nnA unit^jsutw
£nv-ronrr«it»l P*o«*Ct»OH
Office of the
Administrator
EPA 240-F-14-001
August 2014

-------
RetSrospfficMve Study
off mine Cosfts off EIPA
A Repoirt of Four
Case Studies
August 2014
National Center for Environmental Economics
Office of Policy
U.S. Environmental Protection Agency

-------
Acknowledgements
In 2010, then Deputy Administrator Bob Perciasepe inquired about research on retrospective cost analysis,
particularly of past EPA regulations. An investigation of the literature revealed that the collection of
retrospective analyses of EPA regulations is thin and no generalized conclusions could be drawn. Bob
Perciasepe asked the National Center for Environmental Economics (NCEE) to design and launch a
retrospective cost analysis with the goal of improving EPA's cost assessments. This report would not have
been completed without the support of Bob Perciasepe and the authors are grateful for his participation in
the various stages of the project.
Many economists from NCEE were involved; the authors of individual case studies are noted in the report. Al
McGartland provided helpful direction and Alex Barron's review and advice was particularly valuable. Many
EPA economists and engineers also reviewed portions of this report.
Eric Burneson of EPA/OGWDW; Ken Davidson and Don Kopinski of EPA/OTAQ; Tim Kiely of EPA/OPP, David
Donaldson, and Jeremy Arling of EPA/OAP ; Larry Sorrels of EPA/OAQPS and Julie Hewitt of EPA/OST; and Jim
DeMocker of EPA/OAR served important roles both reviewing and coordinating review contributions from
staff in their respective offices.
A number of activities in the case studies were undertaken by contractors, including project managers Anna
Belova and Albert Acquaye of Abt Associates; Rebecca Nicholson, Tom Holloway, and Corey Gooden of RTI
International; and Christopher Weaver of EF&EE, Inc.
Science Advisory Board - Environmental Economics Advisory Committee (SAB-EEAC) review of the draft
report was supervised by Holly Stallworth, Designated Federal Officer. The SAB-EEAC was chaired by Madhu
Khanna of the University of Illinois at Urbana-Champaign throughout the review of the draft
report. Members who participated in the review of the draft report are Nicholas Flores of the University of
Colorado, Boulder, Wayne Gray of Clark University, Karen Palmer of Resources for the Future, George
Parsons of the University of Delaware, James Shortle of Pennsylvania State University, Laura Taylor of North
Carolina State University, Peter Wilcoxen of Syracuse University, JunJie Wu of Oregon State University, Jinhua
Zhao of Michigan State University, and David Zilberman of the University of California - Berkeley.

-------
Table of Contents
PREFACE	V
EXECUTIVE SUMMARY	VII
CHAPTER 1: BACKGROUND	1
1.1.	LITERATURE REVIEW OF PREVIOUS RETROSPECTIVE COST STUDIES	2
1.1.1.	Overview of the Survey Articles	3
1.1.2.	Why Ex Ante and Ex Post Cost Estimates May Differ	8
1.1.2.1.	Strategic factors affecting ex ante costs	8
1.1.2.2.	Technological innovation and unforeseen compliance options	10
1.1.2.3.	Unanticipated exogenous changes	11
1.2.	METHODOLOGY	12
1.2.1 Conceptual Framework for Ex Post Cost Assessment	12
1.2.2.	Selection of Rules	15
1.2.3.	Strategies for Ex Post Data Collection	19
CHAPTER 1 REFERENCES	21
CHAPTER 2: CLUSTER RULE AND MACT II RULE	26
2.1.	IMPETUS AND TIMELINE FOR REGULATORY ACTION	27
2.2.	EX ANTE COST ESTIMATES	29
2.3.	INFORMATION AVAILABLE TO CONDUCT EX POST EVALUATION	32
2.4.	EX POST ASSESSMENT OF COMPLIANCE COSTS	33
2.4.1.	Regulated Universe	33
2.4.2.	Baseline	34
2.4.3.	Methods of Compliance	37
2.4.3.1.	Compliance Costs for MACT I and BAT/PSES	41
2.4.3.2.	Compliance Costs for MACT II	43
2.5.	CONCLUSIONS	52
CHAPTER 2 REFERENCES	54
APPENDIX 2.1: CLOSED MILLS AND MILLS NO LONGER CLASSIFIED AS BPK OR PS	58

-------
CHAPTER 3: RETROSPECTIVE EVALUATION OF THE COSTS ASSOCIATED
WITH METHYL BROMIDE CRITICAL USE EXEMPTIONS FOR
OPEN FIELD STRAWBERRIES IN CALIFORNIA	67
3.1.	IMPETUS AND TIMELINE FOR THE REGULATORY ACTION	69
3.1.1. Overall Trends in U.S. Critical Use Exemptions	70
3.2.	EPA EX ANTE COST ESTIMATES FOR OPEN FIELD STRAWBERRIES IN CALIFORNIA	72
3.2.1.	Main Drivers of Ex Ante Cost Estimates	74
3.2.2.	Main Sources of Uncertainty in Ex Ante Cost Estimates	77
3.2.3.	Exogenous Factors that May Affect Estimated Ex Ante Costs	78
3.3.	LITERATURE AND DATA AVAILABLE TO CONDUCT EX POST EVALUATION	80
3.3.1.	Ex Post Literature	80
3.3.2.	Data for Evaluating Costs Ex Post	81
3.4.	ASSESSING COSTS OF MBR CRITICAL USE EXEMPTIONS RETROSPECTIVELY	83
3.4.1.	Regulated Universe	84
3.4.2.	Baseline Information	84
3.4.3.	Methods of Compliance	85
3.4.4.	Compliance Costs	90
3.4.4.1.	Gross Revenues	91
3.4.4.2.	Operating Costs	94
3.4.4.3.	Indirect Costs	97
3.4.4.4.	Opportunity Costs	99
3.5.	OVERALL IMPLICATIONS AND STUDY LIMITATIONS	101
CHAPTER 3 REFERENCES	104
APPENDIX 3.1: REVIEW OF THE EX ANTE LITERATURE	108
APPENDIX 3.2: THE ROLE OF THE STOCKPILE	114
CHAPTER 4: NATIONAL PRIMARY DRINKING WATER REGULATION FOR -
ARSENIC	115
4.1.	IMPETUS AND TIMELINE FOR THE REGULATORY ACTION	115
4.2.	EPA EX ANTE COST ESTIMATES	116
4.2.1. Main Components of Ex Ante Compliance Costs	116
4.2.1.1.	Identification of Best Available Treatment Technologies	116
4.2.1.2.	Community Water Systems (systems serving 1,000,000 people or fewer)	118
4.2.1.3.	Non-Transient Non-Community Water Systems	119
ii -

-------
4.2.1.4. Community Water Systems (systems serving populations of more than 1,000,000)
119 -
4.2.2. Main Sources of Uncertainty in Ex Ante Cost Estimates	120
4.3.	INFORMATION AVAILABLE TO CONDUCT EX POST EVALUATION	120
4.3.1.	Ex Post Literature	120
4.3.2.	Data for Evaluating Costs Ex Post	122
4.3.2.1.	Publicly Available Data	122
4.3.2.2.	ORD Demonstration Projects	123
4.3.2.3.	Compliance Assistance Engineering Firms	124
4.3.2.4.	Independent Associations	125
4.3.2.5.	State Agencies	126
4.3.2.6.	Summary of Potential Sources of Cost information	130
4.4.	EX POST ASSESSMENT OF COMPLIANCE COST	130
4.4.1.	Regulated Universe	130
4.4.2.	Baseline Information	131
4.4.3.	Methods of Compliance	131
4.4.4.	Compliance Costs	133
4.4.4.1.	Ex Post Compliance Costs	134
4.4.4.2.	Total Reported Capital and O&M Costs	134
4.4.5.	Comparison of Technology Costs	140
4.5.	OVERALL IMPLICATIONS AND STUDY LIMITATIONS	141
CHAPTER 4 REFERENCES:	143
APPENDIX 4.1: PUBLICLY AVAILABLE DATA RELATED TO ARSENIC RULE	145
Potential Sources of Arsenic Occurrence Data	145
Potential Sources of Compliance Cost Data	149
APPENDIX 4.2: EPA COST CURVES FOR COMPLIANCE WITH MCLS FOR ARSENIC IN DRINKING WATER -
152
APPENDIX 4.3: DOCUMENT SENT TO COMPLIANCE ASSISTANCE ENGINEERING FIRMS	158
CHAPTER 5: EPA'S 1998 LOCOMOTIVE EMISSION STANDARDS	180
5.1.	IMPETUS AND TIMELINE FOR REGULATORY ACTION	180
5.2.	EPA EX ANTE COST ESTIMATES	181
5.2.1.	Main Components of the Ex Ante Cost Analysis	182
5.2.2.	Treatment of Uncertainty and Baseline	184
5.3.	INFORMATION AVAILABLE TO CONDUCT EX POST EVALUATION	185
iii -

-------
5.4.	EX POST ASSESSMENT OF COMPLIANCE COST	186
5.4.1.	Regulated Universe	186 -
5.4.1.1.	Locomotive Model Types	186-
5.4.1.2.	Number of Locomotives Affected by the Regulation	187-
5.4.2.	Methods of Compliance	190-
5.4.2.1.	Emission Control Technologies for Tier 0 Locomotives.	192-
5.4.2.2.	Emission Control Technologies for Tier 1 Locomotives.	195-
5.4.2.3.	Emission Control Technologies for Tier 2 Locomotives.	195-
5.4.3.	Per Locomotive Compliance Cost	196 -
5.4.3.1.	Initial Compliance Cost	196-
5.4.4.2.	Remanufacture Costs	199 ¦
5.4.4.3.	Fuel Costs	200 ¦
5.5.	OVERALL IMPLICATIONS AND STUDY LIMITATIONS	204
CHAPTER 5 REFERENCES	208
APPENDIX 5.1: EPA'S EMISSION STANDARDS FOR LOCOMOTIVES AND LOCOMOTIVE ENGINES	210
CHAPTER 6: LESSONS LEARNED AND NEXT STEPS	227 -
6.1.	LESSONS LEARNED	227 -
6.2.	NEXT STEPS	227 -
iv ¦

-------
Preface
The Office of Management and Budget (OMB) has stressed the need for regulatory agencies to conduct
retrospective analyses of their rules. In their draft 2014 Report to Congress on the Benefits and costs of
federal regulation, OMB states: "Retrospective analysis, required by Executive Order 13563 and
institutionalized by Executive Order 13610, can be an important way of increasing accuracy."1 The Executive
Orders instruct: "it is particularly important for agencies to conduct retrospective analyses of existing rules to
examine whether they remain justified and whether they should be modified or streamlined in light of
changed circumstances, including the rise of new technologies."2
Benefit-cost analyses (BCA) of environmental regulations most often involves integrating science from a wide
array of disciplines. Engineering, fate and transport modeling, ecology, toxicology, epidemiology, exposure
modeling, behavioral modeling and economic valuation methods are all needed to assess the benefits and
costs of environmental regulations. All of these sciences have advanced dramatically in the last decade.
With these advances, it is prudent for agencies to periodically assess whether regulations are still appropriate
in their current form.
While new science and the need to quantify more previously unquantified benefits has driven benefits
analysis, comparatively less work has been done examining how well EPA estimates the costs (or benefits) of
regulation. The ex post cost studies that are available in the literature are often based on limited data and
overlap in coverage - many of the same regulations appear in multiple publications. And while the literature
posits a number of hypotheses for why one might expect ex ante and ex post cost estimates to differ, EPA's
current judgment is that ex post analyses are too few in number to draw conclusions regarding general
tendencies to under- or over-estimate costs in ex ante evaluations.
The National Center for Environmental Economics (NCEE) has launched an effort to evaluate the feasibility of
augmenting the existing literature with additional ex post evaluations of costs. Researchers examining the
relationship between ex ante and ex post cost generally used a case study approach. We do too. However,
we develop a common conceptual framework for our ex post cost assessments. In this report we present
the ex post assessments of five EPA regulations: 2001/2004 National Emission Standards for Hazardous Air
Pollutants and Effluent Limitations Guidelines, Pretreatment Standards, and New Source Performance
Standards on the Pulp and Paper industry; Critical Use Exemptions for Use of Methyl Bromide for Growing
Open Field Fresh Strawberries in California for the 2006-2010 seasons; the 2001 National Primary Drinking
Water Regulations for Arsenic; and the 1998 Locomotive Emission Standards. These case studies were
developed with extensive support and input from staff in EPA program offices, economists in EPA's NCEE, and
1	U.S. Office of Management and Budget (OMB.) 2014. Draft Report to Congress on the Benefits and Costs of
Federal Regulations and Unfunded Mandates on State, Local, and Tribal Entities. Page 3.
2	E.O. 13610, "Identifying and Reducing Regulatory Burdens." FR 77(93), May 14, 2012. (available at:
http://www.whitehouse.gov/sites/default/files/docs/microsites/omb/eo 13610 identifying and reducing reeula
torv burdens.pdf)
v -

-------
members of the Science Advisory Board - Environmental Economics Advisory Committee. They represent a
first step by EPA in generating a larger body of evidence on key drivers of compliance costs.
The purpose of the case studies (and the case studies done by other researchers) is NOT to estimate ex post
costs reliably. Rather, it is to see if sufficient information can be gathered to make a "weight of evidence"
determination about whether ex ante cost estimates tend to be higher or lower than ex post cost estimates.
We cannot emphasize this enough. The case studies in this report do not aim to estimate ex post costs of
these EPA actions. Rather, we investigate the key drivers of compliance costs across regulations to see if
informed judgments can be made on the general accuracy of ex ante estimates and what underlying factors
contribute to differences (or similarities) between ex ante and ex post estimates.
If the case study approach is successful, there is much that can be learned from this effort. A careful
assessment of ex post cost drivers could help identify systematic differences between ex post and ex ante
compliance cost estimation and, ultimately, allow for improvements in the way in which ex ante analyses are
done. For instance, if unanticipated changes in market conditions, energy prices, or available technologies
regularly result in an over or underestimate of costs, EPA can invest in improving methods such as expanded
uncertainty analysis that better capture these effects on costs ex ante. It is also possible that industry
overstates the expected costs of compliance (EPA often has to rely on industry to supply it with otherwise
unavailable information on expected compliance costs). Even if such specific differences between ex ante
and ex post cost studies cannot be identified, a sizable set of retrospective analyses can offer broader
insights, such as whether certain cost categories are particularly uncertain or how to better incorporate
behavioral responses to regulation into ex ante analyses.

-------
Executive Summary
While advancements in research on benefit and cost estimation methods and models is continually applied to
EPA analyses of new regulations, there also is significant potential to learn from analysis of the benefits and
costs of past regulations. New science and the desire to capture previously un-quantified benefits has driven
periodic ex post reviews of how benefits analyses are conducted, but comparatively less work has been done
retrospectively examining how well EPA estimates the costs of regulation. The literature posits a number of
hypotheses for why one might expect ex ante and ex post cost estimates to differ, yet ex post cost case
studies are too few in number and narrow in scope to lend strong support for one hypothesis over another.
Existing case studies are often based on limited data and overlap in coverage, with many of the same
regulations appearing in multiple publications.
For these reasons, EPA launched an effort to augment the existing literature with additional ex post
evaluations of costs. Similar to previous studies, we examine the relationship between ex ante and ex post
cost using a case study approach. However, unlike prior literature, we develop a common conceptual
framework for our ex post cost assessments to more consistently investigate the key drivers of compliance
costs across regulations to see if informed judgments can be made on the general accuracy of ex ante
estimates and what underlying factors contribute to differences (or similarities) between ex ante and ex post
estimates.
A careful assessment of ex post cost drivers may help identify systematic differences between ex post and ex
ante compliance cost estimation and, ultimately, allow for improvements in the way in which ex ante
analyses are done. For instance, if unanticipated changes in market conditions, energy prices, or available
technologies regularly result in an over or underestimate of costs, EPA can invest in improving methods such
as expanded uncertainty analysis that better capture these effects on costs ex ante.
After outlining the conceptual framework, this report applies it to four case studies that retrospectively
examine the compliance costs of five EPA regulations. These five rules were not chosen randomly, but rather
are chosen as pilot case studies to help us understand which methodologies are appropriate to measure ex
post compliance costs for a range of rules. Given this purpose, the five rules cover various media, source
categories, and types of regulations (e.g., performance standard versus prescriptive regulation). The five
regulations evaluated in this report are: the 2001/2004 National Emission Standards for Hazardous Air
Pollutants and Effluent Limitations Guidelines, Pretreatment Standards and New Source Performance
Standards on the Pulp and Paper industry; Critical Use Exemptions for Use of Methyl Bromide for Growing
Open Field Fresh Strawberries in California for the 2004-2008 seasons; the 2001 National Primary Drinking
Water Regulations for Arsenic; and the 1998 Locomotive Emission Standards.
For each case study, we assessed whether it would be possible to collect sufficient ex post compliance cost
information using only publicly-accessible data sources. In general, we found that while data for some
necessary components are readily available, the cost information is generally lacking. The critical use
vii -

-------
exemption for methyl bromide use for California strawberries fared the best of the five with regard to the
availability of cost information, and was designated as the case study that would be based on publicly-
available data alone. For the remaining rules, we also consulted industry compliance experts to gather
information on compliance strategies and ex post cost data.
While several of the case studies are suggestive of overestimation of costs ex ante, we do not consider the
current evidence to be conclusive. First, they only represent a small subset of regulatory and other policy
actions taken by EPA. Second, conducting ex post analysis has proven more challenging than anticipated.
With regard to data, these challenges have included the inability to identify qualified industry experts that did
not also work on the rule as well as limited access to industry data. Analytic challenges have included how to
evaluate a highly heterogeneous industry with a limited set of information, how to form a reasonable
counterfactual, and how to disentangle the costs of compliance from other factors, to name a few.

-------

The Environmental Protection Agency (EPA) conducts benefit-cost analyses (BCA) for many of the new rules it
proposes.3 While there are a number of factors that can influence regulatory decisions (including
environmental justice, statutory direction, enforceability of specific options, and uncertainty and precaution),
the benefits and costs of regulatory options are important criteria. Furthermore, BCA informs stakeholders,
policy makers, Congress, and the public of how society will be better off from an environmental regulation
and how much it will cost. Given the prominent role of BCA, EPA strives to use the best available science,
data and methods when conducting its analyses. While research on benefit and cost estimation methods and
models is continually applied to EPA analyses of new regulations, there is significant potential to learn from
additional analysis of the benefits and costs of past regulations. New science and the desire to capture
previously un-quantified benefits has driven innovations in benefits analyses, but comparatively less work has
been exploring how to improve cost estimation techniques.
A careful assessment of ex post cost drivers could help identify systematic differences between ex post and
ex ante compliance cost estimation and, ultimately, allow for improvements in the way in which ex ante
analyses are done. For instance, if unanticipated changes in market conditions, energy prices, or available
technologies regularly result in an over or underestimate of costs, EPA can invest in improving methods such
as expanded uncertainty analysis that better capture these effects on costs ex ante. It is also possible that
industry overstates the expected costs of compliance (EPA often has to rely on industry to supply it with
otherwise unavailable information on expected compliance costs).
Even if such specific differences between ex ante and ex post cost studies cannot be identified, a sizable set
of retrospective analyses can offer broader insights, such as whether certain cost categories are particularly
uncertain or potential ways to better incorporate behavioral responses to regulation into ex ante analyses.
Using a common conceptual framework described in more detail below, we examine the key components of
the estimated cost (e.g., the number of regulated facilities, baseline definition, compliance strategies
employed, capital costs) using a case study approach. The purpose of the case studies is not to estimate ex
post costs at the same level of rigor employed in the original economic analyses, but rather to assess
available information on the key drivers and make a "weight of evidence" determination about whether and
how ex ante cost estimates differ from ex post cost estimates. When a substantial difference exists, we seek
to understand the reasons for the discrepancy and to determine if there are any systematic reasons for the
differences. Ultimately, the goal is to identify areas in which to improve EPA's ex ante cost estimation.
3 Since 1981, EPA has been required to conduct benefit cost analyses of all economically significant regulations
(i.e., those that have an annual effect on the economy of$100 million or more, or meet other criteria).
1 -

-------
While additional retrospective case studies are underway and still more are planned, this report summarizes
the results of the first four case studies that examine five EPA regulations: the 2001/2004 National Emission
Standards for Hazardous Air Pollutants and Effluent Limitations Guidelines, Pretreatment Standards, and New
Source Performance Standards on the Pulp and Paper industry; Critical Use Exemptions for Use of Methyl
Bromide for Growing Open Field Fresh Strawberries in California for the 2004-2008 seasons; the 2001
National Primary Drinking Water Regulations for Arsenic; and the 1998 Locomotive Emission Standards.
The remaining sections are organized as follows. Section I reviews the literature on the accuracy of a variety
of ex ante regulatory cost estimates and discusses competing hypotheses about what causes a divergence
between ex ante and ex post compliance costs. Section II describes our methodology: the conceptual
framework we apply to structure each of the case studies, the rule selection process, and the ex post cost
estimation strategies as well as potential sources of ex post cost data. Results from four case studies are
presented in Sections III through VI. Section VII presents concludes.
1.1. Literature Review of Previous Retrospective Cost
Studies
A number of researchers have reviewed ex ante estimates of the costs of environmental and other forms of
regulation in light of ex post estimates of such costs. We focus here largely on survey articles that review a
number of individual studies, each of which have attempted to compare ex ante to ex post cost estimates,
and then try to draw out more general lessons concerning the accuracy of the ex ante estimates.4 As these
survey articles themselves may incorporate results from a dozen or more individual studies (although there is
some overlap between survey articles in the original studies they include), we should emphasize that the
overall literature is significantly larger than one might infer simply from counting the number of papers we
cite. We begin with a brief overview of the types of regulations that have been examined retrospectively as
well as the general findings with regard to the accuracy of ex ante cost estimates. We then discuss the main
reasons why ex ante costs may be under or overestimated based on this literature.
4 Simpson (2011) assesses the published literature on comparing ex ante and ex post costs and discusses the
different treatment of costs across the studies. He also considers what we can infer from the findings of these
studies, noting in particular that not all of the studies actually conduct a numeric comparison of the ex ante to ex
post costs. Regrettably, for our purposes, Simpson finds only a relative handful of analyses considering the total
(as opposed to unit) costs of regulations in the U.S. Thus, while he attempts a statistical analysis to test if ex ante
estimates are biased, the limitations of his sample severely compromise the power of his test.
2 -

-------
1,1,1, Overview	ey Articles
Table 1.1 summarizes many of the studies described in this section with regard to the types of regulations
reviewed and general findings of accuracy. Most of the underlying studies focus on U.S. regulations, with
EPA regulations featured prominently. We omit retrospective studies of the Title IV S02 cap-and trade

-------
Table 1.1: Summary of Accuracy of Ex Ante Costs from Existing Studies5
Authors (Date of
Publication)
Description
Accuracy of ex ante cost estimation
Putnam, Hayes, and
Bartlett (1980)6
Compare US EPA and industry ex ante
estimates of capital expenditures to
actual expenditures for five EPA
regulations promulgated from 1974-
1977
In four of five cases, industry over-
estimated costs. In three of five cases,
EPA over-estimated costs
Jantzen(1989)and
RIVM (2001) as
reported in Oosterhuis
et al. (2006)
Evaluate costs of compliance for 8
regulations associated with the first
Dutch National Environmental Plan of
1988
Costs were overestimated ex ante for 5
regulations, but only one ex ante
estimate was as much as 2x the ex post
estimate; in aggregate ex ante was only
13% higher than ex post
Office of Technology
Assessment (1995)7
Eight OSHA regulations promulgated
from 1974-1989 in the chemical,
service, and manufacturing industries
OSHA overestimated costs ex ante in
every case. In two cases, costs may
have been negative
Hodges(1997)
Compare industry ex ante estimates
to ex post cost estimates for 12 US
environmental and workplace safety
regulations from the 1970s to 1990s
In every case evaluated, costs were
overestimated ex ante; in 11 of 12
cases, ex ante estimates were more
than double ex post costs
Harrington,
Morgenstern, and
Nelson (2000)
28 US regulations promulgated by
EPA, OSHA, and other regional and
international regulators (13 were EPA
regulations)
Total costs overestimated for 14,
underestimated for 3, and reasonably
accurate (within +/- 25%) for 11
regulations; unit costs overestimated
as often as underestimated. (For EPA
regulations, 7 overestimated costs ex
5	There are several studies of public programs and procurement that raise issues similar to those we consider. -
Boardman, et al. (1994) study the accuracy of ex ante cost estimates for a road-building project and find that costs -
were underestimated. Conventional wisdom has it that the costs of regulation tend to be overestimated. The -
direction of the bias is readily explained in both instances. In the road-building example private firms profit from -
public construction activities, and so would want to make such activities seem more attractive by understating -
costs. In the case of regulation, private firms bear the cost, and would thus have an incentive to exaggerate costs -
of compliance to make the regulation seem less desirable (in both the public works and the regulation cases public -
officials often must rely on private entities for cost estimates). Other studies of the accuracy of estimates of the -
costs and benefits of public programs include Dayton (1998) on HIV/AIDS intervention, Rideout and Omi (1995) on -
fuel reduction measures in public forests, Lindner and Jarrett (1978) on publicly sponsored research, LaFrance and -
Gorter (1985) on dairy price supports, and, very generally, Sappington and Stiglitz (1987) on privatization. -
6	Putnam, Hayes, and Bartlett (1980) also examined a sixth case of the effect of environmental regulations on new -
car prices but results were somewhat more ambiguous. -
7	In two cases, OTA suggests costs may actually have been negative - i.e., in finding ways to reduce risks, -
producers may have identified processes that operate more efficiently. However, while environmental regulations -
may induce some firms to experiment with pollution reduction technologies they would not otherwise have tried, -
and some experimenting firms may be surprised to find in some instances that costs actually decline as a result, -
this does not mean that costs would, generally speaking, be expected to decline in response to tougher regulation. -
There may well be offsetting instances in which other firms try technologies that reduce pollution but, as expected, -
increase costs. Moreover, there are costs of experimentation, which are not always reported accurately. -
4-

-------
Authors (Date of
Publication)
Description
Accuracy of ex ante cost estimation


ante, 2 underestimated, and 3
reasonably accurate.)
Anderson and
Sherwood (2002)
11 vehicle emission and 6 fuel-quality
US EPA regulations
In most cases ex ante estimates of
induced price increases overestimated
actual changes; EPA estimates tended
to be more accurate than industry
Thompson et al. (2002)
US consumer safety regulation
requiring air bags in automobiles
Cost estimated were reasonably
accurate: ex ante exceeded ex post
cost estimates by less than 5%
Grosse et al. (2005)
Evaluated the accuracy of 3 different
ex ante studies of a US FDA regulation
to fortify cereal grains with folic acid
Ex ante estimates overestimated costs
by 3.5 to 9x actual costs
OMB (2005)8
47 US regulations initiated between
1976 and 1995 (18 EPA regulations)
Of 40 regulations for which data were
available, 16 overestimated costs ex
ante, 12 underestimated them, and 12
were reasonably accurate.
McLeod et al. (2006)
Eight UK regulations
Five overestimated costs ex ante, 2
underestimated, and one was
reasonably accurate (within +/- 25%)
Oosterhuis et al
(2006)9
Five EU environmental regulations
Costs were overestimated ex ante by a
factor of two or more in 4 cases, and
reasonably accurate in 1 case
Dale et al. (2009)
Used a hedonic regression approach
to evaluate ex ante costs of US DOE
energy efficiency regulations on
consumer appliances
Ex ante estimates over estimated costs
National Research
Council (NRC) (2012)
Evaluated EPA's estimates of costs for
a proposed EPA water regulation to
establish nutrient criteria
Inconclusive since ex post data were
not yet available
8	The OMB study was not completely independent of earlier work. For instance, results for nine of the studies in
its sample were taken from Harrington, et al. 2000.
9	Oosterhuis et al.(2006) actually consider six environmental directives, addressing large combustion plants,
integrated pollution prevention and control, ozone control, ozone depleting substances packaging, and nitrates,
but are unable to develop ex ante compliance cost estimation numbers for the packaging directive.
5

-------
program from the table; the relatively large literature on this topic is instead summarized in Text Box 1. In a
few cases, ex ante estimates of the cost of regulation are available from both the regulator and industry
offering another point of comparison to ex post estimates. In general, ex ante cost estimates are more often
found to overestimate than underestimate realized costs, and in cases where industry estimates are available
it appears that the regulator is often more accurate in its assessments of costs ex ante.
While Table 1.1 may give the reader the impression that much work has already been done to evaluate
the costs of regulation retrospectively, there are two reasons why this evidence is less compelling than it
may appear. First, given both the paucity of ex post data on compliance costs and the large variation in
methodology and scope of analysis, many of the predictions of over or under-estimation likely have
large error bounds. Studies differ in the approaches they take to the estimation of ex ante costs, and
the elements they include in such estimates. Some have considered only capital costs, others capital
and operating costs. To the extent that the times at which costs are incurred differ across studies,
differences in the discount rates their authors assume may have affected cost estimates. Moreover,
analysts also have to apply their best judgment to distinguish costs that might be incurred in the course
of business as usual from those that would need to be incurred to meet regulatory requirements. If, for
example, the general trend in an industry is toward the availability of cleaner production technology
over time, the cost associated with a regulation might best be measured as the incremental cost of
accelerating capital replacement, rather than the total cost of a new capital investment.
Second, the collection of regulations for which any comparison of ex ante to ex post cost estimates has
been performed is small and is unlikely to form a representative sample of the universe of
environmental rules that have been promulgated. Many of the survey articles summarize the same sets
of underlying studies, which means that there is substantial overlap. Within the regulations that have
been analyzed, there is reason to believe that those for which unforeseen technological breakthroughs
occurred might be overrepresented because they are often the most celebrated and visible regulations.
For example, it came as a considerable surprise to many industry compliance experts that coal-fired
power plants were able to substitute between coal types as easily as they were to meet emission limits
under the S02 cap-and trade program.10 It was, then, natural to further investigate the divergence
between ex ante and ex post estimates of the cost of regulation in this case (see Text Box 1). It is also
likely that economists prefer to study regulations where regulated parties were given flexible options for
compliance. Harrington et al. (2000) suggest that this is because more data are available for rules that
establish markets, as prices are easily observed, and because "economists ... have a proprietary
interest in the performance of economic incentives."
10 To give one example Joskow (1988) argued that electric utilities entered into long-term contracts with coal
providers because the need for a specific grade of coal made for an obligate relationship between a mine and a
plant. As it turned out, this relationship was not nearly as restrictive as had been thought in many cases.
6-

-------
Text Box 1.1 Title IV and the 1990 Clean Air Act Amendments
Title IV of the 1990 Clean Air Act Amendments (CAAA) called for large reductions in sulfur dioxide (SO2)
emissions by coal-fired electrical generating units (EGUs). The aim of Title IV was to cut aggregate annual SO2
emission levels to approximately 9 million tons by 2010, roughly 50% of the recorded 1980 emission levels
from EGUs. To help EGUs make these large SO2 reductions, Title IV created a cap-and-trade program that
established a cap on total SO2 emissions, allocated allowances to EGUs equal to that cap, and permitted
EGUs to freely trade these allowances or to bank them for future use.
Ex post analyses of the trading program tend to be some of the most analytically rigorous: boiler-level data
on emissions, the price of permits and methods of compliance utilized allow for the use of sophisticated
econometric evaluation techniques. Researchers studying the compliance costs for the first phase (1995-
1999) of Title IV, which targeted the dirtiest 110 power plants, have shown that actual costs decreased
substantially over time, particularly once the program began and data became available that documented
how EGUs were responding to the regulation. Table 1 below provides a comparison of some of the Title IV's
cost estimates. Rows that report ex ante estimates are shaded gray while rows reporting ex post costs
remain unshaded. More recently, Chan et al. (2012) provide a summary of the vast ex-post literature
focused on the trading program under Title IV.
Table 1 - Estimates of Compliance Costs for the SO2 Program*
Author
Annual Costs

Marginal Costs
Average costs per

(Billions)

per ton S02
ton of S02
Ex ante Studies
ICF (1995)
$2.3
$532
$252
White et al. (1995)
1.4-2.9
543
286-334
GAO (1994)
2.2-3.3
n/a
230-374
Van Horn Consulting et al. (1993)
2.4-3.3
520
314-405
ICF (1990)
2.3-5.9
579-760
348-499
Ex post Studies
Carlson et al. (2000)
$1.1
$291
$174
Ellerman et al. (2000)
1.4
350
137
Burtraw et al. (1998)
0.9
n/a
239
Goulder et al. (1997)
1.09
n/a
n/a
White (1997)
n/a
436
n/a
* Based on Table 2-1, Burtraw and Palmer (2004). n/a - not reported
Title IV proved less costly than originally estimated due to a number of factors, including unanticipated
changes in the market for coal due to railroad deregulation, technological improvements and input price
changes. The ability of facilities to "bank" SO2 allowances allowed even greater flexibility in meeting the SO2
cap, and also helped to contribute to additional reductions in actual compliance costs. Ex post cost estimates
by Carlson et al. (2000) and Ellerman et al. (2000) take into consideration the discounted savings from
banking. According to Ellerman et al., savings from banking are a relatively minor source of overall savings of
Title IV, but are important in developing a picture of the program's overall effectiveness. Absent banking,
some EGUs would have had to make larger pollution control investments and/or accelerated their
investments in emission controls to be certain of meeting their emission targets. Because of banking, firms
were able to "avoid the much larger losses associated with meeting fixed targets in an uncertain world"
(Burtraw and Palmer 2004).
7 -

-------
1.1.2. Why Ex Ante and Ex Post Cost Estimates May Differ
There are many reasons that potentially explain why ex post and ex ante cost estimates might diverge. The
degree to which the studies included in Table 1.2 reflect on these reasons vary. They also vary with regard to
the level of insight they provide on why ex ante and ex post cost estimates differ. With these caveats in
mind, we briefly summarize potential reasons ex ante and ex post costs may differ. We are particularly
concerned with factors that might lead to ex ante cost estimates being systematically too high (or too low),
as opposed to those that would result in their being less accurate while still, on average, correct.
1.1.2.1. Strategic factors affecting ex ante costs
Much of the cost information used in regulatory analyses comes directly from industry (Hodges 1997;
Harrington et al. 2000; Bailey et al. 2002). This is unavoidable since industry generally has the best
information concerning expected compliance costs. It is not uncommon for EPA to solicit industry compliance
cost data through surveys of individual firms or interactions with their trade organizations (Harrington et al.
2000). However, reliance on industry provided information leads to the possibility that compliance costs are
either strategically over or understated by industry or by regulators when using this information to generate
estimates.
A number of studies argue that regulated entities may overstate their costs of compliance (Hodges 1997,
Bailey et al. 2002, MacLeod et al. 2006). Firms facing regulation might misrepresent their costs in a strategic
attempt to influence regulator's actions and thwart what they see as onerous regulations by providing a
signal that costs are prohibitive (see Bailey et al. (2002), and more generally, Sappington and Stiglitz
(1987)).11 Ex ante cost estimates are also typically based on the application of existing technologies, rather
than relatively untested innovative approaches to inputs, abatement, and processes. While the best source
for information about existing technologies is the people who use them, industry is likely to describe one or
more plausible ways of complying rather than evaluate all alternatives before identifying those that are likely
to minimize compliance costs. Harrington et al. (2000) also suggest that firms are more likely to describe
"off-the-shelf" technologies in their cost estimates rather than examining opportunities for innovation.
Not all firms within an industry, or across different industries, have the same motivations, however. An
alternate explanation may be that industry is providing conservative estimates given the numerous
uncertainties associated with estimating compliance costs of regulations. For example, Oosterhuis, et al.
(2006) note that, while "[t]here seems to be little evidence of industry knowingly providing biased cost
estimates ... in the face of uncertain future technological development, the affected industry will tend to
come up with relatively high cost figures."
Industry interventions may also interact with process and timing issues to affect the accuracy of initial cost
estimates. Proposed rules published in the Federal Register often receive a disproportionate number of
comments, particularly of a technical nature, from the regulated industries and groups that support them
11 Bailey, et al., (2002) make an interesting related observation: in addition to overstating costs, industry groups
may also question the benefits of a proposed regulation.
8 -

-------
highlighting specific issues. Environmental and public interest groups, in contrast, may submit fewer
comments, which tend to be less specific—and therefore may be less useful to regulators for revising cost
estimates. EPA's internal Action Development Process and the Administrative Procedures Act, which covers
all Executive Branch agencies, require EPA to consider and respond to comments received in the open
comment period and explain why cost estimates differ.
As a result of this asymmetric distribution of comments, final rules often prove less stringent than the
versions initially proposed (Magat et al. 1986). Morgenstern and Landy (1997) find that industry
interventions led to less stringent final standards in all twelve of the rules they considered. To the extent that
Agency analyses were based on a version of a rule that was later made less stringent in response to industry
comment, cost estimates may be higher than realized costs.
Environmental regulation might impose a restraint on the competition that can arise when some firms
cannot operate as cleanly as others. Salop and Scheffman's (1983) depiction of "raising rivals' costs" could
provide a rationale for why some firms would prefer regulation that would increase their own level of
regulation because it would hurt others more.12 Bailey et al. (2002) raise an interesting example of divergent
incentives between the petroleum and the automobile industries. The former may oppose tighter regulation
on fuels, whereas the latter might see the reformulation of fuel as a motivation for consumers to purchase
new cars which perform better on the reformulated fuel.
A related concern may be that a regulatory agency may be less rigorous in estimating the costs of rules that
appear likely to pass a benefit-cost test. Under such circumstances there may be reduced incentive for
regulators to refine their cost estimates or to investigate alternative pathways to compliance, such as process
changes or alternative technologies. Further, regulators might conservatively overstate costs in cases when
affordability criteria must be met on the grounds that if a regulation is found to be affordable when stated
costs are higher than expected, the regulation will be affordable using more refined estimates of costs as
well. Harrington et al. (2000) noted that EPA provided upper bound cost estimates in their effluent
guidelines program. It might also be counterproductive for regulators to strive to establish a more refined
precise cost estimate, as the regulated industry might then feel compelled to protest, perhaps on the
grounds that they do not want to see such cost estimates applied in subsequent rule-making.
In addition to the technologies regulators assume when predicting the costs of regulation, they also typically
make assumptions concerning compliance and coverage. While it is common for regulators to assume full
compliance with a proposed rule, actual compliance may be less-than-perfect. Although it is now dated,
Putnam et al. (1980) found compliance rates of only 54 percent in the iron and steel industry and 83 percent
in petroleum. MacLeod et al. (2006) cite imperfect compliance as one reason for finding costs overestimated
in ex ante studies.
12 Maloney and McCormick (1982) argue that tighter OSHA regulation of cotton dust and EPA regulations to
prevent significant deterioration near existing factories both had the effect of restricting new competition and
enhancing the profits of incumbent firms that were well suited to avoid the impact of the regulations or exempted
from meeting it. See Adler (1996) and Bailey et al. (2002) for other examples.
9 -

-------
A related consideration concerns the regulators' assumptions concerning baselines. EPA's Guidelines for
Preparing Economic Analyses (2010) instruct analysts to compare the benefits and costs of regulations
relative to a "baseline," which is defined as "a reference point that reflects the world without the regulation".
Some authors have suggested, however, that regulatory agencies have their own strategic objectives which
could, in theory, lead to incentives both to overstate the benefits and understate the costs of regulation
(James 1998, Harrington et al. 2000, OMB 1998, Hahn 1996, MacLeod et al. 2006). Harrington et al. (2000)
find that agencies may overstate the baselines relative to which subsequent costs under regulation are
compared, but that the data do not support a purposeful underestimation of costs per se. Moreover, there
may be limits to the ability of agencies to pursue cost underestimation. Industry groups with relatively
concentrated membership and relatively closely aligned interests are likely to challenge unrealistically low
estimates.
1,1,2,2, Technological innovation and unforeseen compliance options
As Bailey et al. (2002) write, "Ex ante estimates are forecasts and, like all forecasts, their accuracy will be
limited due to uncertainty." There are a number of potential sources of uncertainty in cost analyses.
Perhaps the most prominent is the prospect for the development of new technology to meet regulatory
requirements. Almost all earlier literature surveys highlight that ex ante estimates of the cost of regulation
do not carefully consider the role of innovation, or more broadly, the full range of options open to regulated
entities in complying with tighter standards.
There is a vast literature on the "induced innovation hypothesis," and environmental regulations are listed as
one factor among many that may induce innovation (see, in particular, Jaffe et al. (2003) for a survey of
environmental regulation and innovation). When firms are forced to rethink production processes and
become more efficient, the result may be both environmental improvement and competitive advantage
(Porter 1991, Heinzerling and Ackerman 2002). While it is a recognized best practice to at least attempt to
factor "learning curve" effects into estimates of the costs of regulation (EPA 2010),13 analysts may not
incorporate potential technological innovation into ex ante cost estimates. Even if they do include their best
estimates of future technological improvements, there would still be random variation in how quickly or
completely such improvements are realized.
Different assumptions concerning technological progress, requirements arising from regulations other than
that under consideration in the analysis, and market conditions could all affect the estimated cost of
regulations. For instance, when EPA estimated costs under its Enhanced Inspection and Maintenance
program for automobiles, analysts assumed a high level of effectiveness of repairs and the incorporation of
56 million cars into the program. After implementation, however, it was determined that the repairs were
13 These effects are due to compliance costs tending to decrease over time as regulated entities learn how to more
easily comply with the regulation.
10-

-------
less effective at reducing emissions than EPA analysts assumed. Only four states actually implemented the
program (Harrington et al 2000).
There are numerous cases where technological innovation following a new regulation was underestimated.
In EPA's Chlorofluorocarbon (CFC) Rule, for example, the ex post costs of the CFC phase-out were 30 percent
less than the ex ante estimates, even though an expedited phase-out occurred (Hodges 1997). Analysts
estimating costs prior to the CFC phase-out's implementation did not account for process changes, reliance
on blends of chemicals, and substitutes (e.g., existing hydrofluorocarbons or HFCs). While estimates
suggested that substitutes would be unavailable for almost a decade, industrial efforts led to their availability
after about two years (Hodges 1997; Harrington et al. 2000).14 In this case, while the CFC rule was under
development (for approximately two years), industry researched alternatives. After substitutes and new
practices were identified, firms faced new costs, lower than those anticipated under ex ante estimates
(Hammitt 2000, Harrington et al 2000).
As another example, cost estimates prior to the implementation of the Title IV of the 1990 CAAA failed to
predict technological and process evolution that ended up lowering compliance costs considerably. Original
estimates predicted compliance costs between $4 billion and $5 billion per year (Hodges 1997). In the ex
ante analysis, scrubbers -- the SO2 treatment technology -- were assumed to be less efficient than ex post
studies show. Original estimates rested on assumptions that scrubbers were 85 percent reliable and
removed between 80 and 85 percent of sulfur produced by an electric utility. In actuality, scrubbers have
been more than 95 percent reliable and remove approximately 95 percent of total sulfur (Harrington et al
2000). Moreover, Popp (2003) concluded that Title IV, which was designed to provide incentives to install
scrubbers with higher removal efficiencies, was successful in promoting the introduction of higher efficiency
scrubbers into the market, thereby leading to lower operating costs. The ex ante analysis also did not account
for fuel mixing—the blending of low and high sulfur coal—that lowered sulfur dioxide emissions (Harrington
et al 2000). At the time of the estimates, blending fuels seemed impractical (Hodges 1997).
Finally, it makes sense to suppose that technological innovation is more likely to occur in response to
regulations that affect a large number of facilities. Developing an improved technology is a fixed cost, and so
investment in such technologies will be more attractive the greater the number of production units and cost
savings over which it can be amortized.
1,1,2,3, Unanticipated exogenous changes
Even an analysis that might have proved to be reasonably accurate at the time it was prepared may be well
wide of the mark by the time the rule actually enters into force. The EPA Action Development Process is
often time-intensive. In 2006, the mean action development time for "significant" rules (those requiring
14 CFC-12, used in refrigeration, was replaced with HFC-134a, an existing chemical used in automobile air
conditioners starting back in 1991. Use of CFC-113 in foam-blowing applications has been replaced by HFC, a
substitute; additionally, process changes and chemical blends were essential to decreased consumption of CFC-113
(Harrington et al. 2000).
11 -

-------
benefit-cost analyses) was 1,088 days, or nearly three years.15 Even if we confine our attention to the period
between the proposal of a regulation and the publication of a final rule, Kerwin and Furlong (1992) found
that 523 days elapse on average. Regulatory processes are also often subject to significant amendment and
delay. Cost estimates based on early versions of a rule may no longer apply to the rule that eventually
emerges (Putnam, etal., 1980, Morgenstern and Landy 1997, Harrington, et al., 2000; see also Oosterhuis, et
al., 2006, who note a similar tendency in European regulation).16
Lower costs may arise from factors not directly tied to the regulation, but perhaps indirectly linked to it. In
the case of the SO2 rule, for example, changing market conditions affected the accuracy of ex ante cost
estimates. Cost estimates did not anticipate the impacts of a deregulated railroad industry on the reduction
of sulfur dioxide pollution. Deregulation of railroads allowed for low-cost shipping of low sulfur coals from
the Powder River Basin in Wyoming to the East, decreasing eastern facilities' costs of consuming low sulfur
coal. (Hodges 1997; but see also Busse and Keohane 2007, who argue that the two railroads serving the
Powder River Basin retained some market power). This reduction in price of low-sulfur coal, coupled with
low-cost technological improvements, reduced compliance costs by allowing EGUs in the East to lower SO2
emissions by expanding their use of low-sulfur coal (Hodges 1997; Carlson et al. 2000; Harrington et al. 2000;
Burtraw and Palmer 2004, Busse and Keohane 2007). This change in a related but separate market enabled
electricity generators to alter production processes and fuel sources to achieve SO2 reduction goals. While it
cannot be proved that railroad deregulation was driven by heightened demand for Western coal under the
CAAA, the benefits of railroad deregulation certainly increased with the increase in demand for low-sulfur
coal.
1,2, Methodology
1,2,1 Conceptual Framework for Ex Post Cost: Assessment
Developing a standardized framework provides a systematic way to identify the key components of
compliance costs relevant to a regulation, to assess whether each of the components turned out to be larger
or smaller than the ex ante estimates, and to understand the characteristics of the regulation that influenced
the divergence. While the aim here is not to produce ex post cost estimates, or reproduce the ex ante
estimates, using the same level of rigor employed in the RIAs, we hope to glean enough information on the
drivers of compliance costs to make a weight of evidence determination on the direction of our ex ante
15	See http://www.epa.gov/regstat/development time office2.html
16	Other authors suggest that such delay may be part of the design of the regulatory process. Bailey, et al. (2002)
describe the process of regulatory development in the European Union as a sort of extended negotiation between
regulators and the firms they oversee, with each staking out negotiating positions from which they expect to be
budged over time. This may, however, represent a distinction between European and US practice, the latter of
which they characterize as "adversarial and legalistic."
12 -

-------
estimates - were they likely too high, too low, or about right? - and to identify underlying factors that
contributed to differences (or similarities) between ex ante and ex post costs.
The degree to which an ex post evaluation of the costs of regulation is able to determine the accuracy of the
initial assessment by EPA will vary by rule. We focus the ex post evaluation on costs that - if incorrect - could
fundamentally alter the findings of the ex ante cost assessment. The scope of the ex post analysis is informed
by a brief review of the ex ante cost assessment to identify:
•	The main drivers of costs ex ante: If these drivers were misidentified, ex ante cost estimates might
be flawed.
•	The main sources of uncertainty in estimating costs ex ante: The less that was known with certainty,
the less accurate we would expect cost estimates to be.
•	Unanticipated exogenous changes that occurred after completion of the ex ante analysis that have
significant implications for the costs of the rule: If the "state of the world" changed in unpredicted, or
perhaps unpredictable, ways, estimates would again be less accurate.
Sources of uncertainty are often rule specific but may include: lack of knowledge about who is in the
regulated universe; lack of knowledge about the effectiveness of certain types of control technologies or
processes in reducing pollutants; lack of information about the costs of relatively untried control technologies
or processes; behavioral responses by industry or consumers to changing rules or incentives, including the
possibility of non-compliance (NRC 2012). In general, we maintain a timeline for implementation consistent
with ex ante assumptions. However, in some cases there is uncertainty as to when regulated entities begin to
undertake investment to comply with the rule. Thus, baseline assumptions may themselves be a source of
uncertainty.17
Exogenous changes are often difficult to anticipate ex ante but may have significant implications for the cost
of meeting rule requirements. Examples include unrelated changes in market demand, higher than expected
oil prices, industry wide changes in manufacturing processes (unrelated to the rule), and other regulations,
legal or political decisions that occurred concurrent with or after the ex ante assessment took place but
affected rule implementation.
Using the information gleaned from the scoping exercise, we proceed to an ex post assessment of costs. For
each regulation analyzed, we evaluate likely drivers of identified differences between ex ante and ex post
cost estimates using a broad categorization of cost components consistent with EPA's Guidelines for
Preparing Economic Analyses (2010). Table 1.2 summarizes the key components of the cost analysis and the
main questions we pose as part of the ex post assessment.
17 OMB defines the baseline as "the best assessment of the way the world would look absent the proposed action"
(OMB 2003).
13 -

-------
Table 1.2: Summary of Conceptual Framework to Guide Ex Post Cost Assessment -
Cost Component
Assessment Questions Posed
Regulated Universe
What types of entities are required to comply with the rule? How
many entities of each type are required to comply?
Baseline
To what extent are these technologies already in use prior to the
rule?
Methods of Compliance
What types of technologies or methods are used to comply? How
often are these compliance strategies used?
Direct Compliance Costs
What are the initial or one-time compliance costs (fixed or variable
components)? What are the ongoing compliance costs (operation and
maintenance)?
Indirect Compliance Costs
What are the indirect compliance costs (e.g., quality trade-offs)?
Opportunity Costs
Are there other major opportunity costs associated with the rule (for
instance, in related markets)?
To evaluate unit compliance costs, we combine information about direct costs per unit of abatement (direct
compliance costs) associated with each identified method of compliance (methods of compliance) plus any
additional indirect compliance costs per unit of abatement (indirect compliance costs). When possible, we
also offer an assessment of total compliance costs. To do this, we need to understand whether EPA correctly
identified who has to comply (regulated universe), netting out any facilities already in compliance (baseline).
While ideally we would measure the social cost of regulation (i.e., the sum of all opportunity costs incurred),
most ex ante regulatory analyses only quantify compliance costs. As such, the first five components of the
conceptual framework in Table 1.2 focus on the basic components for quantifying compliance costs. The final
component (opportunity costs) leaves open the possibility of broader ex post evaluation of social cost when
possible.
For each case study, we provide a summary of our assessment by cost component in one table to make it
easy to understand how the ex ante and ex post costs compare and to aid the reader in making comparisons
across case studies (Table 1.3). Table 1.3 includes some sub-categorization of the main cost components to
mirror the assessment questions posed in Table 1.2. However, while we strive for consistency across the case
studies, sub-categories are sometimes modified to reflect unique aspects of a particular rule.
14-

-------
Table 1.3: Generic "Summary of Findings" Table -
Components of Cost Estimate
Source of Ex Post
Information
Assessment (Compared
to Ex Ante)
Regulated
Universe
Types of Entities


Number of Entities


Baseline


Methods of
Compliance
Types


Usage


Compliance
Costs
Direct, One-
Time
Fixed Cost


Variable Cost
Direct, On-
Going
Operating


Maintenance
Indirect


Opportunity
Costs



Per Unit Costs



TOTAL COSTS



1,2,2, Select!
To select the five rules for the case studies presented in this report, we first assembled an inventory of all
EPA regulations coded in the Agency's Rule and Policy Information and Development System (RAPIDS)
database as "economically significant" and promulgated since January 1995. RAPIDS is the Agency's tracking
database for regulatory and significant non-regulatory actions.18 Typically, these are actions that will involve
notice and comment rulemaking, or are major work products that require significant cross-Agency
collaboration. "Economically significant" rules are those anticipated to have an annual effect on the costs or
benefits associated with the rule of $100 million or more as stated in Executive Order 12866.19 We focus on
recent regulations because rules promulgated decades ago will likely have been overridden by new
regulations, making it more difficult to isolate the compliance strategies and costs associated with the old
18	In February 2012, ADP TRACKER replaced RAPIDS as the system EPA uses to track its Action Development
Process (ADP).
19	Regulatory impact analyses are unlikely to be performed if the annual effect is predicted to be less than $100
million.
15 -

-------
rule. Furthermore, the lessons learned from examining older regulations may be less relevant going forward
due to advancements in benefit-cost analysis methodologies that have been adopted since that time.
The RAPIDS search generated a list of 111 entries. We reviewed the list and gathered preliminary
information on each rule (e.g., compliance dates) to determine which rules could feasibly be studied. We
discarded any duplicate entries and rules that were:
•	not yet implemented
•	remanded by the courts
•	consisting of minor amendments to existing rules
•	noted to be "Other significant action" but not meeting $100 million benefit-cost criteria for
E.0.12866, or
•	difficult to analyze (e.g. multi-sector nature of NAAQS).
To that list, we added effluent limitation guidelines, a category of rules that routinely undergoes OMB review
for which detailed cost analyses are produced. The resulting eligible inventory (shown in Table 1.4) consists
of 42 rules promulgated between 1995 and 2005. We circulated this list to EPA program offices for their
feedback to ensure that there were no inadvertent omissions or rules that should not be included. The list
does not include chemical actions as these are not tracked in the RAPIDS database.
Five rules were selected to serve as pilot studies to inform which methodologies are most appropriate to
measure ex post compliance costs for a range of rules. The five rules analyzed in this report are:
•	National Primary Drinking Water Regulation for Arsenic (2001)
•	Integrated NESHAP and Effluent Guidelines for Pulp and Paper - known as the Cluster Rule
(1998)
•	NESHAP: Chemical Recovery Combustion Sources at Kraft, Soda, Sulfite and Stand-Alone
Semichemical Pulp Mills (2001)
•	Locomotive Emission Standards (1998)
•	Methyl Bromide Critical Use Nomination for Preplant Soil Use for Strawberries Grown for
Fruit in Open Fields on Plastic Tarps (2004 - 2008)
These rules were not chosen randomly, but rather were chosen to cover various media, source categories,
and types of regulations (e.g., performance standard versus prescriptive regulation). Four of the rules were
taken from the master list shown in Table 1.4 and described above. The critical use exemption nomination of
a fumigant was identified separately by the Office of Pesticides Program (OPP) as a good candidate for
study.20 Because the two NESHAPs for the Pulp Mills were so closely related, we opted to combine the two
rules into one case study and that case study is part of this report. Future case studies will be chosen using
stratified random sampling (see Chapter 6).
20 We also selected the NSPS for Nitrogen Oxide Emissions but decided to postpone the analysis in a subsequent
report.
16-

-------
Table 1.4. Final EPA Regulations Eligible for Retrospective Study -

Title
Program
Office
Year
1
Final Effluent Limitations Guidelines and Standards for the Coastal Subcategory of
the Oil and Gas Extraction Point Source Category (RIN 2040-AB72)
OW
1996
2
Pharmaceutical Manufacturing Category Effluent Limitations Guidelines,
Pretreatment Standards, and New Source Performance Standards (RIN 2040-AA13)
ow
1998
3
National Primary Drinking Water Regulations: Stage 1 Disinfectant/Disinfection By-
Products Rule (RIN:2040-AB82SAN:2772; Tier:l; Stage:COMPLETED)
OW
1998
4
National Primary Drinking Water Regulations: Interim Enhanced Surface Water
Treatment Rule (RIN:2040-AC91SAN:2304; Tier:N/A; Stage:COMPLETED)
ow
1998
5
NPDES Comprehensive Storm Water Phase II Regulations (RIN:2040-
AC82SAN:3785; Tier:3; Stage:COMPLETED)
ow
1999
6
Effluent Limitations Guidelines, Pretreatment Standards, and New Source
Performance Standards for the Landfills Point Source Category (RIN 2040-AC23)
ow
2000
7
Effluent Limitations Guidelines, Pretreatment Standards, and New Source
Performance Standards for the Commercial Hazardous Waste Combustor
Subcategory of the Waste Combustors Point Source Category (RIN 2040-AC23)
ow
2000
8
Effluent Limitations Guidelines, Pretreatment Standards, and New Source
Performance Standards for the Transportation Equipment Cleaning Point Source
Category (RIN 2040-AB98)
ow
2000
9
Effluent Limitations Guidelines, Pretreatment Standards, and New Source
Performance Standards for the Centralized Waste Treatment Point Source
Category (RIN 2040-AB78)
ow
2000
10
Effluent Limitations Guidelines and New Source Performance Standards for the Oil
and Gas Extraction Point Source Category (RIN 2040-AD14)
ow
2001
11
National Primary Drinking Water Regulations: Arsenic and Clarifications to
Compliance and New Source Contaminant Monitoring (RIN:2040-AB75SAN:2807;
Tier:2; Stage:COMPLETED)
ow
2001
12
Coal Mining Point Source Category; Amendments to Effluent Limitations
Guidelines and New Source Performance Standards (RIN 2040-AD24)
ow
2002
13
Effluent Limitations Guidelines, Pretreatment Standards, and New Source
Performance Standards for the Iron and Steel Manufacturing Point Source
Category (RIN 2040-AC90)
ow
2002
14
Effluent Limitations Guidelines and New Source Performance Standards for the
Metal Products and Machinery Point Source Category (RIN 2040-AB79)
ow
2003
15
Effluent Limitations Guidelines and New Source Performance Standards for the
Concentrated Aquatic Animal Production Point Source Category (RIN 2040-AD55)
ow
2004
16
Effluent Limitations Guidelines and New Source Performance Standards for the
Meat and Poultry Products Point Source Category (RIN 2040-AD56)
ow
2004
17
Land Disposal Restrictions - Phase III: Decharacterized Wastewaters, Carbamate
Wastes, and Spent Aluminum Potliners (RIN:2050-AD38SAN:3365; Tier:l;
Stage:COMPLETED)
OSWER
1996
17 -

-------

Title
Program
Office
Year
18
Risk Management Program for Chemical Accidental Release Prevention (RIN:2050-
AD26SAN:2979; Tier:N/A; Stage:COMPLETED)
OSWER
1996
19
Land Disposal Restrictions - Phase IV: Treatment Standards for Metal Wastes and
Mineral Processing wastes; Mineral Processing Secondary Materials and Bevill
Exclusion Issues (RIN:2050-AE05SAN:3366; Tier:2; Stage:COMPLETED)
OSWER
1998
20
PCBs; Polychlorinated Biphenyls (PCBs) Disposal Amendments (RIN:2070-
AD04SAN:2878; Tier:2; Stage:COMPLETED)
OPPTS
1998
21
Lead; Identification of Dangerous Levels of Lead Pursuant to TSCA Section 403
(RIN:2070-AC63SAN:3243; Tier:l; Stage:COMPLETED)
OPPTS
2001
22
TRI; Reporting Threshold Amendment for Certain Persistent and Bioaccumulative
Toxic Chemicals (PBTs) (RIN:2070-AD09SAN:3880; Tier:l; Stage:COMPLETED)
OEI
1999
23
Emission Standards for Marine Tank Vessel Loading Operations (RIN:2060-
AD02SAN:3104; Tier:N/A; Stage:COMPLETED)
OAR
1995
24
NSPS: Municipal Waste Combustion-Phase II and Phase III (Large Units) (RIN:2060-
AD00SAN:2916; Tier:l; Stage:COMPLETED)
OAR
1995
25
NSPS: Municipal Solid Waste Landfills Amendments (RIN:2060-AC42SAN:2535;
Tier:3; Stage:COMPLETED)
OAR
1996
26
Regulation of Fuel and Fuel Additives: Certification Requirements for Deposit
Control Additives (RIN:2060-AG06SAN:3597; Tier:2; Stage:COMPLETED)
OAR
1996
27
Control of Emissions of Air Pollution: Emission Standards for Gasoline Spark-
Ignition and Diesel Compression-Ignition Marine Engines (RIN:2060-
AE54SAN:3350; Tier:N/A; Stage:COMPLETED)
OAR
1996
28
Federal Test Procedure for Emissions From Motor Vehicles and Motor Vehicle
Engines; Review (RIN:2060-AE27SAN:3323; Tier:N/A; Stage:COMPLETED)
OAR
1996
29
Acid Rain Program: Nitrogen Oxides Control Regulation (RIN:2060-AD45SAN:2888;
Tier:N/A; Stage:COMPLETED)
OAR
1996
30
Acid Rain Program: Phase II Nitrogen Oxides Reduction Program (RIN:2060-
AF48SAN:3575; Tier:3; Stage:COMPLETED)
OAR
1996
31
Hospital/Medical/lnfectious Waste Incinerators (RIN:2060-AC62SAN:2719;
Tier:N/A)
OAR
1997
32
Control of Emissions of Air Pollution From Nonroad Diesel Engines (RIN:2060-
AF76SAN:3645; Tier:l; Stage:COMPLETED)
OAR
1997
33
Compliance Assurance Monitoring Rule (Previously Enhanced Monitoring Rule)
(RIN:2060-AD18SAN:2942; Tier:l; Stage:COMPLETED)
OAR
1997
34
Integrated NESHAP and Effluent Guidelines: Pulp and Paper (RIN:2060-
AD03SAN:3105; Tier:l; Stage:COMPLETED)
OAR
1998
35
Locomotive Emission Standards (RIN:2060-AD33SAN:2961; Tier:2;
Stage:COMPLETED)
OAR
1998
36
NSPS: Municipal Solid Waste Landfills Amendments (RIN:2060-AI09SAN:4150;
Tier:3; Stage:COMPLETED)
OAR
1998
18 -

-------

Title
Program
Office
Year
37
NSPS: Nitrogen Oxide Emissions From Fossil-Fuel Fired Steam Generating Units-
Revision (RIN:2060-AE56SAN:3352; Tier:2; Stage:COMPLETED)
OAR
1998
38
Control of Emissions of Air Pollution From Nonroad Diesel Engines (RIN:2060-
AF76SAN:3645; Tier:l; Stage:COMPLETED)
OAR
1998
39
Control of Emissions from Nonroad Diesel Engines (RIN:2060-AH50SAN:4014;
Tier:l; Stage:COMPLETED)
OAR
1998
40
Finding of Significant Contribution and Rulemaking for Certain States in the Ozone
Transport Assessment Group (OTAG) Region for Purposes of Reducing Regional
Transport of Ozone (RIN:2060-AH10SAN:3945; Tier:2; Stage:COMPLETED)
OAR
1998
41
NESHAP: Source Categories: (SOCMI) and Other Processes Subject to Negotiated
Regulation for Equipment Leaks (HON) (RIN:2060-AC19SAN:2363; Tier:N/A;
Stage:COMPLETED)
OAR
1999
42
Tier II Light-Duty Vehicle and Light-Duty Truck Emission Standards and Gasoline
Sulfur Standards (RIN:2060-AI23SAN:4211; Tier:l; Stage:COMPLETED)
OAR
2000
1,2,3, Strategies for Ex > , ,	i-
Various methodologies exist for collecting the information needed to conduct the ex post assessments,
ranging from using publicly available data sources, reaching out to industry compliance experts, conducting
site visits to facilities, and/or to administering a comprehensive industry survey such as the Pollution
Abatement Costs and Expenditures (PACE) survey (see Text Box 1.2.21 For each case study, we assessed
whether it would be possible to collect sufficient ex post compliance cost information using only publicly-
accessible data sources. For example, in the case of the 1998 Locomotive Rule, we assessed whether there
are any databases from which we could determine the number of locomotives in operation (based on data of
original manufacture or remanufacture) to compare with EPA's ex ante estimate. Similarly, we explored the
availability of public data on the control mechanisms used for each locomotive to come into compliance with
the rule requirements and the cost of such mechanisms.
In general, we found that while data for some necessary components are readily available, cost information is
generally lacking. The critical use exemption for methyl bromide use for California strawberries fared the
best of the five with regard to the availability of cost information, and was designated as the case study that
would be based on publically available data alone.22 We also explored the applicability and usefulness of the
other methodologies for each rule to help inform analysis of future rules for this project.
21	In the future it may be possible to collect ex post cost data for a particular rule by targeting the affected
regulated entities directly using the PACE survey - this hinges on the PACE survey once again becoming an annual
survey. The PACE survey has not been conducted since 2005 and has not been conducted annually since 1994.
22	Ultimately, publicly available data were used to augment other sources for the Arsenic rule and the MACTII rule.
19 -

-------
For four rules -the combined Cluster Rule and MACT 11, Arsenic rule, and Locomotive rule - we consulted -
industry compliance experts with contractor assistance to gather information on compliance strategies and
ex post cost data. The process used to identify appropriate industry compliance experts with sufficient
knowledge about the ex post regulatory compliance costs of the selected rules consisted of several steps. For
each rule, we began by examining the rulemaking docket, the primary source for the initial set of potential
industry compliance experts. This set includes organizations that supplied data and information during the
original rulemaking and/or commented on it during the comment period. The initial set of potential industry
compliance experts was circulated to relevant EPA staff for review. In some instances, the relevant program
office was able to suggest additional potential industry compliance experts. We also allowed for identification
of industry compliance experts through discussions with other entities or targeted internet searches. In
some cases, for example, independent associations suggested appropriate engineering compliance assistance
firms. We approached the following types of organizations during the information collection process for a
given rule: engineering compliance assistance firms, compliance technology vendors, compliance assistance
firms or consultants, independent associations of entities affected by regulations, independent information
publishers, state regulatory agencies, and EPA contractors who supported the rule.23
Screening and securing commitment from the identified experts to participate in our study required
considerable effort. In most instances, it took at least 2 to 3 rounds of phone calls to reach an individual
within each organization who would be able to provide relevant feedback. Even after finalizing information
provision agreements with the experts, weekly email and phone reminders were sometimes necessary to
ensure their timely participation. To aid in the conversations with the experts, we developed a pilot
questionnaire about each rule based on our review of EPA's ex ante cost estimation methodology. This
questionnaire was also circulated to relevant EPA staff for comment and feedback. Each expert was also
asked to provide documentation for any calculations he or she made to answer the cost questions during the
interview. Summaries of the outreach effort for particular rules are described within each case study below
together with the questionnaires.
23 Any information provided for the RCS by contractors who helped EPA develop the rule was extensively
documented.
20-

-------
Text Box 1.2. Pollution Abatement Costs and Expenditures Survey (PACE) -
The PACE survey was conducted annually between 1973 and 1994 (with the exception of 1987), but was
discontinued after 1994 by the U.S. Census Bureau for budgetary reasons. Recognizing the need for this type
of data, EPA provided the necessary financial and technical support to enable the Census Bureau to conduct
additional surveys and collect PACE data for 1999 and 2005, but limitations on resources and other priorities
have limited more recent data collection to these two years.
The PACE survey is the only comprehensive publicly available source of pollution abatement (operating) costs
and (capital) expenditures spending for the U.S. manufacturing sector. The PACE survey collects
establishment-level information on pollution abatement capital expenditures and operating costs associated
with compliance with local, state, and federal regulations, as well as voluntary or market-driven pollution
abatement activities. The PACE survey intends to capture only incremental costs of pollution abatement.
The pollution abatement capital expenditures and operating costs are disaggregated into four "activity"
categories: treatment, recycling, disposal, and pollution prevention, and by three types of media: air
emissions, water discharges, and solid waste. Total pollution abatement operating cost are separated into
five cost categories: (1) salaries, wages, and benefits; (2) energy costs; (3) materials and supplies; (4) contract
work, leasing, and other purchased services; and (5) depreciation.
While EPA uses the PACE data to estimate the aggregate costs of its regulations, the data collected by the
PACE survey contains information that could be useful in estimating the ex post cost of specific EPA
regulations on the manufacturing sector in several ways. First, if EPA regulates an entire industry, EPA could
approximate the incremental cost of a regulation by comparing pollution abatement costs for the entire
industry before and after a regulation becomes effective. Second, if EPA knows which manufacturing facilities
need to comply with a new regulation, EPA could estimate the incremental cost of the regulation using the
establishment-level data at the US Census Bureau. Finally, if the PACE survey were to become an annual
survey once again, EPA could use it to estimate the incremental cost of a new or more stringent regulation by
developing a very specific set of questions that would only be sent to manufacturing facilities that EPA
believed to be covered by the rule. Also since EPA would have the ability to collect cost data for several years
before and after the regulation became effective it would provide more information on how pollution
abatement costs change over time. This would also allow EPA to estimate how regulations induce
technological change and affect employment.
Chapter 1 References
Adler, Jonathan. 1996. Behind the green curtain. Regulation 4.26-34.
Anderson, J.F., and Sherwood, T. 2002. Comparison of EPA and Other Estimates of Mobile Source Rule Costs
to Actual Price Changes. Paper presented at the SAE Government Industry Meeting, Washington, DC,
May 14, 2002.
Ando, A., V. McConnell and W. Harrington. 2000. Costs, Emission Reductions and Vehicle Repair: Evidence
from Arizona. Journal of the Air and Waste Management Association 50: 174-185.
21 -

-------
Bailey, Peter D., Gary Haq, and Andy Goudson. 2002. Mind the gap! Comparing ex ante and ex post
assessments of the costs of complying with environmental regulation. European Environment 12 (5):
245 - 256.
Boardman, Anthony E., Wendy L. Mallery and Aidan R. Vining. 1994. Learning from ex ante/ex post cost-
benefit comparisons: the coquihalla highway example. Socio-Economic Planning Sciences, 28(2):69-
84.
Burtraw, D., A. Krupnick, E. Mansur, D. Austin and D. Farrell. 1998. Costs and Benefits of Reducing Air
Pollutants Related to Acid Rain. Contemporary Economic Policy 16(4): 379-400.
Burtraw, D. and K. Palmer. 2004. S02 Cap-and-Trade Program in the United States: A 'Living Legend' of
Market Effectiveness. In Choosing Environmental Policy: Comparing Instruments and Outcomes in the
United States and Europe, ed. Winston Harrington, Richard D. Morgenstern, and Thomas Sterner.
Washington D.C.: Resources for the Future.
Busse, Meghan R., and Nathaniel O. Keohane. 2007. Market Effects of Environmental Regulation: Coal,
Railroads, and the 1990 Clean Air Act. RAND Journal of Economics 38 (4): 1159-1179.
Carlson, C., D. Burtraw, M. Cropper, and K. Palmer. 2000. S02 Control by Electric Utilities: What Are the Gains
from Trade? Journal of Political Economy 108(6): 1292-326.
Chan, G., R. Stavins, R. Stowe, and R. Sweeney (2012). The S02 Allowance Trading System and the Clean Air
Act Amendments of 1990: Reflections on Twenty Years of Policy Innovation. White paper. Kennedy
School of Government, Harvard University.
Dale, Larry, Camille Antinori, Michael McNeil, James E. McMahon and K. Sydny Fujita. 2009. Retrospective
evaluation of appliance price trends. Energy Policy 37,2 (February). 597-605.
Dayton, Julia. 1998. World Bank HIV/AIDS Interventions: Ex Ante and Ex Post Evaluations. World Bank
Publication 389.
Ellerman, A.D., P.L. Joskow, R. Schmalensee, J.P. Montero, and E. Bailey. 2000. Markets for Clean Air: The U.S.
Acid Rain Program. Cambridge: Cambridge University Press.
Gallaher, M., C. Morgan and R. Shadbegian. 2008. Redesign of the 2005 Pollution Abatement Costs and
Expenditure Survey. Journal of Economic and Social Measurement 33: 309-360.
Goulder, L., I.W.H. Parry, and D. Burtraw. 1997. Revenue-Raising Versus Other Approaches to Environmental
Protection: The Critical Significance of Preexisting Tax Distortions. RAND Journal of Economics 28(4):
708-731.
Greenstone, M., and T. Gayer. 2009. Quasi-experimental and experimental approaches to environmental
economics. Journal of Environmental Economics and Management 57: 21-44.
Grosse, Scott D., Norman J. Waitzman, Patrick S. Romano, and Joseph Mulinare. 2005. Reevaluating the
Benefits of Folic Acid Fortification in the United States: Economic Analysis, Regulation, and Public
Health. American Journal of Public Health. 95(11):1917-1922.
Hahn, Robert, and Rohit Malik. 2004. Is regulation good for you? Harvard Journal of Law and Public Policy
27, 3 (October). 893-916.
22 -

-------
Hammitt, J. 2000. Are The Costs of Proposed Environmental Regulations Overestimated? Evidence from the
CFC Phaseout. Environmental and Resource Economics 16(3): 281-302.
Harrington, W., R. D. Morgenstern, and P. Nelson. 2000. On the Accuracy of Regulatory Cost Estimates.
Journal of Policy Analysis and Management 19(2): 297-322.
Heinzerling, Lisa, and Richard Ackerman. 2002. Pricing the priceless: Cost benefit analysis of environmental
protection. University of Pennsylvania Law Review 150: 1553 - 1584.
Hodges, Hart. 1997. Falling prices: Cost of complying with environmental regulations almost always less
than advertised. Economic Policy Institute Briefing Paper. Washington, DC: Economic Policy
Institute.
ICF. 1995. Economic Analysis of Title IV Requirements of the 1990 Clean Air Act Amendments. Prepared for
U.S. EPA, Washington, DC. September.
Jaffe, Adam B., Richard G. Newell, and Robert N. Stavins. 2003. Technological Change and the Environment.
In Volume 1 of Handbook of Environmental Economics, ed. Karl-Goran Maler and Jeffrey Vincent,
461-516 (Chapter 11), New York, North-Holland.
James, H. S. j'r, 1998. Estimating OSHA Compliance Costs. Policy Sciences - Special Issue on Regulatory
Budgeting 31: 321-4.
Jantzen, J. (1989), Costs of Environmental Management, 1988-2010, 3 policy scenarios. Report for the
Ministry of VROM, The Hague, 25 May 1989.
Joskow, Paul. 1988. Asset specificity and the structure of vertical relationships: empirical evidence. Journal
of Law, Economics and Organization 4 (1): 95-117.
Kerwin, C. M., and S. Furlong. 1992. Time and Rulemaking: An Empirical Test of Theory. Journal of Public
Administration Research and Theory 2(2): 113-38.
LaFrance, Jeffrey T. and Harry de Gorter. 1985. Regulation in a Dynamic Market: The U.S. Dairy Industry.
American Journal of Agricultural Economics, 67(4):821-832.
Lindner, R. K. and F. G. Jarrett. 1978. Supply Shifts and the Size of Research Benefits. American Journal of
Agricultural Economics, 60(l):48-58.
MacLeod, Michael, Dominic Moran, Manuel Lago Aresti, Winston Harrington, and Richard Morgenstern.
2006. Comparing the ex ante and ex post costs of complying with regulatory change. Final report to
DEFRA. London: Department for Environment, Farms, and Rural Affairs.
Magat, W. A., A. J. Krupnick, and W. Harrington 1986. Rules in the Making: A Statistical Analysis of
Regulatory Agency Behavior. Washington, DC: Resources for the Future.
Maloney, Michael T., and Robert E. McCormick. 1982. A positive theory of environmental quality regulation.
Journal of Law and Economics 25, 1 (April). 99 - 123.
Morgenstern, Richard D. and Mark K. Landy. 1997. Economic analysis: Benefits, costs, implications," in
Richard D. Morgenstern, ed., Economic Analyses at EPA: Assessing Regulatory Impact (Washington,
D.C.: Resources for the Future).
23 -

-------
NRC (National Research Council). 2012. Review of the EPA's Economic Analysis of Final Water Quality
Standards for Lakes and Flowing Waters in Florida. Washington, DC: The National Academies Press.
Office of Technology Assessment (OTA). 1995. Gauging Control Technology and Regulatory Impacts in
Occupational Safety and Health: An Appraisal of OSHA's Analytic Approach. OTA-ENV-635.
Washington DC: Government Printing Office.
Oosterhuis, F., C. des Abbayes, B. Gorlach, D. Huybrechts A. Jarvis, O. Kuik, J. Medhursi, E. Meynaerts, V.
Monier, P. Ekins, J. Jantzen, R. Vanner, P. Vercaemst, and H. van der Woerd. 2006. Ex post estimates
of costs to business ofEU environmental legislation: Final report. Amsterdam: Institute for
Environmental Studies. Available at:
http://ec.europa.eu/environment/enveco/ex post/pdf/costs.pdf.
Popp, D. 2003. Pollution Control Innovations and the Clean Air Act of 1990. Journal of Policy Analysis and
Management 22(4): 641-660.
Porter, Michael. 1991. America's green strategy. Scientific American 264 (4). 96.
Putnam, Hayes, & Bartlett, Inc. 1980. Comparisons of estimated and actual pollution capital expenditures for
selected industries. Report Prepared for the Office of Planning and Evaluation, U.S. Environmental
Protection Agency, Cambridge, MA. (mimeo).
Rideout, Douglas B. and Philip N. Omi. 1995. Estimating the Cost of Fuels Treatment. Forest Science, 41(4):
664-674(11).
RIVM (2000), Techno 2000; Modellering van de doling van eenheidskosten van technologieen in de tijd.
Rapportnummer 773008003, Bilthoven, April 2000.
Salop, Steven, and David T. Scheffman. 1983. Raising rivals'costs. American Economic Review 73 (2). 267-
271.
Sappington, David E. M. and Joseph E. Stiglitz. 1987. Privatization, information and incentives. Journal of
Policy Analysis and Management, 6(4): 567-585.
Simpson, R. D. 2011. Do Regulators Overestimate the Costs of Regulation? U. S. Environmental Protection
Agency, National Center for Environmental Economics. Working Paper #11-07.
Thompson, Kimberley M., Maria Sequi-Gomez, and John D. Graham. 2002. Validating benefit and cost
estimates: the case of airbag regulation. Risk Analysis 22 (4). 803-811.
U. S. Environmental Protection Agency (U.S. EPA), National Center for Environmental Economics. -2010.
Guidelines for Preparing Economic Analyses. Online at:
http://yosemite.epa.gov/ee/epa/eed.nsf/pages/guidelines.html
U.S. Office of Management and Budget, Office of Information and Regulatory Affairs (OMB). -2005.
Validating Regulatory Analysis: 2005 Report to Congress on the Costs and Benefits of Federal
Regulations and Unfunded Mandates on State, Local, and Tribal Entities.
U.S. Office of Management and Budget (OMB). 2003. Circular A-4. Online at:
http://www.whitehouse.gov/sites/default/files/omb/assets/regulatorv matters pdfZa-4.pdf
U.S. Office of Management and Budget (OMB). 1998. Report to Congress on costs and benefits of federal
regulations: Appendix. Washington, DC: GPO.
24-

-------
White, K. 1997. SO2 Compliance and Allowance Trading: Developments and Outlook. Prepared for the Electric
Power Research Institute (EPRI) EPRI TR-107897 (April).
White, K., Energy Ventures Analysis, Inc., and Van Horn Consulting. 1995. The Emission Allowance Market
and Electric Utility SO2 Compliance in a Competitive and Uncertain Future. Prepared for the Electric
Power Research Institute (EPRI), TR-105490, Palo Alto, CA. Final Report: September.
25 -

-------
Chapter 2: Cluster Rule and MACT II Rule
Cynthia Morgan, Carl Pasurka and Ron Shadbegian
On April 15,1998, U.S. Environmental Protection Agency (EPA) published new National Emission Standards
for Hazardous Air Pollutants from the Pulp and Paper Industry (subpart S) as well as Effluent Limitations
Guidelines, Pretreatment Standards, and New Source Performance Standards: Pulp, Paper, and Paperboard
Point Source Category. Because the promulgated rule integrated air and water rulemakings, the combined
standards and guidelines became known as the "Cluster Rule." The Cluster Rule, EPA's first integrated, multi-
media regulation, set limits to reduce releases of toxic (e.g., dioxin, furans, chloroform) and nonconventional
(e.g., adsorbable organic halides, chemical oxygen demand) pollutants to both air and water from the pulp
and paper industry. According to EPA, 155 of the 565 pulp, paper and paperboard mills in the U.S. needed to
comply with the new maximum achievable control technology (MACT I and III) standards for hazardous air
pollutants. Of those 155 mills, 96 mills were also required to comply with either a new set of best available
technology (BAT) economically achievable effluent guidelines or pretreatment standards for existing sources
(PSES) (see U.S. EPA 1997b, p. 4-5).24 Most requirements of the Cluster Rule became effective April, 2001.
Later, on January 12, 2001, EPA published the MACT II (combustion sources) rule to regulate chemical
recovery combustion sources in the pulp and paper industry. This rule, which had to be met by January 12,
2004, established standards for sources annually emitting at least 10 tons of a hazardous air pollutant (HAP)
or 25 tons of total HAPs. At rule proposal, it was anticipated that 149 of the mills subject to MACT I would
also be subject to MACT II (see U.S. EPA 1998a, p. 18579). By the time of the promulgation of the final rule,
EPA (2001b, Appendix B) identified 133 mills that would be subject to MACT II. A provision of the MACT II
that improved the efficiency of the regulation for existing sources was a "bubble compliance alternative"
allowing mills to reduce emissions at any unit as long as the mill-specific bubble limit was achieved.
In this paper, we compare EPA's ex ante cost analyses of the Cluster and MACT II rules to an ex post
assessment of costs. This is not an evaluation of how well EPA conducted its ex ante analyses at the time of
the rulemaking. Instead we attempt to gather enough information on the key drivers of compliance costs to
make an informed judgment as to whether ex post costs are higher or lower than the estimates of ex ante
24 U.S. EPA (1997b, p. 2-6) summarizes the mill subcategories (i.e., pulping processes) subject to the air and water
provisions of the Cluster Rule. According to the U.S. EPA (1997b, p. 1-3), the technological basis for PSES is "... the
same as the basis for the BAT limitationswith the exception of biological treatment." Hence, in this paper we
often refer to the effluent limitation guidelines (ELGs) of the Cluster Rule as BAT.
26-

-------
costs for these rules. This allows us to observe whether actual costs diverged from ex ante costs and, if so,
what factors caused this divergence (e.g., changing market conditions, technological innovation, etc.).
The remainder of this chapter is organized as follows. Section 2.1 details the impetus and timeline for
regulatory action. Section 2.2 presents EPA's ex ante cost estimates of the Cluster Rule and MACT II, while
Section 2.3 discusses the information available to conduct the ex post evaluation of costs. Section 2.4
presents the results of our ex post assessment of compliance costs. Finally, Section 2.5 summarizes our
findings and discusses limitations of our analysis.
2.1. Impetus and Timeline for Regulatory Action
A citizen's petition filed in October 1984 by the Environmental Defense Fund (EDF) and the National Wildlife
Federation (NWF) represents the origin of the Cluster Rule and MACT II regulations. After EPA denied the
petition, the EDF and NWF filed a lawsuit against EPA that ended when EPA signed a consent decree in 1988.
The consent decree required EPA to address the issue of discharges of dioxins and furans into surface waters
by October 31,1993, while the Clean Air Act (CAA) amendments of 1990 required EPA to set MACT standards
for the industry by 1997. As a result, EPA decided to combine the rulemakings and design the most cost-
effective rule and reduce cross-media pollution transfers.26 EPA proposed its regulations on December 17,
1993 and solicited comments and data on the rule.
The 1993 proposed Cluster Rule required complete substitution of elemental chlorine-free bleaching (ECF),
which uses chlorine dioxide (CIO2) as the bleaching agent, for elemental chlorine bleaching as well as the use
of oxygen delignification (i.e., 02 delig) and/or extended delignification (i.e., extended delig) for 77 bleached
papergrade kraft mills in mid-1995 (see U.S. EPA 1997b, p. 4-5). 02 delig reduces the amount of lignin in the
pulp before bleaching process, minimizing the bleaching chemicals required to brighten the pulp. In addition,
10 papergrade sulfite mills were required to use totally chlorine-free bleaching (TCF). EPA anticipated 300
pulp and paper mills would incur costs due to the proposed 1993 Cluster rule, with 11-13 mills confronting
the possibility of closure. This led EPA to project that capital expenditures associated with an integrated (i.e.,
air and water) regulatory strategy would approach $4 billion (in 1992 dollars) with annual operating and
maintenance (O&M) costs of $401 million (see U.S. EPA 1993a, pp. 66153- 66154). Non-EPA sources
estimated the Cluster Rule would cost $11.5 billion (see Pauksta, 1995, p. 51), while the cost of the combined
Cluster Rule and MACT II rule would be $13.2 billion (see Barton, et al., 1995). An important component of
the cost of the proposed regulation was the requirement of 02 delig or extended delig. Barton et al. (1995,
p. 104) estimated the combined cost of the 02 delig systems and improved brown stock washing would be
25The discussion of the origins of the Cluster Rule and MACT II Rule is drawn from Powell (1997, pp. 1-12), and the
U.S. EPA (1993c, Chapter 2).
26 By promulgating the air and water standards simultaneously, EPA was able to develop control options that
included process change technology that would control both emissions to air and pollutant discharges to water.
27 -

-------
$2.3 billion, while CIO2 upgrades and conversions would cost another $530 million.27 In the ensuing years,
00
the Cluster Rule underwent substantial modification before the final rule was promulgated in 1998. In
addition to fewer mills being affected by the 1998 final rule compared to the 1993 proposal, the final rule
dropped the 02 delig / extended delig requirement, which led some companies to petition EPA and request
incentives/rewards for mills that installed 02 delig (EPA Asked 1996).
In the final Cluster Rule for air pollutants, EPA set MACT standards (referred to as MACT I & III) that required
pulp and paper mills to capture and treat toxic air pollutant emissions produced during the pulping and
bleaching stages of the manufacturing process. The MACT I (non-combustion sources) rule covers mills that
chemically pulp wood using kraft, semi-chemical, sulfite, or soda processes, while the MACT III rule covers
mills that mechanically pulp wood, or pulp secondary fiber or non-wood fibers, or produce paper or
paperboard. EPA estimated that HAPs emissions would decline by 139,000 megagrams (one ton equals
0.908 megagrams) per year.29 These standards could be met in a variety of ways including performance
standards (percent reductions in emissions, mass reductions in emissions, and concentration or mass limits),
design standards (use of specific technologies operated in a certain way), and routing of emissions to
combustion or control devices.
The effluent limitation guidelines (ELGs) established in the Cluster Rule covered two subcategories of mills:
bleached papergrade kraft and soda (BPK) and papergrade sulfite (PS). The ELGs and pretreatment standards
set technology-based limits on dioxins, furans, chloroform, 12 chlorinated phenolics, and adsorbable organic
halides (AOX), requiring a 96 percent reduction in dioxin and furan, and a 99 percent reduction in chloroform.
These requirements were based on substituting chlorine dioxide for chlorine in the bleaching process (i.e.,
using ECF or TCF bleaching). The options for the BPK subcategory (listed in terms of increasing stringency)
were 100 percent substitution of chlorine dioxide for elemental chlorine (ECF), 100 percent substitution of
chlorine dioxide for elemental chlorine (ECF) plus oxygen delignification and/or extended delig, and total
chlorine free (TCF) bleaching. EPA only estimated costs for: TCF bleaching for the calcium- and magnesium-
based processes; and 100 percent substitution of chlorine dioxide (ECF) for elemental chlorine ammonium-
based processes and specialty grade pulps.
The Cluster Rule encouraged additional pollutant reductions through the Voluntary Advanced Technology
Incentives Program (VATIP). Mills who were interested in this program were given extended compliance time
in order to explore all technology options or make process changes that would reduce pollution beyond the
27	The goal of brown stock washing is to remove the maximum amount of spent cooking liquor from the pulp using
the minimum amount of wash water. The solids left in the pulp can interfere with the bleaching process and
increase the costs of bleaching.
28	Rule and implementation information for the air portion of the Cluster Rule can be found at:
http://www.epa.gov/ttnatw01/pulp/pulppg.html. Information on the Effluent Guidelines for the Cluster Rule can
be found at: http://water.epa.gov/scitech/wastetech/guide/pulppaper/index.cfm
29	The HAPs covered by the Cluster Rule included compounds such as methanol, chlorinated compounds,
formaldehyde, benzene, and xylene.
28 -

-------
discharge limits required by the rule. The program was voluntary and only available to mills that discharged
directly to surface waters. Mills that chose to participate received six years to comply with the air standards
(April 15, 2004) and an extension of up to eight years for high volume low concentration (HVLC) system vents
at kraft mills (April 17, 2006). This extension was designed to encourage mills to install technology to reduce
toxic air pollutant emissions as well as discharges of pollutants to air and water from the bleaching process.
In addition to the MACT I and MACT III standards, on January 12, 2001 EPA published the MACT II rule that
regulates chemical recovery combustion sources in the pulp and paper industry. The MACT II rule covers
kraft, soda, sulfite, and stand-alone semi-chemical pulp mills. The MACT II standards covered HAP metals and
gaseous organic HAPs using particulate matter (PM) as a proxy for HAP metals and methanol, and total
hydrocarbons as proxies for gaseous organic HAPs. For existing kraft and soda mills, a PM bubble compliance
alternative allowed mills to set PM limits for each emission point, as long as the aggregate of these PM limits
was equal to the aggregated promulgated PM limits of the individual emission points.
2.2, Ex Ante Cost Estimates
At the proposal the baseline was 1992; however, EPA later updated the baseline pollutant loadings to mid-
1995 (U.S. EPA 1997a, p. 4-1). The updated baseline values are reported in Table 2.1. The updated baseline is
also reflected in the EPA ex ante cost estimates of the Cluster Rule and MACT II Rule reported in Table 2.2.
With the publication of the final MACT II rule (U.S. EPA 2001a, pp. 3188-3189), EPA revised its estimate of
the MACT II capital expenditures to $241 million (in 1997 dollars), and its estimate of the annual cost of
MACT II to $32.2 million (in 1997 dollars). According to EPA (1997a, pp. 2-2 and 2-3), "The MACT III rule
contains National Emission Standards for Hazardous Air Pollutants (NESHAP) for mechanical pulping,
secondary fiber pulping, and non-wood pulping mills. No emission reductions or control costs, however, are
associated with the MACT III rule ..." Table 2.2 is supplemented by Table 2.3 and Table 2.4, which show
estimated ex ante costs from several non-EPA sources.
Table 2.3 is divided into three parts based on which rule(s) is associated with the corresponding cost
estimate. First, we list two non-EPA estimates that combine the cost of the Cluster Rule and MACT II rule.
Next, we list three non-EPA estimates of the Cluster rule, and finally we list two non-EPA estimates of
portions of MACT I. Table 2.4 lists three non-EPA estimates of MACT II. Comparing Tables 2.2, 2.3, and 2.4
reveal: (1) both EPA and the pulp and paper industry believed the Cluster Rule would be more costly than the
MACT II rule and (2) industry believed EPA ex ante cost estimates substantially underestimated the cost of
the Cluster Rule and MACT II rule.
30 In exchange for mills reducing discharges beyond BAT levels, the VATIP offered mills "... additional time to
comply with the Cluster Rules, ... reduced monitoring requirements, and public recognition." (see U.S. EPA 2006,
p. 9-5)
29 -

-------
Table 2.1. Pre-Regulation and Post-Regulation Releases of Selected Pollutants (mid-1995 baseline)
Air Pollutants
Baseline
Air Emission Reductions (Mg/year)


(Mg/year)
Final Cluster Rules
Final Cluster Rules



and Proposed MACT II
Hazardous Air Pollutants
240,000
139,000
142,000

Volatile Organic Compounds
900,000
409,000
440,000

Total Reduced Sulfur
150,000
79,000
79,000

Particulate
NA
(83)
24,000

Carbon Monoxide
NA
(8,700)
49,000


Water Units
Baseline
Estimated
Baseline
Estimated
Pollutants
Discharge
Reductions; Final
Discharge
Reductions; Final

(BPK Mills)
BAT/PSES (BPK
(PS Mills)
BAT/PSES (PS


Mills)

Mills)
2,3,7,8 -TCDD g/year
15
11
0.78
0.65
2,3,7,8-TCDF g/year
115
107
6.7
6.4
Chloroform kkg/year
48
40
5.4
5.2
g- grams
kkg-metric ton (1,000 kilograms)
Source: U.S. EPA (1998a, p. 18575)
Table 2.2. U.S. EPA Ex Ante Cost Estimates of the Cluster Rule & MACT II Rule
(thousands of 1995 dollars)
MACT IA
MACT II
(alternate A)
BAT/PSES
Cluster Rule
(MACT I plus
BAT/PSES)
Cluster Rule plus
MACT II
Capital
O&M
Post tax
Annualized
500,758
74,718
81,767
258,389
5,202
23,139
1,039,388
158,413
171,619
1,540,146
233,131
253,386
1,798,535
238,334
276,525
Source: EPA (1997a, p. 5-27) -
30-

-------
Table 2.3. Non-EPA Ex Ante Cost Estimates of the Cluster Rule
Operating
Source	Capital Expenditures Costs
Cluster Rule plus MACT II

American Forest & Paper Association
$2.6 billion $273 million
(see Miller Freeman Publications, Inc. 1999, p. 77)

Pulp & Paper Project Report, April 1998
$3.2+ billion
(see Miller Freeman Publications, Inc. 1999, p. 77)

Cluster Rule

Parthasarathy and Dowd (2000, p. 41)
$2,625 billion*
National Council for Air and Stream Improvement (2003, p. 5)
$3 billion (1999-2005) —
Jensen (1999, p. 72)
$2,675-2.916 billion
MACTI

Garner (2001, p. 44)
$2-3 billion **
Garner (2001, p. 44)
$0,775 billion***
* $1,375 billion for MACT I & III and $1,250 billion for BAT and best management practices (BMP) -
** MACT I (April 2001 compliance) -
*** MACT I (HVLC pollutants, April 2006 compliance) -
Table 2.4. Non-EPA Ex Ante Cost Estimates of MACT II
Capital	Operating
Source	Expenditures Costs
Parthasarathy and Dowd (2000, p. 41)	$0.35 billion
Garner (2001, p. 45)	$0.90 billion
National Council for Air and Stream Improvement $1 billion or less
(2003, p. 5)	
Treatment of Uncertainty and Baseline
One factor affecting cost estimates of the Cluster and MACT II rules is the number of mills that closed after
the introduction of the new regulations. Hence, it is useful to know EPA's ex ante forecast of how many mills
would have closed in the absence of the Cluster and MACT II regulations, and its forecast of the number of
mill closing as a result of the new rules. According to EPA (1997a, p. 3-23), "A baseline closure is a mill that
31 -

-------
fails the salvage value test before the addition of incremental pollution control costs."31 Of the 96 mills
expected to bear incremental costs due to ELGs, the available data allowed closure analyses to be performed
on 94 mills. EPA determined about 9 of these mills would be baseline closures (see U.S. EPA 1997a, p. 3-24).
In addition, EPA projected two mill closures due to the final BAT/PSES and final MACT I. Under all MACT II
options, a third mill closure was expected (see U.S. EPA 1997a, pp. 6-16 and 6-18).
2.3, Information Available to Conduct Ex Post
Evaluation
Data for our ex post assessment come from several sources. We use data acquired from BECA - a consulting
firm - on when 02 delig and extended delig systems were installed and the extent of CIO2 substitution as a
bleach alternative starting in 1997 for mills subject to the BAT provisions of the Cluster Rule. Data on when
air pollution control devices (APCD) were installed are acquired from the 2011 survey for the Risk and
Technology Review (RTR) of the technology-based standards for hazardous air pollutants (HAPs).
For ex post cost estimates, we rely on publicly available data from the National Council for Air and Stream
Improvement, Inc. (NCASI), which produced an annual survey of capital expenditures borne by pulp and
paper industry from 1970 through 2002. ' The survey requested information on each firm's capital
expenditures, including capital expenditures for pollution abatement. The questionnaire also asked firms to
separate their pollution abatement capital expenditures by media (air, water, and solid waste) and by the
type of mill (i.e., integrated or non-integrated).34 Finally, firms divided their pollution abatement capital
expenditures into those (1) for "sole-purpose" equipment (e.g., new secondary clarifier) and (2) incremental
pollution abatement costs for equipment that would have been purchased in the absence of environmental
regulations (e.g., incremental cost of kraft recovery furnace electrostatic precipitator upgrade that increases
particulate capture efficiency from 90 to 99.5 percent).
31	According to EPA (1997a, p. 3-21), "A facility is projected to close if the salvage value exceeds the present value
of future earnings after increased pollution control costs."
32	We use data from the NCASI survey because only 1 mill reported compliance cost data on EPA's FY2011 survey
(see Nicholson et al. 2012, p. 1). This survey included the MACT Subpart S Risk & technology Review (RTR), the
MACT Subpart MM RTR, and the Kraft Pulp Mill NSPS (Subpart BB) Review. These reviews are required by the
Clean Air Act as part of the process of regulating emissions of HAPs.
33	Another potential source of data is the annual Pollution Abatement Costs and Expenditures (PACE) survey (U.S.
Department of Commerce, various years). The PACE survey collects facility-level data on pollution abatement
capital expenditures and operating costs associated with compliance to local, state, and federal regulations and
voluntary or market-driven pollution abatement activities. Because the PACE Survey was discontinued in 1994 and
was only conducted in two subsequent years (1999 and 2005), it cannot be used for the ex post portion of our
analysis.
34	An integrated mill produces at least 20 percent of its total pulp consumption from on-site wood pulping
operations (see NCASI 2003, p. 1).
32 -

-------
The 1998 to 2002 NCASI surveys collected information from firms that accounted for 84 to 94 percent of
wood pulping capacity and 68 to 79 percent of paper and paperboard capacity. From 1973 to 1986, the
NCASI survey found pollution abatement capital expenditures values for air, water, and solid waste pollution
abatement were approximately 4 percent higher than the PACE values for SIC 26 (Paper and Allied Products).
However, its values for 1988 to 1994 were approximately 15 percent higher than PACE. Unlike the PACE
survey, which assigned values for missing observations to be able to produce national estimates of pollution
abatement costs, NCASI treated missing observations as zero costs. Table 2.5 shows the NCASI pollution
abatement capital expenditure data for 1990-2002.
Cost information on MACT II and the implementation of a PM bubble strategy was provided byAbt
Associates / RTI International. These sources are supplemented with firm-level data found in the U.S.
Securities and Exchange Commission (SEC) 10-K form, which provides some firm-level data for ex ante and ex
post costs of Cluster Rule compliance, and data on mill closures during the implementation of the Cluster
Rule and MACT II Rule. The SEC 10-K information on mill closures is augmented by the U.S. EPA (2001b,
Appendix B and 2006, Appendix), USDA (2005), the Pulp & Paper North American Fact Book (Miller Freemen
Publications 1998, and Paperloop.com 2000, 2002, and 2003) and internet searches.
2.4. Ex Post Assessment of Compliance Costs
2,4,1, Regulated Universe
According to EPA (1997a, p. 2-1), of the 158 mills that used kraft, soda, sulfite, or semi-chemical processes at
the time of the ex post analysis, 155 were expected to incur pollution abatement costs as a result of MACT I
and MACT III. In addition, 96 of these mills would incur additional abatement costs as a result of the new
35 The only other source of data was Selected Air Pollution Control Equipment (see U.S. Department of Commerce,
2000). This survey provided data on expenditures for particulate emissions collectors by selected industries
including pulp and paper and pulp mill operations. Unfortunately, expenditures on (1) gaseous emissions collectors
and (2) other types of industrial air pollution control devices were withheld to avoid disclosing data for individual
companies. These data show a 41 percent increase in 1998 expenditures on particulate emissions collectors
relative to 1997. Unfortunately, the survey was discontinued after the 1998 survey.
33 -

-------
Table 2.5. Pollution Abatement Capital Expenditures (NCASI)
(millions of 1995 dollars)
Year
Water
Air
Solid
Waste
Total
Percent of Total Capital
Expenditures
1990
669
553
272
1,494
12
1991
765
542
214
1,521
19
1992
533
416
201
1,150
18
1993
354
289
131
774
14
1994
289
252
188
729
14
1995
309
219
97
625
12
1996
343
244
133
720
13
1997
305
142
105
552
12
1998
288
119
172
579
13
1999
340
294
65
699
17
2000
364
633
74
1,071
23
2001
170
287
72
529
12
2002
105
170
29
304
9
Note: current dollar value values are deflated to 1995 dollar values using the Engineering News
-Record Construction Cost Index (NCASI 2003, pp. A2-A3).
ELGs and pretreatment standards. This constituted the basis of the industry size when ex ante cost
estimates of the Cluster Rule and MACT II Rule were generated. By 2001, EPA (2001b, Appendix B) estimated
133 mills would be subject to the MACT II emission standards.36
2,4,2, Baseline
It has been argued that some mills undertook pollution abatement actions in anticipation of the Cluster Rule.
The 1993 proposal used a 1992 baseline (see U.S. EPA 1997, p. 8-24), which was updated to mid-1995 for the
final rule. After the rule was proposed in 1993, "... a number of pulp mill owners and operators announced
plans to install new technologies at their facilities ...' (see U.S. EPA 1997b, 10-16). Some mills addressed
concerns about dioxin releases by installing extended delignification or 02 delig systems (see U.S. EPA 1993b,
pp. 4-5 to 4-7 and 4-12). Figure 2.1 shows the number of mills that installed their first 02 delig systems
36 As of 2004 (see U.S. EPA 2006, Appendix), 77 of the 96 mills subject to the new ELGs and pretreatment
standards reported bleached chemical pulp operations.
34-

-------
Figure 2.1. Number of Mills Installing 02 Delig for First Time, by Year
59









Z4
1






43
No 02 Delig Pre-1995 1995-1997 1998-2001 2002-2008
Year
Source: BECA (2013b)
during selected time periods. It can be seen that over half of the mills that installed 02 delig did so by 1993.37
Only 4 mills installed 02 delig during 1995-1997, the years prior to 1998, the year the rule was promulgated.
This trend was anticipated by Johnson (1995) when he observed the growth of 02 delig installations
stagnated in North America during 1993-1994 and few new systems were anticipated prior to 1997. In
addition to poor industry profitability, Johnson believed "... a strong industry stand that oxygen
delignification is not a required strategy to meet Cluster Rule objectives" was the other reason for the lack of
growth in 02 delig installations. Johnson concluded the "... industry position that ECF (full substitution)
bleaching alone will accomplish these objectives and the capital expenditures this avoids has dramatically
reduced the motivation for employing oxygen delignification."
Unlike 02 delig systems, where we have a complete inventory of installed systems at mills subject to the ELG
provisions of the Cluster Rule, lack of data on extended delig systems precludes developing a complete
inventory of installed extended delig systems. Nevertheless, EPA (1993b, pp. 4-5 and 4-6) provided a list of
installed extended delig systems through 1994. In addition, BECA (2013b) provides a partial list of extended
37 U.S. EPA (1997b, p. 10-30) provides additional information on changes in mill use of 02 delig and extended
cooking between the proposal and final (mid-1995) Cluster Rule.
35 -

-------
Figure 2.2. Minimum Number of Mills Installing Extended Delig for First Time, by Year
No Extended Pre-1995 1995-1997 1998-2001 2002-2008
Source: BECA (2013b)
delig systems installed through 2013. By combining the two sources, we compiled a complete list of mills
that were subject to the ELG provisions of the Cluster Rule and installed extended delig systems prior to
1995. In addition, BECA provides the minimum number of mills that installed extended delig systems starting
in 1995. It is worth noting that the last installation of an extended delig system on the BECA list occurred in
2003. Remembering the post-1994 information on extended delig systems is incomplete, Figure 2.2 shows a
dramatic decline in the installation of extended delig systems after 1997. While not included in Figure 2.2,
the Valdosta (GA) mill owned by Packaging Corp, the Jacksonville (FL) mill owned by Jefferson Smurfit, and
the Savannah (GA) mill owned by Union Camp were not subject to the ELG provisions of the Cluster Rule, yet
choose to install extended delig systems. This coincides with our finding that several mills not subject to the
ELG provisions of the Cluster Rule installed 02 delig systems.
38 EPA included three mills subject to the Cluster Rule - Alabama Pine Pulp mill in Clairborne (AL), Port Wentworth
(GA), and Quinnesec (Ml) - that were not on the BECA list, while BECA included the Mobile (AL) mill owned by
Kimberly Clark that was not on the EPA list.
36-

-------
The first survey of CIO2 substitution by U.S. pulp and paper mills was the 104 Mill Study conducted by NCASI
and the U.S. EPA (1990, pp. 8-10). Data was collected for 165 lines at 86 kraft mills in 1988. Of the 165 lines,
59 used no CIO2 substitution. Of the lines employing CIO2, 99 lines used between 0 and 30 percent, 4 used
between 30 and 50 percent, 2 used between 50 and 70 percent, and 1 used more than 70 percent. In
addition, of the 18 lines at 16 sulfite mills only one used CIO2 - at a rate of less than 5 percent. CIO2
substitution increased rapidly in the following years. According to the U.S. EPA (1997b, p. 10-30), in 1992
(baseline of the Cluster Rule proposal) 6.6 percent of bleached papergrade kraft and soda mill production as
39
total ECF. By 1994, approximately 22 percent of all bleached chemical production was ECF (AET, 2002).
This increased to 33.2 percent of bleached papergrade kraft and soda mill production in mid-1995 (see U.S.
EPA 1997b, p. 10-30).
While Table 2.5 shows higher pollution abatement expenditures during 1990-1994, we cannot determine
whether this reflected pollution abatement undertaken in anticipation of the Cluster Rule and MACT II Rule
or was a reaction to local concerns about the undesirable by-products generated by pulp and paper mills.
Table 2.1 shows anticipated reductions in releases of key air and water pollutants as a result of the Cluster
Rule and MACT II Rule. This is in addition to a substantial decline in releases of dioxins (TCDD) and furans
(TCDF) between the proposal (1992 baseline) and the final rule (mid-1995 baseline). The baseline releases of
TCDD declined from 70 grams per year in 1992 to 16 grams per year in mid-1995, while TCDF declined from
341 grams per year in 1992 to 122 grams to year in mid-1995 (see U.S. EPA 1997a, p. 8-24 - there are slight
discrepancies between these mid-1995 values and those reported in Table 2.1). However, it has been
suggested the pulp and paper industry abstained from aggressive abatement efforts until the Cluster rule was
finalized (Ferguson, 1995). Ferguson's hypothesis was supported by Maynard and Shortle (2001), who used a
double hurdle model and found the uncertainty associated with an irreversible investment (i.e., installing 02
delig, extended delig, or ECF) resulted in a value of waiting that led some bleached kraft mills to delay their
investment in cleaner technologies. In addition, Maynard and Shortle found "public pressure" variables were
statistically significant in explaining the adoption of cleaner technologies.
2,4,3, Metho	npliance
Under the Cluster Rule, BAT involves switching to elemental chlorine free (ECF) or total chlorine free (TCF)
bleaching. The data in Table 2.6 show that from 1990 to 2001 there was a substantial switch to ECF
bleaching. Both Figure 2.3 and Table 2.6 reveal that approximately half the switch to ECF occurred prior to
1998, which is the first year the Cluster Rule was implemented for some mills. Among the mills covered by
the water provisions of the Cluster Rule, only the Samoa (CA) mill opted for TCF bleaching.
39 The Paper Task Force (1994, p. 5) found 22 percent of bleached chemical production in 1994 was traditional,
enhanced, or ozone ECF. Johnson (1994) reported that in 1994 between 20 and 25 percent of U.S. mills had no
CIO2 substitution, while 10 to 15 percent of U.S. mills had 100 percent CIO2 substitution.
37 -

-------
Table 2.6. United States Bleached Chemical Pulp Production
(millions of tones; 1 tonne = metric ton = 1000 kg = 2204.62 lb)
Year
ECF
TCF
Other
1990
0.5
0.0
26.8
1991
1.6
0.0
25.6
1992
2.8
0.0
24.4
1993
4.0
0.2
23.0
1994
6.0
0.2
21.0
1995
9.1
0.3
17.9
1996
10.4
0.2
16.6
1997
13.3
0.2
13.8
1998
15.5
0.2
11.4
1999
18.1
0.2
8.9
2000
20.7
0.2
6.3
2001
25.9
0.1
0.9
Source: Alliance for Environmental Technology (2002)
Figure 2.3. Percent CIO2 Substitution (1997-2005)
120
97 95 95
¦ Percent CI02
Substitution
	1	1	1	1	1	1	1	1	1
1997 1998 1999 2000 2001 2002 2003 2004 2005 -
Year -
Source: BECA (2013a)
38 -

-------
Starting with 1997, BECA (2013a) provided us with information on the percent of CIO2 substitution used on
lines at mills subject to the ELGs of the Cluster Rule. Weighting the percent of CIO2 substitution by the
production of each line allows us to construct a weighted average of CIO2 substitution for each year. It should
be noted that during 1997 to 2005, the number of active mills subject to the ELGs of the Cluster Rule declined
from 95 to 76.40 Figure 2.2 shows the weighted average of CIO2 substitution for active mills increased from 55
percent in 1997 to 99 percent in 2005. In order to observe the variation in CIO2 substitution among mills,
Figure 2.4 reports the percentage of active mills that fall in various ranges of CIO2 substitution. While half of
active mills subject to ELGs undertook at least 50 percent CIO2 substitution in 1997, only 28 percent
undertook 100 percent CIO2 substitution. By 2000, 91 percent of active mills had at least 50 percent CIO2
substitution, while 67 percent reported 100 percent CIO2 substitution. In 2002, 90 percent of active mills had
100 percent CIO2 substitution, and 95 percent of mills had 100 percent CIO2 substitution in 2005. Although
Franklin (VA) reported 100 percent CIO2 substitution in 2002, it along with two other mills that participated in
VATIP - Spring Grove (PA), Catawba (SC), and Franklin (VA) - did not permanently convert to 100 percent CIO2
substitution until 2005. On the other extreme, 20 percent of the mills undertook no CIO2 substitution in
1997, which declined to 5 percent in 2005.41
In response to EPA's 2011 technology review survey (Spence and Bradfield 2011, p. 3), which included mills
not subject to ELGs, EPA found that in 2009 "...98 facilities reported pulp bleaching with 164 bleaching lines.
Elemental chlorine free processing was used in 104 bleaching lines, while TCF was used in 31 lines, and
processed chlorine free (PCF) was used in 22 lines. The remaining 7 lines utilized peroxide, sodium sulfate,
hypochlorite, chlorine, or a combination of these bleaching chemicals. Oxygen delignification was utilized on
42 of the ECF bleaching lines to reduce emissions and bleaching chemical cost and consumption."
Two previous studies examined the effect of "chlorine" regulations on technological innovation. Snyder, et al.
(2003) conducted an econometric analysis of the effects of the Cluster Rule on the diffusion of technological
change in the chlorine manufacturing industry. Using plant-level data, their study focused on the diffusion of
a new, cleaner production process within the chlorine industry. Snyder, et al.'s results indicate that chlorine
facilities affected by the reduction in demand for chlorine resulting from the Cluster Rule (and the Montreal
Protocol) were more likely to close than were other facilities. This factor along with the adoption of new
technology at existing plants led to an increase in the share of chlorine plants employing a cleaner production
technology. Popp and Hafner (2008) used information on regulations affecting dioxins and patents from
40	For example, in 1997 information on CIO2 substitution is unavailable for 2 of the original 96 mills - (1) the
Longview Fiber (WA) mill, which curtailed chlorine-based bleaching in March 1994 (see U.S. EPA 1997b, p. 4-5),
was not included and (2) no production was reported for the Peshitgo (Wl) mill. In 1998, the Samoa (CA) mill was
added to the list of mills with no reported production.
41	In 2005, the four mills that did not report 100 percent CIO2 substitution undertook no CIO2 substitution. These
mills were Somoa (CA), and three mills in Wisconsin: Park Falls, Port Edwards, and Rothschild. Because Somoa
(CA) employed TCF bleaching, CIO2 was not required. Park Falls, Port Edwards, and Rothschild were Segment B
papergrade sulfite mills and not required to monitor dioxin under the Cluster Rule (see U.S. EPA 2006, pp. 9-10 to
9-11).
39 -

-------
Figure 2.4. Extent of CIO2 substitution, by percent of mills (1997-2005)
(D
>
u
<
£
CD
U
&_
CD
Q.
¦	% of mills = 0% -
% of mills >= 50%
¦	% of mills = 100%
	1	1	1	1	1	1	1	1	1
1997 1998 1999 2000 2001 2002 2003 2004 2005-
Year ¦
Source: BECA (2013a)
Canada, Finland, Japan, Sweden, and the United States, to investigate the association between regulations
and patent activity. They found "substantial innovation" to reduce chlorine use in the bleaching technology
occurred as a response to the implementation of environmental regulations.
Summarizing the technology employed to meet the air provisions of the Cluster Rule is more difficult than
summarizing the technology used to meet the water provisions. The 2011 technology review survey (Hanks
et al. 2013) collected information on air pollution control devices (APCDs) installed at 98 kraft mills in 2009.
Of these mills, 67 were subject to both the air and water provisions of the Cluster Rule. Most mills reported
multiple emission units (i.e., sources of emissions) and multiple APCDs, sometimes more than one APCD for
an emission unit. Hence, summarizing when these devices were installed is challenging. In this paper, we
focus on the last year a mill installed or updated an APCD. These results are summarized in Figure 2.5.
According to the survey, only one mill reported no installed APCD. For the years prior to the Cluster Rule, 40
mills report their last installation/update prior to 1995, while 14 mills reported their last installation/update
during 1995-97. Thirteen mills reported their last installation/update during 1998-2001, which covers the
period for implementing the Cluster Rule. Finally, 29 mills reported their last installation/update during 2002-
40-

-------
Figure 2.5. Number of Mills, by Year, of Last Installed or Updated APCD
50
AD
40

10



P* jU
<4—
o
1—

-------
Cluster Rule. Since the share of the abatement capital expenditures assigned to the Cluster Rule depends
upon the baseline, we construct three baseline scenarios.
EPA established a mid-1995 baseline for its economic analysis of the Cluster Rule and MACT II Rule (U.S. EPA
1997a, p. 4-1). Because we want to avoid the possibility of selecting an arbitrary base year in which capital
costs may be unusually high (low) which will result in underestimating (overestimating) ex post costs, we use
the average capital expenditures for air and water pollution abatement between 1995 and 1997 as our
preferred baseline. Since no additional regulations were promulgated on the pulp and paper industry
between 1995 and 2001, we assume all increases in air and/or water pollution abatement capital
expenditures during 1998 to 2001 relative to the 1995-1997 baseline costs reflect the incremental capital
costs of the Cluster Rule.43
During 1998 to 2001, the time between the promulgation of the Cluster Rule and its compliance date, capital
expenditures for air and water pollution abatement were $2.5 billion (in 1995 dollars). Our preferred baseline
yields an estimate of $65 million in Cluster Rule water pollution abatement capital costs and $610 million in
Cluster Rule air pollution abatement capital costs during 1998 to 2001 (all values in constant 1995 dollars).
This ex post Cluster Rule capital cost estimate of $675 million is 55 percent lower than ex ante capital cost
estimate of $1.54 billion. We investigate the sensitivity of our results to the baseline year by repeating the
analysis using 1996 and 1997 pollution abatement capital expenditures as alternate baselines.44 45 Using
1996 and 1997 as the baseline yields ex post Cluster Rule capital expenditure estimates of $503 million and
$882 million respectively, which are 67 percent and 43 percent lower than the EPA ex ante capital
expenditure estimate.46
One important caveat is that while most of the compliance dates for the Cluster Rule occurred on or before
April 15, 2001, compliance for two MACT provisions: bleaching systems in the voluntary advanced technology
incentives program (VATIP) (of which only 4 mills (see U.S. EPA 2006, p. 9-7) participated)47 and the HVLC
system compliance, were not required until April 15, 2004 and April 17, 2006, respectively (see U.S. EPA
1998b, p. 47). While we would prefer to include these MACT provisions in our analysis, the NCASI survey
43	For cases when capital expenditures during 1998-2002 were less than the baseline capital expenditures, we
assume no capital costs are associated with the Cluster Rule (i.e., ex post costs are nil).
44	1996 and 1997 are selected as baseline years because they are both prior to the promulgation of the Cluster
Rule. NCASI (see Paperloop.com 2003, p. 85) anticipated the pulp and paper industry would experience its highest
levels of capital expenditures associated with the Cluster Rule in 1999 and 2000.
45	Our results could also be sensitive to which mills are included in the NCASI survey, but since we have no access
to the underlying micro-data we cannot test this sensitivity.
46	NCASI estimates of air and water pollution abatement capital expenditures in 1993 and 1994 (in 1995 dollars)
are slightly higher than the 1996 value. Hence, if we include expenditures from 1993 and 1994 in the baseline this
will lead to a lower ex post cost estimate of the Cluster Rule.
47	The four mills participating in VATIP were Eastover (SC), Catawba (SC), Spring Grove (PA), and Franklin (VA).
Other over-complying mills were Oglethorpe (GA) which participated in the XL program and Samoa (CA) which
employed TCF bleaching.
42 -

-------
stopped in 2002. Unfortunately, we do not have any ex post cost estimates of these two MACT provisions to
adjust our ex post cost estimates. Therefore, our ex post cost estimate is biased downwards resulting in EPA's
ex ante cost estimate appearing to be more of an over-estimate than we found.
2,4,3,2, Compliance Costs for MACT II
In order to meet the HAP metals standards of MACT II, approximately 32 pulp and paper mills employed a
"PM bubble compliance alternative" strategy, which uses PM as a proxy for HAP metals (Nicholson et al.
AO
2012, p. 15) . The "PM bubble compliance alternative" gives mills the flexibility to set site-specific PM
emissions limits for each existing source in the chemical recovery area (i.e., recovery furnaces, smelt
dissolving tanks, and lime kilns), as long as the total emissions from all the existing sources are less than or
equal to the total of the promulgated emissions rates for each existing source.49 This improvement in the
efficiency of pollution abatement resulted in lower ex post pollution abatement costs. Although EPA
anticipated the PM bubble compliance alternative would improve the efficiency of pollution abatement, it
was unable to develop ex ante estimates of cost and emission reductions for this alternative because it could
not determine which mills would take advantage of the alternative or what limits the mills would set. The
limits mills set determined which, if any, of the emission units in the bubble would require upgrading and
which would be unchanged. Table 2.7 provides the EPA ex ante engineering estimates of MACT II, plus
BE&K's ex post engineering estimates of the cost of complying with MACT II.
The EPA ex ante cost estimates are based on projected compliance costs presented in the compliance cost
memorandum for the MACT II rule (Holloway 20 00).50 The ex ante capital expenditure estimate of $231
48	Nicholson et al. 2012 final white paper is available upon request.
49	The mill-specific bubble limit is calculated based on the promulgated emissions standards (referred to in the rule
as reference concentrations or reference emissions rates) for each process unit and mill-specific gas flow rates and
process rates.
50	"The ex-ante costs for the MACT II rulemaking were first developed on a model process unit basis (e.g., model
recovery furnaces, model SDTs, model lime kilns), with applicable control option costs developed for each model
process unit.... These ex-ante model costs were then assigned to the individual process units at each mill in the
NCASI MACT survey database, based on whether the process unit was expected to be impacted under the control
option (i.e., whether or not available emissions data showed the mill to be above the emission limit in the control
option).... The mill-specific ex-ante costs for each process unit type were then averaged, and those average costs
were extrapolated nationwide to determine nationwide ex-ante cost estimates for each process unit type..." (see
Nicholson et al. 2012, p. 4)
43 -

-------
Table 2.7. Ex Ante and Ex Post Cost Estimates for MACT II
(millions of 2001 dollars) -

Total Capital Investment
Total Annual Costs
Ex Ante (EPA, 1997)
$231
$80.6
Ex Post (BE&K)
$188
$24.2
Source: Research Triangle Institute (2012, pp. 15-16)
million (in 2001 dollars) reported in Table 2.7 is less than the ex ante EPA estimate of $258 million (in 1995
dollars) reported in Table 2.2. Because ex post cost information was not available for individual mills, ex post
costs are estimated by combining information on the actual (ex post) compliance methods selected by
individual mills with estimated costs of these compliance strategies from the engineering firm BE&K. Thus,
the ex post cost estimates are derived from ex ante costs provided by BE&K, applied to actual ex post mill-
specific compliance information provided by MACT II mills in their responses to EPA's 2011 RTR survey. These
estimates constitute the best ex post compliance cost data for the MACT II rule.
Despite the limitations of this approach, Table 2.7 shows EPA's ex ante total capital investment cost estimate
was nearly 25 percent higher than the ex post cost estimate. Furthermore, EPA's ex ante total annual cost
estimate was roughly three times higher than the ex post cost estimate. The main reason for the lower ex
post cost is the use of the "PM bubble compliance alternative" strategy, which allowed for much more cost-
effective strategy for abating PM emissions than command-and-control.51 In particular, a significant
percentage of sources subject to MACT II did not require upgrades or replacements of existing air pollution
controls, primarily due to the use of the PM bubble compliance alternative. For example, 19 non-direct
contact evaporator (NDCE) recovery furnaces were expected to upgrade or replace their existing electrostatic
precipitators (ESP) units, but only 5 were actually upgraded or replaced. In addition, of the 29 direct contact
evaporator (DCE) recovery furnaces that expected to upgrade or replace ESP units, only 8 were upgraded or
replaced (see Nicholson et al. 2012, p. 15). This is further evidence that more flexible pollution abatement
strategies lead to substantially lower abatement costs.
SEC 10-K Cluster Rule Capital Expenditure Data
The U.S. Securities and Exchange Commission (SEC) collects financial information on firms via its Form 10-K
(Annual Report Pursuant to Section 13 or 15(d) of the Securities Exchange Act of 1934). Because of the
importance of the Cluster Rule, many firms reported anticipated and actual expenditures associated with the
Cluster Rule on the Form 10-K. Unfortunately, the Cluster Rule was implemented in several phases (e.g.,
April 2001 compliance, VATIP, and HVLC system compliance) and some firms were not specific about which
51 It is possible that regulatory-induced technical change played a role in lowering the cost of the MACT II rule. Mill
and equipment shutdowns and consolidations also played a role.
44-

-------
costs were incurred for the different phases of the Cluster Rule and which were incurred for MACT II. As a
result, the cost estimates reported by some firms on the Form 10-K cannot be assigned with certainty to
either different portions of the Cluster or MACT II rules.
The Cluster Rule cost estimates from the Pulp & Paper North American Fact Book (1999 to 2002) provide an
overview of the SEC 10-K data. While the 1999 Fact Book provides estimates of Cluster Rule costs based on
another source, the 2000 to 2002 Fact Books report data collected from the SEC 10-K forms for 30+ pulp and
paper companies. These data are reported in Table 2.8. Using the SEC 10-K forms is further complicated
when publicly-owned U.S. firms are purchased by foreign firms or by private U.S. companies, neither of which
need to submit 10-K forms.
Table 2.9 provides several examples where the SEC data provide a relatively complete picture of the ex ante
and ex post costs of the Cluster Rule. While the ex ante cost estimates of Boise Cascade, Pope & Talbot, and
Wausau were close to their reported actual ex post costs, the ex ante cost estimates of Gaylord Containers,
Potlatch, Smurfit-Stone, and Temple Inland were substantially higher than their ex post costs. Thus, the
anecdotal evidence on the accuracy of ex ante cost estimates of the Cluster Rule based on the SEC 10-K forms
is a bit mixed - some firms accurately predicted the compliance costs, while others substantially
overestimated them. However, since no firms clearly underestimated their actual costs, based on the firms
that did provide ex ante and ex post costs estimates, the aggregate ex ante cost estimates are higher than
52
the aggregate ex post cost estimates, which is consistent with our findings above.
There are several instances in which firms commented on the costs associated with the Cluster Rule. In its
1999 10-K report, Wausau stated "The Company believes that capital expenditures associated with
compliance with the Cluster Rules and other environmental regulations will not have a material adverse
effect on its competitive position, consolidated financial condition, liquidity, or results of operation." In its
1999 10-K report, Potlatch stated "In early 1998 the Environmental Protection Agency published the 'Cluster
Rule' regulations applicable specifically to the pulp and paper industry ... the company estimates that
compliance will require additional capital expenditures in the range of $20 million to $30 million, the majority
of which will be expended over the next 2 to 3 years. The company does not expect that such compliance
costs will have a material adverse effect on its competitive position." Based on these statements and our
inability to locate any statements in the SEC 10-K forms indicating pulp and paper firms believed the Cluster
Rule had a substantial impact on their profitability or would cause them to close any facilities, it appears
paper firms did not believe the costs of Cluster Rule had a substantial impact on their bottom line.
52 Because firms were not obligated to disclose specific data regarding their capital expenditures associated with
the Cluster Rule, firms such as Rayonier and Kimberley-Clark and Westvaco opted to provide only projected
expenditures. As a result it is not possible to draw any conclusions about ex ante and ex post costs for those firms.
45 -

-------
Table 2.8. Projected Capital Expenditures to comply with Cluster Rule -
	(millions of dollars) -	
Fact Book 1999	Fact Book 2000	Fact Book 2001	Fact Book 2002
Company
Cost
Time Frame
Cost
Time Frame
Cost
Time Frame
Cost
Time Frame
Boise Cascade
100-150
1998-2001
85
2000-2001
32
2001
20.6
2002
powater
60-75

1998-2000
150-200

2000-2004

175

2001-2003
30
| 2002-2003
Alliance Forest
Products1








-
-
45
2000
8.7

2001-2002










Buckeye Technologies
40
1998-2000
40
2000-2005
35
2001-2004
40
2002-2005
Chesapeake
5-6
1998-2000






Consolidated Papers2
25
1998-2001
2.6
2000




Crown Vantage3
40
1998-2005
8
22
2000
1999-2003




Donohue
-
-
52
1999-2001




port James4
100

1998-2001
40

2000-2001







George-Pacific5
300

1998-2000
160

1998-1999










-














550

1998-2006

135

2002-2006
118
| 2002-2006
P.H. Glatfelter
21
1998-1999
30
2000-2004
30
2001-2004
30
2002-2004
International Paper
230

1998-2000
229

2000-2001

116

2000-2001
82
2002

180

2001-2006
150-195

2002-2006

330-370


2003-2006
138
2003











123
2004
Champion Intl.
20-40

1998-2004
25-50

2000-2005







|llnion Camp6
125-150

1998-2001









Kimberly-Clark
279
1998-2000
15
2000
0.4
2001
98
2002







99
2003
Longview Fiber
10-20
1998-2000
10-12
2000-2001





20-30
2001-2005
10-20
2002-2006
15-20
2001-2005
3.6
2002
llVlead
110

1998-2002
55

2000-2006

54

2001-2003



Westvaco
257

1998-1999
100-150

2000-2005

100-150


2001-2004




-

-










MeadWestvaco Corp.7









47
2002
46

-------
Fact Book 1999
Fact Book 2000
Fact Book 2001
Fact Book 2002
Company
Cost
Time Frame
Cost
Time Frame

Cost

Time Frame
Cost
Time Frame











35

2003

Packaging Corp. of
America











-


48
2000-2005

2.1

2001
5.8
2002
















¦



25.7

2001-2005
1

2002-2005


Tenneco Packaging8
105
1998-2008











Pope & Talbot
30-35
1998-2000
27
2000-2001

2.8

2001
3
2002-2006
Potlatch
70-95
1998-2006
15
2000
16

2001
5
2002-2006
Rayonier
35
1998-1999
80
2000-2004
70

2001-2005
30
2002-2005

80
1998-2002








Riverwood Intl.9
55
1998-2005
55
2000-2006

55

2000-2006
55
2000-2006

Schweitzer-Mauduit
Intl.
8-16
1998-1999









Smurfit Stone













200
2000
43

2001




















290

2001-2004

60


2002-2005
100-125

2002-2007


Stone Container
180
1998-2005

180

1998-2005









Jefferson Smurfit10
175
1998-2002












Temple Inland
110
1998-2000

20

2000-2001

27

2001-2003
27

2001-2003


Gaylord Container11
5-7
1998-2000

10

2000














20

2001-2008

26

2001-2007




S.D. Warren (part of
Sappi)
70-112
1998-2000
10-35
2000-2001

10

2001


Wausau-Mosinee
Paper
-
-
20
2000-2001

2.3

2001


Weyerhaeuser
80
1998-2001

87

2000-2003
50

2001-2003
50

2001-2003


Willamette












120
1998-2002

100

2000-2004

115

2001-2005




Industries12

















Total


$1,751-2,600


$1,584-1,679


$1,156

Primary source of
Pulp &

10-K forms


10-K forms


10-K forms

data:
Paper









47 -

-------
Fact Book 1999
Fact Book 2000
Fact Book 2001
Fact Book 2002
Company	Cost	Time Frame	Cost	Time Frame	Cost	Time Frame	Cost	Time Frame
Project
Report
(some firms
report
values for all
environmen
tal
	spending)	
NOTES:
1.	Alliance Forest Products was acquired by Bowater in 2001
2.	Consolidated Papers was acquired by Stora Enso Oyj of Finland in 2000
3.	Crown Vantage declared bankruptcy in 2001
4.	Fort James was acquired by Georgia-Pacific in 2000 (see 2001 Fact Book, p. 40 + p. 41 discusses rules for listing capacity after shutdown)
5.	Georgia-Pacific was acquired by Koch Industries in 2005
6.	International Paper purchased Union Camp in 1999 and Champion International in 2000
7.	Mead merged with Westvaco in 2002
8.	Until 1999, the Packaging Corporation of America was known as Tenneco Packaging, a subsidiary of Tenneco
9.	Riverwood Holding purchased Graphic Packaging in 2003 and combined Graphic Packaging with Riverwood International
10.	Jefferson Smurfit merged with Stone Container to form Smurfit Stone Container in 1998
11.	Gaylord Container was acquired by Temple Inland in 2002
12.	Willamette Industries was acquired by Weyerhaeuser in 2002
NOTE: Shaded entries indicate a merger occurred during 1999-2002 among firms in contiguous shaded area. -
Sources: Miller Freeman Publications (1998, p. 26), Paperloop.com (2000, p. 29), Paperloop.com (2002, p. 33), Paperloop.com (2003, p. 33) -
48 -

-------
Table 2.9. Cluster Rule Capital Expenditure Estimates (in millions of dollars) from SEC 10-Ks for Firms with Complete Data
(millions of dollars)


1998

1999
2000

2001

Company
Cost
Time Frame
Cost
Time Frame
Cost
Time Frame Cost

Time Frame
Boise
Cascade
120
Next 4 years
40
85
Through 1999
(actual)
Next 2 years
(projected)
96
32
Through 2000
(actual)
2001
(projected)
117
Through 2001
(actual)
Gaylord
Containers
22.5
First 3 years -
MACT 1 and III, no
BAT costs
(projected)
10
For April 2001
standards - MACT 1
and III, no BAT costs
(projected)
O no
For April 2001
standards -
MACT 1 and III
(projected), no
BAT costs
Through fiscal
2000 (actual)
10
Through fiscal
2001 - MACT 1
and III, no BAT
costs (actual)
Pope &
Talbot
35
Through first
quarter of 2001
(projected)
35
8.2
Through first
quarter of 2001
(projected)
Through 1999
38.6
Through Nov.
2000-
completed
(actual)


Potlatch
20-30
Next 2-3 years
(projected)
15
2000 (projected)
12
Total cost of
project (most
spent in 2000)

Phase 1 of
Cluster Rule is
completed
Smurfit-
Stone
310
2-4 years
(projected)
310
Next several years
(projected)
204
179
through 2000
(actual)
2000 (actual)
232
28
Through 2001
(actual)
2001 (actual)
Temple-
Inland
<110
1999 - 2001
(projected)
1
Through 1999
(actual)
11
Through 2000
(actual)
15
Through
December 31,
2001 (actual)
Wausau
16-20
1999-2001
(projected)
20-22
1999-2001
(predicted)
20-22
1999-2001
(projected)
19.1
1999-2001
(actual)
49 -

-------
Mill Closures
One factor contributing to ex ante cost estimates exceeding ex post costs are mill closings or a reduction in
mill capacity through the shutdown of a machine. Obviously, if a mill shuts down instead of complying with
the Cluster Rule this reduces observable ex post costs. We attempt to identify mills affected by the Cluster
Rule that permanently closed between 1997 and 2004 and provide documentation on the reason for the mill
closing. Complicating this task is the fact that a mill can close, then sold, and reopened under new
management. Table 2.10 provides summary statistics on the number of mills that closed between 1971 and
2001.
Table 2.10 shows 26 mills closed in 1998 and 1999, 12 mills closed in 2000, and 23 mills closed in 2001. In
contrast, the Paper, Allied-Industrial, Chemical and Energy Workers International Union (see Paperloop.com
2003, p. 69) claimed 36 paper mills permanently closed in 2001. However, these totals do not list the specific
mills or the rationale for closing. As a result, we cannot determine how many of these mills were subject to
the Cluster Rule nor how many of the closures were permanent.
Jensen (1999, pp. 71-72) discussed claims of mill shutdowns in response to meeting provisions of the Cluster
Rule by April 2001. For example, Kimberley-Clark decided against undertaking expenditures to bring its
Mobile, AL mill into compliance. In addition, the decision by Sappi to close its Westbrook, ME mill was
partially due to pending Cluster Rule expenditures. Finally, Donohue decided against bringing its Champion
mill in Sheldon, TX into compliance with the Cluster Rule. Some of this story was confirmed by Miller and
Freeman's (1998, p. 26) statement that Proctor & Gamble, Kimberly-Clark, and Donohue, Inc. had closed kraft
mills due to the costs of environmental regulations.
We attempted to independently identify the number of closures among the mills that were subject to the
Cluster Rule and MACT II rule. Starting with the list of the 155 mills subject to the Cluster Rule
(http://www.epa.gov/ttn/atw/pulp/mi 11 tah.pdf). we compiled a list of mills that appear to have
permanently closed by 2004. This list was compiled from several sources. First, we identified the mills not
included in the list of 133 mills subject to MACT II (see EPA 2001b, Appendix B), a 2004 list of the status of the
96 mills subject to the ELG component of the Cluster Rule (see U.S. EPA 2006, Appendix), and annual
information on the 96 mills provided by BECA (2013b) . Next, the USDA (2005) provided an inventory of the
status of pulp mills in the 2005. This list was supplemented with information from the 1999 to 2002 editions
of the Pulp & Paper North American Fact Book, and information on mill closures provided on SEC 10-K forms.
53 Hanks et al. (2013, p. 3) reported trends in the number of chemical pulp mills from 1976 to 2011.
50-

-------
Table 2.10. Pulp and Paper Mill Closures 1971-2001



1991-
1991-
1991-
1991-

1971-1980
1981-1990
1997e
1999e
2000e
2001e
Paper
Newsprint
1
0
0
0
1
1
Printing / writing
13
10
7
12
16
25
Packaging / industrial
converting
15
11
2
7
10
15
Tissue
12
18
9
15
15
15
Total paper
41
39
18
34
42
56

Paperboard
Unbleached kraft
1
0
0
4
4
4
Solid bleached
1
0
0
0
0
1
Semichemical
2
2
0
1
1
1
Recycled
26
15
5
10
14
22
Total Paperboard
30
17
5
15
19
28
Total Paper/
Paperboard
71
56
23
49
61
84
e= estimated
Sources: Miller Freeman Publications (1998, p. 32), Paperloop.com (2000, p. 37), Paperloop.com (2002,
p. 41), Paperloop.com (2003, p. 40)
Finally, we sought confirmation of mill closures via searches on the internet. Based on this information, of the
155 mills subject to the Cluster Rule, we identified approximately 18 permanent mill closures through 2004.
We were unable to locate any statements in the SEC 10-K forms filed by pulp and paper firms that linked mill
closures to environmental regulation, let alone the Cluster Rule. In fact, the most common reasons provided
for mill closures were reduced demand for paper products and excess capacity.
The following examples demonstrate some of the additional challenges confronting efforts to identify mill
closures. First, there were numerous mill sales and firm mergers during the years the rules were
implemented. For example, Scott Paper Co. operated a pulp and mill in Mobile, AL. In December 1994 Scott
sold the paper mill to South African Pulp and Paper Inc. (SAPPI), while its pulp mill was sold to Kimberly-Clark
in 1995. The Scott Paper Co/SAPPI mill was subject to both the air and water provisions of the Cluster Rule.
Kimberly-Clark closed the pulp mill in 1999, and then purchased the paper mill from SAPPI in 2002. Second,
the U.S. EPA (2006, p. 7-7 and Appendix) identified the status of the mills subject to the ELG provisions of the
Cluster Rule when it was promulgated in 1998. While some mills were either temporarily or permanently idle,
6 were operating but no longer classified as BPK or PS mills. Instead, they were classified as non-bleached
chemical mills (e.g., unbleached kraft) and no longer subject to the Cluster Rule.
How did mill closings affect the aggregate ex post costs of complying with the Cluster Rule? Since we do not
have mill specific ex post cost data we cannot provide a precise answer to this question. We use the number
51 -

-------
of mill closures to estimate their effect on ex post costs. The observed number of mill closures (18)
represents an upper limit on the number of mill closures associated with the Cluster Rule.54 Deriving the
number of mill closures due to the Cluster Rule requires subtracting the 9 baseline mill closures from the
observed number of mill closures after implementing the Cluster Rule. These 15 mills - the 9 permanently
closed plus the 6 no longer using bleached chemical processes - represent approximately 10 percent of the
mills affected by the Cluster and MACT II Rules. If we assume that they are typical mills and we increase our
ex post cost estimate by 10 percent we find that EPA over-estimated the costs of the Cluster and MACT II
Rules by 1.5 to 2.5 times depending on the baseline. Based on this, we conclude mill closures did not account
for EPA's over-estimating the costs of the Cluster and MACT II Rules.55
2.5. Conclusions
Our findings suggest EPA's ex ante cost estimates overstated the costs of both the Cluster Rule and the MACT
II rule. Using publicly available data from NCASI, we found that EPA overestimated the capital cost of the
Cluster Rule by 30 to 100 percent, depending on the choice of baseline year from which we derived the
incremental cost. Among the reasons for EPA's overestimates of these capital costs are the mills' use of the
clean condensate alternative (CCA), flexible compliance options, extended compliance schedules, site-
specific rules, use of equivalent-by-permit, and equipment/mill shutdowns and consolidations.56 However,
the lack of detail in the available data means we can only speculate on which reason(s) is primarily
responsible for EPA's overestimate.
Furthermore, our findings show that EPA also overstated the compliance costs of the MACT II rule.
Specifically, EPA overestimated the capital cost by approximately 25 percent and overestimated the annual
cost by 200+ percent. It appears the primary reason for the lower ex post cost is the use of the "PM bubble
compliance alternative" strategy, which is a more efficient policy to abate the same level of PM emissions
and required fewer mills to upgrade or install new pollution abatement equipment than anticipated by EPA.
Anecdotal evidence of the realized costs of the Cluster Rule provided by the SEC Form 10-K is a bit mixed with
some firms accurately predicting their compliance costs, while others substantially overestimating their
actual costs. Because no firm dramatically understated its realized costs, the aggregate ex ante costs are
likely higher than the aggregate ex post costs. While equipment/mill shutdowns and consolidations also
played a role, they are not enough to account for EPA's over-estimate of the actual costs of compliance.
54	Appendix 2.1 lists the 18 mill closures, plus 6 mills no longer using bleached chemical processes.
55	U.S. EPA (http://www.epa.gov/ttn/atw/pulp/milltab.pdf) lists the 155 chemical (kraft, soda, sulfite, standalone
semi-chemical) pulp and paper mills in the United States initially subject to the Cluster Rule, while the U.S. EPA
(2001b, Appendix B) lists the 133 chemical mills subject to MACT II.
56	Bradfield and Spence (2011) provide information on the adoption of the clean condensate alternative, which is a
pollution prevention option available to kraft HVLC systems that allows the control of HAP emissions without
resorting to controlling HVLC system vent streams via combustion devices.
52 -

-------
Defining the baseline remains a challenge for assessing not only the ex post costs of the Cluster Rule and
MACT II Rule , but the ex post analysis of the costs of any regulation. The baseline determines which pollution
abatement expenditures are considered a direct consequence of a regulation and which expenditures would
have been incurred in a counter-factual world without the regulation. When determining the cost associated
with the final Cluster rule, the U.S. EPA (1997b, p. 10-16):
"excluded the incurred costs of process changes that were already implemented as of mid-1995 in
the cost estimates used to analyze the economic achievability of the rules. However, EPA included
the costs of the announced process changes not underway as of July 1, 1995 in the cost estimates
used to analyze the economic achievability of the rule. Although EPA included the costs of the
process changes announced but not yet underway as of mid-1995 in its final cost estimates, EPA
nevertheless evaluated the impact of these costs in an alternative analysis reflecting announced
corporate commitments that were not underway as of mid-1995."
The 1995-97 period, which serves as the baseline for our ex post analysis, represents a lull between 1990-94
period when discussions about the Cluster Rule and MACT II Rule were initiated and 1998-2003 period when
the rules were implemented. If some (or all) of the increase in pollution abatement expenditures during
1990-1994 can be attributed to actions taken in anticipation of the Cluster Rule, it does not invalidate the
findings of this paper. While including expenditures from 1990-1994 would increase the total cost of the
Cluster Rule and MACT II Rule, the objective of the paper was to compare our ex post estimate of the Cluster
Rule and MACT II rule with the ex ante cost estimates that were derived using a mid-1995 baseline. As a
result, because pre-1995 pollution abatement expenditures related to the Cluster Rule and MACT II Rule
were excluded from the ex ante cost estimate, consistency requires they be excluded from our ex post cost
estimates.
While our findings do suggests that EPA overestimated the cost of both the Cluster and MACT II Rules, we
encounter several issues that limit the accuracy of our conclusions: 1) for the Cluster Rule, we only have
access to industry level data, so our results are somewhat sensitive to how we construct the baseline and the
exact mills included in this data; 2) for the Cluster Rule, we have no annual ex post pollution abatement
operating cost data, which means conclusions on ex post compliance costs are limited to capital costs ; 3)
for MACT II, the only industry compliance expert who could provide us with ex post cost information also
supported the ex ante cost analysis for the rule and we could not independently verify the accuracy of the
data; and 4) for MACT II, the ex post cost data was estimated by RTI, the contractor that supported the ex
ante analysis, using a combination of ex ante engineering cost data developed by BE&K based on experience
57 Although we did not use the PACE data to determine the capital costs associated with the Cluster and MACT II
rules, we can use the PACE data on O&M costs in 1994 and 2005 to provide an estimate of how pollution
abatement O&M costs were affected by the Cluster and MACT II rules. Assuming 1994 is representative of the
baseline (pre-Cluster and MACT II rules) and 2005 represents the period of compliance (post-Cluster and MACT II
rules), we calculate the ratio of 2005 to 1994 pollution abatement O&M costs for both air and water (in constant
dollars). The ratio for air is approximately 0.8 and the ratio for water is around 0.7. The decline in O&M costs
suggests O&M costs are not a significant component of the Cluster and MACT II rule compliance costs.
53 -

-------
of similar projects in the pulp and paper industry and the actual (ex post) compliance methods chosen by the
mills.
Chapter 2 References
Alliance for Environmental Technology (2002), "Trends in World Bleached Pulp Production: 1990 - 2001,"
www.aet.org/reports/market/trends90-01.html (May).
Barton, Douglas, John Pinkerton, Rob Kaufman, Mike Jones, David Forbes, and Georgeann Johnson (1995),
"Cluster Rule Would Place Unrealistic Demand on Engineering Resources, Pulp & Paper, 69, No. 13
(December), 103-108.
BECA (2013a), "Annual US pulp line MTPY" spreadsheet, Tony Johnson, ed.
BECA (2013b), "Pulp Mill Data - Delig Study" spreadsheet, Tony Johnson, ed.
Bradfield, John and Kelley Spence (2011), "Summary of Clean Condensate Alternative Technology Review,"
Memorandum to Docket EPA-HQ-OAR-2007-0544-0129, October 20.
"EPA Asked to Reward 02 Delig Users" (1993), Pulp & Paper, 70, No. 6 (June), 21.
Ferguson, Kelly (1995), "Stuck in Environmental Limbo," Pulp & Paper, 69, No. 8 (August), 9.
Garner, Jerry (2001), "Air Emission Control Regulations Pose New Challenges for Mills," Pulp & Paper, 75, No.
10 (October), 44-46.
Hanks, Katie, Tom Holloway, and Corey Gooden (2013), "Projections of the Number of New, Modified and
Reconstructed Emissions Units for the Kraft Pulp Mill NSPS," Memorandum to Kelly Spence, U.S. EPA,
from Katie Hanks, Tom Holloway, and Corey Gooden, RTI International, February 4.
Holloway, Thomas (2000), "Revised Nationwide Costs, Environmental Impacts, and Cost Effectiveness of
Regulatory Alternatives for Kraft, Soda, Sulfite, and Semichemical Combustion Sources,"
Memorandum to Project file from Thomas Holloway, Midwest Research Institute, EPA Docket A-94-
67, Item IV-B-12, October 26.
Jensen, Karl P. (1999), "U.S. Bleached Pulp Mills Move towards Compliance of Phase 1 of Cluster Rule," Pulp &
Paper, 73, No. 9 (September), 71-75.
Johnson, Tony (1994), "Chlorine Dioxide Useage - A Status Report on the Look of the North American
Bleaching Scene," TAPPI1994 Pulping Conference Proceedings, Book 1, pp. 769-771.
Johnson, Tony (1995), "02 Delig System Update," Pulp & Paper, 69, No. 2 (February), 41.
Maynard, Leigh J. and James S. Shortle (2001), "Determinants of Cleaner Technology Investments in the U.S.
Bleached Kraft Pulp Industry," Land Economics, 71, No. 4 (November), 561-576.
54-

-------
Miller Freeman Publications, Inc. (1998), Pulp & Paper North American Fact Book 1999, San Francisco, CA:
Miller Freeman Publications, Inc.
National Council for Air and Stream Improvement, Inc. (1999), "A Survey of Pulp and Paper Industry
Environmental Protection Expenditures - 1998," Special Report No. 99-05, Research Triangle Park:
National Council for Air and Stream Improvement, Inc.
National Council for Air and Stream Improvement, Inc. (2002a), "A Survey of Pulp and Paper Industry
Environmental Protection Expenditures - 1999," Special Report No. 02-01, Research Triangle Park:
National Council for Air and Stream Improvement, Inc.
National Council for Air and Stream Improvement, Inc. (2002b), "A Survey of Pulp and Paper Industry
Environmental Protection Expenditures - 2000," Special Report No. 02-02, Research Triangle Park:
National Council for Air and Stream Improvement, Inc.
National Council for Air and Stream Improvement, Inc. (2002c), "A Survey of Pulp and Paper Industry
Environmental Protection Expenditures - 2001," Special Report No. 02-07, Research Triangle Park:
National Council for Air and Stream Improvement, Inc.
National Council for Air and Stream Improvement, Inc. (2003), "A Survey of Pulp and Paper Industry
Environmental Protection Expenditures - 2002, Special Report No. 03-07, Research Triangle Park:
National Council for Air and Stream Improvement, Inc.
Nicholson, Rebecca, Tom Holloway, and Corey Gooden (2012), "Final White Paper," Memorandum to Anna
Belova, Abt Associates, from Rebecca Nicholson, Tom Holloway, and Corey Gooden , RTI
international, February 28.
Paperloop.com, Inc. (2000), Pulp & Paper North American Fact Book 2000, San Francisco, CA: Paperloop.com,
Inc.
Paperloop.com, Inc. (2002), Pulp & Paper North American Fact Book 2001, San Francisco, CA: Paperloop.com,
Inc.
Paperloop.com, Inc. (2003), Pulp & Paper North American Fact Book 2002, San Francisco, CA: Paperloop.com,
Inc.
Paper Task Force (1995), "Economics of Kraft Pulping and Bleaching," Working Paper No. 7 (19 December), p.
5 (http://c.environmentalpaper.org/documents/1628 WP7.pdf)
Parthasarathy, Perry and Steve Dowd (2000), "Impact of the Cluster Rule on the Cost Competitiveness of U.S.
Papermaking Industry in the Global Market," TAPPI Journal, 83, No. 9 (September), 39-45.
http://www.tappi.org/Downloads/unsorted/UNTITLED—00Sep39pdf.aspx
Pasurka, Carl (2008), "Perspectives on Pollution Abatement and Competitiveness: Theory, Data, and
Analyses," Review of Environmental Economics and Policy, 2, No. 2 (Summer), 194-218.
Pauksta, Patricia M. (1995), "The Cluster Rule: What's at Stake for the Industry?" Tappi Journal, 78, No. 9
(September), 50-51.
55 -

-------
Popp, David and Tamara Hafner (2008), "Policy versus Consumer Pressure: Innovation and Diffusion of
Alternative Bleaching Technologies in the Pulp Industry," (Chapter 3), in Environmental Policy,
Technological Innovation and Patents, OECD Studies on Environmental Innovation, pp. 107-138,
OECD Publications: Paris, France.
Powell, Mark R. (1997), "Control of Dioxins (and other Organochlorines) from the Pulp and Paper Industry
under the Clean Water Act and Lead in Soil at Superfund Mining Sites: Two Case Studies in EPA's Use
of Science," RFF Discussion Paper 97-08.
Shadbegian, Ronald and Wayne Gray (2005), "Pollution Abatement Expenditures and Plant-Level
Productivity: A Production Function Approach," Ecological Economics, 54, 196-208.
Snyder, Lori D., Nolan H. Miller, and Robert N. Stavins (2003), "The Effects of Environmental Regulation on
Technology Diffusion: The Case of Chlorine Manufacturing," American Economic Review, 93, No. 2
(May), 431-435.
Spence, Kelley and John Bradfield (2011), "Summary of Pulp Bleaching Technology Review," Memorandum to
Docket EPA-HQ-OAR-2007-0544-130, November 22.
U.S. -Bureau of the Census (various issues), Pollution Abatement Costs and Expenditures, Current Industrial
Reports (MA200), U.S. Government Printing Office: Washington, DC.
U.S. Department of Commerce, Bureau of the Census (2000), Selected Air Pollution Control Equipment, 1998,
Current Industrial Reports MA333J, U.S. Government Printing Office: Washington, DC.
U.S. Department of Agriculture, Forest Service, Southern Research Station (2005), "U.S. Wood-Using Mill
Locations - 2005," access mill2005p.xls file at http://www.srs.fs.usda.gov/econ/data/mills/
U.S. Environmental Protection Agency (1990), U.S. EPA /Paper Industry Cooperative Dioxin Study "The 104
Mill Study," Summary Report, Office of Water Regulations and Standards, Washington, DC 20460.
U.S. Environmental Protection Agency (1993a), "Effluent Limitations Guidelines, Pretreatment Standards, and
New Source Performance Standards: Pulp, Paper, and Paperboard Category; National Emission
Standards for Hazardous Air Pollutants for Source Category: Pulp and Paper Production," Federal
Register, 58, no. 241 (December 17), 66078-66216.
U.S. Environmental Protection Agency (1993b), Handbook on Pollution Prevention Opportunities for the
Bleached Kraft Pulp and Paper Mills, Office of Research and Development and the Office of
Enforcement, Washington, DC 20460, EPA-600-R-93-098.
U.S. Environmental Protection Agency (1993c), Regulatory Impact Assessment of Proposed Effluent Guidelines
and HESHAPfor the Pulp, Paper, and Paperboard Industry, Office of Water, EPA-821- R-93-020.
U.S. Environmental Protection Agency (1997a), Economic Analysis for the National Emission Standards for
Hazardous Air Pollutants for Source Category: Pulp and Paper Production; Effluent Limitations
Guidelines, Pretreatment Standards, and New Source Performance Standards: Pulp, Paper, and
Paperboard Category-Phase 1, Prepared for U.S. Environmental Protection Agency, Office of Air and
Radiation, Office of Air Quality Planning and Standards, Innovative Strategies and Economics Group,
56-

-------
Research Triangle Park, NC 27711, and Office of Water, Office of Science and Technology,
Engineering and Analysis Division, Washington, DC 20460.
U.S. Environmental Protection Agency (1997b), Supplemental Technical Development Document for Effluent
Limitations Guidelines and Standards for the Pulp, Paper, and Paperboard Category: Subpart B
(Bleached Papergrade Kraft and Soda) and Subpart E (Papergrade Sulfite), EPA-821-R-97-011, Office
of Water, Washington, DC 20460.
U.S. Environmental Protection Agency (1998a), "National Emission Standards for Hazardous Air Pollutants for
Source Category: Pulp and Paper Production; Effluent Limitations Guidelines, Pretreatment
Standards, and New Source Performance Standards: Pulp, Paper, and Paperboard Category," Federal
Register, 63, no. 72 (April 15), 18504-18751.
U.S. Environmental Protection Agency (1998b), Pulp and Paper NESHAP: A Plain English Description, EPA-
456/R-98-008, Office of Air Quality Planning and Standards, Group, Research Triangle Park, NC
27711.
U.S. Environmental Protection Agency (2001a), "National Emission Standards for Hazardous Air Pollutants for
Chemical Recovery Combustion Sources at Kraft, Soda, Sulfite, and Stand-Alone Semichemical Pulp
Mills," Federal Register, 66, no. 9 (January 12), 3180-3203.
U.S. Environmental Protection Agency (2001b), Pulp and Paper Combustion Sources for National Emission
Standards for Hazardous Air Pollutants (NESHAP): A Plain English Description, EPA-456/R-01-003,
Office of Air Quality Planning and Standards, Research Triangle Park, NC 27711.
U.S. Environmental Protection Agency (2006), Final Report: Pulp, Paper, and Paperboard Detailed Study, EPA-
821-R-06-016, Office of Water, Engineering and Analysis Division, Washington, DC 20460.
U.S. Securities and Exchange Commission (various years and firms), "Form 10-K, Annual Report Pursuant to
Section 13 or 15(d) of the Securities Exchange Act of 1934,"
http://www.sec.gov/answers/formlOk.htm
57 -

-------
Appendix 2.1: Closed Mills and Mills No Longer Classified as BPK or PS
Company
Mill
Listed
Status in
Status in
Shutdown Date
BECA
Other Sources

Location
in EPA
(2001)
2002 EPA
(2006)
mill2005p.xls
(USDA-Forest
Service, 2005)

(Status in
2013)

Ketchikan Pulp
Ketchikan,
No
Not listed
Not listed
1997
Closed
htto://en. wikioedia.org/wiki/Ketc
Co (Louisiana-
AK





hikan. Alaska
Pacific)







International
Mobile, AL
Yes
Idle in 2002
Not listed
2000
Demolished
2000 Factbook (p. 416),
Paper




The Printing Papers business
announced the indefinite
closure of the Mobile,
Alabama mill and permanent
closure of the Lock Haven,
Pennsylvania mill. The
announcement was in
conjunction with the
business's plan to realign and
rationalize papermaking
capacity to benefit future
operations." (2000 K-10)

2010Closures spreadsheet; 1999
K-10 MACHINE SHUTDOWN
htto://www. theDinevwoods.com
/iocloses.htm
httD://www. tidewaternews.com/
2009/11/21/other-shuttered-iD-
mills-reDurDosed/
International
Camden, AR
Yes
N/A
Not listed
2000
Not listed
2001 Factbook (p. 40); 2000 K-10
Paper




"The Camden mill, which
produced unbleached kraft
and multi-wall paper, was
closed in December 2000 due
to the declining kraft paper
market, excess internal
capacity and shrinking
customer demand."
(also not
listed in
2011 RTR
survey)
MILL CLOSURE
htto://www. arktimes.com/arkans
as/camden-comeback-
slowed/Content?oid=1217215
httD://articles.latimes. com/2001/
mar/02/business/fi-32251
httD://www. theDinevwoods.com
/iocloses.htm
58 -

-------
Company
Mill
Location
Listed
in EPA
(2001)
Status in
2002 EPA
(2006)
Status in
mill2005p.xls
(USDA-Forest
Service, 2005)
Shutdown Date
BECA
(Status in
2013)
Other Sources
Simpson
Anderson,
CA
No
Idle in 2002
Closed
2001
Closed
htto://www. theshastamill.com/

Jacksonville,
FL
No
N/A
Not listed
1998
Not listed
(also not
listed in
2011 RTR
survey)
Wigmore Street (linerboard) -
closed 1998
httD://iacksonville.com/tu-
online/stories/120298/bus lfls
murf.html
httD://www.metroiacksonville.co
m/article/2012-mav-a-historical-
stroll-down-tallevrand-
avenue/oage/4
httD://www.nvtimes. com/1998/1
2/02/business/comDanv-news-
smurfit-stone-to-close-four-us-
containerboard-mills.html
htto://www. siteselection.com/th
eEnergvReoort/2010/dec/coal.cf
m
Florida Coast
Paper Co., LLC
Port St. Joe,
FL
No
Idle in 2002
Not listed
1998
Demolished
Jensen (1999), Master
2010Closures spreadsheet
httD://en.wikiDedia.ore/wiki/St.







Joe ComDanv

St. Mary's,
GA
Yes
Idle after
2002.
According
to AF&PA,
closed
October
2002.
Closed
2002
Closed
httD://iacksonville.com/tu-
online/stories/111502/met 1097
4474.shtml
httD://onlineathens. com/stories/
091402/bus 20020914026.shtml
httD://www.eeoreiaencvcloDedia.
org/articles/counties-cities-
neighborhoods/st-marvs
59 -

-------
Company
Mill
Listed
Status in
Status in
Shutdown Date
BECA
Other Sources

Location
in EPA
(2001)
2002 EPA
(2006)
mill2005p.xls
(USDA-Forest
Service, 2005)

(Status in
2013)








htto://www. voutube.com/watch







?v=FMzWk0s-d4E

Millinocker,
No
Phase II (no
Idle

Not listed


ME

bleached
chemical
pulp
operations)


(also not
listed in
2011 RTR
survey)


Westbrook,
No
Phase II (no
Not listed

Operating


ME

bleached
chemical
pulp
operations)




International
Moss Point,
Yes
Idle in 2002
Closed
2001
Closed
Master 2010Closures
Paper
MS



"The policy of balancing
International Paper
production with customer
demand resulted in taking
approximately 1.7 million tons
of market-related downtime
across the mill system.
Additionally in 2001, the
closure of paper mills in Erie,
Pennsylvania, and Moss Point,
Mississippi, four wood
products manufacturing
operations and certain
consumer packaging facilities,
and the down-sizing of the
Savannah, Georgia mill and
the Hudson River mill located

spreadsheet, 1999 Factbook (p.
534); 2000 K-10 MACHINE
SHUTDOWN;
htto://www. tidewaternews.com/
2009/11/21/other-shuttered-io-
mills-reDumosed/
60-

-------
Company
Mill
Location
Listed
in EPA
(2001)
Status in
2002 EPA
(2006)
Status in
mill2005p.xls
(USDA-Forest
Service, 2005)
Shutdown Date
BECA
(Status in
2013)
Other Sources





in Corinth, New York were
announced." (2001 K-10)







"In June 2001, the Consumer
Packaging business shut down
the Moss Point, Mississippi
mill and announced the
shutdown of its Clinton, Iowa
facility due to excess internal
capacity." (2001 K-10)


International
Paper
Natchez, MS
Yes
N/A
Closed
2003
"In January 2003,
International Paper
announced that it would close
the Natchez, Mississippi
dissolving pulp mill by mid-
2003 and exit the Chemical
Cellulose Pulp business."
"Specialty Businesses
recorded a severance charge
of $16 million associated with
the termination of 447
employees in connection with
the July 15th shutdown of the
Natchez, Mississippi mill."
(2005 K-10)
Closed
Master 2010Closures
spreadsheet, 2000 Factbook (p.
416) and 2000 Factbook (p. 36)
and 2002 Factbook (p. 39); 2003
K-10 MILL CLOSURE
htto://www. clarionledger.com/ar
ticle/20130812/NEWS01/308120
003/
httD://www.natchezdemocrat.co
m/2013/08/04/former-mill-
emDlovees-recall-life-without-
international-Daoer/
httD://msbusiness.com/blog/201
3/08/06/communitv-learns-hard-
lessons-from-iD-mill-closure/

Missoula,
MT
Yes
Phase II (no
bleached
chemical
pulp
operations)
Not listed

Closed

61 -

-------
Company
Mill
Location
Listed
in EPA
(2001)
Status in
2002 EPA
(2006)
Status in
mill2005p.xls
(USDA-Forest
Service, 2005)
Shutdown Date
BECA
(Status in
2013)
Other Sources

Lyons Falls,
NY
No
N/A
Open
2000 /2001
Closed
httD://oabonnv.com/indexoage6
9.htm
httD://www.watertowndailvtime
s.com/article/20130627/NEWS04
/706279878
httD://www.lcdeveloDment.net/D
ost/ docs/LvonsFallsMillRedevel
ooment.Ddf
Smurfit-Stone
Container Corp.
Circleville,
OH
No
N/A
Not listed
1998 (containerboard mill)
Not listed
(also not
listed in
2011 RTR
survey)
Jensen (1999), Master
2010Closures spreadsheet
httD://www.nvtimes. com/1998/1
2/02/business/comDanv-news-
smurfit-stone-to-close-four-us-
containerboard-mills.html
httD://www. disDatch.com/conte
nt/stories/business/2008/10/13/
ZONE1013.ART ART 10-13-
08 C12 7FBIGN9.html
International
Paper
Gardiner,
OR
No
N/A
Not listed
1998/1999
"Management indefinitely
closed the Gardiner, Oregon
mill because of excess
capacity in International
Paper's containerboard
system." (2000 K-10)
Not listed
(also not
listed in
2011 RTR
survey)
Master 2010Closures
spreadsheet
htto://en. wikioedia.org/wiki/Gar
diner. Oregon
httD://www. brian894x4.com/LPa
ndNrailroad.html
httD://www. katu.com/entertain
ment/3624111.html
62 -

-------
Company
Mill
Location
Listed
in EPA
(2001)
Status in
2002 EPA
(2006)
Status in
mill2005p.xls
(USDA-Forest
Service, 2005)
Shutdown Date
BECA
(Status in
2013)
Other Sources
Weyerhaeuser
North Bend,
OR
No
N/A
Not listed
2003
Not listed
(also not
listed in
2011 RTR
survey)
htto://www. cobis.gatech.edu/dat
a/mills-online/changed
httD://www. DaDerstudies.org/mil
Isonline/datachange.htm
(CLOSED 2003)
httD://offc-
online.com/news/1921-DaDer-
weverhaeuser-shutdown-
containerboard
Pulp mill ceased operation in
1995; operated as recycle paper
mill until 2003
htto://www. dea. state.or. us/la/E
CSI/ecsidetail.asD?seanbr=1083
International
Paper
Erie, PA
Yes
According
to AF&PA
mill closed
June 2002.
Closed
2002
"The Printing Papers business
approved a plan to shut down
the Erie, Pennsylvania mill due
to excess capacity in pulp and
paper and non-competitive
cost of operations." (2001 K-
10)
CLOSED (2002)
"The Printing Papers business
approved a plan to shut down
Closed
Master 2010Closures
spreadsheet, 2002 Factbook (p.
39), mill2005p.xls, 2002 10-K
MILL CLOSURE
htto://www. goerie.com/aDDs/Db
cs.dll/article?AID=/20020517/FR
ONTPAGE/105170219
httD://connection. ebscohost.com
/c/articles/5642488/iD-closes-
erie-mill
(other IP mills closed - Mobile,
Camden, and Moss Point)
63 -

-------
Company
Mill
Location
Listed
in EPA
(2001)
Status in
2002 EPA
(2006)
Status in
mill2005p.xls
(USDA-Forest
Service, 2005)
Shutdown Date
BECA
(Status in
2013)
Other Sources





the Erie, Pennsylvania mill due
to excess capacity in pulp and
paper and non-competitive
cost of operations."







Closed June 2002 (EPA 2006,
A-6)



Mehoopany,
PA
No
Phase II (no
bleached
chemical
pulp
operations)
Not listed

Operating


Lufkin, TX
Yes
Idle after
2002
Open
2003
Closed
httD://www. theDinevwoods.com
/Lufkin PlantJ05.htm
Donohue
Industries Inc.
Sheldon, TX
No
Idle after
2002
Open
2004
Closed
(listed as
Houston
mill)
Jensen(1999)
htto://business. high beam, com/5
874/article-lGl-
119881663/abitibiconsolidated-
shut-down-sheldon-texas-
newsDrint
Closed in 2004
httD://www. recvclinetodav.com/
Author.asox?AuthorlD=2825
htto://www. cdrecvcler.com/Artic
le.asoxParticle id=81528

Pasadena,
TX
No
Phase II (no
bleached
chemical
pulp
operations)
Not listed

Closed

64-

-------
Company
Mill
Location
Listed
in EPA
(2001)
Status in
2002 EPA
(2006)
Status in
mill2005p.xls
(USDA-Forest
Service, 2005)
Shutdown Date
BECA
(Status in
2013)
Other Sources
Georgia-Pacific
Bellingham,
WA
No
Idle after
2002
Closed
2001
"On March 30, 2001, the
Corporation announced that it
would permanently close its
pulp mill and associated
chemical plant at Bellingham,
Washington. This decision was
based on the age
of the facility and the
extraordinarily high energy
costs on the West Coast in
late 2000 and because diesel
generators and other power
alternatives were not cost
effective
at this facility. These
operations had been
temporarily closed since
December 2000."
Closed
Master 2010Closures
spreadsheet; 2001 K-10 MILL
CLOSURE
Pulp mill closed 2001
httD://www. theslowlane.com/bh
aminfo/eD.html
Rayonier
Port
Angeles, WA
No
N/A
Open
1997
"The Company concluded
that the mill was not
competitive in world markets
because of long-term high
wood costs due to federal
environmental restrictions on
Northwest timber harvests,
viscose pulp capacity
additions in lower cost regions
of the world and anticipated
Closed
Master 2010Closures
spreadsheet, 1999 Factbook (p.
547)
1997 K-10 MILL CLOSURE
httD://wa.sierraclub.ore/northolv
mDic/Daees/ravonier.html
65 -

-------
Company
Mill
Location
Listed
in EPA
(2001)
Status in
2002 EPA
(2006)
Status in
mill2005p.xls
(USDA-Forest
Service, 2005)
Shutdown Date
BECA
(Status in
2013)
Other Sources





large expenditures for new
environmental regulations."



Peshtigo, Wl
No
Phase II (no
bleached
chemical
pulp
operations)
Not listed

Operating

66-

-------
Chapter 3: Retrospective Evaluation of
the Costs Associated with
Methyl Bromide Critical Use
Methyl bromide (MBr) has been widely used as a fumigant to effectively control pests in a variety of
agricultural sectors (e.g. tomatoes, walnuts, strawberries, nursery crops, and forest seedlings). It is used to
fumigate the soil before planting and in some post-harvest applications as well as to meet export
requirements (e.g. quarantine and pre-shipment purposes). However, MBr was identified as a significant
ozone-depleting substance in 1992, which brought it under the auspices of the Clean Air Act and the
Montreal Protocol, an international treaty to protect the stratospheric ozone layer in the atmosphere. The
amount of MBr produced and imported by developed countries was phased out between 1993 and 2005
(Table 3.1).58 Developing countries agreed to begin phasing out methyl bromide use beginning in 2002 with a
complete phase-out by 2015. Carter et al. (2005a) note that a major objective of the long phase-out was to
allow time for users to develop competitive substitutes for MBr.
After developed countries reached 100 percent phase-out of methyl bromide in 2005, MBr for controlled (e.g.,
non-quarantine) uses could only be produced when a critical use exemption (CUE) had been agreed to by the
Parties (i.e. signatories) to the Montreal Protocol.59 This provision was included "in recognition of the
uncertainty of the innovation process" to further lengthen the phase-out for critical users such as agriculture
when feasible alternatives had not been identified (Carter et al. 2005a). Specifically, under the Protocol, a
critical use exemption can be granted to a developed country on behalf of farmers of a particular crop if:
58	Methyl bromide used for quarantine and pre-shipment purposes is exempt from this phase out schedule.
59	Title VI of the 1990 Clean Air Act Amendments allows for critical use exemptions for the production, import, or
consumption of methyl bromide that are consistent with the Montreal Protocol.
Exemptions for Open Field

Strawberries in California
Ann. Wolverton
67 -

-------
(i) - "The specific use is critical because the lack of availability of methyl bromide for that use would result
in a significant market disruption; and
Table 3.1: Methyl Bromide Phase-Out Schedule for Developed Countries
Years	Level of Phase-Out
1993 to 1998	Production frozen at 1991 baseline levels1
1999 and 2000	25% reduction from baseline levels
2001 and 2002	50% reduction from baseline levels
2003 and 2004	70% reduction from baseline levels
100% phase out - except for critical use (and a few other)
2005	exemptions
Source: EPA website, http://www.epa.gov/ozone/mbr/
(ii) - There are no technically and economically feasible alternatives or substitutes available to the user that
are acceptable from the standpoint of environment and public health and are suitable to the crops
and circumstances of the nomination" (UNEP 2006).
The threshold for economic infeasibility - or significant market disruption - is not defined by the Montreal
Protocol.60 However, beginning in 2010 the MBTOC indicated that alternatives that "lead to decreases in gross
margins of more than around 15 to 20 percent or more are not financially feasible" (MBTOC 2010). The MBTOC
(2010) also specifies that economic feasibility should be assessed by comparing the effects of using MBr and
its alternatives on "the 'bottom line of individual firms."
This paper examines ex post the per-acre operating cost estimates provided to the nomination process for MBr
critical use exemptions by the U.S. Environmental Protection Agency (EPA) to evaluate the economic feasibility
of MBr alternatives. In particular, this paper examines how EPA's ex ante cost analyses for open field fresh
strawberries grown in California for the 2006-2010 seasons (conducted annually two years prior to the
applicable season, 2004 - 2008) compare to an ex post assessment of costs and identifies possible reasons for
disparities. It does not attempt to evaluate how much MBr was exempted by the Parties to the Montreal
60 DeCanio and Norman (2005) calculate a "political willingness-to-pay" to identify possible economic feasibility
criteria for CUEs. Since MBr is a global pollutant - a ton of MBr emitted has the same effect on the ozone layer
regardless of where it is emitted - the authors argue that continued adherence to the agreement is predicated on
the benefits to signatory countries of phasing out MBr being at least as great as the incremental costs of projects
financed through a multilateral fund. The projects funded from 1993 to 2001 cost almost $16,000 per ton reduced
when weighted by project size (due to economies of scale, larger projects tend to be cheaper per ton than smaller
projects) or $32,000 per-ton unweighted. They compare this willingness-to-pay measure with the estimated cost
per ton of MBr found in the CUE nominations. The median cost increase is almost $24,000 per ton, indicating that
many of the requested CUEs would be economically feasible under the definition offered by the authors.
68 -

-------
Protocol for critical use for this time period. It also does not evaluate the extent to which EPA accurately
characterized regulatory and technical constraints faced by growers, though it does discuss how they may have
affected costs.61
While EPA uses the best available science to conduct its ex ante assessments, there are a variety of reasons
why ex ante and ex post estimates may differ. For instance, market conditions, energy prices, or the cost and
availability of technology may change in unanticipated ways. It is also possible that industry under or
overestimated the costs of compliance (EPA often has to rely on industry to supply it with otherwise unavailable
information on expected compliance costs). Finally, year-to-year variability of production in the agricultural
sector and challenges of estimation in general introduce significant uncertainty into ex ante cost estimates. For
this reason, we choose to examine multiple years of EPA analyses conducted in support of the critical use
exemption nomination process.
The ex post data also is limited in several key respects. We only have information on operating costs for a
typical farmer. We do not have information on the prices of specific fumigant formulations. Data on yield
losses associated with methyl bromide alternatives are based on field trials. While we have detailed annual
data on what fumigants farmers used, we do not have information on other management practices such as
the type of tarp used. It is also analytically challenging to evaluate a counterfactual of what would have
farmers done if they had not received the same level of MBr exemptions for the 2006-2010 seasons.62 Any
insights offered herein should be viewed with these limitations in mind.
3.1. Impetus and Timeline for the Regulatory Action
EPA solicits applications for MBr critical use exemptions from agricultural (and a few other) users on an annual
basis several years prior to the growing season to which the exemption would apply. As part of the
determination of whether and how much methyl bromide is nominated for critical use exemption, EPA
conducts a technical assessment, including a cost analysis, to evaluate all applications. Once the evaluation is
completed, the U.S. Government submits its critical use exemption nominations by commodity category to the
Ozone Secretariat for the Montreal Protocol. This occurs two years in advance of the season to which it will
apply. For instance, the 2006 nomination package was submitted in 2004, while a new nomination package for
the 2007 season was submitted in 2005. The packages are forwarded to an advisory group set up by the
Montreal Protocol, the Methyl Bromide Technical Options Committee (MBTOC), which reviews the packages
and makes a recommendation to the Parties for the amount of methyl bromide needed for each critical use.
61	While CUE nominations include a cost assessment to help determine economic feasibility, the amount of MBr
nominated for exemption is also based on an assessment of technical and regulatory constraints: Is an alternative
registered for use? Do state or local governments have buffer zone requirements or caps on use within a given
area? Are there terrain and temperature considerations that inhibit use of particular alternatives?
62	Analyses continue to be conducted annually in support of the critical use exemption nomination process. We
stop with the 2010 season due in large part to the availability of ex-post cost information.
69 -

-------
The United States has historically been granted about 90 percent of the total amount it has nominated for
exemption.63
3.1.1. Overall Trends in U.S. Ci	e Exemptions
U.S. critical use exemptions nominations for methyl bromide declined substantially from 2005 to 2010.
For instance, the U.S. submitted exemptions for 17 commodity categories for the 2006 growing season,
ranging from forest seedling nurseries to strawberries and tomatoes. These submissions represented 35
percent of U.S. baseline use. U.S. nominated critical use exemptions for the 2010 growing season also
covered a myriad of categories but constituted 13.4 percent of baseline use (Table 3.2).64
Several trends are worth noting (Figure 3.1). First, the aggregate amount of methyl bromide requested by
industry for agricultural use was far higher than what the U.S. nominated for exemption for the 2005 - 2010
growing seasons, though it also generally followed a downward trend. Second, the amount approved by the
Parties was lower than the U.S. government nominated amount. Third, while the amount of methyl bromide
nominated for exemption each year declined, the U.S. sometimes increased nominated amounts for specific
crops or regions between years. Finally, the amount of methyl bromide allowed under the critical use
exemption was met in part by drawing down the stockpile.65
The USDA (2000) notes that prior to phasing out methyl bromide, Florida and California growers accounted
for over 75 percent of pre-plant use for fumigation of soils, with California alone accounting for almost 50
percent of the total in the United States. The best disaggregated data on fumigant use and unit costs for fruit
and vegetable crops are available for California. No equivalent data are available for Florida. For these
reasons, we focus on assessing the ex post costs of critical use exemptions for particular crops in the state of
California when the amount granted is less than what was originally requested.
At the national level, five open field crops were granted critical use exemptions at levels substantially below
what was originally requested: cucurbits (i.e., squash and melons), eggplant, tomatoes, strawberries, and
peppers. In California, cucurbit and eggplant farmers did not request an exemption for MBr use over this time
frame. The three remaining crops were responsible for about 62 percent of U.S. methyl bromide use in 1991,
just prior to the beginning of the phase-out (Ferguson and Yee 1997; USDA 2000). They constituted 68
percent of the total amount of MBr nominated for critical use exemption in 2009. From these, we choose to
63	See "2005-2013 Critical Use Exemption Authorizations" at http://www.epa.gov/ozone/mbr/cueinfo.html.
64	For the 2014 season, the amounts nominated and authorized for U.S. use decreased to 1.7 and 1.7 percent,
respectively, and covered four commodity categories, one of which was California strawberries. Note that EPA
anticipates that California strawberry growers will completely transition out of methyl bromide by 2017 through
the use of straight choloropicrin at rates up to 350 pounds per acre, steam and anaerobic soil disinfestation. See
http://www.epa.gov/ozone/mbr/CUN2016/2016CUNStrawberries.pdf.
65	The stockpile consists of MBr produced prior to the 2005 phase-out. Use of stockpiled MBr for the replanting of
turf was not allowed after April 2014, though the deadline was extended for golf courses through November 2014.
For more information on the potential role of the stockpile, see Appendix 3.2.
70-

-------
Table 3.2: Percent of Baseline MBr Consumption in U.S. Exempted for Critical Use by Year
Calendar Year
Growing Season
U.S. Nominated Amount
(percent of baseline)
Amount Authorized by Parties
for Use in U.S.
(percent of baseline)
2005
2006
2007
2008
2009
2010
39
35
29
23
19.5
13.4
37
32
26
21
16.7
12.7
Source: http://www.epa.gov/ozone/mbr/cueinfo.html accessed 03/19/12.
Figure 3.1: U.S. MBr Production, Imports, and Drawdown of Stockpile for Critical Use (2005 - 2010)
16,000
14,000
12,000
10,000
8,000
6,000
4,000
2,000
2005
2006
2007
2008
2009
2010
Annual Drawdown of U.S. Stockpile for Critical Use
Production + Imports
—	Nominated by U.S. for Exemption
—	Approved by Parties for U.S. Use
Industry Requests (not double counting)
Source: US EPA1 and UNEP (2010).
71 -

-------
focus on California open field strawberries for the 2006 to 2010 seasons, though at times we are only be able
to evaluate a subset of these years.66,67
Table 3.3 illustrates the amount of MBr the United States nominated for exemption for use on strawberry
fields and what this represents in terms of the amount originally requested by growers for the 2006 - 2010
seasons. California made up the vast majority of the requested amount each year (67 percent in 2006 and 80
percent in 2010). This is not surprising as more than 85 percent of all fresh and processed strawberries
grown in the United States came from California in 2007. In each state/region that requested a critical use
exemption, the amount requested by farmers was almost always higher than what EPA nominated for
exemption. However, the rate of decrease in the amount nominated was markedly slower in California than
in other parts of the country, mainly due to regulatory constraints discussed later in the paper.68 Between
the 2006 and 2010 growing seasons, the amount of methyl bromide nominated for exemption for strawberry
fields in California declined by 12 percent, while it declined by 45 percent and 67 percent in Florida and
Eastern states, respectively, over the same period.
3,2. EPA Ex Ante Cost Estimates for Open Field
Strawberries in California
In keeping with MBTOC (2010) guidance, three years prior to the year for which the MBr is approved for use
EPA evaluates the per acre impacts of using methyl bromide and a set of alternatives on the bottom line
finances of a typical farmer on a per-crop basis. As part of this process, EPA assesses the rate at which MBr is
applied (e.g., pounds per acre) and the total amount of land where economic, technical, and regulatory
constraints inhibit the use of alternatives to determine the aggregate amount of methyl bromide to nominate
for critical use exemption in a given year.69
66	By 2001, MBr use had declined substantially per the Montreal Protocol. Carter et al. (2005b) find that California
farmers of non-strawberry crops reduced MBr use by 59 percent between 1996 and 2001, while California
strawberry growers only decreased MBr use by 14 percent over the same time period. About 88 percent of
California strawberries acreage in 2001 continued to use MBr (Carter et al. 2005a).
67	Tomatoes grown in open fields also appear to be a good candidate for study. Florida and Eastern U.S. farmers
continue to request critical use exemptions for tomatoes. While California tomato growers requested MBr critical
use exemptions for hilly terrain for the 2006 and 2007 seasons, they did not apply for CUEs for subsequent
seasons. This raises the question of whether California tomato farmers relied on the MBr stockpile, switched to
growing other crops, or discovered affordable alternatives to MBr that would work effectively on hilly terrain.
68	A review of the CUE nomination packages suggests that EPA initially underestimated California regulatory
constraints faced by farmers for MBr alternatives and that it modified its requests to account for them in CUEs for
subsequent growing seasons (i.e. the amount nominated for exemption jumped by 17 percent from 2006 to 2007).
69	EPA also tries to eliminate any double counting from the requested amount and subtracts out land that
represents growth since 2005 in the industry, since it does not qualify for exemptions.
72 -

-------
Table 3.3: Amount of MBr Requested by Industry and Nominated for Critical Use Exemption in California, -
Florida and Eastern U.S. for Strawberries1

Year
2006
2007
2008
2009
2010
California
Amount (kg)
Nominated
1,086,777
1,267,880
1,244,656
1,064,556
952,543

% of Amount
67%
87%
98%
90%
100%

Requested by






Industry





Florida
Amount (kg)
Nominated
295,853
297,909
220,302
176,333
163,440

% of Amount
51%
51%
38%
30%
28%

Requested by






Industry





Eastern
Amount (kg)
230,332
165,735
137,334
93,488
75,832
U.S.
Nominated






% of Amount
66%
46%
36%
34%*
28%*

Requested by






Industry





Source: EPA Critical Use Exemption Nominations for open field strawberries for the 2006- 2010 seasons.
Because EPA assesses the burden associated with switching to methyl bromide alternatives, the baseline
against which these alternatives are assessed is the continued use of MBr (i.e., continued exemption) instead
of zero MBr use (i.e., no exemptions to the phase-out for critical use). Operating costs and gross revenues are
calculated for methyl bromide and what were deemed feasible alternatives by EPA at the time of the
assessment on a per acre basis.70 The net revenues from using an alternative are then compared to those for
methyl bromide to generate a loss per acre.71 No aggregate estimates of costs (and revenue loss) are
provided by EPA as part of the CUE nomination package, though one could calculate them from the
information available assuming that all acreage to which MBr is applied resembles a typical acre.
In the CUE nomination packages for the 2006-2008 seasons, EPA evaluated the operating cost and revenues
for methyl bromide combined with chloropicrin (PIC) in a 67:33 formulation and three alternatives that
combined fumigants, 1,3-dichloropropene + chloropicrin (1,3-D + PIC), chloropicrin + metam sodium (PIC +
MS), and metam sodium alone (MS).72 For the 2009-2010 seasons, EPA dropped PIC+MS as an evaluated
70	EPA considers all known chemical and non-chemical alternatives to methyl bromide but focuses the analysis on
the subset of the most likely alternatives based on CUE applications, the published literature, and grower input.
71	For purposes of submitting the package to the MBTOC, all values are converted into kilograms and hectares. The
numbers reported here are expressed in pounds and acres.
72	Methyl bromide is typically combined with other chemicals before being applied to a field. A common
formulation in California for use on strawberries in 2000 was 67 parts MBr and 33 parts PIC, though ratios of 57:43
and 75:25 were also regularly used (California Pesticide Information Portal database). How MBr is used to treat
strawberry fields varies to some degree by region. In California, the entire surface of the field is typically
73 -

-------
alternative from the economic analysis.73 While EPA recognized several other potential MBr alternatives, it
did not analyze them for the 2006 - 2010 seasons (see Table 3.4) because they were not yet registered for
use in the United States.
3,2,1, K«> > -v, Ante Cost Estimates
Gross revenues per acre for MBr and its alternatives are calculated by multiplying the market price of the
fruit times the yield. They depend on three main components: potential yield loss due to use of an
alternative, the expected producer price of strawberries, and the potential loss of revenue due to a planting
delay that results in a missed market window. Changes in product quality that could result in lower revenues
and additional fixed costs from the use of an alternative (e.g. a drip system for applying it), while discussed,
are not quantified. While EPA included an estimate of the effect of missing a market window on revenues in
its assessment for the 2006 - 2008 seasons, it dropped it in later year analyses due to lack of evidence of a
harvesting delay (i.e. for the 2009 - 2010 seasons, the market price for strawberries was identical whether
using MBr or an analyzed alternative).74
For the 2006-2010 season CUE nominations, the range of yield losses associated with the three evaluated
MBr alternatives were based on a review of the available literature. The estimate of yield loss for a typical
California farmer from switching to PIC+MS was drawn from an unpublished report (Locascio et al. 1999). EPA
used the mean estimates from Shaw and Larson (1999) to represent yield losses from switching to 1,3-D + PIC
or to MS (see Table 3.5). Shaw and Larson used meta-analysis techniques to compare yield estimates for
methyl bromide-chloropicrin with four other soil treatments applied to California strawberries in three
distinct locations. The test years for the 45 studies underlying the meta-analysis ranged from 1987 to 1997.
EPA retained the same yield loss estimates in the critical use exemption nomination packages for the 2006 -
2010 seasons. The key reason cited by Office of Pesticides Program (OPP) experts for retaining this
assumption was a desire to rely on multi-year studies, as many factors can influence realized yield losses
fumigated, covered by a tarp, and left to sit for a period of time. After the tarp is removed, farmers form planting
beds and then again cover them with plastic. Planting begins 2-6 weeks after fumigation. After harvest, new crops
are planted that benefit from the initial fumigation. In Florida, MBr is applied when raised beds for strawberries
are constructed. The beds are covered with plastic mulch. Two weeks later strawberry plants are transplanted and
fed via drip irrigation. After harvest, existing beds are often used to produce a second crop (EPA 2008).
73	OPP experts indicate that MS+PIC was dropped mainly for technical reasons: It does not distribute evenly or
deeply enough in the soil to be effective against nematodes or pathogens and thus is used mostly for weed
management after 1,3-D + PIC is applied.
74	Industry CUE applications stopped mentioning harvesting delays when switching to MBr alternatives.
74-

-------
Table 3.4: Recognized but Non-Registered MBr Alternatives for Growing Strawberries in CA (as
reported in the CUEs for the 2006-2010 Growing Seasons) -

First


Mention


ed in

Unregistered Alternatives
CUE
Status as of 2009 Growing Season CUE
Basamid
2006
Registration being considered
Methyl iodide
2006
Registration being considered; trial use only
Propargyl bromide
2006
Under proprietary development for future registration
Sodium azide
2006
Under proprietary development for future registration
Furfural
2007
Registration being considered (used for greenhouse
ornamentals)
Muscador ablus strain QST
20799
2008
Registration package received but not yet for sale in US
dimethyl disulfide (DMDS)
2009
Under proprietary development for future registration
Source: CUE nomination packages for the 2006-2010 seasons; OPP provided spreadsheet.
Table 3.5: Estimates of California Strawberry Yield Loss from the Published Literature, as Reported in
the CUE Nomination Packages
MBr Alternative
Range of Yield Loss
Relative to MBr
"Best" Estimate
PIC+MS
6.6% - 47% loss
27%
1,3-D+PIC
1% gain -14% loss
14.4%
MS
16% - 29.8% loss
29.8%
Source: EPA CUE Nominations for 2006-2010 seasons.
(e.g., weather, pest pressure) in a given year.75 The yield per acre for each alternative was derived by
multiplying the estimated yield from methyl bromide by the "best" estimate of yield loss for an alternative.76
The difference in the price of strawberries per pound between baseline and policy was based on an
assessment of the potential for a decrease in price from delaying harvest by several weeks for the 2006 -
2008 seasons. EPA used USDA data to estimate that market prices for strawberries would decline by about 5
percent due to such a harvesting delay ($0.69 per pound x 5% = $0.66). EPA updated the market prices for
strawberries to evaluate the CUE for the 2009 and 2010 growing seasons.
75	Multi-year studies better reflects yield losses from changes in pesticide controls versus seasonal factors.
76	While not included in the nomination, EPA also evaluated cases where yield loss was at the low end of the range
for all three alternatives (7% for PIC+MS; 1% for 1,3-D+PIC; and 16% for MS) and a high case for 1,3-D+PIC.
75 -

-------
EPA used a bottom-up approach to estimate operating costs.77 Based on information provided by industry in
their CUE applications, EPA estimated the labor and material costs associated with land preparation (e.g.,
seed, fertilizer, pesticide, and fumigant), weeding, irrigation, and harvest when using MBr versus its main
alternatives.78 The same basic approach to estimating operating costs was used for each of the 2006-2010
season CUEs, with only slight changes (for instance, updating fumigant prices) for the 2009 and 2010 seasons.
To assess the amount of active ingredient applied during fumigation, EPA used the average number of annual
applications of methyl bromide or its alternatives (i.e., one application) used to treat the crop.
Overall operating costs were nearly identical for methyl bromide and the analyzed alternatives but differed in
three specific areas: the cost of the fumigant, manual labor needed to apply the fumigant, and harvest labor
(due to its relationship to yield). According to spreadsheets provided by OPP, the application of MBr
alternatives were estimated to require a bit less manual (5 percent less for all alternatives) and harvest labor
(between 7 and 15 percent less) than MBr. The analysis assumed that all other aspects of growing
strawberries remained unchanged.79
Table 3.6 presents a summary of the operating cost and gross revenue information that underlies the critical
use exemption analysis for the 2006 planting season. For open field strawberries, the losses per acre from
switching to a MBr alternative were driven primarily by the difference in yield, with EPA predicting based on
its data and assumptions that 1,3-D+PIC was the next best alternative to methyl bromide.80 The loss per acre
was calculated by examining the change in net revenues relative to using MBr. The alternative that resulted
in the lowest loss was determined to be the most likely substitute.
The overall conclusion of the cost analyses for the 2006 - 2010 seasons was that use of the most viable
alternative, 1,3-D + PIC, instead of methyl bromide would result in about a 16 percent loss on a per acre basis
as a percent of gross revenues.
77	EPA did not include fixed costs due to wide variability in factors that influence them (e.g., farm size, technology
adoption). Applicants were asked to provide this information on their exemption request forms.
78	EPA does not quantify the effect of switching to a MBr alternative on the costs of growing a rotation crop. For
example, if a lettuce field that has soil pathogens is leased to a strawberry grower who fumigates the soil prior to
planting and the field is rotated back to lettuce after three years (the soil pathogen has been controlled), both
crops benefit from the strawberry crop soil fumigation. However, only the effects on strawberries are considered.
79	EPA mentioned but did not include other costs of switching to MBr alternatives. For example, 1,3-D + PIC (the
alternative with lowest yield loss) is reportedly less effective with broadcast fumigation than drip fumigation and
would therefore requires 40 percent more fumigant (EPA 2005).
80	On net, the estimated loss from using a MBr alternative was similar across CUEs for the 2006-2010 seasons for
1,3-D + PIC but somewhat higher in 2009 for MS (32 percent instead of 26 percent), according to EPA calculations.
76-

-------
Table 3.6: Yields, Revenues, and Operating Costs for Open Field CA Strawberries
(2006 - 2008 Growing Seasons) -
Fumigant
Methyl Bromide
Alternatives
PIC+MS
1,3-D+PIC
MS
yield loss
0%
27%
14%
30%
yield (pounds per acre)
43,215
31,547
37,165
30,251
strawberries price per pound
$0.69
$0.66
$0.66
$0.66
gross revenue per acre
$29,818
$20,679
$24,362
$19,829
operating costs per acre
$24,334
$22,395
$23,659
$22,226
net revenue per acre
$5,484
($1,716)
$702
($2,396)
loss per acre
$0
$7,200
$4,782
$7,881
loss as percent of MBr gross revenue
-
24%
16%
26%
Source: EPA CUE Nominations, converted to pounds and acres. Note that the CUEs express application rates
and land in kilograms and hectares.
3,2,2, Main Sou'	_ Ex Ante Cost Estimates
Ex ante analyses are subject to many challenges and uncertainties. Recall that EPA conducts its analyses three
years prior to when a CUE is approved, making it difficult to precisely estimate how much methyl bromide
will actually be needed in a given growing season, what MBr alternatives will be available for use, and the
yield loss and operating costs associated with each option.
At the time the phase-out began, the USDA (2000) reported that the most promising alternatives to
methyl bromide for agricultural use were a combination of the fumigants 1,3-dichloropropene and
chloropicrin (1,3-D + PIC), or chloropicrin combined with metam sodium, napropamide (an herbicide
registered for use on eggplant), or pebulate (also an herbicide, now de-registered for use on tomatoes).
Metam sodium was viewed as a potentially viable alternative in areas where the use of 1,3-D was
restricted (see section E.3).81 As many of the studies up to that time had focused on the performance of
MBr alternatives with regard to California strawberries or Florida tomatoes, the USDA document noted
even greater uncertainty regarding alternatives for use on other crops.
Other factors that could affect the rate at which MBr alternatives were adopted include use restrictions
to protect workers and bystanders from health effects associated with their toxicity, and U.S. EPA and
state registration requirements. The USDA (2000) noted that several possible alternatives-for instance,
81 Noling et al. (2010) note that a key challenge to transitioning out of MBr has been its effectiveness against
nematodes (i.e., roundworms), disease, and weeds. Many of the registered alternatives are only effective against a
subset of these problems. For instance, chloropicrin is effective against disease, but far less effective in fighting
nematodes or weeds. 1,3-D is effective against nematodes but does less well in fighting disease or weeds. Metam
sodium is good for weed control but does little to guard against disease or nematodes. As a result, farmers often
use these chemicals in combination.
77 -

-------
methyl iodide and propargyl bromide - were not registered under the Federal Insecticide, Fungicide,
and Rodenticide Act (FIFRA) at the time that the phase-out began.
In addition, EPA faced the challenge of generating ex ante estimates based on somewhat limited data and
poor documentation in source reports on yield loss associated with various MBr alternatives. Without
assessing the relative quality of the studies upon which EPA relied, one can observe that the yield loss
estimates used to evaluate costs for the 2006 - 2010 seasons were based on fairly old data. The "best"
estimates used by EPA in its analysis reflected the high end of the range reported in the CUE nomination
packages for yield loss for two of the three alternatives (see Table 3.5). According to OPP experts, EPA used
conservative assumptions in the early years of the critical use exemption process because the literature
contained a wide range of yield loss estimates for which researchers often did not clearly describe what
impacts were included.
Finally, while EPA did not evaluate how lack of a critical use exemption would affect the ability of
California farmers to compete in the global marketplace it is a relevant consideration and a source of
uncertainty with regard to the ultimate financial welfare of farmers. In particular, it is important to
understand how switching to a MBr alternative impacted the ability of conventional production in
California to compete with organic production in California and imports from Mexico.
3,2,3, Exogenoi tors that Mav Affect Estimated Ex Ante Costs
Regulatory and technical constraints or unexpected innovation could result in greater or less use of MBr
alternatives by California strawberry farmers. While the focus of the economic analysis is an assessment of
per acre costs, EPA also makes a recommendation of the total amount of methyl bromide that should be
exempted for use by crop and region based on an assessment of what alternatives are likely to be available
and regulatory and technical restrictions. Specifically, the requested amount is based on the rate at which
methyl bromide is applied times the total amount of land where technical and regulatory constraints prevent
switching to MBr alternatives (and therefore, are eligible for exemption).
Table 3.7 shows that strawberry farmers' initial requests were based on a higher rate and acreage than EPA
nomination, but in later years they were similar or identical. Also, note that while the application rate used
by EPA initially declined (from 2006 to 2007), it increased for the 2008 -2010 seasons. The amount of land
deemed eligible for MBr use followed a similar trend.
EPA noted that the rate of adoption of MBr alternatives was limited by a combination of transitional and
regulatory issues. In general, the amount of land assumed to face technical constraints stayed about the
same across all five growing seasons - approximately 10-15 percent of land used to grow strawberries in
California was assumed to be on hilly terrain that does not support the drip systems required to apply many
MBr alternatives. However, EPA accounted for the use of strip fumigation (i.e. about 10 percent of land used
this form of fumigation, which has a lower application rate) and the change in the ratio of MBr to PIC from
67:33 to 50:50 in its analysis of the 2009 and 2010 seasons.
78 -

-------
Table 3.7: Application Rates and Acreage Underlying Methyl Bromide Exemption Nominations for Open
Field CA Strawberries -
Growing Season
2006
2007
2008
2009
2010
Application
Rate (lbs/acre)
Requested by farmers
Used by EPA in nominations
180
175
160
160
180
180
174
174
175
175
Acres
Requested by farmers
Qualified for nomination
20,000
13,720
20,000
17,470
15,555
15,244
14,925
13,472
12,000
12,000
Source: EPA CUE nominations, converted to pounds and acres. Note that the CUEs express application rates
and land in kilograms and hectares.
The impact of regulatory constraints on the use of alternatives is not easy to determine and can be different
for every strawberry growing area in California. The main regulatory constraint accounted for by EPA was
California's restrictions on the use of 1,3-D. in the CUE nomination package for 2006, EPA assumed these
restrictions applied to a smaller subset of the total acreage (47-67 percent, which is identical to the California
Strawberry Commission's (CSC) estimate noted in the 2009 CUE) than what it assumed for the subsequent
seasons (82-94 percent).82 This was based on the assumption that some townships would be allowed to
exceed the cap by up to 2 times. However, uncertainty regarding how the process would work resulted in
EPA interpreting the caps strictly for the 2007 season. EPA noted that fewer townships would find the cap on
1,3-D binding if farmers switched to drip irrigation, as less chemical would be required (also, see Carpenter et
al. 2001). However, this could result in a 3-4 week planting delay. According to OPP experts, there were also
county-level restrictions on the use of chloropicrin and metam sodium, though the effects of these
restrictions were not quantified.
With regard to an ex post assessment, we ask whether the regulatory or technical constraints differ from
what was expected and their implications for the cost of alternatives. Ideally, it would be useful to know if
any new alternatives had been registered for use; whether state and/or local governments initiated new
regulatory requirements; and whether terrain and temperature considerations still inhibited the use of
particular alternatives during the 2006 - 2010 seasons.
82 Information available in the CUE nomination packages indicated that 19,550 to 20,900 acres used MBr pre-phase
out between 2000 and 2003. MBr application rates had already begun to decline from 218 pounds per acre in 2000
to 170-179 pounds per acre in 2001-2003. In 2004, the amount of land requiring methyl bromide decreased
substantially - to 17,680 acres - and about 10 percent of farmers using MBr switched from flat fumigation to strip
fumigation, which had a lower application rate (129 lbs/acre) than flat fumigation (172 lbs/acre).
79 -

-------
33, Literature and Data Available to Conduct Ex Post
Evaluation
There are several key components of costs that can be potentially examined ex post (See the first
chapter in this report for a discussion of the conceptual framework for costs): what types and how many
entities comply with the regulation; what technologies or strategies are used to comply; the initial costs
and the ongoing costs of compliance; any indirect costs such as quality tradeoffs or missed market
windows; and other opportunity costs such as costs in related markets. For this case study we largely
rely on publically available data and resources. Specifically, we review the existing literature to identify
any ex post studies on MBr critical use exemptions as well as available data sources on key inputs to the
ex ante cost analysis (e.g., availability of MBr alternatives, fumigant use by type, input and strawberry
prices, production, yields, and operating costs).
It is also important to note several data limitations that will affect the extent to which we can opine on
some aspects of the ex ante cost analysis. First, ex post evaluations of MBr critical use exemptions are
rare in the literature. Second, market data on fruit and vegetable crops are not as widely available as for
row crops, particularly at a geographically disaggregated scale. Third, publically available data to
evaluate the operating costs associated with switching to a MBr alternative in California are also limited.
3.3.1. Ex Post
A number of papers have evaluated the potential impact of banning MBr use in the United States and, in
some cases, have analyzed to what extent critical use exemptions may alleviate this impact. However, a
search of the literature and emails to key researchers who have studied the economic impacts of
banning methyl bromide uncovered only one published ex post analysis of the impact of critical use
exemptions for MBr use by California strawberry farmers (Mayfield and Norman 2012).83 The authors
find little evidence of negative impacts on farmers of the phase out, in part due to exemptions. While no
formal counterfactual is evaluated, they point to rising yields, acreage, exports, revenues, and market
share as evidence that the industry has not faced substantial negative impacts. A review of the main ex
ante studies of a MBr ban or phase-out is available in the appendix.
A number of recent studies also estimate yield effects of various chemical combinations compared to methyl
bromide + chloropicrin for strawberries based on field trials. We also identify a meta-analysis covering studies
from 1997 - 2006 sponsored and approved by the MBTOC (Porter et al, 2006). The MBTOC also discusses
recent evidence in its 2010 assessment report (UNEP 2010). Recent studies by Othman et al. (2009) and
83 Catherine Norman, Rachel Goodhue, Colin Carter, and Lori Lynch did not know of any other ex-post analyses of
the MBr phase out, but they suggested we search for information from the Annual Methyl Bromide Alternatives
Conference, the UC-Davis Cost and Impact Estimates Database, and the University of Florida methyl bromide
research group. We found no ex-post analysis at any of these forums.
80-

-------
Fennimore and Ajwa (2011) are particularly relevant because of their focus on California. Since yield loss is
one of the key uncertainties identified in the ex ante analysis, we discuss these studies in greater detail in
section E.4 as part of the ex post evaluation.
. vting Costs E*>, ,-t
The EPA critical use exemption nomination packages are a good starting point for information on MBr usage.
In particular, EPA included information on the amount of methyl bromide used in prior years as part of each
annual nomination package. For instance, information in the nominating package for the 2013 season may
reveal actual application rates and overall usage for the 2006 - 2010 seasons.
EPA relied on the 2002 NASS Agricultural Chemical Usage Vegetables Survey for information on the
proportion of acreage in California using methyl bromide in the 2006 - 2008 CUE nomination packages. Since
that time, 2006 and 2010 data have been published. The survey also reported the average application rate
and total pounds applied for several states, including California, which can be used in combination with what
was reported in the most recent critical use exemption nominating packages.
To shed light on whether EPA accurately characterized likely MBr alternatives in its ex ante analysis, we rely
on two data sources: future year (post 2010) CUE nomination packages indicate if and when new alternatives
other than those identified in the original package became available (i.e., were registered for use) and what
the experiences of farmers have been with respect to their use. Spatially disaggregated data, available from
the state of California by month, year, and crop from 1989 to 2009 through its Pesticide Information Portal
(PIP), indicate the amount of a specific chemical used and the acreage treated.84 Methyl bromide and many
of the alternatives in EPA's ex ante cost analyses are in the database. Conversations with experts in EPA's
Office of Pesticide Programs as well as discussions in the literature indicate some level of data error in PIP.
Carpenter et al. (2001) state that acreage treated with MBr may be overstated for perennial crops due to
spot treatments on small areas that are reported as though they are full-acre treatments. On the other hand,
a certain amount of MBr use is not reported in the database. Carpenter et al. (2001) note that in 1999 about
2 million pounds used on 8,000 acres (about 13 percent of the total area treated with MBr in California at the
time) were not included in the database. Finally, Carpenter et al. (2001) identify a number of duplicate
entries for MBr and its alternatives. Possible errors also are flagged in the database based on a statistical
analysis of outliers.
Net compliance costs consist, in this case, of gross revenues minus operating costs. As previously stated,
gross revenues were estimated using information about the market price of strawberries and yield for MBr
and several alternatives. Yield information is mainly drawn from the literature. The market price of
strawberries can be evaluated based on national and state-level monthly data on prices received by growers,
and retail prices for 1970-2009 from the United States Department of Agriculture (USDA).85The USDA's 2009
Fruit and Tree Nut Yearbook also reports the national-level monthly average retail and grower prices by year
84	The data can be downloaded from http://calpip.cdpr.ca.gov/main.cfm.
85	For a list, see http://usda.mannlib.Cornell.edu/MannUsda/viewDocumentlnfo.do?documentlD=1381
81 -

-------
back to the mid-1990s. Gross revenues could also be affected by changed in product quality. EPA was not
able to assess ex ante the effect of eliminating methyl bromide on the quality of strawberries due to lack of
data. For similar reasons, we also are not able to assess any effects on product quality in the ex post analysis.
We explore several sources of information on ex post operating costs.86 The first source of information is
crop budgets for open field strawberries assembled by UC-David researchers for the South Coast region in
Santa Barbara and Ventura Counties in 2006 and 2011 and for the Central Coast region in Santa Cruz and
Monterrey Counties in 2010, respectively. These studies generated sample operating (and to some extent,
fixed) costs and revenues for a representative farm. Operating costs are differentiated by stage of production
- land preparation, plant establishment, fertilization, irrigation, pests, harvesting, and end-year clean-up.
While the 2011 reports were issued after the last year of ex ante estimates to which we make comparisons,
they are useful because they evaluate alternative fumigants to methyl bromide. The 2006 and 2010 studies
use MB+PIC as the default fumigant.
We are cognizant of the limitations of using crop budgets for estimating ex post operating costs. They are not
necessarily indicative of actual costs for any individual farmer. Instead, they are produced to help farmers
assess the profitability of growing particular crops and may include cost categories that do not apply to many
growers. That said, the crop budgets are the only ex post information we have on costs, are produced for
strawberry-growing regions in California that overlap with areas seeking critical use exemptions, and are
described by the authors as representative of costs faced by a typical farmer, which is also the focus of EPA
ex ante cost analyses.
The second source of information is the CUE requests for the 2012 growing season. The third source is
proprietary data purchased by EPA Office of Pesticides Program from a private pesticide marketing company.
It is based on a survey of farmers. This database has information on fumigant and pesticide use, total area
treated with methyl bromide or an alternative, total amount of chemical used, average application rates,
crop yields, and chemical application expenditures by crop. Sample size is reported but is often small
depending on the crop and region of the country, making much of the data not useful for our purposes. In
particular, we are only able to rely on fumigant prices for the ex post assessment. Because the database is
proprietary, we report data only in highly aggregate form. The final source of information is studies that take
a bottom-up approach to estimating costs associated with using methyl bromide or one of its alternatives
based on field experiment data.
Information on the role played by California regulatory constraints is limited. We only have information from
the literature that pre-dates or coincides with the critical use exemption nomination packages for the 2006 -
2010 seasons. While we discuss these studies as part of our evaluation, it is not possible to come to any
definitive conclusions ex post regarding the role of regulatory restrictions on the pace and types of MBr
alternatives utilized over this time period.
86 A search through the Federal Register for reregistering MBr alternatives did not uncover any additional sources
of cost information (i.e., the regulatory notices are largely focused on evaluating exposure risk and health impacts).
82 -

-------
Finally, to investigate whether reducing MBr use resulted in unanticipated competitive disadvantages either
in the conventional, organic, or international markets for strawberries we examine national level data from
USDA on production, utilization, acreage, shipments, and yield per acre for strawberries for 1970-2009.
National level monthly data is also reported on imports and exports by country. State-level information is
available for a subset of these variables annually: for instance, harvested acreage, yield per acre, and
production.87 The USDA's 2009 Fruit and Tree Nut Yearbook also includes supply, utilization, and trade
statistics by year at the national level.88 89
*'l ^	5 .-is n	: a	M
Retrospectively
Comparing ex ante compliance costs to ex post estimates of actual compliance costs is challenging for all the
usual reasons - limited access to cost data in the post-regulatory period, few retrospective analyses, etc.
However, a retrospective review of the cost analyses conducted by EPA for MBr critical use exemptions faces
additional challenges. Unlike regulations that seek to control a substance, MBr critical use exemptions allow
for the use of a substance that is otherwise banned. In a typical ex post evaluation, we compare what
analysts estimated ex ante to the cost of actions taken by regulated entities to comply with the rule ex post.
In the case of methyl bromide, however, the market does not reveal the cost of actions that would have
otherwise been taken in the absence of the exemption - moving to a more expensive or less effective
substitute, for example. In other words, we do not have a measurable and quantifiable counterfactual based
on real world revealed market behavior. With this limitation in mind, it may still be possible in some cases to
learn something useful without having to estimate an approximate counterfactual. In particular, because
strawberry farmers request far more than what the U.S. nominates for exemption, we can examine whether
growers faced larger than expected costs of switching to non-MBr substitutes by comparing EPA estimates to
what we know about costs observed in the marketplace.90
While regulations are often revised, the timeframe over which this occurs is typically longer, allowing -in
theory - for an ex post analysis to isolate the effect of a regulation on costs from other factors, including
previous or subsequent rulemakings that apply to the same industry. In the case of methyl bromide, it is
challenging to isolate the cost implications of a CUE in a given year from those of future CUEs. In addition,
87	For a list, see http://usda.mannlib.Cornell.edu/MannUsda/viewDocumentlnfo.do?documentlD=1381
88	The USDA also has a publically available database of own and cross-price demand elasticity estimates from the
published literature by commodity. The database was last updated in September 2009 and is available at:
http://www.ers.usda.gov/Data/Elasticities/querv.aspx. No equivalent database is available for fumigants.
89	In addition, the USDA publishes typical planting and harvesting dates, the most active growing season by crop
and state for strawberries, and the principal producing counties in each state (USDA 2006). However, these data
are of limited use since a more recent version of these data are not produced for fruits and vegetables.
90	Another case where we might potentially be able to estimate the counterfactual is when the amount authorized
by the Parties is non-binding for particular agricultural uses; MBr alternatives turned out to be cheaper than
anticipated and more growers moved to substitutes than originally anticipated.
83 -

-------
some researchers have speculated that there may be a strategic element embedded in the requests made by
industry, particularly since it is repeated annually.91 We choose to examine the cost analyses in the CUE
nominations for the 2006 - 2010 growing seasons as a group, given the unique nature of the CUE process.
This also makes some sense because EPA did not substantially alter the assumptions or inputs to its cost
analyses for California strawberries over this timeframe.
This remainder of this section compares EPA's ex ante cost estimates to available ex post estimates for each
cost component, identifying possible reasons for substantial differences. A summary of the main sources of
information, study findings for each main cost category and limitations of the study appear in Section X.X.X.
3.4.1.	Regulated Universe
EPA ex ante cost analyses conducted for MBr critical use exemption packages only estimated per acre costs
for a typical California strawberry farmer, not total costs. For this reason, we have limited information on the
potentially regulated universe. For instance, we do not know what types of farms were expected to make use
of exempted MBr. However, the ex ante analyses presented some information on overall methyl bromide
use, expressed as total strawberry acreage relying on methyl bromide, which sheds some light on the
regulated universe affected by critical use exemptions.
It appears that California farmers used slightly less MBr to grow strawberries than requested but that this
was approximately in line with EPA expectations. Growers requested MBr for use on 75-85 percent of the
California strawberry crop in the 2006-2008 seasons, falling to 50-60 percent in the 2009 and 2010 seasons.
Information on how much of this amount was expected to be met from the stockpile in any given year is not
available.92 Actual use in 2006 - 2010 from the USDA and California PIP indicate that farmers used methyl
bromide on 67 percent and 40 percent of the acres dedicated to strawberries, respectively, assuming no
growth in acreage (EPA assumed strawberry acreage stayed at 2000 levels).
3.4.2,	Baseline Information
Typically the baseline for a cost analysis would attempt to identify what emission-reducing technologies or
process changes have already been adopted by the regulated universe absent regulation. Voluntary adoption
of emission-reducing practices by industry is not typically attributed to the cost of the regulation (US EPA
2010). In the case of CUEs, this would manifest itself as switching to a MBr alternative for economic reasons.
In these cases, there would be no reason to request a critical use exemption. That said, proper
characterization of baseline conditions is still important for evaluating costs associated with switching away
91	Mayfield and Norman (2012) point to the possibility of rent seeking at the Federal and state levels by California
strawberry industry groups to avoid the costs of switching to MBr alternatives.
92	USDA data (from 2002) cited in the CUE for the 2006 season, indicate that approximately 55 percent of California
strawberry acreage used methyl bromide at the time. These data inform the assumptions for the 2007 - 2009
season CUEs, as more recent data were unavailable.
84-

-------
from MBr use. In particular, estimates of yield loss associated with alternatives are predicated on
assumptions about strawberry yields when using methyl bromide.
EPA's ex ante MBr baseline yield of 43,000 pounds per acre was based on USDA data and was only about 10
percent lower than the national average yield of about 47,000 pounds per acre between 2006 and 2010 (see
Figure 3.2). However, USDA data also indicate that California strawberry farmers were generally much more
productive than the average: The average yield for a California strawberry farmer between 2006 and 2010
was 62,000 pounds per acre. Compared to the California average, EPA estimate was about 44 percent lower
over this time period.
While using the national average underestimates baseline yields for the "typical" California farmer, it does
not affect the bottom-line financial assessment since it affects operating costs and gross revenues equally
(i.e., thus cancelling out its effect). It is worth noting that our ability to draw conclusions about baseline yields
is limited since we only have state and national averages. We have no information on how yields vary by
farmer. It is possible that farmers seeking critical use exemptions are less productive on average. For
instance, yields may be lower or production costs higher due to hilly terrain, complicating the transition away
from methyl bromide.
' ; Vfethck,\ ."Compliance
A key input into estimating the cost of a regulation are the types of technologies or approaches used to
comply. In the case of critical use exemptions, we evaluate the available evidence on the use of MBr
alternatives in the California strawberry sector for the 2006-2010 growing seasons. Were these alternatives
used as frequently as expected? Did any new alternatives become available that were not anticipated by EPA
at the time of the ex ante analysis? We also assess the rate of MBr application for those that continue to use
it since it is possible that farmers found a way to use less than anticipated. Finally, we examine the role of
state regulatory restrictions in slowing the transition to some MBr substitutes, which is also discussed in the
CUEs.
Use of Methyl Bromide Alternatives, Recall that EPA analyzed three alternatives to methyl bromide in its 2006
-2010 CUE nomination packages, 1,3-D + PIC, PIC + MS, and MS alone. It identified 1,3-D + PIC as the lowest
85 -

-------
Figure 3.2: Fresh Strawberry Yield per Acre in the United States and California, 2003 - 2010
United States
California
80,000
70,000





1
i
i
i
A
~

1
li
n m
p
i n
r
i n
r
i
i
i
30,000
2003 2004 2005 2006 2007 2008 2009 2010
Source: http://usda.mann!ib.cornell.edu/MannUsda/viewDocumentlnfo.do?documentlD=1381; See -
table04.xls. -
cost MBr alternative. Ex post data confirm that 1,3-D + PIC was the most commonly used alternative to
methyl bromide for strawberry production in California over this time period.
According to the NASS Agricultural Chemical Usage Vegetables Survey, PIC was used on about 17,600
California acres of strawberries in 2000. By 2006, this amount had increased to 18,300 acres.
Dichloropropene (1,3-D) was not separately identified in the 2000 USDA survey. In 2006, the USDA reported
that it was used on 6,400 acres of strawberries in California. The California PIP tracks the use of specific
products, so that it is possible to eliminate double counting (PIC is used alone and in combination with both
1,3-D and MBr). In 2000, 1,3-D + PIC and PIC alone were rarely used by California strawberry farmers. Fewer
than 500 acres were treated with one of a variety of possible formulations. By 2006, nearly 10,000 acres were
reportedly treated with 1,3-D + PIC, while another 1,700 acres were treated with chloropicrin in 96 and 100
percent formulations. Acreage treated with 1,3-D+ PIC rose to more almost 16,000 acres in 2010, while the
amount treated with chloropicrin grew to more than 4,700 acres.
Metam sodium use by California strawberry growers also increased, though it was still not widely used. In
2000, only 313 acres were treated with metam sodium. This increased to 1,500 acres by 2006 (USDA reports
a similar estimate of 2,100 acres in 2006) and 2,600 acres by 2010.
86-

-------
It is also possible that other alternatives not analyzed by EPA in the CUEs have since become available. As of
March 2011, 10 methyl bromide alternatives were registered at the Federal level for use in the United States
(see Table 3.8). The alternatives analyzed in the CUE nomination packages for the 2006 -2010 planting
seasons are highlighted in dark grey. Alternatives that were recognized at the time of the CUE request but
either not analyzed or not yet registered at the Federal level are highlighted in light grey. Of these, three -
dazomet, dimethyl disulfide, and methyl iodide - have been registered since the time that analysis was
originally conducted.93 The UNEP (2010) also notes that several chemicals that showed initial promise were
no longer considered viable alternatives to methyl bromide, such as propargyl bromide and sodium azide.94
However, federal-level registration is not sufficient for use: fumigants must also be approved via a state level
registration process. California is particularly strict in this regard. Of the chemicals listed in Table 3.8,1,3-
dichloropropene - with or without chloropicrin-, chloropicrin, metam sodium, and dazomet were registered
for use in California as of 2010.95
The MBTOC observed in its 2010 assessment report that much progress has been made in replacing methyl
bromide in pre-plant uses, "particularly due to improved performance of new formulations of existing
chemical fumigants (e.g., 1,3-D + PIC, PIC alone, metam sodium) and new fumigants (e.g., methyl iodide,
dimethyl disulfide), but also due to increased uptake of non-chemical alternatives." The California PIP data
demonstrate that only three potential chemical alternatives to methyl bromide were used in California
between 2006 and 2010,1,3-D, PIC, and metam sodium, and that strawberry farmers generally did not
recombine them in new or novel ways (for instance, they did not utilize the three-way fumigant system of
1,3-D + PIC + MS, increasingly common in Florida).
Methyl iodide (also called iodomethane) has long been recognized as a "near perfect substitute" for methyl
bromide, meaning it results in little or no yield loss when compared to methyl bromide (e.g., Hueth et al.
2000; Sances 2000; Goodhue et al. 2004). While it was registered as a fumigant in the United States in 2007,
California did not register methyl iodide until December 2010 (and since that time, methyl iodide has been
taken off the U.S. market by its producer and therefore is unavailable (Rubin 20 12)).96 Thus, it did not play a
role as a MBr substitute in the time frame we analyze.
93	At the federal level, methyl iodide was first registered for use as a fumigant in 2007. Dazomet was registered in
2008 for use in California only, while dimethyl disulfide was registered federally in 2010 though it is not yet
registered in California. See http://ucanr.edu/sites/PAWMBA/Nursery_Projects/Perennial/Challenges/.
94	Research into non-chemical alternatives (e.g., solarization, steam treatment, natural herbicides) has increased in
recent years (e.g., Samtani et al. 2011). Preliminary data show that some alternatives may hold promise with
regard to yield performance and weed control, but it is unclear whether results would continue to hold on a larger
scale.
95	See http://www.cdpr.ca.gov/docs/emon/vocs/vocproi/desc fieldfum mthd.htm.
96	In spite of the more favorable financial implications, recent experience suggests that public concern regarding
associated health effects may continue to limit its use, at least in the near term. For instance, see
www.panna.org/blog/ca-brings-heat-methvl-iodide and an August 30. and www.grist.org/scarv-food/2011-08-
29/methvl-iodide-mock-fumigation.
87 -

-------
Table 3.8: Federally Registered and Non-Registered Methyl Bromide Alternatives for Strawberries
Federally Registered Alternatives
Known Alternatives that Are Not
Available
Federally Registered
1,3-Dichloropropene
Furfural
Chloropicrin
Propargyl Bromide
Metam Sodium

1,3-Dichloropropene + Chloropicrin

1,3-Dichloropropene + Chloropicrin + Metam Sodium

Metam Sodium + Chloropicrin

Terbacil
Dazomet (Basamid)
Dimethyl Disulfide
Methyl iodide (iodomethane)
Dark grey: alternatives analyzed in the 2006-2010 season CUE nomination packages; Light grey: Alternatives
recognized at the time of the CUEs but either not analyzed or not yet registered; White: Currently registered
for use but not recognized in the CUEs.
If methyl iodide was once again available on the U.S. market, what role might it play going forward? While
the CUE nomination package for the 2012 season continued to assume that 1,3-D + PIC was the most
economic MBr alternative for California strawberries, methyl iodide was considered viable in the CUE for the
2013 growing season. EPA estimated that methyl iodide would be financially feasible according to the criteria
set out by the MBTOC (the per acre loss was estimated to be 6 percent of the gross revenue per acre
compared to MBr, well below the 15-20 percent threshold the MBTOC suggests) and more attractive from a
financial perspective than 1,3-D + PIC (which EPA estimated would result in a 15 percent loss in gross revenue
per acre for the 2013 growing season). The key reason for a predicted loss in gross revenue from methyl
iodide use was higher costs stemming from additional input requirements (i.e. impermeable films are
required with methyl iodide applications in California).97,98 Fennimore and Ajwa (2011) also point out that
97	Hueth et al. (2000) point out that it is difficult to predict what will occur to the price of methyl iodide as it is
becomes more widely used, as its high price at the time of publication could be due to its relatively specialized use.
98	While Noling (2005) note that virtually impermeable films were initially very expensive in the United States due
in part to high transportation costs and were sometimes subject to long delays because only a few European
manufacturers produced them, Noling et al. (2010) report that over a dozen firms manufacturer virtually
impermeable films, including several in the U.S. and Canada.
88 -

-------
totally impermeable films are approved for use with methyl iodide and that trial results show these films to
be effective at retaining the fumigant in the soil."
How Methyl Bromide Is Used, It is possible that farmers that continue to rely on methyl bromide found a way
to use less of it than anticipated while maintaining its effectiveness. Ex post evidence indicates that this has
not been the case for California strawberry farmers. In its assessment of the 2006-2010 growing seasons, EPA
assumed that MBr would be applied at a rate of 175 pounds per acre. USDA chemical usage data
demonstrates that it was actually applied to California strawberries at an average rate of about 190 pounds
per acre in 2006 (EPA underestimated the application rate by about 8 percent). USDA chemical usage data
indicate an average rate of 180 pounds per acre for methyl bromide applied to California strawberries in
2010, while California PIP data show that the average application rate for methyl bromide in 2010 was about
185 pounds per treated acre (an underestimate of 3 - 5 percent).
Regulatory and Other Restrictions, Historically, California farmers tended to use MBr + PIC at a 67:33, 57:43,
or 75:25 ratio. The nomination package for the 2012 growing season notes two factors that have complicated
California's ability to reduce the proportion of methyl bromide in a given formulation: First, for farmers who
continued to use methyl bromide, California restrictions on chloropicrin meant that the lowest formulation
likely allowed in California at the time was 57 part methyl bromide to 43 parts chloropicrin. Data from the
California PIP confirm that about 94 percent of the methyl bromide used in the 2009 and 2010 growing
seasons was formulated at 57:43 or higher. A small amount (about five percent) was available at a 50:50 or
45:55 formulation. Second, two new diseases emerged in fields treated with MBr alternatives, which resulted
in some farmers using MBr once every three years to manage these diseases. The reason for these diseases is
not known, but it has been posited that it could be the result of switching from broadcast to drip fumigation,
the lower rates of fumigant applied via drip, or fundamental differences between methyl bromide and its
alternatives.
The most recent technical assessment by the MBTOC points to a third possible reason why California farmers
did not reduce methyl bromide use at a faster rate (UNEP 2010). It notes that low permeability barrier films
allow for methyl bromide to be applied at significantly lower rates (25-50 percent less than when used with
conventional films) without loss of effectiveness or any discernible impact on yields (e.g., Noling 2005, Noling
et al. 2010).100 Planting is typically delayed, however, to allow enough of the chemical to dissipate so that
residues in the soil do not injure the plant. While required in the European Union, during our period of study
California did not allow virtually impermeable films to be used with methyl bromide due to concerns about
worker exposure to the chemical.
99	While there is far less data available to evaluate the experience of Florida strawberry farmers, they reportedly
were successful at reducing the rate at which MBr was applied by relying on virtually impermeable films (US EPA
2009). Also, methyl iodide was registered for use in Florida shortly after it was federally registered. The CUEs for
the 2011-2012 seasons note that the uptake of methyl iodide could be rapid if early adopters met with success.
100	With more permeable films, 20-90 percent of methyl bromide is allowed to escape into the atmosphere. The
wide range is due to the interaction between the chemical, soil and other environmental factors (Noling 2005).
89 -

-------
California regulations also limited the use of viable MBr alternatives. For instance, EPA (2006) reported that
township caps on the use of 1,3-dichloropropene (1,3-D) were binding for 40-62 percent of California acreage
planted in strawberries in 2005 and were one of the main reasons for granting continued critical use
exemptions to strawberry farmers.101102 In addition to township caps on 1,3-D use, Nolling and Botts (2010)
also credit uncertainty regarding authorization for practices such as virtually impermeable films and bed
shank fumigation with slowing the transition away from methyl bromide in California. In addition, the CUE
nomination packages for the 2006-2010 and subsequent seasons consistently mention restrictions on
application rates for volatile organic compounds (VOCs) such as chloropicrin and metam sodium, and buffer
zone requirements for some chemicals (e.g., 1,3-D) in California as complicating factors.103 104 Finally, farmers
cannot use a chemical until it has also been approved for use in California.105
3,4,4, Compliance Costs
In this section, we examine the ex post evidence on compliance costs, which in the case of critical use
exemptions is defined by EPA to be the net of changes in gross revenue and operating costs of switching
away from MBr to other alternatives. Recall that the EPA ex ante cost analyses focused on net operating
101	California began to allow use of 1,3-D on a restricted basis after 1995. Most townships, defined as a 36 square
mile area, were allowed to use up to 90,250 pounds annually if applied between February and November at a soil
depth of 18 inches or more. Beginning in 2002, California began to allow townships to exceed the cap by up to
twice the allowable amount. The degree to which a township is allowed to exceed the cap is proportional to how
far below the cap it has been in previous years (i.e., previous over-compliance with the cap is used as a bank), so
that on average the original limit is met. If the chemical is applied in December or January or at shallower depths,
then the cap is more restrictive. See www.cdpr.ca.gov/docs/ emon/methbrom/telone/mgmtplan.pdf and
www.cdpr.ca.gov/docs/emon/pubs/ehapreps/analvsis memos/4327 sanders.pdf.
102	Carpenter et al. (2001) estimate what demand for 1,3-D would be after the MBr phase-out absent township
restrictions. At the time of the study, it was assumed that annual township caps were strictly enforced (i.e. no
exceedances are allowed). They estimate that demand for 1,3-D would be 10 million pounds higher absent the
limits on its use, affecting 47 townships and almost 27,000 acres (about 32 percent of total acreage likely to
demand 1,3-D). The vast majority of this demand is driven by strawberries. If strawberries are not included, then
demand is estimated to be 1.5 million pounds over what is allowed, affecting 23 townships and about 6,300 acres.
103	California requires a buffer zone around an occupied structure and has maximum allowable application rates for
1,3-D and other fumigants, including methyl bromide, to protect workers health (Carter et al. 2004).
104	Carter et al. (2004) examine the combined effect of 1,3-D township caps and buffer zone requirements. When
township caps are binding, increasing buffer zones has little effect on fumigant choice. When there is a close
substitute for 1,3-D that can be used in buffer zones (e.g., chloropicrin), growers see little impact on net revenues
but when no good alternative is available returns are lower and the growers' choice of fumigant is affected.
105	This is in contrast to Florida, where there were fewer regulatory constraints on MBr alternatives. The CUE
nomination packages for the 2006-2010 seasons mention buffer zone requirements for some chemicals and
restrictions on the use of 1,3-D where karst geology is present due to the risk of groundwater contamination.
Florida also has a separate process for approving chemicals apart from the Federal registration process but allowed
the use of methyl iodide shortly after it was registered at the Federal level.
90-

-------
costs - they do not evaluate one-time or fixed costs of switching away from methyl bromide - but do
consider indirect costs such as missing the market window for selling strawberries.
3,4,4,1, Gross Revenues
The accuracy of estimates of gross revenues is driven by the ability to anticipate future strawberry prices and
changes in yields. An ex post assessment reveals that EPA's estimates of prices received by California growers
for the 2006-2010 harvest are a reasonable approximation of actual prices. However, while EPA relied on the
best data available at the time, recent literature indicates that early studies likely overestimated the yield loss
associated with switching from methyl bromide to 1,3-D + PIC. EPA also did not update its yield loss estimates
over time (e.g., it maintained the same assumption for 1,3-D+PIC throughout the CUEs for the 2006-2010 and
subsequent seasons). This would result in an overestimate of the potential loss in gross revenues ex ante, all
else equal.
Strawberry Prices, In general, the prices for strawberries used in the CUE nomination packages for the 2006-
2010 seasons are consistent with historical (2000-2003) and contemporaneous (2006-2010) prices received
by growers in California (see Table 3.9). Using data available at the time, EPA assumed strawberry prices
would be $0.69 per pound in the 2006 nomination package (assembled in 2003) and $0.79 per pound in the
2009	nominating package (assembled in 2006).106 While the prices received by strawberry producers
fluctuate from year-to-year - the average annual price was $0.65 per pound in 2006 and $1.01 per pound in
2010	(in 2006 dollars) - the average was $0.65 per pound and $0.86 per pound over the 2003-2006 and
2006-2010 time periods, respectively.
Yield Loss Associated with MBr Alternatives, Recall that the yield losses used by EPA in its ex ante cost
analyses for the 2006-2010 seasons were 14 percent and 30 percent for 1,3-D + PIC and metam sodium,
respectively. The CUE nomination packages for later growing seasons are one potential source of ex post
information. However, the yield loss estimate for 1,3-D + PIC (as well as other aspects of the analysis) were
not updated in the CUEs that occur in the time frame most relevant for this analysis. Thus, we must look to
other data sources.
A number of recent studies on yield loss of MBr alternatives for growing open field strawberries demonstrate
the possible availability of competitive substitutes. The MBTOC discusses some of this recent evidence in its
2010 assessment report, noting that 1,3-D + PIC, methyl iodide + PIC, and DMDS + PIC (as well as other
106 From USDA NASS, the national average from 2000-2003 is about $0.69 per pound in 2000 dollars. When
adjusted to 2006 dollars, it is about $0.79 per pound.

-------
Table 3.9: Strawberry Prices Received by California Growers (2000-2009)
Year
California Grower's Price
(cents per pound)
2000
0.84
2001
0.77
2002
0.59
2003
0.71
2004
0.64
2005
0.60
2006
0.65
2007
0.80
2008
0.91
2009
0.90
2010
1.01
2000-2003 (average)
0.65
2006-2010 (average)
0.86
2000-2010 (average)
0.74
Source: http://usda.mannlib.cornell.edu/MannUsda/viewDocumentlnfo.do?documentlD=1381 See
table06.xls. Prices adjusted to 2006 dollars using BLS Producer Price Index for strawberries.
chemical combinations) performed as well as MBr + PIC in field trials in the United States, Australia, and
Spain (UNEP 2010). However, it also notes that California has restricted the maximum rates at which many of
these chemicals can be used to a level lower than what was tested in the field trials. (Also, recall that DMDS
is not registered for use in California.)
Based on California field trials, Othman et al. (2009) suggest that 1,3-D + PIC (with or without a sequential
application of metam potassium), chloropicrin alone, and iodomethane + PIC all perform competitively with
67:33 MBr+PIC (measured as average total yield per acre) when used in conjunction with virtually or totally
impermeable films. Fennimore and Ajwa (2011) examine the effectiveness of 1,3-D+PIC under standard and
totally impermeable films in California. They find that fumigant retention is substantially higher with totally
impermeable films, such that less 1,3-D + PIC (i.e., about 33 percent less than under standard films) is needed
to achieve strawberry yields comparable to standard MBr + PIC applications.
The UNEP also sponsored a meta-analysis to summarize what the literature has found with regard to the
yield performance of various alternatives relative to methyl bromide for strawberries and tomatoes (Porter
et al. 2006).107 A total of 42 studies published between 1997 and 2006 were identified for strawberries, for
107 Note that while the underlying studies evaluated in the UNEP-sponsored meta-analysis have been published in
peer-reviewed journals, the meta-analysis itself has not - to our knowledge - been externally peer-reviewed.
92 -

-------
which there was information on 101 field trials. The majority of the field trials (about 90 percent) took place
prior to 2002. Twenty-eight percent of these trials were conducted in California. Because the authors could
not express yield loss across the various studies using a common unit of measure, they expressed the results
in terms of the within-study yield response of a given treatment (e.g., a given chemical formulation applied at
a similar rate using a similar method) relative to methyl bromide. They then examined variation in relative
yields of various treatments across studies.
The results show that about one-third of the treatment combinations had average relative yield estimates
"either greater or not statistically different from the estimated yield for the standard [MBr-PIC at a 67:33
ratio] by more than 5 percent," including 1,3-D + PIC and methyl iodide + PIC (see Figure 3.3).108 The estimate
for metam sodium, the other main alternative analyzed, was about a 22 percent reduction in relative yield on
average, though when combined with other chemicals (e.g. 1,3-D or PIC) it was estimated to be much more
effective.
While consistent with the finding of other studies with regard to yield loss, it is difficult to translate the
results of this study - expressed in terms of average relative yield - into specific yield loss estimates
associated with the methyl bromide alternatives analyzed by EPA in its CUE nominations for the 2006-2010
seasons. The meta-analysis looks at the variability of treatment not at actual harvest weight.109 It is also not
clear the extent to which the meta-analysis results are applicable to California farmers for two reasons. First,
more than half of the studies were conducted in Florida, Spain or New Zealand. Second, the results only
compare average relative yields derived under specific conditions. It is possible that the field trials conducted
in California are still not representative of the soils, terrain, pest profiles, and regulatory constraints of
individual farmers requesting critical use exemptions.
What can we learn from the limited ex post evidence available on yield loss with respect to the likely impact
of MBr alternatives on farmers for the 2006 - 2010 seasons? If, in fact, switching to 1,3-D + PIC would have
resulted in less yield loss than anticipated for this time period, then the ex ante and ex post estimates of the
loss in net revenue would differ by 28-87 percent for a yield loss of 10 percent or 0 percent, respectively, for
the 2006 - 2010 seasons.
108	It is worth noting that the average relative yield results for methyl iodide +PIC are much more variable across
trials than for many other alternatives. Also, many of the 22 alternatives included perbulate, an herbicide that is no
longer registered. Nine of the alternatives that fared well when compared to MBr+PIC did not include perbulate.
109	In addition, a review of the data by the Office of Pesticides Programs found that some of the individual data
points may not to have been correctly inputted into the statistical analysis.
93 -

-------
Figure 3.3: Relative Yield for MBr Alternatives for Fresh Strawberries: 1997 to 2005 -
120
CO
CO
100
o
Q-
s
80
X 60
JH
0
a>
M
C
o
a

-------
estimates in the CUE nomination packages are averages across regions applying for exemption. We compare
these sample costs to the ex ante EPA cost estimates for the 2006-2008 seasons. (Aside from updating
fumigant prices, little changed in the underlying assumptions that inform the 2009-2010 CUE cost estimates.)
A difference in average cultivation costs (which do not vary by yield) exists between the UC-Davis sample cost
and EPA estimates. UC-Davis cultivation costs were $8,500-$ll,000 per acre across the two regions (Table
3.10 presents sample costs for one of the two regions110), while EPA estimated them as $16,000 per acre (in
2006 dollars).111 No one category of costs stands out as the sole reason for this discrepancy. EPA included
$6,500 per acre in general material costs, while material costs in the 2006 UC-Davis study totaled to $5,600
per acre. Also, the EPA estimate of MBr fumigation costs per acre was about twice what was assumed in the
2006 UC-Davis study ($1,500 instead of $800 per acre).
When we match the baseline yield used in the CUE nomination packages to that in the sample cost studies
we find that per acre harvesting costs are very similar. (UC-Davis estimated harvesting costs as $13,000-
$15,000 per acre compared to EPA's ex ante estimate of $13,000.) For a strawberry farm that produces at the
California average instead of the national average, the UC-Davis researchers estimated harvesting costs to be
about $19,000-23,000 per acre across the two regions (expressed in 2006 dollars). They assumed, however,
that harvesting costs increase linearly with yield: the cost per pound of strawberries harvested did not
change.
Even with these differences, from the information we have it appears that EPA's ex ante estimates of
operating costs - defined as cultivation plus harvesting costs - are relatively close to ex post estimates (i.e.,
EPA used an estimate of $29,000 while ex post data indicate an estimate of $21,500-$26,000 per acre) for a
baseline yield similar to the national average.
MBr Alternative Fumigation Costs. Did EPA do a reasonable job of anticipating the actual fumigant costs of
the MBr alternatives analyzed? Information on the cost of using MBr alternatives is scarce. Carter et al.
(2005a) note that fumigation of strawberry fields prior to planting accounts for a substantial proportion of
total production costs - about 10 percent for bed fumigation and 20 percent for flat fumigation. While the
2010 sample cost study for the Central Coast region suggests that a grower applying 1,3-D + PIC via drip
irrigation would incur a cost of $900-$l,600 per acre (in 2006 dollars), it does not evaluate the crop budget
110	The UC-Davis sample costs include several cost categories that are excluded from the table because they were
not considered by EPA in the CUEs - for instance, the cost of cooling picked strawberries and interest on operating
capital - that add up to about $2,700-$4,400 per acre for a farm that produces at the national average. EPA
considered them to be fixed costs, which would be difficult to adequately capture as they vary widely with acreage
and the technologies adopted. As we have no ex-ante estimates to which we can compare the UC-Davis estimates,
we also do not include them here.
111	The EPA cost estimates are adjusted to 2006 dollars, assuming they were reported in nominal terms in 2003. To
translate the costs expressed on a tray per acre basis (the 2010 and 2011 sample costs are both reported this way)
to pounds per acre, we use the UC-Davis provided average of about 10 pounds per tray.
95 -

-------
Table 3.10: Operating Costs per Acre - UC-Davis Sample Cost Study for South Coast Region
Yield (pounds per acre)
Cultivation
Harvesting
44,300
50,600 56,900

63,200

69,500 75,900 82,200
8,446
13,095
8,446 8,446
14,982 16,869

8,446
18,757

8,446 8,446 8,446
20,644 22,531 24,419
Source: Takele et al. (2006). Sample Costs to Produce Strawberries: South Coast Region - Santa Barbara
County. -
using this alternative. We can gather a bit more information from the 2011 sample cost studies for the South
Coast region because they are based on using 1,3-D+PIC as the fumigant. Note that the 2011 sample costs for
Santa Barbara and San Luis Obispo Counties continue to use 90 acres as the size of a typical farm in this
region. The sample costs for Ventura County use a somewhat smaller size of 70 acres to represent a typical
farm (which is the same as the last time this county was analyzed by UC-Davis researchers in 2004).112
The direct fumigant cost for 1,3-D+PIC applied through drip irrigation was $1,000-$1,100 (adjusted from 2011
to 2006 dollars) across the two 2011 studies with the slightly higher value used for Ventura County. The
2006-2008 CUE nomination packages used a higher fumigant cost for 1,3-D + PIC - of about $1,700 per acre -
but assumed it was applied using a shank (or broadcast) system. Use of 1,3-D+PIC applied by drip irrigation
reportedly requires less of the fumigant (overall) because the delivery system is more efficient than
broadcast application (CSC 2012b).113 Unfortunately, however, the difference in the method of application
112	Ex-ante studies such as Goodhue et al. (2003) also identified 1,3-D applied alone or in combination with metam
sodium as having slightly lower costs per acre than methyl bromide based on the cost of fumigant application,
weeding, and tarp material. Likewise, Goodhue et al. (2004) find evidence based on field experiments that drip-
applied chloropicrin and 1,3-D "may potentially be economically feasible" when compared to MBr+PIC (applied at
a 67:33 ratio) for fumigating strawberry fields in California. The range of application rates over which they appear
economically feasible increases with a change in the type of tarp used (i.e., virtually impermeable films perform
better than high-density polyethelyne films). At the time of the study, it was common to apply fumigants broadly
with some of what is applied escaping from permeable tarps into the air as volatile organic compounds. The
authors note that, if instead farmers use virtually impermeable film (VIF) and apply fumigants through a drip
system, substantially less of the fumigant would escape into the atmosphere allowing them to use less of the
chemical and to lower costs. EPA estimated ex-ante that the MBr alternatives analyzed had slightly lower
operating costs per acre than MBr, which is consistent with these studies.
113	Sydorovych et al. (2006) note that applying 1,3-D + PIC by a drip system results in lower labor and machinery
costs, but somewhat higher material costs than a shank fumigation system (but this study examines its use in
North Carolina, not California).
96-

-------
that underlies the UC-Davis and EPA cost estimates renders a comparison of limited use and makes it difficult
to draw solid conclusions.114
Data indicate that 1,3-D+PIC was applied via drip irrigation with some regularity in counties where farmers
sought critical use exemptions for the 2006-2010 growing seasons. According to the CUE for the 2014
growing season (EPA 2012), 55 percent of strawberry acreage in Ventura and Oxnard counties in 2009
reportedly used a drip system for applying 1,3-D+PIC, decreasing to 30 percent in 2010 (some farmers
returned to using methyl bromide every three years to control unanticipated diseases).
Fumigant Prices, We obtain fumigant prices in California from a proprietary pesticide marketing database
available through the OPP. Nominal prices are available for methyl bromide and three of its alternatives. We
convert these to real prices using the Producer Price Index (PPI) and measure them against methyl bromide in
1999 (which receives a value of 1). Since these chemicals were often combined for use when applied to
strawberry fields and the application rates at which they were applied differ, the prices do not indicate the
relative difference in cost between the MBr alternatives evaluated by EPA, 1,3-D + PIC, MS alone, and MS +
PIC. They are still instructive, however. First, note that methyl bromide has been consistently more expensive
per pound than its alternatives (see Figure 3.4). Second, while several authors note that MBr prices will begin
to increase relative to other fumigants as exemptions decline and the stockpile is drawn down, it appears
that a more than proportional increase in the price for methyl bromide relative to its alternatives has not yet
occurred. Prices for 1,3-D and PIC both increased by slightly more than methyl bromide over this time
period.115
3,4.4,3, Indirect Costs
In the CUE requests for the 2006 - 2008 growing seasons, farmers argued that the use of MBr alternatives
would result in a planting delay of several weeks. As a result, the prices they received for the strawberry crop
would be lower than otherwise, all else equal. The main explanation offered for the delay was the use of drip
irrigation to apply 1,3-D. (According to the California Strawberry Commission, unlike with broadcast
fumigation, equipment has to be set up for the entire field before the chemicals can be applied (see EPA
114	Combined, cultivation and harvesting costs in Santa Barbara and San Luis Obispo counties are similar to UC-
Davis estimates for 2006 when using MBr ($22,000 vs. $21,500). The combined cultivation and harvesting costs for
Ventura County when 1,3-D+PIC is used are higher, almost $25,000 per acre. A recent ex-post estimate for Ventura
County using MBr is not available. The 2006-2008 CUE nomination packages used a slightly lower harvesting cost
while cultivation costs remained nearly identical when 1,3-D+PIC was used instead of MBr +PIC. Combined they
added about $28,000, $1,000 less than what was estimated when MBr was used. However, it is difficult to draw
conclusions given the difference in assumptions about how the chemical is applied (shank vs. drip).
115	Prices for dichloropicrin only begin in 2001 in the proprietary pesticide marketing data while prices are not
reported in 2000, 2004, and 2006 for metam sodium. For purposes of the Figure, metam sodium prices in
intervening years are linearly interpolated.
97 -

-------
Figure 3.4: Real Prices of Fumigants in California Relative to Methyl Bromide in 1999 -
Chloropicrin
Dichloropicrin
Metam Sodium
Methyl Bromide
Source: Proprietary pesticide marketing data, with data masked by index.
2005).)116 EPA did not analyze the effect of a missed market window on California growers in the CUEs for
the 2009-2010 growing seasons since the industry supplied no evidence that it had actually occurred.
However, it noted the possibility of a planting delay due to the use of tarps (i.e., it takes longer for the
fumigant to dissipate). Carpenter et al. (2000) also indicate that a planting delay of about a week could occur
due to phytotoxicity concerns.
In the CUE nomination packages for the 2006-2008 growing seasons, EPA assumed that missing the market
window by a few weeks would result in about a 5 percent (or 3 cent per pound) penalty in terms of foregone
revenue. This appears to be an accurate characterization of the average monthly differential in national
prices received by producers between 2005 and 2010. However, it is worth noting that, because the
harvesting season varies markedly by region in California, when a delay occurs could matter greatly from the
perspective of the individual farmer.117 Figure 3.5 illustrates the differences in the prices growers receive by
116	It could also delay planting of rotation vegetable crops planted after strawberries. The California Strawberry
Commission contends that this could result in a reduction from two rotation crops to one (US EPA 2005).
117	For instance, data indicate that the peak harvesting months in California are April-August (CSC 2009; USDA
2006). However, this masks considerable variation by region. Orange and San Diego counties produce fresh
strawberries September-early June, but peak harvest is in March-April. Peak harvest in Santa Maria and Salinas-
Watsonville is in May-June, and July-August, respectively.
98 -

-------
Figure 3.5: National Grower Prices (2006 - 2010) for Strawberries by Month
$3.25
$0.25
S*
^ ^/////
^	o	vf

-------
strawberries would no longer be grown in northern California and that production would experience a
decline in southern California. The EPA ex ante analyses assumed that the amount of California land planted
in strawberries would remain fixed at 27,600 acres (the 2000 level).
Data indicate that land dedicated to growing strawberries in California continued to increase. Figure 3.6
illustrates the longer-term trends in growth in overall strawberry acreage in California from 1970 - 2010.
Recall that the methyl bromide phase-out began in 1993 with a freeze at 1991 levels, reducing MBr until it
was no longer in use in 2005 unless an exemption was granted. There are no obvious changes in the overall
trend before or after the phase-out began, nor does growth in strawberry acreage seem impacted in the
post-2005 period. Likewise, while some strawberry growing areas increased acreage and others decreased in
strawberries over time, this trend appears unrelated to the timing of the phase-out. Perez et al. (2011) point
to strong U.S. demand for strawberries as the largest driver of growth in production, which could disguise the
incremental effect of the MBr phase-out.
When we examine the data by region, we find that the majority of the growth in strawberry acreage from
2006 - 2010 stemmed from two districts, one in the south - Santa Maria - and the other in the north - San
Joaquin-Watsonville-Salinas - both of which historically have grown a substantial portion of strawberries on
hillsides where MBr alternatives are reportedly less effective (CSC 2009). These districts were also
presumably the main beneficiaries of critical use exemptions given the technical challenges of switching to
another fumigant. Acreage dedicated to strawberries in two other southern districts - Orange-San Diego-Los
Angeles and Oxnard - remained relatively flat over this time frame.118
Organic Strawberry Production. Goodhue et al. (2005) points out that there will likely be very limited
opportunities to switch from conventional to organic strawberry production for California farmers. Data
confirm that farmers did not engage in large-scale switching to organic strawberry production in response to
the phase-out of methyl bromide. According to the California Strawberry Commission (2005), there were
about 300 acres planted in organic strawberries in California in 2001. Organic strawberry production had
increased to just under 1,000 acres in 2006 and to almost 1,800 acres in 2010. While the rate of increase was
high, the total amount of land dedicated to organic production was still relatively small, about 5 percent of
total California strawberry acreage in 2010 (California Strawberry Commission 2012a).
Strawberries imported from Mexico, According to USDA data, imports of fresh strawberries from Mexico
almost tripled from 124 million pounds in 2001 to 342 million pounds in 2010. However, domestic
consumption of strawberries also increased substantially, from 1.2 billion to 2.2 billion pounds (about an 85
percent increase). Domestic production largely kept pace with demand over the same timeframe, so that
118 In Orange County, this may have been due to increased competition for land. The CSC (2006) notes that land
development and rising property costs in Orange County resulted in lower strawberry acreage in 2006 vs. 2005.
100-

-------
Figure 3.6: California Strawberry Acreage by Major Growing Area: 1970 - 2010

45,000

40,000

35,000

30,000
U)
25,000
a>
*_
<
20,000

15,000

10,000

5,000
aO a'V sSo	qlV	q9? o[V rife eft*	^
oA oA cx* o>» op op oP c?5 op cjv dr> ofl or>	cv*
Year
& ^ $> $> ^ ^ ^ ^ ^ ^
I Orange, San Diego, and Los Angeles I Oxnard
I Santa Maria	I San Joaquin, Watsonville, and Salinas
Fresno, Manteca
Source: -http://usda.mannlib.cornell.edu/MannUsda/viewDocumentlnfo.do?documentlD=1381. See -
table05.xls. -
Mexico's share of total U.S. demand only increased from 10 to 15 percent.119 Without controlling for other
factors, it is difficult to say what role the phase-out of MBr in the United States has had in encouraging
increased imports from Mexico, but it does seem to be far less than what some in the literature had
predicted (e.g., VanSickle and NaLampang 2002) and in line with studies that pointed out various factors that
would limit growth in Mexican imports (e.g. Norman 2005).
3.5. Overall Implications and Study Limitations
Based on the ex post information available, we find that net operating costs on the typical California
strawberry farmer from banning methyl bromide for the 2006-2010 growing seasons was likely less than
119 Data are compiled by USDA. These statistics are taken from tables 12 and 16 and are available at
http://usda.mann! ib.cornell.edu/MannUsda/viewDocumentlnfo.do?documentlD=1381.
101 -

-------
anticipated ex ante (see Table 3.11). It appears that a number of viable MBr alternatives - either new
fumigants or new ways of applying existing fumigants - may have become available more quickly and
resulted in lower yield loss than initially anticipated. Using what ex post information we have on yield losses
associated with 1,3-D +PIC, for example, we find that the ex ante and ex post estimates of the loss in net
revenue may differ by 28-87 percent for the 2006 - 2010 growing seasons, all else equal. Likewise, it appears
that farmers who substituted away from methyl bromide did so without imposing large negative impacts on
production in prime California strawberry growing areas.
We also confirm the effect of California regulatory restrictions in limiting the use of various economically
competitive alternatives. For instance, adoption of 1,3-D + PIC has been slowed by township caps on its use.
Uncertainty about the effect of regulatory restrictions on the feasibility of some fumigant combinations
makes it difficult to precisely identify the extent to which yield losses may have differed from EPA's ex ante
estimates. It is also worth noting that unanticipated complications after switching away from MBr, such as
new diseases, slowed the transition to alternatives, in particular 1,3-D+PIC applied via drip irrigation.
As previously mentioned, conclusions drawn from the ex post evaluation come with significant caveats. First,
we are limited to an evaluation of per acre costs. Second, we only have information on operating costs from
crop budgets designed to reflect a typical farmer. Third, yield losses associated with various MBr alternatives
are based on field trial research. Fourth, while we have detailed annual data on what fumigants farmers used,
we do not have information with regard to other management practices such as the type of tarp used. Fifth,
the prices of specific fumigant formulations are not publically available. Finally, it is analytically challenging to
evaluate the counterfactual: what would have farmers done if they had not received the same level of MBr
exemptions for the 2006-2010 seasons? To draw more robust conclusions, we would need these types of
detailed data.
102 -

-------
Table 3.11: Summary of Findings -
Components of Cost Estimate
Source of Ex Post
Information
Assessment (Compared to Ex
Ante)
Regulated
Universe
Farm types
-
-
Strawberry acreage using MBr
USDA and CAPIP
data
Reasonable
Baseline Yields using MBr
USDA data
May be underestimate but
based on data for typical
farmer
Methods of
Compliance
MBr alternatives used (Types)
CA PIP data
Reasonable but adopt faster
than assumed; no data on
some practices
Rate of application (Usage)
USDA and CAPIP
data
MBr application - slight
under estimate
Compliance
Costs
Direct,
One-Time
Fixed Cost
Variable Cost
-
-
Direct, On-
Going Net
Cost
Gross Revenues
USDA + journal
articles + UN meta-
analysis
Strawberry prices -
reasonable
Yield loss for MBr
alternatives - likely
overestimate
Operating Costs
Crop budgets +
CUE requests +
proprietary data
Reasonable
Indirect - missed market
window
USDA data
Inconclusive; also cannot
evaluate quality trade-offs
Other
Opportunity
Costs
Conventional strawberry
production loses to imports,
organic production
CSC + USDA
Reasonable
PER ACRE
NET COSTS
Likely lower than anticipated - driven by yield loss assumptions
TOTAL COSTS
-
103 -

-------
Chapter 3 References
Bolda, M., L. Tourte, K. Klonsky, and R. De Moura. 2010. Sample Costs to Produce Strawberries. Central Coast
Region: Santa Cruz and Monterey Counties. University of California Cooperative Extension.
California Strawberry Commission. 2005. Strawberry Review: 2005 Acreage Survey Results.
California Strawberry Commission. 2006. Strawberry Review: 2006 Acreage Survey.
California Strawberry Commission. 2009. California Strawberry Revised 2009 Acreage Survey.
California Strawberry Commission. 2012a. Strawberry Review: 2012 Acreage Survey.
California Strawberry Commission. 2012b. The Facts about Methyl Bromide.
Carpenter, J., L. Gianessi, and L. Lynch. 2000. The Economic Impact of the Scheduled U.S. Phaseout of Methyl
Bromide. National Center for Food and Agricultural Policy.
Carpenter, J., L. Lynch, and T. Trout. 2001. Township Limits on 1,3-D Will Impact Adjustment to Methyl
Bromide Phase-Out. California Agriculture 55(3): 12-18.
Carter, C., J. Chalfant, R. Goodhue, K. Groves, and L. Simon. 2004. Impacts of Pesticide Regulation on the
California Strawberry Industry. Working Paper.
Carter, C., J. Chalfant, R. Goodhue, F. Han, and M. DeSantis. 2005a. The Methyl Bromide Ban: Economic
Impacts on the California Strawberry Industry. Review of Agricultural Economics 27(2): 181-197.
Carter, C., J. Chalfant, R. Goodhue, G. McKee. 2005b. Costs of 2001 Methyl Bromide Rules Estimated for
California Strawberry Industry. California Agriculture 59(1): 41-46.
Dara, S., K. Klonsky, and R. De Moura. 2011. Sample Costs to Produce Strawberries. South Coast Region:
Santa Barbara and San Luis Obispo Counties. University of California Cooperative Extension.
Daugovish, O., K. Klonsky, and R. De Moura. 2011. Sample Costs to Produce Strawberries. South Coast
Region: Ventura County. University of California Cooperative Extension.
DeCanio, S., and C. Norman. 2005. Economics of the 'Critical Use' of Methyl Bromide under the Montreal
Protocol. Contemporary Economic Policy 23(3): 376-393.
Deepak, M., T. Spreen, and J. VanSickle. 1996. An Analysis of the Impact of a Ban of Methyl Bromide on the
U.S. Winter Fresh Vegetable Market. Journal of Agricultural and Applied Economics 28: 433-443.
Fennimore, S., and H. Ajwa. 2011. Totally Impermeable Film Retains Fumigants, Allowing Lower Application
Rates in Strawberry. California Agriculture 65(4): 211-215.
Ferguson, W., and J. Yee. 1997. Phasing out Registered Pesticide Uses as an Alternative to Total Bans: A Case
Study of Methyl Bromide. Journal of Agribusiness 15(1): 69-84.
104-

-------
Goodhue, R., S. Fennimore, and H. Ajwa. 2003. Economic Feasibility of Methyl Bromide Alternatives: Field-
Level Cost Analysis. White Paper.
Goodhue, R., S. Fennimore, and H. Ajwa. 2005. The Economic Importance of Methyl Bromide: Does the
California Strawberry Industry Qualify for a Critical Use Exemption from the Methyl Bromide Ban?
Review of Agricultural Economics 27(2): 198-211.
Goodhue, R., S. Fennimore, K. Klonsky, and H. Ajwa. 2004. After Methyl Bromide: The Economics of
Strawberry Production with Alternative Fumigants. Working Paper. Giannini Foundation of
Agricultural Economics.
Goodhue, R., P. Howard, and R. Howitt. 2010. Costs of Methyl Iodide Non-Registration: Economic Analysis.
Agricultural and Resource Economics Update 13(5): 5-8.
Harrington, W., R. Morgenstern, and P. Nelson. 2000. On the accuracy of regulatory cost estimates. Journal
of Policy Analysis and Management 19 (2): 297-322.
Hueth, B., B. McWilliams, D. Sunding, and D. Zilberman. 2000. Analysis of an Emerging Market: Can Methyl
Iodide Substitute for Methyl Bromide? Review of Agricultural Economics 22(1): 43-54.
Locascio S., S. Olson, C. Chase, T. Sinclair, D. Dickson, D. Mitchell, and D. Chellemi. 1999. Strawberry
production with alternatives to methyl bromide fumigation. Annual International Research
Conference on Methyl Bromide Alternatives and Emissions Reductions.
Lynch, L. 1996. Agricultural Trade and Environmental Concerns: Three Essays Exploring Pest Control,
Regulations, and Environmental Issues. Dissertation.
Mayfield, E., and C. Norman. 2012. Moving Away from Methyl Bromide: Political Economy of Pesticide
Transition for California Strawberries since 2004. Journal of Environmental Management.
Noling, J. 2005. Reducing Methyl Bromide Field Application Rates with Plastic Mulch Technology. Paper
ENY046. Institute of Food and Agricultural Science, University of Florida.
Noling, J., and D. Botts. 2010. Transitioning to Methyl Bromide Alternatives: A Current U.S. Assessment.
White Paper. Unpublished.
Noling, J., D. Botts, and A. MacRae. 2010. Alternatives to Methyl Bromide Soil Fumigation for Florida
Vegetable Production. Vegetable Production Handbook. University of Florida, IFAS Extension.
Norman, C. 2005. Potential Impacts of Imposing Methyl Bromide Phaseout on U.S. Strawberry Growers: A
Case Study of a Nomination for a Critical Use Exemption under the Montreal Protocol. Journal of
Environmental Management 75: 167-176.
Othman, M., H. Ajwa, S. Fennimore, F. Martin, K. Subbarao, G. Browne, and J. Hunzie. 2009. Strawberry
Production with Reduced Rates of Methyl Bromide Alternatives Applied under Retentive Film.
Proceedings for 2009 Annual International Research Conference on Methyl Bromide
Alternatives and Emission Reductions.
Perez, A., K. Plattner, and K. Baldwin. 2011. Fruit and Tree Nuts Outlook. FTS-347. National Agricultural
Statistical Service, U.S. Department of Agriculture.
105 -

-------
Porter, I., L. Trinder, and D. Partington. 2006. Validating the Yield Performance of Alternatives to Methyl
Bromide for Pre-Plant Fumigation. Special report commissioned by the Methyl Bromide Technology
Options Committee, Technology and Economic Assessment Panel, UNEP.
Rubin, S. 2012. Arysta to pull methyl iodide from U.S. Monterey County Weekly. March 22.
Samtani, J., H. Ajwa, J. Weber, G. Browne, S. Klose, J. Hunzie, and S. Fennimore. 2011. Evaluation of non-
fumigant alternatives to methyl bromide for weed control and crop yield in California strawberries.
Crop Protection 30(1): 45-51.
Sances, F. 2000. Conventional and organic alternatives to methyl bromide on California strawberries.
Working paper. Presented at Methyl Bromides Alternatives Conference.
Shaw, V., and K. Larson. 1999. A Meta-Analysis of Strawberry Yield Response to Pre-plant Soil Fumigation
with Combinations of Methyl Bromide-chloropicrin and Four Alternative Systems. HortScience 34:
839-845.
Spreen, T., J. VanSickle, A. Moseley, M. Deepak, and L. Mathers. 1995. Use of Methyl Bromide and the
Economic Impact of Its Proposed Ban on the Florida Fresh Fruit and Vegetable Industry. University of
Florida. Institute of Food and Agricultural Science. Bulletin 898.
Sydorovych, O., C. Safley, L. Ferguson, E. Poling, G. Fernandez, P. Brannen, D. Monks, and F. Louws. 2006.
Economic Evaluation of Methyl Bromide Alternatives for the Production of Strawberries in the
Southeastern United States. HortTechnology 16(1): 1 - 11.
Takele, E., K. Klonsky, and R. De Moura. 2006. Sample Costs to Produce Strawberries. South Coast Region:
Santa Barbara County, Santa Maria Valley. University of California Cooperative Extension.
UNEP (United Nations Environmental Programme). 1997. Montreal Protocol on Substances that Deplete the
Ozone Layer: Technology and Economic Assessment Panel. Volume II, Annex I.
UNEP. 2006. Handbook for the Montreal Protocol on Substances that Deplete the Ozone Layer. 7th Edition.
UNEP. 2010. 2010 Report of the Methyl Bromide Technical Options Committee.
U.S. Department of Agriculture. 1993. The Biologic and Economic Assessment of Methyl Bromide. National
Agricultural Pesticide Impact Assessment Program.
U.S. Department of Agriculture. 2000. Economic Implications of the Methyl Bromide Phaseout. Economic
Research Service. Agriculture Information Bulletin 756.
U.S. EPA. 2004. Methyl Bromide Critical Use Nomination for Preplant Soil Use for Strawberries Grown for
Fruit in Open Fields on Plastic Tarps. Submitted for 2006 season.
U.S. EPA. 2005. Methyl Bromide Critical Use Nomination for Preplant Soil Use for Strawberries Grown for
Fruit in Open Fields. Submitted for 2007 season.
U.S. EPA. 2006. Methyl Bromide Critical Use Nomination for Preplant Soil Use for Strawberries Grown for
Fruit in Open Fields. Submitted for 2008 season.
106-

-------
U.S. EPA. 2007. Methyl Bromide Critical Use Renomination for Preplant Soil Use (Open Field or Protected
Environment). Submitted for 2009 season.
U.S. EPA. 2008. Methyl Bromide Critical Use Renomination for Preplant Soil Use (Open Field or Protected
Environment). Submitted for the 2010 season.
U.S. EPA. 2009. Methyl Bromide Critical Use Nomination for Preplant Soil Use for Strawberry Fruit Grown in
Open Fields. Submitted for 2011 season.
U.S. EPA. 2010. Methyl Bromide Critical Use Nomination for Preplant Soil Use for Strawberry Fruit Grown in
Open Fields. Submitted for 2012 season.
U.S. EPA. 2011. Methyl Bromide Critical Use Nomination for Preplant Soil Use for Strawberry Fruit Grown in
Open Fields. Submitted for 2013 season.
U.S. EPA. 2012. Methyl Bromide Critical Use Nomination for Preplant Soil Use for Strawberry Fruit Grown in
Open Fields. Submitted for 2014 season.
VanSickle, J., C. Brewster, and T. Spreen. 2000. Impact of a Methyl Bromide Ban on the U.S. Vegetable
Industry. University of Florida, Institute of Food and Agricultural Science. Bulletin 333.
VanSickle, J., and S. NaLampang. 2002. The Impact of the Phase Out of Methyl Bromide on the U.S. Vegetable
Industry. Policy Brief 02-1. University of Florida. International Agricultural Trade and Policy Center.
VanSickle, J., S. Smith, and R. Weldon. 2009. Impacts of EPA Proposed Buffer-Zone Restrictions on
Profitability of Florida Strawberry Growers. Paper FE795. University of Florida. Institute of Food and
Agricultural Science.
107 -

-------
Appendix 3.1: Review of the Ex Ante Literature
The ex ante literature disagrees regarding the likely impact of banning methyl bromide on U.S. farmers and
the economy more generally. Initial studies tend to predict larger impacts than later studies in part because
they often evaluate an immediate and complete ban and assume no technological innovation over time. In
contrast, later studies tend to allow for the phase-out of methyl bromide over a longer time period and
account for the role of innovation. Another key difference across studies stems from assumptions regarding
Mexico's ability to rapidly increase strawberry exports to the U.S. market.120 (As a developing country,
Mexico does not have to fully phase out methyl bromide use until 2015. However, some researchers argue
that competition from Mexican imports will likely be limited due to little overlap in growing seasons, the
perishable nature of strawberries, and seasonal differences in prices.) We summarize the findings of the main
ex ante studies of the methyl bromide phase-out below.
Spreen et al. (1995) produce an extensive report on the impacts of a methyl bromide ban on Florida fruit and
vegetable growers. The authors build a partial equilibrium model of the U.S. winter vegetable market,
allowing Mexico and Texas to act as alternate suppliers, and extend a Florida grapefruit model to evaluate
the effects of a ban. The impacts analyzed are predicated on a complete and immediate national ban of MBr
use, the substitution of methyl bromide with the next best technology available as of 1993, and no
improvements in technology over time.121 The report finds that planted acreage would decrease by 43
percent as a result of the ban. Florida strawberry production would decline by almost 70 percent, while
tomato production in Florida to supply the winter market would decline by 60 percent. The total economic
impact of a ban for the state of Florida alone was estimated to be about $1 billion (including an export
multiplier). A previous study by USDA (1993) found that banning methyl bromide would result in an economic
loss to all U.S. farmers of $800 million - $1 billion. The lower estimate was predicated on the availability of a
substitute (i.e., Vorlex) that was later withdrawn from the EPA registration process. Tomatoes, peppers, and
strawberries were expected to face the largest impacts. The report notes that a phase-in of the ban would
substantially reduce predicted losses.
120	Signatories to the Montreal Protocol agreed to a certain level of payment into a multilateral fund when
they ratified the agreement. Noting that most countries have complied with promised payments into the
fund, Decanio and Norman (2005) find that the cost-per-ton of ozone depleting substances declined by
almost $600 per year purely as a function of time (2-4 percent per year) after controlling for factors such as
project scale and sector.
121	Spreen et al. (1995) discuss the known alternatives to MBr, including 1,3-D, metam sodium, and changes
in production practices such as changes in the size of the crop bed (which initial studies showed could, alone,
reduce MBr use by 33 percent), more frequent crop rotation, and changes in the formulation of MBr and PIC.
It is unclear which of these is included as an alternative to MBr and whether the options available vary by
crop in the models.
108 -

-------
UNEP (1997) updates the Spreen et al. (1995) analysis to consider the role of learning. When relatively small
improvements in technologies are incorporated into the model (through smaller impacts on yields), the
researchers find that crop production decreases by far less than originally predicted (e.g., a 22 percent
decline in U.S. tomato production instead of 60 percent). Cost impacts also are mitigated: Spreen et al (1995)
estimated a loss in revenues to farmers of almost $625 million, while the UNEP analysis lowers the loss in
revenues to $300 million.
VanSickle et al. (2000) combine a full-year version of the model for winter vegetables used in Spreen et al.
(1995) with new information to re-evaluate the impact of a MBr ban on the 1993-1994 season. They note
that research has yielded better information on alternatives than was available in the mid-1990s. Their
results indicate that impacts would be largest for strawberry farmers with almost $200 million in lost
revenues. The authors predict that strawberries will no longer be grown in northern California and that
production in southern California will decline slightly, while production in Florida will increase. In aggregate,
this results in about a 10 percent decline in California's share in the U.S. strawberry market. The authors do
not account for the possibility that Mexico could enter the strawberry market in seasons where it has not
previously done so. In total, growers in the United States are expected to see an aggregate loss of revenue of
$264 million with some areas of the country - such as South Carolina and Texas - benefiting slightly and
California and Florida being most heavily impacted (each experience about a $218 million loss in revenues).
Consumer surplus is expected to decline by about $110 million as a result of lower production and higher
prices.
VanSickle and NaLampang (2002) use this same model to estimate the impact of phasing out methyl
bromide, as opposed to an outright and immediate ban (the focus of Van Sickle et al. 2000). In particular,
they evaluate the model's ability to correctly predict the effect of the 50 percent reduction in MBr use
between 1991 and 2000 as required by the Montreal Protocol. Once they have confirmed the broad accuracy
of the model with regard to production trends, they use it to project the impacts of a further reduction in use
between 2000 and 2005 (when complete phase-out is to have occurred). They find that the largest impacts
are expected in the strawberry market, where the authors predict that production will decline by about 20
percent and revenues will decrease on net by about $140 million. When comparing these results with the
older Van Sickle et al. study, they find that the phase-out delays a substantial portion of the impact
associated with an outright ban. They also note the use of new technologies that enable farmers to maintain
the effectiveness of MBr while using less of it per acre.
Lynch (1996) also examines the impact of a U.S. ban on methyl bromide for growing strawberries and
tomatoes on consumer and producer surplus, based on the assumption that in 2001 methyl bromide
production and imports will cease. She builds a regionally disaggregated model with fixed proportions
technology122 that treats prices as endogenous. She finds that a ban on methyl bromide use for growing
122 This technology assumption allows the author to assume away any cross-price elasticity between crops so
that the price of a commodity is a function only of its own quantity. It is a typical assumption applied by
Spreen, VanSickle, and others when they use a fixed proportion or Leontief cost curve.
109 -

-------
strawberries would result in a decline in U.S. producer welfare of about $314 million and U.S. consumer
welfare of about $70 million. Mexican producers would benefit by about $90 million. Mexican producers are
expected to increase methyl bromide use, but by a relatively small amount compared to what was used by
U.S. growers. It is also worth noting that the impacts on the strawberry markets depend to some degree on
what is assumed about Mexico's ability to respond to U.S. demand. If Mexico cannot adjust quickly enough,
then the author expects much higher agricultural prices.
Ferguson and Yee (1997) examine the short-run effect of a ban on methyl bromide use on farmer net
revenues and consumer surplus due to changes in production costs and yields. They find that a ban will
result in gains to growers that did not rely on methyl bromide prior to the ban, a mix of losses and gains to
growers that use methyl bromide that varies by crop based on the price elasticity of demand, and the
availability and cost of MBr alternatives. As with previous studies, cross-price elasticities are assumed to be
zero. Imports were accounted for in the case of three crops where it was deemed possible that they could
increase in the short term: strawberries, tomatoes, and tobacco. Relying on USDA production, price, and
acreage data for 21 different crops and demand elasticities from the literature, they estimate an annual
increase in production costs of $26 million, almost a third of which is borne by tomato growers.123 In
aggregate, the authors estimate a short-term welfare loss due to banning methyl bromide of $1 billion due to
reduced production and changes in prices. The authors point to the wide variation in welfare effects by crop
as justification for a gradual phase-out of methyl bromide instead of an outright ban. Peppers, tomatoes, and
strawberries all rank in the middle with regard to the estimated economic effect of methyl bromide use on a
per pound basis (ranging from about $19 - $30 per pound).124
Carpenter et al. (2000) conduct detailed crop-specific analyses for the National Center for Food and
Agricultural Policy (NCFAP) to evaluate the economic impacts of banning MBr use in agriculture immediately.
They begin by surveying the literature to identify the next best feasible alternative to methyl bromide from a
suite of known technologies. Estimated yield and costs effects of switching to this alternative are used as an
input into a regionally disaggregated, fixed proportions economic model to estimate changes in producer and
consumer surplus. Consumer surplus declines by $160 million due to higher prices and lower availability of
particular fruits and vegetables, with 75 percent of the decline stemming from strawberries. The model does
123	They also note that yield declines are expected to be particularly large for fresh strawberries and
tomatoes due to the limited availability of good substitutes for MBr.
124	Deepak et al. (1996) evaluate the economic impact of a MBr ban on the winter market in the United States
for six major fresh vegetables, including tomatoes and peppers. They focus on the effects of the ban on
Florida farm revenues, accounting for competition from Mexico and, to a limited extent, Texas. Using fixed
proportions technology on the supply side, they build a spatially explicit mathematical programming model
to solve for acreage planted, and market clearing prices and quantities. A MBr ban was simulated through a
loss in yield. Results suggest that a ban would eliminate or reduce production of several commodities in
Florida with Mexico making up much of the difference in lost supply. For instance, the authors project that
tomato acreage in Mexico would double as a result of the ban. The authors estimate that revenues of Florida
farmers will decline by 53 percent, while prices will increase by 1 -11 percent depending on the particular
wholesale market.
110-

-------
not predict much of an acreage response for many producers: Higher prices allow many growers to remain
profitable in spite of increased costs. On net, producers see a decrease in revenues of about $77 million. The
USDA (2000) points out that impacts in the NCFAP study are likely overstated to some degree - particularly as
one goes further out in time - because the authors assume that there are no improvements in technology, no
new MBr alternatives available than those currently on the market, and no exemptions granted going
forward. Even with possible overstatement of impacts, the USDA (2000) notes that the estimates of the
impact of a MBr ban by Carpenter et al. (2000) are substantially lower than earlier estimates by the USDA
(1993). The USDA (1993) estimated that banning methyl bromide use would result in $1 billion in impacts for
pre-plant uses, while Carpenter et al. (2000) estimated impacts for pre-plant uses of $400-$450 million.
Carter et al. (2005a) examined the short-term impact of the MBr ban on California strawberry farmers. Since
fresh strawberries are perishable, they assume that supply in a given season is fixed and cannot be easily
shifted into the processed strawberry market. Thus, to estimate the impact of the ban the authors only need
to know the expected reduction in strawberries harvested due to changes in yield and acreage, and the price
elasticity of demand. The authors evaluate a wide range of yield and acreage changes based on interviews
with farmers and field trial data, but consider the most likely scenario to be a decline in acreage of about 10
percent (over about a five-year period) and a decline in yields of 10-15 percent. Using a range of price
elasticities from the literature that range from -1.2 to -2.8 (with a "best" estimate of-1.9), they estimate that
industry revenue would decline by 6 - 17 percent. When the full distribution is taken into account, revenue is
estimated to decline by about 12 percent, on average (with a 90 percent probability that the loss is between
4 and 21 percent). These estimates do not account for the possibility that farmers use land previously
dedicated to strawberries to grow other crops, which would result in some additional revenue.
Carter et al. (2005a) also note that California competes with Florida and Mexico during the winter months,
but that by mid-March only California continues to supply fresh strawberries to the U.S. market due to
warmer temperatures that affect fruit quality in these other regions. How these markets interact is an
important consideration for estimating the national impact of a ban, particularly since California has its own
process for registering MBr alternatives. If Mexico can completely compensate for the decline in domestic
production, then strawberry prices would remain unchanged (instead of increasing), which would increase
the impact on U.S. strawberry farmers (but impact consumers less). The authors see such a dramatic increase
in Mexican exports as unlikely.
Norman (2005) examines the costs to U.S. strawberry growers of switching to MBr alternatives without any
exemptions, arguing that farmers will face much smaller net costs as a result of the ban than what growers
have suggested in their critical use exemption requests based on production costs alone. For instance, the
critical use exemption nomination for California strawberry growers for the 2006 growing season estimated
an overall loss of $1,600 to $4,000 per hectare due to lower yields and higher production costs. Norman
notes that this translates to 20-57 percent of net returns if market effects are not taken into account.
However, she finds that limited price responsiveness by consumers means that much of the cost of the ban
111 -

-------
will be passed on in the form of higher prices.125 Using price elasticities from the literature, Norman (2005)
finds that producers are expected to pass along about 75 percent of the increase in the cost of fumigation to
consumers, reducing farmer losses to $400 - $1,000 per hectare or 5 - 14 percent of net revenues.126 She also
points out that with an increase in the cost of fumigation, growers will seek to substitute toward other inputs
to further reduce the cost of the ban (e.g., while many papers start from a fixed proportions supply curve,
this may not be a valid assumption). Similar to Carter et al. (2005a), Norman (2005) argues that competition
from Mexican imports will likely be limited. She points to several reasons why this is expected to be the case:
little overlap between U.S. and Mexico growing seasons, the perishable nature of strawberries, and seasonal
differences in prices. Norman finds that seasonal variations in strawberry prices are much larger than the
additional costs from phasing out MBr use, making it likely that U.S. farmers will retain a competitive
advantage during the peak domestic growing season.127
Goodhue et al. (2005) evaluate whether California strawberries qualify for a critical use exemption according
to the criteria in the Montreal Protocol (i.e., lack of available alternatives would cause a significant market
disruption and/or no technically or economically feasible alternatives that also meet health and safety
standards). In evaluating the economic impacts of no longer using methyl bromide, the authors considered
three alternatives: 1,3-D, chloropicrin, and metam sodium.128 They do not evaluate possible changes in crop
production practices, such as more integrated pest management techniques or conversion to organic
production.129 Data were taken from field experiments that generated material and weed control costs for
methyl bromide and its alternatives. As a result, differences in application costs and effects on yields are not
considered. The effect of changes in demand for methyl bromide substitutes on fumigant prices and of costs
on total strawberry acreage are also not considered, though the authors acknowledge that these types of
effects are likely. Whether an alternative is technically feasible will vary by soil type, climate, and other
factors, but for this analysis the authors assume growers have identical production costs to conduct a break-
even analysis under different yield loss assumptions. In other words, they evaluate how much price and/or
acreage would need to change for farmers to break even using a given methyl bromide alternative. They find
125	Decanio and Norman (2005) note that demand is fairly price inelastic for most fruits and vegetables.
126	Norman (2005) calculates that a cost increase of $2,800 per hectare would translate to a price increase of
about $0.50 annually for the average U.S. household. However, price increases are most likely to occur during
months when imports from Mexico are less available, which is also when strawberry prices tend to be the
lowest.
127	In addition, Mayfield and Norman (2011) point out that Mexico consumed less MBr than it was allowed
under the Montreal Protocol in 2008. Mexico plans to expedite its phase out such that MBr is no longer in
use by 2012, three years earlier than required, by switching to methyl iodide.
128	While not analyzed, they note that methyl iodide and propargyl bromide could be competitive alternatives
in the future if they are successfully registered in the United States and California.
129	The authors view the opportunities for switching to organic production as limited, due to the substantially
higher hand weeding costs, lower yields, and land and planting requirements to qualify as organic. In
addition, large shifts into organic production would inevitably have price effects.
112 -

-------
that for the most likely yield declines (10-15 percent), prices would have to increase by 13 - 23 percent for
profits to be unaffected, while acreage would have to decline by 13 - 34 percent.
Finally, Carter et al. (2005b) evaluate the impact on strawberry farmers of additional buffer zone restrictions
and notification requirements for MBr fumigation put into place in California in 2001.130 While not a study of
the impacts of a national MBr ban, it is indicative of the way farmers may adapt to use restrictions. The
authors find that for some acreage farmers no longer grew strawberries and instead switched to less valuable
crops. Farms that bordered non-agricultural uses were most affected - they had larger amounts of acreage
where strawberries could no longer be grown (assuming application rates and other factors remained
unchanged). Smaller fields lost a greater proportion of acreage due to buffer zone restrictions. Using UC-
Davis crop budgets combined with expert opinion and surveys of growers, the authors estimated short-run
impacts. The buffer zone requirements lengthened the amount of time it took to fumigate a field, delaying
harvest and reducing production. Fumigation costs were estimated to increase by about 40 percent due to
additional labor and equipment requirements. The authors estimated a loss to the strawberry industry due to
inability to fumigate certain pieces of land. Finally, growers that relied on bed fumigation instead of flat
fumigation were required to establish larger buffer zones due to higher application rates. This resulted in
some switching from bed to flat fumigation by farmers (flat fumigation is about $1,000 per acre more
expensive).
130 EPA finalized new restrictions on the use of many fumigants as part of the re-registration process,
including buffer zone requirements and lower maximum allowable application rates to protect air quality and
the health of workers and nearby residents. Noling et al. (2010) point out that these new requirements are
likely to spur a greater transition into less permeable plastic mulch, which allows for lower application rates
without compromising fumigant effectiveness. Most of the new requirements take effect in 2010-2011. See
VanSickle et al. (2009) for a discussion of the impacts of these new buffer-zone requirements on Florida
strawberry farmers.
113 -

-------
Appendix 3.2: The Role of the Stockpile
Agricultural users that receive a critical use exemption can also rely on the MBr stockpile. While EPA tracks
the draw-down of the stockpile and the overall amount of methyl bromide used for critical and non-critical
uses, it does not know which specific users purchase from it. Figure V-Al shows declines in the stockpile from
2003 to 2010. Experts note that, as critical use exemptions decline and the stockpile is drawn down, they
expect MBr shortages and markedly higher prices in some regions (Noling et al. 2010; Goodhue et al. 2010).
Due to a paucity of data, we are not able to say what role the stockpile played in fumigant decisions for
California strawberry growers for the 2006 - 2010 seasons.
Figure 3.A1: U.S. Methyl Bromide End-of-Year Stockpile (in metric tons): 2003 - 2010
16,422
18,000
16,000
12,994
14,000
12,000
9,974
10,000
7,941
8,000
6,000
4,000
2,000
6,458
4,271
1MB
3,064
2003 2004 2005
2006
1,803
2007 2008 2009 2010
Source: EPA website, http://www.epa.gov/ozone/mbr/otherreginfo.html
114-

-------
Chapter 4: National Primary Drinking
Water Regulation for Arsenic
Cynthia Morgan, and Nathalie Simon
On January 22, 2001, EPA published new National Primary Drinking Water Regulations for Arsenic (the
"Arsenic Rule"). EPA used sound science and the best available information to estimate the costs associated
with the rule in its benefit-cost analysis. The purpose of this paper is to examine how EPA's ex ante cost
analysis of the Arsenic Rule compares to an ex post assessment of costs. This is not an evaluation of how well
EPA conducted the ex ante analysis at the time of the rulemaking, but rather it is an examination of the key
drivers of compliance costs in an effort to make an informed judgment as to whether ex post costs are higher
or lower than the estimates of ex ante costs for this Rule. We are interested to see if actual costs diverged
from ex ante costs and, if so, what factors caused this divergence (e.g., changing market conditions,
technological innovation, etc.) as described in Chapter 1 of this report.
This case study is organized as follows. Section 4.1 describes the 2001 Arsenic Rule and the types and size of
water systems that were expected to be affected. Section 4.2 summarizes the methods EPA used to produce
ex ante compliance costs for the final rule by water system type and size (number of people served). Section
4.3 describes sources of information available to conduct an ex post cost assessment of the Arsenic Rule.
Section 4.4 presents a very limited comparison of ex ante and ex post compliance costs using data from a
limited set of demonstration projects designed to show the effectiveness of various treatment technologies
at reducing arsenic levels. And lastly, Section 5.5 summarizes the analytic challenges we faced in conducting
an ex post cost assessment for this Rule.
4.1. Impetus and Timeline for the Regulatory Action
The 2001 National Primary Drinking Water Regulations for Arsenic lowered the Maximum Contaminant Level
(MCL) for arsenic in drinking water from 50 micrograms/liter (|j.g/L) to 10 pig/L. The rule applied to 54,000
Community Water Systems (CWSs) and 20,000 other systems known as Non-Transient Non-Community
Water Systems (NTNCWSs) that serve non-residential communities (e.g., schools, churches). Water systems
had to comply with this standard by January 23, 2006. EPA estimated that approximately 3,000 CWSs and
1,100 NTNCWSs would initially not meet the 10 pig/L standard and would need to treat their drinking water
to reduce the arsenic levels. Of those systems affected, 97 percent were considered "small systems" serving
10,000 people or fewer.
115 -

-------
The Arsenic Rule was particularly important in that it was the second drinking water rule in which EPA used
the discretionary authority afforded by §1412(b)(6) of the Safe Drinking Water Act to adjust the MCL to a
level above that which is technically feasible if the benefits do not justify the costs. While the Agency initially
proposed an MCL of 5 (J-g/U EPA ultimately set the drinking water standard for arsenic at 10 (J-g/U concluding
that this level maximized health risk reduction at a cost justified by the benefits (US EPA 2001).131 The
technically feasible level for arsenic removal from water was established at 3 (ig/L.
4.2. EPA Ex Ante Cost Estimates
The costs associated with the Arsenic Rule include: 1) the costs to water systems to comply with the
standard, including treatment costs, monitoring costs and administrative costs of compliance and 2) costs to
States to implement and enforce the rule. The total annual costs of the rule were estimated at
approximately $181 million, with treatment costs comprising the bulk at about $177 million. The total costs
to CWS were estimated at approximately $172 million, 98 percent of which were expected to accrue to
systems serving populations of 1,000,000 people or less. Costs for NTNCWS were estimated to be $8.1
million.132
The cost implications for households were dependent on the size of their community water system. For
households served by small community water systems (those serving fewer than 10,000 people), the annual
increase in cost was expected to range between $38 and $327. For those served by community water
systems that serve greater than 10,000 people, the estimated annual household costs for water were
expected to increase from $0.86 to $32. The disparity in household costs between systems sizes was due to
economies of scale, with larger systems able to spread the costs they would incur over a larger customer
base.
4,2,1, Main Con	Ante Compliance Costs
4.2.1.1. Identification of Best Available Treatment Technologies
EPA's ex ante compliance cost estimates for the Arsenic Rule required the identification of the "Best Available
Technologies" (BAT) effective at removing arsenic and bringing water systems into compliance with the MCL.
In the Technologies and Costs for Removal of Arsenic from Drinking Water (EPA, 2000), the various arsenic
removal technologies under different conditions are described. These technologies include
coagulation/filtration, greensand filtration, activated alumina, ion exchange, and membrane processes such
131	Based on the available science at the time, EPA quantified and monetized health benefits associated with the
rule included expected reductions in bladder and lung cancers with estimates ranging from $140 to $198 million
($1999). However, a number of health outcomes associated with arsenic exposure remained unquantified,
including cancers of the kidney, skin, and prostate, endocrine disorders (e.g., diabetes) and other cardiovascular,
pulmonary, and neurological effects.
132	EPA also estimated total annual treatment costs by system size across CWS and for NTNCWS systems by
NTNCWS system service type (see USEPA 2001, Chapter 6).
116-

-------
as reverse osmosis. In addition to the traditional arsenic removal treatment technologies, alternative
technologies still in the experimental stages such as sulfur-modified iron, iron filings, iron oxide coated sand,
and granular ferric hydroxide are discussed. Included in the discussion of each technology are ways to
improve the effectiveness of the technology for the removal of arsenic. The impact on arsenic removal
efficiencies from factors such as pH, arsenic oxidation state, and the effect of competing ions are also
discussed for each technology. As a result of this assessment, the following technologies were identified by
EPA as BAT:
•	Modified Lime Softening
•	Modified Coagulation/Filtration
•	Ion Exchange
•	Coagulation Assisted Microfiltration
•	Oxidation Filtration (Greensand)
•	Activated Alumina
In addition to these centralized treatment technologies, EPA identified point-of-use (POU) devices as
appropriate for small systems to achieve compliance with the arsenic MCL. POU involves treatment at the
tap such as a water fountain or kitchen sink. However, the Safe Drinking Water Act requires that POU devices
be maintained by the public water system which means additional recordkeeping and maintenance costs.
The POU treatment options considered were:
•	POU Reverse Osmosis
•	POU Activated Alumina
Cost equations and the resulting cost curves for both capital and O&M costs for each of these technologies
are presented in the Technologies and Costs for Removal of Arsenic from Drinking Water (EPA, 2000) and
serve as major inputs to EPA's estimation of compliance costs. The capital cost curves are a function of the
system design flow (mgd, million gallons per day) while operating and maintenance (O&M) cost curves are a
function of the average flow (mgd) of the system. Some of these technologies generate wastes that require
disposal or pre-treatment (e.g., pre-oxidation or corrosion control) in order to be effective. The associated
costs of waste disposal and pre-oxidation were included in the costs of treatment when relevant (See
Appendix 4.2).133
With the best available technologies and their unit costs defined, EPA employed different methods to
estimate compliance costs for each of three different system categories: NTNCWSs, CWSs serving 1,000,000
133 EPA's economic analysis of the arsenic rule captured only the predicted costs of the federal regulation and did
not account for disposal costs resulting from state regulations that are more stringent than federal requirements.
117 -

-------
people or fewer, and CWSs serving more than 1,000,000 people. In the Economic Analysis (EA), EPA used a
Monte Carlo Simulation model (the Safewater XL model) to estimate compliance costs for the CWSs serving
1,000,000 or fewer people and a deterministic spreadsheet analysis to assess compliance costs for the
NTNCWSs. EPA estimated compliance costs individually using system specific information for the very large
systems (those serving more than 1,000,000 people) with baseline levels of arsenic expected to exceed the
10 pig/L MCL. Total national compliance costs were then calculated by summing the compliance costs for the
three system categories. Each methodology is discussed in more detail below.
4,2,1,2, Community Water Systems (systems serving 1,000,000 people or fewer)
To estimate compliance costs for this size category of CWSs, EPA used the Safewater XL model. The model
uses a combination of individual system data and distributional data (e.g., arsenic occurrence, number of
entry points per system) to estimate costs. The data required for Safewater XL include a list of all water
systems, system source type (groundwater or surface water), population served by the system grouped into
one of eight size categories (<100; 101-500; 501-1,000; 1,100-3,300; 3,301-10,000; 10,001-50,000; 50,001-
100,000; 100,001-1,000,000), and flow rate of the system. These data are available from EPA's Safe Drinking
Water Information System (SDWIS) which contains data on all public water systems as reported by States and
EPA Regions. Additionally, the model contains probability distributions of the data for the number of entry
points per system and the concentration of arsenic in untreated water.134
EPA estimated the number of entry points for each water system and its corresponding population size
category using data from the 1995 Community Water Supply Survey. Arsenic occurrence data are based on
EPA's "Arsenic Occurrence in Public Drinking Water Supplies" report (US EPA 2000b). Mean arsenic
distributions for each system were estimated by sampling from observed data for actual systems with the
same water source type in eight geographic regions of the country. Each system was assigned a random
concentration from the arsenic occurrence distribution. The arsenic concentration for each system was then
distributed (preserving the assumed mean) across each of the entry points in the system so that each entry
point had its own assumed arsenic concentration.
The Safewater XL model then compared the arsenic concentration at each entry point to the 10 pig/L MCL
standard. Entry points with predicted arsenic concentrations above the MCL were assumed to reduce the
site concentration to 80 percent of the MCL, while entry points with predicted arsenic concentrations below
the MCL were assumed not to employ any treatment.135 For those entry points that required treatment, the
Safewater XL model used a series of decision trees to assign a treatment technology to the entry point
appropriate for the size and type of system.136 Each decision tree assigned a probability to the application of
134 Entry points are points at which water enters a water system's distribution network; in general, groundwater systems have
more entry points than surface water systems and larger systems have more entry points than smaller systems.
135SafewaterXL calculates the percent reduction in arsenic concentration required to reduce the site concentration to 80
percent of the MCL standard (this is a safety factor that includes a 20 percent excess removal to account for system over-
design).
136 OW created sixteen decision trees: two source types for each of the eight group sizes.
118 -

-------
a specific treatment technology at a given entry point, with the probability dependent on the source water
type, population size, and effectiveness across options based on the amount of arsenic requiring mitigation.
Using the design flow and average flow of the system and the cost curves and equations developed in the
Technologies and Costs for Removal of Arsenic from Drinking Water (EPA, 2000), capital and operating and
maintenance (O&M) costs at the site level were calculated for each treatment technology. A system's
compliance cost was then determined by summing across the treated entry points in the system. By
performing this analysis for each system expected to violate the MCL, EPA calculated a national estimate of
compliance costs for CWSs.
4.2.1.3,	Non-Transient Non-Community Water Systems
For the NTNCWSs, EPA estimated compliance costs using a deterministic spreadsheet rather than the
Safewater XL model. Similar to the methodology employed for the CWSs described above, the spreadsheet
relied on the SDWIS data for information on the number of systems affected and the population served and
used the same arsenic occurrence distribution developed above. Based on the design flow of the system,
one of two treatment technologies was selected: (1) point of entry activated alumina or (2) centralized
activated alumina. Point of entry activated alumina was selected for NTNCWSs with design flows less than
2,000 gallons per day and the centralized active alumina was selected for all other systems. Capital and O&M
costs were calculated based on the treatment technology selected and the design and average flow of the
NTNCWS.
4.2.1.4.	Community Water Systems (systems serving populations of more than
1,000,000)
For each of the nation's 25 largest drinking water systems - those serving more than 1,000,000 people, EPA
developed individual compliance cost estimates using system specific information including entry point water
quality parameters, system layouts, design and average flow, and treatment facility diagrams.137 The
resulting estimates were sent to each of the utilities for review and approximately 30 percent submitted
revised cost estimates or additional arsenic occurrence data. EPA revised the cost estimates for those
systems using these additional data. Of the 25 drinking water systems, three were expected to exceed the
arsenic MCL - those located in Houston, Los Angeles and Phoenix. The cost estimates developed for these
three systems accounted for approximately 2 percent of the total compliance costs estimated for the Arsenic
Rule.
137 Some sources of these data included the Information Collection Rule, the Community Water Systems Survey, the Association
of Metropolitan Water Agencies Survey, the Safe Drinking Water Information System, the American Water Works Association
WATERSTATS Survey as well as discussions with system operators.
119 -

-------
4,2,2, Main Sources	rtainty in Ex Ante Cost Estimates
Ex ante analyses are subject to many challenges and uncertainties. Selection of the most effective mitigation
strategy depends on conditions that are specific to each system. Source of water (e.g., groundwater versus
surface water), size of system (population served), and water quality conditions vary across systems. Water
quality parameters such as pH, iron, sulfate and even the type of arsenic have implications for the
effectiveness of a given treatment technology. However, EPA lacked information on exactly which systems
would be out of compliance with the new MCL and relied on modeled outcomes. EPA based its cost
estimates for these systems on predicted mitigation strategies. Over 90 percent of compliance costs were
derived from, a regulatory cost model, SafeWater XL. Modeled outcomes by design introduce uncertainty.
Location may also affect the choice of mitigation strategy. Proximity to neighboring drinking water systems
or other alternative sources of water may favor blending or abandonment of the problem source. Further,
waste streams containing arsenic resulting from the use of some technologies may be considered hazardous
waste and subject to disposal regulations138, with some states imposing their own requirements in addition
to federal regulations. These waste disposal restrictions may further constrain the choice of technologies and
ultimately affect the associated costs. In addition, some states may require pilot testing before the
installation of a treatment technology, increasing the costs of compliance with the new MCL (EPA, 2006).
Technological innovation or regulatory or technical constraints could result in water systems using different
treatment technologies for arsenic removal than the BATs listed by EPA. The SafeWater XL Model is not able
to capture these potential exogenous factors that may influence how a water system will reduce their arsenic
concentration.
4.3, Information Available to Conduct Ex Post
Evaluation
4,3,1, Ex Post L:
Prior to and after promulgation of the Arsenic rule, a number of studies reviewing EPA's ex ante cost
estimates were prepared - some in general support of the Agency's estimates (e.g., Gurian, NDWAC 2001)
and others contesting them (e.g., Bitner et al., 2001, Frey et al. 2000). Shortly following the promulgation of
the rule, EPA engaged NDWAC in an extensive, independent review of EPA's cost analysis. In spite of the
interest the Arsenic Rule generated at the time, our search of the literature identified only two studies that
have made comparisons of ex ante and ex post costs of compliance with the arsenic rule: Gurian et al. (2006)
and Hilkert Colby et al. (2010).
Gurian et al. (2006) presents some limited comparisons of EPA's ex ante cost estimates and realized ex post
cost estimates for the Arsenic rule. Specifically, using information from the first round of EPA demonstration
138 See http://www.epa.gOv/nrmrl/pubs/600s05006/600s05006.pdf
120-

-------
projects reported in Chen et al. (2004), they make comparisons of ex ante and ex post capital costs for small
systems. A number of the demonstration projects utilized iron-based adsorptive media, an emerging
technology at the time that was not a BAT in EPA's economic analysis of the rule. Plotting the realized capital
costs for the 12 demonstration projects against EPA's cost curves for ion exchange and activated alumina,
considered the best options for small systems at the time the rule was promulgated, they find that in 10 out
of 12 cases capital costs for the demonstration projects fell below the 1999 estimates. While the
demonstration projects do provide seemingly good news related to costs experienced by small systems to
mitigate their arsenic levels, Gurian et al. caveat their results by noting potential biases embedded in the
demonstration project cost estimates (e.g., biased vendor bids, tendency toward treatment technologies
rather than non-treatment solutions, availability of additional expertise in devising a solution, etc.).
Gurian et al. also present the results of a small survey of six large water systems conducted in 2003 in which
they ask about the progress each has made in coming into compliance with the new arsenic MCL. Rather
than compare these realized costs with EPA ex ante estimates, however, they make comparisons with pre-
regulatory estimates derived and presented for these same six systems in Frey et al. 2000.
Hilkert Colby et al. (2010) perform a somewhat more comprehensive comparison of ex ante and ex post costs
in their paper looking at costs of arsenic mitigation in the state of California. With help from the California
Department of Public Health, they contacted the 43 systems in the state using treatment technologies to
mitigate arsenic levels in drinking water. Each system was asked to report on cost and performance metrics
for the technologies installed, including capital and O&M costs. They compared these reported costs with
those of 13 EPA Demonstration projects from Rounds 1 and 2 that use Adsorptive media (specifically
Bayoxide E33). In addition, they compare the realized costs with EPA's affordability threshold (i.e., the total
annual household water bill considered affordable) as well as the available expenditure margin for a revised
MCL (i.e., the remainder of the threshold amount after subtracting off estimates of annual household water
bills) reported in the economic analysis.
Although they find that the median annualized costs for California systems fall within the expected household
cost for compliance with the Arsenic Rule of $0.01-$5.05/1,000 gallons (2008$), they report that 22 percent
of the systems had annualized costs that exceeded these amounts; 19 percent had costs greater than EPA's
expenditure margin; 15 percent had costs greater than EPA's affordability threshold for drinking water.
However, in making these comparisons, they admit their assumption that the treatment technology in
operation at each location is used to treat all water sources on the property. This assumption could result in
an overestimate of costs as "not all the water for the system requires arsenic treatment." They also find that
compared to California systems using similar technologies, the selected EPA demonstration sites reported
lower median and maximum annualized costs. Specifically, compliance costs among systems in California
employing similar technologies were $0.09/1,000 gallons higher than the 13 selected EPA demonstration
projects, with the demonstration projects enjoying somewhat lower labor costs but higher media
replacement costs than California systems.
121 -

-------
4	ting Costs Ex Post
We explored several source categories for ex post cost data including publicly available data on water
systems and arsenic contaminant levels, EPA's Office of Research and Development (ORD) Demonstration
Projects, consultations with industry compliance experts as well as information provided by state authorities
and associations in areas known to have levels of arsenic in drinking water exceeding the MCL. Each of these
source types and the data uncovered in each category are described below.
4.3,2,1, Publicly Available Data
Working with Abt Associates, we identified ten sources of publicly available data collected on levels of
contaminants in U.S. drinking waters and four potential data sources on compliance costs.139 The potential
sources on arsenic contaminant levels in drinking water and ambient levels are as follows:
•	Safe Drinking Water Information System (SDWIS)
•	Arsenic Occurrence and Exposure Database (AOED)
•	Consumer Confidence Reports (CCRs)
•	National Tap Drinking Water Database (NTWQD)
•	EPA's STORET Data Warehouse - arsenic ambient levels
•	National Water Information System (NWIS) - arsenic ambient levels
•	National Water-Quality Assessment (NAWQA) Program - arsenic ambient levels
•	Community Water System Survey (CWSS)
•	National Contaminant Occurrence Database (NCOD)
•	National Environmental Public Health Tracking Network
Although not specific to arsenic, potential sources of compliance cost data include:
•	Drinking Water Infrastructure Needs Survey and Assessment (DWINSA)
•	Community Water System Survey (CWSS)
•	Drinking Water Cost Rate Data
A detailed description of each database can be found in Appendix 4.1.
A considerable amount of basic operating information on public water systems appears to be available from
SDWIS and CWSS. These data potentially could be combined with arsenic occurrence data from USGS's NWIS
and NAWQA, EPA's NCOD and STORET as well as compliance cost estimates from EPA's DWINSA. However,
the 2007 DWINSA collections information is on the systems' anticipated capital improvements and associated
needs to meet the new arsenic standard, so the focus is on anticipated projects not on actual strategies
employed. Still, the data may be useful in identifying small systems that had to address the new arsenic
standard, the treatment projects planned by those systems, and the anticipated capital cost of those
139 "Background and Data Sources for Five Selected Rules," memo from Abt Associates to Nathalie Simon, August
17, 2010. Note that this list was later augmented with additional information by EPA.
122 -

-------
projects. Because the focus of the DWINSA is on capital projects, O&M costs associated with those projects
would not be captured, not to mention some non-treatment options.
Even using the data collected in the various arsenic occurrence databases and DWINSA, gaps still remain in
the publicly-available data that prevent us from being able to produce a robust estimate of the realized costs
of complying with the Arsenic rule. These gaps include mitigation strategies pursued by each system out of
compliance with the new arsenic standard and the costs associated with installation and operation of these
technologies (O&M costs and capital expenditures).
4,3,2,2,	* in o n s tra ti o n P ro j e cts
In October 2001, EPA embarked on a project to help small community water systems (<10,000 customers)
research and develop cost-effective technologies to meet the new arsenic standard. As part of the Arsenic
Rule Implementation Research Program, EPA's ORD conducted three rounds of demonstration projects that
applied full-scale, onsite demonstrations of arsenic removal technology, process modifications and
engineering approaches for small systems.
EPA funding in combination with additional funding from Congress provided support for the three rounds of
demonstration projects from 2005-2007. Treatment technologies were selected from solicited proposals.
EPA conducted 50 arsenic removal demonstration projects in 26 states in the US. Treatment systems
selected for the projects included 28 adsorptive media (AM) systems, 18 iron removal (IR) systems (including
two systems using IR and iron addition (I A)) and coagulation/filtration (CF) systems (including four systems
using IR pretreatment followed by AM), two ion exchange (IX) systems, and one of each of the following
systems: reverse osmosis (RO), point-of-use (POU) RO, POU AM, and system/process modification. Of the 50
projects, 42 were community water systems (CWS) and eight were non-transient non-community water
systems (NTNCWS).
The report "Costs of Arsenic Removal Technologies for Small Water Systems: U.S. EPA Arsenic Removal
Technology Demonstration Program" (Wang and Chen, 2011) summarizes the cost data across all
demonstration projects grouped by the type of technology. Total capital costs and operating and
maintenance (O&M) costs are presented for each treatment system. Capital costs are broken down by
equipment, site engineering, and installation costs. Factors affecting capital costs include system flow rate,
construction material, media type and quantity, pre- and/or post-treatment requirements, and level of
instruments and controls required. The O&M costs for each treatment system are broken down by media
replacement, chemical use, electricity and labor.
Although the number of projects and types of treatment technology represented is limited, the ORD
Demonstration projects provide detailed information on the capital and O&M costs associated with select
arsenic mitigation technologies. However, due in part to the goals of the program and the use of emerging
technologies, a number of biases may be present in the data. Arsenic treatment technologies, especially
iron-based adsorptive media were in a developmental stage at the start of the Demonstration program. As
such, vendors were still developing an understanding of the effects of various aspects of water quality on
their technologies as well as techniques for mitigating these impacts. In addition, the price point for the
123 -

-------
adsorptive media was not well-established and, because of the speed at which EPA needed to implement the
demonstration program, there may not have been sufficient time to negotiate the most competitive media
prices. Generally, little to no pilot testing was conducted at Demonstration sites to optimize the design and
installation of the technologies at a given facility prior to the selection of a technology and its
implementation. On the other hand, vendors wishing to establish their technologies as cost-effective
alternatives may have offered EPA more appealing prices. Again, because the goal of the program was to
demonstrate the effectiveness of various alternative treatment technologies, non-technological treatment
alternatives were not considered and are therefore not represented in the data. However, because of the
detailed nature of the data, they nevertheless provide useful information for this exercise.
4,3,2,3, Compliance Assistance Engineering Finns
Water systems needing to respond to new standards often hire engineering firms to aid in designing and
installing appropriate water treatment systems. This was the case with some systems needing to comply
with the Arsenic Rule. As such, compliance assistance engineering firms have information on the capital cost
of projects that they support and may have professional judgment-based estimates of the operating and
maintenance costs required for the installed equipment.140 Depending on the geography covered by a
particular engineering firm, it may have access to the cost information for projects in one or more states.
With assistance from Abt, we identified and contacted seven engineering firms as potential industry experts:
Malcolm Pirnie, Wright-Pierce, Farr West, Black and Veatch, CH2MHMI, Brown and Caldwell, and Brady
Associates. To guide the collection effort, we prepared a detailed template that captured inputs to the cost
estimate methodology used by the Office of Water as well as a separate document with more general
questions on the assumptions and cost estimate framework (See Appendix 4.3). Of the seven, two
engineering firms, Malcolm Pirnie and Wright-Pierce, provided information on the technologies used by
water systems they assisted and the associated compliance costs as well as providing responses to the
general questions.141,142
Specifically, Malcolm Pirnie provided information on the technologies used by water systems and the costs
incurred to comply with the Arsenic Rule for projects on which they worked. In addition to answering
questions designed to collect feedback on the assumptions and cost estimation equations used by EPA to
estimate the costs of treatment technologies, Malcolm Pirnie provided cost information for seventeen water
systems located in California and Arizona ranging in size from 0.4 mgd (million gallons per day) to 6 mgd. The
140At the outset of the process for engaging engineering firms in this effort, firms indicated that they may have
information and insight on the costs of installing treatment technologies at specific water systems, but would
usually not have information on the operation and maintenance costs for those installations.
141	Malcolm Pirnie provided technical support to EPA during the development of the Technology and Cost
Document for the Arsenic Rule.
142	Internal review of this document raised concerns about the potential bias associated with capital cost estimates
provided by engineering firms in that they might capture other capital improvements unrelated to arsenic
mitigation.
124-

-------
treatment technologies for these systems included three ion exchange (10), one reverse osmosis (RO) and
one point-of-use reverse osmosis (POU-RO), one activated alumina (AA), five granular ferric oxide (GFO),
three granular iron media (GIM), one iron-enhanced media and one blending plan
Wright-Pierce provided cost information for two water systems which used greensand filtration as the
treatment technology. The two water systems are located in Maine - one in the town of Lisbon and the
other in the town of South Berwick. The Willow Drive Pump station in the South Berwick water district serves
a population of 3,280 while the Moody River Road Filter plant located in the Lisbon water district serves a
population of 6,250.
4,3,2,4, Independent Associations
We considered independent associations of water systems, including national, regional or those covering
specific types of water systems, as potential sources of information for this effort. To support their own
initiatives, we expected that these associations might sometimes collect information on compliance
strategies and costs from their members. Based on this possibility, we asked Abt to investigate whether
these associations would be able to share information relevant to our study.
With Abt's assistance, we identified and contacted the following four independent associations: the
Association of Metropolitan Water Agencies (AMWA), the American Water Works Association (AWWA), the
National Rural Water Association (NRWA), and the Association of State Drinking Water Administrators
(ASDWA). For the most part, these associations did not have detailed information readily available on the
compliance strategies pursued by their constituents. Nevertheless, discussions with these associations
yielded references to other entities that could have the necessary information.
Specifically, AMWA, an organization of large, publicly-owned metropolitan drinking water systems, provided
some anecdotal information on the costs of compliance with the arsenic rule for their constituents and,
further, suggested we contact the Association of California Water Agencies (ACWA). ACWA is the largest
state-wide coalition of public water agencies in the country, with nearly 450 public agency members.
Collectively, ACWA's constituents are responsible for 90 percent of the drinking water delivered in California.
ACWA had conducted a member survey on compliance with the Arsenic Rule for a different initiative that
occurred before our project launched. ACWA was able to share some of the findings of that survey with us
and pointed us to peer-reviewed publications they had sponsored using the data collected (Hilkert Colby et
al., 2010).
Even though AMWA and ACWA did not provide actual cost data, they both alleged that the costs of
complying with the new arsenic MCL were higher than EPA had estimated in its economic analysis, with
AMWA reporting that the majority of systems relied on iron-based adsorptive media - a technology that was
not yet demonstrated under field conditions at the time the arsenic rule was promulgated and therefore not
considered in the EA (correspondence with Erica Brown, AMWA 2011). AMWA also indicated that a number
of the technologies included in the EA -- activated alumina, ion exchange, greensand filtration, and reverse
osmosis - are not widely used by utilities needing to mitigate arsenic levels. Further, they claimed that there
have been a number of reports of system failures due to poor design, misrepresentations by vendors
125 -

-------
regarding the effectiveness of their technologies, the application of technologies inappropriate for specific
systems, and the application of systems that are too complex for small systems to maintain.
ACWA, on the other hand, contended that EPA's EA failed to account for additional compliance costs
imposed at the state level as a result of California's laws regulating the characterization, generation and
disposal of hazardous waste residuals resulting from the arsenic removal process (correspondence with Abby
Schneider, ACWA 2011). According to ACWA, more stringent requirements in California related to the
management of arsenic residuals were a key driver in the selection of treatment technologies and often
resulted in significantly higher compliance costs in California.143
In addition, ACWA found fault with EPA's assumption regarding the use of point-of-use (POU) devices by
small systems (those serving 500 or less service connections (ACWA 2011)). In California, use of this
technology is no longer an option for long-term, permanent treatment of arsenic due to stricter state
regulation. Effective December, 2010, POU devices are allowed in CA for a 3-year period in public water
systems serving 15-200 service connections. However, these temporary systems need to be replaced with
another treatment technology following that period, resulting in higher compliance costs for the small water
systems in that category. ACWA did not provide actual cost data to substantiate their claims.
Other independent agencies, specifically NRWA and ASDWA, were helpful in identifying other potential
sources of ex post information. Specifically, they suggested that we reach out to individual state agencies
with systems known to have exerted a great deal of effort to mitigate arsenic levels in response to the
revised MCL. In particular, they suggested we reach out to agencies in Arizona, California, Nevada, New
Mexico, and Michigan.
4.3,2,5, State Agencies
Forty nine State agencies and one tribe have primary enforcement responsibility (e.g. primacy) under the
Safe Drinking Water Act and, as such, have state-level information on the number of water systems that had
to take compliance actions in response to the Arsenic Rule. Specifically, these agencies tend to track the sizes
of the systems in question, in addition to general compliance strategy information (i.e., how many systems
complied; how many systems installed treatment equipment; and how many opted for non-treatment
compliance strategies). Although some state agencies may even have specific information on the arsenic
treatment technologies installed, they typically do not have information on their associated costs as tracking
costs is outside of their purview.
Through Abt's contact with independent agencies discussed above, we identified five states -- Arizona,
California, Michigan, Nevada, and New Mexico-where significant effort was exerted and/or much difficulty
143 EPA's economic analysis of the arsenic rule captures only the costs of the federal regulation, not the costs of
more stringent state regulations.
126-

-------
was experienced in mitigating arsenic levels in response to the new MCL for arsenic. Initial contacts with
these states yielded another 4 states with similar experiences, namely Maine, Ohio, Texas and Washington.
Before proceeding with our data gathering efforts, we compared the list of nine states against those
identified in two studies on arsenic occurrence - a study by United States Geological Survey (USGS) and a
study by Natural Resources Defense Council (NRDC). Each of these studies was carried out prior to the
effective date of the Arsenic Rule. The USGS study evaluated arsenic concentration data from ground water
sources, a subset of which were located in public water supply sources. The NRDC study examined arsenic
compliance monitoring data from ground and surface water community water systems in 25 states that
supplied the relevant data. Based on the state-level arsenic occurrence information in the USGS study and
the NRDC study, 32 states were identified where the water treatment systems were likely to have had ground
water or surface water arsenic levels above the proposed MCL when the Arsenic rule was promulgated ("high
arsenic"). We confirmed that all nine states identified through contact with state agencies and independent
associations appeared on the "high arsenic" list in at least one of these two studies.
With Abt's assistance we contacted each of the nine states and sent them both a list of general questions
related to compliance with the Arsenic MCL as well as a detailed template to give them a sense of the
information we were seeking. Abt asked the contacts to provide as much of the information contained
therein as they could about their state's experience in complying with the Arsenic MCL. Although none were
able to provide cost information, we received responses regarding the types of treatments installed from 4 of
the 9 - Maine, Michigan, Nevada and Washington.
Maine, Maine's Drinking Water Program in the Department of Health and Human Services provided some
information in response to our inquiries about what transpired in the state in response to the new arsenic
MCL but did not otherwise answer the general questions provided. In their response, they indicate that
Maine's arsenic compliance issues revolved around public water systems using groundwater and provided
some detail on the types of media installed at the various systems needing to mitigate their arsenic levels.
These are summarized in Table 4.1 below. Each of the 82 systems listed serve a population of less than
10,000 people, with 78 of the 82 serving populations of less than 1,000. As shown, the majority of systems
(67 percent) employed adsorptive media. Anion exchange, installed at 15 percent of systems, was the
second most popular compliance technology employed. They also offered, however, that adsorptive media
did not last as long as originally estimated by vendors, resulting in more frequent media replacement.
Connecting to municipal water systems and installation of new wells accounted for another 6 and 5 percent,
respectively.
127 -

-------
Table 4.1: Arsenic Mitigation Strategies Employed in Maine
Type of Treatment
Number of Systems
Percentage of Systems Needing to

Mitigating Arsenic Levels
Mitigate Arsenic Levels
Adsorptive Media
55
67
Anion Exchange
12
15
Combination of Adsorptive
2
2
Media/Anion Exchange


Reverse Osmosis
2
2
New Wells
4
5
Connected to Municipal Water
5
6
System


Blending Sources
1
1
Unresolved
1
1
TOTAL
82
99*
*Does not sum to 100% due to rounding error
Michigan. According to Michigan's Department of Environmental Quality, 116 systems in Michigan needed to
mitigate their arsenic levels. Like Maine, the majority of these systems serve populations of less than 10,000
people, with 96 of the 116 (or roughly 83 percent) serving populations of less than 1,000. Sixty-three of the
systems (or 54 percent) opted for the installation of some sort of technology with most utilizing either iron-
based adsorptive media, coagulation/filtration or manganese dioxide/greensand process (See Table 2.2).144
An additional 23 systems (20 percent) found new sources of groundwater and 9 (or 8 percent) connected to
municipal water systems. Although we do not know the extent of this problem, a major issue in Michigan
involved the disposal of arsenic laden backwash water from arsenic removal systems. Because of the high
levels of arsenic in the backwash, disposal options were limited, especially for those systems that did not
have access to a sanitary sewer. Even so, industrial pretreatment, bio-solids or NPDES concerns of the
wastewater treatment facility often precluded systems from utilizing the sanitary sewers for disposal of
backwash. Even though Michigan did not provide any cost data to substantiate this statement, they contend
that disposal of backwash "in many cases doubled the cost amount of original arsenic removal system."
Nevada, Nevada's Division of Environmental Protection (NDEP) provided responses to the general questions
we provided as well as providing some statistics on their Public Water Systems (PWSs). As of December
2010, a total of 326 PWSs were subject to the Arsenic Rule in Nevada with a total of 105 reporting levels
greater than lOpig/l. Of these, 75 were community water systems while the remaining 30 were Non-
144 Michigan did not provide detailed information regarding the frequency with which each specific technology was
installed.
128 -

-------
Table 2.2: Arsenic Mitigation Strategies Employed in Michigan
Type of Mitigation
Number of Systems
Percentage of Systems Needing

Mitigating Arsenic Levels
to Mitigate Arsenic Levels
Installation of Treatment
63
54
Technology


New Wells
23
20
Connected to Municipal Water
9
8
System


Blending Sources
1
1
Unresolved
14
12
Other
6
5
TOTAL
116
100
Transient Non-Community Systems. Although 62 of the 105 (or 59 percent) achieved compliance by
December 2010, 64 systems were granted state exemptions along the way allowing them more time to
comply, with 34 of the 64 receiving additional state extensions. NDEP reported that, as of December 2010,
total of 43 of the 105 have not yet achieved compliance. As in the other states, adsorptive media figured
prominently in the treatment strategies employed especially among systems without access to a sanitary
sewer for disposal of backwash. They also offered that Nevada has a pilot testing regulation in place that
may serve as something of a deterrent to the application of new innovative technologies. Essentially, it
requires that any technology that is not proven successful under similar water quality scenarios must be
subject to pilot testing prior to being implemented. As a result proven technologies may get an advantage
over alternative technologies since they may be approved without a pilot test.
Washington, In their responses to our general questions, Washington State's Office of Drinking Water
(WODW) (within its Department of Health) provided some information on the mitigation strategies utilized
the state as of 2009. Although adsorptive media figured prominently among the strategies employed (25
percent) as in the other states, the most widely used strategy was oxidation/filtration (33 percent). Non-
treatment options (including abandoning a contaminated source, drilling new wells, etc.) represented
another 17 percent of the mitigation strategies utilized with blending not far behind at 14 percent.
WODW also noted that the volume of water that could be treated by adsorbents was "greatly over
predicted." As a result, some water systems using this technology have not had the financial resources to
replace the media once exhausted.
In addition, they allege that state rules may have influenced the choice of technologies pursued in that the
state requires that treated water samples be collected on a monthly basis to test for the efficacy of
treatment. This monitoring requirement and issues regarding access to treatment devices "have been
significant barriers to implementation of POU treatment for community water systems" although the issues
were not defined in more detail by the state.
129 -

-------
4,3,2,6, Summary of Potential Sources of Cost information
Unfortunately, the data available to compare ex post and ex ante costs are very limited. Comprehensive cost
information for the treatment technologies installed or other mitigation strategies pursued by water systems
affected by the Arsenic Rule is not available. Instead, this case study makes use of ex post cost data from
EPA's ORD Demonstration Projects. A total of 50 systems across the U.S. are captured by these data - 8
NTNCWS and 42 CWS. These data represent less than one percent of the NTNCWS and less than 2 percent of
the CWSs initially expected to exceed the new standard. These data also reflect costs of treatment
technologies and do not capture the frequency of use or the costs associated with non-treatment options
such as blending or source switching. While we did obtain cost information for another nineteen water
systems from two engineering firms (Malcolm Pirnie and Wright Pierce), we have opted to not compare the
reported realized costs with ex ante cost estimates since we cannot verify that the reported costs are specific
to arsenic mitigation and do not capture costs associated with other unrelated activities (e.g., control of
other contaminants, system improvements, system maintenance, etc.).
Although the states and independent associations provided interesting information on arsenic mitigation
strategies employed and related shortfalls, they did not provide the detailed cost information required to
make a comparison with ex ante estimates. That said, the information relayed to us through the states and
associations reveals an interesting story and suggests some potential reasons why ex ante and ex post costs
would diverge. For instance, state regulations governing disposal of backwash contaminated with arsenic
had implications on the ex post costs.
4.4. Ex Post Assessment of Compliance Cost
4.4.1. Regulated Universe
All public water systems, which include publicly- and privately-owned CWS and NTNCWS, could potentially be
affected by the Arsenic Rule. In addition to being classified by the number of people served by a water
system (system size), public water systems are also classified by their water source: surface water vs. ground
water. EPA primarily used a December 1998 freeze of SDWIS to characterize the universe of water systems
that could potentially be affected by the Arsenic Rule. At the time of the rulemaking, there were a total of
63,984 public/private ground water systems and 11,843 public/private surface water systems that could be
potentially affect by the rule. Most of these systems were CWS - 54,352 - while the remaining 20,255 were
NTNCWS. The majority (greater than 90 percent) of the CWS serve fewer than 10,000 people.
Recall that the Arsenic Rule was promulgated in 2001 but water systems had until 2006 to meet the new
MCL. Looking at the SDWIS summary data for these years, it appears that the size of the regulated universe
has decreased from the 1998 baseline. While the differences are not substantial, decreases are apparent for
both CWSs and NTNCWSs. In 2001 there were a total of 53,783 CWSs and 20,095 NTNCWSs while in 2006
there were a total of 52,339 CWSs and 19,045 NTNCWSs. Most of the decreases in both years were for
systems that serve 500 or fewer people.
130-

-------
4,4,2. Baseline Information
EPA relied on MCL compliance monitoring data from 25 states to develop an estimate of national baseline
arsenic occurrence in CWS and NTNCWS (U.S. EPA, 2000b). When EPA was developing the Arsenic Exposure
and Occurrence Database (AEOD), they examined other arsenic data sources but each database had
limitations. Some of the databases contained old arsenic samples that were considered obsolete while others
were used as comparisons for the AEOD. Ultimately, EPA used the state compliance monitoring data
voluntarily submitted to EPA from 25 states for several reasons. First, for many of the states, the data
collected were representative of almost every ground and surface water CWS in the state in addition to many
NTNCWSs. The data sets also contained multiple samples from the individual systems that showed how
arsenic levels varied over time or across locations within the system.
EPA then used statistical techniques to estimate the arsenic concentration levels at CWSs and the percentage
of those systems that would have one source above the various MCLs. While less than one percent of surface
and groundwater systems were predicted to have an arsenic concentration greater than 50 ug/L, 27 percent
of groundwater systems and 10 percent of surface water systems were predicted to have an arsenic
concentration greater than 2 ug/L. From that, EPA estimated the number of water systems expected to
exceed various MCLs.
In their development of the baseline arsenic concentrations, EPA examined other databases such as the
National Arsenic Occurrence Survey (NAOS), the United States Geological Society (USGS) ambient ground
water arsenic databases, the National Inorganics and Radionuclides Survey (NIRS), and the Metropolitan
Water District of Southern California Survey (Metro). However, each database had a drawback. For example,
the NAOS and NRIS were not useful because they did not provide arsenic concentrations within the range
being considered by EPA for the arsenic MCL. The Metro database only had information for the larger public
water systems (those serving greater than 10,000 people). The USGS database was the most comprehensive,
collecting samples from 20,000 locations across the U.S. However, some of the samples were taken from
wells used for research or used by agriculture and industry. While the USGS database provided significant
information, the samples were not collected to inform the development of a national estimate of arsenic
concentrations in drinking water supplies. However, EPA used these databases as comparison tools to check
the arsenic concentrations predicted by the AEOD (U.S. EPA, 2000b). To the best of our knowledge, EPA has
not updated the AEOD.
' ' ; Vfetho* ,\ ."Compliance
In the Economic Analysis (EA) for the Arsenic rule, EPA presented estimates of unit costs and national system
treatment costs separately for three system categories: small and large CWSs and NTNCWSs.145 In order to
obtain these estimates, EPA made assumptions about the number and types of systems that would need to
treat their water; the type of treatment technology they would adopt; and the cost of installing and
145 The economic analysis was prepared by Abt Associates, Inc., for the Office of Water and is available here:
http://water.epa.gov/lawsregs/rulesregs/sdwa/arsenic/upload/arsenicdwrea.pdf. (US EPA 2000a).
131 -

-------
operating that technology. Ultimately, the actual compliance methods chosen by water systems depend not
only on their arsenic concentrations and the size of the system but also on location specific characteristics
(e.g., iron levels in the water, pH, etc.), treatment methods already in use, and availability of alternative
water sources.
At the time of the Arsenic Rule making, iron-based adsorptive media was in the pilot and research phase, so it
was not identified as a BAT nor was it included in EPA's compliance forecast for the cost analysis. However,
the technology's effectiveness has since been demonstrated by EPA and others, water systems can and have
used iron-based adsorptive media for arsenic mitigation. Non-treatment options such as blending, turning
off wells with high arsenic levels and drawing water from another area in the aquifer with low arsenic levels
were also used and are not considered in the EA.
While we were interested in collecting information on the treatment technologies used by water systems and
their costs, we also wanted to know whether new or modified treatment technologies have been used to
meet the arsenic standard. In particular, we were interested in determining if treatment technologies have
changed since the Arsenic Rule was promulgated. As evidenced by the technologies selected for the ORD
Demonstration Projects and responses from the compliance experts, states, and independent associations to
our inquiries, iron-based adsorptive media emerged as the preferred treatment technology for mitigating
arsenic contamination. In particular, Malcolm Pirnie indicated that adsorption to granular iron media (GIM)
has been widely used at wellheads and in POU treatment systems. They also indicated that Granular Ferric
Hydroxide or variations of this media have been used frequently.
Even though the four states that provided us information stated that the majority of their systems utilized
iron-based adsorptive media, certain BATs were also used. In Washington, oxidation/filtration was the most
used technology. This technology was also used by some systems in Michigan. Anion exchange as well as
coagulation/filtration were used by systems in Maine and Michigan. As the states indicated, factors affecting
use of adsorptive media include how the residuals or backwash water will be disposed and the frequency and
cost of media replacement. Systems that did not have access to sanitary sewers to dispose of backwash
containing arsenic residuals generated from BATs tended to use adsorptive media.
In addition to treatment technologies, Malcolm Pirnie asserted that non-treatment options such as blending
with low or arsenic free water, turning off wells with elevated levels of arsenic, or selective well screening to
draw water from regions of the aquifer with low arsenic level were also widely used. Malcolm Pirnie
provided data on one utility in Central Arizona that used a blending plan. The total treatment capital cost
reported by this utility was $15,000. The states also indicated that systems used non-treatment options that
included blending, finding new sources of groundwater and connecting to municipal water sources.
Wright Pierce, on the other hand, indicated that they did not think treatment technologies have changed
since the Arsenic Rule was promulgated. However, their responses indicated that they were most familiar
with greensand filtration. The pilot testing for their two systems showed greensand filtration to be the best
technology for removing arsenic. Wright Pierce did indicate that innovation has occurred within greensand
filtration - their two systems used Pureflow high rate media which allowed for a higher filtration rate and
fewer filters.
132 -

-------
4,4,4, Compliance Costs
The national cost estimates associated with the Arsenic Rule include the costs to the water system to meet
the new standard and the costs to the States to implement and enforce the rule. In this section we focus on
the method EPA used to estimate compliance methods used by systems and their associated costs. As
discussed earlier, EPA considered the following centralized BATs:
•	Modified Lime Softening
•	Modified Coagulation/Filtration
•	Ion Exchange
•	Coagulation Assisted Microfiltration
•	Oxidation Filtration (Greensand)
•	Activated Alumina
as well as the following POU treatments (treatment at the tap) for smaller systems:
•	POU Reverse Osmosis
•	POU Activated Alumina
Cost equations and the resulting cost curves for both capital and operating and maintenance (O&M) costs for
each of the BAT technologies are presented in the Technologies and Costs for Removal of Arsenic from
Drinking Water (EPA, 2000) and serve as major inputs to EPA's estimation of compliance costs in the EA. The
capital cost curves are a function of the system design flow (mgd, million gallons per day) while O&M cost
curves are a function of the average flow (mgd) of the system. Some of these technologies require pre-
treatment (e.g., pre-oxidation or corrosion control) in order to be effective and/or generate wastes that
require disposal. The associated costs of waste disposal and pre-oxidation were included in the costs of
treatment when relevant.146 In the EA a treatment train consisted of the technology along with any pre-
treatment and disposal required by that technology. Capital and O&M costs as well as any treatment or
waste disposal costs for each treatment train are presented in the EA to show the range of costs across the
different treatment trains to achieve the MCL assuming different initial arsenic concentrations.
The Safewater XL model was used by EPA to estimate compliance costs for individual systems. Using
statistical methods, sites within a system were assigned an arsenic concentration and for sites where this
concentration is higher than the MCL, a treatment train was assigned to the site based on the size and type of
system. Capital and O&M costs were calculated for the treatment train selected for this site. By summing
across treated sites, a system's compliance cost was estimated.
To the best of our knowledge, the majority of the BAT's listed by EPA were used by systems to comply with
the arsenic MCL. However, as evidenced by our discussions with compliance engineering firms and states,
there was widespread use by systems of iron-based adsorption media as a treatment technology. It also
appears systems were able to use a non-treatment method to comply by blending finished water with a
146 Appendix A presents the assumptions and cost curves used by EPA in the EA to estimate the costs of these BATs.
133 -

-------
source that had low arsenic levels. Unfortunately, we do not have enough data to opine on the cost-
effectiveness of adsorptive media compared to the best BAT choice for site remediation.
4.4.4.1,	Ex Post Compliance Costs
For the ex post assessment, we focus on the water system information and treatment technology costs
reported by the ORD Demonstration Projects. Using these data, we make some general comparisons with
the ex ante cost estimates. First, we consider the realized capital costs reported for each of the systems and
plot these against the predicted values generated using EPA's cost curves. In so doing, we compare ex post
costs for these systems with the predicted values. As we have access to cost information for all of the
demonstration projects, this is an extension of the work presented in Gurian et al. (2006).
Second, using information on the design flow rate for each of the systems, we estimate a pseudo ex ante
estimate using the cost curves derived by EPA for that given technology. We then compare this estimate
with the realized costs reported for each system. In this way, we attempt to determine how well the cost
curves performed. Because cost curves were not developed by EPA for all of the technologies represented in
the data, we are limited in the comparisons we can make with this methodology.
We also present the water system information and treatment technology costs reported by the two
engineering firms: Malcolm Pirnie and Wright Pierce. However, we do not make comparisons with ex ante
cost estimates since it is possible that capital and O&M costs for other activities conducted concurrently with
the arsenic mitigation are intermingled. For example, construction costs provided by the engineering firms
for some systems may include the costs of upgrades to increase the capacity of the system or replacement of
existing equipment that are unrelated to the Arsenic Rule but are performed while the system is installing a
technology to reduce arsenic. However, even with the addition of the data on these nineteen systems from
Malcolm Pirnie and Wright Pierce, our data remain too limited to draw robust conclusions on whether EPA
over or under-estimated costs associated with specific technologies.
4.4.4.2,	Total Reported Capital and O&M Costs
Adsorptive Media, For the 28 water systems that selected adsorptive media (AM) technology, seven systems
were NTNCWS and 21 systems were CWS (there are 28 water systems because Klamath Lake has three POU
AM systems). Arsenic concentrations ranged from 12.7 to 67.2 pig/L across the sites. Arsenic removal
capacity of AM is highly dependent on pH. Most AM absorb arsenic more effectively at a pH value of 5.5 to
7.5, with adsorptive capacity increasing as pH decreases. Adjusting the pH value of the water can increase
the adsorptive capacity and lower the operating costs but the additional pH control equipment increases
both the complexity of the system as well the capital cost of the system. Source water pH values ranged
from 6.9 to 9.6 across the sites. Source waters at seventeen sites had a pH value greater than 7.5, and seven
of these 17 sites adjusted the pH value of the water. Table 4.3 summarizes design flow rate, average flow
rate, total capital and O&M costs for the 28 water systems.
Iron Removal or Coagulation/Filtration, Of the 50 demonstration sites, eighteen sites used Iron Removal (IR)
or Coagulation/Filtration (CF) as the main treatment technology. Iron removal or oxidation filtration
134-

-------
processes involve passing water through a greensand filter to remove iron and arsenic. Four of the eighteen
systems that used IR also followed treatment with adsorptive media (AM) to remove iron and arsenic. The
four systems primarily used IR as protection against fouling the AM with iron. Table 4.4 summarizes the
location, technologies, design and average flow rate, total capital and O&M costs for the IR/CF water
systems. Two of the eighteen sites were Non-transient Non-Community Water Systems. Arsenic
concentrations in source waters ranged from 11.4 to 84.0 pig/L.
Other Arsenic Treatment Technologies. Table 4.5 summarizes the location, technologies, flow rates, total
capital and O&M costs on two systems which use Ion Exchange (IX), one system which used Reverse Osmosis
(RO), and two point-of-use (POU) demonstration projects. At the Klamath Falls site, eight POU AM units were
installed under a sink or inside a drinking water fountain in eight college buildings. At the Homedale site,
POU RO units were installed in nine homes. Arsenic concentrations in source waters ranged from 18.2 to
57.8 |-ig/L. The presence of co-contaminants in source waters influenced the selection of treatment
technology for the different sites.
Industry Compliance Engineering Firms, Table 4.6 summarizes the location, treatment technology, design
flow rate and total capital costs provided by Malcolm Pirnie. Six of the facilities used BAT options to reduce
arsenic levels-three ion exchange, two reverse osmosis, and one activated alumina. Seven of the utilities
used some form of an adsorption technology while one utility choose blending, a non-treatment option.
Capital costs are actual costs incurred by the utilities. Although we only report either actual or median total
capital costs, when available, Malcolm Pirnie did break down capital costs by treatment equipment and
materials, waste disposal equipment and materials, construction, land, engineering, bench and pilot testing,
permitting, and other. Malcolm Pirnie provided O&M costs for a few facilities but because it was unavailable
for most facilities, we do not report O&M costs here.
135 -

-------
Table 4.3. Summary of ORD Adsorptive Media Demonstration Sites -
State
Demonstration
Technology
Design
Average
Total
Total O&M

Location (Site ID)

Flow
Rate
(gpm)
Flow Rate
(gpm)
Capital
Costs ($)
Costs ($/kgal)
ME
Wales (WA)
Iron Modified Media
(alumina based)
14
10.4
$16,475
$22.88
$10.44
$5.52#
NH
Bow (BW)
Iron Modified Media
(silica based)
40
41
$166,050
$5.11
NH
Goffstown (GF)
Granular Ferric Oxide
10
13
$34,201
$2.34
NH
Rollinsford (RF)
Granular Ferric Oxide
120
82
$131,692
$3.59*
VT
Dummerston
(DM)
Iron Modified Media
(alumina based)
22
6.1
$14,000
$10.86
CT
Woodstock (WS)
Titanium Oxide Media
20
16.4
$51,895
no estimate**
CT
Pomfret (PF)
Iron Modified Media
(resin based)
15
9.6
$17,255
$7.67
MD
Stevensville (SV)
Granular Ferric Oxide
300
207
$211,000
$0.61
OH
Buckeye Lake (BL)
Granular Ferric Oxide
10
On demand
$27,255
no estimate**
Ml
Brown City (BC)
Granular Ferric Oxide
640
564
$305,000
no estimate**
IL
Geneseo Hills
(GE)
Granular Ferric Oxide
200
32
$139,149
no estimate**
SD
Lead (LD)
Iron Modified Media
(resin based)
75
71.5
$87,892
$0.98
TX
Alvin (AL)
Granular Ferric Oxide
150
129
$179,750
$0.61
TX
Bruni (BR)
Granular Ferric Oxide
40
40
$138,642
no estimate**
TX
Wellman (WM)
Granular Ferric Oxide
100
91
$149,221
no
estimate**
NM
Anthony (AN)
Granular Ferric Oxide
320
260
$153,000
$0.75
NM
Nambe Pueblo
(NP)
Granular Ferric Oxide
160
114
$143,113
no estimate**
NM
Taos (TA)
Granular Ferric Oxide
450
503
$296,644
no estimate**
AZ
Rimrock (RR)
Granular Ferric Oxide
45
31
$88,307
$0.86
AZ
Tohono O'odham
Nation (TN)
Granular Ferric Oxide
63
60.1
$115,306
no estimate**
AZ
Valley Vista (VV)
Iron Modified Media
(alumina based)
37
36
$228,309
$2.47
OR
Klamath Falls
(KF)a






(a)
Iron Modified Media
(resin based)
30
On demand
$55,847
no estimate**

(b)
Granular Ferric Oxide
60
On demand
$59,516
$5.37

(c)
Titanium Oxide Media
60
On demand
$73,258
no estimate**
NV
Reno(RN)
Granular Ferric
Hydroxide
350
275
$232,147
$5.69
CA
Susanville (SU)a
Iron Modified Media
(alumina based)
12
9.3
$16,930
$12.06
136-

-------
State
Demonstration
Location (Site ID)
Technology
Design
Flow
Rate
(gpm)
Average
Flow Rate
(gpm)
Total
Capital
Costs ($)
Total O&M
Costs ($/kgal)
CA
Lake Isabella (LI)
Iron Modified Media
(resin based)
50
23
$114,070
no estimate**
CA
Tehachapi (TE)
Zirconium Oxide
Media
150
79.3
$76,840
$1.16
a Non-Transient Non-Community Water Systems
#	associated with three replacement media types: A/I Complex, GFH, and CFH
*	Estimated Cost- did not replace media
** No estimate of total O&M but estimates of media replacement costs, electricity, chemicals and labor costs
are provided.
Table 4.4. Iron Removal (IR) and Coagulation/Filtration (CF) Systems
State
Demonstration
Technology
Design
Average
Total Capital
Total

Location (Site ID)

Flow Rate
Flow Rate
Costs
O&M



(gpm)
(gpm)
($)
Costs
($/kgal)
IN
Goshen (GS)a
IR + AM
25
15.2
$55,423
$2.90
IN
Fountain City (FC)a
IR
60
47
$128,118
$2.26
MN
Sauk Centre (SC)
IR
20
4
$63,547
$0.36
UT
Willard (WL)
IR + AM
30
9.3
$66,362
$1.93
Wl
Delavan (DV)
IR
45
20 (max)
$60,500
$0.26
IL
Waynesville (WV)
IR
96
84
$161,560
$0.65
MN
Climax (CM)
IR/IA
140
132
$270,530
$0.29
PA
Conneaut Lake (CL)
CF
250
153
$216,876
$0.46
MT
Three Forks (TF)
CF
250
206
$305,447
$0.18
MN
Sabin (SA)
IR
250
231
$287,159
$0.43
OH
Springfield (SF)
IR + AM
250
89
$292,252
$0.33
MN
Stewart (ST)
IR + AM
250
190
$367,838
$0.16
Ml
Sandusky(SD)
IR
340
163
$364,916
$0.27
Wl
Greenville (GV)
IR
375
285
$332,584
$0.55
DE
Felton (FE)
CF
375
263
$334,297
$0.31
Ml
Pentwater (PW)
IR/IA
400
350
$334,573
$0.17
WA
Okanogan (OK)
CF
550
538
$424,817
$0.18
LA
Arnaudville (AR)
IR
770
335
$427,407
$0.07
a Non-transient Non-Community Water Systems
IA = supplemental iron addition; AM = adsorptive media; CF = coagulation/filtration using iron salts and direct
pressure filtration, not conventional coagulation-sedimentation-filtration.
137 -

-------
Table 4.5. Other Arsenic Treatment Technologies: Ion Exchange (IX), Reverse Osmosis (RO), and Point-of-
Use (POU)	
State
Demonstration
Technology
Design
Average Flow
Total Capital
Total O&M

Location (Site ID)

Flow Rate
Rate
Costs
Costs



(gpm)
(gpm)
($)
($/kgal)
ME
Carmel (CE)a
RO
1,200 gpd
0.8 (permeate);
1.2 (reject)
$20,542
$12.89
OR
Klamath Falls (KF-
POU)a
POU AM
NA
NA
$1,216

ID
Homedale (HD)
POU RO
NA
NA
$31,877.50
$201.50/yr
(total)
ID
Fruitland (FL)
IX
250
157
$286,388
$0.62
OR
Vale (VA)
IX
540
534
$395,434
$0.35
a Non-Transient, Non-Community Water System
AM = Adsorptive media; NA = not applicable
Table 4.6. Median and Reported Values of Design Flow Rate and Total Capital Costs to Meet the Arsenic
Standard by Treatment Technology for Select Systems in California and Arizona (Malcolm Pirnie)
Type of Value
Treatment
Technology
Design Flow
Rate (mgd)
Total Capital
Costs ($)
Median
Adsorption (10)
3.6
$1,423,440
Ion Exchange(3)
5.76
ANR
Reported
Reverse Osmosis (1)
1.44
Less than $240,000
Reverse Osmosis (1)
POU
$400
Activated Alumina (1)
0.86
Less than $1,575,000
Blending Plan (1)
4.18
$15,000
ANR = available but not reported because we cannot verify that the reported costs are specific to arsenic -
mitigation. -
(#) Either number of facilities used in the median calculation or the number using a treatment technology. -
In addition, Wright-Pierce provided cost information for two water systems in Maine, both of which used -
greensand filtration as the treatment technology. The Willow Drive Pump station in the South Berwick water -
district serves a population of 3,280 and has a design flow rate of 0.792 mgd. Capital costs associated with -
this project were reported as $1, 329,798 in 2003 and O&M costs of $52,906 per year. The Moody River -
Road Filter plant serves a population of 6,250 with a design flow rate of 1 mgd. Capital costs associated with -
this project were reported as $2,582,326 in 2005 and O&M costs of $69,609 per year. -
Our only source of pre-regulatory cost information is the cost curves developed in EPA's "Technologies and -
Costs for Removal of Arsenic from Drinking Water" (US EPA 2000c). At this time we use only one source of -
post-regulatory costs: ORD Demonstration Projects. A significant share of the post-regulatory cost -
information from the ORD Demonstration Projects is on iron-based adsorptive media, a technology that was -
138 -

-------
still in the research and pilot stage at the time the Arsenic Rule was promulgated. However, as we have
learned iron-based adsorptive media were used by many systems to reduce arsenic levels.
To compare ex ante costs with our limited ex post cost data, we plot our ex post cost data against the capital
cost curves used by EPA for treatment technologies recommended for smaller systems - activated alumina,
ion exchange and greensand filtration. The capital costs from the ORD Projects are plotted in Figure 4.1 and
Figure 4.2.147 To keep the graphs visually simple, Figure 4.1 plots the capital cost data for the demonstration
projects that had a design flow rate between 0.01 mgd and 0.5 mgd while Figure 4.2 plots the data for
projects with a design flow rate greater than 0.5 mgd. The results are mixed. In 42 out of 49 demonstration
projects, realized capital costs are below the 2006 cost curve estimates for at least one of the three
technologies.148
Figure 4.1. Capital Cost Comparison by Design Flow Rate (0.01-0.5mgd) - EPA Cost Curves vs. ORD
Demonstration Projects
Capital Cost Comparison
m 400000-
co
o
o
~ 300000"
o
O
~S
'I- 200000-
03
o
"ctf
•*—>
o
l—
100000-
Q
DC
o
0
.2
3
.4
.5
Design Flow Rate (mgd)
Activated Alumina
Ion Exchange
Greensand Filtration
Total Capital Costs (2006$)
147	Total capital costs for the ORD demonstration projects were converted to 2006 dollars from the year of
construction using the Engineering News Record Construction Cost Index. See appendix for cost curve equations in
$2006.
148	Two POU ORD projects did not provide design flow rate so they are not included on the graphs.
139 -

-------
Figure 4.2. Capital Cost Comparison by Design Flow Rate (0.5-1.2 mgd) - EPA Cost Curves vs. ORD
Demonstration Projects
Capital Cost Comparison
1000000-
CD
o
o
— 800000"
CO
~ 600000"
Q.
CO
o
lo
400000-
Q
cc
o
200000-
4
.6
.8
1.2
Design Flow Rate (mgd)
	 Activated Alumina
	 Ion Exchange
Greensand Filtration
ORD Total Capital Costs
4.4.5. Comparison of Technology Costs
This section presents the actual capital costs and O&M costs compared to predicted costs obtained using the
EPA cost curves for two BAT compliance options: Ion Exchange and Greensand Filtration.149 Before
presenting these comparisons, there are a few points to note. First, there is more uncertainty surrounding
operating cost estimates than capital cost estimates because of the difficulties in separating incremental
activities related to rule compliance from general operating activities. Second, and most importantly, we do
not have enough cost data to draw robust conclusions about whether EPA over or under-estimated
149 We only compare the ORD projects that used a BAT. We do not compare the projects that used a combination
BAT and non-BAT (e.g., iron removal (IR) and AM) or a technology that was in the same class but a variation of a
BAT. For example, we do not compare ORD projects that used coagulation filtration (CF) to EPA's BAT because EPA
assumed modified coagulation/filtration and not new installation of the technology. Also Greensand filtration is
the only form of IR or CF that was a BAT. Although similar, other IR technology used by the demonstration projects
was not a BAT.
140-

-------
technology costs. We present the cost comparisons for these technologies here to simply illustrate the
evaluation we could make if we had more data on ex post technology costs.
Ion Exchange. Table 4.7 presents total capital costs (CapEx) and total O&M costs (OpEx) for the two ORD
Demonstration Projects that used Ion Exchange (IX). Using the design flow rate and average flow rate of the
systems, we use EPA's cost equations for IX reproduced in Table A5 in the Appendix 4.2 to predict the capital
and O&M costs for this technology (EPA Estimate). Column 5 represents the percentage error between these
EPA estimates and the realized costs reported by ORD Demonstration Project sites. A positive (negative)
percentage error means that the EPA estimate was higher (lower) than actual costs incurred by the individual
system.
The EPA estimates of capital costs were mixed. For the smaller system, as measured by design flow, the EPA
estimate was lower than the actual cost of the project and higher than the actual cost of the project for the
larger system. For both projects, EPA's cost curves predicted lower O&M costs than the actual project costs.
Greensand Filtration. Two community water system ORD Demonstration Projects used Greensand filtration
(GF) as a treatment technology. Table 4.8 presents total capital costs (CapEx) and total O&M costs (OpEx) for
these two systems. Using the design flow rate and the average flow rate of the systems, we use EPA's cost
equations employed in the EA for GF (see Appendix 4.2) to estimate the capital and O&M costs for this
technology (EPA Estimate). Column 5 represents the percentage error between the EPA estimate and the
costs reported by ORD Demonstration Project sites. A positive (negative) percentage error means that the
EPA estimate was higher (lower) than the actual project costs for those systems. In the case of the GF
technology, one ORD Demonstration Project had capital costs that were slightly higher than the EPA estimate
(-1 percent) while the other had capital costs that were significantly lower than projected (69 percent). For
both projects, predicted O&M cost were slightly lower than the realized cost.
4.5. Overall Implications and Study Limitations
As the introduction and the literature survey (Sections I and III) make clear, even the most credible analysis of
compliance costs (done before implementation) will vary from actual costs for a large number of reasons.
For example, in the case of arsenic, innovation, impossible to forecast, may have reduced the costs. Or, the
number of water systems exceeding the standard could be larger or smaller than predicted before the rule.
This case study was particularly challenging in that the systems affected by the new arsenic standard are
heterogeneous. In addition to the heterogeneity of sites, it is also challenging to distinguish costs
attributable to compliance with the Arsenic Rule from costs incurred by systems as a result of complying with
other regulations or to meet other needs of the system. For example, some treatment technologies, such as
ion exchange, are capable of removing other contaminants (e.g., uranium) in addition to arsenic. The portion
of the treatment cost attributable to arsenic compliance can be difficult to distinguish from the cost of
contaminants being removed for other regulations. Additionally capital costs may also include costs
associated with other projects unrelated to arsenic treatment, including upgrades that increase the overall
141 -

-------
Table 4.7. Cost Comparisons - Ion Exchange (2006$)

Design Flow/Average
Flow (mgd)
ORD Project Costs
EPA Estimate
% Error
CapEx
0.36
$311,988
$275,245
-12%

0.78
$411,632
$477,021
16%





OpEx
0.23
$55,735
$34,180
-39%

0.77
$102,258
$43,180
-58%





Table 4.8. Cost Comparisons - Greensand Filtration (2006$) -

Design Flow/Average
Flow (mgd)
ORD Project Costs
EPA Estimate
% Error
CapEx
0.14
$150,692
$149,082
-1%

0.36
$196,150
$332,473
69%





OpEx
0.12
$26,767
$19,341
-28%

0.22
$33,457
$27,139
-19%
capacity of the system or replace existing equipment at the treatment plant. Because systems may perform
other types of maintenance projects concurrent with their response to the Arsenic Rule, it can be difficult to
isolate the costs attributable to the rule. These factors all add to the analytic challenge of how to evaluate
the costs faced by systems affected by the Arsenic Rule.
With no comprehensive or even representative data on costs or mitigation strategy selected, our options
were limited. Short of conducting a survey of community water systems to gather information on treatment
methods used and the costs associated with those methods, we found no other means of collecting the
necessary data. Instead, we relied on limited information collected from compliance engineering firms and
EPA demonstration projects which have their own potential biases. For example, the ORD projects rely on
emerging technologies that were not entirely understood by the vendors. In addition, the price point for the
adsorptive media was not well-established and, because of the speed at which EPA needed to implement the
demonstration program, there may not have been sufficient time to negotiate the most competitive media
prices Generally, little to no pilot testing was conducted at demonstration sites to optimize the design and
installation of the technologies at a given facility prior to the selection of a technology and its
implementation. On the other hand, vendors wishing to establish their technologies as cost-effective
alternatives may have offered EPA more appealing prices. Again, because the goal of the program was to
demonstrate the effectiveness of various alternative treatment technologies, non-treatment alternatives
were not considered and are therefore not represented in the data. However, because of the detailed nature
of the data, they nevertheless provided useful information.
142 -

-------
While we do make comparisons of EPA predicted costs and realized costs from the ORD Demonstration
Projects, these comparisons are for illustrative purposes only. We plot all of the capital cost data from the
ORD Demonstration Projects against the cost curves for the compliance technologies recommended for
smaller systems and find that the EPA methodology overestimates capital costs in most cases, especially as
the size of the system increases (as measured by the design flow rate). We also compare EPA predicted costs
and realized costs from the four ORD Demonstration Projects for two specific BATs (ion exchange and
greensand filtration) but make no judgments. Because the number of observations in our data set is very
small compared to the number and heterogeneity of the systems affected by the Arsenic Rule, we cannot
draw any conclusions regarding EPA's technology cost estimates. Our data capture the costs of treatment
technologies for a very small percentage of systems affected by the arsenic standard and as such, our results
are not generalizable across affected systems. Instead, our illustrative comparisons offer insights into how
we might proceed if better and more comprehensive data were available.
We find that this effort illustrates the characteristics of an environmental control problem that make case
study analysis extremely difficult and expensive. Despite our best efforts, our data do not provide enough
coverage of CWSs to make any assessment of how ex post costs deviate from EPAs ex ante estimates. As
discussed below, the heterogeneity of the affected water systems presents major obstacles to comparing ex
post and ex ante costs. These factors and our lessons learned from doing this case study should be
considered when designing future case studies assessing ex ante and ex post costs. We do offer limited
comparisons of predicted cost estimates obtained using methodologies employed by EPA in the EA with the
data we collected on realized compliance costs for the 50 systems.
Chapter 4 References:
Bitner, K., B. Thomson, and J. Chwirka. 2001. The cost of compliance with a new drinking water standard for
arsenic in New Mexico. New Mexico Geology 23:10-12.
Chen, A. S. C., L. Wang, J. L. Oxenham, W. E. Condit. 2004. Capital Costs of Arsenic Removal Technologies,
U.S. EPA Arsenic Removal Technology Demonstration Program Round 1. EPA/600/R-04/201.
Fisher, D. C., C. D. Whitehead and M. Melody. 2006. National and Regional Water and Wastewater Rates For
Use in Cost-Benefit Models and Evaluations of Water Efficiency Programs. Lawrence Berkeley
National Laboratory.
Frey, M. M., J. Chwirka, R. Narasimhan, S. Kommineni, and Z. Chowdhury. 2000. Cost Implications of a Lower
Arsenic MCL. AWWA Research Foundation and American Water Works Association, Denver, CO.
Gurian, P.L., R. Bucciarelli-Tieger, M. Chew, A. Martinez, and A. Woocay. 2006. Validating Pre-regulatory Cost
Estimates for the Revised Arsenic MCL. Proceedings of the AWWA Annual Conference and Exposition
in San Antonio, TX.
Harrington, W., R. D. Morgenstern, and P. Nelson. 2000. On the Accuracy of Regulatory Cost Estimates.
Journal of Policy Analysis and Management 19(2): 297-322.
143 -

-------
Hilkert Colby, E. J., T. M. Young, P. G. Green, and J. L. Darby. 2010. Costs of Arsenic Treatment for Potable
Water in California and Comparison to U.S. Environmental Protection Agency Affordability Metrics.
Journal of the American Water Resources Association 46(6): 1238-1254.
National Drinking Water Advisory Council (NDWAC). 2001. Report of the Arsenic Cost Working Group to the
National Drinking Water Advisory Council. Available at:
http://water.epa.gov/drink/info/arsenic/upload/2005_ll_10_arsenic_ndwac-arsenic-report.pdf.
US EPA. 2000a. Arsenic in Drinking Water Rule Economic Analysis. EPA 815-R-00-026. Prepared by Abt
Associates, Inc. for EPA OGWDW, December
US EPA. 2000b. Arsenic Occurrence in Public Drinking Water Supplies. EPA 815R00023. Prepared by ISSI
Consulting Group under contract 68C70005, The Cadmus Group under contract 68C99206 and ICF
Consulting for EPA OGWDW.
US EPA. 2000c. Technologies and Costs for Removal of Arsenic from Drinking Water. EPA 815R00028.
Prepared by Malcolm Pirnie, Inc. under contract 68C60039 for EPA ORD.
US EPA. 2005. Treatment Technologies for Arsenic Removal. EPA 600/S-05/006. EPA's National Risk
Management Research Laboratory, November.
US EPA. 2007. Environmental Technology Verification (ETV) Program Case Studies: Demonstrating Program
Outcomes. EPA 68-C-02-067. Prepared by Science Applications International Corporation for EPA
ORD, January.
US EPA. 2011. Costs of Arsenic Removal Technologies for Small Water Systems: US EPA Arsenic Removal
Technology Demonstration Program. EPA 600R11090. Prepared by Lili Wang and A.S.C. Chen, under
contract EPC05057 for EPA ORD.
144-

-------
Appendix 4.1: Publicly Available Data Related to
Arsenic Rule
Working with Abt Associates, we identified ten sources of publicly available data collected on levels of
contaminants in U.S. drinking waters and four potential data sources on compliance costs.150 The potential
sources on arsenic contaminant levels in drinking water and ambient levels are as follows:
•	Safe Drinking Water Information System (SDWIS)
•	Arsenic Occurrence and Exposure Database (AOED)
•	Consumer Confidence Reports (CCRs)
•	National Tap Drinking Water Database (NTWQD)
•	EPA's STORET Data Warehouse - arsenic ambient levels
•	National Water Information System (NWIS) - arsenic ambient levels
•	National Water-Quality Assessment (NAWQA) Program - arsenic ambient levels
•	Community Water System Survey (CWSS)
•	National Contaminant Occurrence Database (NCOD)
•	National Environmental Public Health Tracking Network
Potential sources of compliance cost data include
•	Drinking Water Infrastructure Needs Survey and Assessment (DWINSA)
•	Community Water System Survey (CWSS)
•	Drinking Water Cost Rate Data
ial Sources of Arsenic: Occurrenc
Safe Drinking Water information System {SDWIS}: EPA's SDWIS federal (SDWIS/FED) and state
(SDWIS/STATE) databases contain basic information submitted by states and EPA regions about public water
systems. States supervise their drinking water systems to ensure that each public water system meets state
and EPA standards for safe drinking water. SDWIS/STATE contains this information and is designed to help
states manage and run their drinking water programs. States are required to report drinking water
information periodically to EPA and this information is maintained in the SDWIS/FED database.
SDWIS/FED contains the information EPA uses to monitor approximately 156,000 public water systems,
including basic information on each water system (e.g., name, location, source of water as groundwater or
surface water, public or private ownership, and population served) as well as information on the reported
violation and enforcement actions. However, until 2011 SDWIS/FED did not contain information on the
150 "Background and Data Sources for Five Selected Rules," memo from Abt Associates to Nathalie Simon, August
17, 2010. Note that this list was later augmented with additional information by EPA.
145 -

-------
observed measurement of contaminants that lead to a given violation. Now the violation measure is
included for each violation in SDWIS.
EPA routinely evaluates state drinking water programs by conducting data verification audits, which evaluate
state compliance decisions and reporting to SDWIS/FED. Every three years, the Agency use to prepare a
report that presents the results of a review and evaluation of the data quality in SDWIS/FED every three years
but due to budget cuts, EPA is currently not preparing these types of reports.
Arsenic Occurrence and Exposure Database (AOED): The AOED was developed to estimate baseline arsenic
occurrence data in the United States. The database is generally based on arsenic data from the following 25
state compliance monitoring dataset and system characteristics from both SDWIS and State compliance
monitoring data.152 The database was published in December 2000 and has not been updated since that
time.
Consumer Confidence Reports (CCRs): CWSs with 15 or more service connections (e.g., houses or other
buildings where drinking water is consumed) or that regularly serve at least 25 year-round residents must
prepare a CCR starting in 1999 (for 1998 calendar year data). CCRs must disclose detected amounts of
contaminants even if no violation has occurred, and as of 2001 systems detecting arsenic above the MCL also
had to include a statement about the health effects of arsenic (but they did not have to report the measured
amount of arsenic). While these reports are to be provided or made available to customers by July 1 of each
year, exactly how they are released and distributed varies by system size and other factors. Systems serving
100,000 or more people (approximately 336 systems) are required to post CCRs online as well as mail them
to customers. On the other hand, smaller systems (serving fewer than 10,000 people) may be able to provide
their customers with this information via other means such as the newspaper (some states have made some
exceptions to these requirements).153
A number of issues arise when attempting to access CCRs. First, systems are not required to submit CCRs to
EPA but only need to submit them to state agencies for compliance monitoring. Second, EPA has a website
that is intended to provide links to state CCRs but very few CCRs are linked to this site.154 CWS that serve
151	Described in the report Arsenic Occurrence in Public Drinking Water Supplies
http://water.epa.gov/drink/info/arsenic/upload/2005 11 10 arsenic occurrence.pdf
152	The states are Alaska, Alabama, Arkansas, Arizona, California, Illinois, Indiana, Kansas, Kentucky, Maine,
Michigan, Minnesota, Missouri, Montana, North Carolina, North Dakota, New Hampshire, New Jersey, New
Mexico, Nevada, Ohio, Oklahoma, Oregon, Texas, and Utah.
153	States governors are empowered to give systems serving fewer than 10,000 customers waivers instead of
mailing the CCRs to customer. Systems serving fewer than 10,000 customers but more than 500 customers may
publish the CCR in the newspaper and notify their customers that the CCR is available. Systems serving fewer than
500 customers may notify their customers that the CCR is available.
154	http://safewater.tetratech-ffx.com/ccr/index.cfm
146-

-------
greater than 100,000 people must post their CCRs online but are not required to use EPA's website. Most of
these systems have their own website.
National Tap Water Quality Database (NTWQD):155 The Environmental Working Group (EWG) - an advocacy
group - assembled 20 million drinking water quality tests performed by water utilities since 2004 into their
National Drinking Water Database to investigate the quality of drinking water across the country. They
requested system monitoring data from each state water office. They received water quality tests conducted
by 47,667 utilities in 44 states (and the District of Columbia) from 2004 to 20 09.156 The data are presented
for all contaminants that are monitored by the system and sent to the state. On the EWG website a drinking
water quality report can be obtained for only one drinking water system at a time. The data presented on
the report for the system summarizes water quality test results. Detailed data files are not available on their
website.
STORET (short for STOrage and RETrieval):lb/EPA maintains all of its ambient water quality data in the
STORET database. STORET also includes data collected and submitted by states, tribes, watershed groups,
other federal agencies, volunteer groups and universities. STORET contains data on physical, chemical, and
biological sampling of waters (including surface water, groundwater, and wetlands) and each observation
also contains information about the sampling procedures used, the submitting organization, and the type of
sampling project (e.g., a long term monitoring project). Historical water quality data (observations collected
before 1999) are contained in the Legacy Data Center. This database contains over 200 million water sample
observations from about 700,000 ground and surface water sampling sites.
National Water Information System (NWIS):m The U.S. Geologic Survey (USGS) collects water-resources data
at approximately 1.5 million sites in the U.S. (including the District of Columbia). Surface-water data are
collected from major rivers, lakes and reservoirs, while ground-water data are collected from wells and
springs. The types of water-quality data collected include temperature, specific conductance, pH, nutrients,
pesticides, volatile organic compounds, and various other contaminants (including arsenic). Both current and
historical data on surface water (water flows and levels), groundwater (water levels), and water quality
(chemical and physical data) are available by geographic area (i.e., county, hydrologic unit,
latitude/longitude).
National Water-Quality Assessment (NAWQAJ Program: The USGS NAWQA Program is designed to provide
an understanding of water-quality conditions in the U.S. Monitoring data are integrated with geographic
information on hydrological characteristics, land use, and other landscape features in order to understand
how water-quality conditions are changing over time and how natural features and human activities affect
155	http://www.ewg.org/tap-water/methodology
156	Some states did not respond, some requested large fees for the data, and one only submitted paper records.
157	http://www.epa.gov/storet/about.html; data collected prior to 1999 is contained in the Legacy STORET
database, while more recent data is contained in the main STORET database.
158	http://waterdata.usgs.gov/nwis/
147 -

-------
those conditions. One of the studies includes a National-Synthesis Assessment on trace elements in
groundwater and surface waters with a particular focus on arsenic. In the Trace Element National Synthesis
Project "Arsenic in Ground Water of the United States," the USGS has developed maps that show the location
and extent of arsenic in groundwater across the U.S.159 The maps are based on arsenic samples taken from
31,350 wells and show widespread high arsenic concentrations across the Midwest, West, and Northeast.
The sample database for the 31,350 wells has information on the location of the well, depth of the well, date
the sample was taken and the concentration of arsenic in the sample.
Prior to the revision of the Arsenic rule in 2001, USGS conducted a retrospective analysis of arsenic
occurrence in groundwater in the U.S.160 For the retrospective study, USGS selected almost 19,000
groundwater sites from their NWIS database.161 If five or more observations were available for a given
county, all observations within 50 kilometers of the county's centroid were combined to construct a
distribution of arsenic concentrations for that county. The arsenic concentrations were associated with data
from SDWIS about the size and number of public water supply systems that use groundwater in each county.
This information was then used to estimate the number and size of public-water supply systems that exceed
different arsenic concentrations in the groundwater source. Targeted arsenic concentrations of 1, 2, 5, 10,
20, and 50 pig/L were exceeded in the ground-water resource associated with 36, 25, 14, 8, 3, and 1 percent
of public water supply systems, respectively.
Community Water System Survey (CWSS): EPA conducted the 2000 CWSS to support development and
evaluation of all drinking water regulations.162 The survey collects information on systems including
operating information such as ownership, population served, water production, water sources, existing
treatment, storage, system distribution as well as contaminant concentrations (including arsenic) from water
sampling. The survey also collects information on revenue, operating and capital expenses, rate structure,
and number of employees. A sample of approximately 1,800 systems was selected from a list of
approximately 53,000 community water systems in SDWIS. Questionnaires were sent to approximately 1,200
medium to large systems, while site visits were conducted on 600 smaller systems. A separate version of the
questionnaire was sent to systems serving more than 500,000 people. Additional questions on contaminant
concentrations in raw and finished water and well depth were requested from these large systems. In 2006,
1,314 systems responded to the survey and EPA published trends and key findings from the survey.
159	http://water.usgs.gov/nawqa/trace/arsenic/
160	"A Retrospective Analysis on the Occurrence of Arsenic in Ground-Water Resources of the United States and
Limitations in Drinking-Water-Supply Characterizations"
http://pubs.usgs.gov/wri/wri994279/pdf/wri994279.pdf
161	Sites that had water samples that were characterized as non-potable (high saline content or high temperature)
were not included in the retrospective analysis.
162	The 1995, 2000 and 2006 surveys are discussed at
http://water.epa.gov/infrastructure/drinkingwater/pws/cwssvr.cfm.
148 -

-------
National Contaminant Occurrence Database (NCOD):163 The NCOD was developed by EPA to meet its
obligation under the Safe Drinking Water Act (SDWA) to review all MCLs every six years and revise them as
necessary. The first six-year review covered 1996-2002 and the second six-year review covered 2003-2009.
Compliance monitoring data were voluntarily submitted by 47 state/primacy agencies (45 states plus Region
8 and 9 tribes) to support this process.164 The NCOD data comprise more than 15 million analytical records
from approximately 132,000 public water systems. Approximately 254 million people are served by these
systems nationally. The dataset for the second six-year review includes the results of all compliance
monitoring data (all sample analytical detections and non-detections) from January 1998 to December 2005
for 69 regulated contaminants, including arsenic.
The NCOD contains approximately 225,000 water samples tested for arsenic between 1998-2005. Each
public water system in the database is identified by system type (CWS or NTNCWS), water source (ground or
surface water), and by the population it serves. The arsenic contaminant information includes a sampling
point identifier established by the state for each sampling location (e.g., source water quality or entry point
to the distribution system), the date the sample was taken, whether arsenic levels were detected in the
sample, and the actual arsenic level.
National Environmental Public Health Tracking Network (NEPHTN): The NEPHTN was developed by the
Centers for Disease Control as a way to integrate health, exposure, and environmental hazard data. Data on
the level of arsenic contamination in community water systems are taken from state databases associated
with the Safe Drinking Water Act while data on arsenic levels in domestic well water were obtained from the
NWAQA program.
Arsenic data are available for sixteen states: California, Connecticut, Florida, Maine, Massachusetts,
Minnesota, Missouri, New Hampshire, New Jersey, New York, Oregon, Pennsylvania, South Carolina, Utah,
Washington, and Wisconsin. Data for CWS are generally available from 1999-2009 for most of these states
while well water data are available for 2000 only. The data for CWSs can be obtained as a quarterly or yearly
distribution of the number of CWSs by mean arsenic concentrations or as a quarterly or yearly distribution of
number of people served by CWSs by mean arsenic concentrations. The data for domestic wells are self-
supplied and are presented as the number of well samples grouped by arsenic concentration levels.
ial Sources of Compliance Cost Data
Drinking Water Infrastructure Needs Surve y and Assessment (DWINSA):166 Every four years, starting in 1995,
EPA surveys local water utilities to obtain information on the anticipated costs of projects to install, upgrade,
and replace equipment to deliver safe drinking water. The purpose of the survey is to estimate the 20-year
163	http://water.epa.gov/scitech/datait/databases/drink/ncod/databases-index.cfm
164	The states not included in the NCOD database are Pennsylvania, Mississippi, Louisiana, Kansas, Washington, and
the District of Columbia.
165	http://ephtracking.cdc.gov/showWaterLandingSolution.action
166	http://water.epa.gov/infrastructure/drinkingwater/dwns/index.cfm
149 -

-------
capital investment needs of public water systems to protect public health. The information is used to
determine the amount of funding each state receives for its Drinking Water State Revolving Fund. In 2007,
EPA mailed questionnaires to each of the 584 largest water systems (serving more than 100,000 people) and
2,266 medium systems (serving between 3,301 and 100,000 people). Approximately 97 percent of the large
systems and 92 percent of the medium systems returned completed questionnaires. For small community
water systems (serving fewer than 3,300 people), EPA contracted water system professionals to conduct in-
person site visits to 600 small systems. Each project listed on the survey had to be accompanied by written
documentation on the scope and necessity of the project, as well as the project cost. Acceptable
documentation for cost estimates included master and capital improvement plans, preliminary engineering
reports, facility plans, bid tabulations and engineering estimates not developed for the assessment. Systems
providing cost estimates were encouraged to submit design parameters regarding size or capacity of the
infrastructure. If a system could not provide acceptable cost documentation, EPA requested that the system
provide the information needed for EPA to model the cost of the project (e.g., design parameters).
Community Water System Survey (CWSS): The CWSS, discussed in greater detail above, collects information
on revenue, operating and capital expenses, rate structure, and number of employee for public water
systems in 2000.
Cost Rate Data: There are several potential sources of drinking water rates for residential and other
customers. Raftelis Financial Consultants have published a survey of drinking water rates biennially since
1986. Since 2004, this survey has been published jointly with the American Water Works Association
(AWWA).167 The most recent survey contains data on over 300 utilities serving 1000 to 9 million customers.
Separately, Black and Veatch collect rate data for water and sewer services for residential, industrial and
commercial customers. The data are published in their "50 Largest Cities Water/Wastewater Rate Survey"
and they find that water and wastewater bills for residential use across the country have increased at a
steady rate since 2001.168
ORD Demonstration Projects: In October 2001, EPA undertook a project to help small community water
systems (<10,000 customers) research and develop cost-effective technologies to meet the new arsenic
standard. As part of the Arsenic Rule Implementation Research Program, EPA's Office of Research and
Development (ORD) conducted three rounds of demonstration projects that conducted full-scale, onsite
demonstrations of arsenic removal technology, process modifications and engineering approaches for small
systems.
EPA program funds in addition to funding from Congress provided support for the three rounds of
demonstration projects from 2005-2007. Treatment technologies were selected from solicited proposals.
EPA conducted 50 arsenic removal demonstration projects in 26 states in the US. Treatment systems
167	In 1996 and 1999, AWWA published the results of their own survey including detailed financial and revenue
data as part of their Water:\Stats series, but discontinued this publication after 1999.
168	http://www.bv.com/Downloads/Resources/Brochures/rsrc_EMS_Top50RateSurvey.pdf
150-

-------
selected for the projects included 28 adsorptive media (AM) systems, 18 iron removal (IR) systems (including
two systems using IR and iron addition (I A)) and coagulation/filtration (CF) systems (including four systems
using IR pretreatment followed by AM), two ion exchange (IX) systems, and one of each of the following
systems: reverse osmosis (RO), point-of-use (POU) RO, POU AM, and system/process modification. Of the 50
projects, 42 were community water systems (CWS) and eight were non-transient non-community water
systems (NTNCWS). The report "Costs of Arsenic Removal Technologies for Small Water Systems: U.S. EPA
Arsenic Removal Technology Demonstration Program" summarizes the cost data across all demonstration
projects grouped by the type of technology. Total capital costs and operating and maintenance (O&M) costs
are presented for each treatment system. Capital costs are broken down by equipment, site engineering,
and installation costs. Factors affecting capital costs include system flow rate, construction material, media
type and quantity, pre- and/or post-treatment requirements, and level of instruments and controls required.
The O&M costs for each treatment system are broken down by media replacement, chemical use, electricity
and labor.
151 -

-------
Appendix 4.2: EPA Cost Curves For Compliance with
MCLs for Arsenic in Drinking Water
The following tables present the assumptions and cost curves used by EPA to estimate the costs of treatment
technologies. Equations were converted to 2006 dollars from 1998 dollars using the Engineering News
Record Construction Cost Index (ENR CCI).
Modified Coagulation/Filtration:
EPA assumed that typical coagulation/filtration treatment plants remove 50 percent of the influent arsenic
prior to enhancement, and that O&M (operation and maintenance) costs would only include power and
materials and not additional labor. EPA used the following design assumptions to develop cost estimates for
small and large drinking water systems:
•	Small Systems (< 1 mgd): Additional ferric chloride dose, 10 mg/L; Additional feed system for
increased ferric chloride dose; Additional lime dose, 10 mg/L for pH adjustment; and Additional feed
system for increased lime dose.
•	Large Systems (> 1 mgd): Additional ferric chloride dose, 10 mg/L; Additional feed system for
increased ferric chloride dose; Additional lime dose, 10 mg/L for pH adjustment; and Additional feed
system for increased lime dose.
Table A1 summarizes the capital and O&M cost equations that EPA used to estimate costs for
modified/enhanced coagulation/filtration treatment.
Table A1 - Cost Equations for Modified Coagulation/Filtration (2006 dollars)
Design Flow (x)
Capital Cost (y) Equation
O&M Cost (z) Equation
Less than 1 mgd
y = -5095.4X2 + 19626x + 9516.5
z = -402.68V2 + 9722v + 294.09
Between 1 mgd and 10 mgd
y = 125208X- 101161
z = 23282V-4639.8
Greater than 10 mgd
y = -8.9397X2 + 8634.2X + 1065469
z = -0.5291V2 + 19913V + 10531.3
Source: U.S. EPA (2000)
mgd = million gallons per day; x = design flow; v = average flow; y = capital cost; z = O&M cost
Coagulation Assisted Microfiltration:
EPA used the following design assumptions to develop cost estimates for small and large drinking water
systems:
152 -

-------
•	Very Small Systems (< 0.10 mgd): Coagulant dosage, ferric chloride, 25 mg/L; No polymer addition;
Filtration rate, 2.5 gpm/ft2; and Sodium hydroxide dose, 20 mg/L
•	Small Systems (< 1 mgd): Package plant for all small systems; filtration rate 5 gpm/ft2; Ferric chloride
dose, 25 mg/L; Sodium hydroxide dose, 20 mg/L; and Standard microfilter specifications, provided by
vendors.
•	Large Systems (> 1 mgd): Ferric chloride dose, 25 mg/L; Rapid mix, 1 minute; Flocculation, 20
minutes; Sedimentation, 1000 gpd/ft2 in rectangular basins; and Standard microfilter specifications,
provided by vendors.
Table A2 summarizes the capital and O&M cost equations EPA used to estimate costs for coagulation assisted
microfiltration treatment.
Table A2 - Cost Equations for Coagulation Assisted Microfiltration (2006 dollars)
Design Flow (x)
Cost Equation
Capital Costs (y)
Less than 0.10 mgd
y = -15898039X2 + 6500208x + 125640
Between 0.10 mgd and 0.25 mgd
y = 3121141x +304566
Between 0.25 mgd and 1 mgd
y = -644143X2 + 3075576X + 363826
Between 1 mgd and 10 mgd
y = 1373039X + 1422220
Greater than 10 mgd
y = 426x2 + 1227399X + 2835987
O&M Costs (z)
Less than 0.03 mgd
z = 262176V + 26992
Between 0.03 mgd and 0.09 mgd
z= 181594V+ 29489
Between 0.09 mgd and 0.35 mgd
z = 106668V + 35933
Between 0.35 mgd and 4.25 mgd
z = 17730V + 67951
Greater than 4.25 mgd
z = 20294V + 56410
Source: U.S. EPA (2000)
mgd = million gallons per day; x = design flow; v = average flow; y = capital cost; z = O&M cost
Modified Lime Softening
EPA assumed that typical lime softening treatment plants remove 50 percent of the influent arsenic prior to
enhancement, and that O&M costs would only include power and materials, not additional labor. EPA used
the following design assumptions to develop cost estimates for small and large drinking water systems:
•	Additional lime dose, 50 mg/L;
•	Chemical feed system for increased lime dose;
•	Additional carbon dioxide (liquid), 35 mg/L for recarbonation; and
•	Chemical feed system for increased carbon dioxide dose.
Table A3 summarizes the capital and O&M cost equations EPA used to estimate costs for modified/enhanced
lime softening treatment.
153 -

-------
Table A3 - Cost Equations for Modified Lime Softening (2006 dollars)
Design Flow (x)
Cost Equation
Capital Costs (y)
Less than 1 mgd
y = -30601x2 + 64217x + 10519.7
Between 1 mgd and 10 mgd
y = 177803X- 133668
Greater than 10 mgd
y = -10.042X2 + 35445X + 1290926
O&M Costs (z)
Less than 0.35 mgd
z = 2986.7V2 + 40659V + 425.80
Between 0.35 mgd and 3.5 mgd
z = 38821V + 1457.6
Greater than 3.5 mgd
z = -0.6031V2 + 34721V + 19921
Source: U.S. EPA (2000)

mgd = million gallons per day; x = design flow; v
= average flow; y = capital cost; z = O&M cost
Activated Alumina
EPA's design assumptions for activated alumina vary based on whether pH adjustment is necessary. For
natural pH (i.e., no pH adjustment), EPA made the following assumptions:
•	pH will not need to be adjusted after the activated alumina process;
•	Empty Bed Contact Time (EBCT) is 5 minutes per column;
•	The density of the activated alumina media is assumed to be 47 Ib/ft3;
•	The bed depth ranged from 3 to 6 feet, depending on the design flow;
•	The maximum diameter per column is 12 feet;
•	50 percent bed expansion during backwash even though backwashing may not be necessary on a
routine basis for smaller systems;
•	Redundant column necessary to allow the system to operate while the media is being replaced in the
old roughing column.
For systems with pH adjustment, EPA used the same assumptions except included cost to adjust pH to the
optimal pH of 6. Table A4 summarizes the capital and O&M cost equations EPA used to estimate costs for
activated alumina treatment.
154-

-------
Table A4 - Cost Equations for Activated Alumina (2006 dollars)
Design Flow (x) and Design Parameters
Cost Equation
Capital Costs (y)
Less than 0.10 mgd; natural pH
y = 686392X + 13605
Greater than 0.10 mgd; natural pH
y = 559821X + 13602
Less than 0.10 mgd; pH adjusted to 6.0
y = 740360X + 56081
Greater than 0.10 mgd; pH adjusted to 6.0
y = 613790X + 56079
O&M Costs (z)
Less than 0.35 mgd; natural pH 7.0 - 8.0
z = 251601V+ 5491.4
Greater than 0.35 mgd; natural pH 7.0 - 8.0
z = 254047V + 13051.2
Less than 0.35 mgd; natural pH 8.0 - 8.3
z = 479114V+ 5809.6
Greater than 0.35 mgd; natural pH 8.0 - 8.3
z = 485379V + 20999
Less than 0.35 mgd; pH adjusted to 6.0; 23,100 BVs
z = 220201V+ 7718.1
Greater than 0.35 mgd; pH adjusted to 6.0; 23,100 BVs
z = 220298V + 15574
Less than 0.35 mgd; pH adjusted to 6.0; 15,400 BVs
z = 273550V + 8425.8
Greater than 0.35 mgd; pH adjusted to 6.0; 15,400 BVs
z = 274543V + 17439
Source: U.S. EPA (2000)
BVs = bed volumes; mgd = million gallons per day; x = design flow; v= average flow; y = capital cost; z =
O&M cost
Ion Exchange
EPA made the following assumptions to estimate costs for ion exchange:
•	Empty Bed Contact Time (EBCT) = 2.5 minutes per column
•	Bed depth ranged from 3 feet to 6 feet depending on the design flow
•	Maximum diameter per column is 12 feet
•	Vessel cost has been sized based on 50% bed expansion during backwash
•	Capital costs include a redundant column to allow the system to operate while the media is being
regenerated in the other column
•	The run length when sulfate is at or below 20 mg/L is 1,500 bed volumes (BV); the run length when
sulfate is between 20 and 50 mg/L sulfate is 700 BV
•	Salt dose for regeneration was 10.2 Ib/ft3.
•	Incremental labor for the anion exchange is one hour per week plus three hours per regeneration.
Table A5 summarizes the capital and O&M cost equations EPA used to estimate costs for ion exchange
treatment.
155 -

-------
Table A5: Cost Equations for Ion Exchange (2006 dollars)
Design Flow (x) and Design Parameters
Cost Equation
Capital Costs (y)
Less than 0.10 mgd; less than 20 mg/L S04
y = 458982X + 26035
Greater than 0.10 mgd; less than 20 mg/L S04
y = -8363.2X2 + 425133x + 48962
Less than 0.10 mgd; 20 mg/L - 50 mg/L S04
y = 605021X + 26035
Greater than 0.10 mgd; 20 mg/L - 50 mg/L S04
y = -12995.lx2 + 497964x + 97662
O&M Costs (z)
Less than 0.35 mgd; less than 20 mg/L S04
z = -90359V2 + 103289V + 6656.5
Greater than 0.35 mgd; less than 20 mg/L S04
z = -2258.4V2 + 49750V + 22021
Less than 0.35 mgd; 20 mg/L - 50 mg/L S04
z = -110306V2 + 126338V + 11255.3
Greater than 0.35 mgd; 20 mg/L - 50 mg/L S04
z = -2455v2 + 64294V + 32786
Source: U.S. EPA (2000)

mgd = million gallons per day; x = design flow; v = average flow; y = capital cost; z = O&M cost
Greensand Filtration
EPA used the following design assumptions to develop cost estimates for greensand filtration:
•	Potassium permanganate feed, 10 mg/L;
•	The filter medium is contained in a ferrosand continuous regeneration filter tank equipped with an
underdrain;
•	Filtration rate, 4 gpm/ft2;
•	Backwash is sufficient for 40 percent bed expansion; and
•	Corrosion control measures are not required because pH is not affected by the process.
EPA used the VSS model to estimate capital and O&M costs because greensand filtration costs are not
included in either the Water Model or the W/W Model. Thus, while this technology could be effectively
operated in larger size systems, the cost equations below may not provide representative costs for large
systems.
Capital Costs = 782662x0838
O&M Costs = 0.0012X2 + 78483X + 9847.3
Point of Use Reverse Osmosis
EPA estimated costs for reverse osmosis (RO) and activated alumina point-of-use (POU) technologies. EPA
used "Cost Evaluation of Small System Compliance Options - Point-of Use and Point-of-Entry Treatment
Units" (Cadmus Group, 1998) to estimate treatment costs. EPA developed cost curves based on the following
assumptions:
•	Average household consists of 3 individuals using 1 gallon each per day (1,095 gallons per year)
•	Life of unit is 5 years
•	Duration of cost study is 10 years (or 2 POU devices per household)
•	Cost of water meter and automatic shut-off valve included.
156-

-------
•	No shipping and handling costs required.
•	Volume discount schedule: retail for single unit, 10 percent discount for 10 or more units, 15 percent
discount on more than 100 units.
•	Installation time -1 hour unskilled labor (POU)
•	O&M costs include maintenance, replacement of pre-filters and membrane cartridges, laboratory
sampling and analysis, and administrative costs.
The capital and O&M cost equations for POU RO are as follows, with x equal to design flow and v equal to
average flow.
Capital = 1151.73X0-9261
O&M = 89.14V09439
The capital and O&M cost equations for POU activated alumina are as follows, with x equal to design
flow and v equal to average flow.
Capital = 395.46x0-9257
O&M = 549.6v°-937S
157 -

-------
Appendix 4.3: Document sent to Compliance
Assistance Engineering Firms
EPA's Arsenic Drinking Water Rule
The purpose of this questionnaire is to collect information and feedback from industry experts on the U.S.
Environmental Protection Agency's analysis of compliance costs for the arsenic drinking water rule as
undertaken for rule development in 2001. The goal of this project is to assess EPA's analysis and
estimates of compliance costs at the time of rule promulgation. We also want to determine whether EPA
accurately identified all the process technologies that were available to reduce arsenic levels.
This questionnaire summarizes the assumptions and cost estimation frameworks used by EPA to
estimate the costs of treatment technologies that the Agency identified as candidates for compliance with
the arsenic rule. We want to assess whether the actual costs of arsenic treatment differed substantially
from EPA's estimates at the time of rule development. In addition, we hope to understand the reasons for
potential differences in these estimates, including insight into whether new or modified treatment
technologies may have been implemented to meet the arsenic standard, which EPA did not account for in
its cost analysis.
Section 1. Regulatory Background
On January 22, 2001, EPA published a new national primary drinking water regulation for arsenic
(Arsenic Rule), which lowered the maximum contaminant level (MCL) 50 ng/L to 10 ng/L. EPA estimated
that the rule would apply to 54,000 community water systems (CWSs) and 20,000 non-transient non-
community water systems (NTNCWSs) that serve non-residential communities (e.g. schools, churches).
The rule gave water systems until January 23, 2006 to comply with the revised arsenic MCL. EPA had
estimated that approximately 3,000 CWSs and 1,100 NTNCWSs would need to reduce arsenic levels in
their drinking water for compliance with the 10 ng/L standard.
Section 2. Arsenic Treatment Technologies and Costs
EPA identified the following technologies that would effectively remove arsenic and bring a water system
into compliance:
•	Modified Coagulation/Filtration;
•	Coagulation Assisted Microfiltration;
•	Modified Lime Softening;
•	Activated Alumina (with and without pH adjustment);
•	Ion Exchange (groundwater only);
•	Greensand Filtration (groundwater only); and
•	Point-of-Use Reverse Osmosis (for small groundwater systems only).
EPA used three models to develop costs for these treatment technologies (except activated alumina and
ion exchange): Very Small Systems Best Available Technology Cost Document (VSS model; Malcolm
158 -

-------
Pirnie, 1993); the Water Model (Culp/Wesner/Culp, 1984); and the W/W Cost Model (Culp/Wesner/Culp,
1994).
All equations for both capital and O&M costs, as well as all monetary figures are presented in 2006
dollars. Equations and monetary figures were converted to 2006 dollars from 1998 dollars using the
Engineering News Record Construction Cost Index (ENR CCI).
Q1a: Have treatment technologies changed since the rule was promulgated? For example, have
additional or substantially modified treatment technologies or compliance approaches been used
to achieve compliance? If so, please explain how.
A1a: »
Q1b: Based on your professional knowledge and experience, are the treatment technologies that
EPA proposed for groundwater and surface water systems for compliance representative of the
actual treatment technologies employed for compliance with the Arsenic Rule?
A1b: »
Q1c: Based on your professional knowledge and experience, please estimate the frequency with
which these technology options have been used for compliance? To the extent possible, please
identify the principal factors underlying the selection of a particular treatment
technology/compliance approach by different categories of drinking water system - e.g.,
groundwater vs. surface water, small vs. large system.
A1c: »
2.1 Modified Coagulation/Filtration
EPA assumed that typical coagulation/filtration treatment plants remove 50 percent of the influent arsenic
prior to enhancement, and that O&M (operation and maintenance) costs would only include power and
materials and not additional labor. EPA used the following design assumptions to develop cost estimates
for small and large drinking water systems:
• Small Systems (< 1 mgd): Additional ferric chloride dose, 10 mg/L; Additional feed system for
increased ferric chloride dose; Additional lime dose, 10 mg/L for pH adjustment; and
Additional feed system for increased lime dose.
159 -

-------
• Large Systems (> 1 mgd): Additional ferric chloride dose, 10 mg/L; Additional feed system for
increased ferric chloride dose; Additional lime dose, 10 mg/L for pH adjustment; and
Additional feed system for increased lime dose.
Table a summarizes the capital and O&M cost equations that EPA used to estimate costs for
modified/enhanced coagulation/filtration treatment.
Table 1a - Cost Equations for Modified Coagulation/Filtration (2006 dollars)
Design Flow (x)
Capital Cost (y) Equation
O&M Cost (z) Equation
Less than 1 mgd
y = -5095.4X2 + 19626x + 9516.5
z = -402.68x2 + 9722x + 294.09
Between 1 mgd and 10 mgd
y = 125208X- 101161
z = 23282x - 4639.8
Greater than 10 mgd
y = -8.9397X2 + 8634.2x + 1065469
z = -0.5291X2 + 19913x + 10531.3
Source: U.S. EPA (2000)
mgd = million gallons per day;
x = design flow; y = capital cost; z = O&M cost
Table 1b provides capital costs and O&M costs for different design flow thresholds:
Table 1b - Modified Coagulation/Filtration Treatment Costs (2006 dollars)
Design Flow (mgd)
Capital Cost ($)
O&M Cost ($)
0.01
$9,700
$400
0.1
$11,400
$1,300
1
$24,000
$18,600
10
$1,150,900
$228,200
50
$1,474,800
$1,004,900
Notes:


Costs are derived from equations found in U.S. EPA (2000), mgd = million gallons per day
All costs are rounded to the nearest hundred dollars
Q2.1a: Please comment on the estimated costs and assumptions EPA used to estimate the costs
for modified coagulation/filtration.
A2.1a: »
Q2.1b: Have capital and O&M costs for this technology changed significantly from the time
facilities complied with the arsenic rule (i.e., since 2006)? If so, what are the principal reasons for
these changes? To the extent possible, please indicate the approximate amount of difference from
EPA's estimates.
A2.1b: »
2.2 Coagulation Assisted Microfiltration
EPA used the following design assumptions to develop cost estimates for small and large drinking water
systems:
160-

-------
•	Very Small Systems (< 0.10 mgd): Coagulant dosage, ferric chloride, 25 mg/L; No polymer
addition; Filtration rate, 2.5 gpm/ft2; and Sodium hydroxide dose, 20 mg/L
•	Small Systems (< 1 mgd): Package plant for all small systems; filtration rate 5 gpm/ft2; Ferric
chloride dose, 25 mg/L; Sodium hydroxide dose, 20 mg/L; and Standard microfilter
specifications, provided by vendors.
•	Large Systems (> 1 mgd): Ferric chloride dose, 25 mg/L; Rapid mix, 1 minute; Flocculation,
20 minutes; Sedimentation, 1000 gpd/ft2 in rectangular basins; and Standard microfilter
specifications, provided by vendors.
Table a summarizes the capital and O&M cost equations EPA used to estimate costs for coagulation
assisted microtiItration treatment.
Table 2a - Cost Equations for Coagulation Assisted Microfiltration (2006 dollars)
Design Flow (x)
Cost Equation
Capital Costs (y)
Less than 0.10 mgd
y = -15898039X2 + 6500208x + 125640
Between 0.10 mgd and 0.25 mgd
y = 3121141x + 304566
Between 0.25 mgd and 1 mgd
y = -644143x2 + 3075576X + 363826
Between 1 mgd and 10 mgd
y= 1373039X + 1422220
Greater than 10 mgd
y = 426x2 + 1227399X + 2835987
O&M Costs (z)
Less than 0.03 mgd
z = 262176x +26992
Between 0.03 mgd and 0.09 mgd
z = 181594X +29489
Between 0.09 mgd and 0.35 mgd
z = 106668X +35933
Between 0.35 mgd and 4.25 mgd
z = 17730X + 67951
Greater than 4.25 mgd
z = 20294x + 56410
Source: U.S. EPA (2000)
mgd = million gallons per day; x = design flow; y = capital cost; z = O&M cost
Table 2b provides capital costs and O&M costs for different design flow thresholds.
Table 2b - Coagulation Assisted Microfiltration Treatment Costs (2006 dollars)
Design Flow (mgd)
Capital Cost ($)
O&M Cost ($)
0.01
$189,100
$29,600
0.1
$616,700
$142,600
1
$2,795,300
$85,700
10
$15,152,600
$259,400
50
$65,271,600
$1,071,100
Notes:


Costs are derived from equations found in U.S. EPA (2000), mgd = million gallons per day
All costs are rounded to the nearest hundred dollars
Q2.2a: Please comment on the estimated costs and assumptions EPA used to estimate the costs
for coagulation assisted microfiltration.
A2.2a: »
161 -

-------
Q2.2b: Have capital and O&M costs for this technology changed significantly from the time
facilities complied with the arsenic rule (i.e., since 2006)? If so, what are the principal reasons for
these changes? To the extent possible, please indicate the approximate amount of difference from
EPA's estimates
A2.2b: »
EPA also estimated waste disposal costs, which included mechanical and non-mechanical dewatering
with nonhazardous landfill disposal. Table2c summarizes the capital and O&M cost equations that EPA
used to estimate costs for coagulation-assisted microtiItration treatment for waste disposal.
Table 3c: Cost Equations for Coagulation Assisted Microfiltration Waste Disposal (2006 dollars)
Design Flow (x)
Cost Equation
Capital Costs (y)
Less than 0.25 mgd; Mechanical Dewatering
y = -922800X2 + 606498x + 35628
Between 0.25 mgd and 1.75 mgd; Mechanical Dewatering
y = 281887x +56001
Greater than 1.75 mgd; Mechanical Dewatering
y = -2189.9X2 + 200335x + 209890
Less than 0.085 mgd; Non-mechanical Dewatering
y = 4088388x - 1052
Between 0.085 mgd and 1.75 mgd; Non-mechanical
y = 2330137x + 143879
Dewatering

Greater than 1.75 mgd; Non-mechanical Dewatering
y = 2168456X +434903
O&M Costs (z)
Less than 0.085 mgd; Mechanical Dewatering
z = -4631178x2 + 912204x + 7778
Between 0.085 mgd and 1.75 mgd; Mechanical Dewatering
z = 33520x + 49094
Greater than 1.75 mgd; Mechanical Dewatering
z = 106668X +35933
Less than 0.085 mgd; Non-mechanical Dewatering
z = 25058x2 + 6242x + 2829
Between 0.085 mgd and 0.70 mgd; Non-mechanical
z = 148943x - 9257
Dewatering

Greater than 0.70 mgd; Non-mechanical Dewatering
z = 22.599X2 + 80975x + 38308
Source: U.S. EPA (2000)
mgd = million gallons per day, x = design flow, y = capital cost, z = O&M cost
162 -

-------
Table 2d provides capital costs and O&M costs for different design flow thresholds
Table 2d - Coagulation Assisted Microfiltration Waste Disposal Treatment Costs (2006 dollars)
Design Flow (mgd)
Capital Cost ($)
O&M Cost ($)
Mechanical De watering
0.01
$41,600
$16,400
0.1
$87,100
$52,400
1
$337,900
$82,600
10
$1,994,300
$1,102,600
50
$4,751,800
$5,369,300
Non-Mechanical Dewatering
0.01
$39,800
$2,900
0.1
$376,900
$5,600
1
$2,474,000
$119,300
10
$22,119,500
$850,300
50
$108,857,700
$4,143,600
Notes:


Costs are derived from equations found in U.S. EPA (2000), mgd = million gallons per day
All costs are rounded to the nearest hundred dollars
Q2.2c: Please comment on the estimated costs and assumptions EPA used to estimate the costs
for coagulation assisted microfiltration waste disposal treatments.
A2.2c: »
Q2.2d: Have capital and O&M costs for this technology changed significantly from the time
facilities complied with the arsenic rule (i.e., since 2006)? If so, what are the principal reasons for
these changes? To the extent possible, please indicate the approximate amount of difference from
EPA's estimates.
A2.2d: »
2.3. Modified Lime Softening
EPA assumed that typical lime softening treatment plants remove 50 percent of the influent arsenic prior
to enhancement, and that O&M costs would only include power and materials, not additional labor. EPA
used the following design assumptions to develop cost estimates for small and large drinking water
systems:
•	Additional lime dose, 50 mg/L;
•	Chemical feed system for increased lime dose;
•	Additional carbon dioxide (liquid), 35 mg/L for recarbonation; and
•	Chemical feed system for increased carbon dioxide dose.
Table a summarizes the capital and O&M cost equations EPA used to estimate costs for
modified/enhanced lime softening treatment.
163 -

-------
Table 4a - Cost Equations for Modified Lime Softening (2006 dollars)
Design Flow (x)
Cost Equation
Capital Costs (y)
Less than 1 mgd
y = -30601 x2 + 64217x + 10519.7
Between 1 mgd and 10 mgd
y = 177803X- 133668
Greater than 10 mgd
y = -10.042X2 + 35445x + 1290926
O&M Costs (z)
Less than 0.35 mgd
z = 2986.7x2 + 40659x + 425.80
Between 0.35 mgd and 3.5 mgd
z = 38821 x + 1457.6
Greater than 3.5 mgd
z = -0.6031X2 + 34721x + 19921
Source: U.S. EPA (2000)

mgd = million gallons per day; x = design flow; y = capital cost; z = O&M cost
Table 3b provides capital costs and O&M costs for different design flow thresholds
Table 3b - Modified Lime Softening Treatment Costs (2006 dollars)
Design Flow (mgd)
Capital Cost ($)
O&M Cost ($)
0.01
$11,200
$800
0.1
$16,600
$4,500
1
$44,100
$40,300
10
$1,644,400
$367,100
50
$3,038,000
$1,754,500
Notes:


Costs are derived from equations found in U.S. EPA (2000); mgd = million gallons per day
All costs are rounded to the nearest hundred dollars
Q2.3a: Please comment on the estimated costs and assumptions EPA used to estimate the costs
for modified lime softening.
A2.3a: »
Q2.3b: Have capital and O&M costs for this technology changed significantly from the time
facilities complied with the arsenic rule (i.e., since 2006)? If so, what are the principal reasons for
these changes? To the extent possible, please indicate the approximate amount of difference from
EPA's estimates.
A2.3b: »
2.4 Activated Alumina
EPA's design assumptions vary based on whether pH adjustment is necessary. For natural pH (i.e., no
pH adjustment), EPA made the following assumptions:
• pH will not need to be adjusted after the activated alumina process;
164-

-------
•	Empty Bed Contact Time (EBCT) is 5 minutes per column;
•	The density of the activated alumina media is assumed to be 47 Ib/ft3;
•	The bed depth ranged from 3 to 6 feet, depending on the design flow;
•	The maximum diameter per column is 12 feet;
•	50 percent bed expansion during backwash even though backwashing may not be necessary
on a routine basis for smaller systems;
•	Redundant column necessary to allow the system to operate while the media is being
replaced in the old roughing column.
For systems with pH adjustment, EPA used the same assumptions except included cost to adjust pH to
the optimal pH of 6. Table a summarizes the capital and O&M cost equations EPA used to estimate costs
for activated alumina treatment.
Table 5a - Cost Equations for Activated Alumina (2006 dollars)
Design Flow (x) and Design Parameters
Cost Equation
Capital Costs (y)
Less than 0.10 mgd; natural pH
y = 686392X+ 13605
Greater than 0.10 mgd; natural pH
y = 559821 x+ 13602
Less than 0.10 mgd; pH adjusted to 6.0
y = 740360X + 56081
Greater than 0.10 mgd; pH adjusted to 6.0
y = 613790x + 56079
O&M Costs (z)
Less than 0.35 mgd; natural pH 7.0 - 8.0
z = 251601x +5491.4
Greater than 0.35 mgd; natural pH 7.0 - 8.0
z = 254047X + 13051.2
Less than 0.35 mgd; natural pH 8.0 - 8.3
z = 479114x +5809.6
Greater than 0.35 mgd; natural pH 8.0 - 8.3
z = 485379X + 20999
Less than 0.35 mgd; pH adjusted to 6.0; 23,100 BVs
z = 220201x +7718.1
Greater than 0.35 mgd; pH adjusted to 6.0; 23,100
BVs
z = 220298x + 15574
Less than 0.35 mgd; pH adjusted to 6.0; 15,400 BVs
z = 273550X + 8425.8
Greater than 0.35 mgd; pH adjusted to 6.0; 15,400
BVs
z = 274543x + 17439
Source: U.S. EPA (2000)
BVs = bed volumes; mgd = million gallons per day; x = design flow; y = capital cost; z = O&M cost
165 -

-------
Table 4b provides capital costs and O&M costs for different design flow thresholds
Table 4b - Activated Alumina Treatment Costs (2006 dollars)
Design Flow (mgd)
Capital Cost ($)
Natural pH
0.01
$20,500
0.1
$82,200
1
$573,400
10
$5,611,800
50
$28,004,600
pH Adjusted to 6.0
0.01
$63,500
0.1
$130,100
1
$669,900
10
$6,194,000
50
$30,745,600
Design Flow (mgd)
O&M Cost ($)
Natural pH 7.0 - 8.0
0.01
$8,000
0.1
$30,700
1
$267,100
10
$2,553,500
50
$12,715,400
Natural pH 8.0 - 8.3
0.01
$13,300
0.1
$56,400
1
$506,400
10
$4,874,800
50
$24,290,000
pH adjusted to 6.0; 23,100 BVs
0.01
$9,900
0.1
$29,700
1
$235,900
10
$2,218,600
50
$11,030,500
pH adjusted to 6.0; 15; 400 BVs
0.01
$11,200
0.1
$35,800
1
$292,000
10
$2,762,900
50
$13,744,600
Notes:

Costs are derived from equations found in U.S. EPA (2000)
All costs are rounded to the nearest hundred dollars
mgd = million gallons per day

Q2.4a: Please comment on the estimated costs and assumptions EPA used to estimate the costs
for activated alumina treatment.
A2.4a:»
166-

-------
Q2.4b: Have capital and O&M costs for this technology changed significantly from the time
facilities complied with the arsenic rule (i.e., since 2006)? If so, what are the principal reasons for
these changes? To the extent possible, please indicate the approximate amount of difference from
EPA's estimates.
A2.4b: »
EPA also estimated costs for waste disposal which included nonhazardous landfill disposal (for
systems operating without regeneration). EPA assumed zero capital cost for nonhazardous
landfill disposal. O&M cost vary based on pH and BVs as shown in the following equations.
•	Natural pH between 7.0 and 8.0: O&M cost = 10081x
•	Natural pH between 8.0 and 8.3: O&M cost = 19387x
•	pH adjusted to 6.0; 23,100 BVs: O&M cost = 4364x
•	pH adjusted to 6.0; 15,400 BVs: O&M cost = 6547x
Note that the resulting cost estimates from the following equations will be in 2006 U.S. dollars.
Table 4c provides O&M costs for different design flow thresholds for activated alumina waste disposal
treatment including nonhazardous landfill.
Table 4c - Activated Alumina Waste Disposal Treatment Costs Including Nonhazardous Landfill
(2006 dollars)
Design Flow (mgd)
Capital Cost ($)
O&M Cost ($)
Natural pH between 7.0 and 8.0
0.01
$0
$100
0.1
$0
$1,000
1
$0
$10,100
10
$0
$100,800
50
$0
$504,000
Natural pH between 8.0 and 8.3
0.01
$0
$200
0.1
$0
$1,900
1
$0
$19,400
10
$0
$193,900
50
$0
$969,400
pH adjusted to 6.0; 23,100 BVs
0.01
$0
$0
0.1
$0
$400
1
$0
$4,400
10
$0
$43,600
50
$0
$218,200
pH adjusted to 6.0; 15,400 BVs
167 -

-------
0.01
$0
$100
0.1
$0
$700
1
$0
$6,500
10
$0
$65,500
50
$0
$327,300
Notes:


Costs are derived from equations found in U.S. EPA (2000)
All costs are rounded to the nearest hundred dollars

mgd = million gallons per day


Q2.4c: Please comment on the estimated costs and assumptions EPA used to estimate the costs
for activated alumina waste disposal treatment including nonhazardous landfill.
A2.4c: »
Q2.4d: Have capital and O&M costs for this technology changed significantly from the time
facilities complied with the arsenic rule (i.e., since 2006)? If so, what are the principal reasons for
these changes? To the extent possible, please indicate the approximate amount of difference from
EPA's estimates.
A2.4d:»
2.5 Ion Exchange
EPA made the following assumptions to estimate costs for ion exchange:
•	Empty Bed Contact Time (EBCT) = 2.5 minutes per column
•	Bed depth ranged from 3 feet to 6 feet depending on the design flow
•	Maximum diameter per column is 12 feet
•	Vessel cost has been sized based on 50% bed expansion during backwash
•	Capital costs include a redundant column to allow the system to operate while the media is
being regenerated in the other column
•	The run length when sulfate is at or below 20 mg/L is 1,500 bed volumes (BV); the run length
when sulfate is between 20 and 50 mg/L sulfate is 700 BV
•	Salt dose for regeneration was 10.2 Ib/ft3.
•	Incremental labor for the anion exchange is one hour per week plus three hours per
regeneration.
Table a summarizes the capital and O&M cost equations EPA used to estimate costs for ion exchange
treatment.
168 -

-------
Table 6a: Cost Equations for Ion Exchange (2006 dollars)
Design Flow (x) and Design Parameters
Cost Equation
Capital Costs (y)
Less than 0.10 mgd; less than 20 mg/L SO4
y = 458982X + 26035
Greater than 0.10 mgd; less than 20 mg/L SCU
y = -8363.2X2 + 425133x + 48962
Less than 0.10 mgd; 20 mg/L - 50 mg/L SCU
y = 605021 x + 26035
Greater than 0.10 mgd; 20 mg/L -50 mg/L SCU
y = -12995.1X2 + 497964x + 97662
O&M Costs (z)
Less than 0.35 mgd; less than 20 mg/L SO4
z = -90359x2 + 103289X + 6656.5
Greater than 0.35 mgd; less than 20 mg/L SO4
z = -2258.4X2 + 49750x + 22021
Less than 0.35 mgd; 20 mg/L - 50 mg/L SCU
z = -110306X2 + 126338X + 11255.3
Greater than 0.35 mgd; 20 mg/L - 50 mg/L SCU
z = -2455x2 + 64294x + 32786
Source: U.S. EPA (2000)
mgd = million gallons per day; x = design flow; y = capital cost; z = O&M cost
Table 5b provides O&M costs for different design flow thresholds
Table 5b - Ion Exchange Treatment Costs (2006 dollars)
Design Flow (mgd)
Capital Cost ($)
O&M Cost ($)
Less than 20 mg/L SO4
0.01
$30,600
$7,700
0.1
$71,900
$16,100
1
$465,700
$69,500
10
$3,464,000
$293,700
50
$397,500
-$3,136,500
20 mg/L SO4 - 50 mg/L SO4
0.01
$32,100
$12,500
0.1
$86,500
$22,800
1
$582,600
$94,600
10
$3,777,800
$430,200
50
-$7,491,900
-$2,890,000
Notes:


Costs are derived from equations found in U.S. EPA (2000); mgd = million gallons per day
All costs are rounded to the nearest hundred dollars
Q2.5a: Please comment on the estimated costs and assumptions EPA used to estimate the costs
for ion exchange treatment.
A2.5a: »
Q2.5b: Have capital and O&M costs for this technology changed significantly from the time
facilities complied with the arsenic rule (i.e., since 2006)? If so, what are the principal reasons for
these changes? To the extent possible, please indicate the approximate amount of difference from
EPA's estimates.
A2.5b: »
169 -

-------
EPA also estimated waste disposal costs which included discharge to a wastewater treatment plant for
treatment. Table 5c summarizes the capital and O&M cost equations EPA used to estimate costs for ion
exchange treatment.
Table 5c: Cost Equations for Ion Exchange Waste Disposal (2006 dollars)
Design Flow (x) and Design Parameters
Cost Equation
Capital Costs (y)
Less than 0.85 mgd; less than 20 mg/L SO4
y = 5268
Between 0.85 mgd and 25 mgd; less than 20 mg/L SCU
y = 6773
Greater than 25 mgd; less than 20 mg/L SO4
y = 28.6x + 6924
Less than 0.85 mgd; 20 mg/L - 50 mg/L SO4
y = 5268
Between 0.85 mgd and 2.5 mgd; 20 mg/L - 50 mg/L
S04
y = 6773
Greater than 2.5 mgd; 20 mg/L - 50 mg/L SO4
y = 28.6x + 6924
O&M Costs (z)
All flows; less than 20 mg/L SO4
z = 4567x + 500
All flows; 20 mg/L - 50 mg/L SO4
z = 9788x
Source: U.S. EPA (2000)
mgd = million gallons per day; x = design flow; y = capital cost; z = O&M cost
Table 5d provides capital and O&M costs for different design flow thresholds
Table 5d - Ion Exchange Waste Disposal Treatment Costs (2006 dollars)
Design Flow (mgd)
Capital Cost ($)
O&M Cost ($)
Less than 20 mg/L SO4
0.01
$5,300
$500
0.1
$5,300
$1,000
1
$6,800
$5,100
10
$6,800
$46,200
50
$8,400
$228,900
20 mg/L SO4 - 50 mg/L SO4
0.01
$5,300
$100
0.1
$5,300
$1,000
1
$6,800
$9,800
10
$7,200
$97,900
50
$8,400
$489,400
Notes:


Costs are derived from equations found in U.S. EPA (2000)
All costs are rounded to the nearest hundred dollars

mgd = million gallons per day


170-

-------
Q2.5c: Please comment on the estimated costs and assumptions EPA used to estimate the costs
for ion exchange waste disposal treatment.
A2.5c: »
Q2.5d: Have capital and O&M costs for this technology changed significantly from the time
facilities complied with the arsenic rule (i.e., since 2006)? If so, what are the principal reasons for
these changes? To the extent possible, please indicate the approximate amount of difference from
EPA's estimates.
A2.5d: »
2.6 Greensand Filtration
EPA used the following design assumptions to develop cost estimates for greensand filtration:
•	Potassium permanganate feed, 10 mg/L;
•	The filter medium is contained in a ferrosand continuous regeneration filter tank equipped
with an underdrain;
•	Filtration rate, 4 gpm/ft2;
•	Backwash is sufficient for 40 percent bed expansion; and
•	Corrosion control measures are not required because pH is not affected by the process.
EPA used the VSS model to estimate capital and O&M costs because greensand filtration costs are not
included in either the Water Model or the W/W Model. Thus, while this technology could be effectively
operated in larger size systems, the cost equations below may not provide representative costs for large
systems.
Capital Costs = 782662x°838
O&M Costs = 0.0012X2 + 78483x + 9847.3
Table 6a shows the capital and O&M costs for greensand filtration treatment.
171 -

-------
Table 6a - Greensand Filtration Treatment Costs (2006 dollars)
Design Flow (mgd)
Capital Cost ($)
O&M Cost ($)
0.01
$16,500
$10,600
0.1
$113,700
$17,700
1
$782,700
$88,300
10
$5,389,800
$794,700
50
$20,764,000
$3,934,000
Notes:


Costs are derived from equations found in U.S. EPA (2000); mgd = million gallons per day
All costs are rounded to the nearest hundred dollars
Q2.6a: Please comment on the estimated costs and assumptions EPA used to estimate the costs
for greens and filtration treatment.
A2.6a: »
Q2.6b: Have capital and O&M costs for this technology changed significantly from the time
facilities complied with the arsenic rule (i.e., since 2006)? If so, what are the principal reasons for
these changes? To the extent possible, please indicate the approximate amount of difference from
EPA's estimates.
A2.6b: »
EPA also estimated waste disposal costs which included discharge to a wastewater treatment plant for
treatment. EPA assumed capital costs would be $5,300 (in 2006 U.S. dollars), regardless of design flow,
and calculated O&M costs based on the following equations:
•	Flows less than 0.4 mgd: O&M cost = 10054x + 565
•	Flows greater than 0.4 mgd: O&M cost = 10054x + 1505.
Table 6b shows the capital and O&M costs for greensand filtration waste disposal treatment.
Table 6b - Discharge to Wastewater Treatment Plant Treatment Costs (2006 dollars)
Design Flow (mgd)
Capital Cost ($)
O&M Cost ($)
0.01
$5,300
$700
0.1
$5,300
$1,600
1
$5,300
$11,600
10
$5,300
$102,000
50
$5,300
$504,200
Notes:


Costs are derived from equations found in U.S. EPA (2000)
All costs are rounded to the nearest hundred dollars

mgd = million gallons per day


172 -

-------
Q2.6c: Please comment on the estimated costs and assumptions EPA used to estimate the costs
for greens and filtration wastewater treatment.
A2.6c: »
Q 2.6d: Have capital and O&M costs for this technology changed significantly from the time
facilities complied with the arsenic rule (i.e., since 2006)? If so, what are the principal reasons for
these changes? To the extent possible, please indicate the approximate amount of difference from
EPA's estimates.
A 2.6d: »
2.7 Point-of-Use Reverse Osmosis
EPA estimated costs for reverse osmosis (RO) and activated alumina point-of-use (POU) technologies.
EPA used "Cost Evaluation of Small System Compliance Options - Point-of Use and Point-of-Entry
Treatment Units" (Cadmus Group, 1998) to estimate treatment costs. EPA developed cost curves based
on the following assumptions:
•	Average household consists of 3 individuals using 1 gallon each per day (1,095 gallons per
year)
•	Life of unit is 5 years
•	Duration of cost study is 10 years (or 2 POU devices per household)
•	Cost of water meter and automatic shut-off valve included.
•	No shipping and handling costs required.
•	Volume discount schedule: retail for single unit, 10 percent discount for 10 or more units, 15
percent discount on more than 100 units.
•	Installation time - 1 hour unskilled labor (POU)
•	O&M costs include maintenance, replacement of pre-filters and membrane cartridges,
laboratory sampling and analysis, and administrative costs.
The capital and O&M cost equations for POU RO are as follows, with x equal to design flow.
Capital = 1151.73x° 9261
O&M = 89.14x09439
The capital and O&M cost for POU RO treatment are shown in table 7a:
173 -

-------
Table 7a - POU RO Treatment Costs (2006 dollars)
Design Flow (mgd)
Capital Cost ($)
O&M Cost ($)
0.01
$0
$0
0.1
$100
$0
1
$1,200
$100
10
$9,700
$800
50
$43,100
$3,600
Notes:


Costs are derived from equations found in U.S. EPA (2000)

All costs are rounded to the nearest hundred dollars

mgd = million gallons per day


Q2.7a: Please comment on the estimated costs and assumptions EPA used to estimate the costs
for POU RO treatment.
A2.7a:»
Q2.7b: Have capital and O&M costs for this technology changed significantly from the time
facilities complied with the arsenic rule (i.e., since 2006)? If so, what are the principal reasons for
these changes? To the extent possible, please indicate the approximate amount of difference from
EPA's estimates.
A2.7b: »
The capital and O&M cost equations for POU activated alumina are as follows, with x equal to design
flow.
Capital = 395.46x° 9257
O&M = 549.6x09376
The capital and O&M cost POU activated alumina treatment are shown in table 7b.
174-

-------
Table 7b - POU Activated Alumina Treatment Costs (2006 dollars)
Design Flow (mgd)
Capital Cost ($)
O&M Cost ($)
0.01
$0
$0
0.1
$0
$100
1
$400
$500
10
$3,300
$4,800
50
$14,800
$21,500
Notes:


Costs are derived from equations found in U.S. EPA (2000)
All costs are rounded to the nearest hundred dollars

mgd = million gallons per day


Q2.7a: Please comment on the estimated costs and assumptions EPA used to estimate the costs
for POU activated alumina treatment.
A2.7a: »
Q2.7b: Have capital and O&M costs for this technology changed significantly from the time
facilities complied with the arsenic rule (i.e., since 2006)? If so, what are the principal reasons for
these changes? To the extent possible, please indicate the approximate amount of difference from
EPA's estimates.
A2.7b: »
Section 3 Alternative Technologies
Although EPA identified the following alternative treatment technologies at the time of rule development, it
did not consider them in its cost analysis because EPA considered them to be emerging technologies.
Following are the alternative treatment technologies:
•	Sulfur-Modified Iron
•	Granular Ferric Hydroxide
•	Iron Filings
•	Iron Oxide Coated Sand
Q3a: Do you have any knowledge of water systems using these or any other alternative treatment
technologies to comply with EPA's arsenic rule? To the extent possible, please characterize the
approximate frequency with which these alternative technologies have been used for rule
compliance.
A3a: »
175 -

-------
Q3b: Were any of these alternative treatment technologies less costly to install and operate than
the treatment technologies on which EPA based its cost analysis at the time the Arsenic Rule was
promulgated? To the extent possible, please describe cost differences or other factors that may
have favored these alternative technologies compared to the technologies that EPA considered in
the rule analysis.
A3b: »
Section 4 Additional Questions
Q4a: Did technological innovation occur within the treatment systems for which EPA estimated
compliance costs? If so, please indicate which technology or technologies were affected and what
was the impact on the respective capital and O&M costs.
A4a: »
Q4b: Did learning-by-doing play a major role in decreasing O&M compliance costs? If so, please
indicate which technology or technologies were affected by it.
A4b: »
Q4c: Were there any factors that may have caused greater implementation difficulty and higher
costs with the Arsenic Rule? For example, were there:
Any technical challenges to meet compliance requirements?
Issues with financing support for technology installation?
Limitations on compliance in terms of compliance assistance or compliance
schedule?
Terms of regulatory requirements, and specific aspects of the rule requirements?
A4c: »
Q4d: Did treatment technology used by systems you assisted vary based on existing (pre-rule)
arsenic levels (e.g., did systems needing smaller reductions in arsenic concentrations employ
different technologies than systems needing greater reductions)? Explain
A4d: »
176-

-------
Q4e: Did state-level regulations influence the choices treatment technologies that you helped to
install? Explain
A4e: »
Q4e: Do you have any broader knowledge about treatment technologies and their costs installed
by facilities in the region where your projects were located? What treatment technologies did the
systems typically use? Were there differences: by state, system size, source of water (ground/
surface)?
A4c: »
Q4f: Please provide any other comments / suggestions that you feel are not covered in this
questionnaire, but would be helpful in reaching the goals of this project.
A4f: »
References
United States Environmental Protection Agency (U.S. EPA). 2000. Technologies and Costs for Removal
of Arsenic from Drinking Water. EPA 815-R-00-028. December.
177 -

-------
• • *. A Ques - < 4 senic Rule;
EPA identified the following technologies that would effectively remove arsenic and bring a water system
into compliance and developed costs for these treatment technologies:
•	Modified Coagulation/Filtration;
•	Coagulation Assisted Microfiltration;
•	Modified Lime Softening;
•	Activated Alumina (with and without pH adjustment);
•	Ion Exchange (groundwater only);
•	Greensand Filtration (groundwater only); and
•	Point-of-Use Reverse Osmosis (for small groundwater systems only).
1.	- Have treatment technologies changed since the rule was promulgated? For example, have additional or
substantially modified treatment technologies or compliance approaches been used to achieve
compliance? If so, please explain how.
2.	- Based on your professional knowledge and experience, are the treatment technologies that EPA
proposed for groundwater and surface water systems for compliance representative of the actual
treatment technologies employed for compliance with the Arsenic Rule?
3.	- Based on your professional knowledge and experience, please estimate the frequency with which these
technology options have been used for compliance? To the extent possible, please identify the principal
factors underlying the selection of a particular treatment technology/compliance approach by different
categories of drinking water system - e.g., groundwater vs. surface water, small vs. large system.
4.	- Did technological innovation occur within the treatment systems for which EPA estimated compliance
costs? If so, please indicate which technology or technologies were affected and what was the impact on
the respective capital and O&M costs.
5.	- Did learning-by-doing play a major role in decreasing O&M compliance costs? If so, please indicate which
technology or technologies were affected by it.
EPA identified the following alternative treatment technologies that EPA knew existed but did not consider
since these were emerging technologies:
•	Sulfur-Modified Iron
•	Granular Ferric Hydroxide
•	Iron Filings
•	Iron Oxide Coated Sand
•	Others, please describe
178 -

-------
6.	- Do you have any knowledge of water systems using these or any other alternative treatment
technologies to comply with EPA's arsenic rule? To the extent possible, please characterize the
approximate frequency with which these alternative technologies have been used for rule compliance.
7.	- Were any of these alternative treatment technologies cheaper to install and operate than treatment
technologies that existed at the time the Arsenic Rule was promulgated? To the extent possible, please
describe cost differences or other factors that may have favored these alternative technologies
compared to the technologies that EPA considered in the rule analysis.
Additional Questions:
8.	- Were there any factors that may have caused greater implementation difficulty and higher costs with the
Arsenic Rule? For example, were there:
•	Any technical challenges to meet compliance requirements?
•	Issues with financing support for technology installation?
•	Limitations on compliance in terms of compliance assistance or compliance schedule?
•	Terms of regulatory requirements, and specific aspects of the rule requirements?
9.	- Did treatment technology used by systems you assisted vary based on existing (pre-rule) arsenic levels
(e.g., did systems needing smaller reductions in arsenic concentrations employ different technologies
than systems needing greater reductions)? Explain
10.	-Did state-level regulations influence the choices treatment technologies that you helped to install?
Explain
11.	-Do you have any broader knowledge about treatment technologies and their costs installed by facilities
in the region where your projects were located? What treatment technologies did the systems typically
use? Were there differences: by state, system size, source of water (ground/ surface)?
Please provide any other comments / suggestions that you feel are not covered in this questionnaire, but
would be helpful in reaching the goals of this project
179 -

-------
Chapter 5: EPA's 1998 Locomotive
Emission Standards
Elizabeth Kopits
This paper examines how EPA's ex ante cost analysis of the 1998 Locomotive Emission Standard Final Rule
compares to an ex post assessment of costs. This is not an evaluation of how well EPA conducted the ex ante
analysis at the time of the rulemaking. As Chapter 1 makes clear, even the most credible analysis of
compliance costs (done before implementation) will vary from actual costs for a large number of reasons.
For instance, it is possible that market conditions, energy prices, or available technology change in
unanticipated ways. It is also possible that industry overstated the expected costs of compliance. (EPA often
has to rely on industry to supply it with otherwise unavailable information on expected compliance costs.) A
key analytic question we attempt to address is whether ex ante and ex post cost estimates vary by a
substantial degree and why. We organize the discussion according to the conceptual framework outlined in
Section III. An important challenge we face in conducting this assessment is that information to evaluate
costs ex post is quite limited. Any insights offered herein should be viewed with this limitation in mind.
This chapter is organized as follows. Section 5.1 describes the 1998 locomotive rulemaking and the timing of
the engine emission standards. Section 5.2 summarizes the methods EPA used to produce ex ante estimates
of the compliance costs for the final rule. Section 5.3 describes the information sources available to conduct
an ex post cost assessment. Section 5.4 provides our assessment of how the assumptions and estimates used
for each part of EPA's ex ante analysis compare to what occurred in the locomotive industry in the first
decade of the program. Section F offers some preliminary conclusions and summarizes the data limitations
and remaining methodological challenges we face on the parts of the cost analysis where our ex post
assessment is still inconclusive at this time.
5.1. Impetus and Timeline for Regulatory Action
The focus of EPA's 1998 rulemaking was on reducing oxides of nitrogen (NOx) emissions. Since most
locomotives in the U.S. are powered by diesel engines, they have significant NOx emissions, as well as
hydrocarbon (HC) and particulate matter (PM) emissions, all of which have significant health and
environmental effects. At the time of the rulemaking, locomotive NOx emissions were estimated to
represent about 5.5 percent of NOx emissions from all mobile and stationary sources in the U.S. On April 16,
1998, EPA published a rule for a comprehensive emission control program that subjected locomotive
180-

-------
manufacturers and railroads to emission standards, test procedures, and a full compliance program. The rule
was applicable to all locomotives manufactured in 2000 and later, and any remanufactured locomotive
originally built after 1973. The rule exempted locomotives powered by an external source of electricity,
steam-powered locomotives, and locomotives newly manufactured prior to 1973.
The rule established three separate sets of emission standards (Tiers), with applicability of the standards
dependent on the locomotive's date of manufacture:
•	Tier 0 applied to locomotives originally manufactured from 1973 through 2001
•	Tier 1 applied to locomotives and locomotive engines originally manufactured from 2002 to
2004; and
•	Tier 2 applied to locomotives and locomotive engines originally manufactured in 2005 or
later.
Table 5.1 lists the HC, CO, NOx, and PM emission standards and smoke standards for each locomotive tier.
Companies were allowed to meet these performance standards using any technology available to them. The
rule also included average, banking and trading provisions to allow manufacturers and remanufacturers the
flexibility to meet overall emissions goals at lower cost.
In 2008, EPA adopted a new set of emission standards, Tier 3 and Tier 4, for locomotives newly manufactured
or remanufactured after 2008. The revised standards for remanufacturing existing locomotives took effect by
January 1, 2010 for some models, or as soon as certified remanufacture systems were available, and the
requirements for newly-built locomotives were phased-in starting in 2011. Therefore, the universe of
locomotives that were subject to the 1998 rule is limited to locomotives originally built or remanufactured
between 2000 and 2009, after which the 2008 revisions began taking effect.
5,2, EPA Ex Ante Cost Estimates
Table 5.2 summarizes EPA's ex ante estimate of the total costs and emission reductions of the 1998 rule. EPA
estimated these impacts over a forty-one year program run to ensure complete fleet turnover, due to the
extremely long service life of the typical locomotive. Over 2000-2040, the new standards were estimated to
cost $1.33 billion (NPV, 7 percent discounting, 1997$), and reduce NOx emissions from locomotives by nearly
two-thirds, and HC and PM emissions by half. EPA did not monetize the health and environmental benefits
from these emission reductions. The lifetime cost per locomotive was estimated to be approximately
$70,000 for the Tier 0 standards, $186,000 for the Tier 1 standards and $252,000 for the Tier 2 standards. The
average annual cost of this program was estimated to be $80 million per year, or about 0.2 percent of the
total freight revenue for railroads in 1995. The average cost-effectiveness of the standards was expected to
be about $163 per ton of NOx, PM and HC (EPA 1997).
Because the 1998 rule no longer applies to all the locomotives for which EPA estimated costs due to the
promulgation of the 2008 rule, we limit our assessment in this paper to the compliance costs incurred over
roughly the first decade of the program (2000-2009). EPA's ex ante analysis projected that approximately
181 -

-------
Table 5.1. Summary of Emission and Smoke Standards for the 1998 Locomotive Rule
Locomotive Type	Gaseous and Particulate	Smoke Standards
Emissions (% Opacity-Normalized)
	(g/bhp-hr)	

HC2
CO
NOX
PM
Steady
30-sec
3-sec





State
Peak
Peak
Tier 0 Line-haul Duty-cycle
1.00
5.0
9.5
0.60
30
40
50
Tier 0 Switch Duty-cycle
2.10
8.0
14.0
0.72
30
40
50
Tier 1 Line-haul Duty-cycle
0.55
2.2
7.4
0.45
25
40
50
Tier 1 Switch Duty-cycle
1.20
2.5
11.0
0.54
25
40
50
Tier 2 Line-haul Duty-cycle
0.30
1.5
5.5
0.20
20
40
50
Tier 2 Switch Duty-cycle
0.60
2.4
8.1
0.24
20
40
50
Source: EPA (1998).
Notes: EPA set standards for emissions weighted by typical in-use duty cycle. Duty-cycle is a usage pattern
expressed as the percentage of time in use in each of the predetermined throttle notches of a locomotive. The
two distinct types of duty-cycles for freight locomotives are line-haul and switching. Line-haul locomotives,
which perform the line-haul operations, generally travel between distant locations, such as from one city to
another. Yard locomotives, which perform yard operations, are primarily responsible for moving railcars within
a particular railway yard.
$600 million (NPV, 7 percent), or 45 percent of the total program costs, would occur over this period,
achieving 12 percent of the expected NOx reductions. To calculate what EPA estimated the cost per
locomotive to be over 2000-2009, we limit operating costs (fuel and remanufacturing costs) to 10 years, as a
way to approximate the operating costs incurred until each locomotive is remanufactured to the revised (Tier
3 and 4) standards. Using this approach, EPA's ex ante analysis implies the cost per locomotive over 2000-
2009 was approximately $50,000 for the Tier 0 standards, $100,000 for the Tier 1 standards and $98,000 for
the Tier 2 standards.
5.2.1. Main Components of the Ex Ante Cost Analysis
To estimate costs of the Locomotive rule, EPA developed model locomotive categories for each tier to
represent different locomotive model types.169 For each model locomotive, EPA estimated the incremental
per locomotive compliance costs including:
• Initial compliance costs - initial equipment costs (i.e., hardware needed to comply with the
standards initially, but which are not typically replaced at remanufacture), and other costs
such as research and development, engineering, certification, and testing costs.
169 All descriptions of EPA's ex-ante estimates come from the regulatory support document for the rulemaking (US
EPA 1998).
182 -

-------
• Remanufacture costs - maintenance and other costs associated with keeping locomotives in
compliance with the standards through subsequent remanufactures.
Table 5.2. Total Costs and Emission Reductions of the 1998 Locomotive Rule
(EPA Ex-Ante Analysis)
(1997$)
Category
Total Program Costs
(2000-2040)
TIERO
INCREMENTAL COSTS:
Initial Manufacture
$470,446,480
Fuel consumption
$435,742,226
Maintenance
$217,159,792
TOTAL (undiscounted)
$1,123,348,498
NPV (7%)
$584,926,672

TIER 1
INCREMENTAL COSTS:
Initial Manufacture
$102,890,062
Fuel consumption
$79,754,324
Maintenance
$32,013,080
TOTAL (undiscounted)
$214,657,446
NPV (7%)
$132,572,277

TIER 2
INCREMENTAL COSTS:
Initial Manufacture
$669,994,839
Fuel consumption
$1,186,615,407
Maintenance
$78,433,920
TOTAL (undiscounted)
$1,935,044,166
NPV (7%)
$613,541,238

TOTAL COSTS (undiscounted)
$3,273,050,130
NPV (7%)
$1,331,040,187

TOTAL NOx REDUCTIONS (metric tons)
20,052,552
TOTAL PM REDUCTIONS (metric tons)
275,000
TOTAL HC REDUCTIONS (metric tons)
400,000
Source: Locomotive Rule Regulatory Support Document, Table 7-4 (EPA 1998).
• Fuel cost - the cost of any fuel economy penalties associated with compliance.
EPA assumed the initial compliance cost (i.e., fixed and variable costs), together with a manufacturer markup
for overhead and profit, comprise the total manufacturing costs and thus represent the initial cost increase to
the operator. The annual remanufacture and fuel costs calculated over the service life of the locomotive
183 -

-------
comprised the additional operating costs incurred by the operator due to the rule. The per locomotive initial
cost plus the per locomotive operating costs equaled the total per locomotive compliance cost estimate. The
total per locomotive compliance cost, together with the estimated number of locomotives subject to the
rule, was used to calculate the total costs of the program.
>>	M •* > • Mseline
The ex ante compliance costs were based in part on materials supplied by locomotive manufacturers and the
railroad industry, contractor studies of the most likely compliance technologies, and public comments on the
proposed rule or other information available to EPA. The EPA contractors and subcontractors included ICF,
Incorporated, Acurex Environmental Corporation, and Engine, Fuel, and Emissions Engineering, Incorporated
(EF&EE). The regulatory support document does not include a separate formal uncertainty analysis of these
various inputs to the cost estimates, but it does state that the final cost estimates "tend to be somewhat
conservative; that is, for those costs with significant uncertainty, EPA used the higher end of the estimated
range" (US EPA 1998). In some areas, EPA presented a range of costs, especially when contractor estimates
or public comments differed from EPA's initial estimates. A high cost case is included as a sensitivity analysis
to show the effects of modifying base case assumptions regarding some components of the fixed costs
(engineering costs, testing costs, number of suppliers) and the fuel economy penalty (which determines the
additional fuel cost incurred from the added control equipment). These are discussed in greater detail in
Section E below.
It should also be noted that for the most part, the regulatory support document did not include a detailed
discussion of the counterfactual for each component of the cost analysis - e.g., to what extent that more
efficient line-haul locomotives would have been developed and adopted over time in the absence of the rule.
Baseline assumptions about technology (availability, cost, fuel economy), fuel costs, and other inputs (e.g.,
annual fuel consumption) used in EPA's ex ante analysis reflected current conditions rather than a forecast of
future conditions in absence of the regulation. EPA estimated the number of newly manufactured and
remanufactured locomotives of each model type based on information on the number of locomotives
currently in service and existing production, remanufacture, and retirement rates for Class I, II, and III and
passenger rail locomotives.170 For projections of newly manufactured locomotives, the ex ante estimates do
reflect an expectation that the two largest western railroads will purchase large numbers of Tier 2
locomotives during 2005-2010 in order to accelerate their introduction into Southern California. The EPA ex
ante analysis did not discuss other potential exogenous factors that could influence the size of the regulated
universe - e.g., demand side factors that could shift railroad market share relative to trucking and hence the
number of new locomotives purchased.
170 In 1994, Surface Transportation Board (STB) classified a railroad as Class I if its revenue was higher than $255.9
million. Railroads with revenue between $20.5 and $255.8 million were considered Class II, while railroads with
annual revenue less than $20.5 million were Class III.
184-

-------
5.3. Information Available to Conduct Ex Post
Evaluation
To conduct the ex post assessment, we explored the following avenues for collecting ex post compliance
information. We first assessed whether it would be possible to collect compliance cost information using
only publicly-accessible data sources, such as the Census of Manufactures (CMF), Annual Survey of
Manufactures (ASM), American Association of Railroad (AAR) publications, EPA's AirControlNet database, and
Railinc Equipment Registration and Information System (Umler). Overall, we found that while some data is
readily available to help us determine the number of locomotives affected by the regulation, information on
the realized cost of particular control mechanisms is generally lacking.
Next we sought to identify appropriate industry experts with sufficient information about the ex post
regulatory compliance costs. We approached numerous independent associations, including the
Manufacturers of Emission Controls Association, the Association of American Railroads (AAR), the American
Shortline and Regional Railroad Association (ASLRRA), and the Engine Manufacturers Association, but they
were unresponsive to our information requests. We then contacted two engineering consulting firms: Power
Systems Research and Engine, Fuel, and Emissions Engineering, Incorporated (EF&EE). Power Systems
Research is the leading global supplier of business information to the engine, power products and
components industries. We identified its PartsLink database as a potentially useful source for obtaining
information on the historical locomotive fleet but in the end we did not pursue a subscription to this
database due to funding constraints. EF&EE is a research, development, and consulting firm specializing in
motor vehicle emissions and emissions control. The president and founder of EF&EE, Mr. Chris Weaver, was
responsive to our requests and willing to respond to all parts of a questionnaire we prepared based on our
review of EPA's ex ante cost estimation methodology. A copy of the questionnaire is provided in Appendix
5.1.
Ultimately, our analysis below is based on information provided by EF&EE, the sole respondent to the
questionnaire (under a contract with Abt Associates), augmented by publicly available data where possible.
Since Mr. Weaver's firm helped develop EPA's 1997 ex ante cost estimates for this regulation, efforts were
made to provide as much documentation and supporting evidence for his input as possible. Any assessment
and statements based on his professional experience and expert opinion are referenced as such throughout
the paper. A summary of the information sources we relied on for assessing each main component of the cost
estimate is provided in Section 5.5 below.
185 -

-------
5.4, Ex post Assessment of Compliance Cost
5,4,1, Regulated Universe
5.4.1.1. Locomotive Model Types
Railroads can be separated into three classes based on size: Class I, Class II, and Class III. Class I railroads
represent the largest railroad systems in the country, carry most of the interstate freight and passenger
service, and buy almost all of the new locomotives. Class II and III railroads represent the remainder of the
rail transportation system and generally operate within smaller, localized areas, and their fleet of locomotives
tends to be older. Locomotives in each class can perform two different types of operations: line-haul and
yard (or switch). Line-haul locomotives, which perform the line-haul operations, generally travel between
distant locations, such as from one city to another. Switch locomotives, which perform yard operations, are
primarily responsible for moving railcars within a particular railway yard. Switchers make up a relatively
small share of the locomotive market, accounting for approximately 7-8 percent of total Class I fuel
consumption in recent years.171
For the 1998 rulemaking, EPA assumed that the Tier 0 locomotives could be grouped into 5 model categories
(or engine families): switch locomotives from Electro-Motive Diesel (Model A), older and newer line-haul
locomotives from the Electro-Motive Diesel (Model B and C), and older and newer line-haul locomotives from
General Electric Transportation Systems (Model D and E).172 For Tier 1 locomotives, EPA believed that early
versions of the new engine designs used to meet the Tier 2 standards made their appearance during the Tier
1 period. Thus, EPA assumed there would be two Tier 1 models for each of the two manufacturers. Models A
and B are Tier 1 line-hauls from EMD and GE respectively, and Models C and D are early version Tier 2 design
line-hauls from EMD and GE, respectively. EPA assumed that for Tier 2, each manufacturer would have a
single model (Model A- EMD, Model B - GE).
Each manufacturer deployed more versions or types of their locomotive models than estimated by EPA.173
However, for the most part the model categories used by EPA were sufficient for purposes of estimating
compliance costs (EF&EE expert opinion). EMD and GE both deployed direct current (DC) and alternating
current (AC) versions of their basic line-haul locomotives at each Tier level, but the engines and emission
control systems in the DC and AC engines were essentially the same, so it is not clear that these should count
as separate models. EMD also deployed passenger locomotive models for each Tier, generally with twelve-
171	In 2008, 7.7% of Class I fuel consumption was for switchers; 7.4% in 2009-2010 (STB Schedule 750 of Annual
Report Form R-l). Switchers mad up about 7.3% of Class I locomotive fuel consumption in 2007 (ERTAC 2012).
172	GE did not make switch locomotives at that time, or since.
173	Rather than the number of locomotive models offered, another measure would be the number of locomotive
engine families certified. In 2005 and 2008, EMD certified two new locomotive engine families, and GE certified
only one (twelve and 16-cylinder versions of each engine were presumably included in the same family). In 2006
and 2007, they certified one each. Smaller manufacturers such as National Railway Equipment Co. also certified a
number of new as well as remanufactured models. These were probably all genset switchers.
186-

-------
cylinder engines rather than 16 cylinders. GE also deployed a 6000 hp, 16-cylinder version of its GEVO
engine.
5,4,1,2, Number of Locomotives Affected by the Regulation
EPA estimated the number of newly manufactured and remanufactured locomotives affected by the
regulation based on information on the number of locomotives currently in service and existing production,
remanufacture, and retirement rates for Class I, II, and III and passenger rail locomotives.
EPA obtained information on Class I locomotives from the Association of American Railroads Annual Railroad
Facts publication. About 17,500 of Class I locomotives were manufactured post 1972, most of which were
used in line-haul service (Tier 0, Models B through E). The 3,500 older locomotives that were manufactured
prior to 1972 are used as switchers (Tier 0, Model A). EPA assumed that by 2008, almost all 1973 through
1999 line-haul locomotives (13,200) would be remanufactured to meet EPA's standards. EPA also assumed
there would be 400 newly manufactured line-haul locomotives for years 2000-2004, 600 for years 2005-2010,
and 300 new units for all subsequent years.
For Class II and III locomotives, EPA obtained information from American Short Line Railroad Association,
which represents most Class II and Class III railroads. EPA projected that there would be about 600 post-1972
locomotives and 3600 older locomotives in the 1999 Class II and III fleet (Tier 0, Models A through C). EPA
assumed that during the first 10 years of the program, Class II and III railroads would bring about 50
locomotives into compliance with Tier 0 standards each year. EPA further assumed that in 2012, these
railroads would purchase about 150 complying Tier 0 locomotives each year from Class I railroads. For
passenger locomotives, EPA primarily relied on information from Amtrak and the American Public
Transportation Association. There were roughly 463 diesel locomotives in commuter rail service in 1995, with
397 of these manufactured after 1972. EPA projected that about 100 locomotives would be brought into
compliance during each of the first five years of the program, and that all uncontrolled locomotives would be
removed from passenger service by 2011.
Table 5.3 includes EPA's ex ante estimate of the total number of locomotives in each Tier for each model
type.
New Locomotives, Class I railroads buy almost all of the new locomotives in the U.S., and in the timeframe
addressed in the 1998 rule, the bulk of the non-Class I railroad locomotives were not covered by the rule. So
we focus here on Class I.
As shown in Table 5.4, actual sales were higher than EPA's estimate. Over 3,800 newly manufactured
locomotives were in the fleet from 2000 through 2004, or an average of 760 per year. Nearly 4000 were
added from 2005 through 2009, or about 790 per year. This increase was likely driven at least in part by
demand side factors. As fuel prices increased, railroads gained a lot of market share compared to trucks, so
187 -

-------
Table 5.3: Calculation of Per Locomotive Compliance Costs (1997 US Dollars) (EPA Ex Ante Analysis) -
Cost Component
TierO
Tier 1
Tier 2 (2005-2010)
Tier 2 (After 2010)
Model A
Model B
Model C
Model D
Model E
Model A
Model B
Model C
Model D
Model A
Model B
Model A
Model B
Number of Locomotives
3000
4900
2930
2035
2965
360
360
360
360
1700
1700
300
300
Initial Costs
Variable Costs
Hardware Costs













2 deg timing retard
SO
$0
$0
$0
$0
$0
$0
$0
$0
-
-
-
-
4 deg timing retard
SO
$0
-
-
-
-
-
-
-
$0
$0
$0
$0
4 pass aftercooler
-
$5,000
$5,000
$5,000
-
-
-
-
-
-
-
-
-
Improved mechanical injectors
-
$800
-
-
-
-
-
-
-
-
-
-
-
Add electronic fuel injection
-
-
-
$35,000
-
-
-
-
-
-
-
-
-
Improved electronic injectors
-
-
$2,000
-
$2,000
$2,000
$2,000
$2,000
$2,000
$2,000
$2,000
$2,000
$2,000
Increased compression ratio
-
-
-
-
$800
$800
$800
-
-
-
-
-
-
Improved turbocharger
-
-
-
$25,000
$25,000
-
$25,000
-
-
-
-
-
-
Split cooling
-
-
-
-
-
$25,000

$25,000
$25,000
$25,000
$25,000
$25,000
$25,000
High pressure injection
-
-
-
-
-
$2,000
$2,000
$2,000
$2,000
$2,000
$2,000
$2,000
$2,000
Combustion chamber design
-
-
-
-
-
$800
$800
$800
$800
$800
$800
$800
$800
Assembly costs
So
$4,480
$6,720
$4,480
$6,720
$6,720
$6,720
$560
$560
$560
$560
$560
$560
Subtotal Variable cost per locomotive
So
$10,280
$13,720
$69,480
$34,520
$37,320
$37,320
$30,360
$30,360
$30,360
$30,360
$30,360
$30,360
Fixed Costs
Engineering costs
$800,000
$1,700,000
$2,800,000
$1,700,000
$2,800,000
$3,600,000
$3,600,000
$3,600,000
$3,600,000
$4,000,000
$4,000,000
-
-
Testing costs
$422,783
$422,783
$845,566
$422,783
$845,566
$4,227,829
$4,227,829
$4,227,829
$4,227,829
$8,455,659
$8,455,659
$582,900
$582,900
Tooling
-
-
-
-
-
$1,000,000
$1,000,000
$1,000,000
$1,000,000
$1,000,000
$1,000,000
-
-
Technical support
$200,000
$350,000
$500,000
$350,000
$500,000
$500,000
$500,000
$350,000
$350,000
$350,000
$350,000
-
-
Total fixed costs per supplier
$1,422,783
$2,472,783
$4,145,566
$2,472,783
$4,145,566
$9,327,829
$9,327,829
$9,177,829
$9,177,829
$13,805,659
$13,805,659
$582,900
$582,900
Total Fixed Costsl
$4,268,409
$7,418,409
$12,436,818
$2,472,803
$4,145,606
$9,328,029
$9,328,029
$9,178,029
$9,178,029
$13,806,059
$13,806,059
$582,915
$582,915
Subtotal Fixed cost per locomotive2
$1,423
$1,514
$4,245
$1,215
$1,398
$25,911
$25,911
$25,495
$25,495
$8,121
$8,121
$1,943
$1,943
Initial Cost Per Locomotive3
$1,707
$14,153
$21,558
$84,834
$43,102
$75,877
$75,877
$67,025
$67,025
$46,177
$46,177
$38,764
$38,764
Fuel Costs
Average Fuel Consumption
104000
104000
297000
104000
297000
297000
297000
350000
350000
350000
350000
350000
350000
FE Penalty
2%
1%
1%
1%
2%
1%
1%
1%
1%
2%
2%
2%
2%
Gallons of fuel/year4
2,080
1,040
2,970
1,040
5,940
2,970
2,970
3,500
3,500
7,000
7,000
7,000
7,000
Cost per year (@ $0.70/Gal.)
$1,456
$728
$2,079
$728
$4,158
$2,079
$2,079
$2,450
$2,450
$4,900
$4,900
$4,900
$4,900
Fuel Costs Per Locomotive
$21,840
$10,920
$43,659
$10,920
$87,318
$83,160
$83,160
$98,000
$98,000
$196,000
$196,000
$196,000
$196,000
Remanufacture Costs
Cost per year
$0
$400
$846
$400
$846
$1,000
$1,000
$240
$240
$240
$240
$240
$240
Service life
15
15
21
15
21
40
40
40
40
40
40
40
40
Remanufacture Cost Per Locomotive
$0
$6,000
$17,766
$6,000
$17,766
$40,000
$40,000
$9,600
$9,600
$9,600
$9,600
$9,600
$9,600
TOTAL COST PER LOCOMOTIVE
$23,547
$31,073
$82,983
$101,754
$148,186
$199,037
$199,037
$174,625
$174,625
$251,777
$251,777
$244,364
$244,364
1. Represents the fixed cost per supplier multiplied by the number of suppliers for each model type (e.g., 3 suppliers for Tier 0 Models A, B and C, and 1 supplier for the remaining model types).

2. Total fixed costs for all suppliers divided by the number of locomotives in each model category.








3. Sum of total hardware (variable) cost per locomotive and total fixed cost per locomotive plus 20% manufacturer markup.






4. Represents average fuel consumption multiplied by the fuel economy penalty.




Source
: US EPA (1998)


188 -

-------
Table 5.4. Class I Rail Statistics, 1990-2010 -


Average




Revenue ton-miles
Average fuel

Gallons fuel
fuel cost
Number of
Number of
Number of
Revenue
per gallon fuel
consumed per

consumed
(1997$)
locomotives
locomotives
locomotives
ton-miles
consumed
locomotive
Year
(millions)
($/gal)
in service
new
rebuilt
(billions)
(millions)
(thousand gallons)
1990
3134
69.22
18835
530
176
1034
330
166
1991
2926
67.24
18344
472
112
1039
355
160
1992
3022
63.29
18004
321
139
1067
353
168
1993
3112
63.05
18161
504
203
1109
356
171
1994
3356
59.87
18496
821
393
1201
358
181
1995
3503
60.01
18810
928
201
1306
373
186
1996
3601
67.66
19267
761
60
1356
377
187
1997
3603
67.82
19682
743
68
1349
374
183
1998
3619
57.00
20259
889
172
1377
380
179
1999
3749
55.45
20254
709
156
1433
382
185
2000
3720
87.46
20026
640
81
1466
394
186
2001
3730
85.54
19743
710
45
1495
401
189
2002
3751
73.33
20503
745
33
1507
402
183
2003
3849
89.25
20772
587
34
1551
403
185
2004
4082
106.98
22015
1121
5
1663
407
185
2005
4120
151.42
22779
827
84
1696
412
181
2006
4214
192.11
23732
922
158
1772
421
178
2007
4087
218.24
24143
902
167
1771
433
169
2008
3911
312.05
24003
819
129
1777
454
163
2009
3220
177.12
24045
460
103
1532
476
134
2010
3519
224.29
23893
259
181
1691
481
147
2000-01 Average
3725
87
19885
675
63
1481
397
187
2002-04 Average
3894
90
21097
818
24
1574
404
185
2005-09 Average
3910
210
23740
786
128
1710
439
165
Data is for Class I railroads. Class I railroads represent 70 percent of the U.S. rail mileage.
Source: AAR Railroad Facts 2002 and 2011 editions
189

-------
railroads purchased more new locomotives as a result. In addition, improvements in fuel efficiency and/or a
slowdown in the number of rebuilds may have played a role. If companies opted to retire old locomotives
earlier instead of remanufacturing them to comply with Tier 0 requirements during a rebuild, this could have
contributed to an increase in new locomotives in compliance with Tier 1 standards. Similarly, improvements
in fuel efficiency and lower maintenance costs could have led to a rebound effect for locomotive travel, thus
contributing to the robust sales of Tier 2 locomotives.
RemanufacturedLocomotives. As shown in Table 5.4, a total of 839 Class I locomotives were rebuilt during
the first decade of the program (2000-2009), and far fewer rebuilds occurred over 2000-2004 than during the
previous or following five year periods. There were only 40 rebuilds per year on average over 2000-2004,
but about 130 per year on average over 1995-1999 and 2005-2009. The slowdown in rebuilds may reflect a
strategic decision on the part of the railroads in response to the 1998 standards. Typically, line-haul
locomotives are overhauled about every eight years and repowered at least once174, but because the
emission limits were mandated at the time of remanufacture, rather than on a fixed schedule, railroads may
have found it cheaper to deal with the inefficiencies/costs associated with delaying rebuilds or retiring
locomotives earlier and buying more new ones than rebuilding older models to comply with Tier 0
requirements. Continuous improvements in engine durability, improved maintenance practices, and other
factors may have also played a role in increasing the remanufacturing interval over time even absent
emission standards. The increase in rebuilds in the second half of the decade could reflect strategic behavior
in anticipation of the revised locomotive standards. (The advanced notice of proposed rulemaking for the
Tier 3/4 standards was published in mid-2004.) Operators may have opted to rebuild older locomotives
ahead of schedule to Tier 0 standards before the more stringent emission standards took effect.
The number of switch locomotives that were affected by the 1998 rule is likely much less than the number
EPA assumed. Any new switch locomotives sold will be of the genset type, but the large supply of old
locomotives that can be kept running at low cost limits the potential sales of new switchers and old switchers
can be run for a long time without remanufacturing.
In sum, the number of remanufactured locomotives complying with Tier 0 over the first decade of the
program is likely lower than EPA anticipated, and the number of new locomotives complying with Tier 0, 1
and 2 standards is higher than EPA anticipated, by about 140 percent, 70 percent, and 16-23 percent,
respectively.
5,4,2, Methods of Compliance
This section discusses the emission control technologies that EPA expected would already be available at the
time the locomotive emissions standards would take effect. Among those, EPA considered use of the
following technologies:
174http://www.fhwa.dot.aov/environment/air aualitv/conformitv/research/mpe benefits/mpe06.cfm
190-

-------
Table 5.4. Class I Rail Statistics, 1990-2010 -


Average




Revenue ton-miles
Average fuel

Gallons fuel
fuel cost
Number of
Number of
Number of
Revenue
per gallon fuel
consumed per

consumed
(1997$)
locomotives
locomotives
locomotives
ton-miles
consumed
locomotive
Year
(millions)
($/gal)
in service
new
rebuilt
(billions)
(millions)
(thousand gallons)
1990
3134
69.22
18835
530
176
1034
330
166
1991
2926
67.24
18344
472
112
1039
355
160
1992
3022
63.29
18004
321
139
1067
353
168
1993
3112
63.05
18161
504
203
1109
356
171
1994
3356
59.87
18496
821
393
1201
358
181
1995
3503
60.01
18810
928
201
1306
373
186
1996
3601
67.66
19267
761
60
1356
377
187
1997
3603
67.82
19682
743
68
1349
374
183
1998
3619
57.00
20259
889
172
1377
380
179
1999
3749
55.45
20254
709
156
1433
382
185
2000
3720
87.46
20026
640
81
1466
394
186
2001
3730
85.54
19743
710
45
1495
401
189
2002
3751
73.33
20503
745
33
1507
402
183
2003
3849
89.25
20772
587
34
1551
403
185
2004
4082
106.98
22015
1121
5
1663
407
185
2005
4120
151.42
22779
827
84
1696
412
181
2006
4214
192.11
23732
922
158
1772
421
178
2007
4087
218.24
24143
902
167
1771
433
169
2008
3911
312.05
24003
819
129
1777
454
163
2009
3220
177.12
24045
460
103
1532
476
134
2010
3519
224.29
23893
259
181
1691
481
147
2000-01 Average
3725
87
19885
675
63
1481
397
187
2002-04 Average
3894
90
21097
818
24
1574
404
185
2005-09 Average
3910
210
23740
786
128
1710
439
165
Data is for Class I railroads. Class I railroads represent 70 percent of the U.S. rail mileage.
Source: AAR Railroad Facts 2002 and 2011 editions
191

-------
•	Retarding fuel injection - optimizing injection timing and duration to achieve significant NOx
emissions reductions at minimal cost (2 degree or 4 degree timing retard depending on potential fuel
economy impacts);
•	4 pass after cooler - changing from two-pass to a four-pass aftercooler to lessen the degree of timing
retard needed through enhanced charge air cooling;
•	Improved mechanical and electrical injectors - optimizing spray pattern from the nozzle in
conjunction with the configuration of the combustion chamber and induction swirl to achieve
emission reductions;
•	Add electronic fuel injection - to improve control of injection rate and timing;
•	Engine Modifications - reduction in engine size to achieve the desired lower power rating;
•	Improved turbocharger -ensuring that fuel consumption and emissions formation are minimized,
including preventing smoke generation due to turbo lag; changing the geometry of the gas flow
passages in the turbine to improve the response time of the turbocharger;
•	Split cooling - an aftercooler that uses a coolant system separate from the engine coolant system;
•	High pressure injection - to shorten the duration of the fuel injection event, which allows a delay in
the initiation of fuel injection causing lower peak combustion temperatures and reduced NOx
formation, and also reduces fuel economy penalties associated with retarded injection timing; and
•	Combustion chamber design - redesign of the shape of the combustion chamber and the location of
the fuel injector to optimize the motion of the air and the injected fuel with respect to emission
control.
The effective use of some of these technologies can be optimized through the use of other technologies, and
adverse effects of some technologies can be limited or eliminated through the application of other
technologies. For this reason, in estimating compliance costs EPA considered use of multiple technologies
together to form a larger emission reduction system.
Table 5.5 presents EPA's ex ante crosswalk between the expected compliance technologies, their usage, and
the locomotive model types by tier. We discuss the emission control technologies used for each of the Tiers
in turn.
5,4,2,1, Emission Control Technologies for Tier 0 Locomotives.
The main emission control technologies that EPA expected to be used to comply with Tier 0 were:
•	Locomotives equipped with turbocharged engines would be able to employ: modified/improved fuel
injectors, enhanced charge air cooling, injection timing retard, and in some cases, improved
turbochargers, to reduce NOx emissions.
•	EPA expected that engine coolant would continue to be the cooling medium in most cases, rather
than a separate cooling system, and that it would be cost-effective to replace two-pass aftercoolers
with four-pass aftercoolers during the remanufacturing process.
192 -

-------
Table 5.5: Control Options, Expected Usage and Locomotive Models (EPA Ex Ante Analysis) -

Expected Technology
Usage and Models
Developed for Cost
Analysis

2 deg timing retard


4 deg timing retard


4 pass aftercooler

Improved mechanical
injectors
Add electronic fuel
injection
Improved electronic
injectors
Increased compression
ratio

Improved turbocharger


Split cooling


High pressure injection

Combustion chamber
design
TierO
(1973-2001)
Percent locomotives
using technology

50


50


60

30
13
27
20

30


-


-

-
Models using technology
A
X
X









B
X
X
X
X







C
X

X


X





D
X

X

X


X



E
X




X
X
X



Tier 1
(2002-2004)
Percent locomotives
using technology

100


-


-

-
-
100
50

25


75


100

100
Models using technology
A
X




X
X

X
X
X
B
X




X
X
X

X
X
C
X




X


X
X
X
D
X




X


X
X
X
Tier 2
(2005-2010)
Percent locomotives
using technology
-
100
-
-
-
100
-
-
100
100
100
Models using technology
A

X



X


X
X
X
B

X



X


X
X
X
Tier 2
(after 2010)
Percent locomotives
using technology

-


100


-

-
-
100
-

-


100


100

100
Models using technology
A

X



X


X
X
X
B

X



X


X
X
X
Source: U.S. EPA (1998).
193 -

-------
The tools available to manufacturers to reduce emissions for naturally-aspirated and Roots-blown
engines would be modifications to the fuel system, modifications to the combustion chamber and
injection timing.
All of the technologies listed by EPA were actually used to comply with Tier 0, except for engine modifications
to reduce power output, where the approach was instead to substitute smaller non-road engines (EF&EE
expert opinion). For low-power switch locomotives, the EPA regulatory support document discussed two
approaches that appeared to be available to manufacturers: "One approach would be the continued use of
large displacement naturally aspirated engines employing electronic control of the fuel system, improved fuel
injection and improved combustion chambers. Another approach wouldbe to use turbocharging and other
technologies used on line-haul locomotives, but with a reduction in engine size to achieve the desired lower
power rating. A reduction in engine size could be achieved either through the use of fewer power assemblies
of the same configuration as those used on line-haul locomotives or by the use of a different engine design
than that used in line-haul applications. Locomotive manufacturers could also use large non-road engines
(1000-2000 hp) that were originally designed for use in non-locomotive applications" (US EPA 1998).
After the rule was enacted, the two major locomotive manufactures abandoned the switch locomotive
market, and with it, the market for naturally aspirated and Roots-blown engines, leaving it to smaller
companies. The preferred approaches of those smaller companies were the "Hybrid" and "Genset Switcher".
The hybrid substitutes one smaller non-road engine plus a large battery back for the large locomotive engine,
while the genset switcher substitutes (typically) two or more small non-road engines. EPA correctly predicted
the potential to substitute non-road engines for locomotive engines in switchers, but did not foresee the use
of batteries or two or three smaller non-road engines in place of a single larger one.
Two other technologies that were used to meet Tier 0 requirements were increasing the compression ratio
and modifying the cylinder liner and piston rings to reduce lubricating oil consumption. EPA had expected
compression ratio changes to be introduced for compliance with Tier 1, but GE did so for Tier 0 as well (Chen
et al. 2003).
Finally, the usage frequencies assumed by EPA for several technologies for Tier 0 were too low because they
were used by more models than anticipated. For example, EF&EE reports that Model B used electronic fuel
injectors (EFI) (Fritz et al. 2005). Note these EFI systems may not have been absolutely necessary to meet the
emission standards themselves. Rather, they were likely used to minimize the loss in fuel economy from
retarding injection timing to meet the NOx standards. In addition, EF&EE reports that new Tier 0 locomotives
(Models C and E) used split cooling (Uzkan and Lenz, 1999), increased compression ratios, and combustion
chamber design, and Chen et al. (2003) comment in their conclusions that the same technology package can
also be used to upgrade baseline engines to the same standards. As with EFI, EF&EE expects that it was not
strictly necessary to add split cooling in order to meet the standards. Rather, it was used to minimize the
need to retard injection timing, with the resulting adverse impact on fuel economy and mechanical
reliability.
194-

-------
5,4,2,2, Emission Control Technologies for Tier 1 Locomotives.
The main emission control technologies that EPA expected to be used for to comply with Tier 1 were:
•	Tier 1 locomotives would be able to incorporate all the technologies available for Tier 0 locomotives.
•	Additionally, electronic controls and enhanced aftercooling could be used for Tier 1 compliance.
Further, timing retard could be used to reduce NOx emissions without a negative impact on PM.
•	In addition, some models could use in-cylinder and turbocharger modifications.
•	Increased compression ratios could be used to reduce PM emissions and ignition delay. Upgraded
turbocharger designs would reduce smoke emissions.
All of the technologies listed were, in fact, used on line-haul locomotives in order to comply with Tier 1
standards (Dillen and Gallagher 2002). In addition, changes were made to the cylinder liner and piston rings
to reduce lubricating oil consumption and all Tier 1 units used 4-pass aftercooling (EF&EE expert opinion). As
for switch locomotives, the principal compliance mechanism was to employ non-road engines certified to Tier
1 or Tier 2 standards in genset switchers.
5,4,2,3, Emission Control Technologies for Tier 2 Locomotives.
The main emission control technologies that EPA expected to be used for compliance with Tier 2 were:
•	With the change from DC to AC traction motors, manufacturers would be using new four-stroke
engines, which would have lower PM emissions as they achieve better oil control.
•	EPA expected additional NOx and PM emission reductions to be possible through continued
refinements in charge air cooling, fuel management, and combustion chamber configuration.
•	Improved fuel management would include increased injection pressure, optimized nozzle hole
configuration, and rate-shaping.
•	Potential combustion chamber redesigns would include the use of reentrant piston bowls and
increased compression ratio.
All of the technologies considered by EPA for Tier 2 compliance were, in fact, used, in both two-stroke and
four-stroke engines (Flynn et al. 2003). Combustion chamber designs were extensively optimized, but this
optimization did not include the use of re-entrant combustion chambers. For engines in the size and speed
range, the optimal combustion chamber has been found to be wide and flat (the so-called Mexican hat
shape) rather than re-entrant. The usage frequencies noted in Table 5.5 for each technology were
reasonable, the one exception being that all Tier 2 units ended up using 4-pass aftercooling (EF&EE expert
opinion).
There were some other changes in the locomotive market in the years following the rulemaking that were
unanticipated by EPA, but for the most part these did not impact the cost of meeting Tier 2. For example, the
anticipated migration from 2-stroke to 4-stroke engine designs for EMD did not occur, but this did not create
a cost divergence because the rulemaking did not ascribe the switch to 4-strokes as being due to EPA's
program in the first place. EMD wound up using the same technologies on its two-stroke engine, and they
195 -

-------
were equally effective. Similarly, the widespread change from 4400 HP DC locomotives to 6000 HP AC
locomotives that was anticipated in 1998 has largely failed to occur. Although a substantial number of AC
locomotives are in service, line-haul locomotives with DC propulsion continue to make up a substantial
fraction of new locomotive sales. Those AC locomotives that are sold are primarily in the 4300 to 4400
horsepower range. EMD locomotives in this power range have 16-cylinder two-stroke engines, while GE
units have 12-cylinder four-stroke GEVO engines. Although DC and AC locomotives differed in their electrical
systems, there was little or no difference in the engine and emission control systems. The same engine
families were used in DC and AC locomotives, so this also should not have altered the compliance cost of
meeting the Tier 2 standards (EF&EE expert opinion).
In sum, except for the use of Tier 2 and Tier 3 non-road engines in genset switchers, we are not aware of any
major emission control technologies not considered by EPA that were actually employed in a significant
number of locomotives (EF&EE expert opinion).175
comotive Compliance Cost
5,4,3,1, Initial Compliance Cost
EPA estimated the initial cost increase to the operator as the sum of the fixed costs and variable costs of
hardware needed for compliance, adjusted by a 20 percent manufacturer's markup for overhead and profit.
Fixed Costs, EPA's fixed costs of manufacturing locomotive models compliant with the emissions standards
included costs of testing, engineering, tooling, and technical support.
•	The testing costs included developmental testing, as well as certification testing, production line
testing and in-use testing. Testing costs also included the costs of any necessary additional facilities
and equipment for emissions testing, plus engineering, operating and maintenance costs for the
testing facility. These costs, when allocated over the estimated testing requirement, were estimated
to amount to about $21,000 per test prior to 2010 and about $39,000 per test after 2010 when the
developmental testing would be completed (U.S. EPA 1998).
•	The engineering costs category represented the estimated average cost for the number of
engineering work years EPA projected to be required to develop the calibrations and hardware
necessary for meeting the emission standards. This also included the effort for any ancillary changes
made to the locomotives to accommodate the required new hardware.
•	The tooling costs included costs for any additional or modified tooling necessary to produce the
emission control hardware, as well as for any required setup changes. Because EPA estimated that
175 In the public comments on the proposed rule, EMD stated that exhaust gas recirculation (EGR) would be the
likely technology of choice for meeting Tier 2 standards. EMD also projected a 5-10 percent fuel economy penalty,
rather than the 1 percent estimated by EPA, based on the experience of others in the use of EGR. EGR was not
used to meet Tier 2 (EF&EE expert opinion).
A very small number of switch locomotives were built using alternative fuels such as LNG for demonstration
purposes, but they were not offered as commercial products.
196-

-------
Tier 0 compliance would be achieved through calibration changes or hardware obtained from
suppliers (particularly in the case of aftermarket remanufacturers), EPA did not estimate specific
tooling costs for Tier 0.
• The technical support costs included the costs of any changes that would be required in the technical
support that manufacturers provide to users, including any necessary operator or maintenance
training and changes to technical publications that provide operating and maintenance guidance.
EPA estimated these fixed costs for each locomotive supplier, multiplied by the number of suppliers for each
model type, and divided by the total number of locomotives (assuming suppliers would recover costs from
the locomotives) to derive the total per locomotive fixed cost by model type. EPA assumed that there were
three suppliers each for Tier 0 Model A, B, and C locomotives, and one supplier each for Tier 0 Model D and E,
Tier 1 Model A, B, C, and D, and Tier 2 Model A and B locomotives. EPA based this assumption on the
numbers of independent part suppliers and remanufacturers for the various locomotive models at the time
of the analysis. The number of suppliers EPA estimated for each model category was less than the total
number of suppliers in existence at the time because EPA assumed that the manufacturers for which initial
costs were cost prohibitive would pay other manufacturers with the ability to incur initial costs to perform
the necessary services.
Because the fixed costs were for goods and services that are useful for more than one year of production,
EPA amortized initial costs over 5 years (i.e., manufacturers would recover costs within the first five years of
production). For Tier 2, because the standards were to be in effect for longer than 5 years, EPA developed
two sets of unit costs (because initial fixed costs would be recovered by 2010). EPA did not calculate separate
compliance costs reflecting fully-recovered fixed costs for Tier 0 and Tier 1 as it did for Tier 2, because the
initial hardware costs occur only at original manufacture (for Tier 1) or the first remanufacture (for Tier 0),
and thus are applicable only during the first few years of the program.
Table 5.3 above summarizes the fixed costs of manufacturing for each Tier and model type that were
estimated by EPA.
Certification data published in 2005 shows that the number of suppliers, and especially the number of
different Tier 0 remanufacturing systems developed, were higher than EPA estimated. EPA estimated that a
total of 11 remanufacturing systems would be developed and certified for Tier 0 locomotive models, from a
total of three suppliers. In 2005, there were 37 remanufacturing systems certified, from four suppliers (US
EPA 2005). EPA's estimates of the cost per remanufacturing system certified are probably too high, as they
assume that the same level of effort went into certifying remanufacture systems as new engines which is
probably not the case (EF&EE expert opinion). Even taking this into account, however, the large number of
systems certified means that the total costs of certification of Tier 0 remanufacturing systems were probably
about double EPA's estimate (EF&EE expert opinion). This suggests that the total realized fixed costs for the
Tier 0 line-haul locomotives (Models B-E) were closer to $53 million (1997$) than EPA's original estimate of
$26.5 million. What this implies about the realized per locomotive fixed cost depends on how EPA's estimate
of the number of remanufactured locomotives compares to the number of locomotives actually affected by
the rule in each model category. Since the total number of locomotives to be remanufactured was over-
197 -

-------
estimated (see more on this below), the fixed cost per locomotive for remanufactured locomotives were
likely higher than EPA's estimate.
EF&EE's expert opinion indicates that EPA's assumptions regarding the total fixed costs of certification for
newly built locomotives were fairly accurate. Since the total number of newly built locomotives over 2000-
2009 was underestimated (see more on this below), the realized fixed cost per locomotive for new
locomotives were likely lower than EPA's estimate.
Variable Costs, EPA's estimate of the initial incremental variable compliance costs included costs of hardware
and assembly.
The hardware costs represented the emission reduction technologies EPA projected that manufacturers
would employ for compliance with the standards. EPA developed hardware cost estimates for the following
technologies:
•	Retarding fuel injection (2 degree or 4 degree timing retard)
•	4 pass after cooler
•	Improved mechanical and electrical injectors
•	Electronic fuel injection
•	Engine Modification
•	Improved turbocharger
•	Split cooling
•	High pressure injection
•	Combustion chamber design
Table 5.3 shows the costs assumed for each of these technologies and specifies the combinations of these
technologies that were expected to be used for each locomotive model type and Tier.
Assembly costs included the labor and overhead costs for retrofitting (in the case of Tier 0) or for initial
installation of the new or improved hardware. These also varied with the characteristics of individual
locomotives and the type of hardware necessary for compliance with the applicable emission standards.
EF&EE's expert opinion indicates that EPA's estimate of the hardware cost of each emission control
technology was reasonable. However, since the usage frequency of several technologies was higher than
EPA anticipated (as discussed in Section C.2), per locomotive total hardware costs for line-haul locomotives
were likely higher than EPA's ex ante estimate. For Tier 0, the use of electronic fuel injectors would have
added $35,000 in hardware costs for an older line-haul EMD locomotive (Model B), and the use of split
cooling, increased compression ratios, and combustion chamber design would have added about $26,000 in
hardware costs for newer line-hauls (Model C and E locomotives).176 For Tier 1 and 2, the use of 4-pass
176 These price increases are based on EPA assumed costs of these emission control technologies for other Model
types, as shown in Table 5.4.
198 -

-------
aftercooling may not have added to the hardware costs per locomotive since the aftercooling costs may have
already been included in the assumption of split cooling being used in these locomotives (EF&EE expert
opinion).
The industry move to genset switchers instead of remanufacturing old ones to comply with the new
standards means the realized Tier 0 per locomotive compliance cost was likely different that what EPA
estimated for the switch locomotives (Model A). Presumably companies found gensets to be more cost-
effective than remanufacturing to Tier 0 standards. However, it is unclear to what extent genset switchers
were developed in reaction to the rule or other factors. The genset has major benefits in terms of
availability/reliability and fuel consumption, so EF&EE's expert opinion indicates that this technological
change would likely have been undertaken even in the absence of the emission standards. Better reliability
means one unit can often replace two old conventional units, and fuel consumption is at least 50 percent
less.177 The genset switcher is significantly more expensive but costs have come down in recent years. EF&EE
reported that the current price of a new genset switcher is around $700,000 whereas a standard switcher
such as an SW1200 could be sold for about $236,000 (although that does not include the cost of
remanufacturing the engine to Tier 0).
EF&EE's expert opinion indicates that the assembly costs were reasonable for new locomotive but were likely
underestimated by a factor of two or three for remanufactured locomotives. EPA's assembly cost estimates
for remanufactured locomotives in Tier 0 were similar to those for new ones in Tier 1. However,
remanufacturing takes place in locomotive repair shops that perform a variety of activities, rather than in
assembly areas that specialize in only one locomotive model. EF&EE observed that these operations are
much less efficient. If assembly costs were double or triple what EPA estimated, this would add about $4500-
9000 per locomotive for older line-hauls meeting Tier 0 (models B and D) and close to $7000-13000 per
locomotive for newer line-hauls subject to Tier 0 (models C and E) (since remanufactured locomotives make
up most of the ones subject to Tier 0).
5,4.4,2, Remaniifactiire Costs
EPA's estimate of the costs associated with keeping locomotives in compliance with the standards through
subsequent remanufactures included:
•	Costs of replacing electronic fuel injectors every two years;
•	Costs of electronic injection wiring harnesses, which need to be replaced in Tier 0 and Tier 1
locomotives every seven years due to embrittlement of the insulation from the heat generated by
the engine;
•	Cost of improved injector replacement for Tier 2 locomotives every two to three years.
177 Estimates based on EF&EE discussion with a genset switcher company.
199 -

-------
Table 5.3 summarizes the remanufacture cost per locomotive for each Tier and model type that was
estimated by EPA.
For line-haul locomotives, expert opinion indicates that EPA's estimate of the annual remanufacture cost per
locomotive and assumptions about remanufacture frequency were reasonable (EF&EE expert opinion). On
the other hand, most switchers would not be remanufactured at all over the first decade of the program.
5,4,4,3, Fuel. Costs
EPA estimated increases in fuel consumption due to various emission control technologies and the
corresponding incremental fuel costs. Based on past developments in the industry, EPA believed that
manufacturers would make every effort to eliminate any initial fuel consumption penalties, and would have
largely succeeded by 2010. However, EPA included fuel economy penalties for the full 41 years covered by
the analysis.
As shown in Table 5.3, fuel costs made up a large share of EPA's total per locomotive cost estimates for all
model types except older line-haul models (Models B and D, Tier 0). For Tier 0, for switchers (Model A), fuel
cost makes up over 90 percent of cost of compliance. For older line-haul models (B, D), fuel cost make up
smaller share of the per locomotive compliance cost (11-35 percent). For newer line-haul models (C, E), fuel
cost make up about half (42-56 percent) of per locomotive cost. For Tier 1 and Tier 2, fuel costs account for
53-59 percent and 70-80 percent of EPA's total cost per locomotive, respectively.
EPA's estimates of per locomotive fuel costs were calculated as: average annual fuel consumption (gal/yr) *
FE penalty (%) * price ($/gal) *service life (15-21 yrs for TierO, 40 yrs for Tier 1&2). We assess each
component of the annual fuel cost calculation in turn.
Fuel price. EPA assumed a constant fuel price of $0.70 per gallon of diesel consumed (1997$). Actual prices
over the first decade of compliance were substantially higher. See Table 5.4. Locomotive fuel averaged
$1.20/gal (1997$) over 2000-2009178, or over 70 percent more than EPA's estimate (AAR 2002, 2011).179
Most of the increase in diesel price over this period was likely unanticipated. Around the time of the
rulemaking, the Energy Information Administration (EIA) was forecasting a modest increase in fuel prices -
178	This estimate includes the impact of hedging. The railroads use hedging to stabilize the impact of fuel price
volatility. In some cases, hedging saves the railroad money. In other cases, the railroad may have to spend more
for fuel then it would have without hedging. The source for the data is Annual Report Form R-l, Schedule 750.
179	The other potential source of fuel price data is the AAR Monthly Railroad Fuel Price Indexes report. The source
for this report is AAR survey of the largest Class I railroads, using a methodology decided by the Interstate
Commerce Commission. Data from this survey are used for the Rail Cost Adjustment Factor, which is required by
law to be published by the Surface Transportation Board (and earlier, the Interstate Commerce Commission).
The individual railroad pricing information is confidential. A weighted average of the fuel price (total dollars
divided by total gallons) is used to construct our index. Note that estimates based on this index indicate fuel prices
were even higher than the Railroad Facts data suggests - i.e., averaging more than $2/gal (1997$) over 2000-2009
(AAR 2001, 2003, 2006, 2009).
200-

-------
e.g., about 0.4 percent annual growth in the end user price of distillate fuel between 1995 and 2015 (EIA
1997) - but world oil prices, the main determining factor in the price of diesel, increased substantially more
than EIA was projecting at the time. Over 2000-2009, oil prices were on average 76 percent higher than
what EIA had projected in the 1997 Annual Energy Outlook (AEO) (EIA 2011).
Average annual fuel consumption per locomotive. Table 5.3 includes the fuel consumption assumptions used
for calculating fuel costs. For Tier 0, EPA assumed average annual fuel consumption per locomotive of
104,000 gallons for switchers and remanufactured older line-hauls (Models A, B, and D), 297,000 for newer
(mostly remanufactured) line-hauls (Models C and E). Average annual fuel consumption per locomotive was
assumed to be 297,000 gallons for the Tier 1 line-hauls (Models A and B), and 350,000 gallons for the
remaining Tier 1 line hauls (early versions of Tier 2 design) and all Tier 2 locomotives.
EPA assumed that fuel consumption remained constant. EPA recognized that there was a short-term trend of
increasing fuel consumption, but was not confident that the trend would continue. The long-term trend up to
that time was for fuel consumption to remain fairly constant as a result of continual improvements in
locomotive fuel economy, which offset the significant increase in ton-miles of freight hauled.
EF&EE's expert opinion is that EPA's estimates of average annual per locomotive fuel consumption were
reasonable, but there is little data available against which to check this claim. The data in Table 5.4 shows
that on a fleetwide basis per locomotive fuel consumption fluctuated in the early years of the program and
declined more significantly after 2004. Annual per locomotive fuel consumption for all Class I locomotives in
use averaged about 187,000 gallons over 2000-2001,185,000 gallons over 2002-2004, and 165,000 gallons
over 2005-2009. These fleetwide averages are lower (at least for 2002-09) than the annual fuel consumed
per locomotive assumed in EPA's analysis, but without more information on the share of fuel consumption
coming from new locomotives, it is difficult to draw ex post conclusions about this element of EPA's analysis.
The fleetwide averages could be consistent with the EPA assumptions if operators run the newest line-haul
engines more per year than the older ones in their fleet (outweighing any fuel efficiency gains from newer
models). It is also possible that annual per locomotive fuel consumption was lower than EPA estimated due
to fuel efficiency improvements in the new engines. (Since fuel efficiency of newer models is likely better
than that of older models, and since the newest engines are likely to handle more ton-miles per year than the
fleetwide average180, all we can reasonably conclude based on existing data is that annual fuel consumption
of a new locomotive was more than 186,000 gallons over 2000-2004 and more than 165,000 gallons over
2005-2009).
For switch locomotives, there is little data available with which to estimate annual fuel consumed by a new or
remanufactured switcher over 2000-2009. However, it is likely that average annual fuel consumption of
genset switchers was lower than EPA's assumed 104,000 gallons per year for a switch locomotive (Tier 0,
180 Over 2000-2006, new locomotives comprised approximately 25% of the fleet, but given the higher power and
more intensive use of newer locomotives, they probably handled 35-40% of total gross ton-miles (FRA 2009).
201 -

-------
Model A). Gensets were introduced around 2005 (EF&EE expert opinion), and currently, switcher fuel
consumption is about 40,000 to 70,000 gallons a year, or 30-60 percent lower than EPA's estimate.181
Fuel Economy Penalty, EPA used the existing engines as the fuel-economy baseline and then estimated
increases in fuel consumption due to various emission control technologies and the corresponding
incremental fuel costs. EPA assumed fuel penalties of:
•	2 percent for Tier 2 locomotives,
•	1 percent for Tier 1 locomotives, and
•	1-2 percent for Tier 0 locomotives.
Based on past developments in the industry, EPA believed that manufacturers would make every effort to
eliminate any initial fuel consumption penalties, and would have largely succeeded by 2010. However, EPA
included fuel economy penalties for the full 41 years covered by the analysis. EPA also conducted a high case
sensitivity analysis with 2-4 percent fuel economy penalties (but did not adjust assumptions about fuel price
or fuel consumption in the sensitivity analysis).
To determine the realized fuel economy penalty from compliance with the rule, one needs to compare the
actual fuel economy of new and remanufactured locomotives over 2000-2009 with the fuel economy of new
and remanufactured locomotives that would have been achieved in absence of the rule. Both of these are
extremely difficult to estimate - the former because in use, model specific fuel economy information is not
readily available from manufacturers, and the latter because locomotive manufacturers are constantly
striving to reduce fuel consumption, as this is one of the principal decision for Class I railroads in selecting a
locomotive.
For competitive reasons, locomotive manufacturers generally do not release fuel consumption data,182 and
our ability to glean anything about the realized fuel economy using existing aggregate data is extremely
limited. For example, one common measure of the fuel efficiency of freight rail is revenue ton-miles per
gallon of fuel consumed. By this measure, as shown in Table 5.4, the overall fuel efficiency of Class I rail has
consistently improved over time, especially after 2005. As with the fuel consumption estimates discussed
above, however, these measures provide an underestimate of the fuel economy of locomotives subject to
the rule, since newer (and rebuilt) engines will have higher fuel efficiency than the fleetwide average. A
slowdown in rebuild frequency would also be reflected in the observed fleetwide change in fuel efficiency. If
we could make reasonable assumptions about the percentage of total fuel consumed and travel done by new
line-haul locomotives, then we could apply these shares along with data on the number of new locomotives
to get rough estimates of how much fuel economy of new line-haul locomotives improved over 2000-2009.
181	Estimate based on EF&EE discussion with a genset switcher company.
182	See, for instance, Figure 2 of Flynn et al. (2003), which shows the general relation between NOx and fuel
economy, but omits the units from the fuel-economy axis.
202 -

-------
Even so, the challenge of constructing the counterfactual would remain. Given the long term trend of
improved fleetwide rail efficiency observed before the rule,183 and projections made in the year before the
rule was promulgated,184 the fuel economy of new locomotives may have increased even more than
observed over 2000-2009 in absence of the emission standards. However, with other changes going on in the
industry over this period (e.g., increasing share of unit train service, increasing congestion),185 we are
skeptical that it will be possible to identify a fuel economy change attributable to the rule based on aggregate
data.
Model specific information from the trade press indicates that manufacturers were able to develop new
locomotives and remanufacture kits to meet emission standards without sacrificing fuel economy. For
example, in 2009 EMD Tier 0+ kits offered up to 2 percent fuel savings versus previous engine
configurations.186 It is unclear, however, to what extent fuel economy improvements would have been
implemented in the absence of the rule. It is therefore also unclear to what extent fuel economy
improvements actually achieved were motivated by the rule and associated actions to comply. Locomotive
suppliers would have had incentive to continue to look for ways to offer improvements in fuel efficiency,
especially in the face of rising fuel prices, so it is possible that they would have been able to tweak existing
models or introduce even more fuel-efficient ones in the absence of pollution controls.
Compared to a counterfactual case in which the locomotive manufacturers were able to use the latest
technical advances to optimize fuel consumption without regard to NOx or PM emissions, EF&EE expert
opinion is that the fuel consumption penalty was higher than anticipated, probably about 2 to 4 percent. This
is based on experience and professional judgment, and interpretation of optimization studies undertaken on
an EMD 710-series locomotive engine (Dolak and Bandyopadhyay 2011), however, and not on public-domain
data. Dolak and Bandyopadhyay (2011) show that even for engines developed to meet Tier 2 standards,
there remains a tradeoff between NOx and fuel-efficiency. The results shown in the paper suggest that, for
183	Based on data in Table 5.4, revenue ton-miles per gallon fuel consumed increased on average nearly 2%
annually between 1990 and 2000 (AAR 2002).
184	EIA forecast in the year before the rule was promulgated projected a continued increase in efficiency. Overall
rail efficiency (ton miles per BTU) was forecast to achieve on average a 1% improvement annually between 1995
and 2015 (EIA 1997).
185	Unit train service, typically 100 cars or more, is loaded at the origin point with one commodity follows a direct
route to the destination point without passing through yards or terminals on the way and remains intact. Most unit
trains are either intermodal or coal trains, It is more fuel efficient than carload service which is a fuel-intensive
operation because of the need for switch engines in breaking up trains and making new ones in every terminal
through which the shipment passes. In recent years, there has been a strong trend towards unit trains—partly due
to the growth of intermodal traffic from West-coast ports and coal traffic from the Powder River Basin (FRA 2009).
186	See, for example, article in Progressive Railroading, August 2009, "Locomotive Manufacturers Offer
Information on their Fuel-Saving Models",
http://www.progressiverailroading.com/mechanical/article/Locomotive-Manufacturers-Offer-lnformation-on-
their-FuelSaving-Models-21139#.
203 -

-------
the range of plausible injection timing settings, the difference between lowest NOx (subject to PM
limitations) and lowest fuel consumption fuel efficiency is roughly 2 to 4 percent in fuel efficiency.
In addition, it is important to keep in mind that efforts to control emissions may lead to other improvements
in production processes and/or equipment which would not have occurred in the absence of the regulation.
Manufacturers could have added technologies to new locomotives and remanufacture kits that were not
strictly needed to comply with the emission standards but helped to offset any fuel economy loss from the
pollution controls. The Tier 0 discussion in Section C.2 above and the locomotive manufacturer's own
assessment187 suggest that this occurred. In this case, the fuel penalty associated with operating costs would
be offset to some unknown extent, though an additional hardware cost would be attributable to the
regulation.
As for switch locomotives, EPA assumed this group could be brought into compliance with Tier 0 by retarding
injection timing alone, with a fuel economy penalty of only 2 percent. EF&EE's expert opinion is that
additional changes were also needed - i.e., improvements in fuel injectors at a minimum. In practice,
however, very few if any, of these units were remanufactured. Some operators instead moved to genset
switchers which, as already mentioned, had significant fuel savings compared to conventional older
switchers. One industry source reports fuel cost savings with a genset are at least 50 percent (EF&EE);
another reports "fuel savings of more than 20 percent, compared to existing diesel locomotive technology in
side-by-side use, have been demonstrated."188 However, most purchases of gensets or hybrids to date have
been financed in part with air quality improvement grants, and it may be hard to compete with existing four-
axle locomotives on the second-hand market (FRA 2009).
« < ^ -a- v. ^ ' • *• • v :
As stated at the outset, the purpose of this paper is not to review the ex ante cost analysis of the 1998
Locomotive rule. Rather, the goal is to explore available data to gauge whether actual compliance costs may
have diverged from ex ante cost estimates and, if so, what factors might have contributed to any divergence
(e.g., changing market conditions, technological innovation, etc.) as described in Chapter 1 of this report. Our
findings are summarized in Table 5.6 and discussed briefly below.
We encountered significant methodological challenges in conducting an ex post assessment of the 1998
Locomotive rule. There is a paucity of data needed to calculate various components of the realized costs,
especially information on the actual costs of individual control technologies, and data on fuel consumption
and fuel economy of new and remanufactured locomotives. We are also extremely limited in our ability to
187	Lawson, Pete, General Electric Transportation Systems, Faster Freight Cleaner Air Conference, Long Beach, CA,
February 27, 2007, www.fasterfreightcleanerair.com/presentations.html#California2007. Also see GE's
promotional materials for the Evolution Series locomotive:
http://www.qetransportation.com/resources/doc download/275-evoloution-series-enqine.html
188	http://www.gwrr.com/about_us/community_and_environment/gwi_green/genset_locomotives.be
204-

-------
Table 5.6: Summary of Findings -
Components of Cost Estimate
Source of Ex Post Information
Assessment (Compared to Ex Ante)
Regulated
Types of Entities
EF&EE
Reasonable
Universe
Number of Entities
AAR for all Class 1
EF&EE for switch
New - Higher
Remanufactured - Lower
Switch - Lower
Methods of
Types
EF&EE + journal articles
Reasonable
Compliance
Usage
EF&EE + journal articles
Higher than anticipated for some technologies on some model
types
Per
Locomotive
Direct,
One-Time
Per Locomotive
Fixed Cost
EF&EE + EPA certification data
New- Reasonable
Remanufactured - Higher than projected
Compliance
Costs

Per Locomotive
Variable Cost
Hardware Costs:
EF&EE + journal articles
Hardware Costs:
Line Haul - Higher than projected
Switch - Inconclusive



Assembly Costs:
EF&EE
Assembly Costs:
New- Reasonable
Remanufactured - Higher than projected

Direct, On-
Operating
Fuel price: AAR
Fuel price: Higher than projected

Going
(Additional Fuel
Costs)
Annual Fuel Consumption:
EF&EE for line haul,
genset websites for switch
Annual Fuel Consumption:
Line Haul - Reasonable
Switch - Lower



Fuel economy penalty:
EF&EE+ journal articles, AAR, FRA,
manufacturer promotional materials
Fuel economy penalty:
Line Haul - Likely higher
Remanufactured Switch -Likely higher


Maintenance
EF&EE
Reasonable

Indirect


Opportunity Costs


Total Per Locomotive Cost

Line Haul - Likely higher
Switch - Inconclusive (difficult to assess whether alternative
technology would have been developed in absence of the rule)
TOTAL COSTS

Line Haul - INCONCLUSIVE
Switch - LIKELY LOWER (very few remanufactured and new units
adopted alternate technology, but with some support from air
quality grants)
205 -

-------
construct a reasonable counterfactual for each component of the cost analysis. For example, to the extent
that more efficient line-haul locomotives (through advancements in engine design, cooling systems, etc.)
would have been developed and adopted overtime in the absence of the rule, the costs of these
technologies should not be attributed to the 1998 rule, and the costs of the Tier 1 and Tier 2 standards were
less than EPA's ex ante estimate. Due to data limitations and our minimal ability to speculate about what
would have occurred in the absence of the rule, most of our assessment is limited to comparing the opinion
of one industry expert about how industry complied with the emission standards and some ex post
information to what EPA assumed. Finally, examining whether EPA's method for building up the fixed costs
of compliance provides an accurate reflection of the true initial cost is outside the scope of our preliminary
analysis. We have not investigated the extent to which the 20 percent manufacturer markup on per
locomotive initial compliance cost was appropriate. We are also not able to determine to what extent
manufacturers and remanufacturers used average, banking and trading provisions of the rule to meet overall
emissions goals at lower cost.
Keeping the above caveats in mind, a number of EPA's ex ante estimated or assumed cost factors were fairly
similar to the limited ex post empirical data and EF&EE opinion. These assumptions include: locomotive
model types, the types of compliance technologies, fixed costs and assembly costs for newly manufactured
locomotives, hardware costs of each emission control technology, and annual remanufacture costs per
locomotive. However, our assessment identified other areas in which the ex ante estimates differed from
the realized per-unit compliance costs over the first decade of the program (2000-2009). First, the initial per-
unit costs for remanufactured line-haul locomotives (Tier 0) were likely higher than EPA estimated because
the large number of remanufactured engine families certified and the smaller number of units
remanufactured increased the fixed cost per locomotive. Second, increased usage rates for some
technologies caused variable costs for remanufactured locomotives to be higher than the EPA estimates for
most model types. Third, operating costs per locomotive (new or remanufactured) imposed by the rule may
have been higher than anticipated because actual fuel prices were much higher than EPA assumed. This
implies, the same percentage fuel consumption penalty could have contributed to higher dollar cost due to
higher fuel prices; over the first decade of the program, total per locomotive costs could have been 5-32
percent higher for Tier 0 (line-hauls built 2000-2001 or remanufactured), 14-19 percent higher for newly built
line-haul locomotives over 2002-2004 (Tier 1), and 36 percent higher for newly built line-haul locomotives
over 2005-2009 (first five years of Tier 2).189
The impact of the higher fuel price may have been offset to some extent by lower fuel consumption and/or
lower fuel penalties than anticipated by EPA. The information available to us suggests that manufacturers
were able to reduce fuel penalties from the pollution controls by designing more fuel efficient locomotives,
but we are unable to quantitatively assess how the additional costs incurred to bring about these fuel
efficiency improvements compare to the ex ante fuel economy penalty costs of the rule. In addition, the
189 These percentages are calculated with only 10 years of the fuel and remanufacture costs as a way to
approximate the operating costs incurred until each locomotive is remanufactured to the revised standards.
Attributing all operating costs over the remaining life of the locomotive to the 1998 rule would be inappropriate
given the 2008 revisions to the standards.
206

-------
difficulty in constructing the counterfactual remains. Given the strong incentive for manufacturers to
improve fuel efficiency, especially in the face of rising fuel prices as occurred in the 2000s, it is likely that fuel
efficiency improvements would have occurred over time in the absence of the regulation. In fact, compared
to the counterfactual case in which the locomotive manufacturers would have used the latest technical
advances to optimize fuel consumption without regard to NOx and/or PM emissions, it is possible that the
fuel economy penalties were higher than EPA's assumptions, which would further increase the fuel costs of
compliance. Taken together, these issues suggest that, given the information currently available to us, it is
extremely difficult to estimate the extent to which the impact of higher fuel price may have been offset by
changes in other components of the fuel cost of the rule. However, even setting aside the operating cost
impact of the rule, EF&EE expert opinion and accompanying information about the variable and fixed costs of
compliance suggest that the total per locomotive cost was likely higher than EPA's ex ante analysis projected
for most new line-haul and especially most remanufactured line-haul locomotives subject to the rule over
2000-2009.
Our ex post assessment of the total cost of bringing line-haul locomotives into compliance with the 1998 rule
is inconclusive. This is because total compliance cost depends not only on the per locomotive compliance
cost but also on the number of locomotives affected by the regulation. Over 2000-2009, the number of
newly built line-haul locomotives was higher but the number of remanufactured line-haul locomotives was
lower than EPA's estimate. It is difficult to tease out the extent to which this was driven by an industry
reaction to the 1998 rule (or the 2008 rule) or by external factors. If operators found it to be more cost-
effective to buy new rather than remanufacture the old units to Tier 0 standards, then it would be
inappropriate to conclude that the higher-than-expected sales of new Tier 2 locomotives added to the cost of
complying with the standards without accounting for the offsetting savings from lower maintenance and
fewer remanufactures over this time period. It is possible that the lower costs due to far fewer
remanufactures taking place than anticipated may have outweighed the higher compliance costs from new
line-hauls.
The total costs of bringing switch locomotives into compliance with the 1998 rule was likely lower than
anticipated by EPA, but this has not had a major impact on overall costs of the 1998 locomotive rule because
switchers comprise a relatively minor part of the overall locomotive market. Any new switch locomotives
sold would be of the genset type, which have higher initial costs but lower fuel and maintenance costs than
the conventional switchers EPA anticipated would be remanufactured to meet emission standards, but
without knowing to what extent the development of gensets would have occurred in absence of the rule, it is
difficult to draw conclusions about the total per locomotive cost of compliance for this segment of the
market. Regardless, the large supply of old locomotives that can be kept running at low cost limits the
potential sales of new switchers and old ones can be run for a long time without remanufacturing so very few
switch locomotives were likely remanufactured over 2000-2009.
207

-------
Chapter 5 References
American Association of Railroads (AAR). 2009. Monthly Railroad Fuel Price Indexes.
American Association of Railroads (AAR). 2006. Monthly Railroad Fuel Price Indexes.
American Association of Railroads (AAR). 2003. Monthly Railroad Fuel Price Indexes.
American Association of Railroads (AAR). 2001. Monthly Railroad Fuel Price Indexes.
American Association of Railroads (AAR). 2011. Railroad Facts, 2011 Ed.
American Association of Railroads (AAR). 2002. Railroad Facts, 2002 Ed.
Chen, G., P.L. Flynn, S.M. Gallagher, and E.R. Dillen. 2003. Development of the Low-Emission GE 7-FDL High-
Power Medium-Speed Locomotive Diesel Engine. J. Engineering for Gas Turbines and Power 125:
505-512.
Dillen, E.R., and S.M. Gallagher. 2002. Development of the Tier 1 Emission Reduction Control Strategy for GE-
7FDL High-Power Medium-Speed Locomotive Diesel Engine. Proceedings of the 2002 Fall Technical
Conference oftheASME Internal Combustion Engine Division, ICEF2002-504.
Dolak, J., and D. Bandyopadhyay. 2011. A Computational Investigation of Piston Bowl Geometry for a Large
Bore Two-cycle Diesel Engine. Proceedings of the ASME Internal Combustion Engine Division's 2011
Fall Technical Conference, ICEF2011-60155.
Eastern Regional Technical Advisory Committee (ERTAC). 2012. Rail Emissions Inventory. Available at:
http://www.gaepd.org/air/airpermit/downloads/planningsupport/regdev/locomotives/class2 3 doc
2012.pdf.
Energy Information Administration (EIA). 2011. AE02010 Retrospective Review. Available at:
http://205.254.135.7/forecasts/aeo/retrospective/.
Energy Information Administration (EIA). 1997. Annual Energy Outlook. Available at:
ftp://ftp.eia.doe.gov/forecasting/038397.pdf.
Federal Railroad Administration (FRA). 2009. Comparative Evaluation of Rail and Truck Fuel Efficiency on
Competitive Corridors. Available at:
http://www.fra.dot.gov/Downloads/Comparative Evaluation Rail Truck Fuel Efficiency.pdf.
Flynn, P., P. Hupperich, S. Napierkowski, and E. Reichert. 2003. General Electric GEVO Engine For Tier 2
Locomotive Application. Proceedings of the 2003 Fall Technical Conference of the ASME Internal
Combustion Engine Division, ICEF2003-708.
Fritz, S., J. Hedrick, and B. Smith. 2005. Exhaust Emissions from a 1,500 kW EMD 16-645-E Locomotive Diesel
Engine Using Several Ultra-Low Sulfur Diesel Fuels. ASME Paper No. ICEF2005-1228.
Harrington, W., R.D. Morgenstern, and P. Nelson. 2000. On the accuracy of regulatory cost estimates.
Journal of Policy Analysis and Management 19(2): 297 - 322.
US EPA. 2005. Locomotive Compression-Ignition Engines Certification Data. Available at:
http://www.epa.gOv/otaq/certdata.htm#locomotive.
208

-------
US EPA. 1998. Locomotive Emission Standards Regulatory Support Document. Available at:
http://www.epa.gov/oms/regs/nonroad/locomotv/frm/locorsd.pdf.
US EPA. 1997. Regulatory Announcement: Final Emissions Standards for Locomotives. Available at:
http://www.epa.gov/otaq/regs/nonroad/locomotv/frm/42097048.pdf.
Uzkan and Lenz 1999. On the Concept of Separate Aftercooling for Locomotive Diesel Engines. Journal of
Engineering for Gas Turbines and Power 121:205-210.
209

-------
Appendix 5.1: EPA's Emission Standards for
Locomotives and Locomotive Engines
The purpose of this questionnaire is to collect information and feedback from industry experts on the U.S.
Environmental Protection Agency's analysis of compliance costs for the emission standards rule for
locomotives as undertaken for rule development in 1998. The goal of this project is to assess whether
EPA's estimates of compliance costs at the time of rule promulgation were accurate. We also want to
determine whether EPA correctly identified all the process technologies that were available to reduce
emissions from locomotives.
This questionnaire summarizes the assumptions and cost estimation framework used by EPA to
determine the costs of treatment technologies that were identified as candidates for compliance with the
locomotives emissions standards rule. We want to assess whether the actual costs of emission reduction
treatments differed substantially from EPA's estimates at the time of rule development. In addition, we
hope to understand the reasons for potential differences in these estimates, including insight into whether
new or modified treatment technologies may have been implemented to meet the emission standards,
which EPA did not account for in its cost analysis.
According to the Paperwork Reduction Act of 1995, an agency may not conduct or sponsor, and a person
is not required to respond to, a collection of information unless it displays a valid OMB control number.
The valid OMB control number for this information collection is 2090-0028.
On April 16, 1998, EPA published a rule for a comprehensive emission control program that subjected
locomotive manufacturers and railroads to emission standards, test procedures, and a full compliance
program. The rule was applicable to all locomotives manufactured in 2000 and later, and any
remanufactured locomotive originally built after 1973. The rule exempted locomotives powered by an
external source of electricity, steam-powered locomotives, and locomotives newly manufactured prior to
1973.
The rule established three separate sets of emission standards (Tiers), with applicability of the standards
dependent on the locomotive's date of manufacture:
•	Tier 0 applied to locomotives and locomotive engines originally manufactured from 1973
through 2001;
•	Tier 1 applied to locomotives and locomotive engines originally manufactured from 2002
to 2004;and
•	Tier 2 applied to locomotives and locomotive engines originally manufactured in 2005 or
later.
Table 1 presents the emission and smoke standards for each locomotive tier.
210

-------
Table 1. Summary of Emission and Smoke Standards for Locomotive Rule
Locomotive Type
Gaseous and Particulate Emissions
Smoke Standards


(g/bh
p-hr)

(% Opacity-Normalized)

HC2
CO
NOX
PM
Steady
State
30-sec
Peak
3-sec
Peak
Tier 0 Line-haul Duty-cycle
1.00
5.0
9.5
0.60
30
40
50
Tier 0 Switch Duty-cycle
2.10
8.0
14.0
0.72
30
40
50
Tier 1 Line-haul Duty-cycle
0.55
2.2
7.4
0.45
25
40
50
Tier 1 Switch Duty-cycle
1.20
2.5
11.0
0.54
25
40
50
Tier 2 Line-haul Duty-cycle
0.30
1.5
5.5
0.20
20
40
50
Tier 2 Switch Duty-cycle
0.60
2.4
8.1
0.24
20
40
50
In 2008, EPA adopted a new set of emission standards, Tier 3 and Tier 4, for locomotives newly
manufactured or remanufactured after 2008. Therefore, the universe of locomotives that were subject
to the 1998 rule would be limited to locomotives originally built or remanufactured between 2001
and 2008, after which the 2008 revision took effect for newly manufactured or remanufactured
locomotives. The 1998 rule's emission standards continue to apply to locomotives built or
remanufactured between 2001 and 2008 after 2008 until they are remanufactured or taken out of service.
EPA estimated the costs for the 1998 rule through 2040 to ensure complete fleet turnover due to the long
service life of the typical locomotive. However, because the 1998 rule no longer applies to all the
locomotives for which EPA estimated costs due to the promulgation of the 2008 rule, the most relevant
costs for this analysis are likely the annual per locomotive costs (and not the total 40-year or net present
value costs).
Section 2 Compliance Technologies
To estimate costs of the proposed rule, EPA projected the number of new and remanufactured
locomotives for several categories defined by emission standard and locomotive model type. This section
discusses the emission control technologies that EPA expected would already be available at the time the
locomotive emissions standards would take effect. Among those, EPA considered use of the following
technologies:
•	Retarding fuel injection - optimizing injection timing and duration to achieve significant
NOx emissions reductions at minimal cost (2 degree or 4 degree timing retard depending
on potential fuel economy impacts);
•	4 pass after cooler - changing from two-pass to a four-pass aftercooler to lessen the
degree of timing retard needed through enhanced charge air cooling;
•	Improved mechanical and electrical injectors - optimizing spray pattern from the nozzle in
conjunction with the configuration of the combustion chamber and induction swirl to
achieve emission reductions;
•	Add electronic fuel injection - to improve control of injection rate and timing;
•	Engine Modifications - reduction in engine size to achieve the desired lower power rating;
•	Improved turbocharger -ensuring that fuel consumption and emissions formation are
minimized, including preventing smoke generation due to turbo lag; changing the
geometry of the gas flow passages in the turbine to improve the response time of the
turbocharger;
211

-------
•	Split cooling - an aftercooler that uses a coolant system separate from the engine coolant
system;
•	High pressure injection - to shorten the duration of the fuel injection event, which allows
a delay in the initiation of fuel injection causing lower peak combustion temperatures and
reduced NOx formation, and also reduces fuel economy penalties associated with
retarded injection timing; and
•	Combustion chamber design - redesign of the shape of the combustion chamber and the
location of the fuel injector to optimize the motion of the air and the injected fuel with
respect to emission control.
The effective use of some of these technologies can be optimized through the use of other technologies,
and adverse effects of some technologies can be limited or eliminated through the application of other
technologies. For this reason, in estimating compliance costs EPA considered use of multiple
technologies together to form a larger emission reduction system.
The emission control technologies that EPA expected to be used for each of the Tiers are discussed
below.
Emission Control Technologies for Tier 0 Locomotives
•	Locomotives equipped with turbocharged engines would be able to employ:
modified/improved fuel injectors, enhanced charge air cooling, injection timing retard, and
in some cases, improved turbochargers, to reduce NOx emissions.
•	EPA expected that engine coolant would continue to be the cooling medium in most
cases, rather than a separate cooling system, and that it would be cost-effective to
replace two-pass aftercoolers with four-pass aftercoolers during the remanufacturing
process.
•	The tools available to manufacturers to reduce emissions for naturally-aspirated and
Roots-blown engines would be modifications to the fuel system, modifications to the
combustion chamber and injection timing.
Q1a: Were all Tier 0 emission control technologies captured by EPA? Were there any emission
control technologies that were never used to achieve compliance? Were there any additional
emission control technologies or substantially modified emission control technologies used to
achieve compliance? If so, please explain.
A1a:»
Emission Control Technologies for Tier 1 Locomotives
•	Tier 1 locomotives would be able to incorporate all the technologies available for Tier 0
locomotives.
•	Additionally, electronic controls and enhanced aftercooling could be used for Tier 1
compliance. Further, timing retard could be used to reduce NOx emissions without a
negative impact on PM.
212

-------
•	In addition, some models could use in-cylinder and turbocharger modifications.
•	Increased compression ratios could be used to reduce PM emissions and ignition delay.
Upgraded turbocharger designs would reduce smoke emissions.
Q2a: Were all Tier 1 emission control technologies captured by EPA? Were there any emission
control technologies that were never used to achieve compliance? Were there any additional
emission control technologies or substantially modified emission control technologies used to
achieve compliance? If so, please explain.
A2a:»
Emission Control Technologies for Tier 2 Locomotives
•	With the change from DC to AC traction motors, manufacturers would be using new four-
stroke engines, which would have lower PM emissions as they achieve better oil control.
•	EPA expected additional NOx and PM emission reductions to be possible through
continued refinements in charge air cooling, fuel management, and combustion chamber
configuration.
•	Improved fuel management would include increased injection pressure, optimized nozzle
hole configuration, and rate-shaping.
•	Potential combustion chamber redesigns would include the use of reentrant piston bowls
and increased compression ratio.
Q3a: Were all Tier 2 emission control technologies captured by EPA? Were there any emission
control technologies that were never used to achieve compliance? Were there any additional
emission control technologies or substantially modified emission control technologies used to
achieve compliance? If so, please explain.
A3a: »
Q3b: Were selective catalytic reduction and/or alternative-fueled engines used as emission
control strategies? How often were they used?
A3b: »
EPA assumed that the Tier 0 locomotives could be grouped into 5 model categories (or engine families):
switch locomotives from Electro-Motive Diesel (Model A), older and newer line-haul locomotives from the
Electro-Motive Diesel (Model B and C), and older and newer line-haul locomotives from General Electric
Transportation Systems (Model D and E). For Tier 1 locomotives, EPA believed that early versions of the
new engine designs used to meet the Tier 2 standards made their appearance during the Tier 1 period.
Thus, EPA assumed there would be two Tier 1 models for each of the two manufacturers. EPA assumed
that for Tier 2 locomotive each manufacturer would have a single model.
213

-------
Table 2 presents a crosswalk between the expected compliance technologies, their usage, and the
locomotive model types by tier.
Table 2: Control Options, Expected Usage and Locomotive Models
Tier
Expected
Technology Usage
and Models
Developed for Cost
Analysis
2 deg timing retard
4 deg timing retard
4 pass aftercooler
Improved mechanical
injectors
Add electronic fuel
injection
Improved electronic
injectors
Increased
compression ratio
Improved
turbocharger
Split cooling
High pressure
injection
Combustion chamber
design

Percent locomotives
using technology
50
50
60
30
13
27
20
30
-
-
-
Tier 0
(1973-
2001)
Models using technology
A
X
X









B
X
X
X
X







C
X

X


X






D
X

X

X


X




E
X




X
X
X




Percent locomotives
using technology
100
-
-
-
-
100
50
25
75
100
100
Tier 1
Models using technology
(2002-
A
X




X
X

X
X
X
2004)
B
X




X
X
X

X
X

C
X




X


X
X
X

D
X




X


X
X
X
Tier 2
Percent locomotives
using technology
-
100
-
-
-
100
-
-
100
100
100
(2005-
Models using technology
2010)
A

X



X


X
X
X

B

X



X


X
X
X
Tier 2
(after 2010)
Percent locomotives
using technology
-
100
-
-
-
100
-
-
100
100
100
Models using technology
A

X



X


X
X
X

B

X



X


X
X
X
Source: U.S. EPA (1998).
Q4a: Based on your professional knowledge and experience, were the expected usage
frequencies for each technology considered by EPA for each Tier representative of actual
technology usage frequencies over the time period 1998 to 2008? If not, please explain.
A4a: »
214

-------
Q4b: Based on your professional knowledge and experience, were models used by EPA to
estimate costs for each Tier representative of the actual locomotive models employed for
compliance with the Locomotive rule over the time period 1998 to 2008?
A4b: »
Q4c: If not, were there any other locomotive models (aside from the ones used by EPA) that were
compliant with the rule? If so, please describe.
A4c: »
Section 3 Estimated Number of Locomotives
EPA estimated the number of newly manufactured and remanufactured locomotives based on information
on the number of locomotives currently in service and existing production, remanufacture, and retirement
rates for Class I, II, and III and passenger rail locomotives190.
EPA obtained information on Class I locomotives from the Association of American Railroads Annual
Railroad Facts publication. About 17,500 of Class I locomotives were manufactured post 1972, most of
which were used in line-haul service (Tier 0, Models B through E). The 3,500 older locomotives that were
manufactured prior to 1972 are used as switchers (Tier 0, Model A). EPA assumed that by 2008, almost
all 1973 through 1999 line-haul locomotives (13,200) would be remanufactured to meet EPA's standards.
EPA also assumed there would be 400 newly manufactured line-haul locomotives for years 2000-2004,
600 for years 2005-2010, and 300 new units for all subsequent years.
For Class II and III locomotives, EPA obtained information from American Short Line Railroad
Association, which represents most Class II and Class III railroads. EPA projected that there would be
about 600 post-1972 locomotives and 3600 older locomotives in the 1999 Class II and III fleet (Tier 0,
Models A through C). EPA assumed that during the first 10 years of the program, Class II and III railroads
would bring about 50 locomotives into compliance with Tier 0 standards each year. EPA further assumed
that in 2012, these railroads would purchase about 150 complying Tier 0 locomotives each year from
Class I railroads.
For passenger locomotives, EPA primarily relied on information from Amtrak and the American Public
Transportation Association. There were roughly 463 diesel locomotives in commuter rail service in 1995,
with 397 of these manufactured after 1972. EPA projected that about 100 locomotives would be brought
into compliance during each of the first five years of the program, and that all uncontrolled locomotives
would be removed from passenger service by 2011.
190 In 1994, Surface Transportation Board (STB) classified a railroad as Class I if its revenue was higher than $255.9
million. Railroads with revenue between $20.5 and $255.8 millions were considered Class II, while railroads with
annual revenue less than $20.5 million were Class III.
215

-------
Table 2 shows the estimated total number of locomotives in each Tier for each model type.
Table 2: Estimated Number of New and Remanufactured Locomotives Affected by the Rule
Tier
Model
Number of Locomotives
Tier 0 (1973-2001)
A
3,000
B
4,900
C
2,930
D
2,035
E
2,965
Total
15,830
Tier 1 (2002 - 2004)
A
360
B
360
C
360
D
360
Total
1,440
Tier 2 (2005-2010)
A
1,700
B
1,700
Total
3,400
Tier 2 (after 2010)
A
300
B
300
Total
600
Source: U.S. EPA (1998).
Note that because EPA adopted new standards applicable to any locomotives manufactured after 2008,
EPA's estimate of Tier 2 locomotives after 2010 is not relevant.
Q5a: Was EPA's estimate of the number of locomotives affected by each Tier of standards
accurate? If not, please explain why or how the estimate is inaccurate.
A5a: »
Q5b: If possible, please provide an estimate of the number of locomotives affected by each Tier of
standards for each model type in the table below
A5b: »
216

-------
Tier
Model
Number of
Class I
Locomotives
Number of
Class II
Locomotives
Number of
Class III
Locomotives
Number of
Passenger
Locomotives

A





B




Tier 0 (1973-2001)
C




D





E





Total





A





B




Tier 1 (2002 - 2004)
C





D





Total





A




Tier 2 (2005-2010)
B





Total




Section 4 Costs
Manufacturers who produce new locomotives incurred fixed costs (initial investments made before the
beginning of production) and variable costs (production costs proportional to the number of locomotives
manufactured) that were dependent on the technology and emission standard.
The incremental costs incurred by the manufacturers (along with the assumed 20% manufacturer
markup)
-increased the prices of the new locomotives that were purchased by the operators. This increase in price
was the initial cost of compliance experienced by the operators. In addition to the initial costs, the
operators were expected to incur the following operation and maintenance costs: remanufacture costs
(i.e., costs associated with keeping the locomotive in compliance with the standards through subsequent
remanufactures) and fuel costs (i.e., cost of fuel economy penalties associated with compliance).
Detailed descriptions of each type of cost and EPA's assumptions are provided in the sub-sections below.
Table 3 summarizes the cost per locomotive estimated by EPA for each Tier and model type.
217

-------
Table 3: Calculation of Per Locomotive Compliance Costs (1997 US Dollars)

TierO
Tier 1
Tier 2 (2005-2010)
Tier 2 (After 2010)
Cost Component
Model A
Model B
Model C
Model D
Model E
Model A
Model B
Model C
Model D
Model A
Model B
Model A
Model B
Number of Locomotives
3000
4900
2930
2035
2965
360
360
360
360
1700
1700
300
300
Initial Costs
Variable Costs
Hardware Costs













2 deg timing retard
$0
$0
$0
$0
$0
$0
$0
$0
$0
-
-
-
-
4 deg timing retard
$0
$0
-
-
-
-
-
-
-
$0
$0
$0
$0
4 pass aftercooler
-
$5,000
$5,000
$5,000
-
-
-
-
-
-
-
-
-
Improved mechanical injectors
-
$800
-
-
-
-
-
-
-
-
-
-
-
Add electronic fuel injection
-
-
-
$35,000
-
-
-
-
-
-
-
-
-
Improved electronic injectors
-
-
$2,000
-
$2,000
$2,000
$2,000
$2,000
$2,000
$2,000
$2,000
$2,000
$2,000
Increased compression ratio
-
-
-
-
$800
$800
$800
-
-
-
-
-
-
Improved turbocharger
-
-
-
$25,000
$25,000
-
$25,000
-
-
-
-
-
-
Split cooling
-
-
-
-
-
$25,000

$25,000
$25,000
$25,000
$25,000
$25,000
$25,000
High pressure injection
-
-
-
-
-
$2,000
$2,000
$2,000
$2,000
$2,000
$2,000
$2,000
$2,000
Combustion chamber design
-
-
-
-
-
$800
$800
$800
$800
$800
$800
$800
$800
Assembly costs
$0
$4,480
$6,720
$4,480
$6,720
$6,720
$6,720
$560
$560
$560
$560
$560
$560
Subtotal Variable cost per locomotive
$0
$10,280
$13,720
$69,480
$34,520
$37,320
$37,320
$30,360
$30,360
$30,360
$30,360
$30,360
$30,360
Fixed Costs
Engineering costs
$800,000
$1,700,000
$2,800,000
$1,700,000
$2,800,000
$3,600,000
$3,600,000
$3,600,000
$3,600,000
$4,000,000
$4,000,000
-
-
Testing costs
$422,783
$422,783
$845,566
$422,783
$845,566
$4,227,829
$4,227,829
$4,227,829
$4,227,829
$8,455,659
$8,455,659
$582,900
$582,900
Tooling
-
-
-
-
-
$1,000,000
$1,000,000
$1,000,000
$1,000,000
$1,000,000
$1,000,000
-
-
Technical support
$200,000
$350,000
$500,000
$350,000
$500,000
$500,000
$500,000
$350,000
$350,000
$350,000
$350,000
-
-
Total fixed costs per supplier
$1,422,783
$2,472,783
$4,145,566
$2,472,783
$4,145,566
$9,327,829
$9,327,829
$9,177,829
$9,177,829
$13,805,659
$13,805,659
$582,900
$582,900
Total Fixed Costs1
$4,268,409
$7,418,409
$12,436,818
$2,472,803
$4,145,606
$9,328,029
$9,328,029
$9,178,029
$9,178,029
$13,806,059
$13,806,059
$582,915
$582,915
Subtotal Fixed cost per locomotive2
$1,423
$1,514
$4,245
$1,215
$1,398
$25,911
$25,911
$25,495
$25,495
$8,121
$8,121
$1,943
$1,943
Initial Cost Per Locomotive3
$1,707
$14,153
$21,558
$84,834
$43,102
$75,877
$75,877
$67,025
$67,025
$46,177
$46,177
$38,764
$38,764
Fuel Costs
Average Fuel Consumption
104000
104000
297000
104000
297000
297000
297000
350000
350000
350000
350000
350000
350000
FE Penalty
2%
1%
1%
1%
2%
1%
1%
1%
1%
2%
2%
2%
2%
Gallons of fuel/year4
2,080
1,040
2,970
1,040
5,940
2,970
2,970
3,500
3,500
7,000
7,000
7,000
7,000
Cost per year (@ $0.70/Gal.)
$1,456
$728
$2,079
$728
$4,158
$2,079
$2,079
$2,450
$2,450
$4,900
$4,900
$4,900
$4,900
Fuel Costs Per Locomotive
$21,840
$10,920
$43,659
$10,920
$87,318
$83,160
$83,160
$98,000
$98,000
$196,000
$196,000
$196,000
$196,000
Remanufacture Costs
Cost per year
$0
$400
$846
$400
$846
$1,000
$1,000
$240
$240
$240
$240
$240
$240
Service life
15
15
21
15
21
40
40
40
40
40
40
40
40
Remanufacture Cost Per Locomotive
$0
$6,000
$17,766
$6,000
$17,766
$40,000
$40,000
$9,600
$9,600
$9,600
$9,600
$9,600
$9,600
TOTAL COST PER LOCOMOTIVE
$23,547
$31,073
$82,983
$101,754
$148,186
$199,037
$199,037
$174,625
$174,625
$251,777
$251,777
$244,364
$244,364
1.	Represents the fixed cost per supplier multiplied by the number of suppliers for each model type (e.g., 3 suppliers for Tier 0 Models A, B and C, and 1 supplier for the remaining model types).
2.	Total fixed costs for all suppliers divided by the number of locomotives in each model category.


3.	Sum of total hardware (variable) cost per locomotive and total fixed cost per locomotive plus 20% manufacturer markup.
4.	Represents average fuel consumption multiplied by the fuel economy penalty.







218

-------
4.1 Initial Costs
4.1a Fixed Costs
Fixed costs of manufacturing locomotive models compliant with the emissions standards included costs of
testing, engineering, tooling, and technical support.
•	The testing costs included developmental testing, as well as certification testing,
production line testing and in-use testing. Testing costs also included the costs of any
necessary additional facilities and equipment for emissions testing, plus engineering,
operating and maintenance costs for the testing facility. These costs, when allocated over
the estimated testing requirement, were estimated to amount to about $21,000 per test
prior to 2010 and about $39,000 per test after 2010 when the developmental testing
would be completed (U.S. EPA, 1998).
•	The engineering costs category represented the estimated average cost for the number
of engineering work years EPA projected to be required to develop the calibrations and
hardware necessary for meeting the emission standards. This also included the effort for
any ancillary changes made to the locomotives to accommodate the required new
hardware.
•	The tooling costs included costs for any additional or modified tooling necessary to
produce the emission control hardware, as well as for any required setup changes.
Because EPA estimated that Tier 0 compliance would be achieved through calibration
changes or hardware obtained from suppliers (particularly in the case of aftermarket
remanufacturers), EPA did not estimate specific tooling costs for Tier 0.
•	The technical support costs included the costs of any changes that would be required in
the technical support that manufacturers provide to users, including any necessary
operator or maintenance training and changes to technical publications that provide
operating and maintenance guidance.
EPA estimated these fixed costs for each locomotive supplier and divided by the total number of
locomotives (assuming suppliers would recover costs from the locomotives) to derive per locomotive
costs. EPA assumed that there were three suppliers each for Tier 0 Model A, B, and C locomotives, and
one supplier each for Tier 0 Model D and E, Tier 1 Model A, B, C, and D, and Tier 2 Model A and B
locomotives. EPA based this assumption on the numbers of independent part suppliers and
remanufacturers for the various locomotive models at the time of the analysis (U.S. EPA, 1998). The
number of suppliers EPA estimated for each model category was less than the total number of suppliers
because EPA assumed that the manufacturers for which initial costs were cost prohibitive would pay
other manufacturers with the ability to incur initial costs to perform the necessary services.
Because the fixed costs were for goods and services that are useful for more than one year of production,
EPA amortized initial costs over 5 years (i.e., manufacturers would recover costs within the first five years
of production). For Tier 2, because the standards were to be in effect for longer than 5 years, EPA
developed two sets of unit costs (because initial fixed costs would be recovered by 2010). EPA did not
calculate separate compliance costs reflecting fully-recovered fixed costs for Tier 0 and Tier 1 as it did for
Tier 2, because the initial hardware costs occur only at original manufacture (for Tier 1) or the first
remanufacture (for Tier 0), and thus are applicable only during the first few years of the program.
219

-------
Table 3 summarizes the fixed costs of manufacturing for each Tier and model type that were estimated by
EPA.
Q6a: Were EPA's assumptions regarding number of suppliers and distribution of fixed costs
reasonable?
A6a: »
Q6b: Based on your professional knowledge and experience, were the fixed costs per locomotive
for the various control options and Tiers in Table 3 over- or under-estimated? If so, please explain
why.
A6b: »
4.1b Variable Costs
Initial incremental variable compliance costs included costs of hardware and assembly.
• The hardware costs represented the emission reduction technologies EPA projected that
manufacturers would employ for compliance with the standards. EPA developed hardware
cost estimates for the following technologies:
Retarding fuel injection (2 degree or 4 degree timing retard)
4 pass after cooler
Improved mechanical and electrical injectors
¦	Electronic fuel injection
¦	Engine Modification
Improved turbocharger
Split cooling
High pressure injection
Combustion chamber design
Table 3 specifies combinations of these technologies that were expected to be used for each
locomotive model type and Tier.
• Assembly costs included the labor and overhead costs for retrofitting (in the case of Tier 0) or
for initial installation of the new or improved hardware. These also varied with the
characteristics of individual locomotives and the type of hardware necessary for compliance
with the applicable emission standards.
Q7a: Based on your professional knowledge and experience, were the per locomotive hardware
costs for each technology in Table 3 over- or under-estimated? If so, please explain why.
A7a: »
220

-------
Q7b: Based on your professional knowledge and experience, were the per locomotive assembly
costs for each model and Tier in Table 3 over- or under-estimated? If so, please explain why.
A7b: »
4.2 Remanufacture Costs Incurred by the Train Operators
The costs associated with keeping locomotives in compliance with the standards through subsequent
remanufactures included:
•	Costs of replacing electronic fuel injectors every two years;
•	Costs of electronic injection wiring harnesses, which need to be replaced in Tier 0 and Tier 1
locomotives every seven years due to embrittlement of the insulation from the heat generated
by the engine;
•	Cost of improved injector replacement for Tier 2 locomotives every two to three years.
Q8a: Based on your professional knowledge and experience, were the annual per locomotive
remanufacture costs for each model type and Tier in Table 3 over- or under-estimated? If so,
please explain why.
A8a: »
Q8b: Were EPA's assumptions about replacement frequencies reasonable? If not, please explain
why.
A8b: »
4.3 Fuel Costs Incurred by the Train Operators
EPA estimated increases in fuel consumption due various emission control technologies and the
corresponding incremental fuel costs. EPA assumed fuel penalties of:
•	2% for Tier 2 locomotives,
•	1% for Tier 1 locomotives, and
•	1 %-2% for Tier 0 locomotives.
Based on past developments in the industry, EPA believed that manufacturers would make every effort to
eliminate any initial fuel consumption penalties, and would have largely succeeded by 2010. However,
EPA included fuel economy penalties for the full 41 years covered by the analysis.
221

-------
Q9a: Were EPA's assumptions regarding fuel penalties reasonable, including the average fuel
consumption rate and fuel costs (in $ per gallon)?
A9a: »
Q9b: Based on your professional knowledge and experience, what can you say about elimination
of initial fuel consumption penalties by 2010? If this occurred, did learning by doing play a role?
A9b: »
The last line of Table 3 presents the total per locomotive cost estimated by EPA for each model type and
Tier.
Q10a: Did actual total per locomotive compliance costs differ significantly from EPA's estimates
over the time period in which this rule was applicable (1998 to 2008)? If so, what are the principal
reasons for these changes? To the extent possible, please indicate the approximate amount of
difference from EPA's estimates.
A10a:»
Q10b: Did technological innovation occur within the emission control technologies? If so, please
indicate which technology or technologies were affected and what the compliance cost
implications were.
A10b:»
Section 5 Emission Reductions
EPA first calculated baseline national emissions for each type of locomotive service (line-haul and switch)
by multiplying fuel consumption rates (gal/yr) by a conversion factor of 20.8 bhp-hr/gal to obtain total fleet
bhp-hr/yr values. EPA then multiplied these fleet bhp-hr/yr numbers by the applicable fleet average
emission rates to calculate emissions inventories (tons/yr). EPA estimated the fleet average emission
rates for each year based on the number of each type of locomotive it projected to be in the fleet at the
end of the respective year. EPA estimated the total reductions expected for each future year by
subtracting the expected controlled inventory from the estimated 1999 baseline inventory.
EPA calculated fleet average emission rates as weighted averages of uncontrolled, Tier 0, Tier 1, and
Tier 2 emission rates based on estimated relative class- and service type-specific fuel consumption rates
(e.g., the percent of total fuel consumed by Tier 1 line-haul locomotives in Class I for a given year).
222

-------
Assumptions Used for Class I Analysis:
The relative fuel consumption rates used to create average emission rates for Class I
line-haul locomotives were proportional to the product of the number of locomotives
(Nioc), average horsepower (HPavg), and a relative use rate factor (Fru) based on average
locomotive age, as shown below:
N HP F
Relative Fuel Consumption = ¦ loc RU

RU
•	EPA assumed 7.5% of fuel consumption by Class I railroads is for switching.
•	Calculations of the relative fuel consumption rates used to create average emission rates
for Class I switch locomotives did not account for differences in average horsepower and
relative use rates due to a lack of specific information. (Emission rates were weighted by
numbers of locomotives only.) EPA believed that this simplification did not significantly
affect the overall analysis because the differences in locomotive horsepower and usage
rates for this class, as a function of the tier of applicable standards, were less significant
than for Class I freight locomotives.
•	EPA assumed that fuel consumption remained constant at the 1996 level of 3.601 billion
gallons per year. EPA recognized that there was a short-term trend of increasing fuel
consumption, but was not confident that the trend would continue. The long-term trend
was for fuel consumption to remain fairly constant as a result of continual improvements
in locomotive fuel economy, which offset the significant increase in ton-miles of freight
hauled.
Table 4 shows the estimated emission rates of various pollutants for Class I locomotives.
Table 4: Estimated Emission Rates
g/bhp-hr) for Class 1 Locomotives
Pollutant
Tier
Line-Haul Locomotive
Switch Locomotive
Hydrocarbons
Uncontrolled
0.48
1.01
Tier 0
0.48
1.01
Tier 1
0.47
1.01
Tier 2
0.26
0.51
Carbon Monoxide
Uncontrolled
1.28
1.83
Tier 0
1.28
1.83
Tier 1
1.28
1.83
Tier 2
1.28
1.83
Nitrous Oxides
Uncontrolled
13.0
17.4
Tier 0
8.6
12.6
Tier 1
6.7
9.9
Tier 2
5.0
7.3
Particulate Matter
Uncontrolled
0.32
0.44
Tier 0
0.32
0.44
Tier 1
0.32
0.43
Tier 2
0.16
0.19
Source: U.S. EPA (1998)
223

-------
Q11a: Was EPA's method of determining relative fuel consumption for Class I locomotives by
service type (line-haul and switch) for each Tier reasonable? If not, please explain why.
A11a:»
Q11b: Was EPA's assumption about constant fuel consumption reasonable? Was the amount of
fuel consumed by Class I locomotives per year over- or under-estimated on average for the time
period 1998-2008? If so, please explain why.
A11b: »
Q11c: Was EPA's assumption about the share of fuel consumed by Class I switch locomotives
reasonable? If not, please explain why.
A11c: »
Q11d: Were the estimates of emission rates for each pollutant and locomotive type and Tier
reasonable given your knowledge and professional experience?
A11d:»
Assumptions used for Class ll/lll Analysis
•	For Class ll/lll locomotives, EPA did not account for differences in average horsepower
and relative use rates in calculating relative fuel consumption rates due to lack of specific
information for these classes (emission rates were weighted by numbers of locomotives
only).
•	EPA used information from the American Short Line Railroad Association (which
represents most of the Class II and Class III railroads) to estimate that the 4,200
locomotives in service with the Class II and III railroads in service in 1994 consumed
about 215 million gallons of diesel.
•	Due to a lack of specific information, EPA assumed that average Class II and III emission
rates were the same as the average emission rates for Class I line-haul locomotives.
EPA acknowledged that actual emission rates could be somewhat higher since smaller
railroads typically have lower power duty-cycles (i.e., more time at idle and low power
notches, and less at notch 8), especially those railroads performing primarily switch and
terminal services.
224

-------
Q12a: Was EPA's method of determining relative fuel consumption for Class ll/lll locomotives for
each Tier reasonable? If not, please explain why.
A12a:»
Q12b: Was the amount of fuel consumed by Class ll/lll locomotives per year over- or under-
estimated on average for the time period 1998-2008? If so, please explain why.
A12b: »
Q12c: Was EPA's assumption that the emission rates of each pollutant (by Tier) for Class ll/lll
locomotives was same as emission rates for Class I line-haul locomotives reasonable given your
knowledge and professional experience? If not, please explain why.
A12c: »
Assumptions used for Passenger Locomotives Analysis
•	For passenger locomotives, EPA did not account for differences in average horsepower and
relative use rates in calculating relative fuel consumption rates due to lack of specific information
for these classes. (Emission rates were weighted by numbers of locomotives only.)
•	EPA estimated that 463 passenger locomotives consumed about 61 million gallons of diesel fuel
per year.
•	EPA estimated that the 315 diesel Amtrak locomotives in service consumed about 72 million
gallons of diesel fuel per year.
•	EPA assumed that average passenger locomotive emission rates were the same as the average
emission rates for Class I line-haul locomotives.
Q13a: Was EPA's method of determining relative fuel consumption for passenger locomotives for
each Tier reasonable? If not, please explain why.
A13a:»
Q13b: Was the amount of fuel consumed by passenger locomotives per year over- or under-
estimated on average for the time period 1998-2008? If so, please explain why.
A13b:»
225

-------
Q13c: Was EPA's assumption that the emission rates of each pollutant (by Tier) for passenger
locomotives was same as emission rates for Class I line-haul locomotives reasonable given your
knowledge and professional experience? If not, please explain why.
A13c: »
Section 6 Additional Questions
Q14a: Since the time of rule development and promulgation, have technological innovations
occurred within the compliance technology options considered by EPA? If so, what innovations
occurred and approximately what impact did these innovations have on the cost of complying
with the rule?
A14a:»
Q14b: Did any learning by doing in development and use of the new technologies occur since the
time of rule development and promulgation? If so, what impact did these innovations have on the
cost of complying with the rule?
A14b: »
Q14c: Were there factors that may have caused greater implementation difficulty and higher costs
with the Rule? For example, were there:
•	Any technical challenges in designing process changes to meet compliance
requirements?
•	Issues with financing support for technology installation?
•	Technical performance issues in operating and maintaining the equipment?
•	Limitations on compliance in terms of compliance assistance or compliance
schedule?
•	Terms of regulatory requirements, and specific aspects of the rule requirements?
A14c: »
References
United States Environmental Protection Agency (U.S. EPA). 1998. Locomotive Emission
Standards: Regulatory Support Document. April.
226

-------
Chapter 6: Lessons Learned and Next
Steps
The four case studies presented in this report represent a first step in generating a larger body of evidence on
key drivers of compliance costs. While individual case studies of particular regulations are informative,
perhaps the more significant contribution of this effort is the application of a common conceptual framework
to the ex post assessments. Applying this framework to our case studies underscores the difficulties and
impediments to conducting consistent, comprehensive retrospective analyses of regulatory costs.
6.1.	Lessons Learned
Our retrospective analyses proved more challenging than originally anticipated and were often limited by the
paucity of evidence on how facilities chose to comply with the selected regulations and their associated
costs. In short, each of the case studies suffer from a lack of comprehensive cost information on treatment
technologies and mitigation strategies at the facility level, limiting our ability to make definitive statements
on the reasons for differences between ex ante and ex post cost estimates. Instead, the case studies either
rely on accessible industry level data (as opposed to facility level data), bottom-up cost estimates for a typical
"model" facility, or information from a limited number of industry experts. Each of these approaches, while
useful, also met with its own problems. For some case studies, the arsenic rule in particular, the regulated
sources were quite heterogeneous, varying by size, attributes, compliance technology, and vintage, giving
rise to complicated decision strategies for identifying appropriate technologies.
Disentangling the expenditures made expressly for pollution control was a challenge for several of the case
studies. Compliance expenditure data sometimes include expenditures - referred to as "might as well do
this" costs - those that occur at the same time as the compliance costs but that are, in truth, unrelated to
the regulation (i.e., upgrades, maintenance, etc.). For others, namely the pulp and paper rules, the methyl
bromide critical use exemption analyses, and the locomotive rule, defining the counterfactual was difficult
(i.e., what would have occurred had the rule not been promulgated).
In cases where EPA relied on outside experts, it also sometimes proved challenging to find qualified industry
experts. They were sometimes few in number to begin with, making it particularly difficult to identify
individuals who had not offered expertise during the development of the rule. In reaching out to trade
associations for assistance, we found that some were helpful but others were reluctant to become involved.
6.2,	Next Steps
While informative, the added evidence provided by the four case studies in this report is insufficient to draw -
broad conclusions about EPA cost estimation practices. As already noted, the rules were not selected to be -
227

-------
representative but rather to shed light on the process of conducting ex post analyses and the challenges
analysts engaged in these activities may face.
As a next step in this process, EPA has selected additional rules for retrospective analysis from the list of
eligible rules described in Chapter 1. Unlike the rules discussed in the case studies presented here, these
rules were selected using a stratified random selection process and include:
•	Control of Emissions of Air Pollution From Nonroad Diesel Engines (1997);
•	NSPS for Nitrogen Oxide Emissions
•	NESHAP: Surface Coating of Automobiles and Light-Duty Trucks (2004);
•	Effluent Limitations Guidelines, Pretreatment Standards, and New Source Performance Standards for
the Commercial Hazardous Waste Combustor Subcategory of the Waste Combustors Point Source
Category (2000);
•	Effluent Limitations Guidelines, Pretreatment Standards, and New Source Performance Standards for
the Transportation Equipment Cleaning Point Source Category (2000).
•	NSPS: Municipal Waste Combustion-Phase II and Phase III (Large Units) (1995)
As we pursue additional case studies, we will continue to explore the feasibility of other data collection
strategies including site visits, focus groups, and industry surveys to augment the publically available
information we are able to identify. Eventually, we hope to amass enough information to draw generalizable
conclusions about factors that cause ex ante and ex post cost estimates to differ, as discussed in Chapter 1,
with the ultimate goal of informing improvements to our cost estimation methodologies.
228

-------