OCPA Working Paper Series1
Working Paper # 2017-05
May 24, 2017
Do more frequent inspections improve compliance?
Evidence from underground storage tank facilities in
Louisiana
Karen A. Sullivan2
Office of Communications, Partnerships and Analysis
Office of Land and Emergency Management
U.S. Environmental Protection Agency
1200 Pennsylvania Ave., N.W.
Washington, DC 20004
(202) 566-1259
sulllivan.karen@epa.gov
Achyut Kafle
ORISE Research Fellow
Hosted by the U.S. Environmental Protection Agency
kafle.achyut@epa.gov
1 U.S. Environmental Protection Agency's Office of Communications, Partnerships and Analysis (OCPA)
publishes a working paper series authored by scientists in the Office of Land and Emergency
Management (OLEM) or produced with OLEM funding on research related to the impacts of OLEM
programs and policies. The working papers are distributed for purposes of information and discussion
while they undergo the peer review process at academic journals. Any opinions and conclusions
expressed herein are those of the author(s) and do not necessarily represent the views of the U.S.
Environmental Protection Agency.
2Questions or comments on the draft paper can be emailed to: sullivan.karen@epa.gov.
-------
Do more frequent inspections improve compliance?
Evidence from underground storage tank facilities in Louisiana
Karen A. Sullivan (Corresponding Author)
Office of Communications, Partnerships and Analysis
Office of Land and Emergency Management
U.S. Environmental Protection Agency
1200 Pennsylvania Ave., N.W.
Washington, DC 20004
(202) 566-1259
sulllivan. karen@epa. gov
Achyut Kafle
ORISE Research Fellow
Hosted by the U.S. Environmental Protection Agency
kafle. achyut@epa. gov
Disclaimer: The views expressed in this paper are those of the authors and do not necessarily reflect the views
or policies of the U.S. Environmental Protection Agency.
Abstract
This paper examines the effect of inspection frequency on compliance decisions in an environmental pollution
prevention context by capitalizing on policy changes occurring under the Energy Policy Act of 2005 that
increased inspection frequency requirements for underground storage tank [UST] facilities to at least once
every three years. A censored bivariate probit model is estimated using data from Louisiana on inspection,
compliance, releases and other socioeconomic and biophysical characteristics of UST localities to examine the
relationship between increased inspection frequency and compliance. We find that increased inspection
frequency improved compliance with UST requirements in Louisiana and this impact is heterogeneous based
on a facility's compliance status at the last inspection—larger impact for those facilities that were compliant
than those that were noncompliant at the last inspection.
KEYWORDS: compliance; inspection; environment; government policy; impact analysis; firm behavior
ACKNOWLEDGEMENTS: Researchers conducted this analysis while supported by the AAAS Science and
Technology Policy Fellowship Program, the Oakridge Institute for Science and Education Research Participation
Program and the U.S. Environmental Protection Agency (US EPA). ArcGIS data work supported by funding from
the US Environmental Protection Agency (contract GS-10F-0061N via Industrial Economics, Inc - Kate Doiron).
We would like to acknowledge valuable information and comments provided by Sam Broussard at the Louisiana
Department of Environmental Quality and Linda Gerber and Tim Smith at the US EPA as well as comments from
participants at the National Tanks Conference and Expo 2015, Northeastern Agricultural and Resource Economics
Association Meeting 2015, Southern Economics Association Meeting 2015, and the Society for Environmental Law
and Economics Meeting 2016.
JEL CODES: K32 Energy, Environmental, Health, and Safety Law, Q5 Environmental Economics; Q53 Air
Pollution -Water Pollution - Noise - Hazardous Waste - Solid Waste - Recycling; Q58 Government Policy
2
-------
1. Introduction
Across the United States approximately 561,000 underground storage tanks (UST) store
petroleum or hazardous substances at approximately 202,000 sites, and are regulated by the U.S.
Environmental Protection Agency's (EPA) UST Program (US EPA 2016). The majority of USTs
are located at gas stations and some are located at facilities in other industries such as the
commercial sector, manufacturing, transportation, wired telecommunications, electric utilities,
and hospitals (US EPA 2011). The greatest potential hazard from a leaking UST is that
petroleum or other hazardous substances can seep into the soil and contaminate groundwater, the
source of drinking water for nearly half of all Americans (USGS 2003). A release from an UST
can also present other health and environmental risks, including potential for fire and explosion.
EPA, states, and tribes work in partnership with industry to protect the environment and
human health from potential UST releases. In 1984, Congress established the UST program to
monitor the approximately 2.1 million tanks that were active at that time.1 The U.S. EPA UST
program is designed to prevent releases of petroleum and hazardous substances into the
environment, detect releases when they occur, and clean up any contamination from releases. To
monitor the large number of tanks, EPA enlisted states' assistance in implementing and
enforcing the program.2 Despite early efforts, releases were common—from the beginning of
the program until 2000 over 400,000 releases were reported. At an average cleanup cost of
$152,000 that roughly represents an estimated 60.8 billion dollars in cleanup costs (US EPA
2015b).3 This only represents a lower bound estimate of the costs from these releases as it does
not include negative impacts on nearby property values, human health or ecosystem services
1 Since the 1984 inception of the UST program, 1,832,048 USTs have been properly closed.
2 As of 2016, 38 states and the District of Columbia and the Commonwealth of Puerto Rico have approved state
UST programs. To obtain EPA approval, state programs must be at least as stringent as the federal requirements (US
GPO e-CFR 2013)
3 2012 USD
3
-------
(Jenkins et al. 2014; Guignet et al. 2016). In 2001, the United States General Accounting Office
(GAO) investigated concerns raised by the United States Senate Committee on Environment and
Public Works that the UST program was not effectively preventing leaks (US GAO 2001). One
aim of the investigation was to determine the breadth of EPA's and the states' tank inspections.
Physical inspections confirm whether tanks have been updated and are being properly operated
and maintained to prevent and detect releases. The GAO's survey of state UST programs showed
that at the time 42% of states did not inspect all tanks on a regular basis and 20% of states
inspected at intervals of 4 years or longer. EPA managers recommended that inspections take
place annually or where resources are limited at a minimum of every three years but only 38% of
states inspected all tanks at an interval of three years or less. Based on their findings, the GAO
recommended that Congress may want to authorize EPA to establish a federal requirement for
the physical inspection of all tanks on a periodic basis.
On August 8, 2005, President Bush signed the Energy Policy Act (EPAct) of 2005. With
this came amendments to Subtitle I of the Solid Waste Disposal Act (SWDA), which is the
original 1984 legislation that required the U.S. Environmental Protection Agency (EPA) to create
a comprehensive regulatory program for USTs storing petroleum or certain other hazardous
substances. Among other provisions, the UST provisions of EPAct added the requirement that all
regulated UST facilities must be inspected to evaluate compliance with UST requirements at
least once every three years.4 More frequent UST inspections are intended to improve facilities'
compliance with UST release detection and prevention requirements, and in doing so prevent
accidental releases of harmful substances into the environment. This study examines how the
resulting changes in inspection frequency impact compliance at UST facilities.
4 Other provisions include operator training, delivery prohibition, secondary containment, financial responsibility,
and cleanup of releases that contain oxygenated fuel additives.
4
-------
Today national compliance rates are higher than they were before the 3-year inspection
requirement. At the end of fiscal year 2005, 66 percent of facilities were in operational
compliance but by the end of fiscal year 2015 compliance rates reached 81 percent (US EPA
2005; US EPA 2015a). This trend, as depicted in Figure 1, represents a significant achievement
but the extent to which this improvement in compliance is due to the increase in inspection
frequency is unclear. There may be other factors affecting compliance rates such as the
establishment of UST operator trainings or changes in individual state UST regulations that in
some cases are more stringent than the UST federal requirements. Without controlling for other
factors that may impact compliance, the role that increased inspection frequency has taken in
these improvements cannot be clearly identified.
Previous empirical analyses consistently show that inspections combined with penalties
for violations improve compliance across a variety of environmental regulations such as the
Clean Air Act, the Clean Water Act, and hazardous and toxic waste regulations (Shimshack
2014). However, these studies do not explicitly investigate how changes in inspection frequency
affect facilities' compliance behavior. This analysis explicitly examines the effect of inspection
frequency on compliance in a pollution prevention context by capitalizing on changes in
inspection frequency occurring as a result of an exogenous policy change—the EPAct of 2005.5
Given the significant resources devoted to compliance inspections in the UST program,
determining the effect of those inspections is critical to making future policy and funding
5 The impact of increasing inspection frequency on compliance has been studied in other contexts. Ko, Mendeloff,
and Gray (2010) examine the effect of repeated Occupational Safety and Health Administration (OSHA) inspections
and the time between inspections on noncompliance and find that the number of violations cited increased with each
additional year since the prior inspection after controlling for other variables. The increases totaled approximately
15% over five years. Alberini et al. (2008) examine FDA inspections of seafood processors' compliance with
sanitation requirements and a new Hazard Analysis and Critical Control Points (HACCP) requirement. Anticipated
inspection frequency represented by the hazard rate predicted from the inspection model increases the likelihood of
compliance with the sanitation program but not with the newer HACCP program.
5
-------
decisions. Similarly, this analysis may be useful to other environmental programs relying heavily
on compliance inspections to monitor and enhance compliance.
A national analysis of the impact of the EP Act's 3-year inspection requirement on
compliance would be ideal but the data needed to do so is not available. States report aggregated
state-level UST information periodically throughout the year to EPA. This reported data is useful
to EPA for measuring UST performance, however, it cannot be used to conduct a national
analysis of the impact of increasing inspection frequency on compliance due to limited data on
inspection frequency in each state over time (i.e., the total annual number of inspections in each
state was not reported to EPA by states until 2008—three years after the enactment of the EPAct
of 2005). Furthermore, many state UST programs do not have inspection and compliance
databases that contain sufficient data from prior to EPAct to be able to examine the impact of
changes in inspection frequency on compliance.
This empirical analysis uses an UST facility-level dataset from the Louisiana Department
of Environmental Quality (LADEQ) that includes facility characteristics and information on
inspection, compliance, and releases from before and after EPAct (2001 to 2012) combined with
data on the socioeconomic and biophysical characteristics of the facilities' locations. Prior to the
EPAct of 2005, Louisiana inspected all tanks at an interval of 4 years or longer, which allows us
to capitalize on the exogenous implementation of the EPAct of 2005's 3-year inspection mandate
to examine the impact of increased inspection frequency on compliance (US GAO 2001). Results
from a censored bivariate probit model show that increasing inspection frequency improved
compliance of owners and operators at regulated UST facilities in Louisiana and this effect is
heterogeneous based on the facility's compliance status at the last inspection—larger impact for
those facilities that were compliant than those that were noncompliant at the last inspection.
6
-------
The remainder of this paper is organized as follows. Section 2 provides a brief discussion
of related literature. The data used in this analysis as well as background information on
Louisiana's UST Program is described in Section 3. Section 4 describes the empirical approach
taken, and the choice of explanatory variables is explained in Section 5. Results of the analysis
are presented in Section 6, followed by conclusions in Section 7.
2. Related literature on inspection and compliance
The theoretical foundation for investigating the impact of environmental monitoring and
enforcement on regulated firms' compliance decisions is grounded in the economic theory of
crime and punishment first formalized by Becker (1968). Becker's model of crime and
punishment assumes that a rational, risk-neutral agent evaluates the expected benefits and the
expected costs of a private action, and then acts if the expected benefits exceed the expected
costs.6 Becker's model was later adapted to an environmental context by Russell, Harrington and
Vaughan (1986), and there have since been extensions to examine different aspects of
environmental enforcement regimes (e.g., avoidance behavior (Malik 1990), self-reporting
(Malik 1993; Kaplow and Shavell 1994; Innes 1999a, 1999b, 2001), and more frequent versus
more thorough inspections (Heyes 1994)). Since the late 1980s, an extensive empirical literature
on the impacts of environmental monitoring and enforcement has also developed. Interested
readers may refer to Shimshack (2014) for a comprehensive review of the environmental
monitoring and enforcement literature.
6 An alternative theory—the behavioral model of compliance—from the social-legal tradition theorizes that
inspections reduce accidents or injuries by spurring firms to pay more attention to safety. Firms may be found out of
compliance and be eager to return to compliance due to the social norms—the desire to be a law abiding citizen
(Cyert and March 1963; Scholz and Gray 1990; Mendeloff and Gray 2005).
7
-------
Specific-deterrence based enforcement is a widely used enforcement regime in which
regulators use inspections or threats of inspection and penalties for identified violations as
mechanisms to enforce environmental regulations. Empirical studies focused on the impact of
specific-deterrence based enforcement on compliance consistently show that inspections
combined with penalties for violations improve compliance across a variety of environmental
regulations such as the Clean Air Act (CAA), the Clean Water Act (CWA), and hazardous and
toxic waste regulations.7 For example, Gray and Deily (1996) and Deily and Gray (2007) showed
that EPA monitoring and enforcement actions under the CAA led to improved compliance at
steel mills in the early 1980s. At paper and pulp mills, Nadeau (1997) and Gray and Shadbegian
(2005) found that EPA and state environmental monitoring and enforcement actions resulted in
reductions in both duration and rate of air pollution noncompliance during the 1980s. Also, the
threat of lawsuits reduced air pollution emissions at coal-fired power plants and various
manufacturing industries (Keohane, Mansur and Voynov 2009; Hanna and Oliva 2010).
In the context of the CWA, Magat and Viscusi (1990) found that increased threats of
inspections at U.S. pulp and paper mills improved water pollution compliance and Shimshack
and Ward (2005; 2008) showed that formal enforcement actions with monetary penalties reduced
water pollution discharges. At water treatment plants during the 1990s and at chemical facilities
during the late 1990s and early 2000s, federal fines were also found to reduce pollution (Earnhart
2004a, 2004b; Glicksman and Earnhart 2007).
7 This study focuses on specific-deterrence effect, which is the effect that inspections, sanctions or increased threats
of inspections or sanctions have on an evaluated or sanctioned facility, as opposed to general deterrence. General
deterrence effects occur when inspections or sanctions on a targeted facility spillover to other non-targeted facilities
and lead to compliance improvements at the facilities that were not directly evaluated or sanctioned. In
environmental context, Shimshack and Ward (2005; 2008) detected general deterrence effects of government
enforcement whereas Langpap and Shimshack (2010) found that private enforcement (or private citizen suits)
significantly crowded out government enforcement in case of wastewater treatment. Qualitative survey results also
tend to support the evidence of general deterrence effects in environmental settings (Carlough 2004; Thornton,
Gunningham and Kagan 2005).
8
-------
Monitoring and enforcement has also been shown to affect hazardous and toxic waste
emissions and compliance. Stafford (2002; 2003) found that increased liabilities or penalties
under the Resource Conservation and Recovery Act (RCRA) have been shown to reduce plants'
violation probabilities and Alberini and Austin (1999; 2002) showed that increased threats of
lawsuits and strict liability rules affected toxic waste releases. At regulated facilities in Michigan,
Liu (2012) showed that RCRA inspections have a significantly positive effect on compliance, as
well as evidence of positive cross-program effects (i.e., inspections under the CAA have a
positive and significant effect on facility compliance with RCRA). Also, monitoring efforts have
reduced oil spill frequency and spill size (Epple and Visscher 1984, Cohen 1987, Grau and
Groves 1997) and federal cases against gas and liquid pipeline operators may have improved
environmental performance in the late 2000s and early 2010s (Stafford 2014). Eckert (2004)
examined the impact of inspections and warnings on compliance with storage inventory
reconciliation regulations at above and underground petroleum tanks in Canada and found a
small but positive impact (i.e., inspections and warnings deter future violations).
This brief summary of the environmental monitoring and compliance literature suggests
that specific-deterrence enforcement mechanisms are an effective means to improve compliance.
In these analyses, a variety of factors that may influence compliance such as compliance history,
facility characteristics, and changes in penalties are examined, however, to the best of our
knowledge these studies do not explicitly investigate how changes in the length of time between
inspections may impact regulated facilities' compliance behavior.8 This analysis aims to address
8 Liu (2012) examines the impact of inspection frequency on compliance at facilities regulated under Resource
Conservation and Recovery Act (RCRA) and the Clean Water Act (CAA). Liu finds that increasing the total number
of RCRA inspections in the last year increases RCRA compliance, and that there is evidence of cross-program
effects (i.e., increasing the total number of CAA inspections at a facility in the last year also increases RCRA
compliance). However, we believe that Liu's measure of inspection frequency, which is defined as the total number
of inspections at a facility in the last four quarters, is limited because typically RCRA facilities are not inspected
more than once in a given year. In fact, only 50% of facilities in her sample were formally inspected at all during the
9
-------
this gap. More specifically, this analysis aims to quantify the impact that more frequent
inspections have on compliance by capitalizing on the policy changes occurring under the
exogenous implementation of the EPAct of 2005 that established a requirement for all states
receiving Subtitle I funding for their UST programs that the time between a facility's concurrent
compliance inspections cannot exceed three years.
3. Data description and background
3.1 Data description
This analysis uses Louisiana Department of Environmental Quality's (LADEQ) data on
inspection, compliance, and releases at 4,424 UST facilities from fiscal year 2001 to 2012. The
data includes information on facility specific characteristics, results of compliance inspections,
and releases. The facilities' addresses were geocoded and matched with location specific
socioeconomic data obtained from the 2009-2013 U.S. Census American Community Survey 5-
year estimates and biophysical data obtained from the Soil Survey Geographic Database (US
American Communities Survey 2010; U.S. Department of Agriculture 2015). The final sample is
an unbalanced panel that consists of 108,281 quarterly observations on 4,424 facilities that had at
least one active petroleum UST subject to federal UST regulations between 2001 and 2012. On
average each facility has 2.82 USTs with an average tank capacity of approximately 8,500
gallons. The average age of the oldest tank at a facility is 21.7 years.
study time period (2001-2010) and the mean number of RCRA and CAA inspections at a facility in the past four
quarters were 0.19 and 0.17, respectively. Repeat RCRA inspections within a one year time period may be triggered
by endogenous factors that simultaneously impact facilities' compliance behavior. In our analysis, we are able to use
an exogenous policy change affecting inspection frequency to explicitly examine the impact of changes in inspection
frequency on compliance.
10
-------
3.2 Background: Underground storage tank inspection and compliance in Louisiana
Louisiana's UST state program was approved in 1992. During the late 1980s and 1990s,
Louisiana focused on closure of substandard tanks and remediation activities. In 2000, the
Louisiana State Legislature established requirements that 15% of active USTs be inspected each
year. UST inspections are announced usually one week in advance in Louisiana. This notice is
given with the purpose of providing the tank owner time to gather the required paperwork that
will need to be examined. An inspection typically takes one to three hours, and the inspector
goes through each step of the inspection with the facility owner or operator if they are available.
All USTs at the facility are inspected. The inspector checks to see if the facility is compliant with
a comprehensive list of requirements aimed at preventing and detecting releases such as
standards for tanks and piping, spill and overfill prevention equipment, operation and
maintenance of corrosion protection systems, release detection, record keeping, and so on. If a
violation is identified during an inspection, the inspector will document the violations and confer
with the LADEQ Enforcement Division to determine the appropriate type of enforcement action
that will be issued. Usually a Notice of Deficiency (NOD) or Notice of Potential Delivery
Prohibition (NOPDP) is issued.9
Facilities that do not return into compliance or do not respond to NODs or NOPDPs will
receive Compliance Orders from the LADEQ Enforcement Division and those that had been
issued a NOPDP are subject to having their tanks prohibited from receiving product deliveries,
which is referred to as red tagged. When facilities refuse to return into compliance or certain
egregious violations occur, the Enforcement Division has the discretion to issue either a formal
9 If the facility has a temporarily closed tank or is an abandoned facility, the Enforcement Division may opt to issue
a Compliance Order immediately rather than a Notice of Deficiency.
11
-------
penalty notice or an Expedited Penalty Agreement.10 One limitation of the UST data from the
LADEQ is that data on enforcement actions beyond the initial compliance citations, NODs or
NOPDPs, is not available until fiscal year 2004, and therefore we are only able to account for the
initial NODs and NOPDPs in the analysis.11
Figure 2 shows the percent of facilities inspected (dotted line), the percent of inspected
facilities that received at least one noncompliance citation (solid line), and the percent of
facilities at which a release was confirmed (dashed line) in each year from 2001 to 2012. Prior to
the EPAct of 2005, roughly 7-15% of Louisiana's UST facilities were inspected each year. This
coincides with the time frame during which the Louisiana State Legislature had a requirement
that 15% of active USTs be inspected each year.12 The EPAct of 2005 that included provisions
on UST inspection frequency requirements was signed on August 8, 2005. The provisions
included a transition phase from August 8, 2005 to August 8, 2007 during which states were
required to inspect all active UST facilities that had not been inspected since 1998. The LADEQ
began to focus inspections on these facilities just before Hurricanes Katrina and Rita hit
Louisiana (August and September 2005, respectively). LADEQ diverted resources to deal with
the hurricanes' aftermath, and as a result only 7.6% of facilities were inspected in 2006. Once
10 A facility has the option to sign an Expedited Penalty Agreement, which allows them to settle the violations for a
reduced penalty by certifying that violation(s) was corrected within the 30 day timeframe allowed for in the
agreement. Signing the agreement is strictly voluntary on the part of the regulated facility. Louisiana has a Delivery
Prohibition (Red Tag) program that allows inspectors to red tag tanks at facilities that have certain egregious
violations. The delivery prohibition can happen simultaneously with the enforcement actions listed above.
11 If we were to include enforcement action data in the analysis, the sample would be reduced by approximately 25%
from 5,769 to 4,324 observed inspections, and the observations lost would largely be inspections from prior to the
change in inspection frequency that occurred as a result of EPAct. The loss of these pre-EPAct observations would
significantly reduce the variation in inspection frequency in the sample and our ability to identify the impact of
changes in inspection frequency on compliance.
12 From approximately 2000 to the passing of EPAct in 2005, the Louisiana Regional Department of Environmental
Quality staff identified 15% of the active UST in their region to inspect. Each region had their own systems of
selecting the 15 percent. For example, some just went alphabetically down the site list while others went numerical
by facility number. Also, if one region was overloaded with work and could not inspect 15% of their USTs facilities,
then Louisiana would do more inspections in another region instead.
12
-------
resources could be directed back to inspections, the LADEQ worked on inspecting those
facilities that had not been inspected since 1998, and then towards meeting the requirement of
inspecting each UST facility at least once every three years.
More frequent UST inspections are intended to improve facilities' compliance with UST
release detection and prevention requirements, and in doing so prevent accidental releases of
harmful substances into the environment. From 2001 to 2012, on average each year 2.55 percent
of facilities in the sample had a release confirmed. No clear trend in the percent of facilities with
a release each year is visible, however, interestingly confirmed releases spike in 2008 when the
LADEQ was focused on inspecting those facilities that had not been inspected since 1998.
Trends in noncompliance are more apparent. In the years immediately following EPAct, the
percent of inspected facilities that had at least one noncompliance citation issued increased—
reaching a high of 56% in fiscal year 2008. This increase is likely due to the fact that many of the
facilities inspected during those years were ones that had not been inspected since 1998. From
2009 to 2012, there is a downward trend in the percent of inspected facilities identified as
noncompliant—reaching a low of 33.6% in 2012. Overall this improvement in compliance
coincides with the establishment of the ongoing 3-year inspection requirement in Louisiana but
in order to substantiate that this observed improvement is due to increased inspection frequency
our empirical analysis will account for other factors that may also have impacted compliance.
4. The empirical model
There are three common challenges that arise when measuring the deterrence effects of
environmental monitoring and enforcement: omitted variable bias, measuring facilities'
perceptions about the likelihood of inspections and enforcement, and reverse causality (Gray and
13
-------
Shimshack 2011). The first, omitted variable bias can occur if factors not included in the model
simultaneously affect both regulatory activity and facility compliance. Measuring facilities'
perceptions about the likelihood of inspection and enforcement is difficult because perceptions
are not observable to researchers. The last issue—reverse causality—arises if there is targeting of
facilities by regulators. To minimize these concerns, we use facility-level panel data in a
censored bivariate probit model as detailed in Greene (1992) and Stafford (2002; 2012), include
temporal lags (i.e., examine relationship between current compliance and an UST facility's
compliance status at the last inspection) and control for a variety of facility and location
characteristics that may also affect compliance.
UST violations are detected when a facility is inspected, as opposed to the majority of the
existing empirical work on environmental compliance that has used datasets with self-reported
more frequent (e.g., monthly) observations of whether the firm is in compliance with a regulation
(e.g., Hanna and Oliva 2010; Shimshack and Ward 2005, 2008). If an UST facility is not
inspected, then there is no information about whether or not the facility is in compliance with
regulated UST requirements. Because the data are censored and selection bias may arise if there
is any inspection targeting by regulators, we construct a censored bivariate probit model. The
censored bivariate probit model addresses the selection bias that could occur if there was any
targeting of inspections based on unobserved characteristics of the facilities that would make
them, for example, both more likely to be inspected and more likely to violate, particularly, in
the pre-EPAct years when each regional office in Louisiana had their own systems of selecting
facilities for inspections.
The censored bivariate probit consists of two equations—the selection equation and
outcome equation. Here the selection equation is the probability of an inspection, and the
14
-------
outcome equation is the probability of a noncompliance. As in Stafford (2002; 2012), we model
the probability of noncompliance as the latent variable, IyJt = XVjC(iw + %it, as well the
probability of inspection as the latent variable, Fyt = Jytft + £yt. We define YVjt and Yljt as
binary variables that we observe at UST facility j in quarter l with respect to noncompliance and
inspection, respectively. If facilities are selected for inspection based on unmeasured
characteristics that also make them more likely to violate the UST regulations, the error terms
(%Jt and £iJt) should be positively correlated. Given that a facility is inspected, the likelihood
that a violation will be detected (i.e., YV)t = Yl)t = 1) is expressed as:
LrVjt=i, rjjt-i =1iYVjt=i, Yijt=i PvA/»^iA»p]}j
where Ovl is the bivariate normal cumulative distribution function and p is the covariance
between error terms, and £iJt. The likelihood that no violation will be detected when the
facility is inspected (i.e., YVjt = 0, Yljt = 1) is given by the following expression:
Lryjt=0. Yijt=l =1,YVjt=Q, Yijt=l [—¦XvPv'XiPb —P]}-
If a facility is not inspected, the compliance status of the facility cannot be observed. Therefore,
facilities that are not inspected (i.e., 0), regardless of whether they are compliant or
noncompliant, are observationally equivalent and can be represented as:
£rjjf=o = E YXjt=o logfl — i IXiA.]}-
where Oj is the univariate normal distribution for the inspection equation. The maximum
likelihood function for the censored bivariate probit model can be given by:
15
-------
1 - LrVjt=l. Yljt=i + ^^=0, Y1Jt=1 + LYijt=0
= ErVj't=i, Vi;t=i log{#Vi [Xv(Jv,Xi/?j,p]}
+ ZyV;t=0. yI;t=1 logl^VI [~XvPv>XlPl> —p]} + X Yijt=0 I°g{l Kft]}
For the censored bivariate probit model to be identified, at least one variable that affects the
probability that a facility will be inspected but that does not affect the probability that a facility is
noncompliant should be included in the inspection equation (Wooldridge 2002). The specific
variable used in this case will be discussed at the end of the next section.
5. Explanatory Variables: Definitions and Hypothesized Effect on Noncompliance
As discussed the censored bivariate probit consists of two equations—here the inspection
equation and the noncompliance equation. The dependent variable in the inspection equation is a
dummy variable that is equal to one if facility j is inspected in quarter t. In the violation equation,
the dependent variable is a dummy variable that equals 1 if facility j inspected in quarter l
received at least one noncompliance citation (i.e., if a Notice of Deficiency (NOD) or Notice of
Potential Delivery Prohibition (NOPDP) is issued). For brevity, from here forward, we will refer
to a facility as noncompliant if it had at least one noncompliance citation issued at its inspection.
Both the probability of inspection and noncompliance is expected to depend on a
facility's characteristics and history (i.e., inspection, compliance, and release history) as well as
socioeconomic and biophysical attributes of the facility's location. Table 1 presents descriptive
statistics for the variables used in the inspection and noncompliance equations. We include a
common set of variables (rows 1 to 21) in both equations to account for these factors as well as
variables that are unique to each equation (rows 22 to 24). In this section, we define these
variables and describe their expected relationships with noncompliance, which is the equation of
primary interest in this analysis.
16
-------
The main variable of interest, inspection frequency, is defined as the number of years
since the last inspection (Years Lastlnspection). This continuous measure of inspection
frequency is used rather than a dummy variable that would indicate if the inspection was before
or after the EPAct of 2005 3-year inspection requirement because there is no clear date that
establishes a before period, when inspections were less frequent than three years, and an after
period, when inspections were at least once every three years (i.e., the transition to the 3-year
inspection requirement took several years). Increasing inspection frequency is expected to
improve compliance. As more time passes since a facility's last inspection, owners and operators
may become more lax about keeping up with required standards and procedures, and therefore
would be more likely to have a violation identified when inspected. The estimated coefficient on
Years Lastlnspection is expected to be positive, ceteris paribus.
Last Noncompliance is a dummy variable that takes the value 1 if at least one violation
was detected at a facility's last inspection. The effect of noncompliance at the last inspection on
the probability of noncompliance at the current inspection is not obvious. A noncompliance
citation at the last inspection is expected to have a deterrent effect and to reduce the likelihood
that an UST facility will violate at the current inspection, which means that the expected sign on
the estimated coefficient on Last Noncompliance is negative. However, if the facility believes
that the cost of complying is greater than the benefits of complying, then the facility may return
to a noncompliant state, which means we would expect the estimated coefficient on
Last Noncompliance to be positive. Therefore, the expected sign of the coefficient on
Last Noncompliance is ambiguous.
The magnitude of the effect that the time since the last inspection has on compliance at
the current inspection may differ depending on whether or not the facility had a violation at the
17
-------
last inspection. To allow for the heterogeneous effect of increasing inspection frequency for
those that were identified as noncompliant at last inspection and those that were compliant, the
interaction of inspection frequency and whether the facility was noncompliant at the last
inspection (Years LastInspection*Last Noncompliance) is included.
The empirical model also accounts more broadly for the effect that a facility's past
experience with inspections may have on the facility owner's or operator's compliance behavior.
Overall we would expect that the more compliance inspections a facility had experienced in the
past, the lower the likelihood that facility would receive a noncompliance citation at the current
inspection. The estimated coefficient on the cumulative number of previous inspections is
expected be negative (Total Inspection) because with each additional inspection we would
expect the facility owner's knowledge and understanding of the UST requirements and how to
meet them may improve.
We also include the variable Past Noncompliance, to account for the total number of past
inspections at which a facility had at least one violation detected. Past Noncompliance does not
include the results of the last compliance inspection. While we would expect that over time as
violations are identified at consecutive inspections, a facility's compliance behavior would
eventually improve, it may also be that those facilities with a high number of past inspections at
which violations were identified are chronic offenders that will habitually violate. It may be that
for that facility the cost of complying is greater than the benefits of complying. If the effect of
facilities that violate all the time dominates, then the coefficient on Past Noncompliance will be
positive. If the learning effect dominates, then the coefficient will be negative.
After a facility has an accidental release from an UST, the owners may become more
vigilant about following UST requirements to prevent and detect releases in order to avoid the
18
-------
potential costs of a release that are now more concrete in the facility owner's expected costs of
noncompliance. To account for this, we include Last Release, a dummy variable that is equal to
1 if there has been a release since the last inspection, and expect the coefficient on Last Release
to be negative.
The empirical model also accounts for the effect that UST facility characteristics may
have on the likelihood of noncompliance. The expected cost of noncompliance with UST
requirements depends on expected penalties and expected cleanup costs if a release occurs.
Expected cleanup costs depend on the probability of a release and the size of a release, which
depends on both tank technology and preventive measures taken by the owner or operator (e.g.,
reconciling inventories, inspecting and maintaining sumps and spill buckets, etc.). For each
facility, we include the number of tanks {Number Tanks), age of oldest tank (Age Oldesi 'l'ank),
and average capacity of the tanks (Mean TankCapacity) at the facility.
Expected cleanup costs also depend on the likelihood that a release will contaminate
groundwater. The average cleanup cost of a leaking UST is approximately $152,000 (2012
dollars) but the cost can be significantly higher if groundwater is affected (U.S. EPA 2015). A
release is more likely to contaminate groundwater if the soil in the area is more permeable
(SoiljVlostPermeable) or if the water table is closer to the surface (Depth Water Table). We
expect the coefficient on Soil MostPermeable to be negative (i.e., facilities located in areas with
the most permeable soils are less likely to be noncompliant), and the coefficent on
Depth WaterTable to be positive (i.e., facilities located in areas where the water table is further
from the surface are more likely to be noncompliant).
Socioeconomic attributes of the communities surrounding UST facilities are included to
proxy for the demand for petroleum product from nearby consumers. The higher the demand for
19
-------
the product, the more frequent withdrawals and deliveries will be at the facility, and thus the
more inventory oversight required (i.e., higher compliance costs). We include population density
{Density Population) and median income (Income Median) in the census block group where the
facility is located. Based on the increased demand for petroleum products that higher population
density or median income would generate, we would expect the signs of the coefficent on both of
these variables to be positive, however, in high income or highly populated areas there may also
be higher demand for environmentally responsible businesses as well as higher potential costs to
a facility should a release of petroleum occur. The sign of the coefficient on Density Population
and Income Median will depend on which effect dominates—the increased demand for product
leading to higher compliance costs or the increased demand for environmentally responsible
businesses. Furthermore, we include Louisiana fiscal year quarterly dummy variables to control
for seasonality in the demand for products at the UST facilities.
Staff at regional LADEQ offices are responsible for coordinating compliance inspections
in their respective regions. We include regional dummy variables to account for any potential
regional differences in how inspections are conducted by inspectors. In defining the regional
dummy variables, we exclude the capital region (Region Capital). We also include the distance
of the facility to the LADEQ regional field office (Distance FieldOffice). A facility located
closer to the field office may have greater knowledge of UST requirements and awareness of
regulator presence. We expect the coefficient on Distance FieldOffice to be positive, that is; the
closer a facility is to the field office, the lower the likelihood that the facility will violate.
Inspectors are either employees of the LADEQ or contracted by the LADEQ. Both types
of inspectors receive the same training and use the same compliance evaluation inspection
checklist to conduct inspections. The only difference is that the LADEQ does not allow contract
20
-------
inspectors to issue compliance order letters or apply delivery prohibition red tags—instead the
LADEQ will do that for them. To capture the effect that the type of inspector may have on the
probability of a violation we include Contract Inspector that takes the value 1 when the
inspector was contracted by the LADEQ. We exclude Contract Inspector from the inspection
equation because we only observe inspector type when there is an inspection.
State OperatorTraining is a dummy variable that takes the value 1 if a compliance
inspection occurred after the start of operator trainings in Louisiana. The EPAct of 2005 also
required state and territorial UST programs receiving federal funds to require UST systems to
have designated UST system operators and to develop state-specific operator training
requirements that meet EPA's grant guidelines. The federal deadline to have designated UST
system owners and operators trained was August 8, 2012, though some states, such as Louisiana,
established earlier deadlines.13 The first operator training was held in Louisiana on March 9,
2010.14 Unfortunately, facility specific data on operator training status was not available.
Therefore, to account for the fact that during this time period some operators and owners may
have learned additional information on UST maintenance, testing, and recordkeeping, we include
the dummy variable State OperatorTraining in the compliance equation. We expect that after
operator trainings were held in Louisiana the probability of a violation would be lower. We
exclude State OperatorTraining from the inspection equation as we would not expect the fact
that operator trainings are being held in Louisiana to affect the probability of inspection at any
given facility.
13 Louisiana had a phase-in period for operator training based on compliance inspection dates. Facilities inspected
between February 20, 2010 and November 8, 2011 had to have their operators trained within 9 months of their
inspection date. Everyone else had to be trained by August 8, 2012, which was the federal deadline to have
designated UST system owners and operators trained.
14 After the state deadline, operator training requirements became part of compliance inspections. Since this added a
new major component to the compliance inspection, we do not include compliance inspections conducted after the
federal deadline (August 8, 2012).
21
-------
For the censored bivariate probit model to be identified, at least one variable that affects
the probability that a facility will be inspected but that does not affect the probability that a
facility will be noncompliant should be included in the inspection equation (Wooldridge 2002).
For identification purposes, we include the total annual number of hurricane related visits made
by LADEQ UST inspectors to facilities, State TotalHurricaneVisits, in the inspection equation
but exclude it from the noncompliance equation. State TotalHurricaneVisits reflects changes in
the resource constraint of the LADEQ. We expect that when the total number of hurricane
related visits is higher the resources available to conduct compliance inspections is reduced but
that the total number of hurricane related visits would not affect the probability of a violation at
an inspected facility. In fact, when State TotalHurricaneVisits was included in the
noncompliance equation, it was not significant (Coef.= -0.0002; p=0.254).
6. Results
Results of the censored bivariate probit regression are presented in Table 2.15 In this
section, we briefly discuss estimates from the inspection equation before turning the focus to the
main results from the noncompliance equation. Louisiana's inspection strategy over the study
period, particularly post-EPAct, should be largely determined by the time since last inspection
(Years Lastlnspection). This is evident in the inspection equation estimates, where the
coefficient on Years Lastlnspection is positive and significant at the 1% level. Interestingly,
results suggest that other factors affect the probability of an inspection. The probability of an
inspection is higher when a facility has an older tank, a higher mean tank capacity, or fewer
15 We used Stata 13 's heckprobit command, which estimates the censored bivariate probit model. We also use
clustering to account for non-independence of inspections and compliance outcomes from a single UST facility in
our unbalanced panel dataset to allow for potential within-groups (facilities) correlation while modeling econometric
error (Rogers 1993; Williams 2000; Wooldridge 2002).
22
-------
tanks; is located in an area where the water table is further from the surface; cumulatively has
had more inspections; or has had a release since the last compliance inspection. The probability
of an inspection is lower when the LADEQ's UST inspector resources are constrained (i.e., when
a higher total annual number of hurricane related visits were conducted by LADEQ inspectors at
UST facilities).
6.1 Effect of Inspection Frequency on Compliance
We now turn to the main hypothesis of the paper: Does increasing inspection frequency
improve compliance? Results suggest that increasing inspection frequency to at least every three
years as required by EPAct has improved compliance with UST requirements in Louisiana. The
results also show that the magnitude of the effect differed depending on whether or not the
facility had a violation at the last inspection. For those facilities that were compliant at the last
inspection (i.e., Years Lastlnspection* Last Noncompliance=0), the coefficent on
Years Lastlnspection is positive and statistically significant at the 10% level (Table 2). For those
facilities that were noncompliant at the last inspection, the effect (i.e., the linear combination of
the Years Lastlnspection and Years Lastlnspection* Last Noncompliance) is also positive and
statistically significant but lower in magnitude. This suggest that increasing inspection frequency
increases the likelihood of noncompliance more at facilities that were compliant at the last
inspection than those that were noncompliant. While this may seem counterintuitive, it suggests
that for those facilities that were in compliance at the last inspection it may be easier for them to
maintain compliance with UST prevention and release requirements than for facilities with a
compliance violation at the last inspection that would need to take new actions to achieve
compliance. To better illustrate the impact of the 3-year inspection requirement of EPAct on
UST owners and operators compliance with UST regulations based on our results, we estimate
23
-------
how changes in inspection frequency at a hypothetical representative facility in Louisiana affect
the probability of noncompliance. The hypothetical representative facility has the mean values
for all continuous explanatory variables, the mean value for noncompliance at the last inspection,
and median values for all other discrete explanatory variables (see Table 1 for details). We use
the estimates of the censored bivariate probit model presented in Table 2 and the representative
facility's characteristics to estimate the predicted probability that a facility will be noncompliant
at the time of inspection for years since last inspection ranging from three to six years (Figure 3).
We will focus on the difference in the predicted probability of noncompliance when the time
since the last inspection has been six years, a representation of pre-EPAct, versus three years, the
requirement under the EPAct of 2005. We use six years as representative of pre-EPAct because
prior to the EPAct 3-year requirement, Louisiana had a requirement that 15% of the UST
facilities in Louisiana be inspected each year, which is approximately equivalent to a 6-year
cycle if they did not return to the same facility for a second inspection until all other facilities
had been inspected. Predicted probabilities from the censored bivariate probit model estimated
coefficients show that moving from a 6-year to a 3-year inspection cycle reduces the likelihood
that a representative facility will receive a noncompliance citation at the time of inspection by
about 11% (Table 3).
To illustrate the differing effect that increasing inspection frequency has on compliance
depending on the results of a facility's last compliance inspection, we also estimate predicted
probabilities of noncompliance for a hypothetical representative facility that was noncompliant at
the last inspection and for one that was compliant at the last inspection. The reduction in the
likelihood of noncompliance moving from a 6-year to a 3-year inspection cycle is larger for
24
-------
facilities that were compliant at their last inspection (about 13%) relative to the facilities that
were noncompliant at their last inspection (about 9%).
6.2 Effect of Other Explanatory Variables on Compliance
As expected, Total Inspection had a negative and statistically significant coefficient,
suggesting that the more compliance inspections a facility has experienced in the past, the lower
the probability that the facility would be noncompliant at the current inspection. The coefficients
on Last Noncompliance and Past Noncompliance were both positive and statistically significant
at the 1% level. If a facility was noncompliant at the last inspection or cumulatively had a greater
number of past inspections where it was noncompliant, the probability of noncompliance at the
current inspection is higher. This suggests that the effect of chronic offenders (i.e., those that
habitually violate) dominates. It may also be that these variables are capturing an unobserved or
omitted variable such as corporate culture that makes a facility consistently more or less likely to
comply with requirements.
UST facilities with an older tank (Age OldestTank) or a lower mean tank capacity
(Mean TankCapacity) were more likely to violate UST regulations. Older tanks and tanks with
lower average capacity may have less advanced technologies that make it more challenging for
facilities to maintain a compliant status. For example, older or smaller tanks may use a dip stick
to reconcile tank inventory rather than an electronic inventory reconciliation device. Also, single
facility owners are more likely to have older and smaller tanks and it may be challenging for
single facility owners to meet UST requirements given all the other requirements simultaneously
placed on them as a small business (e.g., Occupational Safety and Health Administration laws
and regulations, fire prevention codes, food codes, tobacco and liquor sale laws, etc.).
25
-------
We found no significant effect of the socioeconomic characteristics of communities and
bio-physical characteristics around UST localities on the likelihood of noncompliance except for
water table depth. As expected, a facility located in an area where the water table is further away
from the surface was more likely to be noncompliant.
A facility inspected by a contracted inspector was more likely to be noncompliant than
one inspected by a state-employed inspector (Contract Inspector). This seems counterintuitive,
however, state-employed inspectors may feel that they have the authority to allow a facility some
leeway whereas the contracted inspector may not have such sense of authority. For example, at
the time of inspection if there was only one minor issue that could be resolved while the
inspector was onsite, a state-employed inspector may not issue a citation whereas the contracted
inspector may feel obligated to cite the facility.
As expected the coefficient on State Operator Training, which indicates whether or not
an inspection occurred after Louisiana began holding operator trainings, was negative. This
suggests that even though all owners and operators were not yet trained the presence of operator
trainings in Louisiana reduced the likelihood that an inspected facility would have a violation
detected. This effect on the likelihood of UST noncompliance is attributable to operator trainings
to the extent that the dummy for the time period is not capturing other unobservable factors that
are unique to that timeframe and influence UST compliance decisions.
In the censored bivariate probit model, rho measures the correlation of the residuals from
the two equations. Here the coefficient on rho is not statistically significant. This suggests that
the regulators decision to inspect a given UST facility in Louisiana may not in fact be
26
-------
endogenous to inspection or compliance outcomes (Wald Test of Independent Equations (p=0),
Chi-squared (1) = 0.47; p=0.4925).16
6.3 Sensitivity Analysis
To assess the robustness of our findings with respect to inspection frequency and
compliance, we explored a number of alternative models.17 First, given the insignificance of the
correlation coefficient between the residuals from the compliance and inspection equations, we
estimate a probit model for the compliance equation. Second, given potential measurement error
from defining a binary measure of noncompliance (i.e., at least one compliance citation indicates
noncompliant) rather than the number of citations, we estimated a poisson regression for the
noncompliance equation where the outcome variable was the number of citations. The results
for both of these estimations are qualitatively similar and consistent with our main results. Most
importantly, the coefficient on years since the last inspection remains positive and statistically
significant. Lastly, one potential limitation of this analysis is our inability to account for
enforcement actions beyond the initial compliance citations (NODs and NOPDPs) due to lack of
data in pre-EPAct years. We estimate a censored bivariate probit model for the reduced sample
(primarily consisting of post-EPAct inspections) with and without these enforcement action
variables. The results using the reduced sample suggest that excluding these enforcement action
variables does not change the effect that other explanatory variables have on noncompliance.
Therefore, it is a reasonable assumption that our main results are robust to the exclusion of the
additional enforcement action data.
16 We also estimated the noncompliance equation using a probit model. Results are available upon request from
authors. Note that while there are small changes in the magnitude of coefficients and significance level, the overall
conclusions do not change between the probit and the censored bivariate probit estimation of the noncompliance
equation.
17 Results of these analyses are available upon request from the authors.
27
-------
7. Conclusions
This paper examines the effect of inspection frequency on compliance decisions in an
environmental pollution prevention context by capitalizing on policy changes occurring under
EPAct of 2005 that increased inspection frequency requirements for UST facilities to at least
once every three years. Specifically, we used facility-level data on inspection, compliance,
releases and other socio-economic and biophysical characteristics of UST localities to examine
this in Louisiana. A censored bivariate probit model was used to account for the censored nature
of the inspection and compliance data and to account for potential bias in estimates due to
inspection targeting that may have occurred, particularly in pre-EPAct years. Results suggest that
increasing inspection frequency improved UST facilties' compliance in Louisiana and this
impact is heterogeneous based on a facility's compliance status at the last inspection—larger
impact for those facilities that were compliant than those that were noncompliant at their last
inspection. This result is consistent with previous empirical literature that has consistently shown
that inspections improve compliance across a variety of environmental regulation contexts
(Shimshack 2014). Furthermore, our study illustrates the effect of more frequent inspections and
finds a heterogeneous effect across facilities based on the compliance status at the last
inspection.
The aim of increasing inspection frequency at UST facilities is to prevent and to reduce
the size of accidental releases of petroleum and other hazardous substances into the environment.
The greatest potential hazard from a leaking UST is that petroleum or other hazardous substances
can seep into the soil and contaminate groundwater, the source of drinking water for nearly half
of all Americans (USGS 2003). A release from an UST can also present other environmental and
health risks, including potential for fire and explosion, neurological damage, blood disorders,
28
-------
cancer, and other adverse health outcomes (Jenkins et al. 2014; Marcus 2016). Furthermore,
when releases are prevented, remediation costs are avoided, which represents cost savings that
accrue to owners, operators and public entities charged with remediating contaminated media at
regulated facilities. Remediation costs will vary depending on site size and the media
contaminated. The average cleanup cost of a leaking UST is approximately $152,000 but the cost
can be significantly higher if groundwater is affected (US EPA 2015b).
While beyond the scope of this analysis, it would be informative for future research to
examine the relationship between increased inspection frequency and the prevention of UST
releases. Many empirical studies have examined the role of Occupational Safety and Health
Administration inspections in the context of preventing workplace injuries (e.g., Viscusi 1979;
Scholz and Gray 1990; Gray and Mendeloff 2005; Haviland et al. 2010; Levine, Toffel and
Johnson 2012, etc.) but there are few in the context of preventing accidental releases of
hazardous materials (Epple and Visscher 1984; Cohen 1987; Grau and Groves 1997; Talley, Jin
and Kite-Powell 2005). Conclusions in the context of preventing workplace injuries have been
mixed, while those in the context of reducing vessel oil transfer spills have been more
consistent—generally, that Coast Guard inspection and enforcement activities have been
effective in reducing spills during vessel oil transfer. Due to data limitations, it is difficult to
identify the impact of increased inspection frequency on the prevention of UST releases. A key
difference between data on releases at UST facilities and data on workplace injuries and oil
transfer spills is that the dates recorded for workplace accidents and oil transfer spills are
typically the actual date of occurrence whereas for UST releases the date recorded is the
confirmed (or discovery) date of the release (i.e., the actual date when the UST release occurred
is often unknown or uncertain). Therefore, it may be challenging to determine if a release
29
-------
occurred before or after a particular inspection date and to identify the direct impact of increased
inspections frequency on preventing releases at UST facilities.
Given the limited ability to analyze the impact of inspections on UST releases, a cost-
benefit analysis is not feasible and beyond the scope of this analysis. However, for some
perspective on the relative costs of inspections and potential benefits, consider the following
back-of-the envelope calculation. The cost of conducting inspections is estimated at $96,348 per
inspector with each completing 200 compliance inspections (US EPA 2000).18 In Louisiana,
there are roughly 4,400 UST facilities to be inspected. To inspect all of those facilities, the
inspector cost is estimated to be roughly $2.12 million dollars. From fiscal year 2014-2016, 487
UST cleanups were completed in Louisiana with an average cleanup cost of $297,448 (Louisiana
Department of Environmental Quality, personal communication, December 7, 2016). If the
improved compliance from increased inspection frequency led to just 7.13 fewer releases that
required cleanups over the course of 3 years, then the potential benefits of avoided cleanup costs
would outweigh the direct cost of compliance inspections. Note that this comparison is for
illustrative purposes only as it does not capture the full costs and benefits of UST compliance
inspections. Specifically, it neither includes costs associated with training inspectors,
enforcement, state administrative oversight nor UST owners' compliance costs.19 Furthermore, it
does not include additional potential benefits accruing from avoided product loss and negative
impacts on nearby property values, human health and ecosystem services that may be
18 In 2000, annual inspector cost were estimated at $70,000 includes salary, travel costs, benefits, managerial and
secretarial support, and inspector equipment. To compare to the cleanup costs, which are the average from 2014-
2016, this inspector cost of $70,000 was adjusted using the Consumer Price Index, and is equivalent to $96,348 in
2015 dollars.
19 Estimated direct compliance costs for individual facilities with UST release detection and prevention requirements
in the final revisions to EPA's Underground Storage Tank Regulations are small at approximately $715 per year for
the average facility (US EPA 2015b).
30
-------
substantial.20 This analysis provides evidence that increased inspection frequency due to the 3-
year inspection mandate of the Energy Policy Act of 2005 improved compliance in Louisiana.
Future research should examine the relationship between improved compliance and impact on
UST release prevention.
20 For more information on estimated benefits of compliance with UST release detection and prevention
requirements based on expert elicitations, refer to the "Assessment of the Potential Costs, Benefits, and Other
Impacts of the Final Revisions to EPA's Underground Storage Tank Regulations" (US EPA 2015b).
31
-------
Funding: US Environmental Protection Agency
Conflict of Interest: Karen Sullivan is an employee of the US Environmental Protection
Agency. Achyut Kafle is an ORISE Post-doctoral Fellow at the US Environmental Protection
Agency.
32
-------
References
Alberini, A., and D. Austin. 1999. "Strict liability as a deterrent in toxic waste management:
Empirical evidence from accident and spill data. "Journal of Environmental Economics
and Management 38:20-48.
Alberini, A., and D. Austin. 2002. "Accidents waiting to happen: Liability policy and toxic
pollution releases." Review of Economics and Statistics 84:729-74.
Alberini, A., E. Lichtenberg, D. Mancini, and G. Galinato. 2008. "Was it something I ate?
Implementation of the FDA seafood HACCP program." American Journal of
Agricultural Economics 90 (1):28—41.
Becker, G.S. 1968. "Crime and punishment: An economic approach." Journal of Political
Economy 76:169-217.
Cohen, M.A. 1987. "Optimal enforcement strategy to prevent oil spills: An application of a
principal-agent model with moral hazard." Journal of Law and Economics 30:23-51.
Cyert, R., and J. G. March. 1963. A Behavioral Theory of the Firm. Englewood Cliffs, NJ:
Prentice-Hall.
Deily, M.E., and W.B. Gray. 2007. "Agency structure and firm culture: OSHA, EPA, and the
steel industry." Journal of Law and Economics 23:685-709.
Earnhart, D. 2004a. "Panel data analysis of regulatory factors shaping environmental
performance." Review of Economics and Statistics 86:391-401.
Earnhart, D. 2004b. "Regulatory factors shaping environmental performance at publicly-owned
treatment plants." Journal of Environmental Economics and Management 48:655-81.
Eckert, H. 2004. "Inspections, warnings, and compliance: the case of petroleum storage
regulation." Journal of Environmental Economics and Management 47:232-59
Epple, D., and M. Visscher. 1984. "Environmental pollution: modeling occurrence, detection,
and deterrence." Journal of Law and Economics 27:29-60.
Glicksman, R.L., and D.H. Earnhart. 2007. "Comparative effectiveness of government
interventions on environmental performance in the chemical industry." Stanford
Environmental Law Journal 26:317.
Greene, W. H. 1992. "A statistical model for credit scoring." Working Paper no. EC-92-29,
Department of Economics, Stern School of Business, New York University, New York,
NY.
33
-------
Grau M., and T. Groves. 1997. "The oil spill process: the effect of coast guard monitoring on oil
spills." Environmental and Resource Economics 10:315-39.
W., and M. Deily. 1996. "Compliance and enforcement: air pollution regulation in the US
steel industry." Journal of Environmental Economics and Management 31:96-l 11.
W., and J. Mendeloff 2005. "The declining effects of OSHA inspections on
manufacturing injuries, 1979-1998." Industrial and Labor Relations Review 58: 571-87.
W., and R. Shadbegian. 2005. "When and why do plants comply? Paper mills in the
1980s." Law and Policy, 27:238-61.
W., and J. Shimshack. 2011. "The effectiveness of environmental monitoring and
enforcement: A review of the empirical evidence." Review of Environmental Economics
and Policy 5:3-24.
Guignet, Dennis, Robin R. Jenkins, Matthew Ranson, and Patrick J. Walsh. 2016. "Do Housing
Values Respond to Underground Storage Tank Releases? Evidence from High-Profile
Cases across the United States." National Center for Environmental Economics Working
Paper Number 2016-01.
Hanna, R., and P. Oliva. 2010. "The impact of inspections on plant-level air emissions." B.E.
Journal of Economic Analysis and Policy 10:1-31.
Haviland, A., R. Burns, W. Gray, T. Ruder, and J. Mendeloff. 2012. "A new estimate of the
impact of OSHA Inspections on Manufacturing Injury Rates, 1998-2005. "American
Journal of Industrial Medicine 55: 964-975.
Heyes, A. G. 1994. "Environmental enforcement when "inspectability" is endogenous: a model
with over-shooting properties." Environmental and Resource Economics 4:479-94.
Heyes, A. G. 2000. "Implementing environmental regulation: enforcement and compliance."
Journal of Regulatory Economics 17:107-29.
Innes, R. 1999a. "Remediation and self-reporting in optimal law enforcement." Journal of Public
Economics 72:379-93.
Innes, R. 1999b. "Self-policing and optimal law enforcement when violator remediation in
valuable." Journal ofPolitical Economy 107: 1305-25.
Innes, R. 2001. "Violator avoidance activities and self-reporting in optimal law enforcement."
Journal of Public Economics 72:379-93.
Jenkins, Robin R., Dennis Guignet, and Patrick J. Walsh. 2014. "Prevention, Cleanup, and Reuse
Benefits from the Federal UST Program." National Center for Environmental Economics
Working Paper No. 2014-05.
34
Gray,
Gray,
Gray,
Gray,
-------
Kaplow, L., and S. Shavell. 1994. "Optimal law enforcement with self-reporting of behavior."
Journal of Political Economy 102:583-606.
Keohane, N. E. Mansur, and A. Voynov. 2009. "Averting regulatory enforcement: evidence from
new source review." Journal of Economics and Management Strategy 18:75-104.
Ko K., J. Mendeloff, and W. Gray. 2010. "The role of inspection sequence in compliance with
US Occupational Safety and Health Administration's (OSHA) standards: Interpretations
and implications." Regulation & Governance 4: 48-70.
Langpap C., and J. Shimshack. 2010. "Private citizen suits and public enforcement: Substitutes
or complements?" Journal of Environmental Economics and Management 59: 235-249.
Levine, D.L., M.W. Toffel, and M.S. Johnson. 2012. "Randomized Government Safety
Inspections Reduce Worker Injuries with No Detectable Job Loss." Science, May 18.
Liu, L. 2012. "Spillover effects across environmental programs: the case of hazardous waste
regulation in Michigan." Environmental Economics 3(2):35-43.
Magat,W., and W.K. Viscusi. 1990. "Effectiveness of the EPA's regulatory enforcement: The
case of industrial effluent standards." Journal of Law and Economics 33: 331-60.
Malik, A.S. 1990. "Avoidance, screening and optimum enforcement. " Rand Journal of
Economics 21:341-53.
Malik, A.S. 1993. "Self-reporting and the design of policies for regulating stochastic pollution."
Journal of Environmental Economics and Management 24:241-57.
Marchus, M. "Going Beneath the Surface: Petroleum Pollution, Regulation, and Health" Job
Market Paper, Brown University, Department of Economics. November 8, 2016.
Mendeloff, J., and W. Gray. 2005. "Inside the Black Box: How do OSHA Inspections Lead to
Reductions in Workplace Injuries?" Law and Policy 27(2): 219-237.
Nadeau, L. 1997. "EPA effectiveness at reducing the duration of plant-level non-compliance."
Journal of Environmental Economics and Management 34:54-78.
Rogers, W. H. 1993. "Regression standard errors in clustered samples." Stata Technical Bulletin
13: 19-23. Reprinted in Stata Technical Bulletin Reprints, vol. 3, 88-94.
Russell, C.W. Harrington, and W. J. Vaughan. 1986. Economic models of monitoring and
enforcement: Enforcing pollution control laws. Resources for the Future, Washington,
DC.
Scholz J. and W.B. Gray. 1990. OSHA enforcement and workplace injuries: A behavioral
approach to risk assessment. Journal of Risk Uncertainty 3(3): 283-305.
35
-------
Shimshack, J. 2014. "The economics of environmental monitoring and enforcement." Annual
Review of Resource Economics 6(1): 339-360.
Shimshack, J., andM.B. Ward. 2005. Regulator reputation, enforcement, and environmental
compliance. Journal of Environmental Economics and Management 50: 519-40.
Shimshack, J., and M.B. Ward. 2008. "Enforcement and over-compliance." Journal of
Environmental Economics and Management 55 (1): 90-105.
Stafford, S.L. 2002. The effect of punishment on firm compliance with hazardous waste
regulations. Journal of Environmental Economics and Management. 44: 290-308.
Stafford, S.L. 2003. Assessing the effectiveness of state regulations and enforcement of
hazardous waste. Journal of Regulatory Economics 23: 27-41.
Stafford, S.L. 2012. Do carrots work? Examining the effectiveness of EPA's compliance
assistance program. Journal of Policy Analysis and Management 31(3): 533-555.
Stafford, S.L. 2014. Will additional federal enforcement improve the performance of pipelines in
the U.S.? International Review of Law and Economics. 37(C): 137-146.
Talley, Wayne K., Di Jin, and Hauke Kite-Powell. 2005. Post OPA-90 Vessel Oil Transfer Spill
Prevention: The Effectiveness of Coast Guard Enforcement. Environmental & Resource
Economics 30: 93-114.
Thornton, D., N., Gunningham, and R. Kagan. 2005. General deterrence and corporate
environmental behavior. Law and Policy, 27:262-288
U.S. Census Bureau. "Summary File." 2009-2013 American Communities Survey. U.S. Census
Bureau's American Community Survey Office, 2015.
U.S. Department of Agriculture. Natural Resources Conservation Service, Soil Survey Staff.
Web Soil Survey, http://websoilsurvey.nrcs.usda.gov/ (accessed February 2015).
U.S. Government Accountability Office. 2001. Improved Inspections and Enforcement Would
Better Ensure the Safety of Underground Storage Tanks. GAO-Ol-464, Washington, D.C.
U.S. Geological Survey. 2003. Ground-Water Depletion across the Nation. U.S. Geological
Survey Fact Sheet 103-03. http://pubs.usgs.gOv/fs/fs-103-03/#pdf (last updated January 9,
2013).
U.S. Environmental Protection Agency. 2000. Report to Congress on a Compliance Plan for the
Underground Storage Tank Program. EPA 510-R-00-001.
U.S. Environmental Protection Agency. 2005. Semiannual Report of UST Performance
Measures: End of Year Fiscal Year 2005 (April 1, 2005-September 30, 2005).
Washington, D.C.
36
-------
U.S. Environmental Protection Agency. 2011. Assessment of the Potential Costs, Benefits, and
Other Impacts of the Proposed Revisions to EPA 's Underground Storage Tank
Regulations. Prepared by Industrial Economics, Incorporated. August.
http://www.epa.gov/sites/production/files/2015-07/documents/regs2015-ria.pdf (accessed
January 28, 2016).
U.S. Environmental Protection Agency. 2015 a. Semiannual Report of UST Performance
Measures: End Fiscal Year 2015 (October 1, 2014-September 30, 2015). Washington,
DC.
U.S. Environmental Protection Agency. 2015b. Assessment of the Potential Costs, Benefits, and
Other Impacts of the Final Revision to EPA 's Underground Storage Tank Regulations.
Prepared by Industrial Economics, Incorporated. April.
http://www.regulations.gOv/#! documentDetaii;D=EP A-HQ-UST-2011-0301 -0472
(accessed February 18, 2016)
U.S. Environmental Protection Agency. 2016. Semiannual Report of UST Performance
Measures: End Fiscal Year 2016 (October 1, 2015-September 30, 2016). Washington,
DC.
US Government Printing Office. 2013. Electronic Code of Federal Regulations Part 280.
Accessed Dec 1, 2106 at http://www.ecfr.gov
Viscusi, W.K. 1979. :"The Impact of Occupational Safety and Health Safety and Health
Regulation." Bell Journal of Economics 10(1): 117-140.
Williams, R. L. 2000. "A note on robust variance estimation for cluster-correlated data."
Biometrics 56: 645-646.
Wooldridge, J. M. 2002. Econometric Analysis of Cross Section and Panel Data. Cambridge,
MA: MIT Press.
37
-------
Table 1: Summary statistics of explanatory variables
Std.
Variable Mean Dev. Median
Y earsLastlnspection
3.65
1.57
3.02
T otal Inspection*
1.48
0.64
1
LastN oncompli ance
0.46
0.50
0
PastNoncompliance'
0.19
0.43
0
LastRelease'
0.05
0.23
0
NumberTanks'
2.82
1.17
3
Age OldestTank (years)
21.69
9.77
21.68
MeanTankCapacity (1000's of gallons)
8.36
3.97
8
Depth WaterTable (meters)
0.47
0.37
0.31
SoilMostPermeable'
0.43
0.50
0
Di stanceFiel dOffi ce
20.94
16.16
17.08
Density Population (100's people/sq mile)
13.82
19.53
6.10
IncomeMedian (100's of USDs)
43.83
19.63
40.83
Regi on_A cadi ana '
0.18
0.38
0
Region NE"'
0.19
0.39
0
RegionNW'
0.11
0.31
0
RegionSE'
0.23
0.42
0
RegionSW'
0.10
0.30
0
LAFiscalYear Q2^
0.20
0.40
0
LAFiscalYear Q3^
0.26
0.44
0
LAFiscalYear Q4^
0.30
0.46
0
Contractlnspector
0.49
0.50
0
StateOperatorTraining"'
0.46
0.50
0
State TotalHurricaneVisitsb
20.14
117.09
0
Notes: The variables marked by a were included only in the noncompliance equation whereas
the variables marked by b was included only in the inspection equation. The variables marked
by t are the discrete variables for which the median value was used in calculating the
predicted probabilities in table 3.
38
-------
Table 2: Estimation results of the censored bivariate probit model
Inspection Noncompliance
Equation Equation
Variable
Coefficient
Standard
Error
Coefficient
Standard
Error
Constant
-2.9141*"
0.053
-0.5047
0.549
Y earsLastlnspection
0.3322*"
0.007
0.0814*
0.049
Y ears_LastInspection* Last_Noncompliance
0.0306**
0.012
-0.0407*
0.022
T otallnspection
0.2077***
0.012
-0.1109"
0.047
Last_Noncompliance
-0.0272
0.030
0.5484***
0.090
Past_Noncompliance
-0.0004
0.014
0.1444***
0.051
LastRelease
0.3975***
0.038
-0.1126
0.093
Number_Tanks
-0.0212***
0.006
-0.0205
0.015
AgeOlde stTank
0.0050***
0.001
0.0120***
0.002
MeanTankCapacity
0.0135***
0.002
-0.0240***
0.005
DepthWaterTable
0.0397**
0.018
0.1560***
0.052
Soil_MostPermeable
-0.0221
0.015
-0.0318
0.039
DistanceFieldOffice
-0.0002
0.000
0.0013
0.001
DensityPopulation
-0.0003
0.000
0.0007
0.001
Income Median
-0.0002
0.000
-0.0002
0.001
RegionAcadiana
-0.0631***
0.021
0.0698
0.059
Region NE
0.0385*
0.022
0.0809
0.058
Region_NW
0.1100***
0.021
0.1973***
0.069
RegionSE
0.0608***
0.023
0.1611***
0.058
Region_SW
-0.0415*
0.025
0.3287***
0.072
LAFiscalY ear_Q2
-0.0956***
0.022
0.0049
0.055
LAFiscalY ear_Q3
-0.0116
0.020
-0.0261
0.049
LAFiscalYear Q4
0.0316
0.020
-0.0023
0.049
Contractlnspector
0.1003***
0.037
StateOperatorT raining
-0.1835***
0.045
StateT otalHurricane V isits
-0.0009***
0.000
P
-0.1125
0.164
Log-likelihood: -22,169
Number of facilities: 4,424
Censored Observations: 102,512
Uncensored Observations: 5,769
Notes: Cluster-robust standard errors. Statistical significance at the 1%, 5% and 10% are represented by ***, **, and
*, respectively.
39
-------
Table 3: Predicted probability of noncompliance at a hypothetical representative facility
Predicted Pr{Noncompliance)
Years Since Last Inspection Change m Predlcted
6 Years 3 Years pr (Noncompliance)
Last Noncompliance=Mean
0.49***
0.38***
-0.11
(0.026)
(0.023)
Last Noncompliance=0
0.44***
0 3i*«
-0.13
(0.027)
(0.023)
Last Noncompliance=1
r\ — /-***
0.56
0.47***
-0.09
(0.029)
(0.026)
Notes: The hypothetical representative facility has the mean values for all continuous explanatory variables, the
mean value for noncompliance at the last inspection, and the median values for all other discrete explanatory
variables. See Table 1 for means and medians. Standard errors in parentheses. Statistical significance at the 1%, 5%
and 10% are represented by ***, **, and *, respectively.
40
-------
Figures
Figure 1. National uunderground storage tank compliance rate
Fiscal Year
Note: The Energy Policy Act of 2005 provisions for underground storage tanks include a
transition phase from August 8, 2005 to August 8, 2007 during which all states receiving
Subtitle I funding for their UST programs are required to inspect all active UST facilities that
had not been inspected since 1998. After August 8, 2007 the time between concurrent
compliance inspections at an UST facility cannot exceed three years.
41
-------
Figure 2. Louisiana inspection, compliance and confirmed releases (fiscal year 2001-2012)
80
70
60
e 50
1.
30
20
10
0
Hurricane
Katrina &
Rita
Energy
Policy Act
Transition
Period in
Louisiana
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
Louisiana Fiscal Year
Percent of Facilities Inspected
Percent of Inspected Facilities Issued Noncompliance Citation(s)
Percent of Facilities with a Confirmed Release
Note: The Energy Policy Act of 2005 provisions for underground storage tank inspections
included a transition phase from August 8, 2005 to August 8, 2007 during which states were
required to inspect all active UST facilities that had not been inspected since 1998. This
transition period was delayed in Louisiana due to Hurricane Katrina and Rita (August and
September 2005, respectively).
42
-------
Figure 3. Predicted probability of noncompliance at typical facility
4 5
Years Since Last Inspection
Note: Shaded area represnts the 95% confidence intervals.
Note: The hypothetical representative facility has the mean values for all continuous explanatory
variables, the mean value for noncompliance at the last inspection and the median values for all
other discrete explanatory variables. See Table 1 for means and medians.
43
------- |