency); and (6) policymaking processes at state and local levels (such as different capacities
among states and communities to take action on environmental risks).
For the present paper we outline the framework only briefly because we focus below on
one component of firm environmental performance in the fifty states. One dependent variable is
a state's ratio of TRI facilities that reduce production related waste to the number of TRI facilities
that expand production related waste. Another is the ratio of TRI facilities that reduce their total
releases to the number of TRI facilities that expand their total releases.8 A third is the ratio of
facilities that decrease both releases and production related waste to those that increase both. We
seek to explain variation among the fifty states in these measures of environmental performance.
We do so partly because theory suggests that state variations do matter and reflect relevant
differences across states (see the literature review below). In later papers, we will examine the
related dependent variable of actual reduction in health risk to exposed populations using the
U.S. EPA's RSEI model discussed in note 1.
The key assumption in our working model is that TRI information disclosures affect
community and corporate decisionmaking through an alteration in risk perception and a change
in environmental beliefs and values. That is, the release of such information is likely to affect the
way corporate officials and community residents think about environmental and public health
and the value they attach to them. It may be that the information enhances knowledge of
previously uncertain health risks, that it conflicts with established expectations of industry
behavior or community health, or that it raises the saliency of these issues. The new knowledge
and attention given to it within both the corporation and community may also raise levels of
concern, propel the issues to a higher status on corporate and community agendas, mobilize
community leaders and activists, and as a result spur action to reduce emissions. How the
information releases, the environmental risks themselves, and the associated corporate and
community actions are covered by local media should be a significant variable as well.
Decisionmaking within a corporate setting and within a community obviously is complex
and is affected by many factors (including, as reported here, state differences). For example,
pollution reduction by companies may come about because of the actions of green investors and
consumers who cajole them to change their ways, pressures from the local community and
interest groups, competitive business practices (including perception of liability), shared learning
within the industrial sector, the development of new technologies, efforts to forestall anticipated
regulation through voluntary action, or calculations that the benefits of reducing releases of toxic
chemicals exceed the costs (Harrison and Antweiler 2003; Press and Mazmanian 2003). Some
companies may engage in what appears to be voluntary reduction in release of toxic chemicals
but which in reality is part of the firm's compliance with what it believes to be (or soon will be)
required by federal or state law.9
We will explore many of these possibilities during the field work stage of the project
when interviews will be conducted with corporate officials and community leaders and activists.
For this paper, as noted, we focus on variation in facility performance across the fifty states. We
distinguish facilities that decrease production-related waste and toxic chemical emissions (and
6
-------
thus risk) from those that show an increase in either and we look for patterns across the states to
help explain these variations.
As indicated in Table 1, this dimension is one of two that can be used to distinguish the
states by placing them into four categories: greening, browning, safer but still dirty, and cleaner
but riskier. The x-axis represents a continuum of production-related waste where facilities on the
left-hand side increase waste. If they decrease waste, they progress to the right side of the axis,
towards cleaner production. A continuum of risk runs along the y-axis, with facilities that
increase pollution risks on the bottom. If they reduce risk, they progress upwards towards safer
production. When facilities reduce both production-related waste and pollution risk, they move
from the lower-left to the upper-right, reflecting an ideal case of cleaner and safer production. As
discussed above, we will fill out the typology in a related paper that will include estimates of
actual risk reduction using EPA's RSEI model.
Industrial Pollution and Environmental Decisionmaking Within and Across States
That states vary in pollution production, management, and reduction, is not really in
doubt. However, scholars have tried to explain why some states do more than others to manage
or reduce their toxic pollution levels (including hazardous waste production), and why some
states see marked improvement in pollution control, while other states do not exhibit the same
success (Bacot and Dawes 1997; Grant 1997; Lester 1983; Lester and Lombard 1990; Potoski
and Woods 2002; Ringquist 1993; Williams and Matheny 1984; Yu et al. 1998). Some common
themes have emerged over decades of work.
Research that seeks to improve understanding of state-level policy choices or priorities
(as defined by regulatory decisions or budget expenditures) has focused mainly on whether
policy has more to do with political factors (such as interest group strength), administrative
capacities (both in terms of fiscal health and organizational capacities), or the severity of the
pollution problem. Results have varied, suggesting at the least that multiple factors help to shape
the choices made by states. For example, some studies have found that pollution severity is itself
a significant influence on state choices (Bacot and Dawes 1997; Lester et al. 1983; Ringquist
1994), while others have had mixed results (Potoski and Woods 2002) or found no such evidence
(Lombard 1993; Williams and Matheny 1984). A variety of studies have suggested that
administrative or state capacity is critical, whether the factor is measured by a consolidated
environmental bureaucracy (Lester et al. 1983), state size (Potoski and Woods 2002), or the level
of professionalism of state legislatures (Lester et al. 1983; Ringquist 1994). In contrast, Lombard
(1993) suggests that federal actions may be more critical to state decisionmaking than state
capacity factors. In particular, state enforcement activity may follow closely in the wake of
federal enforcement action, regardless of other state-level factors.
The most contradictory results relate to political factors, with some research suggesting
that the strength of interest groups or political parties is critical (Bacot and Dawes 1997;
Ringquist 1994; Sigman 2003; Williams and Matheny 1984), while others present findings that
undermine this thesis (Davis and Feiock 1992; Lester et al. 1983; Lombard 1993). In one of the
most interesting studies to date, Potoski and Wood (2002) find evidence that suggests political
factors influence some state policy choices, but not all, and that different types of environmental
programs will have different sets of drivers. They conclude that programs "that allocate
resources may be influenced primarily by the interplay of affected interests within the larger
political environment"(211). Because different environmental arenas tend to draw different
7
-------
interests, we should not be surprised that the significant factors vary. Another potential political
variable is state opinion liberalism which may have an indirect affect on state choices by
influencing the strength of interest group and party behavior (Erikson, Wright, and Mclver 1989;
Ringquist 1994; Wright, Erickson, and Mclver 1987).
Research focused on understandings changes in industrial pollution levels over time has
tended to focus on factors such as regulatory pressures, non-regulatory pressures (e.g.,
information disclosure programs), and industrial practices. Ringquist (1993) finds that the
stringency of regulation has a significant and negative relationship to sulfur oxide (S02) and
nitrogen oxide (NOx) emissions. In a related vein, Yu et al. (1998) find that as the direct
regulation of pollution emissions increases, the levels of emissions fall over time. Research also
suggests that variations in state-level information disclosure programs are related to changes in
emissions. States that fund information disclosure programs (Grant 1997) and extend their
outreach to affected communities (Yu et al.1998) are able to reduce toxic emissions, even when
controlling for factors such as industry production levels, state wealth, and enforcement
activities.
Regulatory and non-regulatory actions can also move in tandem and interact in important
ways (Yu et al. 1998). To use Schneider and Ingram's (1997) "policy tool" terminology, it may
not be a question of whether "informational tools" are better or worse than "authoritative tools,"
but rather the extent to which both may together have an impact on pollution levels. Though
shifts in industrial practices attributable to technological updates or economic changes
undoubtedly have an affect on pollution levels (Stephan 2003; Yu et al. 1998), the extent to
which these occur as exogenous pressures remains unclear. Arguably improvements in pollution
levels may in turn help to spur the spread of technological practices and may help revive states
dealing with economic downturns.
In sum, previous research suggests a multi-faceted examination of the policy relevant
factors which may influence changes in industrial pollution over time. Key categories of
variables include both political and administrative factors. Regulatory and non-regulatory
variations across states are potentially critical and cannot be ignored. Finally, control variables-
such as the severity of the problem—must be included in order to better assess whether policy
choices are proactive or reactive.
Data and Methods
EPA's online TRI Explorer (www.epa.gov/triexplorer) provided facility-level data for
this study's dependent measures. To characterize state-level TRI trends, the study analyzed a
sample of facilities (11,353) reporting in both 1991 and 1997 and their changes in reported
releases of toxic chemical pollutants as well as production-related-waste (PRW). The PRW is the
sum of all toxic wastes generated across a firm's production processes that a facility reports as
recycled, recovered for energy, treated on and off-site, or released on and off-site. The year 1991
was selected because it was the first year in which PRW was reported in response to the 1990
Pollution Prevention Act. We end with 1997 because it was the last year for which the EPA's
first version of the RSEI model included facility-level data. We will extend the period covered in
later work. The sample included only the 1991 core chemicals to assure consistent comparisons
of facility-level toxic chemical management across the two years.10 These facility characteristics
were then aggregated to explore state-level factors related to environmental waste management
8
-------
changes. However, ten states were excluded from the analysis because their relative smaller
concentrations of TRI facilities distorted statistical comparisons.11
Candidate independent variables encompassed state measures of policy, politics, and
resources (see Table 4). No single variable was critical in our initial data gathering. Rather, we
sought to identify a set of interrelated clusters of variables that captured our understanding of
what the literature suggests are the compelling factors in driving state differences.
A set of state policy measures were obtained from Bob Hall and Mary Lee Kerr's 1991-
1992 Green Index. Their index of "state policy initiatives" encompassed 73 different policies,
and we included this broad measure while also breaking out some of its subcomponents that
addressed toxic waste. Specific policy measures included the presence or absence of a state
program promoting access to information about toxic chemical usage and releases, the
imposition of fees that support community right-to-know (RTK), and laws requiring toxics
reduction plans and reporting. Another variable was included: a measure for state spending on air
pollution per capita. The policy measures were intended to capture the extent to which states
have policies, programs, and budgeted funds in place to deal with pollution.
Political variables included several measures from Erikson, Wright, and Mclver's (1993)
work. Their composite index of policy liberalism was included, as were measures for state
partisan identification (whether citizens lean Democratic or Republican) and ideological
identification (whether citizens lean conservative or liberal). Further variables included a
measure of mass ideological polarization (the extent to which party members in a state contrast
ideologically with the members from the opposing party) and measures for
Democratic/Republican elite ideology (the average ideological stance for party leaders in the
state). Beyond these "statehouse democracy" variables, separate measures were used for the
number of members of conservation groups, community improvement or capacity-building
groups, and philanthropic groups (all per 1,000 population) to further understand the role of
interest groups, both state-level and local. In sum, the variables were meant to capture the extent
to which political forces both within and outside of state governments might either directly or
indirectly influence pollution reduction.
Our resource variables were meant to serve as controls. Our intent was to avoid
attributing to policy or politics what may have more to do with demographic or economic
differences across states (although the three categories of variables are intertwined to some
degree). Population, poverty, unemployment, and income measures all came from the U.S.
Census. Following Potoski and Woods (2002) and Ringquist (1993, 1994), we included
measures of industry group strength, such as the value added by manufacturing (as a percentage
of the state's gross product) associated with state firms most responsible for air pollution.
Results
Forty-one percent of facilities (4,655) reported reductions in chemical releases and
reductions in production related waste, while thirty-four percent of sampled facilities increased
both (see Table 1). On average, 41% of a state's facilities achieved both pollution and waste
reductions, with the top percentile occupied by North Carolina, Maine, and Connecticut, which
had more than 48% of their TRIs moving towards safer and cleaner production.12 A second
group of state performers included New Hampshire, Delaware, New York, Virginia, New Jersey,
Massachusetts and Rhode Island, which saw more greening TRIs than 75% of their peers.
Lagging states, where less than 37% of TRI facilities achieved pollution and waste reductions,
9
-------
included Kentucky, Georgia, Utah, Michigan, Louisiana, and Maryland. Arizona and Tennessee
hosted the fewest TRIs moving towards greener production, with less than 32% reducing waste
and releases (See Table 2 and Figure 1). Table 3 presents state TRI aggregations for total
facilities, the percentage reducing both waste and pollution releases, and the ratio of waste and
release reducers to increasers.
While states, on average, saw more firms reducing releases than increasing them (134 to
99, respectively, or a ratio of 1.41), the opposite was true for production-relate waste trends. An
average of 126 firms per state were managing more hazardous waste in 1997 than in 1991, while
119 saw, on average, less PRW (see Table 4). These figures suggest that the TRI program is
perhaps not as successful as many have assumed it to be. Much depends on which indicator one
selects for analysis. As we show here, substantial decreases in overall emissions can occur at the
same time that many firms are increasing production-related waste. More recycling or energy
recovery may be occurring but firms can move closer to clean production with source reduction.
Bivariate correlations for a variety of socioeconomic, policy, and political measures
(listed in Table 4) on two sets of dependent variables (state concentrations of release and waste
reducer facilities) yielded a diversity of expected and unexpected patterns (see Table 5). As
previous research would suggest (Yu et al., 1998), informational tools in the form of state Right-
to-Know (RTK) initiatives produced a moderate correlation with the ratio of firms in a state
reducing production-related waste (PRW). However, an unexpected negative correlation
appeared between PRW reduction ratios and an index of state environmental policy
initiatives. The percentage of a state's firms reducing PRW also produced a negative correlation
with unemployment trends. However, because the variable was operationalized as the difference
in unemployment percentages from 1990 to 1996, higher positive values meant more job losses
and the negative correlation meant reductions were associated with unemployment decreases.
Similarly, a state's concentration of firms reducing toxic chemical releases also yielded a
negative correlation with unemployment increases and the index of state environmental policy
effort. Additional positive correlations between release reducer concentrations appeared with
environmental conservation membership levels, state ideological identification, a composite
index of policy liberalism, and, unexpectedly, with Republican elite ideology. These patterns are
somewhat consistent with findings in the comparative state policy literature. A slight variation
appeared when the release variable was operationalized in percentage terms instead of a
ratio. State partisan identification emerged with a significant correlation while no significant
relationships emerged for policy liberalism or Republican elite ideology.
When our comparison measure was the sheer amount of toxic releases, bivariate
correlation patterns mirrored relationships indicative of the states' size, and therefore their
pollution levels (see Table 5). The strongest correlations appeared between 1991 release levels
and population, Manufacturing Gross State Product, and Air Polluters Gross State Product, while
moderate relationships appeared with state poverty levels and the number of community and
foundation groups. Conservation group membership density, on the other hand, produced a
negative correlation with total release levels. On a measure of release trends, the correlations
again produced a pattern reflecting the influence of state size, with positive relationships
appearing with population, manufacturing GSP, and air polluters GSP. Finally, when the percent
change relative to a state's 1991 release amount base was examined, only the index of state
pollution prevention effort and conservation membership density produced significant negative
correlations.
10
-------
In the next phase of the study, the correlation analysis and our theoretical framework
guided the exploration of multiple regression models of environmental waste and release
reductions using ordinary least squares (OLS). Since our key dependent variables are interval
measurements, OLS is an appropriate statistical estimation technique. An initial analysis was
performed with all of our candidate predictors, although it was clear that issues of
multicollinearity made this model inappropriate.13 Candidate predictors were eliminated if they
presented multicollinearity problems or they were dropped through a series of likelihood ratio
tests for variables that had a consistently low impact on the model. Combinations of independent
variables from each theoretical grouping (policy, politics, and resources) were followed with the
calculation of a Variance Inflation Factor (VIF).When predictors are highly correlated, standard
errors of fitted coefficients are inflated and commonly diagnosed with the VIF procedure in
statistical software packages. Independent variable combinations were excluded if items
exceeded a VIF of three. Finally, adjusted R-squared were compared across models in order to
better understand what variables seemed to consistently hold up under multiple specifications.14
Although concerns about endogeneity were considered, we were unable to perform
comprehensive tests to cover this contingent. Our main concern was whether our policy factors
might be explained partly by our political or resource variables. Initial analyses suggest that there
is little or no direct relationship between variables such as "citizen right-to-know laws" and our
key political and resource measures.15 Part of the problem is that our measure for pollution
releases (TRI data) is truncated. We do not have good information before 1989, and therefore we
cannot check to see whether changes in pollution output have driven the creation of policies, the
nature of politics, or the quality of resources before the late 1980s. Future research will address
these concerns in more depth.
These iterations left just six independent variables in our models of toxic chemical trends
among facilities aggregated across the states. In the first model of the percentage of firms
reducing releases, conservation group membership produced the largest standardized coefficient
estimate. Ideological polarization was the second largest coefficient, but in a negative direction,
while state initiatives on citizen right-to-know produced a third significant and positive
coefficient. Insignificant factors included population, a pollution prevention index, and an index
of policy liberalism. The adjusted R2 indicated that over 30% of the variance in the dependent
variable could be explained by the model's combination of independent variables (see Table 6).
Model 2 displayed a higher R2 (45.6%) and showed that both conservation group membership
and ideological polarization relate to a state's ratio of release reducer facilities to increasers. In
both models, the statistically significant F-test demonstrated that rejection of the null hypothesis
that each independent variable except the constant are equal to zero could be rejected with 99%
confidence.
The same multiple regression model faired much worse when the dependent variable
encompassed Production-Related Waste (PRW) trends in percentage or ratio form. In model 3, a
state's percentage of facilities reducing PRW did not produce a commonly accepted level of
statistical significance on the F-ratio (see Table 7). However, in bivariate regressions, the citizen
right-to-know measure did produce a statistically significant model with an adjusted R2 of 0.232
and 0.117 (see Table 8).
In three final models, we performed multiple regressions on three dependent variables
that combined facility-level waste and release performance: (1) a ratio measure of facilities
reducing both toxic waste and releases to facilities increasing them; (2) the percentage of toxic
waste and release reducers; and (3) the percentage of toxic waste and release increasers. The
11
-------
model of our sample's ratio of toxic reducers displayed an adjusted R2 indicating that the
combination of independent variables accounted for nearly half (48.8%) of the variance in the
dependent variable (see Table 9). Conservation group membership again achieved the largest
standardized and significant coefficient, followed by ideological polarization in the negative
direction and value added by air polluters with a positive correlation.
When the percentage of toxic reducers (both waste and releases) became the dependent
variable, the regression's performance dropped substantially and achieved an adjusted R2
accounting for less that 18% of the variance. Conversely, our explanatory model of toxic
increasers accounted for almost half of the dependent variable's variance (the adjusted R2 =
0.494) with significant but negative coefficients on conservation group membership and value
added by air polluters and a positive coefficient for ideological polarization.
Discussion
The results, taken in their entirety, are suggestive rather than conclusive. There are at
least three conclusions we take away from them.
The first is that policy factors and political factors may both play a role in driving state
differences. For example, policy liberalism correlates positively with the number of facilities
making release reductions between 1991 and 1997 (see Table 5) while ideological polarization
seems to have a negative influence on a number of the dependent variables, including the
percentage of firms reducing releases and the ratio of toxic reducers to increasers (see Table 6).
Interestingly enough, the same cannot be said of reductions in PRW when observed in isolation
(see Table 7). This actually makes sense given the nature of the TRI as a form of information
disclosure. Pollution releases have a much greater salience in the media and in communities than
does PRW. Moreover, the results relating to ideological polarization may suggest that politically
liberal groups are having more of an impact in less polarized states, or facilities may be more
sensitive to this threat and more likely to be reducing release levels. Though PRW data are
publicly available, it is the subset of pollution releases that gets greater attention from interest
groups and average citizens. Other PRW besides releases are "off the radar" for all but the
closest of examiners of TRI information.
The consistency of statistical significance in our measure of environmental group
membership levels further reinforces the influence of a state's political environment (see Tables
6 through 9). Two interpretations are possible. First, facility pollution reductions in states with
more conservation group participation may be evidence that companies are facing more pressure
from organized environmentalists in some states and not others. Or, companies may be
anticipating pressure from a more vigorous environmentalism and be making efforts to forestall
unwanted attention by reducing emissions. A definitive conclusion one way or another must
await more focused research.
There is some evidence to suggest that the presence of state-level citizens right-to-know
programs helps to increase the percentage of TRI facility reducers in a given state across both
PRW and total releases (see Tables 6 to 8). This is not surprising, given that others have found
related results (Grant 1997; Yu et al. 1998). At the same time, we can only speculate why this is
true. One possibility is that state level right-to-know programs may actually enhance the abilities
of facilities to measure their own performance, which in turn may be driving them to reduce their
overall waste production. Facilities in the 1990s were measuring aspects of their production
processes that they were not measuring ten years previously. Alternatively, state level programs
12
-------
may increase the salience of both total releases and PRW for citizens outside of these firms,
which in turn motivates them to put pressure on the facilities for change. These results raise
further questions about the extent to which information disclosure programs are driving firm
behavior from within rather than firms responding to external demands for change.
The second major conclusion is that results in the correlation matrix suggest the
possibility of an interesting dynamic that defies straightforward explanation (see Table 5). When
comparing results for the ratio or percentage of reducers in total releases to the results for
changes to total pounds of pollution reduced in each state, the findings suggest that two separate
processes may be going on. It is conceivable that key political factors, such as policy liberalism
within a state, improve the likelihood that facilities will move towards pollution reductions rather
than expansions, while at the same time macro-level variables such as population and
manufacturing gross state product are driving the amount of change in pollution output overall.
The former suggests that political factors can influence the direction of change, while non-
political factors may influence the intensity of change.
In the context of theory about the function of information disclosure policies, it may be
that such programs can help to motivate companies to improve their environmental practices, but
the robustness of those improvements may be constrained by factors that no level of information
transfer can influence. Relatedly, the dynamic here may be similar to the findings of Potoski and
Woods (2002), in which different environmental issues draw different sets of interests to them.
The third major finding, the lack of significant results across most of our independent
variables, is itself telling. Arguably, our broad argument about the influence of information
disclosure programs remains viable: decision-making is most critically influenced closer to the
source—either through interactions with the community or within facilities (and companies)
themselves. State differences can mediate some of what happens at the local level, but only
partially.
Future Directions
Measuring and modeling the factors that influence innovative environmental decisions
and outcomes will significantly advance our understanding of their relationship to information
disclosure policies. However, we also intend to augment this modeling with qualitative analysis
through the use of questionnaires, interviews, and case studies. Such a precedent was outlined by
Meier and Kaiser (1996) when they leveled a provocative criticism of traditional regression
techniques. They point out that these focus on average cases when more interest may lie in
unusual cases. For the research under way, this would be communities with high concentrations
of facilities that have undertaken source reduction or have decreased pollution levels beyond
what would have been expected (performers). If the most ideal presumptions about information
disclosure are right, we would expect to find performing firms to indicate that their
environmental management choices were partially or even fully influenced by community
factors. On the other hand, much can be learned from communities hosting facilities that struggle
to change (or regress on) their environmental management and/or their pollution levels
(strugglers).
We can offer a few comments about the direction in which our research efforts are
heading. As noted, these kinds of initial results will guide our sampling of leading and lagging
facilities and the communities in which they are located. We will use a survey questionnaire to
be sent to corporate officials, administrative agency officials, environmental group leaders, and
13
-------
community leaders in 30 communities. The questionnaire will focus on how environmental
decisions have been affected by TRI information disclosures, with attention to the range of
variables identified earlier in the paper. Special consideration will be given to opportunities for
communication between communities and industrial facilities. As noted above, we will also
gather information on corporate environmental commitment, environmental expertise, and
management structure.
Following analysis of the data from the questionnaires, we will select ten communities
for in-depth field research. This effort will involve interviews with corporate officials as well as
environmental agency administrators, environmental group leaders, and community leaders, and
will focus on the effects of information disclosure on decisionmaking. We will match five
communities with leading industrial facilities and five with laggard facilities in terms of TRI
reductions and toxics management. The flexibility inherent in interviews will allow us to probe
more effectively on issues raised by our respondents that expand our understanding of
information disclosure programs and give us some indication of possible reforms or policy
changes.
Our research strategy is to compare leading and lagging firms and the communities in
which those firms are located. Selecting firms from the greening and browning categories should
facilitate this comparative case study approach, and the qualitative phase of our research should
enable us to learn much about why these firms make the kinds of decisions they do about toxic
chemical pollution.
As suggested early in the paper, there are policy implications to work of this kind. Until
we know more about the effect of information disclosure programs we cannot speak with
confidence about either their previous success or what changes in policy design or
implementation might make them more effective in the future. In later reports we hope to be able
to address how the TRI program might be redesigned to provide greater incentives to industrial
facilities to reduce both production-related waste and the volume of high-risk chemicals released
to the environment. We also expect to say more about how communities can use TRI data
(including new data coming from the RSEI model) to become better informed about health and
environmental risks and help to influence corporate environmental decisions that can have a
substantial effect on those risks.
References
Bacot, A. Hunter and Roy A. Dawes. 1996. "Responses to Federal Devolution: Measuring State
Environmental Efforts." State and Local Government Review 28 (2): 124-135.
. 1997. "State Expenditures and Policy Outcomes in Environmental Program
Management." Policy Studies Journal 25 (3): 355-370.
Bouwes, Nicholaas, Steven M. Hassur, and Mark D. Shapiro. 2001. "Empowerment Through
Risk-Related Information: EPA's Risk Screening Environmental Indicators Project." University
of Massachusetts, Amherst, Political Economy Research Institute, Working Paper Series Number
18.
14
-------
Davis, Charles E., and Richard C. Feiock. 1992. "Testing Theories of Social Regulation: A
Reassessment of Hazardous Waste Regulation in the American States." American Politics
Quarterly 20:501-11.
Dietz, Thomas, and Paul C. Stern, eds. 2003. New Tools for Environmental Protection:
Education, Information, and Voluntary Measures. Washington, D.C.: National Academy Press.
Erikson, Robert, Gerald Wright, and John Mclver. 1989. "Political Parties, Public Opinion, and
State Policy in the United States." American Political Science Review 83 (3): 729-750.
. 1993. Statehouse Democracy: Public Opinion and Policy in the American States. New
York: Cambridge University Press.
Feiock, Richard C. and Jonathan P. West. 1993. "Testing Competing Explanations for Policy
Adoption: Municipal Solid Waste Recycling Programs." Political Research Quarterly 46: 41-48.
Graham, Mary, and Catherine Miller. 2001. "Disclosure of Toxic Releases in the United States."
Environment 43 (8): 8-20.
Graham, Mary. 2002. Democracy by Disclosure. Washington: Brookings Institution Press.
Grant, Don. 1997. "Allowing Citizen Participation in Environmental Regulation: An Empirical
Analysis of the Effects of Right-to-Sue and Right-to-Know Provisions on Industry's Toxic
Emissions." Social Science Quarterly 78 (4): 859-73.
Grant, Don, and Andrew W. Jones. 2003. "Are Subsidiaries More Prone to Pollute? New
Evidence from the EPA's Toxics Release Inventory." Social Science Quarterly 84 (1) March:
162-173.
Hadden, Susan G. 1989. A Citizen's Right to Know: Risk Communication and Public Policy.
Boulder, CO: Westview Press.
Hall, Bob, and Mary Lee Kerr. 1991. 1991-1992 Green Index: A State by State Guide to the
Nation's Environmental Health. Washington, D.C.: Island Press.
Harrison, Kathryn, and Werner Antweiler. 2003. "Incentives for Pollution Abatement:
Regulation, Regulatory Threats, and Non-Governmental Pressures." Journal of Policy Analysis
and Management 22 (3): 361-382.
Lester, James P. 1990. A New Federalism? Environmental Policy in the States. In
Environmental Policy in the 1990s, ed. Norman J. Vig and Michael E. Washington, D.C., CQ
Press: 59-79.
Lester, James P. and EmmettN. Lombard. 1990. "The Comparative Analysis of State
Environmental Policy." Natural Resources Journal 30: 301-19.
15
-------
Lester, James P., et al. 1983. A Comparative Perspective on State Hazardous Waste Regulation.
In The Politics of Hazardous Waste Management, ed. James P. Lester and Ann O'M. Bowman.
Durham, NC, Duke University Press: 212-231.
Lombard, EmmettN. 1993. "Determinants of State Air-Quality Management: A Comparative
Analysis " American Review of Public Administration 23: 57-73.
Lowry, William. 1992. The Dimensions of Federalism: State Governments and Pollution Control
Policies. Durham, NC, Duke University Press.
Mazmanian, Daniel A., and Michael E. Kraft, eds. 1999. Toward Sustainable Communities:
Transition and Transformations in Environmental Policy. Cambridge: MIT Press.
Meier, Kenneth J, and L.R. Kaiser. 1996. "Public Administration as a Science of the Artificial: A
Methodology for Prescription." Public Administration Review 56 (5): 459-466.
Natan, Thomas E. Jr., and Catherine G. Miller. 1998. "Are Toxic Release Inventory Reductions
Real?" Environmental Science and Technology 32 (15): 368A-374A.
National Academy of Public Administration. 2000. Environment.gov: Transforming
Environmental Protection for the 21st Century. Washington, D.C.: National Academy of Public
Administration.
Potoski, Matthew, and Neal D. Woods. 2002. "Dimensions of State Environmental Policies: Air
Pollution Regulation in the United States." Policy Studies Journal 30(2): 208-226.
Portney, Kent. 2003. Taking Sustainable Cities Seriously: Economic Development, the
Environment, and Quality of Life in American Cities. Cambridge: MIT Press.
Press, Daniel, and Daniel A. Mazmanian. 2003. "Understanding the Transition to a Sustainable
Economy." In Environmental Policy, 5th ed., ed. Norman J. Vig and Michael E. Kraft.
Washington, D.C.: CQ Press.
Ringquist, Evan J. 1993. Environmental Protection at the State Level: Politics and Progress in
Controlling Pollution. Armonk, New York, Sharpe.
Ringquist, E. J. 1994. "Policy Influence and Policy Responsiveness in State Pollution Control."
Policy Studies Journal 22(1): 25-43.
Santos, Susan L., Vincent T. Covello, and David B. McCallum. 1996. "Industry Response to
SARA Title III: Pollution Prevention, Risk Reduction, and Risk Communication." Risk Analysis
16 (1): 57-66.
Schmidt, Charles W. 2003. "The Risk Where You Live." Environmental Health Perspectives
111 (7) June: A404-A407.
16
-------
Schneider, Anne L., and Helen Ingram. 1990. "Policy Design: Elements, Premises, and
Strategies." In Policy Theory and Policy Evaluation, ed. Stuart S. Nagel. New York: Greenwood.
Schneider, Anne L., and Helen Ingram. 1997. Policy Design for Democracy. Lawrence:
University of Kansas Press.
Sexton, Ken, Alfred A. Marcus, K. William Easter, and Timothy D. Burkhardt, eds. 1999. Better
Environmental Decisions: Strategies for Governments, Businesses, and Communities.
Washington, D.C.: Island Press.
Sigman, Hilary. 2003. "Letting States do the Dirty Work: State Responsibility for Federal
Environmental Regulation." Unpublished manuscript, National Bureau of Economic Research.
Stephan, Mark. 2002. "Environmental Information Disclosure Programs: They Work, but Why?"
Social Science Quarterly 83 (1) (March): 190-205.
Stephan, Mark. 2003. "Information Disclosure Programs: Influencing Elite or Mass Behavior?"
Vancouver, Washington: Washington State University-Vancouver, unpublished manuscript.
Tietenberg, Tom, and David Wheeler. 1998. "Empowering the Community: Information
Strategies for Pollution Control." Frontiers of Environmental Economics Conference, at:
www.worldbank.org/nipr/work paper/ecoenv/index.htm.
U.S. Environmental Protection Agency. 2002a. 2000 Toxics Release Inventory (TRI) Public Data
Release Report. Office of Environmental Information, May.
U.S. Environmental Protection Agency. 2002b. "Factors to Consider When Using TRI Data,"
EPA-260-F-02-017.
Weiss, Janet A., and Mary Tschirhart. 1994. "Public Information Campaigns as Policy
Instruments." Journal of Policy Analysis and Management 13: 82-119.
Williams, Bruce A., and Albert R. Matheny. 1984. "Testing Theories of Social Regulation:
Hazardous Waste Regulation in the American States." Journal of Politics 46 (2): 428-59.
Wright, Gerald, Robert Erikson, and John Mclver. 1987. "Public Opinion and Policy Liberalism
in the American States " American Journal of Political Science 31:980-1001.
Yu, Chilik, Lawrence J. O'Toole, Jr., James Cooley, Gail Cowie, Susan Crow, and Stephanie
Herbert. 1998. "Policy Instruments for Reducing Toxic Releases: The Effectiveness of State
Information and Enforcement Actions." Evaluation Review 22 (5): 571-589.
17
-------
Table 1. Firm Environmental Performance Associated
with Increasing or Decreasing Waste Production and Risk
Risk
Production-Related Waste
Decreasing
(Safer)
Increasing
(Riskier)
Increasing (Dirtier)
Decreasing (Cleaner)
Safer, but Still Dirty
1,898 (16.7%
Example: a firm increases
production but takes no steps to
control the higher volume of toxic
air releases and production-related
waste.
Greening Firms
4,655 (41.0%
Example: a firm could substitute a
more benign chemical for one of its
most toxic air releases but still
generate and even release large
quantities of less toxic pollutants.
Browning Firms
3,911 (34.4%
Example: a firm installs new
pollution control equipment that
decreases the volume of its more
toxic air releases and initiates
source reduction activity that
reduces its production-related
waste.
Cleaner, but Riskier
889 (7.8%)
Examples: a firm targets its biggest
waste streams for reductions while
maintaining or even increasing a
low volume, but highly toxic air
release.
Note: Production-related waste is the sum of all toxic wastes generated across a firm's
production processes that a facility reports as recycled, recovered for energy, treated on and ofif-
site, or released on and off-site. For this paper we use TRI total releases as a surrogate for risk. In
future papers we will apply the U.S. EPA's RSEI model to toxic air emissions to gain a more
useful measure of actual risk to exposed populations.
18
-------
Table 2. State Performance in Hosting TRI Facilities That Reduce Both
Production-Related Waste and Pollution Releases
Brownest
Browns
Yellows
Greens
Greenest
Less Than 37%
37% to 40%
40% to 43%
43% to 46%
More Than 46%
Tennessee
Nebraska
Iowa
New Hampshire
North Carolina
Arizona
West Virginia
South Carolina
Delaware
Maine
Maryland
Oregon
Washington
New York
Connecticut
Louisiana
Pennsylvania
California
Virginia
Michigan
Missouri
Illinois
New Jersey
Utah
Kansas
Massachusetts
Georgia
Oklahoma
Rhode Island
Kentucky
Ohio
Mississippi
Minnesota
Indiana
Wisconsin
Alabama
Texas
Florida
Colorado
Arkansas
Note: Ten states were dropped from the analysis because they had too few facilities to permit
comparative statistical analysis without distorting the results: Alaska, Hawaii, Idaho, Montana,
Nevada, New Mexico, North Dakota, South Dakota, Vermont, and Wyoming.
19
-------
Table 3. Trends in Toxic Chemical Waste and Release Reductions for the Fifty States
State
Reporting
% Waste
Ratio of
State
Reporting
% Waste
Ratio of
TRIs
& Release
Reducers to
TRIs
& Release
Reducers to
1991, 97
Reducers
Increasers
1991, 97
Reducers
Increasers
Alabama
270
40%
0.99
Montana
15
33%
1.00
Alaska
4
25%
.50
Nebraska
72
38%
0.96
Arizona
89
31 %
0.67
Nevada
17
38%
1.29
Arkansas
219
40 %
1.05
New Hampshire
61
46 %
1.40
California
662
43%
1.57
New Jersey
335
44%
1.63
Colorado
78
40 %
1.24
New Mexico
22
41 %
1.29
Connecticut
201
48%
1.81
New York
371
45 %
1.62
Delaware
42
45 %
2.38
North Carolina
471
51 %
1.71
Florida
226
40%
0.85
North Dakota
10
20%
0.33
Georgia
347
37 %
0.90
Ohio
931
42 %
1.27
Hawaii3
5
—
Oklahoma
138
42%
1.21
Idaho
18
39%
1.40
Oregon
135
39%
0.95
Illinois
702
43%
1.30
Pennsylvania
681
39%
1.11
Indiana
565
40%
1.02
Rhode Island
71
44%
1.41
Iowa
201
43%
1.16
South Carolina
293
43%
1.25
Kansas
139
42 %
1.05
South Dakota
29
41 %
1.09
Kentucky
240
37%
1.14
Tennessee
306
31 %
0.69
Louisiana
177
36 %
1.25
Texas
657
40%
1.22
Maine
52
48%
2.08
Utah
63
37%
0.85
Maryland
88
35%
1.03
Vermont
21
48%
2.00
Massachusetts
279
44 %
1.63
Virginia
262
44%
1.19
Michigan
497
36%
0.96
Washington
153
43%
1.38
Minnesota
252
40%
1.10
West Virginia
90
38 %
1.26
Mississippi
170
41 %
1.13
Wsconsin
478
40 %
1.12
Missouri
289
39 %
0.96
Wyoming
10
30 %
0.60
a No percentage is entered for Hawaii because the state has too few cases to permit comparison with the other
states.
20
-------
Figure 1. State Percentage of Pollution
& Waste Reducers
~
>46 %
n
43 to 46 %
~
40 to 43 %
~
37 to 40 %
< 37 %
-------
Table 4. Descriptive Statistics for Variables
N
Mean
Std. Deviation
Minimum
Maximum
Skewness
Kurtosis
Independent Variables
Statistic Std. Error Statistic Std. Error
State census population (90)a
50
4963.14
5462.411
454
29786
2.55
0.337
8.301
0.662
Conservation Members'3
50
8.544
3.58467
2.5
20.2
0.64
0.337
0.831
0.662
Manufacturing GSPC
50
20324.38
22457.41
486
113253
1.98
0.337
5.004
0.662
Air Polluters' GSPC
50
6791.14
7470.428
120
33794
1.80
0.337
3.515
0.662
Value Added by Air Polluters0
50
0.33584
0.140083
0.105
0.751
1.01
0.337
1.139
0.662
State Spending on Air
Pollution (per capita)
50
0.629
0.498317
0.14
3.26
3.24
0.337
15.256
0.662
Comm. Improvement Groupsd
50
159.3
165.6466
18
770
1.94
0.337
3.916
0.662
Foundations (1988)d
50
789.58
1096.155
30
6417
3.33
0.337
14.133
0.662
Median household income0
50
29102.82
5560.394
20136
41721
0.64
0.337
-0.151
0.662
Per capita income in 1989
50
13658.58
2347.937
9648
20189
0.62
0.337
0.057
0.662
State Pollution Prevention8
50
1.80006
1.529701
0
6
1.06
0.337
0.768
0.662
State Policy Initiatives8
50
25.5
14.57738
1
50
0.00
0.337
-1.200
0.662
Green Policies rank8
50
2200.64
670.2726
764
3230
-0.31
0.337
-1.101
0.662
Citizen's Right to Knowf
50
0.4
0.494872
0
1
0.42
0.337
-1.900
0.662
Aid for Right to Knowf
50
0.42
0.498569
0
1
0.33
0.337
-1.969
0.662
Toxic Cuts Lawf
50
0.32
0.471212
0
1
0.80
0.337
-1.425
0.662
Plan and Report Cutsf
50
0.26
0.443087
0
1
1.13
0.337
-0.759
0.662
Focus: Reduce Toxicsf
50
0.06
0.239898
0
1
3.82
0.337
13.124
0.662
Partisan identification
48
7.110417
11.38368
-17.4
35.3
0.16
0.343
-0.125
0.674
Ideological identification
48
-14.3
7.514525
-28
-0.2
-0.05
0.343
-0.748
0.674
Ideological polarization'
48
35.46458
9.118359
18.6
54
-0.08
0.343
-0.684
0.674
Policy liberalismj
48
-0.006875
0.986043
-1.54
2.12
0.19
0.343
-1.093
0.674
Dem. elite ideologyh
46
3.176957
2.045752
-0.71
7.47
-0.19
0.350
-0.806
0.688
Rep. elite ideologyh
46
-3.176087
1.754952
-6.05
-0.4
0.13
0.350
-1.251
0.688
Total TRIs reporting
in 1991 and 1997s
50
284.12
273.4119
5
1117
1.23
0.337
0.956
0.662
Dependent Variables
PRW Increasers9
50
125.92
120.7923
0
492
1.22
0.337
0.881
0.662
PRW Decreasers9
50
118.84
116.9886
1
483
1.30
0.337
1.188
0.662
PRW Unchanged9
50
2.18
2.869189
0
13
2.00
0.337
4.351
0.662
PRW Unreported9
50
37.18
35.62084
1
131
1.11
0.337
0.476
0.662
Percent of State TRIs
reducing PRW9
50
0.416179
0.083539
0.2
0.83
2.19
0.337
12.598
0.662
PRW Ratio9
49
0.928571
0.207385
0.33
1.63
0.54
0.340
3.100
0.668
REL Increasers9
50
99.98
95.26536
1
388
1.17
0.337
0.769
0.662
REL Decreasers9
50
133.68
131.0151
2
551
1.30
0.337
1.246
0.662
REL Unchanged9
50
14.34
17.78627
0
96
2.52
0.337
8.467
0.662
REL Unreported9
50
36.04
34.63274
0
131
1.12
0.337
0.555
0.662
REL Ratio9
50
1.408107
0.589504
0.22
4
2.07
0.337
7.268
0.662
22
-------
a. 1990 census, in thousands.
b. Membership in three environmental groups (Sierra Club, Greenpeace, and the National Wildlife
Federation) per 1,000 in population.
c. As of 1989. Value added by air polluters refers to the percentage of a state's gross product added
by manufacturing industries most responsible for air pollution. The Manufacturing GSP refers to
the manufacturing share of the Gross State Product, expressed in millions of dollars, as of 1989.
Following Ringquist (1993), the Air Polluters GSP refers to the sum of contributions to the Gross
State Product contributed by seven key manufacturing sectors: paper and allied products;
chemicals and allied products; petroleum and coal products; rubber and miscellaneous plastics
products; stone, clay and glass products; primary metal industries; and other transportation
equipment (not including vehicles). The data are taken from a Commerce Department Web site:
www.bea.doc.gov/bea/regional/gsp.
d. As of 1988. The measure of community improvement groups is the number of community
improvement or capacity building groups per 1,000 population. The foundations variable refers
to the number of philanthropic groups per 1,000 population.
e. Indices are taken from the 1991-1992 Green Index (Hall and Kerr 1991).
f. Dichotomous (l=yes, 2=no) measure taken from 1991-1992 Green Index (Hall and Kerr 1991).
The toxic cuts law variable refers to state policies intended to reduce toxic chemical emissions.
The citizen right-to-know variable refers to whether a state promotes public access to
information about toxic chemical usage and emissions. Aid for right-to-know refers to whether a
state provides funding for right-to-know programs. The plan and report cuts variable refers to
state policies that require facilities to plan and report on toxic chemical use and emissions. The
focus: reduce toxics variable refers to state policies that go beyond the planning and reporting
requirements to require source reduction.
g. State aggregations from EPA's TRI Explorer. The PRW measures used in the paper are not
normalized for production because we did not have access to production data. However, we did
control in multiple regressions models for total state releases in 1991, which is a surrogate of a
facility's production level.
h. These measures all use a mean score for the period 1976 to 1988, and are taken from Erickson,
Wright, Mclver (1993). Partisan identification and ideological identification are aggregate
measures based on New York Times/CBS surveys of registered voters in each state over this time
period. Democratic elite ideology and Republican elite ideology are measures built with factor
analysis from surveys of county party chairpersons, state legislators, convention delegates, and
congressional candidates by different researchers, most of which are from the same time period.
They also are drawn from Erickson, Wright, and Mclver (1993, pp. 98-99)).
i. Ideological polarization refers to the differences in mean ideology scores between registered
voters within the state who identified themselves as Democrats or Republicans (but not
independents) in surveys conducted during the 1976 to 1988 period (from Erickson, Wright,
Mclver 1993).
j. Standardized composite index of state policy measures on eight issues, including education
funding, Medicaid coverage, welfare eligibility, consumer protection enactments, criminal
justice liberalism (victim compensation, absence of death penalty, etc.), legalized gambling,
women's rights, and tax progressivity, taken from Erickson, Wright, Mclver (1993).
23
-------
Table 5. Selected Pearson Correlations
State Facility
State Facility
State Facility
State Facility
Total State
Total State
State Release
Release Change
PRW Reducer
Release
PRW Reducers
Release
Releases (91)
Releases (97)
Difference
(% of 91 base)
Ratio
Reducer Ratio
(%)
Reducers (%)
(91-97)
State population
0.113
-0.0032
0.0041
0.0442
0.5526***
0.3943**
0.3476**
0.0022
ST Poverty Percent
-0.167
-0.2184
-0.0700
-0.0473
0.3192*
0.2814*
0.1154
0.0138
ST Unemp Change
-0.337*
-0.4763***
-0.4899***
-0.3377**
0.2011
0.3045*
-0.1298
-0.2004
Median House Inc
0.287
0.4142
0.2031
0.2676
-0.1865
-0.1445
-0.0991
-0.0921
Per capita income
0.0945
0.3543
0.1402
0.2629
-0.1764
-0.1430
-0.0838
-0.1112
Manufacturing GSP
0.1849
0.0184
0.0542
0.0960
0.6211***
0.4774***
0.3362*
-0.0298
Air Pol Val Add
0.0440
0.0412
0.0688
0.2270
0.2396
0.1910
0.1189
0.0850
ST Air Pol Spend
-0.0356
0.1953
0.0592
0.1305
0.0203
-0.1461
0.2685
0.1732
ST Pollute Prevent
0.1343
0.2127
0.0242
0.1613
0.1369
0.2167
-0.1033
-0.3153*
ST Policy Initiatives
-0.2802*
-0.3661**
-0.2261
-0.2917*
0.0462
0.0491
0.0035
0.0808
ST Green Plcy Rank
-0.2786*
-0.3590**
-0.2042
-0.2734*
0.0714
0.0768
0.0038
0.0877
Citizen's RTK
0.4346***
0.2046
0.2613
0.2634
-0.0586
-0.0188
-0.0736
-0.1451
Aid for RTK
-0.0402
0.1263
-0.0106
0.2465
0.1912
0.0816
0.2076
0.1510
Toxic Cuts Law
0.2041
0.1317
0.0803
0.1390
0.2223
0.2610
-0.0230
-0.1768
Plan/Report Cuts
0.2361
0.1839
0.0963
0.1604
0.0478
0.1035
-0.0805
-0.1990
Reduce Toxics
0.0267
0.0367
0.0150
0.0654
-0.1360
0.1862
-0.5364***
-0.5900***
Cons. Members
0.1286
0.4560***
0.1187
0.2893*
-0.4102**
-0.357**
-0.1544
-0.0378
Air Polluters GSP
0.1661
0.0130
0.0578
0.1281
0.7330***
0.5670***
0.3911**
0.0250
Comm. Groups
0.1357
0.0401
0.0278
0.0934
0.5327***
0.4393***
0.2410
-0.0788
Foundations
0.1314
0.0983
0.0343
0.1040
0.362**
0.3029*
0.1583
-0.1095
Partisan ID
-0.0002
0.1005
0.1369
0.3043*
0.2140
0.2754
-0.0594
-0.1764
Ideological ID
0.0975
0.4253***
0.1946
0.3596**
-0.2152
-0.1228
-0.1811
-0.2006
Ideol Polarization
-0.0352
0.0522
-0.0226
-0.0284
-0.0928
-0.1178
0.0232
0.0175
Policy liberalism
0.1167
0.4125**
0.1349
0.2292
-0.1833
-0.1275
-0.1185
-0.1544
Dem. elite ideology
-0.0601
0.1598
-0.1105
-0.0466
-0.1828
-0.1804
-0.0345
-0.0631
Rep. elite ideology
0.0300
0.4247***
0.1223
0.2264
-0.2795
-0.1685
-0.2179
-0.1490
Correlation is significant at the 0.001 level (2-tailed) ***; at the 0.01 level (2-tailed)**; or at the 0.05 level*.
The "reducer" ratios above are simple ratios of reducing facilities to increasing facilities, where any value over 1.0 means a state has more
reducers than increasers, and vice versa for all values below 1.0. The "percent reducers" is a measure of the number of reducing facilities divided
by total number of facilities.
24
-------
Table 6. Ordinary Least Squares Regression on Toxics Release Inventory (TRI) Trends
State Percentage of State Ratio of
Release Reducers Release Reducers
Estimate21 Estimate
Independent Variables (Standard Error) (Standard Error)
Socioeconomic Factors
Population
-0.019
(0.000)
-0.021
(0.000)
State Policy Factors
Pollution Prevention
0.068
(0.007)
0.170
(0.040)
Citizen Right To Know
0.374**
(0.020)
0.164
(0.110)
State Political Factors
Conservation Group Membership
0.717**
(0.004)
0.884***
(0.240)
Ideological Polarization
-0.576**
(0.001)
-0.628***
(0.007)
Index of Policy Liberalism
-0.213
(0.014)
-0.086
(0.078)
Adjusted R2
0.3118
0.456
Standard Error
0.0532
0.2935
F
4.0207
6.588
Significance of F
0.004
0.000
(N)
40
40
a All of the regression coefficients reported here are standardized coefficients.
Table 7. Ordinary Least Squares Regression on Production-Related Waste (PRW) Trends
State Percentage of
State Ratio of
PRW Reducers
PRW Reducers
Estimate21
Estimate
Independent Variables
(Standard Error)
(Standard Error)
Socioeconomic Factors
Population
0.075
(0.000)
0.138
(0.000)
State Policy Factors
Pollution Prevention
-0.023
(0.005)
0.013
(0.020)
Citizen Right To Know
0.547**
(0.015)
0.349
(0.056)
State Political Factors
Conservation Group Membership
0.184
(0.003)
0.201
(0.012)
Ideological Polarization
-0.239
(0.001)
-0.267
(0.004)
Index of Policy Liberalism
-0.071
(0.011)
0.024
(0.040)
Adjusted R2
0.165
0.046
Standard Error
0.0039
0.1494
F
2.314
1.320
Significance of F
0.056
0.275
(N)
40
40
a All of the regression coefficients reported here are standardized coefficients.
25
-------
Table 8. Bivariate Regression on Production-Related Waste (PRW) Trends
State Percentage of
State Ratio of
PRW Reducers
PRW Reducers
Estimate21
Estimate
Independent Variable
(Standard Error)
(Standard Error)
State Policy Factors
Citizen Right to Know
0.501** (0.012)
0.373* (0.047)
Adjusted R2
0.232
0.117
Standard Error
0.0039
0.1455
F
12.768
6.155
Significance of F
0.001
0.018
(N)
40
40
a All of the regression coefficients reported here are standardized coefficients.
Table 9. Ordinary Least Squares Regression on Toxic Reducers and Increasers
Ratio of Toxic Reducers Percentage of Percentage of Toxic
to Increasers Toxic Reducers Increasers
Estimate3 Estimate Estimate
Independent Variables (Standard Error) (Standard Error) (Standard Error)
Socioeconomic Factors
Manufacturing GSP
0.151
(0.240)
0.111
(0.000)
-0.232
(0.000)
State Policy Factors
Value added by air polluters
0.339**
(0.337)
0.024
(0.050)
-0.388**
(0.059)
Citizen Right to Know
0.155
(0.015)
0.161
(0.014)
-0.161
(0.016)
State Political Factors
Conservation Group Membership
0.852***
(0.017)
0.599**
(0.003)
-0.798***
(0.003)
Ideological Polarization
-0.473**
(0.001)
-0.344
(0.000)
0.475**
(0.001)
Adjusted R2
0.488
0.175
0.494
Standard Error
0.253
0.038
0.045
F
8.42
2.66
8.62
Significance of F
0.000
0.039
0.000
(N)
40
40
40
a All of the regression coefficients reported here are standardized coefficients.
Notes
1 In 2003, the U.S. EPA's Office of Pollution Prevention and Toxics released for public comment
a draft version of a Community Air Screening How to Manual that carried the subtitle "A Step-
by-Step Guide to Using a Risk-Based Approach to Identify Priorities for Improving Outdoor Air
Quality." The manual illustrates well how government agencies might take action to improve the
26
-------
public's capacity to understand and use data, in this case Toxics Release Inventory data. The
manual strongly endorses the establishment of partnerships between communities and local
industry, with broad stakeholder involvement, as the best way to establish local priorities and
promote their achievement.
The agency's educational effort focuses on the Risk-Screening Environmental Indicators (RSEI)
model that we used in a paper presented at the annual meeting of the American Political Science
Association in August 2003 that covered only EPA Region V. We will use the model again but
for the entire nation in a paper to be presented at the 2004 annual meeting of the American
Political Science Association. The history of the model's development and its use to date is
described in Schmidt (2003) and in Bouwes, Hassur, and Shapiro (2001). Further information is
available at EPA's RSEI Web site: www.epa.gov/opptintr/rsei.
2 TRI facilities include all industrial firms that are required by the EPA to self-report the release
of any toxic chemical into the environment. The federal guidelines stipulate that a facility must
file a report for the TRI program if it conducts manufacturing operations within Standard
Industrial Classification codes 20 through 39 (with a broader set of categories applicable after
1998, such as metal mining, coal mining, and electric utilities that burn coal); has ten or more
full-time employees; and manufactures or processes more than 25,000 pounds or otherwise uses
more than 10,000 pounds of any listed chemical during the year. For 2000, the TRI was
expanded to include new persistent bioaccumulative toxic (PBT) chemicals, with lower reporting
thresholds. The full TRI list now includes over 650 chemicals.
3 The quote comes from the "overview" section of "The Toxics Release Inventory (TRI) and
Factors to Consider When Using TRI Data": www.epa.uov/tri/tridataytriOO/press/overview.pdf.
4 One striking figure drives home the importance of large manufacturing facilities. In 1999, just
50 facilities out of the 21,000 reporting that year accounted for 31 percent of all the TRI releases
nationwide (cited in Graham and Miller 2001). It also is apparent that larger facilities have been
more successful on the whole in reducing toxic releases than have smaller facilities.
5 One interesting finding is that subsidiaries of major corporations apparently have "significantly
higher emissions rates"; that is, they release a greater percentage of on-site TRI chemicals to the
environment. The study reporting this finding used data for 1990, and many changes in the TRI
program have been made since then. However, if the findings hold up over time, they suggest
that firms with less visibility have fewer incentives to reduce toxic chemical pollution. See Grant
and Jones (2003).
Given the design of the TRI program, a long-time concern is that facility managers have an
incentive to reduce the emission of listed chemicals and thereby to avoid or limit undesired
media attention and public pressures to alter business practices. But they might choose to do so
in a way that merely shifts use of unlisted chemicals for those on the list, or they might seek a
cheap and quick way to reduce the volume of emissions that does not reduce the risks to public
health as much as might be achieved through a more thorough alteration in production processes.
27
-------
6 In a paper delivered at the 2003 annual meeting of the American Political Science Association,
we identified five mediating factors as likely to be most important in shaping the way in which
information disclosure affects community and corporate risk perceptions and environmental
beliefs and values: community social capital, local and state political system characteristics
(including regulatory stringency), local social and economic conditions, firm or facility
characteristics (such as size, age of facility, and profitability), and the nature of media coverage
of the information released as well as coverage of corporate actions.
7 The research reported in this paper is supported by the National Science Foundation under
Grant No. 0306492, Information Disclosure and Environmental Decision Making. Michael Kraft
and Troy Abel are co-principal investigators, and part of the research is being conducted by
Mark Stephan. Any opinions, findings, and conclusions or recommendations expressed in the
paper are those of the authors and do not necessarily reflect the views of the National Science
Foundation.
8 As reported below, "total releases" can be understood as a subset of "production-related
waste." Total releases are simply the releases to land, water, and air, while production related
wastes also include produces that are recycled, recovered for energy, and treated either on or off-
site.
9 In a survey conducted in the early 1990s, Santos, Covello, and McCallum (1996) found that
regulatory compliance was one of the two reasons that facilities cited most frequently for
reduction of their TRI releases and transfers. The other was employee health. That is, where
many observers assume that TRI reductions are made voluntarily because the program is non-
regulatory in nature, this kind of evidence suggests a more realistic explanation would
acknowledge the incentives created by the larger regulatory environment, including company
concern over civil liability and state regulatory action. Without the requirements imposed by
such federal and state environmental regulation, information disclosure programs might be
considerably less effective.
10 EPA doubled the reportable chemical list in 1996 that potentially distorts longitudinal
analyses.
11 A total of 151 TRI facilities were excluded.
12 We use the term "TRIs" here as a shorthand for "TRI reporting facilities."
13 Results are available from the authors upon request.
14 Full results are available upon request.
15 There was some slight evidence that policy liberalism within a given state might influence the
likelihood of citizen right-to-know laws, but the results were not stable across multiple
specifications. Results are available upon request.
28
-------
The Effect of Reporting Thresholds on the Validity of TRI Data as
Measures of Environmental Performance: Evidence from
Massachusetts
by:
Lori D. Snyder
John F. Kennedy School of Government
Harvard University
lori_snyder@ksg.harvard.edu
1 Introduction
The old adage, "You can't manage what you don't measure/' is the primary rationale for
conducting systematic evaluations of the effectiveness of various environmental policy
initiatives. Only if governments, non-profits, industries, and communities have good
measures of environmental outcomes (changes in pollution levels, risk levels, etc.) can they
evaluate what policies work, how well they work, and how to improve their effectiveness.
If the metrics are not valid, then neither are the policy inferences drawn from these metrics.
This paper examines one aspect of the validity of a frequently used measure of
environmental performance — pollution releases reported under the federal Toxics Release
Inventory (TRI) program.
The TRI data are the most comprehensive data available on facility-level releases of toxic
chemicals. Facilities are required to disclose publicly their releases to all media —air, water,
land, and underground injection (on-site releases) —as well as their transfers of chemicals
off-site for recycling or disposal (off-site releases), for over 600 toxic chemicals. Because the
TRI data capture releases to all media and include measures of the environmental impact of
29
-------
a facility's activities beyond the facility's fence (off-site releases), many have argued that the
TRI data provide a more complete picture of facility-level environmental performance than
other available metrics (Karkkainen, 2001).
In fact, the TRI data are used frequently for the purposes of comparing environmental
performance across geographic areas and over time by government agencies, non-profits,
and academic researchers. Environmental Defense uses the TRI data as one of the indicators
of environmental performance in its Scorecard, an online database that allows the public to
compare counties on a number of environmental metrics.1 The Public Interest Research
Group (PIRG) and its state affiliates frequently compile lists of the worst polluters in a state
or region based on releases reported to the Toxics Release Inventory.2 Similarly, EPA ranks
states and industries by their total releases as reported in the TRI (U.S. EPA, 2002a). State
environmental agencies also publish annual progress reports that measure changes in
environmental performance based on changes in releases to the TRI, and label particular
industries and facilities as the top polluters based on these releases.3 Academic researchers
have used the TRI data as outcome variables in evaluations designed to determine what
factors affect environmental performance (Arora and Cason 1998; Grant and Jones 2003;
Helland and Whitford 2003; Khanna and Anton 2002; King and Lenox 2000). Researchers
have also used TRI data to evaluate whether requiring facilities to publicly disclose
pollution leads facilities to decrease pollution. Fung and O'Rourke (2000) and Wolf (1996)
1 The Environmental Defense scorecard can be accessed at www.scorecard.org.
2 See, for example, U.S. Public Interest Research Group 1998.
3 A full list of state TRI programs with links to the state annual reports can be found at
www.epa.gov/ tri/programs/state_programs.htm.
30
-------
argue that the observed 45 percent decline in overall TRI releases from 1988 to 1995
indicates that information disclosure is a valuable regulatory tool for reducing pollution.4
Despite the frequent use of TRI data for policy analyses, there are several known
concerns about the validity of these data as measures of environmental performance. This
paper defines the characteristics of a valid measure of environmental performance and
outlines several known threats to the validity of the TRI data. The paper then focuses on
estimating the magnitude of the measurement error created by the existence of arbitrary
reporting thresholds. The potential for measurement error exists because facilities are only
required to report releases to TRI if their use of a chemical exceeds some threshold. This
creates an incentive for facilities to reduce their use of a listed chemical to a level just below
the reporting threshold. However, this does not necessarily represent a real improvement in
environmental performance, as the facility's release level may remain largely unchanged (or
potentially could even increase). As a result, observed decreases in reported releases might
overstate the true change in environmental performance. This paper asks the question:
How much of any observed decline in reported releases could be artificially created by the
existence of the reporting thresholds?
The TRI data are also used to rank facilities based on their pollution levels.
Truncation at the reporting threshold may also have an effect of the validity of these cross-
sectional rankings. This paper also asks the question: How much would our rankings of
facilities change if we could account for releases by facilities that are below the reporting
threshold?
4 See U.S. EPA 2003 for a list of detailed discussion of how TRI data have been used by government,
business, academics, and citizen groups.
31
-------
While the potential for bias introduced by reporting thresholds is well known, there
has previously been little that users of the data could do to ascertain the magnitude of this
bias. The TRI data provide scant information that would allow a user of the data to
ascertain whether a facility that ceases reporting did so because it went below the reporting
threshold. However, the State of Massachusetts has expanded the disclosure requirements
under TRI in ways that better allow for assessment of the reasons facilities cease reporting.
This paper utilizes data from the Massachusetts Toxics Use Reduction Act (TURA) to
estimate, for facilities in Massachusetts, bounds on the degree of bias introduced by the
reporting thresholds in both time series and cross-sectional analyses.
The TURA data are similar to the TRI data. Indeed for Massachusetts facilities, data
reported to TRI is a subset of the data reported to TURA. However, the TURA data include
two additional features missing from the national data that make possible estimation of
bounds on the truncation bias. First, the TURA reporting forms contain an optional
question on why facilities are no longer reporting a chemical they had previously reported.
Second, the TURA program requires facilities to report their chemical use in addition to their
chemical releases. These two features of the data for Massachusetts allow estimation of the
number of facilities that cease to report because they reduced use below the reporting
threshold, but still use the chemical in positive quantities. These data are also used to
estimate bounds on the amount of "missing" releases that result. Because the TURA data
are in other ways identical to the TRI data, analysis of the TURA data provides some
preliminary evidence on whether truncation bias is likely to be a large or small problem for
the national TRI data.
The paper begins in Section 2.2, by describing the TRI and TURA data. Section 2.3
articulates a specific definition of "validity" of an environmental performance metric and
32
-------
highlights several threats to the validity of the TRI/TURA data as measures of
environmental performance under this definition. This section defines truncation bias at the
reporting thresholds and details how this bias may invalidate both time-series and cross-
sectional comparisons. Section 2.4 estimates the magnitude of the truncation bias created by
the existence of the reporting thresholds using the TURA data. The results suggest that
truncation bias is indeed a serious threat to the validity of these data as measures of
environmental performance, particularly in cross-sectional comparisons. Time-series
estimates are off by roughly 40 percent in Massachusetts in the worst-case scenario. That is,
40 percent of the observed decrease in releases in Massachusetts may be artificial declines
created by the reporting thresholds. For cross-sectional comparisons the results for
Massachusetts indicate that quartile rankings of facilities may be wrong as often as 45
percent of the time when truncation bias is not accounted for. Section 2.4 ends with a
discussion of the implications of the Massachusetts findings for the use of nationwide TRI
data. Given the potential importance of these effects for policy analysis, Section 2.5 presents
suggestions for adjusting policy analysis to account for truncation bias.
2 The U.S. Toxics Release Inventory and the Massachusetts Toxics Use
Reduction Act
In December 1984, a Union Carbide pesticide plant in Bhopal, India accidentally
released 40 metric tons of methyl isocyanate, killing an estimated 2,000 people and injuring
100,000 others. Shortly thereafter, a pesticide release in West Virginia hospitalized 150
people. Partly in response to concerns raised by these high-profile accidental releases,
Congress passed the Emergency Planning and Community Right to Know Act (EPCRA) in
33
-------
1986. EPCRA requires manufacturing firms to report releases of specific toxic chemicals on
an annual basis, and to make these reports available to the public. The U.S. Environmental
Protection Agency's (EPA) implementation of EPCRA resulted in the creation of the Toxic
Release Inventory (TRI), which requires large manufacturing facilities publicly disclose total
releases of listed toxic chemicals to all media on an annual basis.
There are three factors that determine whether a facility is subject to the disclosure
requirements. The first is industrial sector. Originally, only manufacturing facilities were
subject to the TRI reporting requirements. Subsequently, several other industrial sectors
were added including: federal facilities, metal and coal mining, electrical generating
facilities that combust coal or oil, chemical wholesale distributors, petroleum terminals and
bulk storage facilities, and hazardous waste treatment storage and disposal facilities. The
final two determinants of regulatory eligibility concern facility size. Only facilities with 10
or more full-time equivalent employees are required to report pollution levels. In addition,
facilities are only required to disclose releases of listed chemicals for which they either: (1)
manufacture or process more than 25,000 pounds or (2) otherwise use more than 10,000
pounds. Figure 2-1 provides a flow chart of TRI reporting requirements.
Several states have subsequently required additional public disclosure for plants in
their jurisdictions. In 1989, Massachusetts passed the Toxics Use Reduction Act which
expanded the reporting requirements for facilities in that state (Massachusetts Department
of Environmental Protection 2003). There are three main differences between the TURA and
the TRI requirements. First, the list of chemicals for which facilities are required to report is
larger in Massachusetts. Massachusetts facilities must report releases of all chemicals
required by TRI and also any chemical listed under the Comprehensive Environmental
Response, Compensation, and Liability Act (CERCLA), commonly known as Superfund.
34
-------
Is your facility a manufacturing facility (SIC 20-39)?
~
yes
Does your facility have 10 or more full time
equivalent employees?
yes
For Each Listed Chemical Determine:
DO NOT REPORT
RELEASES FOR
ANY CHEMICAL
yes
Does your facility manufacture more
than 25,000 pounds of this chemical?
yes
Does your facility process more than
25,000 pounds of this chemical?
yes
Does your facility otherwise use more
than 10,000 pounds of this chemical?
ESTIMATE AND
REPORT RELEASES
FOR TfflS CHEMICAL
DO NOT REPORT
RELEASES FOR
THIS CHEMICAL
Figure 2-1: Flow Chart of TRI Reporting Requirements
35
-------
The second difference is that the reporting thresholds are weakly lower in
Massachusetts. If a facility triggers the federal reporting threshold for at least one listed
chemical (i.e., manufacturers or processes more than 25,000 pounds or otherwise uses more
than 10,000 pounds), the facility must report for all listed chemicals for which total
manufacture, process, and use is greater than 10,000 pounds. Therefore, for most facilities in
Massachusetts there are not three separate reporting thresholds, but one binding threshold
at 10,000 pounds of total use for each listed chemical.
The final difference is that facilities in Massachusetts are required to report chemical
use in addition to chemical release. The reporting thresholds are based on chemical use, but
the federal program does not require public disclosure of use levels. In Massachusetts
facilities must report both total use of the chemical and total releases of the chemical. Figure
2-2 provides a flow chart for the TURA reporting requirements.
This paper is concerned about the effects of the reporting thresholds present in both
the TRI and the TURA designs, on the valid uses of these data for policy analysis. Before
specifically examining the TRI and TURA data and determining whether or not these data
are valid measures of environmental performance, it is worthwhile to articulate a definition
of valid measurement of environmental performance. That is the subject of the next section.
36
-------
Is your fecility a manufacturing facility (SIC 20-39)?
>
yes
Does your facility have 10 or more full time
equivalent employees?
yes
Does your facility manufacture or
process more than 25,000 pounds or
otherwise use more than 10,000 pounds
of any of the listed chemicals?
yes
For Each Listed Chemical Determine:
Is the total amount your facility
yes
manufectures, processes, and otherwise
uses greater 10,000 pounds for this
chemical?
no
no
no
no
DO NOT REPORT
RELEASES FOR
ANY CHEMICAL
DO NOT REPORT
RELEASES FOR
THIS CHEMICAL
REPORT RELEASES
AND TOTAL USE FOR
THIS CHEMICAL
Figure 2-2: Flow Chart of TURA Reporting Requirements
37
-------
3 The TRI/TURA Data as Measures of Environmental Performance
3.1 Defining a valid measure of environmental performance
There are a number of potential characteristics of a valid measure of environmental
performance (National Academy of Engineering, 1999; White and Zinkl, 1997; U.S. EPA,
1992). The most frequently discussed characteristics of a valid performance metric are:
1. Significance: The measure has a clear relationship to environmental
performance.
2. Precision: The stochastic component of the measure is small relative to the
deterministic component.
3. Verifiability: The measure is publicly available and can be validated by a third
party.
4. Comparability: The measure allows for comparisons across facilities, industries,
countries, and across time.
Focusing on the last of these, comparability, there are potentially two standards of
comparability that could be required—cardinal comparability or ordinal comparability.
Cardinal comparability implies that the measure can be compared in magnitude across
different entities. Define releases by facility i to be X. If X is cardinally comparable, Xi>X2
implies that facility 1 is doing better than facility 2 and is, in fact, doing precisely X1-X2
better.5
5 Cardinal comparability is often present in outcome measures used to evaluate program
effectiveness by many other federal and state agencies such as the Department of Health and Human
Services and the Labor Department. If participants in a pilot job training program earn an average of
38
-------
A weaker comparability standard would be ordinal comparability. Ordinal
comparability implies that if Xi>X2 then facility 1 is doing better than facility 2. However, it
is not the case that the magnitude of the difference in performance is captured by the
magnitude of the difference in the performance metric, X. Ordinal comparability is the
minimum comparability standard for a valid performance metric. The whole point of
program or policy evaluation is to ascertain whether a program or policy made society
better off. If the metric used cannot be relied upon to order outcomes from better to worse,
then the metric cannot be relied upon to measure whether a program resulted in a better
outcome. Cardinal comparability is not necessary if the goal of the evaluation is to
determine whether a specific initiative improved outcomes. However, cardinal
comparability is necessary if the goal of the evaluation is to determine how much of an
improvement was obtained by the firm, industry, state, or policy initiative. For example,
cardinal comparability is necessary for cost-effectiveness analyses that compare policies on
the basis of the amount of risk reduction per dollar spent.
The remainder of this paper evaluates the Toxics Release Inventory data as measures
of plant-level environmental performance in relationship to these four characteristics of
validity. In particular the paper highlights the potential threat to validity presented by the
existence of minimum reporting thresholds and attempts to quantify the degree of the bias
created by these thresholds.
3.2 Validity of TRI Data as a Measure of Environmental Performance
There are several known threats to the validity of the TRI data. The first is that data
are only collected for certain industries and chemicals. Originally data were collected for
$25,000 per year and non-participants earns an average of $20,000 per year, then trainees are better off
than non-trainees by exactly $5,000 per year, on average.
39
-------
manufacturing facilities and for approximately 300 chemicals. The fact that the TRI data are
not comprehensive may reduce the significance of these data as a measure of environmental
performance. For the measure to be significant in needs to have a clear relationship to
environmental performance. If the TRI data only capture releases of a small subset of total
chemicals, the significance of these data may be compromised.
This concern was recognized early in the program (U.S. Government Accounting
Office 1991) and the program has since been expanded to more than double the number of
chemicals and to include federal facilities, metal and coal mining facilities, electric
generating facilities that combust coal or oil, chemical wholesale distributors, petroleum
terminals and storage facilities, and hazardous waste treatment storage and disposal
facilities. Obviously the TRI data cannot be used as a valid measure of environmental
performance for plants that are not obligated to report these data. Thus, these data will
necessarily by limited to use only in evaluations that affect these industrial sectors.
A second threat to the validity of the TRI data as a measure of environmental
performance stems from the fact that TRI data are reported in total pounds of a chemical
released, but it is widely acknowledged that pounds of different chemicals present widely
different levels of risk. Similarly, a pound of a chemical released in rural Oklahoma may
have a different impact than a pound of the same chemical released in downtown
Manhattan. Some of the factors that determine the overall environmental risk presented by
a facility's activities include the total pounds of releases, the toxicity of these releases, the
exposure level in the population, the duration of exposure, and the sensitivity of the
population. Simply adding up total pounds of different chemicals released and comparing
across facilities may provide an incorrect ordering of facilities according to relative risk. For
example, a facility that releases a few pounds of a very toxic, persistent, and bio-
40
-------
accumulating toxin may appear to have much better "environmental performance" than a
facility that releases more pounds of a less hazardous and more degradable substance.
Even if the point estimate of environmental performance provided by TRI is biased
by lack of information on toxicity, one might hope that changes in TRI releases still provide a
measure of changes in environmental performance. That is, perhaps time series
comparability is maintained even if cross-sectional comparability is compromised. For
example, even if pounds of different chemicals are not equal in their risk profile, one might
argue that fewer pounds is better than more pounds, holding all else constant. But the key
is that all else must be held constant. The assumption that the trend in TRI releases is a
valid indicator of the trend in environmental performance is likely to hold if, among other
things, the facility consistently uses and reports on the same chemicals in every year. One
change in reporting that may threaten the validity of changes in TRI releases as measures of
changes in environmental performance is if a facility substitutes from a less toxic to a more
toxic chemical (or vice versa). Even if the facility reduces releases by a substantial amount,
the total risk presented by the facility may remain the same, decrease by less, or even
increase. Thus, not adjusting for toxicity compromises the significance and the comparability
of the TRI data both across facilities in a given year and within a facility over time.
EPA and academic researchers have been working to develop indicators that take
the TRI data as an input, combine it with toxicity, exposure, and degradability information,
and then calculate a risk-based metric. EPA has recently released its Risk Screening
Environmental Indicators (RSEI) model, which combines TRI data and risk information to
develop a risk-adjusted pollution measure for peer review (U.S. EPA, 2004). These changes
will doubtless improve the quality of estimates of environmental performance using TRI
data.
-------
A third threat to the validity of the TRI data stems from the fact that these data are
self-reported by facilities. This affects both the precision and the verifiability of the data as
measures of environmental performance. Changes in how a facility estimates releases or
even how it classifies releases may lead to changes in reported releases that are not reflective
of real changes in environmental performance (Graham and Miller 2001). In the initial years
following the establishment of TRI, the accuracy of the self-reported releases was thought to
be poor (U.S. Government Accounting Office 1991); however, most observers believe it has
improved over time. EPA has issued updated guidance on estimating releases both for
different industries and chemicals,6 and it audits a small number of facilities' TRI reports
each year. Despite these efforts, concerns over the precision and verifiability of the self-
reported data remain.
The final threat to the validity of the TRI data is due to truncation in reported
releases caused by the existence of reporting thresholds. It is this threat that this paper
addresses. Reporting thresholds set a minimum level of chemical use at a facility, below
which the facility is not required to report releases. For the TRI data, the reporting
thresholds are 25,000 pounds for manufacture and processing of a listed chemical and
10,000 pounds for any other use of the chemical. Similarly, the TURA data require reporting
only if total use (manufacturing plus processing plus otherwise use) is greater than 10,000
pounds. These reporting thresholds create incidental truncation in the TRI data. This
truncation threatens the significance and comparability of the TRI data for both ordinal and
cardinal analyses. To illustrate the potential bias, define:
ykt = actual releases of chemical c, by facility z, in time t
6 For example, U.S. EPA (2000) provides detailed guidance for the Textile Processing Industry. The
set of all industry and chemical guidance can be obtained online at www.epa.gov/tri/ guided_docs.
42
-------
Mict = amount of chemical c manufactured by facility i in time t
Pict = amount of chemical c processed by facility i in time t
Uict = amount of chemical c otherwise used by facility i in time t
S" = set of all chemicals, c, that are reportable to TRI
For each facility there exists a true aggregate measure of releases, Y, such that
Y„ = .
ceg
However, we do not observe y^tor Yit. Instead, we observe:
Y. = V v. r.
it / J S let let
ceg
where
r» =Hf(A4, >25,000) or (i>„ > 25,000) or ([/„ >10,000)
= 0 otherwise
Further define:
rit = vector of rict (rat, ri2t, ..., riCt) for all c in c ,
uict = 1 if (Mict > o) or (Plct > o) or (Ulct > 0) and
= 0 otherwise
uit = vector of uict (uin, ui2t, ..., uiCt) for all cine.
Under what circumstances can we use observed data on reported releases, Yit, to
make valid cross-section or time series comparisons? Begin with a cross-sectional
comparison. If we observe: Yu>Y2t, under what circumstances can we be confident that
43
-------
Yu > Y2i ? This inference will be valid if both facilities report for all chemicals for which
they use any positive quantity.7 That is if:
rlt = ult and r2t = u2t.
The key point is that if there are some chemicals which a facility uses in positive quantities,
that is Uict>0, but for which the facility is not required to report because ulcl is below the
reporting threshold, then Yit does not necessarily equal Yit and Yit is not necessarily
ordinally comparable across facilities.8
Turning to time-series evaluation, under what circumstances do the trends in reported
releases provide valid information on the trends in actual releases? That is, under what
circumstances does Y], - Y], , = Yit - Y], , ? As with the cross-sectional comparison, time series
comparisons based on reported releases will only provide valid inferences on true releases if
the facility reports for every chemical that it uses in any positive quantity.9 That is, if:
ru = ult and rlt_x = ult_x
Notice that the facility does not need to report for the same set of chemicals in both years. It
just must report for all chemicals it uses in each year. If a facility stops using a chemical
and, hence, stops releasing this chemical, that is a real change in environmental
7 This is the only condition under which this inference is generally valid. However, this inference
may be valid in certain special cases. For example, if every facility does not report for a chemical, c,
but releases of that chemical are the same for all facilities. There is no reason to think that this will
hold.
8 If ordinal comparability is not preserved, then cardinal comparability is also not preserved.
9 This condition allows for valid time series inference in all cases. There may be special cases in
which require less restrictive conditions. For example, if the amount of non-reported releases is
constant across time, then the absence of these releases does not bias the time series comparison.
However, there is no a priori reason for assuming non-reported releases are constant across time.
44
-------
performance. But if a facility simply uses less of the chemical and is no longer required to
report releases, this is not necessarily a real change in environmental performance.
The bias that results from truncation at the reporting threshold is referred to as
truncation bias. The term truncation bias here has a slightly different meaning than the
classic econometric definition. The truncation bias in this paper most closely resembles
incidental truncation bias, which arises when facilities or individuals are observed on the
basis of the outcome of another decision (Wooldridge, 2002: 552). For example, imagine
one only observes wages for the employed, so observing wages is the result of another
decision, in this case the labor market participation decision. The truncation bias in the TRI
data is incidental truncation bias at the unit for which the data are collected, which is the
facility-chemical-year level. One only observes a facility-chemical-year record if the facility
triggers the reporting threshold for that chemical in that year. So observing data is the result
of the chemical use decision. This incidental truncation bias is further aggravated by
aggregation to other levels of analysis, such as the facility, firm, industry, or state level. The
aggregation essentially treats all unobserved data as zero, which further invalidates
comparisons. Perhaps a more specific name for this bias is "truncation and aggregation
bias," but it will be referred to here as truncation bias for short.
While the first three threats to the validity of TRI data are well recognized and
measures have been taken to address the threats, little has been done to assess or reduce the
problems associated with truncation bias. This lack of attention is not the result of a lack of
understanding of the problem, rather it is largely due to the fact that data have not been
available that would allow analysts to estimate the extent of the problem and correct it. In
contrast, the bias introduced by not weighting TRI pounds by some measure of toxicity is
conceptually easier to address because risk factors can be determined with available data
45
-------
and the TRI data can be corrected using these risk factors. However, since facilities are
generally not required to report chemical use levels, one cannot observe whether a facility is
reporting for all chemicals it uses in positive quantities. Thus, the truncation bias cannot be
corrected for in a systematic way using available data.
While systematic correction of the truncation bias cannot be obtained, this paper
focuses on estimating bounds on the degree of bias presented by truncation of the data at
the reporting thresholds. A reasonable question might be: How much of the observed
decrease in TRI releases is potentially due to truncation bias? Similarly, for a cross-section
one can ask how much the relative rankings of facilities could change if truncation bias were
to be accounted for. A better understanding of the approximate magnitude of the bias
introduced by the reporting thresholds can help ascertain whether this bias is a practical,
rather than purely theoretical, threat to the validity of the TRI data as a measure of
environmental performance.
4 Estimating Bounds on the Truncation Bias
There are several limitations in the TRI data that inhibit systematic estimation of the
degree of potential truncation bias. First, the researcher only observes whether a facility
reports releases of a chemical in a given year. If a facility does not report for a chemical in
year t for which it reported in year t-1, the researcher cannot determine if the facility
eliminated use of the chemical, substituted to a different chemical, or was below the
reporting threshold.10
10 Facilities that report in one year and then cease reporting in future years represent only a fraction of
the truncated observations. There may also be facilities that use a chemical in every year below the
reporting threshold, have releases of these chemicals, but are never legally required to report. This
non-reporting is equally problematic for policy analysis, but little can be done to identify facilities for
which this may be true.
46
-------
In contrast, the TURA data provide two sources of variation that better allow for the
identification of bounds on the truncation bias. First, if a facility does not report for a
chemical in year t for which it had reported in year t-1, the TURA reporting forms include an
optional question that asks the facility to explain the change. Approximately, one-third of
facilities that cease reporting for a chemical answer the optional question. I check to see if
the facilities that respond to this optional question are representative of the set of facilities
that cease reporting and find evidence that responders are not systematically different from
non-responders. Thus, the responses to the optional question are used to gauge the degree to
which facilities are not reporting for chemicals they use in positive quantities.
The second source of variation in the TURA data that can be used to help estimate
bounds on the truncation bias is that facilities are required to report how much chemical
they use in addition to how much they release. The federal TRI has reporting thresholds
based on use, but only requires reporting on releases. The use data combined with the
responses to the optional question about why facilities ceased reporting reveal that the
distance from the reporting threshold is a good predictor of whether a facility ceases
reporting because it went below the reporting threshold, but still used the chemical in
positive quantities. I use this relationship to predict for non-responders the reason why
they ceased reporting.
For facilities that directly reveal that they ceased reporting because they went below
the reporting threshold, and for facilities that are predicted to have ceased reporting because
they went below the reporting threshold, I then estimate the effect of these missing releases
on trends in releases over time and on cross-sectional comparisons of facilities within a
given year. Missing releases are estimated using three different procedures. The first is a
lower bound estimate on total releases at the facility. The lower bound estimate assumes
47
-------
that when a facility ceases reporting its true releases are zero. This is the implicit
assumption currently made by government agencies, non-profits, and academic researchers
when aggregating releases data to the facility level (or higher levels). The second estimate
of missing releases can be considered an upper bound estimate. In the upper bound
scenario, if a facility ceases reporting a chemical because it went below the reporting
threshold, releases are set equal to the most recent level of releases reported for that
chemical at that facility. Moreover, the facility is assumed to continue releasing the
chemical at the same level in perpetuity. The final estimate of missing releases is one that
extrapolates the value of non-reported releases based upon linear trends in reported
releases. These three scenarios present an upper and lower bound and an intermediate
estimate of the degree of bias introduced by the reporting thresholds.
Section 4.1 examines why facilities claim they ceased reporting in Massachusetts
using the optional question from the TURA form. This section also discusses the estimation
used to predict reasons for non-reporting for facilities that did not answer the optional
question. Section 4.2 discusses the estimation of the magnitude of missing releases in
Massachusetts based on three estimation methods. Section 4.3 discusses the implications of
the Massachusetts results for national analyses.
4.1 Why do facilities cease reporting?
Facilities in Massachusetts were required to disclose pollution and chemical use data
to the TURA program beginning in 1990. From 1990 to 1999 there were a total of 23,200
chemical reports filed by 1,092 facilities. During this same time period there were 3,758
cases where a facility reported for a chemical in one year, but did not report the chemical in
48
-------
the following year.11 For these facilities, the TURA form provides an optional question
where the facility can explain why it is no longer reporting for this chemical. The question
is multiple-choice with the following six possible responses: (1) chemical use is below the
reporting threshold but greater that zero, (2) chemical was not used this year, (3) substituted
a different chemical, (4) chemical use eliminated without substitution, (5) decline in
business, and (6) other. If the facility answers other, they are given the option to fill in a
reason. This question is answered for a total of 1,271 (or 33.8 percent) of the cases where
reporting ceases. Table 2-1 provides the distribution of responses.
Table 2-1: Explanation for Non-Reporting Among Respondents to the Option
Question in Massachusetts
Reason for No Longer Number of Respondents Percentage of All Non-
Reporting missing Responses
Chemical use is below the 844 66.40
reporting threshold but
greater than zero
Chemical was not used this 96 7.55
year
Substituted a different 87 6.85
chemical
Chemical use eliminated 60 4.72
without substitution
Decline in business 31 2.44
Other 153 12.04
Approximately two-thirds of all respondents to the optional question answered that
they were no longer reporting because their chemical use was below the reporting
threshold, but greater than zero. Of the respondents that answered "other" the most
11 It is possible that the facility reported that chemical again in future years. This occurs 379 times.
49
-------
frequent explanation was that the chemical in question had been delisted by the state and
reporting was no longer legally required.
The high percentage of respondents that state they ceased reporting because their
chemical use was positive, but lower than the reporting threshold, raises concern that the
degree of truncation bias may not be trivial. However, before estimating bounds on the
truncation bias, it is necessary to determine whether the facilities that responded to the
optional question are systematically different from those that did not respond. If
responders are systematically different from non-responders in ways that may also be
correlated with their reason for not reporting, then the sample of responders cannot be used
to impute explanations for non-reporting. To see whether the responders to the optional
question are a representative sample, I compare the distribution of the data for the two
groups for three key variables — two-digit SIC code, year reporting stopped, and total
releases to the environment in year before reporting stopped. Table 2-2 presents the
distribution of SIC codes, Table 2-3 presents the distribution of years, and Table 2-4 presents
the mean and standard deviation for total pounds released.
50
-------
Table 2-2: Distribution of Standard Industrial Codes by Optional Question
Responders and Non-Responders
Percentage of Total
SIC code
Respondents to Optional
Non-Respondents to
Question
Optional Question
17
0.24
0.04
20
1.83
1.04
22
4.54
4.23
23
0.48
0.41
24
0.08
0.29
25
1.27
1.41
26
5.74
4.39
27
1.83
1.49
28
21.83
21.72
29
0.08
0.17
30
5.10
7.21
31
0.80
1.66
32
0.88
0.99
33
6.22
7.54
34
15.30
14.17
35
2.15
3.69
36
11.24
10.48
37
3.03
2.61
38
6.06
4.77
39
3.51
2.69
45
0.00
0.21
47
0.00
0.08
49
2.79
4.85
51
4.38
2.90
72
0.56
0.70
73
0.00
0.04
75
0.00
0.21
76
0.08
0.04
Bold SIC codes indicated that the difference in percentages is statistically significant at
the 5% level. The t-test assumes unequal variances across groups.
51
-------
Table 2-3: Distribution of Years When Reporting Ceased by Optional Question
Responders and Non-Responders
Year Reporting Ceased
Percent of Total
Respondents to Optional
Question
Non-Respondents to
Optional Question
1990
10.96
11.76
1991
11.36
8.62
1992
15.06
13.05
1993
14.67
9.51
1994
13.56
15.31
1995
8.60
10.68
1996
8.99
8.90
1997
7.57
7.98
1998
9.23
14.18
Bold SIC codes indicated that the difference in percentages is statistically significant at
the 5% level. The t-test assumes unequal variances across groups.
Table 2-4: Distribution of Reported Pollution Releases by Optional Question
Responders and Non-Responders
Mean Standard Error
Respondents to Optional 16,749 4,333
Question
Non-Respondents to 18,102 1,443
Optional Question
Difference between 1,353 4,567
Respondents and Non-
Respondents
A t-test on the difference in average pollution releases between the two groups cannot
reject the null hypothesis that this difference equals zero. The t-statistic is 0.30 allowing
the variance between the two groups to be unequal.
The data suggest that responders to the optional question are not systematically
different from non-responders at least on total pounds of chemicals released. The difference
in the average release levels is 1,353 pounds with a standard error of 4,567 pounds. There
are some systematic differences in industry and year reporting stopped. Of the 28 SIC
codes, the two groups — responders and non-responders — are statistically different for 7 of
52
-------
them.12 There are also some differences among responders and non-responders in the years
for which reported ceased. Of the nine years, the distribution of responders and non-
responders differs in four years.13 In addition, one may be concerned that there are
unobservable differences between the responders and the non-responders that also are
correlated with whether the facility goes below the reporting threshold but still uses the
chemical in positive quantities. In the absence of a valid instrument that can explain
response to the optional question, but not explain going below the threshold, specific
sample selection correction models cannot be employed. Rather in this section, I impute
reasons for non-reporting for those that did not answer the optional question based on data
from the sample of facilities that answered the optional question assuming that responders
are reasonably representative on non-responders. I then conduct sensitivity analysis on
these results which is presented in Section 2.4.3.
There are several ways one might impute these data. A common method is known
as "hot deck" (Ford 1983; Little and Rubin 1987). Essentially, hot deck is a matching
strategy—find a facility that responded to the question that looks like a facility that did not
respond and assign the matching responders value to the non-responder. A variant of hot
deck is to use regression to estimate a relationship between facility characteristics and the
explanation for ceasing reporting for those that respond to the optional question, and then
using this regression function, predict for non-responders what their explanation would
have been.14 This regression-based imputation strategy is employed here.
12 A Pearson's Chi-squared test rejects equality of the industry distributions across the two groups.
13 A Pearson's Chi-squared test rejects equality of the industry distributions across the two groups.
14 These two strategies differ in the degree that matching is "enforced" and the functional form
assumption. Hot deck is a non-parametric strategy that does not impose a specific functional form on
the relationship between the covariates and the response variable. Regression is loose matching, but
53
-------
To construct the prediction relationship, I first create a new dummy variable that
takes a value of one if the facility ceased reporting because it went below the reporting
threshold and zero otherwise. I then use this variable as the dependent variable in a logit
estimation on observable facility characteristics. The task is then to compile a set of
observable characteristics that explain whether a facility will go below the reporting
threshold.
One characteristic that may explain the propensity to go below the reporting
threshold is the facility's distance from the reporting threshold. Degeorge, Patel and
Zeckhauser (1999) demonstrate that the existence of performance thresholds for managers
induces specific changes in their earnings management, with managers managing to the
thresholds. For example, empirically there appears to be a concentration of profits just
above zero for managers that are compensated based on whether their unit earns positive
profits. Degeorge, Patel and Zeckhauser refer to this as threshold-regarding behavior.
Threshold-regarding behavior is particularly pronounced among units that are very close to
the profit threshold. In other words, if your unit is quite far from earning positive profits,
you do not try to manipulate earnings much, because no amount of manipulation will cause
your unit to earn positive profits. But, if your unit is very close to the positive profit
threshold, manipulation of earnings is more valuable.
In the context of the TURA reporting thresholds, we might expect to see similar
threshold-regarding behavior. Facilities that are very close to reporting threshold in year t
have a greater incentive to manage their chemical use so that they fall below the reporting
threshold in t+1. To measure the distance from the reporting threshold, I construct a
does impose a functional form. The regression-based matching will perform well when the assumed
parametric specification is a good approximation to the average response function.
54
-------
variable that measures, for each chemical, how far the facility is from each of the reporting
threshold. In Massachusetts, this process is simplified by the fact that once the facility
triggers the reporting threshold for one chemical, the reporting threshold for all chemicals is
10,000 pounds of combined manufacture, process, and other use amounts. So in
Massachusetts there are not three separate reporting thresholds, but one binding threshold
at 10,000 pounds of total use. The facility's distance from the reporting threshold is then
given by total use minus 10,000 pounds. The hypothesis is that the greater the distance from
the reporting threshold, the less likely a facility is to cease reporting that chemical because
its use of the chemical went below the reporting threshold.
Similarly, if there is a relationship between chemical use and chemical release, then
total releases of the chemical in time t may predict whether the facility goes below the
reporting threshold in year t+1. Thus, total releases are also used as an explanatory
variable. Other potential explanatory variables include industry dummy variables, and year
dummy variables that proxy for changes in industry best practices and exogenous
technological change. The relationship estimated is then given by:
below threshold
ro^
vly
= a + /?, Distance from Threshold + /?2releases + PslcSIC + PtYear + s
The results of this estimation are present in Table 2-5.
55
-------
Table 2-5: Results for Estimation of Threshold Logit for
Respondents to Optional Question
Coefficients
Percent Increase in Baseline
Probability Resulting from
a 10% Decrease from Mean
Constant
1.02
(1.19)
NA
Distance from Threshold
-0.00001 ***
(0.000003)
6.5%
Releases
-0.00003 ***
(0.000006)
3.1%
SIC dummies
Yes +
NA
Year dummies
Yes +
NA
Number of observations
Pseudo R-squared
Baseline Probability
1,251
0.17
60.0 %
Baseline Probability is the probability evaluated at the mean value of all continuous
explanatory variables and at zero for all binary variables.
*** Significant at the 1% level
+ An F-test shows variables are jointly significant at the 1% level
Because the logit estimation is non-linear, the coefficients presented do not convey
information about the marginal effect of a change in one of the explanatory variables on the
dependent variable. To determine what the magnitude of the effect of each explanatory
variable is on the probability that a facility ceases reporting because it goes below the
reporting threshold, I first calculate the baseline probability—that is the probability of going
below the threshold predicted by the logit equation when all of the continuous variables are
set at their mean value and all of the binary variables are set at zero. The baseline
probability evaluated at the mean is 60.0%. I then decrease each covariate in turn by 10%
from its mean value and report the percentage point change from the baseline. Using this
method, a 10 percent decrease from the average distance from the reporting threshold
56
-------
results in a 6.5 percentage point increase in the probability of going below the reporting
threshold. Similarly a 10 percent decrease in total releases to the environment results in a
3.1 percentage point increase in the probability of going below the reporting threshold.
This probability function can then be used to predict whether facilities that ceased
reporting for a chemical, but did not explain why, were likely to have ceased reporting
because they went below the reporting threshold. The above equation was used to predict
the probability of going below the threshold for the 2,482 observations with no answer to
the optional question. Observations with a predicted probability greater than 50 percent
were coded as going below the threshold. Table 2-6 provides a breakdown of the
observations that ceased reporting by explanation. Of the 2,482 observations for which no
explanation for non-reporting was provided, 1,786 (72.0 percent) were predicted to have
stopped reporting because the facility went below the reporting threshold for that chemical,
but still use the chemical in positive quantities.
Table 2-6: Distribution of Observations (Facility-Chemical-Year) that
Cease Reporting by Explanation
Explanation Explanation Total
Provided by Facility Predicted
Below the 844 1,786 2,630
Reporting
Threshold
All Other Reasons 424 696 1,120
Total 1,268 2,482 3,750
In summary, analysis of the Massachusetts TURA data indicates that a substantial
percentage of facilities that cease reporting a chemical do so because they go below the
reporting threshold for that chemical, but still use it and may still have positive releases of
these chemicals. While this frequency of threshold-regarding behavior seems to present
some concern about the validity of the TRI data for making comparisons among facilities or
57
-------
across time, the level of concern may still be low if the total amount of releases that
disappear from the registry is small. That is, perhaps the percentage of observations
affected by the reporting thresholds is large, but the total share of releases represented by
these observations is small. The next section addresses the question of the likely magnitude
of the truncation bias.
4.2 What is the Magnitude of Missing Releases in Massachusetts?
The difficulty in assessing the effect of truncation at the reporting thresholds on the
validity of the TRI data is that it is impossible to know how large a problem non-reportable
releases are, precisely because these releases are no longer reported. The best we can do is
assess how large this problem might reasonably be, given observable information. To that
end, this section focuses on estimating the magnitude of "missing" releases by estimating
bounds on the possible size of these releases —a lower bound estimate, an upper bound
estimate, and a linear projection estimate.
To make these results meaningful, I restrict the analysis to chemicals and industries
that have been subject to the TURA reporting requirements since its inception — the so-called
core group. This ensures that we are examining variation in reported releases due to
facility behavior around the thresholds and not changes in releases due to changes in the
regulatory requirements themselves. The analysis is also done for two different measures of
pollution from the TRI. The first is on-site releases. On-site releases are releases of the
pollutant to air, water, land, or under-ground injection that occur at the facility's location.
The second measure is total on- and off-site releases. Total releases include on-site releases,
but add transfers of waste to offsite locations for disposal or recycling. In general, it is
thought that total releases is over-inclusive as a measure of environmental performance
58
-------
because some of the transfers are not pollution, but are transfers for recycling and reuse.
On-site releases, however, are under-inclusive as a measure of environmental performance
because some waste that is transferred off-site is pollution attributable to the facility.
Researchers have used both on-site and total releases as measures of environmental
performance, so it important to see if the reporting thresholds have different effects for the
two measures.
The lower bound estimate of the value of non-reported releases is that these releases
are zero. This assumes that once a facility drops below the reporting threshold for a given
chemical, they no longer release any of that chemical. A conservative upper bound estimate
of the value of non-reported releases is that they equal the last reported value of releases for
that chemical at that facility. Thus, if a facility reports releases of 500 pounds of a chemical
in year t and then drops below the reporting threshold, the upper bound estimate of missing
releases is that this facility releases 500 pounds a year of that chemical in perpetuity.15
One might argue that assuming unobserved releases are either zero or set at their
most recent value in perpetuity are extreme assumptions. There is some evidence that
suggests that, on average, setting releases equal to the last reported value is not as extreme
an assumption as it may first appear. There are 287 observations (1 percent of the total)
where the facility stops reporting for a chemical in one year because it went below the
reporting threshold and then in the future the facility begins reporting for that chemical
again. How do releases of the chemical in future years compare to the reported releases in
the last reported year? On average, the future reported releases are 796 pounds greater than
15 Occasionally a facility will report for a chemical for a few years, then stop reporting for that
chemical because they are below the reporting threshold, and then begin reporting for the chemical
again in later years. In this case, the upper bound estimate of releases equals observed releases in any
year in which the facility actually reports releases and in non-reporting years are set equal to the
releases in the most recent year in the past for which the facility reported releases for that chemical.
59
-------
the last reported releases for that chemical.16 Of course, this does not rule out the
possibility that assuming releases are constant for non-reporting facilities is a conservative
upper bound. Facilities that waver above and below the reporting threshold are distinct
from facilities that go and remain below the reporting threshold. Thus, we might not expect
the release behavior of the facilities that oscillate around the reporting threshold to be
indicative of the release behavior of facilities that go below the reporting threshold and stay
there forever. But this evidence does suggest that using the last reported releases as an
estimate of future reported releases might be a reasonable upper bound.
Providing upper and lower bounds on the potential bias in reported releases
induced by truncation at the reporting thresholds provides useful information on how large
the bias may be. However, it does not provide any information on the probability
distribution of the true bias within that range. While this range is useful, perhaps, we
would wish to also have an estimate of bias under a more probable scenario. An alternative
assumption about the behavior of releases among non-reporting facilities is to assume that
facilities do not fundamentally behave in different ways with respect to their release
decisions based on whether they are above or below the reporting threshold. If this is the
case, then the trends in non-reported releases might be expected to increase or decrease at
the same rate as the trends in observed releases. Under this assumption, one can project
trends in releases for non-reportable data based on observed trends in releases among
reported chemicals.
How can we predict the trend in reported releases so that it can be extrapolated to
non-reporting facilities? One might hypothesize that the amount of the chemical used, the
16 The largest future drop in reported releases is by 63,571 pounds arid the largest future increase in
reported releases is 105,600 pounds.
60
-------
type of chemical, the industrial sector, and other similar factors, would all be good
predictors of how much chemical a facility will release in a given year. While all of these
factors do explain chemical releases, taken together they only explain about seven percent of
the variation in reported releases.17 The best predictor of reported releases turns out to be a
distributed lag of past releases. A simple one period distributed lag, where current releases
of each chemical are regressed on releases of that chemical from the previous year, explains
80% of the variation in releases. Increasing the number of lags increase the predictive power
slightly, but makes the equation less useful for prediction because a smaller number of
observations have multiple lags. The results for one-, two-, and three-period distributed lag
models are presented in Table 2-7. Using the one-period distributed lag model, current on-
and off-site releases are, on average, 95 percent of previous years' releases. Similarly, using
a one-period distribution lag model, current on-site releases are 80 percent of previous
years' releases.
17 This estimate comes from a regression of releases on total chemical use with two-digit SIC code,
chemical, and year fixed effects. Including facility-chemical fixed effects increases the predictive
power of the regression to 65 percent. However, the distributed lag model, which yields a higher
predictive power, also is preferred because all of the data are observable. That is, it relies only on
past releases, which we observe for all facilities that cease reporting.
61
-------
Table 2-7: Distributed Lag Models of Reported Releases
Total Releases (On- and Off-site)
On-site Releases
Specification
1
2
3
1
2
3
1 year lag
0.947 ***
0.865 ***
0.872 ***
0.798 ***
0.740 ***
0.737 ***
(0.054)
(0.131)
(0.149)
(0.036)
(0.084)
(0.115)
2 year lag
0.105
0.086
0.087
0.130
(0.134)
(0.143)
(0.068)
(0.089)
3 year lag
0.017
-0.021
(0.083)
(0.046)
Observations
8586
6572
4964
8927
6952
5330
Adjusted
0.779
0.795
0.804
0.782
0.807
0.808
R-squared
*** indicates the coefficient is statistically significant at the one percent level.
Using the coefficients from the one-period distributed lag model, I predict releases
for each facility that ceases reporting. This generates trends for facilities that cease
reporting. These projected trends are downward sloping for all facilities that cease to report
for a chemical, and the rate of decrease is larger for on-site releases than for total releases.
Figure 2-3 diagrams the process of estimating lower, upper, and linear projected releases for
a facility and Table 2-8 provides data on the magnitude of missing releases, both total
releases (on and off-site) and on-site releases, using both the upper bound assumption and
the linear projection.
62
-------
On-Site Releases
Time t
Time t+1
Uses chemical c in quantity
greater than 10,000 pounds
Reports On-Site Releases =
OSR,
Does not report for
chemical c.
Estimated on-site releases set equal to:
0 for the lower bound scenario
OSRct for the upper bound scenario
OSRct*0.80 for the linear projection
Estimated on-site releases set equal to: 0
for all scenarios
Total Releases
Uses chemical c in quantity
Xime t greater than 10,000 pounds
Reports Total Releases = Rt,
Time t+1
Does not report for
chemical c.
Estimated total releases set equal to:
0 for the lower bound scenario
Rct for the upper bound scenario
Rct*0.95 for the linear projection
Estimated total releases set equal to: 0
for all scenarios
Figure 2-3: Flow Chart for Imputation of "Missing" Releases
63
-------
Table 2-8: Estimates of TURA Releases in Massachusetts (in 1,000s of pounds)
Total Reported Releases
On-Site Releases
(On- and Off- Site)
°/
/o
Percent
Difference
Difference
Between
Between
Lower and
Lower and
Lower
Linear
Upper
Upper
Lower
Linear
Upper
Upper
Year
Bound
Projection
Bound
Bounds
Bound
Projection
Bound
Bounds
1991
43,294
45,228
45,337
4.7%
17,600
18,870
19,191
9.0%
1992
44,842
48,301
48,774
8.8%
15,563
17,495
18,296
17.6%
1993
38,215
44,110
44,427
16.3%
11,857
14,640
16,060
35.4%
1994
39,052
46,131
46,345
18.7%
9,299
12,158
14,258
53.3%
1995
36,538
44,265
44,845
22.7%
8,711
11,264
13,928
59.9%
1996
36,812
44,852
45,935
24.8%
7,449
9,657
13,134
76.3%
1997
32,430
40,614
41,903
29.2%
6,357
8,337
12,318
93.8%
1998
33,026
41,623
43,495
31.7%
5,599
7,364
12,050
115.2%
1999
19,955
28,778
31,333
57.0%
5,117
6,689
11,966
113.8%
64
-------
The results indicate that the degree of missing releases in the early years is
relatively modest as a percentage of total releases. In 1991, missing releases are only
about four percent of total (on- and off-site) releases and between seven and nine
percent of reported on-site releases. However, over time, missing releases as a
percentage of total reported releases rise dramatically. There are two reasons for this.
First, missing releases are cumulative. In every year, about two to four percent of the
previous year's releases are not reported due to facilities going below the reporting
threshold. But in each year the total amount of missing releases are all those releases
that are no longer being reported by all facilities, these include the releases from facilities
that stopped reporting for the chemical this year, as well as the missing releases from
facilities that went below the reporting threshold two years ago, three years ago, and so
forth. Particularly in the upper bound scenario, when facilities that drop below the
reporting threshold are assumed to continue releasing at the same level in perpetuity,
the cumulative effect can be quite large.
The second explanation is that reported releases fall considerably over time, even
for facilities that continue to report for all chemicals. For the upper bound case, where
non-reporting facilities are assumed to continue to release at the same level forever, the
relative importance of these releases increases over time — the missing releases stay the
same, but the total reported releases decrease — resulting in a sharp increase in missing
releases as a percentage of total reported releases. In fact, for on-site releases the
difference between the upper and lower bound estimates of releases differs by 134
percent in 1999.
The data in Table 2-8 illustrate that the magnitude of missing releases generated
by the existence of the reporting threshold is non-trival. But does this result in
65
-------
significantly biased estimates in the trend in environmental performance over time?
Figures 2-4 graphs the trend in total releases over time for each of three scenarios: (1)
only reported releases (lower bound estimated assumes non-reported releases equal
zero), (2) reported releases plus the upper bound estimate of missing releases (sets
missing releases equal to their last reported value for all years), and (3) reported releases
plus linearly extrapolated estimates of missing releases.
Looking at total releases first, it is clear that the estimate of the trend in
environmental performance is substantially affected by the exclusion or inclusion of the
estimated releases for non-reporting facilities. Using a lower bound assumption, that all
non-reported releases are zero, the change in total releases from 1990 to 1999 is 36.7
percent. However, if one instead uses the upper bound assumption, that facilities that
fall below the reporting threshold continue to release the same amount forever, the
change in total releases over the same ten year period was only 0.6 percent. Using the
more moderate assumption that facilities that are no longer reporting continue to
decrease releases over time at the same rate as the average reporting facility, the change
in total releases over the ten-year period is 8.7 percent.
66
-------
60,000
~ Total Reported Releases (Lower Bound)
-Total Upper Bound Releases
- - ilr - -Total Projected Releases
50,000
40,000
1990-1999
0.6% decrease
30,000
* 1990-1999
8.7% decrease
20,000
1990-1999
36.7% decrease
10,000
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
Year
Figure 2-4: Trends in Total (On- and Off-Site) Releases in Massachusetts
-------
The story is somewhat less stark, but still significant, if one ignores the upward
trend in reported off-site releases between 1990 and 1991 and instead examines trends
only for the years 1991 to 1999. For this period, the lower bound assumption leads to the
conclusion that total releases have fallen by 53.9 percent, the upper bound assumption
estimates the decline at 30.9 percent, and the linear projection leads to an estimate of a
36.4 percent decrease. Using the difference in trends from 1991 to 1999, it appears that in
the worst case, non-reported releases by facilities that fall below the reporting threshold
may account for approximately 23 percentage points of the total 54 percentage point
decrease in reported releases, or roughly 40 percent.
The potential degree of bias introduced by the reporting thresholds in the trends
for on-site releases is less pronounced, although still sizeable. Figure 2-5 provides the
estimated trends in on-site releases. Using the lower bound assumption, the decrease in
on-site releases from 1990-1999 was 75.7 percent. Using the upper bound assumption,
the decrease over this period was 43.1 percent. Finally using the assumption of linearly
projected decreases in non-reported releases, the change over the decade was 68.2
percent. Thus, of the observed 75 percent decline in reported releases, from 1990-1999 as
much as 32.5 percentage points (or 40 percent) of this decrease may be accounted for by
non-reported releases.
68
-------
25,000
~ Total Reported Releases (Lower Bound)
-Total Upper Bound Releases
- - ilr - -Total Projected Releases
20,000
15,000
1990-1999
43.1 % decrease
A.
10,000
1990-1999
68.2% decrease
1990-1999
75.7% decrease
5,000
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
Year
Figure 2-5: Trends in On—Site Releases in Massachusetts
69
-------
As evidenced by these data, our understanding of the magnitude of the change
in environmental performance is substantially affected by the truncation bias introduced
by the reporting threshold. Making different assumptions about what happens to
releases when facilities cease to report because they fall below the reporting threshold
makes a big difference in our understanding of the level of environmental improvement.
Having said that, it is somewhat reassuring that even in the upper-bound scenario, on-
site releases of toxic chemicals have fallen by over 43 percent in Massachusetts. In this
case, using TURA data to conduct policy analysis is very likely to result in erroneous
cardinal estimates of the magnitude of improvement, but unlikely to result in erroneous
ordinal estimates — things have improved by somewhere between 43 and 76 percent.
A separate concern is whether the reporting thresholds induce bias in cross-
sectional comparisons of facilities. Imagine that we rank facilities from lowest releases
to highest releases. If the degree of non-reported releases is small relative to any given
facility's total reported releases, adjusting for these non-reported releases will change
our estimate of how much a facility releases, but may not change the facility's rank
much. If on the other hand, the magnitude of missing releases is large relative to total
reported releases for some set of facilities, then adjusting facility totals to account for this
may substantially changing the rankings of facilities.
To investigate the degree of cross-sectional bias potentially introduced by the
reporting thresholds, I first divide facilities into quartiles based on their reported
releases. Thus, facilities whose reported releases are in the lowest 25 percent of all
facilities in that year are assigned to quartile number one. Facilities whose releases are
among the highest 25 percent of all facilities are assigned to quartile number four.
These are the quartile rankings associated with the lower bound assumption that all
70
-------
non-reported releases are equal to zero. I then recalculate the facility's quartile ranking
based on the upper bound assumption and the linear projection. If the relative
importance of the non-reported releases is small, we would expect to see few facility
quartile rankings change. In this case we would not be terribly concerned about the
potential cross-sectional bias introduced by the reporting thresholds. However, if the
relative importance of missing releases is large, we might expect to see more facility
quartile rankings change dramatically. An alternative way to think about this is to
imagine a policy maker assigning a facility a grade based on its pollution levels relative
to other facilities in a given year. A facility can get a grade of excellent, good, fair, or
poor. If we adjust for missing releases resulting from facilities going below the
reporting threshold, do we change our grades for only a few facilities or for a substantial
number of facilities? The results of the quartile distribution comparison are provided in
Tables 2-9 and 2-10 for total and on-site releases, respectively.
Table 2-9: Quartile Ranking Comparisons for Total Releases
Upper Bound Ranking
1
2
3
4
Lower
1
1,046
960
794
277
Bound
2
349
146
33
6
Ranking
3
256
472
501
105
4
0
67
321
1,257
71
-------
Table 2-10: Quartile Ranking Comparisons for On-Site Releases
Upper Bound Ranking
1
2
3
4
Lower
1
1,279
644
865
515
Bound
2
232
173
30
5
Ranking
3
171
624
338
71
4
0
177
414
1,054
In each table, the diagonal elements (in bold) represent facilities whose quartile
ranking is unaffected by inclusion of estimates of their missing releases. All off-diagonal
entries are facilities whose quartile rankings differ under the upper and lower bound
assessments. For total releases, 45 percent of facilities have different quartile rankings
under the upper bound assessment than under the lower bound assessment. For on-site
releases 43 percent of facilities have different quartile rankings under the two different
scenarios.
For the purposes of using TURA or TRI rankings for regulatory targeting
purposes, perhaps one is most concerned about errors in the top and bottom quartile.
That is, one is concerned most with mislabeling a facility as "poor" or "excellent."
Focusing on these quartiles, we can see from the data in Table 2-9 that 24 percent of the
facilities that are labeled "poor" based on their reported releases would have been
labeled either "fair" or "good" once reported releases are adjusted to include an upper
bound assessment of missing releases. In addition, 66 percent of facilities that were
labeled "excellent" using only reported releases would have been labeled "good," "fair,"
or "poor" if missing releases are included. The results for on-site releases are quite
similar (Table 2-10). For on-site releases 36 percent of facilities labeled "poor" using
72
-------
reported releases should have received a higher grade and 61 percent of facilities labeled
"excellent" should have received a lower grade.
This potential variation in ordinal rankings of facilities within a given year is
large enough to be of substantial concern. Indeed, the degree of variation in the cross-
section rankings of facilities seems more troubling than the variation in trends over time.
In the trends, we were confident in the direction of the change (there had been an
improvement), but not in the magnitude of the change. With the cross-sectional rankings
we are potentially assigning the wrong grade two-thirds of the time in the best quartile
and a quarter of the time within the worst quartile. If regulatory or enforcement resources
are targeted based on rankings of reported releases, these resources are likely to be
misallocated.
4.3 Sensitivity Analysis
One of the key components of the above analysis was imputing whether or not a
facility ceased reporting because it went below the reporting threshold if the facility did
not answer the optional question. As discussed in Section 2.4.1, there were some
observable differences between facilities that responded to the optional question and
those facilities that did not respond. How sensitive are the findings to changes in the
imputation method? I try two alternative imputations and provide information on the
impact of these changes on the estimates of truncation bias both for trend and cross-
sectional analyses.
The first sensitivity analysis is to change the probability cutoff from the logit that
determines whether a facility that does not answer the optional question stopped
reporting because it went below the reporting threshold. In the main analysis, facilities
with a predicted probability of going below the threshold greater than 50 percent from
73
-------
the logit were assumed to have ceased reporting because they went below the reporting
threshold. The first test is to increase this probability cutoff to 75 percent.
The second sensitivity analysis is to only examine missing releases for facilities
that actually respond to the optional question and state that they cease reporting
because they went below the reporting threshold, but still use the chemical in positive
quantities. No imputation is made for facilities that do not answer the optional
question. This is essentially the same as increasing the probability cutoff from the logit
to 100 percent.
Table 2-11 presents the results of the sensitivity analysis for estimates of the
effect on the trend in total and on-site releases. For comparison, the first column
contains the original estimates (where the probability cutoff equals 50 percent). Under
the most conservative scenario, that only used data from facilities that actually answered
the optional question, up to 15 percent of the observed decrease in total releases and up
to 17 percent of the observed decrease in on-site releases from 1991-1999 could be due to
missing releases.
Table 2-11: Maximum Percent of Observed Decline in Releases from 1991-
1999 That Could be Explained By Threshold Regarding
100%
50%
75%
(responders only)
Total Releases
42.7
23.8
14.8
On-Site Releases
46.9
27.2
16.7
74
-------
Sensitivity analysis was also done on the cross-sectional rankings. Table 2-12
presents the results for the differences between the upper bound and lower bound
quartile rankings for all three modifications to the probability cutoff (75%, 100%, and
expected value). The first column of the table provides the main findings based on a 50
percent cutoff for comparison. In the most conservative scenario, the quartile rankings
are wrong about 25 percent of the time. The rankings are only wrong about seven
percent in the bottom quartile. That is only about seven percent of facilities that would
have received a label of "poor" using the reported data should have received a higher
grade if missing data were incorporated. In the upper quartile the percent error is still
higher than average. Approximately 33 percent of facilities that would have been
labeled "excellent" based on reported releases would have received a lower ranking if
missing releases had been accounted for.
Table 2-12: Maximum Potential Error in Cross-Sectional Rankings Due to
Threshold Regarding
50%
75%
100%
(responders only)
Total
45%
42%
25%
Bottom Quartile
24%
7%
7%
Top Quartile
66%
56%
33%
Overall the sensitivity analysis supports the general findings presented in the
main analysis. First, a potentially significant share of the decrease in observed releases
may be due to facilities no longer being legally obligated to report releases because their
use of the chemical is below the reporting threshold. Second, the reporting thresholds
may also skew the cross-sectional rankings of facilities. In particular, facilities that
75
-------
appear to be low releasers or good environmental performers based on TRI releases may
actually not be better than other facilities with higher releases. The error in cross-
sectional ranking diminishes as one moves down the distribution. The rankings are
considerably less wrong about identifying the worst facilities.
4.4 Extrapolating to the National Level
So far we have examined the degree of bias potentially introduced by the
reporting thresholds into both trend and cross-sectional measures of environmental
performance for Massachusetts' facilities only. The reason for focusing on
Massachusetts was one of data availability. Data from TURA are critical in estimating
the magnitude of the truncation bias.
Can we extrapolate the findings from Massachusetts to estimate the degree of
truncation bias in national trends and cross-sectional comparisons? Unfortunately, any
such precise extrapolation would be shaky, at best, and outright misleading at worst.
While the results for Massachusetts do give a strong reason to be concerned about
truncation bias affecting the validity of national TRI data, the actual bias at the national
level could be lower or higher than in Massachusetts. The set of Massachusetts' facilities
is far from a representative sample of national facilities reporting to the TRI. On
average, facilities in Massachusetts have reported releases that are an order of
magnitude smaller than average releases for facilities in other states. The average total
releases between 1990 and 1999 for Massachusetts' facilities were 22,971 pounds while
the average for all other facilities were 166,368 pounds.18 These substantial differences
18 This difference is not only due to a smaller variance among Massachusetts facilities. If one
graphs the distribution of both total and on-site releases for Massachusetts and all other states,
76
-------
in the only observable variable that might be used to both predict which facilities cease
reporting because they went below the reporting threshold and to estimate the
magnitude of missing releases, make valid extrapolation infeasible.
Even on an intuitive level it is difficult to predict how the precise degree of
truncation bias nationally will compare to the estimated degree of bias in Massachusetts.
On the one hand, the reporting thresholds are lower in Massachusetts than for the
federal TRI program. In Massachusetts, once a facility triggers a reporting threshold for
a single chemical, the facility must report for all chemicals for which manufacturing plus
processing plus otherwise use amounts are greater than 10,000 pounds. For the federal
program reportability is calculated separately for each chemical based on the 25,000
pounds manufacture or process and 10,000 pounds otherwise use thresholds. This
might imply that truncation bias is likely to be a larger problem for the federal TRI data
than for the Massachusetts TURA data.
On the other hand, Massachusetts has an aggressive state-level pollution
prevention and reporting program that provide additional incentives for facilities in that
state to reduce use of their chemicals below the regulated level. Thus, we might expect
to see more threshold-regarding behavior in Massachusetts than in the rest of the
country. Similarly, Massachusetts' facilities do have lower average releases. This may
also indicate that Massachusetts' facilities are, on average, closer to the reporting
threshold and we would expect to see more threshold-regarding behavior in that state
than in the nation as a whole.
the entire distribution for Massachusetts' facilities is shifted to left. A Kolmogorov-Smirnov test
rejects the null hypothesis of equality of the distribution at the 1 % level.
77
-------
While I cannot provide any specific estimate of the degree of truncation bias in
the national trends or cross-section comparisons using TRI data, the experience in
Massachusetts does suggest that concern over truncation bias is well-founded. In
addition, some preliminary evidence from decreases in the federal reporting thresholds
for lead provide further evidence that the reporting thresholds may affect inferences
from the federal TRI data.
In 2001 the reporting threshold for lead was lowered to 100 pounds at the federal
level. For the 2000 reporting year, there were 1,997 facilities reporting for lead and lead
compounds and total reported releases of lead were 374 million pounds. In 2001, the
first year of reporting under the lower threshold, there were 8,444 facilities reporting a
total of 443 million pounds. This represents a net increase of 69 million pounds (19
percent). About half of this increase~33.5 million pounds—is attributable to facilities
that did not report on lead and lead compounds in 2000.19
Given these results, what implications should we draw about how to use TRI
data as a measure of environmental performance for policy analysis? That is the topic of
the final section.
5 Policy Implications for Using TRI as a Measure of Environmental
Performance
The above analysis indicates that the existence of the reporting thresholds may
introduce substantial bias in both the trend and cross-sectional estimates of environmental
performance using reported TRI releases. However, the TRI data are currently one of the
best sources of facility-level pollution levels nationwide. What is the policy analyst to
19 Personal communication, Cody Rice, Office of Pollution Prevention and Toxics, U.S.
Environmental Protection Agency, April 26th, 2004.
78
-------
do? The following recommendations are likely to enhance the validity of studies that use
TRI data for policy analysis.
First, the number of chemicals reported by a facility should be treated as an
additional policy outcome.20 If EPA is investigating the effectiveness of a new
regulation, it is not sufficient to examine only the effects on total TRI releases (even if
these releases are adjusted for toxicity). If the policy also has an effect on the number of
facilities that reduce chemical use below the reporting threshold, then estimates of the
policy's effect on TRI releases are likely to be biased upward —that is, the effect of the
policy will be overstated. If however, one estimates that the policy does not have an
effect on the number of chemicals reported, but does have an effect on total releases, one
can feel more confident that the policy has actually improved environmental
performance and not simply reduced reporting.21
If one is only concerned about whether a policy had a positive effect on
environmental performance, then a worst-case estimate of the effect may be appropriate.
For this worst-case estimate one would assume that any facility that ceases reporting for
a chemical did so because it went below the reporting threshold (as opposed to
eliminating the chemical, going out of business, or so forth). The facility's releases could
20 It may be tempting to normalize total releases by how many chemicals the facility reports. For
example, one could compare facilities both cross-sectionally and over time based on their average
releases per report. However, this correction does not remove the truncation bias. For example,
imagine a facility releases uses three chemicals and in the first reporting years reports releases of
200 pounds of Chemical A, 200 pounds of Chemical B, and 150 pounds of Chemical C. In that
year the average releases per chemical reported is 183 pounds. In the second reporting year the
facility only reports for two chemicals. It still releases 200 pounds of chemical B and 150 pounds
of Chemical C. Now the average releases per reporting year are lower at 175 pounds. Releases
per chemical reported declined, but the releases for that facility's reported chemicals do not
change across those two years.
21 See the first chapter of this dissertation for an example of a policy analysis that estimates the
effect of the policy on both releases and number of chemicals reported.
79
-------
then be set to their last reported level in perpetuity. This clearly overestimates total
releases, but if the policy is still found to lower releases even under this extreme
assumption, then one can feel confident in the policy's effectiveness. The magnitude of
the effect will be wrong, but the direction, if positive, will be correct.
One could also do sensitivity analysis on the directional effect of the policy
(although again not on the magnitude of the effect) by comparing the results for the
whole sample to the results for a sample only of facilities that report for the same
chemicals over the relevant time frame. If one estimates positive effects of the policy in
both samples, then the effects are not due only to decreases in reporting.
Such sensitivity analysis would be greatly improved by the addition of a
question on the federal reporting form that asks facilities why they are not reporting for
a chemical in the current year for which they reported in previous years. This question,
similar to the one used on the Massachusetts TURA form, could help EPA and others
assess the potential for truncation bias.
Unfortunately, the only fail-proof way to ensure that truncation bias will not
affect the results, and the only way to ensure the magnitude of policy estimates is
accurate, is to eliminate the reporting thresholds. EPA has the regulatory authority to
change the reporting thresholds, and has done so on two occasions. In 1999, EPA
reduced the reporting threshold for persistent and bio-accumulating toxins (the
threshold was reduced to 10 or 100 pounds depending on the chemical). In 2000, EPA
reduced the reporting threshold for lead to 100 pounds. Obviously, there are costs
associated with lowering-reporting thresholds. These costs include administrative costs
for promulgating a series of notice-and-comment rulemakings that are likely to be
contentious. There are also substantial paperwork compliance costs for facilities that
80
-------
will be required to report for chemicals that were previously unreportable. These costs
were estimated at 80 million dollars for the first year of reporting under the lower
threshold for lead (U.S. EPA, 2001).
Despite the costs, there may be important benefits from reducing or eliminating
the reporting thresholds. For example, EPA argues that responsible use of TRI data
allows "Federal, state, and local governments to compare facilities or geographic areas,
identify hot spots, evaluate existing environmental programs, and track pollution
control and waste reduction progress" (EPA, 2002b). This statement is only correct with
the caveat that analysis must also include an examination of the effect of the policy on
the number of chemicals reported, or in some other way address the potentially serious
issue of truncation bias. EPA has spent considerable resources developing a series of
risk-based weights for the TRI data in an effort to enhance the validity of these data as a
measure of environmental performance. Based on the results presented in this paper, it
is worth asking whether a similar effort on reducing truncation bias is warranted.
81
-------
6 Bibliography
Arora, Seema and Timonth N. Cason (1998), "Do Community Characteristics Influence
Environmental Outcomes? Evidence from the Toxics Release Inventory/' Journal
of Applied Economics 1(2): 413-53.
Degeorge, Francois, Jayendu Patel, and Richard Zeckhauser (1999), "Earnings
Management to Exceed Thresholds/' The Journal of Business 72(1): 1-33.
Ford, B.L. (1983) "An Overview of Hot-Deck Procedures," in: Madow, W.G., I. Olin and
D. B. Rubin, eds., Incomplete Data in Sample Surveys (New York, New York:
Academic Press): 185-207.
Fung, Archon and Dara O'Rourke (2000), "Reinventing Environmental Regulation from
the Grassroots Up: Explaining and Expanding the Success of the Toxics Release
Inventory, Environmental Management 25(2): 115-127.
Graham, Mary and Catherine Miller (2001) "Disclosure of Toxic Releases in the United
States," Environment 43(8): 9-20.
Grant, Don and Andrew W. Jones (2003), "Are Subsidiaries More Prone to Pollute? New
Evidence from the EPA's Toxics Release Inventory," Social Science Quarterly 84(1):
162-73
Helland, Eric and Andrew B. Whitford (2003), "Pollution Incidence and Political
Jurisdiction: Evidence from TRI," Journal of Environmental Economics and
Management 46 (3): 403-24.
Karkkainen, Bradley (2001), "Information as Environmental Regulation: TRI,
Performance Benchmarking, Precursors to a New Paradigm?," Georgetown Law
Journal, 89: 257-370.
King, Andrew A. and Michael J. Lenox (2000), "Industry Self-Regulation Without
Sanctions: The Chemical Industry's Responsible Care Program, " Academy of
Management Journal 3(4): 689-716.
Khanna, Madhu and Wilman R. Q. Anton (2002), "EPA's Voluntary 33/50 Program:
Impact on Toxic Releases and Economic Performance of Firms," Journal of
Environmental Economics and Management 37(l):l-25.
Little, R. J. A. and Rubin D.B. (1987), Statistical Analysis with Missing Data (New York,
New York: John Wiley Press).
Massachusetts Department of Environmental Protection, Bureau of Waste Prevention
(2003), Toxics Use Reporting Instruction (Boston, Massachusetts: Massachusetts
Department of Environmental Protection)
82
-------
National Research Council, National Academy of Engineering (1999), Industrial
Environmental Performance Metrics: Challenges and Opportunities (Washington,
D.C.: National Academy Press).
White, A. and D. Zinkl (1997), Green Metrics: A Status Report on Standardized Corporate
Environmental Reporting (Boston, MA: Tellus Institute).
Wooldridge, Jeffrey M. (2002), Econometric Analysis of Cross Section and Panel Data
(Cambridge, Massachusetts: The MIT Press).
Wolf, Sidney M. (1996), "Fear and Loathing About the Public Right To Know: The
Surprising Success of the Emergency Planning and Community Right-To-Know
Act," Journal of Land Use and Environmental Law 11(2): 217-324.
United States Environmental Protection Agency, Office of Pollution Prevention and
Toxics (2004), EPA's Risk-Screening Environmental Indicators (RSEI) Chronic Human
Health Methodology (Washington, DC: United States Environmental Protection
Agency).
United States Environmental Protection Agency, Office of Environmental Information
(2003) How are the Toxics Release Inventory Data Use? - Government, Business,
Academic and Citizen Uses (Washington, DC: United States Environmental
Protection Agency, Document number EPA 260-R-002-004).
United States Environmental Protection Agency, Office of Environmental Information
(2002a) 2001 Toxics Release Inventory, Executive Summary (Washington, DC:
United States Environmental Protection Agency, Document number EPA 2060-S-
03-001).
United States Environmental Protection Agency, Office of Environmental Information
(2002b) Toxics Release Inventory (TRI) Factors to Consider When Using TRI Data
(Washington, DC: United States Environmental Protection Agency, Document
number EPA 260-F-02-017).
United States Environmental Protection Agency, (January 17, 2001), "Lead and Lead
Compounds; Lowering of Reporting Thresholds; Community Right-to- Know
Toxic Chemical Release Reporting; Final Rule" Federal Register 40 CFR Part
372, 66(11): 4499-4547.
United States Environmental Protection Agency (2000), Emergency Planning and
Community Right-to-Know Act Section 313 Reporting Guidance for the Textile
Processing Industry (Washington, D.C.: United States Environmental Protection
Agency, Document number EPA 745-B-00-008).
83
-------
United States Environmental Protection Agency (1992), Guidance for Data Useability in
Risk Assessment (Part A) (Washington, D.C.: United States Environmental
Protection Agency, Document number 9285.7-09A): 31-34.
United States Government Accounting Office (1991), Toxic Chemicals: EPA's Toxic Release
Inventory in Useful but Can Be Improved (Washington, D.C.: United States General
Accounting Office, Document number GAO/RCED-91-121).
U.S. Public Interest Research Group (1998), Troubled Waters: A Report on Toxic
Releases Into America's Waterways (Washington, D.C.: U.S. Public Interest
Research Group).
84
-------
Transcription of Session VI Discussant Comments by John Dombrowski (U.S. EPA,
Office of Environmental Information)
Thank you. I'm John Dombrowski—I'm the Associate Director of the TRI Program
Division in the Office of Environmental Information. I'm really impressed by the work
that Lori [Snyder] and Michael [Kraft] have done. When I read through the papers, I was
actually thrilled, thinking, "Wow, people are actually using our data, and look at this!" I
think you did very good work, and you should be proud of that.
Just some points of clarification that I'd like to make first, generally about the TRI
Program. Then I have some thoughts, or comments, on the papers themselves and a little
update to what's going on in the TRI Program, because I think some of your work is
relevant to some things that we're looking at within the TRI Program.
As everyone here is aware, the EPCRA [Emergency Planning & Community Right-To-
Know Act] Section 313 set the reporting thresholds; the one thing that I would encourage
is [for everyone] to look at the Pollution Prevention Act [PPA], because our program is
based on this act, and that's where we start encouraging facilities to go towards source
reduction.
There has been a lot of reference to the risk-screening environmental indicators [tool] in
[the] context of the data in risk. That's fine—we agree with that—we don't have any
problem with that, except just keep in mind that the TRI Program is a hazard-based
program. I have my hands full just collecting the information, getting the information out
there, needless to say to worry about the risk of the data, but I don't think it's something
that should be ignored. Again, we provide the data so that communities can make a risk
determination for themselves at the community level.
There's been discussion about how the data represent just an estimate, and I assume
that's meant in a broader sense, because the origin of TRI data can be based on direct
monitoring. The statute does allow for direct monitoring pursuant to other laws—it
doesn't require direct monitoring—or making a reasonable estimate. Then I guess if
there is direct monitoring data, they can extrapolate that data over the course of a year,
and that's where the estimate would come in. I just want to clarify that it does allow for
both, and sometimes, actually, we think the engineering-estimated data are probably
better than some of the direct-monitoring data that are collected out there.
There's another clarification, also, that the total releases—I just want people to be aware,
because sometimes this is totally misunderstood—the total releases that EPA reports do
not include transfers offsite for recycling or treatment, and the TRI Program does not
include reuse in their program. "Total releases" involves either releases to environmental
media onsite or transfers offsite for disposal. I just want to make sure that people are
clear on the terminology that we're using for the program.
The one thing about the TRI Program, as Mike pointed out: It is a very good capacity-
building tool—that's exactly the intention of the law. If you look at EPCRA Section
85
-------
313h, that's what Congress intended it to be. We provide information to the public, to
governments, to researchers, and yes, we think it is a very valuable and very good
measurement tool. We'll stand behind a lot of those statements that are up there.
However, I don't think you're going to be able to find one specific answer to your
questions about why reductions are occurring—I think there will be other factors, a lot of
factors involved in that. Yes, I think public pressure is a big part of that, but I think
you're going to find a lot of different reasons for your answers. Don't look for just one
answer is what I would suggest.
The one thing, Mike, that you pointed out in your paper (this is a big question for us that
we always talk about, and we're starting to think about it a little bit more)—releases are
going down. Why is that? Well, again, that's one of the questions that you're trying to
answer, and we look forward to those answers. Part of it, also—and what I would
suggest you look at, too—is the fact that we allow facilities to come in and revise their
information. We don't have any set limit on that. They could also withdraw their
information. So, as facilities gain better information about their operations and better
information about how they may have estimated their releases, they may go back and
revise their information using better methods. That's something that should be factored
in as well, for more accurate information.
The one thing that you pointed out, Mike, in your paper is that the production-related
waste is increasing. We've noticed that trend, too, but we don't really have an answer for
that. Releases to the environment are going down, and that's a very good story to tell.
We want to tout that story, and the media picks up on it a lot—the public picks up on it,
because they understand "releases"—that's very simple to understand. But then we look
at the production-related waste, and I wonder what is this increase in production-related
waste attributed to? Production-related waste, again, includes releases, transfers offsite,
onsite treatment, and energy recovery and recycling. Hopefully the majority of
production-related waste is in recycling—that's what we're hoping for. Still, when you
look at the fact that production-related waste is increasing, you wonder whether TRI is
achieving the ultimate corporate behavior that you want it to achieve. It's nice that
they're reducing their releases, but you want them also to start going towards source
reduction. Then, you say, let's focus on PPA, because PPA is where we're starting to
look at source reduction. Maybe it's partly something EPA could do more—we focused
a lot on releases in the past years, in our public data releases and our public data release
reports, etc. Maybe we should start putting a little bit more "context" to the data—that's
been a big buzzword for us—and highlighting production-related waste and corporate
behavior, if you will, and the trends that we see. Part of the difficulty of that, though, is
trying to make that relevant to the public—showing what that means for a community.
They understand what a release means, and they can take that information and use it to
determine risk within their community, and that's what we would hope they do. But,
what does recycling, or energy recovery, or treatment mean for that community just
because the facility increased those amounts? So, it's not really easy for us to answer
how to put that all in context.
86
-------
When I was thinking about Lori's paper, and as I told her before the presentation, that's
another big area that we always ask about internally. We look at our data, and sometimes
we see facilities popping in and out of reporting through the years, and we wonder what
they're doing when they're not reporting. They're obviously below the reporting
threshold, but are they still releasing the chemical? Your analysis is very interesting to
see what that impact would be. The one thing I wonder, and maybe this is irrelevant and
wouldn't make a difference, but have you considered looking at where we tried to
minimize the truncation bias with the PBT [Persistent, Bioaccumulative, and Toxic]
chemicals? We lowered the thresholds, so that theoretically minimizes that truncation
bias, and you could start looking at facilities that, for the PBT chemicals, weren't added
as PBTs—they were just made PBTs—meaning they were always on the TRI list—if you
look at facilities that were reporting for those chemicals prior to them being a PBT—and
then maybe they dropped out before they were made a PBT, and then all of a sudden we
made them a PBT and they came back into the universe, what are they reporting as a
release there? Because now you know it's less than 25- or 10-thousand pounds, but it's
greater than 100 or 10 pounds for their use or manufacturing or processing. How do
those releases compare to the higher thresholds when they were reporting? Or how about
facilities that reported for the first time because of the PBT chemicals? How are those
releases comparing to facilities that were reporting prior to the [re-classification of the]
PBT chemicals and the lowering of thresholds? That might just be something to help
with the data here and verifying what you're finding, and I'd be curious to see if it does
relate.
Nothing against your conclusions, but I thought your measure on how effective our
policies are was kind of interesting. Actually, I have to give it some more thought, but
the idea about having the same number of chemicals and the same number of reports
based upon a policy decision, but the releases go down or maybe even the production-
related waste decreases. Trying to apply that as a measure of effectiveness to our policy I
find very interesting, and I'll have to think about how we might use that.
The idea about adding a question on our form to collect information about why they have
no longer reported or actually reducing or eliminating the reporting thresholds—well, to
be honest, that's not going to happen.
As far as the questionnaire, Mike, that you're getting ready to develop, I'm excited to see
the results.
The thing I find interesting about both of these reports that kind of relates to what we're
doing within TRI—we're doing a lot of modernization. One interesting note for the
researchers here is we're working really, really hard to get the data out sooner. We're
hoping that by November of this year we're going to have the 2003 data out on a facility-
specific basis, so you'll be getting the data a lot earlier.
I look at the reports and I look at what we're doing now—we have a burden-reduction
effort underway, and part of the burden reduction effort, for example, is to increase
thresholds as one option. Then I look at Lori's paper and I go, "Oh—okay!" The one
87
-------
thing I find interesting, too, is that on one's really mentioning the Form A, and I'm
curious to know why. Is that just because it's a lack of information now or it's useless?
What's the feedback there? One of our options is to look at the Form A and maybe
increase its use by increasing thresholds. Another option is "no significant change"—a
facility can report what they reported the previous year just by certifying no significant
change. And what is the meaning of that? Well, we have to figure that part out—that's
the challenge.
I intend to read these papers again. There are some people back in my office who are
working on a burden-reduction effort. I want to share these papers with them so they can
consider the information presented here as they go through the rulemaking, because,
again, I thought this was very good work on both of these papers.
88
-------
Session V Discussant Comments
Tom Bierle, Resources for the Future
Overview
These papers deal with two of the most important questions about information disclosure:
¦ What does disclosed data mean? and,
¦ How are information disclosure programs working?
Snyder provides a very interesting addition to the body of research on just what TRI
numbers mean. She adds what is to my knowledge a new and important caveat to the
many qualifications—acknowledged but often unheeded—to the facile assumption that
TRI is a comprehensive and relatively accurate measure of environmental performance.
Kraft, Abel, and Stephan focus in on how disclosure programs work. In a new and
welcome installment to their ongoing and ambitious work on TRI, they move their
analysis down to the state level on the way to examining how individual communities and
facilities translate information into action.
As I turn to the individual papers, I will begin with Snyder's question of what the data
mean and then move to Kraft, Abel and Stephan's analysis of how disclosure works.
Snyder
Snyder focuses on how reporting thresholds may lead to a significant under-reporting of
actual TRI emissions, possibly accounting for 40% of the observed decline in emissions
in Massachusetts. The argument rests on facility managers practicing "threshold-
regarding behavior"—looking for relatively easy opportunities to reduce the use of
chemicals below reporting thresholds and thereby avoid reporting emissions.
The analysis proceeds through three key steps:
¦ First was determining whether the facilities that specified in a questionnaire that they
had ceased reporting because they had fallen below a use threshold were
representative of other firms that had ceased reporting but had not specified why.
¦ Second was estimating why these other firms ceased reporting based on a model of
"threshold-regarding behavior." Snyder estimated that 72% of these firms ceased
reporting because they fell below thresholds, a number similar to the 66% among
firms that answered the questionnaire.
¦ Third, was providing three estimates—upper bound, lower bound, and extrapolated—
of what releases would have been if use thresholds had not existed. The upper bound
used 100%) of the previous year's releases, the lower bound used 0% of the previous
year's releases, and the extrapolated used 79% to 93% of the previous year's releases.
89
-------
Reporting thresholds are there because of the assumption that releases of chemicals used
at levels below thresholds don't add much to the aggregate—they are essentially a de
minimis exemption. Synder's analysis suggests that this assumption may be very wrong,
undermining the validity of TRI as a measure of environmental performance. Not only
do reporting thresholds affect the magnitude of trends but the ability to confidently rate
and compare facilities using a common metric, which many analysts see as the most
important aspect of TRI.
Unfortunately for this discussant, the degree to which these findings are significant rests
on questions of methodology. Rather than dissect that methodology here, I will merely
raise two questions:
¦ Barring some sort of sensitivity analysis, I suspect that one of the key variables here
is the prediction that 72% of non-reporting facilities stopped reporting because they
fell below a threshold. How sensitive are the results to even small changes in this
number? What value would different approaches to estimation produce? What do we
make of the fact that the state program explicitly encouraged companies to show a
decline in use? Was there "program-regarding behavior" that encouraged firms to
choose a particular response in the questionnaire?
¦ When looking at trends under the different scenarios, should they really begin at the
same point in year 1? Wouldn't a worst case scenario involve different assumptions
about what was not being reported in that first year, rather than assuming that all non-
reported chemicals actually accounted for zero emissions?
Finally, I might simply acknowledge that advocates of materials accounting reporting
might take issue with the contention that reducing chemical use "does not represent a real
improvement in environmental performance." Indeed, advocates argued that lowering
use led to lower risks from transportation, accidents at facilities, etc. As Snyder
acknowledges, a specific policy goal of the Massachusetts program was reducing
chemical use. Facility managers who identified a decline in chemical use in their TURA
reports were doing exactly what the state program had encouraged them to do, not
admitting to a crafty way of avoiding TRI-type reporting.
Kraft, Abel, Stephan
Kraft, Abel, and Stephan have undertaken a large project to take TRI data, warts and all,
and understand why it affects firms' behavior. There are two aspects of this overall
project that I particularly like. One is the specific focus on the agency of communities
and firms and the acknowledgement that actions may be very case-specific. Second is
bringing the tools of political science to this area of policy analysis, acknowledging that
these programs are not operating in a political vacuum.
In this particular paper, the authors examine a variety of indicators that may explain state-
level differences in the performance of firms. They find that the key variables that matter
90
-------
are 1) the level of conservation group membership in the state and 2) less ideologically
polarized politics in the state. Political scientists must be very satisfied when political
variables rise to the top.
The question arises: how should we interpret these results? Ultimately, the authors are
going to combine the results of analysis at various levels of generality to come up with
what I expect will be a nuanced interpretation. In the context of this paper, however, a
few questions arise:
¦ How to interpret the results about conservation group membership? Does this mean
that a more conservation-minded ethos permeates the business community; that state
legislators and regulators are more likely to pursue stricter regulation; that NGOs with
large memberships are able to push firms or push their members to push firms, more
easily; or some other explanation? If the NGO explanation, how do we fit in the fact
that most NGOs are nationally-based and don't get involved in state level policy or
local decisions about facility permits, etc.?
¦ How to interpret the results about the polarization of politics? Do less polarized
states have a more unified view about the value of environmental protection? Are
they less prone to gridlock that distracts legislators, regulators, and communities? I
suppose one could make the case that more polarization would have firms doing
better in order to hedge against a change in administration that might punish them
more. Less polarization may create the expectation of continuity and less impetus for
firms to go the extra mile.
Two other questions come to mind as well:
¦ To what extent do these kinds of variables affect TRI differently from other
environmental programs? Are we simply seeing generic differences in environmental
regulation and enforcement among states or is there something different about how
these variables affect community empowerment or facility managers' incentives to go
beyond regulatory requirements in response to TRI.
¦ Finally, given that the data come from the period 1991 to 1997—a good bit of which
was the era of low hanging fruit—how much does the analysis tell us about today's
dynamics and opportunities for improving the program?
It will be interesting to see how these authors' future work can tease out these sorts of
dynamics. I must say I am intrigued by the idea that "political factors can influence the
direction of change, while non-political factors may influence the intensity of change."
Synthesis/Themes
The most challenging task for a discussant is to try to identify common themes in the
papers presented. Although these two papers deal with very different questions about
TRI, I do think there are some common themes.
91
-------
First is the value of getting beyond analysis of aggregated national trends. At the state
level, the smoothly declining slopes in national releases get far more complex. Both
papers do a very good job of acknowledging this complexity, Snyder by being very
cautious about applying her findings beyond Massachusetts and Kraft, Abel, and Stephan
by welcoming idiosyncrasies as possible sources of interesting findings. I do wonder if
state-level analysis will ultimately prove to be the most revealing level of disaggregation
or merely a convenient way to divide up the world. I can see why it would be relevant in
Massachusetts, where a more comprehensive disclosure program is in place. But I
wonder whether Kraft, Abel, and Stephan will find that disaggregating to sub-state
regions or even communities is ultimately a more revealing.
The second way these papers relate to each other is in providing insights into the
motivations of corporations. Snyder discusses a dynamic of "threshold regarding
behavior" by which firm managers see opportunities to "reduce" reported emissions by
simply using slightly less of a particular chemical and thereby falling below reporting
thresholds. This focus inside the firm on the motivations of individual managers may
well be the kind of thing that Kraft, Abel and Stephan find as they move into facility-
level surveys. One would not be surprised to ultimately conclude the TRI "works"
because of a complex assortment of political, policy, community, and organizational
motivators, of which corporate managers' keen eye on reporting thresholds is one.
Finally, Snyder's and related work does provide a bit of a thorn in the side of analysts
like Kraft, Abel, and Stephan, and other people who work on disclosure. To suggest that
the measuring tape we are using doesn't have the feet and inches quite right is not
necessarily a welcome insight. The performers may not be the performers we thought
they were (or at least in the way we thought they were) and likewise the strugglers may
simply be less crafty about gaming the system. I don't think this is an insurmountable
problem. The key is always keeping in mind that TRI numbers are constructs influenced
by human agency, not objective measures of environmental performance. If efforts to
explain why TRI numbers go up or down keep in mind the many ways that reporting
facilities can make those numbers go up or down—some through real changes in the
workings of physical plants and some on paper—the kind of analysis done by Snyder and
the kind of analysis done by Kraft, Abel, and Stephan can be complementary rather than
antagonistic. I should note that Kraft, Abel, and Stephan are taking one of these issues
head on—that TRI reports on total pounds rather than an indicator or risk—in their larger
work.
Ultimately, the key question for those interested in policy and EPA in particular is going
to be what these and other studies tell us about how disclosed information ought to be
used and how disclosure policies ought to be changed. I think both of the lines of
research discussed in this panel will ultimately have that kind of pragmatic value.
92
-------
Summary of the Q&A Discussion Following Session VI
Ann Wolverton (U.S. EPA, National Center for Environmental Economics)
[Note: Dr. Wolverton brought two questions for each presenter. For better cohesion and
flow in this summary, the chronological order will be interrupted, with the responses
from Dr. Kraft coming before the questions posed to Dr. Snyder.]
Dr. Wolverton began by addressing Dr. Kraft's "future plans of defining an analysis at
the community level" and his "confidence about drilling down to the community level"
because of the potential consistency between that and the state-level analysis.
Mentioning that there has been a lot of recent discussion in the literature "about how
results change all the time depending on the way in which you define the community,"
She asked Dr. Kraft how he planned to define "community" in his future research.
The second question addressed to Dr. Kraft asked whether the trend of production-related
waste increasing over time "holds true when you normalize by production."
Michael Kraft (University of Wisconsin at Green Bay)
Dr. Kraft said he probably couldn't answer the second question but would pass it along to
someone who could. Although he didn't believe they had normalized for that, he
asserted, "I think there's still a trend there—it may be altered with normalization."
On the other hand, he said, "The first question is something that we've certainly thought
about. What the question is getting at is: We're looking at state-level comparisons here,
and we've said that gives us confidence that there is some kind of political factor at the
state level, whether it's ideology, conservation group memberships, . . . We're only
interested in how that plays out around the facility, because the whole point of the
information release is that communities—and that's the language that's used,
communities—take cognizance of this information, and, of course, some action follows.
Of course, state groups might also take an interest in what's going on in the community,
and that may be true of environmental justice groups at the state level or environmental
groups more broadly even."
Dr. Kraft continued by saying that the community identification issue is complicated by
the fact that "we're looking at facilities, and then we're looking at communities in which
they are located." He suggested that in different circumstances "community" could mean
"the city," "the county," or "the immediate two miles around a facility." Dr. Kraft went
on to say, "I think we're inclined to take a broader view of surrounding community. That
may get a little messy in a larger urban area—exactly where the boundaries ought to be,"
but he indicated that the risk screening model might help in that regard. Citing the
successful use of this model in working with Region 5 air emissions, he explained that "it
actually models the plume that would follow a factory, and you can identify exactly
which neighborhoods are affected by a given facility. So, it isn't just geographic—for
example, "x" blocks or miles away—you can get very specific. That's something we
would certainly like to look carefully at to know what groups, what people, what citizens,
93
-------
what community leaders we need to talk to, so that's a big part of the next stage of the
project."
Ann Wolverton's comments addressed to Lori Snyder:
Dr. Wolverton commented, "I was just curious about whether the same firms that are
reporting in the U.S. EPA's TRI are also in the Massachusetts TRI and whether there is a
way to make use of that. It may be that there's not a lot of intersection—I was just
curious to know."
The second comment that Dr. Wolverton addressed to Dr. Snyder was: "You're focusing
on the pounds of emissions and sort of what's going on with the threshold and how much
you're missing in terms of pounds. I was curious about how much you might be missing
if you attempted to translate that into a more risk-based measure?"
Lori Snyder (Harvard University)
In response to the first comment, Dr. Snyder clarified that the TRI reports are actually
sort of a subset of what is reported in the other reports. She added that, consequently, one
of the things she could do "is just look at chemicals that are reportable to TRI," although
this could get "a little tricky" due to the fact that "sometimes, compound chemicals that
are in TRI are defined differently in the State and Federal reports. Another complication
would be that she has the "use information based on State thresholds." Dr. Snyder closed
by saying that she did have plans to revisit this and justify her claims.
Elizabeth (Betsy) David (Stratus Consulting, Inc.)
Citing results from some "mid-to-late-90's" interviews of 100 Wisconsin firms, Dr.
David said she and her colleagues found "three important reasons why people reduced
their pollution," with the primary reason being "to stay in compliance." However, she
added that "the second reason was that the firm perceived that its ranking on the TRI
reports was very important to them [i.e., communities], so even though at community
levels I didn't see much in the way of communities taking action on this information,
firms perceived that the communities would, and so they changed their behavior." Dr.
David commented that since this data was compiled at the same time as Dr. Kraft's data,
he should also have gotten indications that "this perception was really a very important
determinant of people's actions." Dr. David closed by citing the story of "a firm called
Fort Hubbard Steel [in Green Bay, Wisconsin] that came out ranked very high on the TRI
list" and subsequently modified its process from etching with acid to grinding (to the
dismay of the Green Bay Metropolitan Sewage agency, who had been using the waste
acid in their process to precipitate out phosphorous).
Dinah Koehler (University of Pennsylvania)
Dr. Koehler commented on John Dombrowski's assertion that engineering estimates
sometimes provide a closer approximation of stack pollution than actual sampling
94
-------
measurements. Citing her research, Dr. Koehler said "there are huge differences between
what the National Toxics Inventory (NTI) estimates were at the 4-digit SIC level and the
TRI." She had held hopes of potentially using the NTI as a substitute for the TRI, "given
that there's all this noise in the TRI," but the fact that the difference is sometimes so great
prevents her from doing so. She wondered whether Mr. Dombrowski could comment on
what might be going on there.
John Dombrowski (U.S. EPA, Office of Environmental Innovation)
In response, Mr. Dombrowski said that if one is looking at "releases for just stack
emissions," in that case monitoring might be the better of the two options. However, he
stated, "In the general sense, though, I would say that when you look at fugitive
emissions, for example, engineering estimates for the mass balance you're doing might
be better than just occasional monitoring of a fugitive emission, because you can account
for wherever the releases may be occurring." He clarified further, "my statement was in
the sense of all monitoring data," and depending on what type of release you're talking
about, one or the other of the methods might be preferable.
Mr. Dombrowski went on to say that he "just felt that people sometimes make an unfair
assessment of the TRI data and discredit it because of the estimation that sometimes
occurs," whereas he feels that "you have to look at the data specifically as to what you're
using and then make an evaluation of that, but not an overall generalization of the TRI
data." He added that "NEI (National Emissions Inventory) sometimes actually
supplements their data with TRI data, so we capture sources that they may not be
capturing. Sometimes we have captured better air data—with mercury releases, for
example, I think we were capturing much more of a complete data set than what the Air
Data Set was capturing at that time. So, again, you have to be very specific as to what
you're looking at for the data element."
Randy Becker (U.S. Bureau of the Census)
Dr. Becker claimed that "the best paper was actually saved for last—it sounds like
required reading for anyone who wants to use the TRI." He added that the paper "sets an
excellent example for all of us, because what's important is getting the questions right
and the econometric specification right—you need to know where the data come from
and what their limitations are." Commenting that "we're spending a lot of money here on
studies that use data to answer questions," Dr. Becker noted, "just because others use data
blindly doesn't mean all of us have to. Lori's paper reminded me that we should
probably also spend money on studies that evaluate the quality of the data—and probably
also spend money on improving the quality of the data, as John said."
Furthermore, Dr. Becker said that in looking at the TRI, "before we chalk everything up
to community activism, there's also regulation running parallel to this. In fact, some of
the pollutants in TRI are actually regulated because they're ozone precursors."
Consequently, in efforts to comply with regulations for cleaning up these precursors or
other pollutants, facilities "may be cleaning up TRI and it has nothing to do with
community activism whatsoever." Dr. Becker added that "while there's no direct
95
-------
measure of how much regulation a plant is facing, you could control for whether it is in a
county that is in non-attainment of the National Ambient Air Quality Standards. Also, in
terms of the conforming community, perhaps more important than TRI is something like
ozone action days, which, again . . . may be driving things rather than community
activism or just the information disclosure."
Michael Kraft
Dr. Kraft responded that "we do make a comment in the paper to the effect that many
facilities clearly do alter their actions based on regulatory pressures or anticipated
regulation—community pressure is not always the explanation—it's a multi-faceted
situation. He also clarified that "there's a good bit of literature that suggests regulation
is, in fact, a driving force, and maybe a lot more so than community pressure." In
closing, Dr. Kraft stated, "TRI is predicated on the notion that the public has the right to
know, and we're looking at what's going on in the community in part because of that, but
that doesn't mean we're going to come to the conclusion that that's the most important
factor."
Eric Orts (University of Pennsylvania)
Dr. Orts asked the following "quick question that also pertains to the reliability of the
TRI data: . . . Have there been enforcement actions taken with respect to companies on
whether they reported these data accurately or not? In another life I'm a securities law
corporate academic, and certainly in the securities field reliability of information, even if
there is enforcement, is sometimes suspect. But, for the most part, it's thought to be
pretty good, in part because there's enforcement." Dr. Orts asked whether there are
actual examples of the EPA going after companies that have reported bad data, and
(because he thinks it's potentially a criminal act) whether a case has ever been referred to
the Department of Justice.
John Dombrowski
In response, Mr. Dombrowski admitted that he didn't know whether there have actually
been any criminal actions, but "there has been, and there continues to be, enforcement
related to TRI data. He expounded by saying, "Every year our Regions do data quality
audits, and they look for facilities that are under-reporting. . . . We've had some Regions
even take actions on over-reporting, because the purpose of TRI is the public's right to
know, and we want to provide the most accurate information. You also want to consider
a level playing field between the facilities that are doing a good job with their data versus
those that are doing a not-so-good job and just getting by."
Mr. Dombrowski also added, "Speaking of enforcement, there is currently an
enforcement initiative ongoing right now on TRI. . . and you'll see a flyer—an
enforcement alert for late reporters. We've done a lot of analysis on facilities that have
reported their data late to EPA . . . and we're not talking about a trivial amount of data
when you add it all up together on the late reporters. And, there is a statutory deadline
for them to report and for us to get the information out to the public, so enforcement is
quite active in the TRI program in various aspects."
96
-------
Eric Orts—a follow-up question
"Have there been any cases of a civil penalty that's been assessed against a company
because of infractions in this area?"
John Dombrowski
"Yes, there have been. Jon [Silberman], you might know better than I, but I have
heard—and I don't know exact numbers—but through EPA's audit policy, for example,
one of the more frequently disclosed violations by companies is TRI reporting
violations."
Matt Clark (U.S. EPA, Office of Research and Development)
Dr. Clark commented, "We had one paper that we really wanted to present today (the
author wasn't able to show up) that I would recommend to you—it's by Michael Vasu.
He looked at the extent to which communities actually knew about TRI, and it turned out
to be in the low teens. Not only that, more people reported knowing about a fictitious
database than knew about the TRI."
Dr. Clark also added, "We are trying to get out a solicitation this year on the benefits of
environmental information disclosure. It's going to cover a lot of stuff directly related to
this."
Sarah Stafford (College of William and Mary)—another question
In a comment "kind of related to this idea of regulatory pressure" Dr. Stafford suggested
to Dr. Kraft that "some potential regulations may be correlated with those political
variables that you have in there," and she said she is "most concerned about pollution
prevention regulations, both mandatory and voluntary." She cited the fact that in 1995
"at least two states had required pollution prevention actions" which, though hazardous
waste based, "clearly had a big impact on TRI. Dr. Stafford added that some of her work
suggested that there "could be some correlations as to which states adopt voluntary
pollution prevention programs and what types of programs they have related to,
specifically, the membership in environmental organizations." In closing, she said, "It's
something worth looking at, just in terms of correlation."
Michael Kraft
Dr. Kraft stated, "We're not going to presume anything in the way that we word the
questionnaire. We're certainly not going to assume that the community knows about the
TRI. . ., and we're aware, of course, that State regulations factor into the equation."
End of Session VI Q&A
97
------- |