Description
This study investigates the relationship between environmental performance and productive efficiency in the United
States electric utility industry before and after the 1990 Clean Air Act Amendments. Using Data Envelopment Analysis
(DEA), cross-sectional examinations reveal lower polluting plants are more efficient than higher polluting plants. Longitudinal
analyses indicate plants can simultaneously reduce pollution and increase relative efficiency
Ecoe?ciency: De?ning a role for environmental
cost management
Royce D. Burnett
a,
*
, Don R. Hansen
b,1
a
The University of Miami, Department of Accounting, 316 Kosar/Epstein Faculty O?ce Wing,
5250 University Drive, Coral Gables, FL 33146, United States
b
Oklahoma State University, William S. Spears School of Business Administration, School of Accounting,
Room 401, Stillwater, OK 74075, United States
Abstract
This study investigates the relationship between environmental performance and productive e?ciency in the United
States electric utility industry before and after the 1990 Clean Air Act Amendments. Using Data Envelopment Analysis
(DEA), cross-sectional examinations reveal lower polluting plants are more e?cient than higher polluting plants. Lon-
gitudinal analyses indicate plants can simultaneously reduce pollution and increase relative e?ciency. Collectively, these
results are evidence that proactive environmental management can reduce environmental costs and thus, lends support
for adopting an environmental cost management system.
Ó 2007 Elsevier Ltd. All rights reserved.
Introduction
Concern for taking care of the environment is
increasing as is interest in accounting for the envi-
ronment (Beets & Souther, 1999; Deegan, 2002;
Gray, Kouhy, & Lavers, 1995; Mathews, 1997).
The response from the business community has
been to gather and report more information about
environmental activities to stakeholders. For
instance, KMPG (2005) indicates that 55 percent
of the Fortune top global ?rms are now issuing
sustainability reports and that the number of top
100 ?rms issuing such reports within each country
surveyed has tripled since 1993. Although the con-
tent and quality of the reports vary widely, the
KPMG survey supports the view that environmen-
tal accounting and reporting is indeed on the rise
and is becoming an integral part of the information
released to shareholders. Moreover, the survey sug-
gests that all stakeholders – internal and external –
bene?t fromthese reports because they focus on the
risk and opportunities associated with corporate
0361-3682/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aos.2007.06.002
*
Corresponding author. Tel.: +1 305 284 8618; fax: +1 305
284 5737.
E-mail addresses: [email protected] (R.D. Burnett),
[email protected] (D.R. Hansen).
1
Tel.: +1 405 744 8626; fax: +1 405 744 1680.
www.elsevier.com/locate/aos
Available online at www.sciencedirect.com
Accounting, Organizations and Society 33 (2008) 551–581
social responsibility. White (2005) indicates that
producing these reports is valuable because they
provide a framework for linking economic, envi-
ronmental and social decision making to strategy.
To date, research studies in environmental
accounting have explored issues relating to envi-
ronmental disclosure such as the characteristics of
disclosers (e.g., Gray et al., 1995; Hackston &
Milne, 1996; Patten, 1992), what is being disclosed
(e.g., De Villiers & Van Staden, 2006; Wiseman,
1982), what should be disclosed (e.g., Dierkes &
Preston, 1977; Deegan, 2002; Moneva, Archel, &
Correa, 2006), and why voluntary disclosures are
made (e.g., Beets & Souther, 1999; Neu, Warsame,
& Pedwell., 1998; Walden & Schwartz, 1997).
Another line of research in environmental account-
ing concerns the relationships among environmen-
tal disclosure, environmental performance, and
economic performance. Studies have examined all
three pair-wise relationships (Berthelot, Cormier,
& Magnan, 2003; Margolis & Walsh, 2003). Much
of the research related to environmental account-
ing in this area has targeted the environmental dis-
closure–environmental performance pair (Fekrat,
Inclan, & Petroni, 1996; Freedman & Wasley,
1990; Hughes, Anderson, & Golden, 2001; Ingram
& Frazier, 1980; Rockness, 1985; Wiseman, 1982).
Patten (2002) notes that evidence concerning this
particular relationship is of special interest because
of socio-political theories concerning environmen-
tal disclosure. Patten (1991) suggests that while
these theories implicitly assume businesses are
responsible for using resources to engage in activi-
ties designed to increase pro?ts, they also call for
an awareness of the adverse consequences that
the pursuit of capitalism has on the environment.
In a similar vein, there is increasing interest con-
cerning the relationship between environmental per-
formance and economic performance because of an
emerging socio-economic theory known as ecoe?-
ciency (Birkin & Woodward, 1997a, 1997b). Ecoef-
?ciency claims that it is possible to increase
productivity and thus reduce costs while simulta-
neously improving environmental performance
(Bebbington, 2001; Lehman, 2002; Stone, 1995).
The ecoe?ciency paradigm has signi?cant implica-
tions for environmental accounting and, in particu-
lar, for environmental management accounting
(EMA). Speci?cally, if ecoe?ciency is valid, then
the likelihood will increase that companies will be
willing to establish formal EMA control systems
that isolate and quantify the costs, bene?ts, and
operational outcomes of proactive environmental
management. In fact, according to the International
Federation of Accountants’ Statement Management
Accounting Concepts, EMA is ‘‘the management of
environmental and economic performance through
the development of appropriate environment-
related accounting systems and practices.’’ (Savage
&Jasch, 2005, p. 19). Indeed, Walden and Schwartz
(1997) indicate that creating and maintaining envi-
ronmental measurement and reporting systems is
an appropriate response to the social, political and
economic pressures surrounding the use of natural
resources. Ilinitch, Soderstrom, and Thomas
(1998) agree and indicate that environmental man-
agement e?ectiveness depends on an output-based
approach that, in part, relies on an internal assess-
ment of how well an organization uses its resources
to gain competitive advantage. Thus, for an EMA
system, the environmental-economic performance
pair is the most important and critical relationship.
The purpose of this paper is to provide empirical
evidence concerning ecoe?ciency in the electric util-
ity industry by examining the relationship between
environmental performance and economic perfor-
mance, where environmental performance is mea-
sured (proxied) by levels of SO
2
and economic
performance is measured by productive e?ciency.
Admittedly, SO
2
is not a comprehensive measure
of environmental performance but it is a very
important pollutant for the industry and therefore
is likely to be a useful proxy for environmental per-
formance.
2
Thus, knowing the nature of this rela-
2
Ilinitch et al. (1998) indicate that de?ning corporate
environmental performance is not a straightforward task and
as such, any one performance indicator, or even an entire
category of indicators, would likely be insu?cient to gauge a
company’s environmental performance with con?dence. Lanen
(1999), in a case study of the assessment of environmental
performance and waste management, suggests that one way to
gauge the relevance of any performance measure is to assess
whether actions adopted result in improvements in the measure
of interest. Given this insight, as well as the intent of the 1990
Clean Air Act Amendments, SO
2
emissions appear to be an
appropriate proxy for environmental performance in this study.
552 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
tionship ought to be of signi?cant interest to man-
agement, accountants, shareholders, regulators,
and other stakeholders. For example, if the same
or more good (intended) output is produced using
fewer inputs while simultaneously reducing pollu-
tion, then costs decrease and overall environmental
and ?nancial performance improves. Such con?r-
matory evidence would support ecoe?ciency and
provide a positive incentive for management to
invest in an environmental cost management system
as part of an agenda to reduce pollution.
Con?rmatory evidence would also a?ect disclo-
sure practices and investor decisions. If managers
believe that good environmental performance is
positively associated with good economic perfor-
mance, then this would tend to be viewed as good
news by investors and managers would have an
incentive to disclose more information about envi-
ronmental performance (Al-Tuwaijri, Christensen,
& Hughes, 2004). Moreover, the signal would be
that investing in ‘‘green’’ companies has a sound
economic rationale. Indeed, Feldman, Soyka,
and Ameer (1997) suggest that improving a com-
pany’s environmental management may result in
improved environmental performance, and that
improved environmental performance is compati-
ble with superior ?nancial performance. Further,
they posit that environmental improvements may
be viewed by the ?nancial community as a reduc-
tion in the overall risk of the ?rm resulting in a
lower cost of capital and an increase in stock price.
Finally, evidence supporting ecoe?ciency would
imply that the principal role of regulatory
approaches is to mandate reductions in pollution,
leaving management free to innovate and ?nd
ways to simultaneously improve e?ciency, reduce
costs, and achieve the mandated reductions. Of
course, if the empirical evidence does not support
ecoe?ciency, this also has implications, but with
opposing directional e?ects.
The outcomes produced in reaction to the Uni-
ted States Federal 1990 Clean Air Act Amend-
ments (1990 CAAA) provide the opportunity to
examine these claims. The 1990 CAAA mandated
a speci?c subset of plants in the electric utility
industry participate in a ‘‘group’’ reduction of pol-
lution within a ?ve-year period without specifying
how the reductions were to be achieved. Fortui-
tously, plant-level data are publicly available and
so this setting a?ords a unique and direct opportu-
nity to evaluate ecoe?ciency by examining the
changes in plant-level productive e?ciency from
1990 to 1995 under a scenario that is devoid of spe-
ci?c operational mandates. At issue is not the time
period of the regulatory act and the associated
plant data but the conditions speci?ed by the Act
and whether management can act under these con-
ditions to pursue process innovations and
improvements to achieve pollution reduction while
maintaining or increasing e?ciency.
The 1990 CAAA identi?es two groups of
plants: (1) Phase-One plants, de?ned as higher pol-
luting plants subject to an overall requirement to
reduce pollution by 1995 (the beginning of
Phase-One); and, (2) Non-Phase-One plants,
de?ned as lower polluting plants exempt from
any pollution reduction requirements for the
Phase-One period. The 1990 and 1995 Non-
Phase-One plants serve as control groups for
cross-sectional analyses while the 1990 Phase-
One plants de?ne a natural longitudinal control
group to assess the e?ect on plant-level e?ciency
as pollution is reduced because of the 1990 CAAA
intervention.
The approaches taken and the results obtained
are as follows. Initially, raw or unadjusted Data
Envelopment Analysis (DEA) scores are used to
measure relative e?ciency on both a cross-sec-
tional and longitudinal basis. For cross-sectional
analysis, we compare the e?ciency of the Non-
Phase-One (lower polluting) plants with that of
the Phase-One (higher polluting) plants in 1990
and 1995 to assess whether the lower polluting
plants are relatively more e?cient than the higher
polluting plants. Using unadjusted DEA scores,
we ?nd that the group of lower polluting plants
is relatively less e?cient than the higher polluting
plants in 1990, an outcome inconsistent with ecoef-
?cient behavior. The 1995 results, however, favor
the Non-Phase-One plants. We then evaluate the
e?ect of intervention (the 1990 CAAA) by investi-
gating changes in the relative e?ciency of both
Phase-One and Non-Phase-One plants between
1990 and 1995. We ?nd that by 1995 the Phase-
One plants have simultaneously decreased pollu-
tion and increased their relative e?ciency. During
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 553
the same time period, we also ?nd that the Non-
Phase-One plants did not reduce pollution but
did show an increase in relative e?ciency.
Next, the DEA scores are adjusted for the
e?ects of potential confounding variables. Then,
the cross-sectional and longitudinal analyses are
repeated using the adjusted e?ciency scores.
Results obtained indicate that the lower polluting
plants are relatively more e?cient than the higher
polluting plants for both 1990 and 1995. Further-
more, the longitudinal results show that the 1995
plants are more e?cient than the 1990 plants.
Finally, we provide evidence that the increases
in e?ciency are associated with such compliance
actions as fuel switching, new fuel mix technology
and the use of scrubbers. This evidence is impor-
tant because it provides insight as to which opera-
tional options are e?ective when developing a
pollution reduction compliance strategy. Thus,
from an overall perspective, our results favor eco-
e?ciency and thus support a positive relationship
between environmental performance and eco-
nomic performance. Accordingly, a strong argu-
ment can be made for the role for an EMA
system as envisioned by the International Federa-
tion of Accountants and others.
We organize the remainder of the paper as fol-
lows. Section ‘‘Relevant literature, motivation and
empirical predictions’’ discusses the relevant litera-
ture and develops the empirical predictions for the
study. Section ‘‘Research design’’ explains the
research design and Section ‘‘Empirical results’’
provides the empirical results. The ?nal section
o?ers some concluding remarks.
Relevant literature, motivation and empirical
predictions
Environmental performance, economic performance,
and ecoe?ciency
The traditional view concerning the relationship
of environmental performance and economic per-
formance argues that improving environmental
performance inevitably raises costs and reduces
productivity (Porter, 1991). According to the tra-
ditional view of environmental economics, envi-
ronmental degradation is a negative output and
decreasing negative output requires additional
inputs (increasing environmental costs).Thus, any
improvement cannot increase economic e?ciency
which implies a win–lose paradigm. Accordingly,
the traditional hypothesis assumes that improving
environmental performance is costly and requires
regulatory intervention to bring about any envi-
ronmental improvements. However, several
researchers have taken exception to this assumed
relationship between environmental improvements
and cost, arguing that pollution is a form of eco-
nomic ine?ciency and, thus, reductions in pollu-
tion actually increase productive e?ciency and
thereby reduce costs (King & Lenox, 2002; Porter,
1991; Porter & van der Linde, 1995a, 1995b). This
competing hypothesis challenges the traditional
model of pollution control and cost management,
re?ects an underlying win–win paradigm, and is
referred to as ecoe?ciency. Managers exhibit eco-
e?cient behavior when they adopt innovative and
creative methods to produce more or the same
level of useful goods and services while simulta-
neously reducing environmental degradation,
resource consumption, and costs (WBCSD,
2000a, 2000b). Thus, the voluntary ecoe?ciency
hypothesis maintains that fully rational managers
will voluntarily engage in an unfettered or pure
form of ecoe?cient behavior. The two competing
hypotheses have the same premise (improving
environmental performance) but have totally
opposite consequences.
While agreeing with the ecoe?ciency paradigm,
Porter (1991) and Porter and van der Linde
(1995a, 1995b) are skeptical that managers will
voluntarily engage in ecoe?cient behavior. They
hypothesize that agents within ?rms generally have
bounded rationality relative to ecoe?ciency and
consequently must be encouraged or stimulated
to engage in ecoe?cient behavior by properly
designed regulatory intervention. It seems curious
that in competitive markets, managers need to be
encouraged to adopt ecoe?ciency (if valid); yet
one possible explanation is that managers are ?x-
ating on the traditional paradigm. Indeed, para-
digm shifts often need some stimulus. Market
forces are one possible stimulus and can often be
very punishing, as evidenced by the experiences
554 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
?rms and industries had in shifting from the tradi-
tional view of quality to total quality management
(Shank & Govindarajan, 1994). In transitioning
from the traditional to the new ecoe?cient para-
digm, it would not be unusual to see some ?rms
experimenting with ecoe?cient investments, while
most do not. Over time, if the ecoe?cient experi-
ments prove fruitful, surviving ?rms will all be
subscribers to the ecoe?cient paradigm.
Properly designed regulatory intervention is
another possible stimulus, and, if timely, may avoid
some of the punishing e?ects of the competitive
market by reducing the uncertainty of investing in
technology and producing an equal playing ?eld
for change (Porter & van der Linde, 1995a). Prop-
erly designed regulations have two components.
First, they focus on environmental outcomes and
not on the means to achieve those outcomes. Sec-
ond, they provide a signal to companies to improve
and trigger innovations that will reduce pollution
while simultaneously reducing environmental
costs. Thus, the Porter Hypothesis (guided ecoe?-
ciency hypothesis) argues that bounded rationality
requires carefully crafted regulatory intervention
to induce ecoe?cient behavior.
Ecoe?ciency – in either of its forms (voluntary
or guided) – has signi?cant implications for an
EMA system and also for environmental account-
ing and reporting in general. There are those who
believe that social and environmental accounting
should have the objective of replacing conven-
tional accounting with new accountings that are
predicated on values broader than the traditional
pro?t-maximization objective (Dierkes & Preston,
1977; Patten, 1991; Rubenstein, 1992; Ullman,
1976). Others view pro?t maximization and elimi-
nating environmental degradation as incompatible
and also believe that conventional accounting
encourages, supports, and even fosters social injus-
tices and environmental degradation (Gray &
Bebbington, 2000). Thus, calls are issued for radi-
cal revisions of accounting, with the new account-
ing taking a lead role in fostering needed change so
that sustainability can be achieved (Bebbington,
2001; Bebbington, Gray, & Owen, 1999; Gray,
1992).
However, if ecoe?ciency is valid, then the
calls for radical revisions of accounting to accom-
modate environmental values are much less com-
pelling because the basic premise that pro?t
maximization and environmental improvement
are incompatible would be in error. As Wildavsky
(1994, p. 463) clearly states regarding this issue:
‘‘The changes sought in accounting are also pre-
mised on factual beliefs about the vast harms done
to the natural environment and life-forms of all
kinds by modern technology. If these beliefs are
unwarranted, the case for accounting change col-
lapses.’’ Wildavsky (1994, p. 462) also agrees, in
principle, with the ecoe?ciency claim that pollu-
tion is a form of economic ine?ciency by noting
that: ‘‘A cleaner environment is, in a signi?cant
way, a function of economic e?ciency.’’
Although the validity of ecoe?ciency would
challenge the need for a radical revision of
accounting, it would also carry with it a need for
signi?cant modi?cations of the way current
accounting and reporting is done. To support the
ecoe?ciency paradigm and the needed internal
decision making, an EMA system must identify
and report two types of information: (1) physical
information relating to uses and ?ows of materials,
water, energy, and wastes, and (2) monetary infor-
mation relating to environmental costs, earnings,
and savings (United Nations Division for Sustain-
able Development, 2001). However, several exist-
ing issues could complicate these e?orts. For
instance, conventional accounting systems typi-
cally underestimate environmental costs. More-
over, most companies lack adequate systems for
measuring and managing environmental costs
(Epstein, 1996; Joshi, Krishnan, & Lave, 2001).
As such, some contend that the economic rational-
ist view of environmental performance may be
lacking because the full costs of environmental
activities are not being captured (Herbohn,
2005). Therefore, the availability and use of a
wider set of accounting information relative to
the economic consequences of the social position
(e.g., environmental management) of a company
could provide a more comprehensive view of the
company’s cost structure (Jonson, Jonson, &
Svensson, 1978).
Given these issues, signi?cant modi?cations are
needed to develop an EMA; yet, for this to actually
happen, managers require an incentive to do so.
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 555
The traditional view of the relationship between
environmental performance and economic per-
formance o?ers, at best, no incentive for changing
the current system of accounting and reporting.
If, however, pollution is equivalent to economic
ine?ciency as the Porter Hypothesis claims,
then reducing pollution will increase economic
e?ciency and the value of the ?rm will increase.
Moreover, if investors and creditors were con-
vinced that environmental degradation is equiva-
lent to economic ine?ciency, their demand for
environmental information would increase the
likelihood that ?rms would develop a comprehen-
sive EMA and issue substantive sustainability
reports (Neu et al., 1998; Solomons, 1991). Conse-
quently, ecoe?ciency, if valid, provides a strong
incentive for managers to invest in an EMA system
and to signal their most in?uential and relevant
stakeholders through substantive sustainability
reporting.
What is needed, then, is convincing evidence of
the validity of ecoe?ciency. The early empirical
support for ecoe?ciency is grounded by a solid
body of case studies (DeSimone & Popo?, 1997;
Ditz, Ranganathan, & Banks, 1995; Epstein, 1996;
Schmidheiny & Zorraquin, 1996). Case studies play
an especially important role in accounting research
because they provide a way to investigate the
boundaries of a particular process and help estab-
lish the importance of patterns within an organiza-
tion (Hagg & Hedlund, 1979). Case studies are
useful initially to identify and explore the validity
of hypothesized relationships and thus, serve as
an important forerunner and input for other types
of empirical testing (Kaplan, 1986; Lillis & Mundy,
2005). Indeed, Porter and van der Linde (1995a)
recognize that given the current ability to capture
the true costs and bene?ts of regulation, a reliance
on case studies to examine environmental issues is
helpful. However, Porter and van der Linde
(1995a, p. 104) themselves admit,‘‘. . .a list of case
studies . . .no matter how long, is not a complete
substitute for [other] careful empirical testing’’
and thus, call for more cross-sectional and multi-
unit testing.
Such empirical examinations of the Porter
Hypothesis and voluntary ecoe?ciency are begin-
ning to appear in the accounting and environ-
mental management literature. Al-Tuwaijri et al.
(2004) examine a sample of 198 ?rms from the
Standard and Poor’s 500 and ?nd a statistically
signi?cant positive association between a toxic-
waste measure (toxic waste recycled/total toxic
waste generated) and market returns. They indi-
cate that these results are consistent with the Por-
ter Hypothesis, noting that innovative solutions to
reduce pollution can simultaneously promote envi-
ronmentalism and industrial competitiveness. In
reality, the results are more consistent with volun-
tary ecoe?ciency than guided ecoe?ciency as
there is no speci?c regulatory act mandating a
reduction of toxic waste for all ?rms in the sample.
There are two studies that explicitly consider
the e?ects of regulation. Murty and Kumar
(2003) study the e?ect of environmental regulation
on the productive e?ciency of water polluting
industries in India. They study 92 ?rms in the
sugar industry in India and ?nd that the degree
of technical e?ciency increases with the degree of
compliance to environmental regulation and water
conservation e?orts, a result consistent with the
Porter Hypothesis. They suggest that ?rm manag-
ers should use the message of the Porter Hypothe-
ses to focus on ways to reduce pollution so that
costs can be reduced. Freedman and Jaggi (1992)
examine the association among three measures of
pollution performance and economic performance
for ?rms in the US. pulp and paper industry. These
?rms were subject to the 1972 Federal Water Pol-
lution Control Amendments and signi?cantly
reduced pollution in the 1978–1986 period. Pollu-
tion performance is measured at that plant level,
but economic performance is measured at the ?rm
or segment level. Although all the ?rms reduced
pollution, they were unable to reject the hypothesis
of no association between the pollution reduction
and economic performance. Nonetheless, reducing
pollution without increasing cost is still an interest-
ing outcome. If we de?ne a weak form of ecoe?-
ciency as pollution reduction with no extra cost,
then the absence of a positive or negative relation-
ship is supportive of this weaker form of
ecoe?ciency.
The early evidence is reasonably encouraging
but much more is needed to build a case that
would be su?ciently convincing to management
556 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
and in?uential stakeholders so that implementa-
tion of an EMA becomes a reality for most ?rms.
Absent the economic incentives of ecoe?ciency,
other approaches such as the calls for radical revi-
sions of accounting may begin to assume greater
legitimacy. Thus, the validity of the Porter
Hypothesis is an important empirical question
for environmental accounting. Its validity also car-
ries with it important implications regarding the
nature of regulatory actions needed to improve
both environmental and economic performance.
The Porter Hypothesis and the Federal 1990 Clean
Air Amendments
To empirically investigate the Porter Hypothe-
sis, regulations that emphasize desired environ-
mental improvements without specifying means
of achievement must be present. The 1990 CAAA
satisfy this requirement (EPA, 2000; Swift, 2001, p.
315; US Department of Energy, Energy Informa-
tion Administration, 1994 and 1997). Title IV of
the 1990 CAAA directed the Environmental Pro-
tection Agency (EPA) to establish a program that
would reduce acid rain (formed largely from emis-
sions of sulfur dioxide (SO
2
) and nitrogen oxides
(NO
x
)). This reduction was to be achieved primar-
ily by the electric utility industry because its fossil
fuel plants were the major sources of SO
2
and NO
x
emissions.
3
Title IV also set an eventual overall cap on util-
ity SO
2
emissions at 8.95 million tons per year, or
roughly half the 1980 baseline emissions (for the
industry as a whole). One of the most salient fea-
tures of Title IV was the establishment of an allow-
ance trading market. An allowance was de?ned as
the authorization granted to an electricity pro-
ducer to emit one ton of SO
2
during a calendar
year (US Department of Energy, Energy Informa-
tion Administration, 1994). The EPA initially allo-
cated a certain number of allowances to each
electric utility plant. However, plants that reduced
their actual emissions below the number of allow-
ances allocated directly to them would have excess
allowances that could be traded within their com-
pany, banked for future use, or sold to other util-
ities. Each utility plant had to either reduce
emissions to the number of allowances issued ini-
tially or acquire allowances from other utilities
and thus avoid having to reduce emissions. There-
fore, the total amount of allowances actually held
at any point de?ned the SO
2
output limits. As
such, compliance was determined by comparing
the number of allowances held by a plant to its
actual level of emissions.
Title IV was also unique in that it granted con-
siderable autonomy and ?exibility to plants rela-
tive to the methods they could use to reduce
emissions. Therefore, compliance strategies were
potentially a combination of active allowance
management and the use of proactive pollution
reduction practices. This approach to controlling
SO
2
was a radical departure from the traditional
‘‘command and control’’ approach of pollution
reduction and thus, for the ?rst time, electric util-
ities had a real opportunity to use process
improvements and innovations to reduce emis-
sions and improve e?ciency (EPA, 2000, p. 3).
The economic bene?ts from such a program struc-
ture are obvious as Swift (2001, p. 315) noted
‘‘. . .the cap-and-trade system makes the economic
sense of a ton of reduction very real and exerts
continuous economic pressure to improve perfor-
mance and transform compliance behavior.’’
Finally, Title IV was administered as a two-
phased plan: Phase One and Phase Two. Phase
One of the program ran from 1995 through 1999
and a?ected power plants known to produce large
amounts of SO
2
and NO
x
.
4
Phase Two started in
2000 and in addition to increasing the annual emis-
sions limits imposed on Phase One plants, also
expanded mandated reductions to Non-Phase-
One electric utility plants.
3
Initially, nitrogen oxide (NO
x
) emission reductions were
included as part of the Act. However, the rules governing these
reductions were vacated in 1994 and not replaced until the latter
part of 1995, thus making NO
x
reductions not required until
1996. Therefore, NO
x
reductions are not included in our study.
4
The 1990 CAAA initially required 110 large plants to reduce
their SO
2
to mandated levels by 1995. However, by the
beginning of 1995, 54 additional plants had been brought into
Phase-One. Thus, in 1995, 164 plants underwent the annual
reconciliation process administered by the EPA to determine
compliance with Phase One.
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 557
There has been some empirical investigation of
the 1990 CAAA and the outcomes it produced.
Hughes (2000) uses ?rms a?ected by Phase-One
of the 1990 CAAA to study the relationship
between their market value of equity and SO
2
emissions for the period 1986–1993. He ?nds that
these emissions have value for the years 1989,
1990, and 1991 (there is a signi?cant negative rela-
tionship between market value and SO
2
). This
relationship disappears subsequent to 1991. The
study o?ers several inviting queries. First, Hughes
(2000) suggests that the decline in value relevance
may be attributable to changes made within the
industry in response to the Act. However, the
study provides no direct evidence to support this
suggestion. Second, the study does not provide a
direct link between how changes in productive e?-
ciency are attributable to the regulation. Third,
Hughes (2000) indicates that the initial cost esti-
mates for compliance were signi?cantly less than
expected, and goes on to suggest that this may
be the result of a lack of understanding about
how the impact of the availability of choice rela-
tive to the adoption of pollution control strategies
would a?ect costs. This last point implies the need
for an EMA that provides information that is
needed to understand the business and societal
bene?ts associated with this type of public policy.
A plant-level study may facilitate a response to
these open issues. Indeed, Freedman and Jaggi
(1994) note that ?ndings at the plant level are use-
ful because they provide comprehensive informa-
tion about performance before and after
regulations are imposed. Swift (2001) concurs
and suggests the unique regulatory design of the
Act provides the chance to produce major emis-
sion reductions, lower compliance costs, and cre-
ate very low ongoing transaction costs at both
the plant and the ?rm level. Therefore, analysis
at the plant level o?ers an opportunity to directly
test the Porter Hypothesis by evaluating the
impacts on plant-level e?ciency when actions are
taken in response to pollution reduction mandates.
Given the above, it is appropriate to suggest
one way plants can produce economic value is
through the strategic use of environmental expen-
ditures. As such, there is an underlying rationale
for a place for environmental cost management
and an EMA system in proactive pollution con-
trol. There are at least two major reasons for this:
(1) environmental costs are signi?cant (Bhat, 1996;
Ditz et al., 1995; Joshi et al., 2001; Wong, 2001),
and (2) there is an increasing body of anecdotal
and empirical evidence suggesting that improving
environmental performance can actually reduce
environmental costs (Al-Tuwaijri et al., 2004;
Epstein, 1996; Schmidheiny & Zorraquin, 1996).
Taking the ?rst reason as a given, our general
research objective is to determine if ecoe?ciency
(i.e., the simultaneous reduction of environmental
costs and improvements in environmental quality,
as measured by reductions in SO
2
) has any empir-
ical content. Evidence validating ecoe?ciency can,
then, be used to establish a role for environmental
cost management.
However, a robust evaluation of ecoe?ciency
also requires that it be examined within the context
of hypotheses that incorporate both environmen-
tal and ?scal policy (e.g., environmental cost man-
agement). Three competing hypotheses regarding
environmental cost management have been identi-
?ed: (1) the traditional hypothesis, (2) the volun-
tary (pure) ecoe?ciency hypothesis, and (3) the
guided ecoe?ciency (Porter) hypothesis. Our spe-
ci?c research objective is to evaluate the validity
of the Porter Hypothesis using the 1990 CAAA.
However, since the three hypotheses are compet-
ing, the 1990 CAAA framework also a?ords an
opportunity to shed light on the other two
hypotheses.
Empirical predictions
If the Porter Hypothesis is valid, then four pre-
dictions can be identi?ed that relate to the empir-
ical framework in which we are operating. Two
predictions are concerned with cross-sectional
relationships and two are concerned with longitu-
dinal relationships. Our predictions are as follows.
Prior to intervention (the 1990 period), ecoe?cient
behavior may already exist to some extent in the
industry. Thus, if plants have reduced pollution
relative to an earlier period, then the lower pollut-
ing plants would be more e?cient than the higher
polluting plants. In fact, if it is accepted that pol-
lution is a form of economic ine?ciency as the
558 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
Porter Hypothesis claims, then it is logical to pre-
dict that lower polluting plants are more e?cient
than higher polluting plants. Therefore, we predict
that in 1990 the Non-Phase-One plants are rela-
tively more e?cient than the Phase-One plants (?rst
cross-sectional prediction).
From 1990 to 1995, we expect regulatory inter-
vention to induce ecoe?cient behavior for the
Phase-One plants. Anecdotal evidence appears to
support this assumption. For example, Phase-
One plants reduced their pollution by only about
8% between 1980 and 1990 (EPA, 1996), whereas
the same plants reduced pollution by about 50%
between 1990 and 1995 (Pelz & Fitzgerald, 2001).
Thus, since Phase-One plants are lower polluting
in 1995 relative to 1990 (by a large amount), we pre-
dict that the 1995 Phase-One plants are relatively
more e?cient than their 1990 counterparts (?rst lon-
gitudinal prediction).
The ?nal two predictions are based on the
underlying logic of the Porter Hypothesis coupled
with evidence concerning pollution reduction and
innovation after implementation of the Act. Spe-
ci?cally, the Porter Hypothesis assumes that lower
levels of pollution can be achieved without com-
promising e?ciency. Therefore, it is logical to con-
tinue to assume that lower polluting plants should
continue to be relatively more e?cient than higher
polluting plants, even after intervention. As such,
whichever group of plants is the lower polluting
group (after intervention) should be more e?cient.
Thus, in 1995 we predict that the Non-Phase-One
(Phase-One) plants are relatively more e?cient than
Phase-One (Non-Phase-One) plants provided they
have lower pollution (second cross-sectional
prediction).
Finally, a key claim of the Porter Hypothesis is
that properly crafted regulatory intervention will
lead to innovations that will reduce pollution and
increase economic e?ciency. Interestingly, innova-
tions appear to be at the root of the pollution
reductions achieved by the Phase-One plants.
Swift (2001) notes that the 1990 CAAA stimulated
the development of new fuel blending technologies
and also stimulated a ‘‘spillover e?ect’’ whereby
the innovations of one industry spur innovations
in another industry. In fact, he estimates that the
1990 CAAA promoted over $12 billion in innova-
tive and widespread technologies designed to
reduce pollution, improve fuel access and trans-
portation, and eliminate waste (Swift, 2001, p.
338). For example, in response to the events taking
place in the electric utility industry, coal producers
developed new mining technologies that increased
coal mining productivity by almost 7 percent
between 1990 and 1995 (US Department of
Energy, Energy Information Administration,
1997, p. 23).
The innovations created in response to the 1990
CAAA are likely to have widespread e?ects and
thus, their impact may extend beyond Phase-One
plants. Accordingly, Non-Phase-One plants may
capture the derived bene?ts from the intervention-
induced innovations such as new fuel blending
technologies and lower prices of coal. Further-
more, we would not expect the Non-Phase-One
plants to signi?cantly increase their pollution.
Therefore, we predict that 1995 Non-Phase-One
plants should be at least as e?cient as 1990 Non-
Phase-One plants because of the derived-bene?ts
from Phase-One, Porter-predicted innovations
(second longitudinal prediction).
Research design
Modeling and measuring relative e?ciency
E?ciency is measured in relative terms where
one ?rm is compared to another. Relative compar-
isons are especially important to environmental
management because they provide a baseline to
set improvement goals for external environmental
regulations, internal business practices and emerg-
ing technological implications (Ruch & Roper,
1992). Our paper uses Data Envelopment Analysis
(DEA), a fractional linear programming technique
initially developed by Farrell (1957) and subse-
quently re?ned by Charnes, Cooper, and Rhodes
(1978) to provide these e?ciency measures.
DEA estimates the relative e?ciency of a group
of decision-making units (DMU) with similar
goals and objectives (Callen, 1991). The usual
empirical approach is to assume that ?rms operat-
ing in the same industry have the same production
function. DEA converts multiple input and output
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 559
measures into a single measure of relative perfor-
mance and uses this summary measure to map
the best-practice envelope (frontier) of the input–
output data (Majumadar, 1998). DMUs operating
on the frontier are technically e?cient and rela-
tively more e?cient than those operating o? the
frontier. Fig. 1 illustrates these concepts for six
?rms represented by points A–F. Technically e?-
cient DMUs (A–D) lie on the e?cient frontier,
AD, and are awarded an e?ciency index of one
(e.g., OB/OB). Technically ine?cient DMUs (E
and F) lie above the frontier and have an index
of less than one (e.g., OF
*
/OF).
The e?ciency indices must be computed for
each DMU. Linear programming models are used
to accomplish this objective. Although there are
various DEA models, our relative e?ciency analy-
sis uses the variable-returns-to-scale model
(referred to as the BCC model) developed by
Banker, Charnes, and Cooper (1984). The BCC
model for n DMUs, m outputs, and p inputs is
represented as follows:
Min e ð1Þ
subject to:
X
n
i¼1
w
i
O
ij
PO
kj
; j ¼ 1; 2; . . . ; m ð2Þ
X
n
i¼1
w
i
I
ir
6 eI
kr
; r ¼ 1; 2; . . . ; p ð3Þ
X
n
i¼1
w
i
¼ 1 ð4Þ
where
e the e?ciency or ‘‘DEA’’ scores assigned
by DEA
O
ij
output j for DMU i
I
ir
input r for DMU i
O
kj
the output j for DMU k (target DMU)
I
kr
input r for DMU k, where k denotes the
DMU being evaluated (target DMU) for
its e?ciency relative to all others
w the weight assigned by DEA
The model selects a set of weights that minimize
e. This is equivalent to minimizing the input
resources required by the composite DMU. These
weights de?ne a hypothetical composite DMU
using the outputs and inputs of all DMUs within
the reference group. The value of e lies between
zero and one. If e < 1, then the composite DMU
requires less input resources to produce the same
or greater output as the target DMU. Thus, the
target DMU is judged as relatively ine?cient. If
e = 1, then the target and composite DMU are
one in the same, indicating that there is no other
DMU or group of DMUs that are relatively more
e?cient. In this case, the target DMU is on the e?-
cient frontier.
DEA model for the plant setting
To calculate DEA scores, outputs and inputs
must be selected and de?ned. Five variables are
used in our plant-level DEA model. These consist
of two outputs [Kilowatt-hours (H) and SO
2
(S)]
and three inputs [Capital (K), Fuel Costs (F), and
Operating Costs (C)].
5
Kilowatt-hours are a
positive or good output whereas SO
2
is a bad or
Units of Input 1
U
n
i
t
s
o
f
I
n
p
u
t
2
A
B
E*
E
F
F*
C
D O
Fig. 1.
5
Because of in?ation, nominal prices from di?erent years
cannot be compared without making adjustments. Therefore,
both fuel and operating costs are expressed in constant
dollars. As such, 1995 prices are de?ated to 1990 prices. This
approach is consistent with other studies that examine the
electric utility industry on a longitudinal basis (e.g., Freedman
& Jaggi, 1994).
560 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
negative output.
6
The three inputs are some of the
most widely used to measure performance in the
electric utility industry (Christensen & Greene,
1976; Cowing & Stevenson, 1981) and are among
those frequently employed to calculate DEA
scores in the industry (e.g., Athanassopoulos,
Lambroukos, & Sieford, 1999; Goto & TsuTsui,
1998; Tyteca, 1999).
Capital is the ability of an individual plant to
produce electricity and is measured by the name-
plate generating capacity of the plant. Nameplate
generating capacity is the full-load rating of a
plant to continuously produce electricity. Its use
as a proxy for capital is consistent with prior stud-
ies (Goto & TsuTsui, 1998; Whiteman, 1995;
Yaisawarng & Klein, 1994). Fuel Costs represent
all of the costs associated with the purchase, han-
dling, and storage of fuel inputs burned to gener-
ate electricity. Three issues relative to fuel costs
are worth noting. First, these costs are the largest
and most variable input of a utility’s list of produc-
tion expenses (Bossong, 1999). Also, in accordance
to Freedman and Jaggi (1994), because of its rela-
tive magnitude, special attention should be paid to
this cost category. Second, fuel cost is a speci?c
type of environmental cost as the fuel consumed
(e.g., gas and coal) represents the consumption of
a nonrenewable resource (Tyteca, 1999). Third,
the cost of fuel di?ers by fuel type, and, thus, fuel
cost should re?ect fuel type di?erences among
groups for both cross-sectional and longitudinal
analyses. Therefore, its inclusion is an important
proxy of the ability of management to alter gener-
ating sources in order to take advantage of lower
costs (Haeri, Perussi, & Khawaja, 1999).
Operating Costs are de?ned as the cost of labor
and other expenses related to operating and main-
taining a plant. Thus, operating costs would
include the economic e?ect of investments in
pollution reduction (i.e., the operating costs of
scrubbers). Moreover, when capturing operating
expenses at the plant rather than the ?rm level,
the direct e?ect of controlling and reducing pollu-
tion from a cost basis is more likely to emerge.
Using these inputs and output, relative e?-
ciency for a plant DMU is measured by the follow-
ing program:
Min e ð5Þ
subject to:
X
n
i¼1
w
i
H
i
PH
k
ð6Þ
X
n
i¼1
w
i
ðÀS
i
Þ PÀS
k
ð7Þ
X
n
i¼1
w
i
F
i
6 eF
k
ð8Þ
X
n
i¼1
w
i
K
i
6 eK
k
ð9Þ
X
n
i¼1
w
i
C
i
6 eC
k
ð10Þ
X
n
i¼1
w
i
¼ 1 ð11Þ
One comment about the model needs to be
made. Constraint Eq. (7) compares the hypotheti-
cal composite DMU and the target DMU relative
to the production of SO
2
, a negative output. For a
negative output, the composite DMU can only
produce the same or more output by producing
the same or lower amount of pollution than the
target DMU.
To test the four empirical hypotheses, measur-
ing relative e?ciency at the power plant level for
both cross-sectional and longitudinal analyses is
required. Utilizing DEA scores as our measure is
appropriate because they provide a gauge of tech-
nical e?ciency for each plant relative to the most
e?cient plants in the sample being studied. Not-
withstanding, there are two major concerns rela-
tive to the use of DEA in our study. First, as
noted earlier, DEA computations assume that all
plants have a common production function. How-
ever, in our case high polluting plants (Phase-One
plants) and low polluting plants (Non-Phase-One
plants) may have a di?erent technology (method)
6
For our purposes, good output is de?ned as a revenue-
producing product and the principal objective of the production
processes; negative or bad output is de?ned as a product that is
produced as a consequence of producing the good product, has
no value, and negatively impacts the environment (causes
environmental degradation). This dichotomous de?nition of
electricity production is consistent with exiting literature (i.e.,
Fare & Tyteca, 1996; Golany, Roll, & Rybak (1994); Tyteca
(1997); Tyteca (1996)).
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 561
to produce electricity and thus, any cross-sectional
DEA analysis must consider and account for this
possibility. A similar observation can be made
for longitudinal analysis. For example, as Phase-
One plants reacted to the 1990 CAA intervention,
such actions as switching fuel types and installing
scrubbers may have caused a change in the pro-
duction function over time. In such cases, the
assumption of a common production function
between 1990 and 1995 is problematic.
The second issue relates to the possibility that
DEA scores may, in fact, re?ect the e?ects of fac-
tors other than pollution as possible determinants
of the level of e?ciency. For example, regulatory
climate and di?erences in states in which plants
operate could be causes of di?erences in DEA
scores. Thus, the DEA scores need to be purged
of these e?ects and the resulting ‘‘?ltered’’ DEA
scores should then be used for e?ciency
comparisons.
Common isoquants and relative e?ciency
comparisons
DEA assumes that all DMUs in a sample have
access to and use the same technology, have the
same isoquant, and that ine?ciency is represented
by any variation from the common isoquant (Far-
rell, 1957; Olatubi & Dismukes, 2000). However,
when two seemingly distinct groups of DMUs
are compared, it is conceivable that each group
employs di?erent production techniques with the
result being di?erent best practices isoquants. Fizel
and Nunnikhoven (1992) and Grosskopf and
Valdmanis (1987) provide a methodology that
not only tests to see if two di?erent groups have
a common isoquant but also provides a means of
assessing relative e?ciency when a common iso-
quant does not exist. Their approach requires
three e?ciency indices: (1) overall indices created
when both groups are pooled (EI
p
), (2) within-
group indices produced when DMUs of the sepa-
rate groups are calculated (EI
w
), and (3)
between-group indices (EI
b
).
Fig. 2 illustrates the relationships between EI
p
,
EI
w
, and EI
b
. The example uses nine DMUs and
assumes that points A–D represent DMUs in one
group and points E–I represent DMUs in a second
group. Suppose AD represents the best practices
isoquant for the entire sample, A–I, and that IH
represents the best practices isoquant for the sub-
sample, E–I. Of course, many other representa-
tions for the best practice isoquants for the entire
sample and subgroup are possible; however, this
particular representation is su?cient to illustrate
the methodology.
DMU F can be used to explain the three indi-
ces. The technical e?ciency of F relative to the
overall pooled isoquant AD is:
EI
p
¼ OF
Ã
=OF ð12Þ
The within group index, EI
w
, is calculated as the
ratio of the subsample optimal use (OF
**
) to the
actual input use (OF):
EI
w
¼ OF
ÃÃ
=OF ð13Þ
Finally, the between-group index, EI
b
, is the ratio
of the overall e?ciency index (EI
p
) to the within-
group index (EI
w
):
EI
b
¼ EI
p
=EI
w
¼ ðOF
Ã
=OFÞ=ðOF
ÃÃ
=OFÞ
¼ OF
Ã
=OF
ÃÃ
ð14Þ
To test for a common isoquant for the two groups,
EI
p
and EI
w
are compared for each subsample
using a Wilcoxon Rank Sum z-score to test for dif-
ferences in distributions between the pooled fron-
tier and the subsample frontier. If there is no
U
n
i
t
s
o
f
I
n
p
u
t
2
Units of Input 1
A
B
E*
E
F
F*
C
D
O
H
I
G
F**
Fig. 2.
562 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
signi?cant di?erence between the pooled and with-
in-group indices for both groups, then it is appro-
priate to assume that the subsamples share a
common isoquant. In this case, it is appropriate
to conduct a Wilcoxon Rank Sum test to compare
the EI
p
for each group. If there is a signi?cant dif-
ference between EI
p
and EI
w
then the group being
tested has a frontier isoquant di?erent from the
pooled frontier. For this outcome, the two groups
do not have a common production function and
an EI
b
versus EI
b
Wilcoxon Rank Sum compari-
son is used to test for di?erences in relative e?-
ciency between the two groups.
The relevance of applying this methodology to
our power plant setting is obvious. Relative e?-
ciency assessments at the power plant level are
required for both cross-sectional and longitudinal
analyses. Cross-sectional analyses are done for
plants in 1990 and 1995. For a cross-sectional
analysis, a power plant is a DMU belonging to
one of two groups: (1) the high polluting plant
group (Phase-One plants), or (2) the low polluting
plant group (Non-Phase-One plants). EI
p
and EI
w
are compared to see if the two power plant groups
have a common production function and, depend-
ing on the outcome, either an EI
p
versus EI
p
or EI
b
versus EI
b
Wilcoxon Rank Sum comparison is exe-
cuted to assess di?erences in relative e?ciencies.
Similar longitudinal analyses are conducted using
1990 and 1995 subgroups for the Phase-One and
Non-Phase-One plants, respectively.
Purging and explaining DEA scores
Comparing relative e?ciencies of two groups
using unadjusted DEA scores ignores the possibil-
ity that di?erences in DEA scores can be caused by
factors other than e?ciency. Furthermore, there
may be other variables that cause di?erences in
e?ciency other than di?erences in pollution levels
(Fizel & Nunnikhoven, 1992; Olatubi & Dismukes,
2000). Conceptually, these variables represent
those external factors over which plants have very
little control and yet, may cause di?erences in
DEA scores. Thus, recognizing and removing the
e?ects of a set of confounding variables should
improve the validity of the empirical tests of the
Porter Hypothesis.
Following Fizel and Nunnikhoven (1992) and
Olatubi and Dismukes (2000), regression analysis
is used to purge the DEA scores of the confound-
ing e?ects in the following manner.
7
First, the
unadjusted DEA scores are regressed on the con-
founding factors thought to in?uence the value
of the e?ciency indices. Then, the residuals from
this regression are harvested and become the
purged e?ciency measures that will be used to
compare the relative e?ciencies between groups.
Using these purged values, a cleaner test of the
Porter Hypothesis can be obtained.
In addition to removing the confounding
e?ects, Fizel and Nunnikhoven (1992) and Olatubi
and Dismukes (2000) suggest that it is also possible
to identify speci?c environmental (compliance) vari-
ables that may re?ect actions taken by plants to
reduce or manage pollution to achieve compliance.
Knowing how plants acted to comply with the
1990 CAAA and how these actions a?ect e?ciency
may shed some additional light on the validity of
the Porter Hypothesis (e.g., evidence of innovative
actions). Again, regression analysis is employed to
gain this insight.
Confounding variables
We have identi?ed four confounding variables:
(1) favorableness of regulatory climate (REG);
(2) stringency of state environmentalism (STATE);
(3) age of the plant (AGE); and (4) scale of opera-
tions at the operator level (MKTSHR).
8
The ?rst
two variables are measures of state environmental
e?ects. The second two variables are longer-term
plant and operator-level economic factors that
can a?ect plant-level e?ciency.
Prior research implies that there is an associa-
tion between a utility’s value and the favorableness
of its regulatory climate (Blacconiere, Johnson, &
7
Because DEA scores are bounded between zero and one,
ordinary least square (OLS) may be biased and thus, may not
be an appropriate estimation technique (Cooper, Seiford, &
Tone, 2000). To avoid this potential bias, a tobit-censored
regression model is used (Claggert & Ferrier, 1998; Olatubi &
Dismukes, 2000; Puig-Junoy, 1998).
8
Clearly, one limitation is that the list of potential con-
founding variables is not exhaustive but re?ects those used in
prior studies and those that have data available. Nonetheless,
we would expect their usage to improve the e?ciency measures.
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 563
Johnson, 1997; D’Souza, Jacob, & Soderstrom,
2000; Loudder, Khurana, & Boatsman, 1996).
However, others have argued that these rate-set-
ting procedures may not provide an incentive to
become e?cient and contend that the entire pro-
cess promotes waste and provides little motivation
to cut costs or improve e?ciency (Bossong, 1999;
Haeri, Khawaia, & Perussi, 1997, p. 26). Thus,
the regulatory climate of the state in which the
plant operates may have some e?ect on plant e?-
ciency. In this study, we follow Hughes (2000) and
develop a measure of regulatory climate for states
that assumes the following values: below average
(À1), average (0), and above average (+1). We
adopt this approach to develop (REG).
King and Lenox (2002) indicate that environ-
mental regulation varies across regions. We recog-
nize this in our study as follows. First, we de?ne a
region as the state in which the ?rms operate. Sec-
ond, we recognize that the stringency of regulation
varies among states based on di?erences in policies,
programs, and statutes adopted to protect the envi-
ronment. Meyer (1992) uses a set of twenty envi-
ronmental policy variables to evaluate di?erences
in environmental stringencies across states using
such criteria. As such, in an approach similar to
Hughes (2000), we use the Meyer (1992) state envi-
ronmentalism assessments (STATE) to develop
the following stringency values that are assigned
to the states: weak (À1), moderate (0), and strong
(+1).
Juskow and Schmalaensee (1987) posit that the
performance of a plant will eventually deteriorate
as the unit ages. Following that study, an age var-
iable (AGE) is measured as the di?erence between
the calendar year and the initial date of operation.
Finally, the scale of operations at the operator
level may also have an impact on the ability of a
single plant to make changes. This may occur
because of the transfer of knowledge that may take
place between the plants as more output is pro-
duced by an operator. Olatubi and Dismukes
(2000) suggest that this may have an impact in
the e?ciency of plant. Following that study, we
proxy market share (MKTSHR) as the percentage
of total kilowatt hours produced by the operator
of the plant in the NERC region in which the plant
is located.
Environmental (compliance) variables
Assuming that the plants react to the 1990
CAAA, there are some speci?c environmental (com-
pliance) variables that can be used to manage the
pollution reduction requirements. We have identi-
?ed ?ve compliance variables: (1) the percentage of
sulfur present in the coal consumed (AVC); (2) the
percentage of sulfur present in the petroleum con-
sumed (AVP); (3) the percentage of coal used as a
fuel (BTUC); (4) the number of scrubbers installed
(SCR); and (5) permits usage (PERMIT). The ?rst
three variables are concerned with managing fuel
types to reduce pollution. The last two are alterna-
tive means of managing pollution. Permit usage, of
course, does not lead to pollution reduction.
One of the ways plants can reduce SO
2
emis-
sions is to convert to coal with lower sulfur con-
tent. In fact, the use of low to medium sulfur
content coal in 1995 accounted for 77 percent of
total coal receipts at electric utilities, which is more
than 20 higher than the receipts of the same type of
coal in 1990 (US Department of Energy, Energy
Information Administration, 1994). Moreover,
evidence suggests that the switch from higher to
lower sulfur coal can occur with little deterioration
to plant performance (US Department of Energy,
Energy Information Administration, 1994). There-
fore, the percent of sulfur contained in the coal
consumed at a plant (AVC) is used as a proxy
for the level of sulfur present in the coal employed
to produce electricity. Another fossil fuel often
used to produce electricity is petroleum and its
use also produces SO
2
. Like AVC, we include the
percent of sulfur contained in the petroleum con-
sumed at a plant (AVP) as a proxy for the level
of sulfur present in the petroleum employed to
produce electricity. Moreover, we do recognize
that the level of SO
2
produced by a plant is likely
to be a?ected by the fuel mix used at the plant. In
general, there are three primary fuels that are
employed: (1) coal, (2) petroleum and (3) gas. Log-
ically, plants that adopt coal as their primary
source will produce more pollutants. Therefore,
to account for this e?ect, we include the percent
of coal used to create a BTU of heat at a plant
(BTUC) as a proxy for fuel mix di?erences.
Olatubi and Dismukes (2000) suggest that the
presence of a scrubber may have an impact on
564 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
the e?ciency of a plant and certainly ought to
a?ect the level of pollution. Thus, the presence of
scrubbers (SCR) seems like a reasonable way to
measure one approach used to reduce pollution.
Finally, one of the key components of Title IV is
allowing plants to use permits to emit di?erent lev-
els of pollution. Information about the initial allo-
cation of permits, the use of permits held (initial
allocation plus those bought, sold, or otherwise
distributed), and the amount of permits remaining
after reconciliation could provide some insight
into how the plant managed its pollution, and
the extent to which permits were used to achieve
compliance. For 1990, (PERMIT) is computed
as: the total allowances on hand at the start of
Phase 1 divided by the total number of allowances
that were initially awarded. For 1995, (PERMIT)
is computed as: the total amount of allowances
actually used divided by the actual amount on
hand at the start of Phase 1. This proxy a?ects
only Phase One plants and is used only for the lon-
gitudinal analysis.
Regression models
The regressions for each category of variables
will be run for cross-sectional and longitudinal set-
tings, using the following basic models:
EI ¼ b
0
þ b
1
REG þ b
2
STATE þ b
3
AGE
þ b
4
MKTSHR þ b
5
DUMMY þ e
1
ð15Þ
EI ¼ b
0
þ b
1
AVC þ b
2
AVP þ b
3
BTUC
þ b
4
SCR þ b
5
PERMIT þ b
6
DUMMY þ e
2
ð16Þ
The dependent variable, EI, corresponds to EI
p
for
pooled samples and to EI
w
for within-group sam-
ples. In addition, an indicator variable, DUMMY,
is included to test whether relative e?ciencies di?er
across the current groups being investigated (either
plant type or year). Speci?cally, DUMMY is used
only for pooled regressions and not for within-
group regressions. For cross-sectional pooled sam-
ple regressions, DUMMY is labeled TYPE and is
equal to one if the plant is a Phase-One plant and
zero if it is a Non-Phase-One plant. For longitudi-
nal pooled sample regressions, DUMMY is la-
beled as YEAR and is equal to one if the period
is 1995 and zero if it is 1990. Finally, the variable,
PERMIT, is included only for the sample longitu-
dinal regressions exclusively involving Phase-One
plants.
Sample and data collection
Data were collected for Phase-One plants for the
periods 1990 and 1995 and for Non-Phase-One
plants for the same periods. Non-Phase-One plants
were identi?ed using the Inventory of Power Plants
in the United States for 1990. Non-Phase-One
plants were drawn from the same North American
Electric Reliability Council (NERC) region as
Phase One plants. NERC region is used in this
manner because it is assumed that plants located
within the same region face a common set of pro-
duction, operating, and regulatory characteristics.
In addition, prior studies have used NERC regions
as a basis for cost comparisons within the electric
utility industry (Blacconiere et al., 1997; Knutson,
1995). Further, to the extent possible, nameplate
capacity is also used to identify Non-Phase One
plants. This approach was adopted based on simi-
lar methodologies present in existing studies that
make comparisons within the electric utility indus-
try (Goto & TsuTsui, 1998; Whiteman, 1995;
Yaisawarng & Klein, 1994).
The Federal Energy Regulatory Commission
(FERC) Form 1 reports were used to gather data
related to the inputs, good (kilowatt hour) output,
and the age of the plants. Only major investor-
owned electric utilities are required to submit a
FERC Form 1. Investor owned utilities are those
electric utilities owned by investors. In contrast,
publicly owned utilities are operated by the federal
government, a state or local municipality, or an
electric cooperative. Data incorporated in a FERC
Form 1 include income and earnings, operating
and maintenance expenses, and various produc-
tion statistics. Plant level ?nancial and operating
data are located in the ‘‘Steam Generating Plant
Statistics for Large Plants’’ section of the FERC
Form 1. The relative e?ciencies (i.e., DEA scores)
for this study were produced using DEA-Solver-
Pro 3.0, a windows based linear programming
model developed by Saitech, Inc.
The Energy Information Administration of the
Department of Energy was the source for SO
2
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 565
emissions. Like Hughes (2000), assessments of the
regulatory environment were drawn from the
Value Line Investment Surveys. The market share
of the operators within the NERC region was
compiled using the Energy Information Adminis-
tration of the Department of Energy Form 906
Database. The average sulfur content of coal and
petroleum, as well as the use of coal to produce
electricity were all obtained from the 1990 and
1995 Cost and Quality of Fuels For Electric Utility
Plants. Information about scrubber activity was
found at the Energy Information Administration
Clean Air Act Database Browser. Finally, informa-
tion about the permits came from the Environmen-
tal Protection Agency’s 1995 Acid Rain Compliance
Report.
Empirical results
Descriptive statistics
Table 1 details and summarizes the sample
selection process used. Phase-One of the 1990
CAAA a?ected 164 plants. Of these 164 plants,
80 were eliminated from the sample for one of sev-
eral reasons: (1) they did not ?le a FERC Form 1;
(2) plant level data were not included in the FERC
Form 1 submission; (3) a plant had missing or
incomplete operating or ?nancial information;
and, (4) a suitable Non-Phase One plant could
not be located. Ultimately 84 Phase-One and
Non-Phase-One plants were included in the study.
Table 2 provides a summary of plant level data
and the descriptive statistics for Phase One and
Non-Phase-One plants for 1990 and 1995. Panels
IA and IB of Table 2 re?ect cross-sectional com-
parisons of Phase One and Non-Phase One plants
at 1990 and 1995, respectively. Panels IIA and IIB
provide a longitudinal comparison of plant types
across 1990 and 1995. Statistical tests of the means
of the data presented in the Panels are also
included to provide a preliminary assessment of
the e?ects of the 1990 CAAA on the plants.
Panel IA of Table 2 reveals that there is no sig-
ni?cant di?erence in size (as measured by name
plate generating capacity), kilowatt hours, or costs
between the Phase-One and Non-Phase-One
plants in 1990. However, there is a signi?cant dif-
ference in SO
2
production between the Phase-One
and Non-Phase-One plants (the higher polluting
plants produce nearly four times as much total
SO
2
). Thus, in 1990, Non-Phase-One plants on
average produce the same good output as Phase-
One plants, but do so with a fraction of the bad
output and no greater cost. The traditional
hypothesis predicts that reducing bad output is
costly while ecoe?ciency maintains that reducing
bad output will increase e?ciency. Thus, the
descriptive evidence provides some weak or mild
support for the empirical prediction that the lower
Table 1
Sample of phase one plants with available FERC Form 1 data
a
Total Phase One Plants
b
164
Plants Not Required to File a FERC Form 1
c
(25)
Initial Set of Plants with FERC Form 1 Data 139
Data Not Located in the FERC Form 1
d
(15)
Data in FERC Form 1 was Unusable
e
(26)
Unable to Locate as Suitable Non-Phase One Plant
f
(14)
Final Phase One Plants Used
g
84
a
The data for the study are obtained from the Federal Energy
Regulatory Commission Form 1 (FERC Form 1). Only major
investor-owned electric utilities are required to submit a FERC
Form 1. FERC Form 1 submissions are required annually.
Data presented in a FERC Form 1 include income and earn-
ings, operating and maintenance expenses, and various pro-
duction statistics. Plant level ?nancial and operating data are
located in the ‘‘Steam Generating Plant Statistics for Large
Plants’’ section of the FERC Form 1.
b
Includes all of the electric utility plants a?ected by Phase
One of Title IV of the 1990 CAAA. These plants are referred to
as Phase One plants. Starting in 1995, the SO
2
emissions of
Phase One plants were evaluated by the Environmental Pro-
tection Agency (EPA) to ensure that they are in compliance
with the 1990 CAAA.
c
Includes those plants that are publicly-owned and thus, are
not required to submit a FERC Form 1. Publicly-owned utili-
ties are operated by (1) the federal government; (2) a state or
local municipality; or (3) an electric cooperative.
d
Includes those Phase One plants whose ‘‘Steam Generating
Plant Statistics For Large Plants’’ data were not included with
the FERC Form 1 ?led with the Commission.
e
Includes those Phase One plants whose ‘‘Steam Generating
Plant Statistics For Large Plants’’ data ?led in the FERC
FORM 1 were incomplete.
f
Includes those Phase One plants that had no similar Non-
Phase-One plant.
g
Re?ects the total number of Phase One Plants: (1) whose
data is complete; (2) can be successfully associated with Non-
Phase One plants. Thus, the grand total of the plants included
in the study is 168 (i.e., 84 Phase One and 84 Non-Phase One).
566 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
Table 2
Summary and descriptive statistics for phase one and non-phase one plants
a
Variable Phase
One
Non-Phase
One
Net
Di?
% Di? Phase One
Mean
Phase One
Std. Dev
Non-Phase
One Mean
Non-Phase
One Std. Dev
T Stat
Part I: Cross-sectional analyses (Hypothesized mean di?erences: Non-Phase-One mean less Phase-One mean)
Panel A: 1990 Comparisons (n = 84 for each group)
Inputs
NPC 72407 73064 657 0.91 861.98 607.60 869.81 613.93 0.08
PC 6995 6837 (158) (2.26) 83 63 81 70 À0.18
FC 5467 5370 (97) (1.78) 65 54 63 56 À0.13
OC 1527 1467 (60) (3.96) 18 11 17 20 À0.28
Outputs
KWH 332428 264437 (67991) (20.45) 3957 3307 3148 3084 À1.64
POLL 5153797 1378377 (3775420) (72.26) 61354 68062 16409 29040 À5.57
*
Panel B: 1995 Comparisons (n = 84 for each group)
Inputs
NPC 71816 72557 741 1.03 854.96 614.16 863.77 614.14 0.09
PC 5331 5008 (323) (6.06) 63 54 59 48 À0.48
FC 3907 3835 (72) (1.84) 46 42 45 36 À0.14
OC 1424 1172 (252) (17.70) 16 14 13 16 À1.27
Outputs
KWH 317387 266246 (51141) (16.11) 3778 3573 3169 2984 À1.20
POLL 2810175 1397005 (1413170) (50.29) 33454 40824 16631 27285 À3.14
*
Operations
NDECREASE 76 44 (32) (42.1) 0.904 0.294 0.524 .499 À5.47
*
Variable 1990 1995 Net Di? % Di? 1990 Mean 1990 Std. Dev 1990 Mean 1995 Std. Dev T Stat
Part II: Longitudinal Analyses (Hypothesized mean di?erences: 1995 mean less 1990 mean)
Panel A: Phase One Plants (n = 84)
Inputs
NPC 72407 71816 (591) (0.82) 861.98 607.60 854.96 614.16 À0.07
PC 6995 5331 (1664) (23.79) 83 63 63 54 À2.16
*
FC 5467 3907 (1560) (28.53) 65 54 46 42 À2.45
*
OC 1527 1424 (103) (6.75) 18 11 16 14 À0.61
Outputs
KWH 332428 317387 (15041) (4.52) 3957 3307 3778 3573 À0.34
POLL 5153797 2810175 (2343622) (45.47) 61354 68062 33454 40824 À3.22
*
Panel B: Non-Phase One Plants (n = 84)
Inputs
NPC 73064 72557 (507) (0.69) 869.81 613.93 863.77 614.14 0.06
PC 6837 5008 (1829) (26.75) 81 70 59 48 2.33
*
FC 5370 3835 (1535) (28.58) 63 56 45 36 2.48
*
(continued on next page)
R
.
D
.
B
u
r
n
e
t
t
,
D
.
R
.
H
a
n
s
e
n
/
A
c
c
o
u
n
t
i
n
g
,
O
r
g
a
n
i
z
a
t
i
o
n
s
a
n
d
S
o
c
i
e
t
y
3
3
(
2
0
0
8
)
5
5
1
–
5
8
1
5
6
7
polluting 1990 Non-Phase-One plants are at least
as e?cient as the 1990 Phase-One plants. Panel
IB of Table 2 compares the Phase-One and Non-
Phase-One plants in 1995 after intervention. The
Phase-One plants continue to have signi?cantly
higher pollution than Non-Phase-One plants with
no signi?cant di?erences in cost and good output.
Thus, the lower polluting plants continue to pro-
duce (on average) the same good output with no
greater cost, an outcome weakly consistent with
ecoe?ciency.
However, Panel IB of Table 2 also reveals that
the di?erence in the levels of pollution has nar-
rowed, with the Phase-One plants producing only
about double the pollution of Non-Phase-One
plants in 1995. Also, as Panels IIA and IIB of
Table 2 indicate, from 1990 to 1995 the Phase-
One plants reduced total pollution by 45.47 per-
cent (a statistically signi?cant decrease) while the
Non-Phase-One plants increased their total pollu-
tion by 1.35 percent (not a signi?cant change).
Furthermore, the pollution reductions of the
Phase-One plants were wide-spread with 76 or
about 90% of the plants reducing pollution from
1990 to 1995 (Panel IB) with an average reduction
of 27,900 tons. Thus, not only did the Phase-One
plants reduce pollution by a large amount but
there were also signi?cant decreases in production
and fuel costs from 1990 to 1995. Reducing pollu-
tion while simultaneously decreasing cost is consis-
tent with ecoe?ciency. The outcome tends to
support the Porter Hypothesis and thus, provides
evidence to support the validity of ecoe?ciency.
Thus, the descriptive data suggest that the
Phase-One plants have no signi?cant di?erence in
output from 1990 to 1995, post signi?cant reduc-
tions in fuel costs and total production costs
and, yet, manage to simultaneously reduce pollu-
tion by a signi?cant amount. The Non-Phase-
One plants experience the same outcomes with
one exception: there is no signi?cant change in pol-
lution. Interestingly, both groups experience rela-
tively the same percentage reduction in fuel costs
(28 percent). Collectively, these results provide sev-
eral additional insights. First, since fuel costs are
an environmental cost (re?ecting the use of nonre-
newable resources), the descriptive evidence por-
trays an environmental management model T
a
b
l
e
2
(
c
o
n
t
i
n
u
e
d
)
V
a
r
i
a
b
l
e
1
9
9
0
1
9
9
5
N
e
t
D
i
?
%
D
i
?
1
9
9
0
M
e
a
n
1
9
9
0
S
t
d
.
D
e
v
1
9
9
0
M
e
a
n
1
9
9
5
S
t
d
.
D
e
v
T
S
t
a
t
O
C
1
4
6
7
1
1
7
2
(
2
9
5
)
(
2
0
.
1
1
)
1
7
2
0
1
3
1
6
1
.
2
4
O
u
t
p
u
t
s
K
W
H
2
6
4
4
3
7
2
6
6
2
4
6
1
8
0
9
0
.
6
8
3
1
4
8
3
0
8
4
3
1
6
9
2
9
8
4
À
0
.
0
5
P
O
L
L
1
3
7
8
3
7
7
1
3
9
7
0
0
5
1
8
6
2
8
1
.
3
5
1
6
4
0
9
2
9
0
4
0
1
6
6
3
1
2
7
2
8
5
À
0
.
0
5
a
T
h
i
s
t
a
b
l
e
r
e
p
o
r
t
s
s
e
l
e
c
t
e
d
d
e
s
c
r
i
p
t
i
v
e
a
n
d
s
t
a
t
i
s
t
i
c
a
l
p
l
a
n
t
l
e
v
e
l
d
a
t
a
o
n
:
(
1
)
a
c
r
o
s
s
-
s
e
c
t
i
o
n
a
l
(
b
e
t
w
e
e
n
)
p
l
a
n
t
t
y
p
e
b
a
s
i
s
;
a
n
d
(
2
)
a
l
o
n
g
i
t
u
d
i
n
a
l
(
w
i
t
h
i
n
)
p
l
a
n
t
t
y
p
e
b
a
s
i
s
.
A
l
l
p
-
v
a
l
u
e
s
r
e
l
a
t
e
t
o
a
t
w
o
-
s
i
d
e
d
t
-
t
e
s
t
w
h
e
r
e
*
r
e
?
e
c
t
s
p
<
0
.
1
0
.
T
h
e
v
a
r
i
a
b
l
e
d
e
?
n
i
t
i
o
n
s
:
N
P
C
T
h
e
n
a
m
e
p
l
a
t
e
(
f
u
l
l
l
o
a
d
)
g
e
n
e
r
a
t
i
n
g
c
a
p
a
c
i
t
y
e
x
p
r
e
s
s
e
d
a
s
m
i
l
l
i
o
n
s
o
f
w
a
t
t
s
c
a
p
a
c
i
t
y
.
K
W
H
T
h
e
k
i
l
o
w
a
t
t
p
e
r
h
o
u
r
o
f
e
l
e
c
t
r
i
c
i
t
y
g
e
n
e
r
a
t
e
d
e
x
p
r
e
s
s
e
d
a
s
m
i
l
l
i
o
n
s
o
f
K
W
H
.
P
C
T
h
e
t
o
t
a
l
p
r
o
d
u
c
t
i
o
n
c
o
s
t
s
(
f
u
e
l
p
l
u
s
a
l
l
o
t
h
e
r
c
o
s
t
s
)
e
x
p
r
e
s
s
e
d
i
n
m
i
l
l
i
o
n
s
o
f
c
o
n
s
t
a
n
t
(
1
9
9
0
)
d
o
l
l
a
r
s
.
F
C
F
u
e
l
c
o
s
t
s
e
x
p
r
e
s
s
e
d
i
n
m
i
l
l
i
o
n
s
o
f
c
o
n
s
t
a
n
t
(
1
9
9
0
)
d
o
l
l
a
r
s
.
O
C
O
p
e
r
a
t
i
n
g
c
o
s
t
s
e
x
p
r
e
s
s
e
d
i
n
m
i
l
l
i
o
n
s
o
f
c
o
n
s
t
a
n
t
(
1
9
9
0
)
d
o
l
l
a
r
s
.
P
O
L
L
T
h
e
a
m
o
u
n
t
o
f
S
O
2
p
o
l
l
u
t
i
o
n
i
n
a
c
t
u
a
l
t
o
n
s
.
N
D
E
C
R
E
A
S
E
N
u
m
b
e
r
o
f
p
l
a
n
t
s
s
h
o
w
i
n
g
a
d
e
c
r
e
a
s
e
i
n
p
o
l
l
u
t
i
o
n
f
r
o
m
1
9
9
0
t
o
1
9
9
5
.
568 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
where environmental costs are simultaneously
reduced with improvements in environmental
quality. Second, this evidence is consistent with
the claims of process innovations within the elec-
tric utility industry and spillover innovations pro-
duced by suppliers. Finally, it also appears that the
Non-Phase-One plants may be capturing bene?ts
produced by the Phase-One plants, an outcome
consistent with the derived-bene?ts assumption.
Results of DEA analyses with unadjusted e?ciency
indices
The distribution of the unadjusted DEA scores
and the summary statistics are provided in Tables
3 and 4 for each of the three e?ciency indices
(pooled, within-group and between-group) that
are needed to test the various hypotheses of inter-
est. Table 5 provides the appropriate statistical
tests using these three indices.
Cross-sectional analysis
The cross-sectional hypotheses predict that the
lower polluting, Non-Phase-One plants will be rel-
atively more e?cient than the higher polluting,
Phase-One plants. To ensure that any di?erences
in e?ciency are not caused by di?erences that arise
because the two types face di?erent frontiers, we
?rst test to see whether the Phase-One and Non-
Phase-One groups share a common production
Table 3
Cross-sectional frequencies and summary statistics unadjusted e?ciency indices
a
E?ciency score ranges EI
p
= pooled index EI
w
= within-group index EI
b
= between–group index
Total Phase One Non-Phase One Phase One Non-Phase-One Phase One Non-Phase-One
Panel A: 1990 Statistics for Phase-One and Non-Phase-One Plants
1.00 26 10 16 19 27 38 24
0.90–0.99 2 0 2 0 1 20 18
0.80–0.89 1 0 1 1 3 10 30
0.70–0.79 10 2 8 3 2 6 9
0.60–0.69 15 8 7 10 9 3 2
0.50–0.59 29 17 12 15 14 2 1
0.40–0.49 47 31 16 25 12 3 0
0.30–0.39 21 12 9 9 6 1 0
0.20–0.29 15 3 12 1 10 1 0
0.10–0.19 2 1 1 1 0 0 0
N 168 84 84 84 84 84 84
Mean 0.5624 0.5350 0.5897 0.6193 0.6612 0.8922 0.8948
Standard deviation 0.2338 0.2021 0.2600 0.2327 0.2770 0.1724 0.0974
Panel B: 1995 Statistics for Phase-One and Non-Phase-One Plants
1.00 25 10 15 13 21 14 59
0.90–0.99 1 1 0 0 0 41 12
0.80–0.89 2 1 1 1 3 25 4
0.70–0.79 8 1 7 5 6 1 7
0.60–0.69 14 9 5 13 4 2 0
0.50–0.59 28 14 14 19 10 1 2
0.40–0.49 43 26 17 18 16 0 0
0.30–0.39 37 16 21 11 20 0 0
0.20–0.29 9 5 4 3 4 0 0
0.10–0.19 1 1 0 1 0 0 0
N 168 84 84 84 84 84 84
Mean 0.5544 0.5345 0.5743 0.5815 0.6105 0.9149 0.9578
Standard deviation 0.2283 0.2158 0.2398 0.2159 0.2655 0.0787 0.0970
a
This table re?ects the distribution and descriptive statistics for the three types of e?ciency indices for Phase One and Non-Phase
One plants within each year before adjusting for confounding variables. Unadjusted indices are bounded between zero and one.
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 569
function (Panel A of Table 5). This question is
addressed by comparing the pooled indices for
1990 and 1995 with the within-group indices for
the two types of plants. Wilcoxon Rank Sum
z-scores are used to compare the distributions of
the two groups.
For each year, at least one frontier of the two
subgroups is statistically di?erent from the pooled
frontier. Thus, the two groups do not share a com-
mon production function in either year. For 1990,
both frontiers are di?erent from the pooled fron-
tier, leaving the e?ciency relationships unclear.
However, for 1995, only the Phase-One plants have
a frontier that is shifted out from the pooled fron-
tier, which immediately supports the hypothesis
that the Non-Phase-One plants are relatively more
e?cient. Based on the absence of a common fron-
tier, between-group tests are used to assess di?er-
ences in relative e?ciency between the two
groups (Table 5, Panel B). For 1990, the evidence
indicates that the Phase-One plants are relatively
more e?cient than the Non-Phase-One plants, an
outcome inconsistent with the 1990 cross-sectional
hypothesis. However, the evidence is strongly sup-
portive of the 1995 cross-sectional hypothesis.
Longitudinal analysis
The longitudinal hypotheses predict that the
1995 plants of each type will be relatively more e?-
cient than their 1990 counterparts. Table 5, Panel
Table 4
Longitudinal frequencies and summary statistics unadjusted e?ciency indices
a
E?ciency score ranges EI
p
= pooled index EI
w
= within-group index EI
b
= between-group index
Total 1990 1995 1990 1995 1990 1995
Panel A: Phase-One Statistics for 1990 and 1995
1.00 23 10 13 19 13 10 81
0.90–0.99 0 0 0 0 0 15 3
0.80–0.89 1 0 1 1 1 38 0
0.70–0.79 7 2 5 3 5 9 0
0.60–0.69 16 3 13 10 13 3 0
0.50–0.59 30 11 19 15 19 4 0
0.40–0.49 50 32 18 25 18 2 0
0.30–0.39 29 18 11 9 11 2 0
0.20–0.29 10 7 3 1 3 1 0
0.10–0.19 2 1 1 1 1 0 0
N 168 84 84 84 84 84 84
Mean 0.5429 0.5044 0.5813 0.6193 0.5815 0.8301 0.9997
Standard deviation 0.2155 0.2094 0.2159 0.2327 0.2159 0.1533 0.0020
Panel B: Non-Phase-One Statistics for 1990 and 1995
1.00 33 15 18 27 21 16 32
0.90–0.99 1 1 0 1 0 6 45
0.80–0.89 3 1 2 3 3 17 4
0.70–0.79 11 6 5 2 6 23 1
0.60–0.69 4 2 2 9 4 15 0
0.50–0.59 4 11 13 14 10 5 1
0.40–0.49 29 13 16 12 16 2 1
0.30–0.39 39 17 22 6 20 0 0
0.20–0.29 20 14 6 10 4 0 1
0.10–0.19 4 4 0 0 0 0 0
N 168 84 84 84 84 84 84
Mean 0.5562 0.2665 0.2896 0.6612 0.6105 0.7955 0.9596
Standard deviation 0.2660 0.2730 0.2585 0.2770 0.2655 0.1433 0.1141
a
This table re?ects the distribution and descriptive statistics for the three types of e?ciency indices for Phase-One and Non-Phase-
One plants across years before adjusting for confounding variables. Unadjusted indices are bounded between zero and one.
570 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
A, indicates that the 1990 frontiers are statistically
di?erent from the pooled frontier, while the 1995
frontiers are not statistically di?erent. Thus, the
1990 frontiers are shifted out from the pooled
frontier, while the 1995 frontiers are not statisti-
cally di?erent from the pooled frontiers. This out-
come supports the hypotheses that the 1995
plants are relatively more e?cient than the 1990
plants (see Fig. 2). The between-group compari-
sons (Table 5, Panel B) strongly support the
hypotheses that the 1995 plants are relatively more
e?cient than the 1990 plants. Thus, the longitudi-
nal hypotheses are consistent with the Porter
Hypothesis.
The evidence of the unadjusted DEA scores is
largely supportive of the Porter Hypothesis and
of ecoe?ciency, with only one cross-sectional
hypothesis failing the test. However, while the
unadjusted indices provide useful evidence, they
may not be completely de?nitive because of the
possibility that the indices are a?ected by contam-
inating variables.
DEA analysis with adjusted indices
Descriptive data for the confounding and com-
pliance variables are given in Table 6 for both
pooled and within-group samples. Tobit regression
Table 5
Statistical analyses of unadjusted productive e?ciency
a
Cross-sectional analysis Distribution
b
Longitudinal analysis Distribution
b
Panel A: EI
p
versus EI
w
1990 EI
p
6308 Phase-One EI
p
5844
1990 Phase-One EI
w
7888 1990 Phase-One EI
w
8352
Z-score 2.51
**
Z-score À3.98
***
1990 EI
p
6574 Phase-One EI
p
7091
1990 Non-Phase-One EI
w
7622 1995 Phase-One EI
w
7104
Z-score 1.67
*
z-Score À0.02
1995 EI
p
6536 Non-Phase-One EI
p
6099
1995 Phase-One EI
w
7660 1990 Non-Phase-One EI
w
8097
Z-score 1.78* z-score À3.19
***
1995 EI
p
6902 Non-Phase-One EI
p
6857
1995 Non-Phase-One EI
w
7293 1995 Non-Phase-One EI
w
7339
Z-score 0.62 Z-score À0.76
Cross-sectional analysis Distribution
c
Longitudinal analysis Distribution
c
Panel B: EI
b
versus EI
b
1990 Phase One EI
b
7691 1990 Phase-One EI
b
4005
1990 Non-Phase One EI
b
6505 1995 Phase-One EI
b
10191
Z-score À1.92
**
Z-score 10.69
***
1995 Phase One EI
b
5291 1990 Non-Phase One EI
b
5005
1995 Non-Phase One EI
b
8905 1995 Non-Phase One EI
b
9191
Z-score 5.98
***
Z-score 6.71
***
a
This table reports the results of tests of comparative relative e?ciency based on the unadjusted productive e?ciency indices. For
this group of tests, Panel A tests for the presence of a common production function. Panel B tests for di?erences between groups when
a common production function is unavailable. Signi?cance levels of p < 0.10 and p < 0.05 and p < 0.001 are denoted as * , **, and ***
respectively.
b
For all tests, the null hypothesis assumes that the expected rank sums (distributions) of the groups being compared is the same.
Results re?ect the cumulative expected rank sum of each group and the z-score when a Wilcoxon Rank Sum test is employed. All
z-scores are two sided.
c
For the cross-sectional hypotheses, the null hypotheses assume that the Phase-One plants are at least as e?cient as the Non-Phase-
One plants. For the longitudinal analyses, the null hypotheses assume that the 1990 plants are at least as e?cient as the 1995 plants.
Results re?ect the cumulative expected rank sum of each group and the Z-score when a Wilcoxon Rank Sum test is employed. All
Z-scores are one sided.
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 571
Table 6
Summary and descriptive statistics
a
Variable A: 1990 Pooled B: 1995 Pooled C: Phase One D: Non-Phase-One
Mean Std Dev Mean Std Dev Mean Std Dev Mean Std Dev
Part I: Overall comparisons (N = 168)
Confounding
REG À0.0952 0.5411 À0.7142 0.5745 À0.3571 0.4756 À.1309 0.6525
STATE 0.3214 0.7285 0.3214 0.7285 0.2619 0.6403 0.3809 0.8027
AGE 32.92 15.5761 37.928 15.5761 35.2380 13.7737 35.61 17.5505
MKTSHR 0.0829 0.0855 0.0821 0.0858 0.0732 0.0671 0.0918 0.1000
Compliance
AVC 0.0133 0.0108 0.0106 0.0093 0.0159 0.0096 0.0081 0.0092
AVP 0.003 0.0051 0.0026 0.0037 0.0026 0.0034 0.0032 0.0053
BTUC 0.7967 0.3779 0.781 0.3843 0.9322 0.2260 0.6516 0.4471
SCR 0.1607 0.5721 0.2380 0.6678 0.2083 0.5979 0.1904 0.6469
PERMIT 0.8958 0.4938
Outcome
INDEX 0.5624 0.2338 0.5544 0.2283 0.5429 0.2155 0.5562 0.2660
E: 1990 Phase One F: 1995 Phase One G: 1990 Non-Phase One H: 1995 Non-Phase
One
Part II: Yearly comparisons (N = 168)
Confounding
REG À0.4761 0.4885 À0.2380 0.4650 À0.1428 0.6427 À0.1190 0.6659
STATE 0.2619 0.6422 0.2619 0.6422 0.3809 0.8051 0.3809 0.8051
AGE 32.73 13.5842 37.73 13.58 33.11 17.4227 38.1190 17.4227
MKTSHR 0.7433 0.0688 0.7220 0.0658 0.9155 0.0992 0.9219 0.1014
Compliance
AVC 0.0184 0.0096 0.0133 0.0089 0.0083 0.0094 0.0079 0.0090
AVP 0.0029 0.0040 0.0022 0.0027 0.0036 0.0061 0.0029 0.0045
BTUC 0.9403 0.140 0.9240 0.2384 0.6532 0.4470 0.6501 0.4498
SCR 0.1309 0.4854 0.2857 0.6867 0.1904 0.6489 0.1904 0.6489
PERMIT 0.5666 0.5471 0.3292 0.2780
Outcome
INDEX 0.6193 0.2370 0.5815 0.2159 0.6612 0.2770 0.6105 0.2655
a
This table re?ects the summary and descriptive data relative to the confounding variables that will be used to adjust the initial
(unadjusted) e?ciency indices. Panels A and B of Part I re?ect data when Phase One and Non-Phase One plants are pooled in 1990 and
1995. Panel C provides results when 1990 and 1995 Phase One are pooled. Panel D re?ects the same information for Non-Phase One
plants. Panels E and F of Part II provide data related to Phase One plants when 1990 and 1995 are considered separately. Panels G and
H re?ect the same information for the Non-Phase One plants. The variable de?nitions are:
REG Measures how favorably a state regulatory commission addresses rate increase requests.
STATE Measures the stringency of state environmentalism.
AGE The age of the plant.
MKTSHR The percentage of kilowatt hours of electricity produced by the operator of the plant in the NERC region in which the plant
operates.
AVC The average sulfur content of the coal used by the plant.
AVP The average sulfur content of the petroleum used by the plant.
BTUC The percent of heat (in BTUs) produced by the use of coal at the plant to produce kilowatt hours of electricity.
SCR The number of scrubbers at the plant.
INDEX The unadjusted data envelopment e?ciency index of the electricity plants. These scores lie between zero and one.
PERMIT Measures the e?ciency of the allocation and use of pollution permits. In 1990 and 1995, permits were allocated only to
Phase One.
572 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
is used to control for these potential contaminat-
ing variables; the residuals of the regressions are
used as the adjusted e?ciency indices. The pooled
and within-group regressions for cross-sectional
and longitudinal settings are presented in Tables
7 and 8, respectively. Table 9 presents the statisti-
cal analyses using the adjusted indices.
Pooled regressions
Pooled regressions are of two types: cross-sec-
tional and longitudinal. The pooled cross-sectional
regressions for 1990 reveal that all confounding
variables are signi?cant with negative coe?cients
(Table 7, Panel A). For 1995, only stringency and
market share are signi?cant. Thus, regulatory
climate, the stringency of state environmental reg-
ulations, plant age and market share have a nega-
tive e?ect on e?ciency levels. This may indicate,
for example, that the wrong kind of regulations
(climate and stringency) can have a depressive
e?ect on e?ciency (a common complaint of ?rms
facing environmental restrictions). It is reasonable
to presume that as a plant ages, e?ciency
decreases. The negative relationship between mar-
ket share and e?ciency is somewhat puzzling –
unless the operator has gone beyond the optimal
size. Finally, it is most interesting to note that when
plant type is added in the equation (Model 2), it is
found to be signi?cant with a negative coe?cient
(for both 1990 and 1995). This outcome provides
Table 7
Pooled regression analysis controlling for confounding variables
a
Model Intercept REG STATE AGE MKTSHR TYPE Log R
2
Model 1 (M1): EI
p
= b0 + b1REG + b2 STATE+b3AGE + b4MKTSHR + e
Model 2 (M2): EI
p
= b0 + b1REG + b2 STATE + b3AGE + b4MKTSHR + b5TYPE + e
Panel A: Cross-sectional pooled sample results, N = 168
M1: 1990 0.7574 À0.0823 À0.0517 À0.0032 À0.7579
15.33
***
À2.42
**
À1.83
*
À2.57
***
À3.35
***
À33.53 0.45
M2: 1990 0.8020 À0.0761 À0.0564 À0.0032 À0.7961 À0.0787
15.00
***
À2.25
**
À2.04
**
À2.60
***
À3.55
***
À2.07
**
À31.42 0.46
M1: 1995 0.6903 0.0142 À0.0849 À0.0010 À0.6199
12.60
***
0.417 À3.018
***
À0.850 À2.57
***
À33.36 0.45
M2: 1995 0.7315 0.0192 À0.0886 À0.0010 À0.6528 À0.0726
12.52
***
0.54 À.3174
***
À0.875 À2.73
***
À1.90
*
À31.57 0.46
Panel B: Longitudinal Pooled Sample Results, N = 168
M1: Phase-One 0.7332 0.0211 À0.0524 À0.0027 À0.9074
13.70
***
0.549 À1.79
*
À2.02
**
À3.37
***
À22.84 0.46
M2: Phase-One 0.7062 0.0204 À0.0496 À0.0033 À0.8931 0.0963
13.31
***
0.54 À1.74
*
À2.53
***
À3.40
***
2.71
***
À19.26 0.46
M1: Non-Phase-One 0.7489 À0.0698 À0.0583 À0.0021 À0.8164
12.75
***
À1.87
*
À1.85
*
À1.57 À3.33
***
À65.80 0.43
M2: Non-Phase-One 0.7254 À0.0704 À0.0573 À0.0024 À0.8235 0.0691
12.02
***
À1.90
*
À1.83
*
À1.78
*
À3.39
***
1.47 À64.71 0.44
a
This table provides the regression results for the regression results using pooled data. These regressions assess the e?ects of
confounding variables. Amounts in the tables are parameter and t-statistic values. Levels of signi?cance of p < 0.10, 0.05, and 0.01 are
denoted as *, **, and ***, respectively. The variable de?nitions are:
EI
p
The unadjusted data envelopment e?ciency score of the electricity plants (pooled indices). These indices lie between zero
and one.
REG Measures how favorably a state regulatory commission addresses rate increase requests.
STATE Measures the stringency of state environmentalism.
AGE The age of the plant.
MKTSHR The percentage of kilowatt hours of electricity produced by the operator of the plant in the NERC region in which the
plant operates.
TYPE Dummy variable equal to one if Phase One plant and zero if a Non-Phase One plant (for Panel A). Dummy variable equal to
one if 1995 and zero if 1990 (for Panel B).
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 573
evidence that the Non-Phase-One plants are more
e?cient than Phase-One plants, an outcome consis-
tent with the two cross-sectional hypotheses.
The pooled longitudinal regressions (Table 7,
Panel B) indicate that both Phase-One and Non-
Phase-One plants are a?ected by the confounding
variables over time, although the e?ects di?er.
The Phase-One plants have signi?cant negative
coe?cients for stringency, age, and market share,
while the Non-Phase-One plants have signi?cant
negative coe?cients for regulatory climate, strin-
gency, age, and market share. The presence of these
variables tends to decrease the relative e?ciency of
the plants, similar to the e?ects noted for cross-
sectional analysis. When year is added as a vari-
able, the 1995 Phase-One plants emerge with a
signi?cant positive coe?cient, supporting the pre-
diction that relative e?ciency is greater for 1995
plants than 1990 plants. However, the absence of
a signi?cant relationship for the Non-Phase-One
plants is inconsistent with the prediction that the
1995 Non-Phase-One plants are relatively more
e?cient than their 1990 counterparts.
Within-group regressions
Table 8 displays the within-group regressions
that allow the assessment of how the confounding
variables a?ect the within-group indices. The
regressions also provide the within-group adjusted
indices (residuals) that will be used in Table 9 for
testing to see if a common production function is
now possible (by comparing the within-group
adjusted indices with the corresponding adjusted
pooled indices). The Non-Phase-One plants have
two confounding variables in each year that are
signi?cant and negative (climate and age in 1990
and stringency and market share in 1995), while
the Phase-One plants only have market share with
a signi?cant negative coe?cient for 1995.
Statistical analyses with adjusted e?ciency indices
In Table 9, Panel A, initial tests results provide
evidence that is in favor of a common production
function for both cross-sectional and longitudinal
pooled data. Thus, the statistical comparisons of
the subsamples by type and by year are made using
the pooled indices (shown in Panel B of Table 9).
The statistical results in Panel B of Table 9 are
fully consistent with the predictions that the
Non-Phase-One plants, which are lower polluting,
are relatively more e?cient than the higher pollut-
ing Phase-One plants. Moreover, the predictions
that the 1995 plants are more e?cient than the
1990 plants are also fully supported. Thus, both
Table 8
Within-group regression analyses controlling for confounding variables
a
Model Intercept REG STATE AGE MKTSHR Log R
2
Model: EI
w
= b0 + b1REG + b2 STATE + b3AGE + b4MKTSHR + e
1990 0.6897 0.0419 À0.0568 0.0005 À0.5514
Phase-One 7.50
***
0.62 À1.09 0.23 À1.18 À32.46 0.45
1995 0.7808 0.0555 À0.0639 À0.0025 À0.9742
Phase-One 9.55
***
0.98 À1.54 À1.28 À2.45
***
À12.65 0.46
1990 0.9965 À0.1528 À0.0382 À0.0055 À1.045
Non-Phase One 10.61
***
À2.52
***
À0.75 À2.43
***
À2.73 À42.15 0.48
1995 0.8098 À0.0322 À0.1302 À0.0012 À0.7039
Non-Phase-One 8.79
***
À0.56 À2.72
***
À0.59 À1.86
*
À36.30 0.46
a
This table provides the regression results for the longitudinal pool (1990 and 1995) for Phase One and Non-Phase One plants.
Amounts in the tables are parameter and t-statistic values. Levels of signi?cance of p < 0.10, 0.05, and 0.01are denoted as *, **, and
***, respectively. The variable de?nitions are:
EI
w
The unadjusted data envelopment e?ciency score (within–group index) of the electricity plants. These indices lie between
zero and one.
REG Measures how favorably a state regulatory commission addresses rate increase requests.
STATE Measures the stringency of state environmentalism.
AGE The age of the plant.
MKTSHR The percentage of kilowatt hours of electricity produced by the operator of the plant in the NERC region in which the
plant operates.
574 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
ecoe?ciency and the Porter Hypothesis are sup-
ported by the statistical results.
DEA analysis using compliance variables
Pooled regressions
Table 10 gives the pooled regression results with
compliance variables. Regressing the DEA scores
on the compliance variables yields insights regard-
ing the measures taken to reduce pollution and
increase e?ciency. For the cross-sectional regres-
sions (Panel A) there are two variables that have
signi?cant negative coe?cients – for both 1990
and 1995: the average sulfur content of the petro-
leum used and the percentage of heat produced by
coal. The fact that these compliance variables
(which are associated with SO
2
releases) decrease
relative e?ciency as they increase in value is inter-
esting as it provides evidence that pollution is
indeed related to economic ine?ciency as claimed
by the Porter Hypothesis.
For the longitudinal regressions (Table 10,
Panel B), the same two variables (AVP and BTUC)
also emerge as negative and signi?cant for the Non-
Phase-One plants. These variables, however, are
not signi?cant for the Phase-One plants.
However, for the Phase-One plants scrubbers are
a signi?cant and positive variable. Thus, as scrub-
bers increase, relative e?ciency increases. Since
scrubbers reduce pollution, this again supports
the claim that as pollution decreases, e?ciency
increases.
Within-group regressions
Regressing the within-group indices on the
compliance variables also supplies some revealing
outcomes. Table 11 displays the results of these
regressions. In 1990, the Non-Phase-One plants
Table 9
Statistical analyses of adjusted e?ciency indices confounding variables
a
Cross-sectional analysis Distribution
b
Longitudinal analysis Distribution
b
Panel A: Adjusted EI
p
versus adjusted EI
w
1990 EI
p
7049 Phase-One EI
p
7591
1990 Phase-One EI
w
7147 1990 Phase-One EI
w
6605
Z-score À0.15 Z-score 1.56
1990 EI
p
7505 Phase-One EI
p
6899
1990 Non-Phase-One EI
w
6691 1995 Phase-One EI
w
7297
Z-score 1.28 Z-score À0.62
1995 EI
p
6701 Non-Phase-One EI
p
6787
1995 Phase-One EI
w
7495 1990 Non-Phase-One EI
w
7409
Z-score À1.25 Z-score À0.98
1995 EI
p
7524 Non-Phase-One EI
p
7409
1995 Non-Phase-One EI
w
6672 1995 Non-Phase-One EI
w
6787
Z-score 1.34 Z-score 0.98
Panel B: Adjusted EI
p
versus adjusted EI
p
1990 Phase One EI
p
6577 1990 Phase-One EI
p
6039
1990 Non-Phase One EI
p
7619 1995 Phase-One EI
p
8157
Z-score 1.65
*
Z-score 3.35
***
1995 Phase One EI
p
6464 1990 Non-Phase One EI
p
6453
1995 Non-Phase One EI
p
7732 1995 Non-Phase One EI
p
7743
Z-score 2.00
**
Z-score 2.04
**
a
This table provides the tests of relative e?ciency using the adjusted productive e?ciency scores when only confounding variables
are considered. Panel A reveals the presence of a common production function. As a consequence, the adjusted EI
p
indices are used to
conduct these tests and the results are reported in Panel B. Signi?cance levels of p < 0.10, p < 0.05, and p < 0.01 and are denoted as
*, **, and *** respectively.
b
For the cross-sectional hypotheses, the null hypotheses assume that the Phase-One plants are at least as e?cient as the Non-Phase-
One plants. For the longitudinal analyses, the null hypotheses assume that the 1990 plants are at least as e?cient as the 1995 plants.
Results re?ect the cumulative expected rank sum of each group and the Z-score when a Wilcoxon Rank Sum test is employed. All
Z-scores are one-sided.
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 575
continue to show signi?cant negative coe?cients
for petroleum sulfur and coal BTUs. Interestingly,
for Non-Phase-One plants the BTUC variable is
no longer signi?cant in 1995, perhaps indicating
a change in fuel mix for these plants, something
that was made possible by new blending fuel tech-
nology. For the Phase-One plants, the sulfur con-
tent of coal is a signi?cant negative coe?cient in
1990 and scrubbers produce a signi?cant positive
coe?cient. For 1995, scrubbers continue as a posi-
tive signi?cant coe?cient but the sulfur content of
coal is no longer signi?cant. The disappearance of
coal sulfur as a negative signi?cant factor from
1990 to 1995 is consistent with the new fuel blend-
ing technologies and other actions such as switch-
ing to lower sulfur coal. The persistence of
scrubbers as a signi?cant variable also signals their
role in achieving pollution reduction. Overall,
these results are consistent with the claims of the
Porter Hypothesis and ecoe?ciency.
Table 10
Pooled sample regression: compliance variables
a
Model Intercept AVC AVP BTUC SCR TYPE Log R
2
Model 1 (M1): EI
p
= b
0
+ b
1
AVC + b
2
AVP + b
3
BTUC + b
4
SCR + e
Model 2 (M2): EI
p
= b
0
+ b
1
AVC + b
2
AVP + b
3
BTUC + b
4
SCR + b
5
TYPE + e
Panel A: Cross-sectional pooled sample results (EI
p
), N = 168 (pooled by plant type)
M1: 1990 0.8123 0.2241 À11.9728 À0.2522 0.0393
14.82
***
0.095 À3.07
***
À3.63
***
1.08 À38.00 046
M2: 1990 0.8126 0.2913 À11.9672 À0.2516 0.0387 À0.0033
14.77
***
0.11 À3.00
***
À3.59
***
1.04 À0.07 À37.99 0.46
M1: 1995 0.7338 1.6398 À20.075 À0.1711 0.0280
13.18
***
0.651 À3.56
***
À2.61
***
0.35 À36.90 0.46
M2: 1995 0.7368 1.8609 À19.9024 À0.1591 0.0280 À0.0305
13.21
***
0.73 À3.54
***
À2.36
**
0.92 À0.72 À36.64 0.46
Model Intercept AVC AVP BTUC SCR PERMIT YEAR Log R
2
Model 1 (M1): EI
p
= b
0
+ b
1
AVC + b
2
AVP + b
3
BTUC + b
4
SCR + e
Model 2 (M2): EI
p
= b
0
+ b
1
AVC + b
2
AVP + b
3
BTUC + b
4
SCR + b
5
YEAR + e
Panel B: Longitudinal pooled sample results (EI
p
), N = 168 (pooled over years)
M1: 0.6192 À1.9423 À6.7058 À0.011 0.1092 À0.0289
Phase-One 5.85
***
À0.85 À1.04 À0.11 3.24
***
À0.70 À26.54 0.46
M2: 0.5647 À1.480 À5.413 À0.0127 0.1041 À0.0066 0.0511
Phase-One 4.92
***
À0.64 À0.83 À0.12 3.09
***
À0.14 1.18 À25.84 0.46
M1: 0.7451 4.992 À15.0425 À0.2182 À0.0421
Non-Phase-One 15.07
***
1.48 À3.28
***
À3.13
***
À1.09 À69.17 0.43
M2: 0.7217 5.0287 À14.7772 À0.2174 À0.0421 0.0434
Non-Phase-One 12.99
***
1.49 À3.22
***
À3.13
***
À1.09 0.90 À68.77 0.43
a
This table provides the regression results for the pooled regressions. Amounts in the tables are parameter and t-statistic values.
Levels of signi?cance of p < 0.10, 0.05, and 0.01 are denoted as *, **, and ***, respectively. The variable de?nitions are:
EI
p
The unadjusted data envelopment e?ciency score of the electricity plants (pooled indices). These indices lie between zero
and one.
AVC The average sulfur content of the coal used by the plant.
AVP The average sulfur content of the petroleum used by the plant.
BTUC The percent of heat (in BTUs) produced by the use of coal at the plant to produce kilowatt hours of electricity.
SCR The number of scrubbers at the plant.
TYPE Dummy variable equal to one if Phase One plant and zero if a Non-Phase One plant.
PERMIT Measures the e?ciency of the allocation and use of pollution permits (for Phase-One plants only).
YEAR Dummy variable equal to one if 1995 and zero if 1990.
576 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
Conclusions, limitations, and implications
This study investigates the relationships between
environmental performance and economic perfor-
mance in the electric utility industry in the United
States, where environmental performance is mea-
sured by levels of SO
2
and economic performance
is measured by relative e?ciency. The setting is of
particular interest because it is one where reduc-
tions were mandated without specifying how the
reductions were to be achieved, thus meeting the
requirements of the Porter Hypothesis. Under these
conditions, the Porter Hypothesis predicts that the
reductions can be achieved while maintaining or
improving economic e?ciency. Furthermore, given
the fact that a set of lower polluting plants was not
subject to the mandated reductions, it a?ords the
opportunity to see if ecoe?ciency is valid by com-
paring the lower polluting plants at a point in time
with the higher polluting plants (subject to the man-
date). The traditional view o?ers a competing
hypothesis which claims that any reductions in
SO
2
necessarily decrease economic e?ciency.
Adjusting for confounding variables, the evidence
indicates that lower polluting plants are relatively
more e?cient, both cross-sectionally and longitudi-
nally. Thus, the evidence is consistent with ecoe?-
ciency and the Porter Hypothesis. The support for
the Porter Hypothesis is reinforced by evidence con-
sistent with the use of new and innovative fuel
blending technologies to reduce SO
2
.
There are two very important limitations that
need to be mentioned. First, the study is for a
single industry and the results may not extend
beyond that industry. Second, we are measuring a
single aspect of environmental performance, albeit
a very important aspect. In this sense we duplicate
other studies that have attempted to assess the rela-
tionship between environmental performance and
economic performance. In our study, only a partial
environmental performance measure is used. Of
course, other studies look at di?erent measures of
environmental performance and it can be argued
that, collectively, the evidence may have more to
Table 11
Within-group sample regression: compliance variables
a
Model Intercept AVC AVP BTUC SCR Permit Log R
2
Within-group sample results (EI
w
) N = 168
Model 1 (M1): E
w
= b
0
+ b
1
AVC + b
2
AVP + b
3
BTUC + b
4
SCR + b
5
PERMIT + e
Model 2 (M2): E
w
= b
0
+ b
1
AVC + b
2
AVP + b
3
BTUC + b
4
SCR + e
Ml:1990 0.7541 0.7093 À7.525 À0.0909 0.1718 À0.0490
Phase-One 4.05
***
À1.83
*
À0.83 0.50 2.24
**
À0.80 À28.15 0.48
M2:1990 0.9863 7.7618 À16.70 À0.3623 À0.0890
Non-Phase-One 11.23
***
1.19 À2.46
***
À2.91
***
À1.22 À45.91 0.46
M1:1995 0.6483 À0.4049 À13.3757 À0.0131 0.0897 À0.0421
Phase-One 4.04
***
À0.11 À1.06 À0.08 2.08
**
À0.41 À15.96 0.46
M2:1995 0.7864 4.2128 À22.4140 À0.1455 À0.0236
Non-Phase-One 10.11
***
0.84 À2.70
***
À1.39 À0.42 À40.33 0.43
a
This table provides the regression results for the within-group regressions. Amounts in the tables are parameter and t-statistic
values. Levels of signi?cance of p < 0.10, 0.05, and 0.01 are denoted as *, **, and ***, respectively. The variable de?nitions are:
EI
w
The unadjusted data envelopment e?ciency score (within–group index) of the electricity plants. These indices lie between
zero and one.
MKTSHR The percentage of kilowatt hours of electricity produced by the operator of the plant in the NERC region in which the plant
operates.
AVC The average sulfur content of the coal used by the plant.
AVP The average sulfur content of the petroleum used by the plant.
BTUC The percent of heat (in BTUs) produced by the use of coal at the plant to produce kilowatt hours of electricity.
SCR The number of scrubbers at the plant.
PERMIT Measures the e?ciency of the allocation and use of pollution permits. In 1990 and 1995, permits were allocated only to
Phase One plants.
YEAR Dummy variable equal to one if 1995 and zero if 1990.
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 577
say than any individual study. Thus, we add to
this collective body further evidence of a positive
relation between environmental performance and
economic performance. Moreover, there is some
merit to arguing that a single very important mea-
sure of environmental degradation may be a good
proxy for the environmental performance of an
industry. Nonetheless, developing a more compre-
hensive measure of environmental performance
would be a signi?cant contribution, something per-
haps future research will produce.
As the evidence builds for ecoe?ciency, the
incentive increases for management to obtain
and use more information about environmental
activities and costs. This implies an increase in the
likelihood that management will demand better
environmental accounting, including the imple-
mentation of formal environmental management
accounting systems. After all, if the possibility exists
that paying more attention to environmental per-
formance will create more opportunities to increase
pro?ts, then a demand for environmental manage-
ment accounting systems seems logical. Indeed,
these systems are needed to facilitate management
decisions and corporate accountability (Bartolo-
meo et al., 2000; Epstein, Flamholtz, & McDon-
ough, 1976) and to aid in both external and
internal decision making (Bouma & Kamp-Roe-
lands, 2000; Everett & Neu, 2000; Gray, 2002).
However, while this study, other empirical studies,
and a number of case studies are supportive of eco-
e?ciency (voluntary or guided), the empirical evi-
dence is in its infancy and more studies are needed.
Like previous endeavors that delve into emerg-
ing cost accounting research agendas, investiga-
tions of the behavior and impact of environmental
costs have a long way to go. For instance, it is pos-
sible that resistance to the Porter view that pollu-
tion is a form of economic ine?ciency is similar in
nature to the much earlier Deming view that poor
product quality was a form of economic ine?-
ciency. While it is nowwidely accepted that it is pos-
sible to simultaneously improve product quality
and reduce quality costs, this view was not readily
accepted in the beginning. Evidence had to accumu-
late. Furthermore, more theoretical understanding
of the ecoe?ciency paradigm would also be useful.
For example, it would be interesting to model and
then empirically investigate the predictions of a bet-
ter-speci?ed environmental cost function – one that
speci?cally recognizes the presence of ecoe?cient
behavior. In addition, there may be value in explor-
ing how and when management decides to make a
commitment to improve the environment. For
instance, it may be that the bounded rationality
assumed by Porter is attributable mostly to many
managers being committed to the traditional envi-
ronmental cost paradigm. If ecoe?ciency is valid,
as evidence is beginning to suggest, a paradigmshift
may be needed to bring managers to a point where
they actively engage in actions that improve the
environment.
Acknowledgements
The authors would like to thank William F.
Bowlin, Marc Epstein, Kevin Hughes, Michael
Stein, and Annie McGowan for their informative
comments on previous drafts. The authors would
like to express a special note of thanks to the En-
ergy Information Administration of the Depart-
ment of Energy for its assistance procuring data
for this project. Finally, Royce D. Burnett is espe-
cially grateful for the ?nancial support provided
by the AIPCA Minority Doctoral Fellowship Pro-
gram and the PhD project.
References
Al-Tuwaijri, A. S., Christensen, T. E., & Hughes, K. E. (2004).
The relations among environmental disclosure, environ-
mental performance, and economic performance: a simul-
taneous equations approach. Accounting, Organizations and
Society, 29, 447–471.
Athanassopoulos, A. D., Lambroukos, N., & Sieford, L.
(1999). Data envelopment scenario analysis for setting
targets to electricity generating plants. European Journal of
Operations Research, 115, 413–428.
Banker, R. D., Charnes, A., & Cooper, W. W. (1984). Some
models for estimating technical and scale e?ciencies in data
envelopment analysis. Management Science, 30, 1078–1092.
Bebbington, J. (2001). Sustainable development: a review of the
international development, business and accounting litera-
ture. Accounting Forum, 25, 128–157.
Bartolomeo, M., Bennett, M., Bouma, J., Heydkamp, P.,
James, P., & Wolters, T. (2000). Environmental manage-
ment accounting in europe: current practice and future
potential. The European Accounting Review, 9(1), 31–52.
578 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
Bebbington, J., Gray, R., & Owen, D. (1999). Seeing the wood
for the trees: taking the pulse of social and environmental
accounting. Accounting, Auditing, and Accountability Jour-
nal, 12(1), 47–51.
Beets, S. D., & Souther, C. C. (1999). Corporate environmental
reports: the need for standards and environmental assurance
service. Accounting Horizons, 13(June), 129–145.
Berthelot, S., Cormier, D., & Magnan, M. (2003). Environ-
mental disclosure research: review and synthesis. Journal of
Accounting Literature, 22, 1–44.
Bhat, V. N. (1996). The green corporation: The next competitive
advantage. Westport, CT: Quorum Books.
Birkin, F., & Woodward, D. (1997a). Management accounting
for sustainable development Part 5: accounting for sustain-
able development. Management Accounting (UK), 75(10),
52–54.
Birkin, F., & Woodward, D. (1997b). Management accounting
for sustainable development Part 2: from economic to
ecological e?ciency. Management Accounting (UK), 75(7),
42–45.
Blacconiere, W. G., Johnson, M. F., & Johnson, M. S. (1997).
Stranded costs and competitive forces in the electric utility
industry. Journal of Accounting, Auditing, and Finance,
12(Summer), 199–221.
Bossong, E. M. (1999). Electric Utilities. Standard and Poors
Industry Survey, 167(32), 1–38.
Bouma, J., & Kamp-Roelands, N. (2000). Stakeholders’ expec-
tations of an environmental management system: some
exploratory research. The European Accounting Review,
9(1), 131–144.
Callen, J. L. (1991). Data envelopment analysis: partial survey
and applications for management accounting. Journal of
Management Accounting Research, 3(Fall), 35–55.
Charnes, A., Cooper, WW., & Rhodes, E. (1978). Measuring
the e?ciency of decision making units. European Journal of
Operation Research(2), 249–444.
Christensen, L. R., & Greene, W. H. (1976). Economies of scale
in U.S. electric power generation. Journal of Political
Economy, 84, 655–676.
Claggert, E. T., & Ferrier, G. D. (1998). The e?ciency of TVA
power distributors. Managerial and Decision Economics, 19,
365–378.
Cooper, W. W., Seiford, L., & Tone, K. (2000). Data
envelopment analysis: A comprehensive text with models,
applications, references and DEA-solver software. Kluwer
Boston MA: Academic Press.
Cowing, T. G., & Stevenson, R. E. (1981). Introduction:
productivity measurement in regulated industries. In T. G.
Cowing & R. E. Stevenson (Eds.), Productivity measurement
in regulated industries (pp. 3–15). New York, NY: Academic
Press.
De Villiers, C., & Van Staden, C. J. (2006). Can less
environmental disclosure have a legitimizing e?ect? Evi-
dence from Africa. Accounting, Organizations and Society,
31, 763–781.
Deegan, Craig (2002). Introduction: the legitimizing e?ect of
social and environmental disclosures – a theoretical foun-
dation. Accounting, Auditing, and Accountability Journal,
15(3), 282–311.
DeSimone, L. D., & Popo?, F. (1997). Eco-e?ciency: The
business link to sustainable development. MIT press.
Dierkes, M., & Preston, L. (1977). Corporate social accounting
reporting for the physical environment: A critical review
and implementation proposal. Accounting, Organizations
and Society, 2(1), 3–22.
Ditz, D., Ranganathan, J., & Banks, R. D. (1995). Green
ledgers: Case studies in corporate environmental accounting.
World Resources Institute.
D’Souza, J., Jacob, J., & Soderstrom, N. S. (2000). Nuclear
decommissioning costs: The impact of recoverability risk on
valuation. Journal of Accounting and Economics, 29, 207–
230.
Environmental Protection Agency. (1996). 1995 Acid Rain
Compliance Report: Overview. Washington, DC.
Environmental Protection Agency. (2000). Progress Report on
the EPA Acid Rain Program. Washington, DC.
Epstein, M., Flamholtz, E., & McDonough, J. (1976). Corpo-
rate social accounting in the United States of America: state
of the art and future prospects. Accounting, Organizations
and Society, 1(1), 23–42.
Epstein, M. J. (1996). Measuring corporate environmental
performance: Best practices for costing and managing an
e?ective environmental strategy. Chicago, IL: Irwin.
Everett, J., & Neu, D. (2000). Ecological modernization and the
limits of environmental accounting. Accounting Forum,
24(1), 5–29.
Fare, R. Grosskopf, & Tyteca, D. (1996). An activity analysis
model of the environmental performance of ?rms – appli-
cation to fossil-fuel-?red electric utilities. Ecological Eco-
nomics, 18, 161–175.
Farrell, M. J. (1957). The measurement of productive e?ciency.
Journal of Royal Statistical Society, 120(Series A, Part 3),
253–281.
Feldman, S. J., Soyka, P., & Ameer, P. (1997). Does improving
a ?rm’s environmental management system and environ-
mental performance result in a higher stock price? The
Journal of Investing(Winter), 87–97.
Fekrat, M., Inclan, I., & Petroni, D. (1996). Corporate
environmental disclosures: Comprehensive disclosure
hypothesis using 1991 annual report data. The International
Journal of Accounting, 175–195.
Fizel, J. L., & Nunnikhoven, T. S. (1992). Technical e?ciency
of for-pro?t and non-pro?t nursing homes. Managerial and
Decision Economics, 13(5), 429–439.
Freedman, M., & Wasley, C. (1990). The association between
environmental performance and environmental disclosure in
annual reports and 10ks. Advances in Public Interest
Accounting, 183–193.
Freedman, M., & Jaggi, B. (1992). An investigation of the long-
run relationship between pollution performance and eco-
nomic performance: The case of pulp and paper ?rms.
Critical Perspectives on Accounting, 3, 315–336.
Freedman, M., & Jaggi, B. (1994). analysis of the association
between pollution performance and input cost factors: the
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 579
case of electric utility plants. Journal of Accounting and
Public Policy, 13, 31–48.
Golany, B., Roll, Y., & Rybak, D. (1994). Measuring e?ciency
of power plants in Israel by data envelopment analysis.
IEEE Transactions on Engineering Management, 41(3),
291–301.
Goto, M., & TsuTsui, M. (1998). Comparison of productive
and cost e?ciencies among Japanese and US electric
utilities. Omega, International Journal of Management Sci-
ence, 26, 177–194.
Gray, R. H. (1992). Accounting and environmentalism: an
exploration of the challenge of gently accounting for
accountability, transparency, and sustainability. Account-
ing, Organizations and Society, 17, 399–425.
Gray, R., Kouhy, R., & Lavers, S. (1995). Corporate social and
environmental reporting: a review of the literature and a
longitudinal study of UK disclosure. Accounting, Auditing,
and Accountability Journal, 8(2), 47–77.
Gray, R., & Bebbington, J. (2000). Environmental accounting
managerialism and sustainability: is the planet safe in the
hands of business and accounting? Advances in Environ-
mental Accounting and Management(1), 1–44.
Gray, R. (2002). The social accounting project and accounting,
organizations and society privileging engagement, imagin-
ings, new accountings and pragmatism over critique?
Accounting, Organizations and Society, 27, 687–708.
Grosskopf, S., & Valdmanis, V. (1987). Measuring hospital
performance: a non-parametric approach. Journal of Health
Economics, 6, 89–107.
Hackston, D., & Milne, M. J. (1996). Some determinants of
social and environmental disclosures in New Zealnad
companies. Accounting, Auditing, and Accountability Jour-
nal, 77–108.
Hagg, I., & Hedlund, G. (1979). Case studies in accounting
research. Accounting, Organizations, and Society, 4(1/2),
135–143.
Haeri, H., Khawaia, M. S., & Perussi, M. (1997). Competitive
e?ciency: do mergers and ‘‘critical mass’’ really make a
di?erence? Public Utilities Forthnightly(June15), 26–33.
Haeri, H., Perussi, M., & Khawaja, M. S. (1999). The 100
fortnightly revisited: do utility stock prices re?ect operational
e?ciency? Public Utilities Fortnightly(February 15), 36–41.
Herbohn, K. (2005). Afull cost environmental accounting exper-
iment. Accounting, Organizations and Society, 30, 519–536.
Hughes, K. E. (2000). The value relevance of non?nancial
measures of air pollution in the electric utility industry. The
Accounting Review, 75(April), 209–228.
Hughes, S., Anderson, S., & Golden, S. (2001). Corporate
environmental disclosures: are they useful in determining
environmental performance? Journal of Accounting and
Public Policy, 3(20), 217–240.
Ilinitch, A. Y., Soderstrom, N. S., & Thomas, T. E. (1998).
Measuring corporate environmental performance. Journal
of Accounting and Public Policy, 17, 383–408.
Ingram, R., & Frazier, K. (1980). Environmental performance
and corporate disclosure. Journal of Accounting Research,
113–144.
Joshi, S., Krishnan, R., & Lave, L. (2001). Estimating the
hidden costs of environmental regulation. The Accounting
Review, 26(2), 171–198.
Jonson, L., Jonson, B., & Svensson, G. (1978). The applica-
tion of social accounting to absentism and personnel
turnover. Accounting, Organizations, and Society, 3(3/4),
261–268.
Juskow, P. L., & Schmalaensee, R. (1987). The performance of
coal-burning electric generating units in the United States,
1960 – 1980. Journal of Applied Economics, 2(April), 85–109.
Kaplan, R. S. (1986). The role for empirical testing in
management accounting. Accounting, Organizations, and
Society, 11(4/5), 429–452.
King, A., & Lenox, M. (2002). Exploring the locus of pro?table
pollution reduction. Management Science, 48(2 February),
289–299.
Knutson, K. (1995). Trends in utility fuel cost. Public Utility
Fortnightly(April 15), 12–13.
KPMG International International Survey of Corporate Sus-
tainability. (2005). , as of 20.1.2007.
Lanen, W. (1999). Waste minimization at 3M company: a ?led
study of non?nancial performance measurement. Journal of
Management Accounting Research, 11, 29–43.
Lehman, G. (2002). Global accountability and sustainability:
research prospects. Accounting Forum, 26(3), 219–232.
Lillis, A., & Mundy, J. (2005). Cross-sectional ?eld studies in
management accounting research – closing the gaps between
surveys and case studies. Journal of Management Accounting
Research, 17, 119–141.
Loudder, M., Khurana, I., & Boatsman, J. (1996). Market
valuation of regulatory assets in public utility ?rms. The
Accounting Review, 71, 161–182.
Majumadar, S. K. (1998). On the utilization of resources:
perspectives from the U.S. telecommunications industry.
Strategic Management Journal, 19, 809–831.
Margolis, J., & Walsh, J. (2003). Miersy loves companies:
rethinking social initiatives by business. Administrative
Science Quarterly, 48, 268–305.
Mathews, M. (1997). Twenty-?ve years of social and environ-
mental research: is there a silver jubilee to celebrate?
Accounting, Auditing, & Accountability Journal, 10(4),
481–531.
Meyer, S. (1992). The economic impact of environmental
regulation. Working Paper. Massachusetts Institute of
Technology.
Moneva, J., Archel, P., & Correa, C. (2006). GRI and the
camou?aging of corporate unsustainability. Accounting
Forum, 30, 121–137.
Murty, M. N., & Kumar, S. (2003). Win–win opportunities and
environmental regulation: testing of the porter hypothesis
for Indian manufacturing industries. Journal of Environ-
mental Management, 67, 139–144.
Neu, D., Warsame, H., & Pedwell., K. (1998). Managing public
impressions: environmental disclosures in annual reports.
Accounting, Organizations and Society, 23(3), 265–282.
580 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
Olatubi, W. O., & Dismukes, D. E. (2000). A data envelopment
analysis of the levels and determinants of coal ?red electric
power generating performance. Utilities Policy, 9, 47–59.
Patten, D. (1992). Intra-industry environmental disclosures in
response to the Alaskan oil spill: a note on legitimacy
theory. Accounting, Organizations, and Society, 471–475.
Patten, D. (1991). Exposure, legitimacy, and social disclosure.
Journal of Accounting and Public Policy, 10, 297–308.
Patten, D. (2002). The relation between environmental perfor-
mance and environmental disclosure. Accounting, Organi-
zations, and Society, 27, 763–773.
Pelz, E., & Fitzgerald, T. J. (2001). From market failure to
market-based solution: policy lessons from clean air legis-
lation. Economic Commentary.
Porter, M. E. (1991). America’s green strategy. Scienti?c
American(April), 168.
Porter, M. E., & van der Linde, C. (1995a). Toward a new
conception of the environment-competitiveness relation-
ship. Journal of Economic Perspectives, 9(Fall), 97–118.
Porter, M. E., & van der Linde, C. (1995b). Green and
competitive: ending the stalemate. Harvard Business
Review(September–October), 120–134.
Puig-Junoy, J. (1998). Technical e?ciency in the clinical
management of critically ill patients. Health Economics, 7,
263–277.
Rockness, J. (1985). An assessment of the relationship between
U.S. corporate environmental performance and disclosure.
Journal of Business Finance and Accounting, 12, 339–354.
Rubenstein, D. (1992). Bridging the gap between green
accounting and black ink. Accounting, Organizations and
Society, 17(5), 501–508.
Ruch, D., & Roper, J. J. (1992). The greening of corporate
Canada. CMA Magazine(December–January), 15–18.
Savage, D. E., Jasch, C., & International Federation of
Accountants, 2005. International guidance document: envi-
ronmental management accounting. 28.2.2007.
Schmidheiny, S., & Zorraquin, F. J. L. (1996). Financing
change. MIT Press.
Shank, J. K., & Govindarajan, V. (1994). Measuring the cost of
quality: a strategic cost management perspective. Journal of
Cost Management(Summer), 5–17.
Solomons, D. (1991). Accounting and social change: a neutralist
view. Accounting, Organizations and Society, 16(3), 287–295.
Stone, D. (1995). No longer at the end of the pipe, but still a long
way fromsustainability: a look at management accounting for
the environment and sustainable development in the United
States. Accounting Forum, 19(2/3), 95–110.
Swift, B. (2001). How environmental laws work: an analysis of
the utility sector’s response to regulation of nitrogen oxides
and sulfur dioxides under the clean air act. Tulane Environ-
mental Law Journal, 14(Summer), 309–424.
Tyteca, D. (1996). On the measurement of the environmental
performance of ?rms – a literature review and a productive
e?ciency perspective. Journal of Environmental Manage-
ment, 46, 281–308.
Tyteca, D. (1997). Linear programming models for the
measurement of environmental performance of ?rms –
concepts and empirical results. Journal of Productivity
Analysis, 8, 183–197.
Tyteca, D. (1999). Sustainability indicators at the ?rm level.
Journal of Industrial Ecology, 2, 61–77.
Ullman, A. (1976). The corporate environmental accounting
system: a management tool for ?ghting environmental
degradation. Accounting, Organizations and Society, 1(1),
71–79.
United Nations Division for Sustainable Development. (2001).
Environmental management accounting, procedures and prin-
ciples. accessed 9.3.2007.
United States Department of Energy, Energy Information
Administration (EIA). (1994). Electric utility phase I acid
rain compliance strategies for the clean air act amendments
of 1990. DOE/EIA-0578(94), Washington, DC.
United States Department of Energy, Energy Information
Administration (EIA). (1997). The e?ects of title IV of the
clean air act amendments of 1990 on electric utilities: an
update, DOE/EIA-0582(97). Washington, DC.
Walden, W. D., & Schwartz, B. N. (1997). Environmental
disclosures and public policy pressure. Journal of Accounting
and Public Policy, 16(2), 125–1554.
White, G. (2005). How to report a company’s sustainability
activities. Management Accounting Quarterly, 7(1), 36–43.
Whiteman, J. L. (1995). Benchmarking developing country
electricity systems using data envelopment analysis. Asia-
Paci?c Economic Review, 1(3), 71–78.
Wildavsky, Aaron (1994). Accounting for the environment.
Accounting, Organizations and Society, 19(4/5), 461–481.
Wiseman, Joanne (1982). An evaluation of environmental
disclosures made in corporate annual reports. Accounting,
Organizations and Society, 7(1), 53–63.
Wong, E. A. (2001). How to develop a database for forecasting
environmental expenditures. Strategic Finance(March),
52–56.
World Business Council for Sustainable Development. (2000a).
Eco-e?ciency: creating more value with less input. North
Yorkshire, UK.
World Business Council for Sustainable Development. (2000b).
Measuring eco-e?ciency: a guide to reporting company
performance. North Yorkshire, UK.
Yaisawarng, S., & Klein, J. D. (1994). The e?ects of sulfur
dioxide controls on productivity changes in the U.S. electric
power industry. The Review of Economics and Statistics,
76(August), 447–460.
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 581
doc_137172020.pdf
This study investigates the relationship between environmental performance and productive efficiency in the United
States electric utility industry before and after the 1990 Clean Air Act Amendments. Using Data Envelopment Analysis
(DEA), cross-sectional examinations reveal lower polluting plants are more efficient than higher polluting plants. Longitudinal
analyses indicate plants can simultaneously reduce pollution and increase relative efficiency
Ecoe?ciency: De?ning a role for environmental
cost management
Royce D. Burnett
a,
*
, Don R. Hansen
b,1
a
The University of Miami, Department of Accounting, 316 Kosar/Epstein Faculty O?ce Wing,
5250 University Drive, Coral Gables, FL 33146, United States
b
Oklahoma State University, William S. Spears School of Business Administration, School of Accounting,
Room 401, Stillwater, OK 74075, United States
Abstract
This study investigates the relationship between environmental performance and productive e?ciency in the United
States electric utility industry before and after the 1990 Clean Air Act Amendments. Using Data Envelopment Analysis
(DEA), cross-sectional examinations reveal lower polluting plants are more e?cient than higher polluting plants. Lon-
gitudinal analyses indicate plants can simultaneously reduce pollution and increase relative e?ciency. Collectively, these
results are evidence that proactive environmental management can reduce environmental costs and thus, lends support
for adopting an environmental cost management system.
Ó 2007 Elsevier Ltd. All rights reserved.
Introduction
Concern for taking care of the environment is
increasing as is interest in accounting for the envi-
ronment (Beets & Souther, 1999; Deegan, 2002;
Gray, Kouhy, & Lavers, 1995; Mathews, 1997).
The response from the business community has
been to gather and report more information about
environmental activities to stakeholders. For
instance, KMPG (2005) indicates that 55 percent
of the Fortune top global ?rms are now issuing
sustainability reports and that the number of top
100 ?rms issuing such reports within each country
surveyed has tripled since 1993. Although the con-
tent and quality of the reports vary widely, the
KPMG survey supports the view that environmen-
tal accounting and reporting is indeed on the rise
and is becoming an integral part of the information
released to shareholders. Moreover, the survey sug-
gests that all stakeholders – internal and external –
bene?t fromthese reports because they focus on the
risk and opportunities associated with corporate
0361-3682/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aos.2007.06.002
*
Corresponding author. Tel.: +1 305 284 8618; fax: +1 305
284 5737.
E-mail addresses: [email protected] (R.D. Burnett),
[email protected] (D.R. Hansen).
1
Tel.: +1 405 744 8626; fax: +1 405 744 1680.
www.elsevier.com/locate/aos
Available online at www.sciencedirect.com
Accounting, Organizations and Society 33 (2008) 551–581
social responsibility. White (2005) indicates that
producing these reports is valuable because they
provide a framework for linking economic, envi-
ronmental and social decision making to strategy.
To date, research studies in environmental
accounting have explored issues relating to envi-
ronmental disclosure such as the characteristics of
disclosers (e.g., Gray et al., 1995; Hackston &
Milne, 1996; Patten, 1992), what is being disclosed
(e.g., De Villiers & Van Staden, 2006; Wiseman,
1982), what should be disclosed (e.g., Dierkes &
Preston, 1977; Deegan, 2002; Moneva, Archel, &
Correa, 2006), and why voluntary disclosures are
made (e.g., Beets & Souther, 1999; Neu, Warsame,
& Pedwell., 1998; Walden & Schwartz, 1997).
Another line of research in environmental account-
ing concerns the relationships among environmen-
tal disclosure, environmental performance, and
economic performance. Studies have examined all
three pair-wise relationships (Berthelot, Cormier,
& Magnan, 2003; Margolis & Walsh, 2003). Much
of the research related to environmental account-
ing in this area has targeted the environmental dis-
closure–environmental performance pair (Fekrat,
Inclan, & Petroni, 1996; Freedman & Wasley,
1990; Hughes, Anderson, & Golden, 2001; Ingram
& Frazier, 1980; Rockness, 1985; Wiseman, 1982).
Patten (2002) notes that evidence concerning this
particular relationship is of special interest because
of socio-political theories concerning environmen-
tal disclosure. Patten (1991) suggests that while
these theories implicitly assume businesses are
responsible for using resources to engage in activi-
ties designed to increase pro?ts, they also call for
an awareness of the adverse consequences that
the pursuit of capitalism has on the environment.
In a similar vein, there is increasing interest con-
cerning the relationship between environmental per-
formance and economic performance because of an
emerging socio-economic theory known as ecoe?-
ciency (Birkin & Woodward, 1997a, 1997b). Ecoef-
?ciency claims that it is possible to increase
productivity and thus reduce costs while simulta-
neously improving environmental performance
(Bebbington, 2001; Lehman, 2002; Stone, 1995).
The ecoe?ciency paradigm has signi?cant implica-
tions for environmental accounting and, in particu-
lar, for environmental management accounting
(EMA). Speci?cally, if ecoe?ciency is valid, then
the likelihood will increase that companies will be
willing to establish formal EMA control systems
that isolate and quantify the costs, bene?ts, and
operational outcomes of proactive environmental
management. In fact, according to the International
Federation of Accountants’ Statement Management
Accounting Concepts, EMA is ‘‘the management of
environmental and economic performance through
the development of appropriate environment-
related accounting systems and practices.’’ (Savage
&Jasch, 2005, p. 19). Indeed, Walden and Schwartz
(1997) indicate that creating and maintaining envi-
ronmental measurement and reporting systems is
an appropriate response to the social, political and
economic pressures surrounding the use of natural
resources. Ilinitch, Soderstrom, and Thomas
(1998) agree and indicate that environmental man-
agement e?ectiveness depends on an output-based
approach that, in part, relies on an internal assess-
ment of how well an organization uses its resources
to gain competitive advantage. Thus, for an EMA
system, the environmental-economic performance
pair is the most important and critical relationship.
The purpose of this paper is to provide empirical
evidence concerning ecoe?ciency in the electric util-
ity industry by examining the relationship between
environmental performance and economic perfor-
mance, where environmental performance is mea-
sured (proxied) by levels of SO
2
and economic
performance is measured by productive e?ciency.
Admittedly, SO
2
is not a comprehensive measure
of environmental performance but it is a very
important pollutant for the industry and therefore
is likely to be a useful proxy for environmental per-
formance.
2
Thus, knowing the nature of this rela-
2
Ilinitch et al. (1998) indicate that de?ning corporate
environmental performance is not a straightforward task and
as such, any one performance indicator, or even an entire
category of indicators, would likely be insu?cient to gauge a
company’s environmental performance with con?dence. Lanen
(1999), in a case study of the assessment of environmental
performance and waste management, suggests that one way to
gauge the relevance of any performance measure is to assess
whether actions adopted result in improvements in the measure
of interest. Given this insight, as well as the intent of the 1990
Clean Air Act Amendments, SO
2
emissions appear to be an
appropriate proxy for environmental performance in this study.
552 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
tionship ought to be of signi?cant interest to man-
agement, accountants, shareholders, regulators,
and other stakeholders. For example, if the same
or more good (intended) output is produced using
fewer inputs while simultaneously reducing pollu-
tion, then costs decrease and overall environmental
and ?nancial performance improves. Such con?r-
matory evidence would support ecoe?ciency and
provide a positive incentive for management to
invest in an environmental cost management system
as part of an agenda to reduce pollution.
Con?rmatory evidence would also a?ect disclo-
sure practices and investor decisions. If managers
believe that good environmental performance is
positively associated with good economic perfor-
mance, then this would tend to be viewed as good
news by investors and managers would have an
incentive to disclose more information about envi-
ronmental performance (Al-Tuwaijri, Christensen,
& Hughes, 2004). Moreover, the signal would be
that investing in ‘‘green’’ companies has a sound
economic rationale. Indeed, Feldman, Soyka,
and Ameer (1997) suggest that improving a com-
pany’s environmental management may result in
improved environmental performance, and that
improved environmental performance is compati-
ble with superior ?nancial performance. Further,
they posit that environmental improvements may
be viewed by the ?nancial community as a reduc-
tion in the overall risk of the ?rm resulting in a
lower cost of capital and an increase in stock price.
Finally, evidence supporting ecoe?ciency would
imply that the principal role of regulatory
approaches is to mandate reductions in pollution,
leaving management free to innovate and ?nd
ways to simultaneously improve e?ciency, reduce
costs, and achieve the mandated reductions. Of
course, if the empirical evidence does not support
ecoe?ciency, this also has implications, but with
opposing directional e?ects.
The outcomes produced in reaction to the Uni-
ted States Federal 1990 Clean Air Act Amend-
ments (1990 CAAA) provide the opportunity to
examine these claims. The 1990 CAAA mandated
a speci?c subset of plants in the electric utility
industry participate in a ‘‘group’’ reduction of pol-
lution within a ?ve-year period without specifying
how the reductions were to be achieved. Fortui-
tously, plant-level data are publicly available and
so this setting a?ords a unique and direct opportu-
nity to evaluate ecoe?ciency by examining the
changes in plant-level productive e?ciency from
1990 to 1995 under a scenario that is devoid of spe-
ci?c operational mandates. At issue is not the time
period of the regulatory act and the associated
plant data but the conditions speci?ed by the Act
and whether management can act under these con-
ditions to pursue process innovations and
improvements to achieve pollution reduction while
maintaining or increasing e?ciency.
The 1990 CAAA identi?es two groups of
plants: (1) Phase-One plants, de?ned as higher pol-
luting plants subject to an overall requirement to
reduce pollution by 1995 (the beginning of
Phase-One); and, (2) Non-Phase-One plants,
de?ned as lower polluting plants exempt from
any pollution reduction requirements for the
Phase-One period. The 1990 and 1995 Non-
Phase-One plants serve as control groups for
cross-sectional analyses while the 1990 Phase-
One plants de?ne a natural longitudinal control
group to assess the e?ect on plant-level e?ciency
as pollution is reduced because of the 1990 CAAA
intervention.
The approaches taken and the results obtained
are as follows. Initially, raw or unadjusted Data
Envelopment Analysis (DEA) scores are used to
measure relative e?ciency on both a cross-sec-
tional and longitudinal basis. For cross-sectional
analysis, we compare the e?ciency of the Non-
Phase-One (lower polluting) plants with that of
the Phase-One (higher polluting) plants in 1990
and 1995 to assess whether the lower polluting
plants are relatively more e?cient than the higher
polluting plants. Using unadjusted DEA scores,
we ?nd that the group of lower polluting plants
is relatively less e?cient than the higher polluting
plants in 1990, an outcome inconsistent with ecoef-
?cient behavior. The 1995 results, however, favor
the Non-Phase-One plants. We then evaluate the
e?ect of intervention (the 1990 CAAA) by investi-
gating changes in the relative e?ciency of both
Phase-One and Non-Phase-One plants between
1990 and 1995. We ?nd that by 1995 the Phase-
One plants have simultaneously decreased pollu-
tion and increased their relative e?ciency. During
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 553
the same time period, we also ?nd that the Non-
Phase-One plants did not reduce pollution but
did show an increase in relative e?ciency.
Next, the DEA scores are adjusted for the
e?ects of potential confounding variables. Then,
the cross-sectional and longitudinal analyses are
repeated using the adjusted e?ciency scores.
Results obtained indicate that the lower polluting
plants are relatively more e?cient than the higher
polluting plants for both 1990 and 1995. Further-
more, the longitudinal results show that the 1995
plants are more e?cient than the 1990 plants.
Finally, we provide evidence that the increases
in e?ciency are associated with such compliance
actions as fuel switching, new fuel mix technology
and the use of scrubbers. This evidence is impor-
tant because it provides insight as to which opera-
tional options are e?ective when developing a
pollution reduction compliance strategy. Thus,
from an overall perspective, our results favor eco-
e?ciency and thus support a positive relationship
between environmental performance and eco-
nomic performance. Accordingly, a strong argu-
ment can be made for the role for an EMA
system as envisioned by the International Federa-
tion of Accountants and others.
We organize the remainder of the paper as fol-
lows. Section ‘‘Relevant literature, motivation and
empirical predictions’’ discusses the relevant litera-
ture and develops the empirical predictions for the
study. Section ‘‘Research design’’ explains the
research design and Section ‘‘Empirical results’’
provides the empirical results. The ?nal section
o?ers some concluding remarks.
Relevant literature, motivation and empirical
predictions
Environmental performance, economic performance,
and ecoe?ciency
The traditional view concerning the relationship
of environmental performance and economic per-
formance argues that improving environmental
performance inevitably raises costs and reduces
productivity (Porter, 1991). According to the tra-
ditional view of environmental economics, envi-
ronmental degradation is a negative output and
decreasing negative output requires additional
inputs (increasing environmental costs).Thus, any
improvement cannot increase economic e?ciency
which implies a win–lose paradigm. Accordingly,
the traditional hypothesis assumes that improving
environmental performance is costly and requires
regulatory intervention to bring about any envi-
ronmental improvements. However, several
researchers have taken exception to this assumed
relationship between environmental improvements
and cost, arguing that pollution is a form of eco-
nomic ine?ciency and, thus, reductions in pollu-
tion actually increase productive e?ciency and
thereby reduce costs (King & Lenox, 2002; Porter,
1991; Porter & van der Linde, 1995a, 1995b). This
competing hypothesis challenges the traditional
model of pollution control and cost management,
re?ects an underlying win–win paradigm, and is
referred to as ecoe?ciency. Managers exhibit eco-
e?cient behavior when they adopt innovative and
creative methods to produce more or the same
level of useful goods and services while simulta-
neously reducing environmental degradation,
resource consumption, and costs (WBCSD,
2000a, 2000b). Thus, the voluntary ecoe?ciency
hypothesis maintains that fully rational managers
will voluntarily engage in an unfettered or pure
form of ecoe?cient behavior. The two competing
hypotheses have the same premise (improving
environmental performance) but have totally
opposite consequences.
While agreeing with the ecoe?ciency paradigm,
Porter (1991) and Porter and van der Linde
(1995a, 1995b) are skeptical that managers will
voluntarily engage in ecoe?cient behavior. They
hypothesize that agents within ?rms generally have
bounded rationality relative to ecoe?ciency and
consequently must be encouraged or stimulated
to engage in ecoe?cient behavior by properly
designed regulatory intervention. It seems curious
that in competitive markets, managers need to be
encouraged to adopt ecoe?ciency (if valid); yet
one possible explanation is that managers are ?x-
ating on the traditional paradigm. Indeed, para-
digm shifts often need some stimulus. Market
forces are one possible stimulus and can often be
very punishing, as evidenced by the experiences
554 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
?rms and industries had in shifting from the tradi-
tional view of quality to total quality management
(Shank & Govindarajan, 1994). In transitioning
from the traditional to the new ecoe?cient para-
digm, it would not be unusual to see some ?rms
experimenting with ecoe?cient investments, while
most do not. Over time, if the ecoe?cient experi-
ments prove fruitful, surviving ?rms will all be
subscribers to the ecoe?cient paradigm.
Properly designed regulatory intervention is
another possible stimulus, and, if timely, may avoid
some of the punishing e?ects of the competitive
market by reducing the uncertainty of investing in
technology and producing an equal playing ?eld
for change (Porter & van der Linde, 1995a). Prop-
erly designed regulations have two components.
First, they focus on environmental outcomes and
not on the means to achieve those outcomes. Sec-
ond, they provide a signal to companies to improve
and trigger innovations that will reduce pollution
while simultaneously reducing environmental
costs. Thus, the Porter Hypothesis (guided ecoe?-
ciency hypothesis) argues that bounded rationality
requires carefully crafted regulatory intervention
to induce ecoe?cient behavior.
Ecoe?ciency – in either of its forms (voluntary
or guided) – has signi?cant implications for an
EMA system and also for environmental account-
ing and reporting in general. There are those who
believe that social and environmental accounting
should have the objective of replacing conven-
tional accounting with new accountings that are
predicated on values broader than the traditional
pro?t-maximization objective (Dierkes & Preston,
1977; Patten, 1991; Rubenstein, 1992; Ullman,
1976). Others view pro?t maximization and elimi-
nating environmental degradation as incompatible
and also believe that conventional accounting
encourages, supports, and even fosters social injus-
tices and environmental degradation (Gray &
Bebbington, 2000). Thus, calls are issued for radi-
cal revisions of accounting, with the new account-
ing taking a lead role in fostering needed change so
that sustainability can be achieved (Bebbington,
2001; Bebbington, Gray, & Owen, 1999; Gray,
1992).
However, if ecoe?ciency is valid, then the
calls for radical revisions of accounting to accom-
modate environmental values are much less com-
pelling because the basic premise that pro?t
maximization and environmental improvement
are incompatible would be in error. As Wildavsky
(1994, p. 463) clearly states regarding this issue:
‘‘The changes sought in accounting are also pre-
mised on factual beliefs about the vast harms done
to the natural environment and life-forms of all
kinds by modern technology. If these beliefs are
unwarranted, the case for accounting change col-
lapses.’’ Wildavsky (1994, p. 462) also agrees, in
principle, with the ecoe?ciency claim that pollu-
tion is a form of economic ine?ciency by noting
that: ‘‘A cleaner environment is, in a signi?cant
way, a function of economic e?ciency.’’
Although the validity of ecoe?ciency would
challenge the need for a radical revision of
accounting, it would also carry with it a need for
signi?cant modi?cations of the way current
accounting and reporting is done. To support the
ecoe?ciency paradigm and the needed internal
decision making, an EMA system must identify
and report two types of information: (1) physical
information relating to uses and ?ows of materials,
water, energy, and wastes, and (2) monetary infor-
mation relating to environmental costs, earnings,
and savings (United Nations Division for Sustain-
able Development, 2001). However, several exist-
ing issues could complicate these e?orts. For
instance, conventional accounting systems typi-
cally underestimate environmental costs. More-
over, most companies lack adequate systems for
measuring and managing environmental costs
(Epstein, 1996; Joshi, Krishnan, & Lave, 2001).
As such, some contend that the economic rational-
ist view of environmental performance may be
lacking because the full costs of environmental
activities are not being captured (Herbohn,
2005). Therefore, the availability and use of a
wider set of accounting information relative to
the economic consequences of the social position
(e.g., environmental management) of a company
could provide a more comprehensive view of the
company’s cost structure (Jonson, Jonson, &
Svensson, 1978).
Given these issues, signi?cant modi?cations are
needed to develop an EMA; yet, for this to actually
happen, managers require an incentive to do so.
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 555
The traditional view of the relationship between
environmental performance and economic per-
formance o?ers, at best, no incentive for changing
the current system of accounting and reporting.
If, however, pollution is equivalent to economic
ine?ciency as the Porter Hypothesis claims,
then reducing pollution will increase economic
e?ciency and the value of the ?rm will increase.
Moreover, if investors and creditors were con-
vinced that environmental degradation is equiva-
lent to economic ine?ciency, their demand for
environmental information would increase the
likelihood that ?rms would develop a comprehen-
sive EMA and issue substantive sustainability
reports (Neu et al., 1998; Solomons, 1991). Conse-
quently, ecoe?ciency, if valid, provides a strong
incentive for managers to invest in an EMA system
and to signal their most in?uential and relevant
stakeholders through substantive sustainability
reporting.
What is needed, then, is convincing evidence of
the validity of ecoe?ciency. The early empirical
support for ecoe?ciency is grounded by a solid
body of case studies (DeSimone & Popo?, 1997;
Ditz, Ranganathan, & Banks, 1995; Epstein, 1996;
Schmidheiny & Zorraquin, 1996). Case studies play
an especially important role in accounting research
because they provide a way to investigate the
boundaries of a particular process and help estab-
lish the importance of patterns within an organiza-
tion (Hagg & Hedlund, 1979). Case studies are
useful initially to identify and explore the validity
of hypothesized relationships and thus, serve as
an important forerunner and input for other types
of empirical testing (Kaplan, 1986; Lillis & Mundy,
2005). Indeed, Porter and van der Linde (1995a)
recognize that given the current ability to capture
the true costs and bene?ts of regulation, a reliance
on case studies to examine environmental issues is
helpful. However, Porter and van der Linde
(1995a, p. 104) themselves admit,‘‘. . .a list of case
studies . . .no matter how long, is not a complete
substitute for [other] careful empirical testing’’
and thus, call for more cross-sectional and multi-
unit testing.
Such empirical examinations of the Porter
Hypothesis and voluntary ecoe?ciency are begin-
ning to appear in the accounting and environ-
mental management literature. Al-Tuwaijri et al.
(2004) examine a sample of 198 ?rms from the
Standard and Poor’s 500 and ?nd a statistically
signi?cant positive association between a toxic-
waste measure (toxic waste recycled/total toxic
waste generated) and market returns. They indi-
cate that these results are consistent with the Por-
ter Hypothesis, noting that innovative solutions to
reduce pollution can simultaneously promote envi-
ronmentalism and industrial competitiveness. In
reality, the results are more consistent with volun-
tary ecoe?ciency than guided ecoe?ciency as
there is no speci?c regulatory act mandating a
reduction of toxic waste for all ?rms in the sample.
There are two studies that explicitly consider
the e?ects of regulation. Murty and Kumar
(2003) study the e?ect of environmental regulation
on the productive e?ciency of water polluting
industries in India. They study 92 ?rms in the
sugar industry in India and ?nd that the degree
of technical e?ciency increases with the degree of
compliance to environmental regulation and water
conservation e?orts, a result consistent with the
Porter Hypothesis. They suggest that ?rm manag-
ers should use the message of the Porter Hypothe-
ses to focus on ways to reduce pollution so that
costs can be reduced. Freedman and Jaggi (1992)
examine the association among three measures of
pollution performance and economic performance
for ?rms in the US. pulp and paper industry. These
?rms were subject to the 1972 Federal Water Pol-
lution Control Amendments and signi?cantly
reduced pollution in the 1978–1986 period. Pollu-
tion performance is measured at that plant level,
but economic performance is measured at the ?rm
or segment level. Although all the ?rms reduced
pollution, they were unable to reject the hypothesis
of no association between the pollution reduction
and economic performance. Nonetheless, reducing
pollution without increasing cost is still an interest-
ing outcome. If we de?ne a weak form of ecoe?-
ciency as pollution reduction with no extra cost,
then the absence of a positive or negative relation-
ship is supportive of this weaker form of
ecoe?ciency.
The early evidence is reasonably encouraging
but much more is needed to build a case that
would be su?ciently convincing to management
556 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
and in?uential stakeholders so that implementa-
tion of an EMA becomes a reality for most ?rms.
Absent the economic incentives of ecoe?ciency,
other approaches such as the calls for radical revi-
sions of accounting may begin to assume greater
legitimacy. Thus, the validity of the Porter
Hypothesis is an important empirical question
for environmental accounting. Its validity also car-
ries with it important implications regarding the
nature of regulatory actions needed to improve
both environmental and economic performance.
The Porter Hypothesis and the Federal 1990 Clean
Air Amendments
To empirically investigate the Porter Hypothe-
sis, regulations that emphasize desired environ-
mental improvements without specifying means
of achievement must be present. The 1990 CAAA
satisfy this requirement (EPA, 2000; Swift, 2001, p.
315; US Department of Energy, Energy Informa-
tion Administration, 1994 and 1997). Title IV of
the 1990 CAAA directed the Environmental Pro-
tection Agency (EPA) to establish a program that
would reduce acid rain (formed largely from emis-
sions of sulfur dioxide (SO
2
) and nitrogen oxides
(NO
x
)). This reduction was to be achieved primar-
ily by the electric utility industry because its fossil
fuel plants were the major sources of SO
2
and NO
x
emissions.
3
Title IV also set an eventual overall cap on util-
ity SO
2
emissions at 8.95 million tons per year, or
roughly half the 1980 baseline emissions (for the
industry as a whole). One of the most salient fea-
tures of Title IV was the establishment of an allow-
ance trading market. An allowance was de?ned as
the authorization granted to an electricity pro-
ducer to emit one ton of SO
2
during a calendar
year (US Department of Energy, Energy Informa-
tion Administration, 1994). The EPA initially allo-
cated a certain number of allowances to each
electric utility plant. However, plants that reduced
their actual emissions below the number of allow-
ances allocated directly to them would have excess
allowances that could be traded within their com-
pany, banked for future use, or sold to other util-
ities. Each utility plant had to either reduce
emissions to the number of allowances issued ini-
tially or acquire allowances from other utilities
and thus avoid having to reduce emissions. There-
fore, the total amount of allowances actually held
at any point de?ned the SO
2
output limits. As
such, compliance was determined by comparing
the number of allowances held by a plant to its
actual level of emissions.
Title IV was also unique in that it granted con-
siderable autonomy and ?exibility to plants rela-
tive to the methods they could use to reduce
emissions. Therefore, compliance strategies were
potentially a combination of active allowance
management and the use of proactive pollution
reduction practices. This approach to controlling
SO
2
was a radical departure from the traditional
‘‘command and control’’ approach of pollution
reduction and thus, for the ?rst time, electric util-
ities had a real opportunity to use process
improvements and innovations to reduce emis-
sions and improve e?ciency (EPA, 2000, p. 3).
The economic bene?ts from such a program struc-
ture are obvious as Swift (2001, p. 315) noted
‘‘. . .the cap-and-trade system makes the economic
sense of a ton of reduction very real and exerts
continuous economic pressure to improve perfor-
mance and transform compliance behavior.’’
Finally, Title IV was administered as a two-
phased plan: Phase One and Phase Two. Phase
One of the program ran from 1995 through 1999
and a?ected power plants known to produce large
amounts of SO
2
and NO
x
.
4
Phase Two started in
2000 and in addition to increasing the annual emis-
sions limits imposed on Phase One plants, also
expanded mandated reductions to Non-Phase-
One electric utility plants.
3
Initially, nitrogen oxide (NO
x
) emission reductions were
included as part of the Act. However, the rules governing these
reductions were vacated in 1994 and not replaced until the latter
part of 1995, thus making NO
x
reductions not required until
1996. Therefore, NO
x
reductions are not included in our study.
4
The 1990 CAAA initially required 110 large plants to reduce
their SO
2
to mandated levels by 1995. However, by the
beginning of 1995, 54 additional plants had been brought into
Phase-One. Thus, in 1995, 164 plants underwent the annual
reconciliation process administered by the EPA to determine
compliance with Phase One.
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 557
There has been some empirical investigation of
the 1990 CAAA and the outcomes it produced.
Hughes (2000) uses ?rms a?ected by Phase-One
of the 1990 CAAA to study the relationship
between their market value of equity and SO
2
emissions for the period 1986–1993. He ?nds that
these emissions have value for the years 1989,
1990, and 1991 (there is a signi?cant negative rela-
tionship between market value and SO
2
). This
relationship disappears subsequent to 1991. The
study o?ers several inviting queries. First, Hughes
(2000) suggests that the decline in value relevance
may be attributable to changes made within the
industry in response to the Act. However, the
study provides no direct evidence to support this
suggestion. Second, the study does not provide a
direct link between how changes in productive e?-
ciency are attributable to the regulation. Third,
Hughes (2000) indicates that the initial cost esti-
mates for compliance were signi?cantly less than
expected, and goes on to suggest that this may
be the result of a lack of understanding about
how the impact of the availability of choice rela-
tive to the adoption of pollution control strategies
would a?ect costs. This last point implies the need
for an EMA that provides information that is
needed to understand the business and societal
bene?ts associated with this type of public policy.
A plant-level study may facilitate a response to
these open issues. Indeed, Freedman and Jaggi
(1994) note that ?ndings at the plant level are use-
ful because they provide comprehensive informa-
tion about performance before and after
regulations are imposed. Swift (2001) concurs
and suggests the unique regulatory design of the
Act provides the chance to produce major emis-
sion reductions, lower compliance costs, and cre-
ate very low ongoing transaction costs at both
the plant and the ?rm level. Therefore, analysis
at the plant level o?ers an opportunity to directly
test the Porter Hypothesis by evaluating the
impacts on plant-level e?ciency when actions are
taken in response to pollution reduction mandates.
Given the above, it is appropriate to suggest
one way plants can produce economic value is
through the strategic use of environmental expen-
ditures. As such, there is an underlying rationale
for a place for environmental cost management
and an EMA system in proactive pollution con-
trol. There are at least two major reasons for this:
(1) environmental costs are signi?cant (Bhat, 1996;
Ditz et al., 1995; Joshi et al., 2001; Wong, 2001),
and (2) there is an increasing body of anecdotal
and empirical evidence suggesting that improving
environmental performance can actually reduce
environmental costs (Al-Tuwaijri et al., 2004;
Epstein, 1996; Schmidheiny & Zorraquin, 1996).
Taking the ?rst reason as a given, our general
research objective is to determine if ecoe?ciency
(i.e., the simultaneous reduction of environmental
costs and improvements in environmental quality,
as measured by reductions in SO
2
) has any empir-
ical content. Evidence validating ecoe?ciency can,
then, be used to establish a role for environmental
cost management.
However, a robust evaluation of ecoe?ciency
also requires that it be examined within the context
of hypotheses that incorporate both environmen-
tal and ?scal policy (e.g., environmental cost man-
agement). Three competing hypotheses regarding
environmental cost management have been identi-
?ed: (1) the traditional hypothesis, (2) the volun-
tary (pure) ecoe?ciency hypothesis, and (3) the
guided ecoe?ciency (Porter) hypothesis. Our spe-
ci?c research objective is to evaluate the validity
of the Porter Hypothesis using the 1990 CAAA.
However, since the three hypotheses are compet-
ing, the 1990 CAAA framework also a?ords an
opportunity to shed light on the other two
hypotheses.
Empirical predictions
If the Porter Hypothesis is valid, then four pre-
dictions can be identi?ed that relate to the empir-
ical framework in which we are operating. Two
predictions are concerned with cross-sectional
relationships and two are concerned with longitu-
dinal relationships. Our predictions are as follows.
Prior to intervention (the 1990 period), ecoe?cient
behavior may already exist to some extent in the
industry. Thus, if plants have reduced pollution
relative to an earlier period, then the lower pollut-
ing plants would be more e?cient than the higher
polluting plants. In fact, if it is accepted that pol-
lution is a form of economic ine?ciency as the
558 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
Porter Hypothesis claims, then it is logical to pre-
dict that lower polluting plants are more e?cient
than higher polluting plants. Therefore, we predict
that in 1990 the Non-Phase-One plants are rela-
tively more e?cient than the Phase-One plants (?rst
cross-sectional prediction).
From 1990 to 1995, we expect regulatory inter-
vention to induce ecoe?cient behavior for the
Phase-One plants. Anecdotal evidence appears to
support this assumption. For example, Phase-
One plants reduced their pollution by only about
8% between 1980 and 1990 (EPA, 1996), whereas
the same plants reduced pollution by about 50%
between 1990 and 1995 (Pelz & Fitzgerald, 2001).
Thus, since Phase-One plants are lower polluting
in 1995 relative to 1990 (by a large amount), we pre-
dict that the 1995 Phase-One plants are relatively
more e?cient than their 1990 counterparts (?rst lon-
gitudinal prediction).
The ?nal two predictions are based on the
underlying logic of the Porter Hypothesis coupled
with evidence concerning pollution reduction and
innovation after implementation of the Act. Spe-
ci?cally, the Porter Hypothesis assumes that lower
levels of pollution can be achieved without com-
promising e?ciency. Therefore, it is logical to con-
tinue to assume that lower polluting plants should
continue to be relatively more e?cient than higher
polluting plants, even after intervention. As such,
whichever group of plants is the lower polluting
group (after intervention) should be more e?cient.
Thus, in 1995 we predict that the Non-Phase-One
(Phase-One) plants are relatively more e?cient than
Phase-One (Non-Phase-One) plants provided they
have lower pollution (second cross-sectional
prediction).
Finally, a key claim of the Porter Hypothesis is
that properly crafted regulatory intervention will
lead to innovations that will reduce pollution and
increase economic e?ciency. Interestingly, innova-
tions appear to be at the root of the pollution
reductions achieved by the Phase-One plants.
Swift (2001) notes that the 1990 CAAA stimulated
the development of new fuel blending technologies
and also stimulated a ‘‘spillover e?ect’’ whereby
the innovations of one industry spur innovations
in another industry. In fact, he estimates that the
1990 CAAA promoted over $12 billion in innova-
tive and widespread technologies designed to
reduce pollution, improve fuel access and trans-
portation, and eliminate waste (Swift, 2001, p.
338). For example, in response to the events taking
place in the electric utility industry, coal producers
developed new mining technologies that increased
coal mining productivity by almost 7 percent
between 1990 and 1995 (US Department of
Energy, Energy Information Administration,
1997, p. 23).
The innovations created in response to the 1990
CAAA are likely to have widespread e?ects and
thus, their impact may extend beyond Phase-One
plants. Accordingly, Non-Phase-One plants may
capture the derived bene?ts from the intervention-
induced innovations such as new fuel blending
technologies and lower prices of coal. Further-
more, we would not expect the Non-Phase-One
plants to signi?cantly increase their pollution.
Therefore, we predict that 1995 Non-Phase-One
plants should be at least as e?cient as 1990 Non-
Phase-One plants because of the derived-bene?ts
from Phase-One, Porter-predicted innovations
(second longitudinal prediction).
Research design
Modeling and measuring relative e?ciency
E?ciency is measured in relative terms where
one ?rm is compared to another. Relative compar-
isons are especially important to environmental
management because they provide a baseline to
set improvement goals for external environmental
regulations, internal business practices and emerg-
ing technological implications (Ruch & Roper,
1992). Our paper uses Data Envelopment Analysis
(DEA), a fractional linear programming technique
initially developed by Farrell (1957) and subse-
quently re?ned by Charnes, Cooper, and Rhodes
(1978) to provide these e?ciency measures.
DEA estimates the relative e?ciency of a group
of decision-making units (DMU) with similar
goals and objectives (Callen, 1991). The usual
empirical approach is to assume that ?rms operat-
ing in the same industry have the same production
function. DEA converts multiple input and output
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 559
measures into a single measure of relative perfor-
mance and uses this summary measure to map
the best-practice envelope (frontier) of the input–
output data (Majumadar, 1998). DMUs operating
on the frontier are technically e?cient and rela-
tively more e?cient than those operating o? the
frontier. Fig. 1 illustrates these concepts for six
?rms represented by points A–F. Technically e?-
cient DMUs (A–D) lie on the e?cient frontier,
AD, and are awarded an e?ciency index of one
(e.g., OB/OB). Technically ine?cient DMUs (E
and F) lie above the frontier and have an index
of less than one (e.g., OF
*
/OF).
The e?ciency indices must be computed for
each DMU. Linear programming models are used
to accomplish this objective. Although there are
various DEA models, our relative e?ciency analy-
sis uses the variable-returns-to-scale model
(referred to as the BCC model) developed by
Banker, Charnes, and Cooper (1984). The BCC
model for n DMUs, m outputs, and p inputs is
represented as follows:
Min e ð1Þ
subject to:
X
n
i¼1
w
i
O
ij
PO
kj
; j ¼ 1; 2; . . . ; m ð2Þ
X
n
i¼1
w
i
I
ir
6 eI
kr
; r ¼ 1; 2; . . . ; p ð3Þ
X
n
i¼1
w
i
¼ 1 ð4Þ
where
e the e?ciency or ‘‘DEA’’ scores assigned
by DEA
O
ij
output j for DMU i
I
ir
input r for DMU i
O
kj
the output j for DMU k (target DMU)
I
kr
input r for DMU k, where k denotes the
DMU being evaluated (target DMU) for
its e?ciency relative to all others
w the weight assigned by DEA
The model selects a set of weights that minimize
e. This is equivalent to minimizing the input
resources required by the composite DMU. These
weights de?ne a hypothetical composite DMU
using the outputs and inputs of all DMUs within
the reference group. The value of e lies between
zero and one. If e < 1, then the composite DMU
requires less input resources to produce the same
or greater output as the target DMU. Thus, the
target DMU is judged as relatively ine?cient. If
e = 1, then the target and composite DMU are
one in the same, indicating that there is no other
DMU or group of DMUs that are relatively more
e?cient. In this case, the target DMU is on the e?-
cient frontier.
DEA model for the plant setting
To calculate DEA scores, outputs and inputs
must be selected and de?ned. Five variables are
used in our plant-level DEA model. These consist
of two outputs [Kilowatt-hours (H) and SO
2
(S)]
and three inputs [Capital (K), Fuel Costs (F), and
Operating Costs (C)].
5
Kilowatt-hours are a
positive or good output whereas SO
2
is a bad or
Units of Input 1
U
n
i
t
s
o
f
I
n
p
u
t
2
A
B
E*
E
F
F*
C
D O
Fig. 1.
5
Because of in?ation, nominal prices from di?erent years
cannot be compared without making adjustments. Therefore,
both fuel and operating costs are expressed in constant
dollars. As such, 1995 prices are de?ated to 1990 prices. This
approach is consistent with other studies that examine the
electric utility industry on a longitudinal basis (e.g., Freedman
& Jaggi, 1994).
560 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
negative output.
6
The three inputs are some of the
most widely used to measure performance in the
electric utility industry (Christensen & Greene,
1976; Cowing & Stevenson, 1981) and are among
those frequently employed to calculate DEA
scores in the industry (e.g., Athanassopoulos,
Lambroukos, & Sieford, 1999; Goto & TsuTsui,
1998; Tyteca, 1999).
Capital is the ability of an individual plant to
produce electricity and is measured by the name-
plate generating capacity of the plant. Nameplate
generating capacity is the full-load rating of a
plant to continuously produce electricity. Its use
as a proxy for capital is consistent with prior stud-
ies (Goto & TsuTsui, 1998; Whiteman, 1995;
Yaisawarng & Klein, 1994). Fuel Costs represent
all of the costs associated with the purchase, han-
dling, and storage of fuel inputs burned to gener-
ate electricity. Three issues relative to fuel costs
are worth noting. First, these costs are the largest
and most variable input of a utility’s list of produc-
tion expenses (Bossong, 1999). Also, in accordance
to Freedman and Jaggi (1994), because of its rela-
tive magnitude, special attention should be paid to
this cost category. Second, fuel cost is a speci?c
type of environmental cost as the fuel consumed
(e.g., gas and coal) represents the consumption of
a nonrenewable resource (Tyteca, 1999). Third,
the cost of fuel di?ers by fuel type, and, thus, fuel
cost should re?ect fuel type di?erences among
groups for both cross-sectional and longitudinal
analyses. Therefore, its inclusion is an important
proxy of the ability of management to alter gener-
ating sources in order to take advantage of lower
costs (Haeri, Perussi, & Khawaja, 1999).
Operating Costs are de?ned as the cost of labor
and other expenses related to operating and main-
taining a plant. Thus, operating costs would
include the economic e?ect of investments in
pollution reduction (i.e., the operating costs of
scrubbers). Moreover, when capturing operating
expenses at the plant rather than the ?rm level,
the direct e?ect of controlling and reducing pollu-
tion from a cost basis is more likely to emerge.
Using these inputs and output, relative e?-
ciency for a plant DMU is measured by the follow-
ing program:
Min e ð5Þ
subject to:
X
n
i¼1
w
i
H
i
PH
k
ð6Þ
X
n
i¼1
w
i
ðÀS
i
Þ PÀS
k
ð7Þ
X
n
i¼1
w
i
F
i
6 eF
k
ð8Þ
X
n
i¼1
w
i
K
i
6 eK
k
ð9Þ
X
n
i¼1
w
i
C
i
6 eC
k
ð10Þ
X
n
i¼1
w
i
¼ 1 ð11Þ
One comment about the model needs to be
made. Constraint Eq. (7) compares the hypotheti-
cal composite DMU and the target DMU relative
to the production of SO
2
, a negative output. For a
negative output, the composite DMU can only
produce the same or more output by producing
the same or lower amount of pollution than the
target DMU.
To test the four empirical hypotheses, measur-
ing relative e?ciency at the power plant level for
both cross-sectional and longitudinal analyses is
required. Utilizing DEA scores as our measure is
appropriate because they provide a gauge of tech-
nical e?ciency for each plant relative to the most
e?cient plants in the sample being studied. Not-
withstanding, there are two major concerns rela-
tive to the use of DEA in our study. First, as
noted earlier, DEA computations assume that all
plants have a common production function. How-
ever, in our case high polluting plants (Phase-One
plants) and low polluting plants (Non-Phase-One
plants) may have a di?erent technology (method)
6
For our purposes, good output is de?ned as a revenue-
producing product and the principal objective of the production
processes; negative or bad output is de?ned as a product that is
produced as a consequence of producing the good product, has
no value, and negatively impacts the environment (causes
environmental degradation). This dichotomous de?nition of
electricity production is consistent with exiting literature (i.e.,
Fare & Tyteca, 1996; Golany, Roll, & Rybak (1994); Tyteca
(1997); Tyteca (1996)).
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 561
to produce electricity and thus, any cross-sectional
DEA analysis must consider and account for this
possibility. A similar observation can be made
for longitudinal analysis. For example, as Phase-
One plants reacted to the 1990 CAA intervention,
such actions as switching fuel types and installing
scrubbers may have caused a change in the pro-
duction function over time. In such cases, the
assumption of a common production function
between 1990 and 1995 is problematic.
The second issue relates to the possibility that
DEA scores may, in fact, re?ect the e?ects of fac-
tors other than pollution as possible determinants
of the level of e?ciency. For example, regulatory
climate and di?erences in states in which plants
operate could be causes of di?erences in DEA
scores. Thus, the DEA scores need to be purged
of these e?ects and the resulting ‘‘?ltered’’ DEA
scores should then be used for e?ciency
comparisons.
Common isoquants and relative e?ciency
comparisons
DEA assumes that all DMUs in a sample have
access to and use the same technology, have the
same isoquant, and that ine?ciency is represented
by any variation from the common isoquant (Far-
rell, 1957; Olatubi & Dismukes, 2000). However,
when two seemingly distinct groups of DMUs
are compared, it is conceivable that each group
employs di?erent production techniques with the
result being di?erent best practices isoquants. Fizel
and Nunnikhoven (1992) and Grosskopf and
Valdmanis (1987) provide a methodology that
not only tests to see if two di?erent groups have
a common isoquant but also provides a means of
assessing relative e?ciency when a common iso-
quant does not exist. Their approach requires
three e?ciency indices: (1) overall indices created
when both groups are pooled (EI
p
), (2) within-
group indices produced when DMUs of the sepa-
rate groups are calculated (EI
w
), and (3)
between-group indices (EI
b
).
Fig. 2 illustrates the relationships between EI
p
,
EI
w
, and EI
b
. The example uses nine DMUs and
assumes that points A–D represent DMUs in one
group and points E–I represent DMUs in a second
group. Suppose AD represents the best practices
isoquant for the entire sample, A–I, and that IH
represents the best practices isoquant for the sub-
sample, E–I. Of course, many other representa-
tions for the best practice isoquants for the entire
sample and subgroup are possible; however, this
particular representation is su?cient to illustrate
the methodology.
DMU F can be used to explain the three indi-
ces. The technical e?ciency of F relative to the
overall pooled isoquant AD is:
EI
p
¼ OF
Ã
=OF ð12Þ
The within group index, EI
w
, is calculated as the
ratio of the subsample optimal use (OF
**
) to the
actual input use (OF):
EI
w
¼ OF
ÃÃ
=OF ð13Þ
Finally, the between-group index, EI
b
, is the ratio
of the overall e?ciency index (EI
p
) to the within-
group index (EI
w
):
EI
b
¼ EI
p
=EI
w
¼ ðOF
Ã
=OFÞ=ðOF
ÃÃ
=OFÞ
¼ OF
Ã
=OF
ÃÃ
ð14Þ
To test for a common isoquant for the two groups,
EI
p
and EI
w
are compared for each subsample
using a Wilcoxon Rank Sum z-score to test for dif-
ferences in distributions between the pooled fron-
tier and the subsample frontier. If there is no
U
n
i
t
s
o
f
I
n
p
u
t
2
Units of Input 1
A
B
E*
E
F
F*
C
D
O
H
I
G
F**
Fig. 2.
562 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
signi?cant di?erence between the pooled and with-
in-group indices for both groups, then it is appro-
priate to assume that the subsamples share a
common isoquant. In this case, it is appropriate
to conduct a Wilcoxon Rank Sum test to compare
the EI
p
for each group. If there is a signi?cant dif-
ference between EI
p
and EI
w
then the group being
tested has a frontier isoquant di?erent from the
pooled frontier. For this outcome, the two groups
do not have a common production function and
an EI
b
versus EI
b
Wilcoxon Rank Sum compari-
son is used to test for di?erences in relative e?-
ciency between the two groups.
The relevance of applying this methodology to
our power plant setting is obvious. Relative e?-
ciency assessments at the power plant level are
required for both cross-sectional and longitudinal
analyses. Cross-sectional analyses are done for
plants in 1990 and 1995. For a cross-sectional
analysis, a power plant is a DMU belonging to
one of two groups: (1) the high polluting plant
group (Phase-One plants), or (2) the low polluting
plant group (Non-Phase-One plants). EI
p
and EI
w
are compared to see if the two power plant groups
have a common production function and, depend-
ing on the outcome, either an EI
p
versus EI
p
or EI
b
versus EI
b
Wilcoxon Rank Sum comparison is exe-
cuted to assess di?erences in relative e?ciencies.
Similar longitudinal analyses are conducted using
1990 and 1995 subgroups for the Phase-One and
Non-Phase-One plants, respectively.
Purging and explaining DEA scores
Comparing relative e?ciencies of two groups
using unadjusted DEA scores ignores the possibil-
ity that di?erences in DEA scores can be caused by
factors other than e?ciency. Furthermore, there
may be other variables that cause di?erences in
e?ciency other than di?erences in pollution levels
(Fizel & Nunnikhoven, 1992; Olatubi & Dismukes,
2000). Conceptually, these variables represent
those external factors over which plants have very
little control and yet, may cause di?erences in
DEA scores. Thus, recognizing and removing the
e?ects of a set of confounding variables should
improve the validity of the empirical tests of the
Porter Hypothesis.
Following Fizel and Nunnikhoven (1992) and
Olatubi and Dismukes (2000), regression analysis
is used to purge the DEA scores of the confound-
ing e?ects in the following manner.
7
First, the
unadjusted DEA scores are regressed on the con-
founding factors thought to in?uence the value
of the e?ciency indices. Then, the residuals from
this regression are harvested and become the
purged e?ciency measures that will be used to
compare the relative e?ciencies between groups.
Using these purged values, a cleaner test of the
Porter Hypothesis can be obtained.
In addition to removing the confounding
e?ects, Fizel and Nunnikhoven (1992) and Olatubi
and Dismukes (2000) suggest that it is also possible
to identify speci?c environmental (compliance) vari-
ables that may re?ect actions taken by plants to
reduce or manage pollution to achieve compliance.
Knowing how plants acted to comply with the
1990 CAAA and how these actions a?ect e?ciency
may shed some additional light on the validity of
the Porter Hypothesis (e.g., evidence of innovative
actions). Again, regression analysis is employed to
gain this insight.
Confounding variables
We have identi?ed four confounding variables:
(1) favorableness of regulatory climate (REG);
(2) stringency of state environmentalism (STATE);
(3) age of the plant (AGE); and (4) scale of opera-
tions at the operator level (MKTSHR).
8
The ?rst
two variables are measures of state environmental
e?ects. The second two variables are longer-term
plant and operator-level economic factors that
can a?ect plant-level e?ciency.
Prior research implies that there is an associa-
tion between a utility’s value and the favorableness
of its regulatory climate (Blacconiere, Johnson, &
7
Because DEA scores are bounded between zero and one,
ordinary least square (OLS) may be biased and thus, may not
be an appropriate estimation technique (Cooper, Seiford, &
Tone, 2000). To avoid this potential bias, a tobit-censored
regression model is used (Claggert & Ferrier, 1998; Olatubi &
Dismukes, 2000; Puig-Junoy, 1998).
8
Clearly, one limitation is that the list of potential con-
founding variables is not exhaustive but re?ects those used in
prior studies and those that have data available. Nonetheless,
we would expect their usage to improve the e?ciency measures.
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 563
Johnson, 1997; D’Souza, Jacob, & Soderstrom,
2000; Loudder, Khurana, & Boatsman, 1996).
However, others have argued that these rate-set-
ting procedures may not provide an incentive to
become e?cient and contend that the entire pro-
cess promotes waste and provides little motivation
to cut costs or improve e?ciency (Bossong, 1999;
Haeri, Khawaia, & Perussi, 1997, p. 26). Thus,
the regulatory climate of the state in which the
plant operates may have some e?ect on plant e?-
ciency. In this study, we follow Hughes (2000) and
develop a measure of regulatory climate for states
that assumes the following values: below average
(À1), average (0), and above average (+1). We
adopt this approach to develop (REG).
King and Lenox (2002) indicate that environ-
mental regulation varies across regions. We recog-
nize this in our study as follows. First, we de?ne a
region as the state in which the ?rms operate. Sec-
ond, we recognize that the stringency of regulation
varies among states based on di?erences in policies,
programs, and statutes adopted to protect the envi-
ronment. Meyer (1992) uses a set of twenty envi-
ronmental policy variables to evaluate di?erences
in environmental stringencies across states using
such criteria. As such, in an approach similar to
Hughes (2000), we use the Meyer (1992) state envi-
ronmentalism assessments (STATE) to develop
the following stringency values that are assigned
to the states: weak (À1), moderate (0), and strong
(+1).
Juskow and Schmalaensee (1987) posit that the
performance of a plant will eventually deteriorate
as the unit ages. Following that study, an age var-
iable (AGE) is measured as the di?erence between
the calendar year and the initial date of operation.
Finally, the scale of operations at the operator
level may also have an impact on the ability of a
single plant to make changes. This may occur
because of the transfer of knowledge that may take
place between the plants as more output is pro-
duced by an operator. Olatubi and Dismukes
(2000) suggest that this may have an impact in
the e?ciency of plant. Following that study, we
proxy market share (MKTSHR) as the percentage
of total kilowatt hours produced by the operator
of the plant in the NERC region in which the plant
is located.
Environmental (compliance) variables
Assuming that the plants react to the 1990
CAAA, there are some speci?c environmental (com-
pliance) variables that can be used to manage the
pollution reduction requirements. We have identi-
?ed ?ve compliance variables: (1) the percentage of
sulfur present in the coal consumed (AVC); (2) the
percentage of sulfur present in the petroleum con-
sumed (AVP); (3) the percentage of coal used as a
fuel (BTUC); (4) the number of scrubbers installed
(SCR); and (5) permits usage (PERMIT). The ?rst
three variables are concerned with managing fuel
types to reduce pollution. The last two are alterna-
tive means of managing pollution. Permit usage, of
course, does not lead to pollution reduction.
One of the ways plants can reduce SO
2
emis-
sions is to convert to coal with lower sulfur con-
tent. In fact, the use of low to medium sulfur
content coal in 1995 accounted for 77 percent of
total coal receipts at electric utilities, which is more
than 20 higher than the receipts of the same type of
coal in 1990 (US Department of Energy, Energy
Information Administration, 1994). Moreover,
evidence suggests that the switch from higher to
lower sulfur coal can occur with little deterioration
to plant performance (US Department of Energy,
Energy Information Administration, 1994). There-
fore, the percent of sulfur contained in the coal
consumed at a plant (AVC) is used as a proxy
for the level of sulfur present in the coal employed
to produce electricity. Another fossil fuel often
used to produce electricity is petroleum and its
use also produces SO
2
. Like AVC, we include the
percent of sulfur contained in the petroleum con-
sumed at a plant (AVP) as a proxy for the level
of sulfur present in the petroleum employed to
produce electricity. Moreover, we do recognize
that the level of SO
2
produced by a plant is likely
to be a?ected by the fuel mix used at the plant. In
general, there are three primary fuels that are
employed: (1) coal, (2) petroleum and (3) gas. Log-
ically, plants that adopt coal as their primary
source will produce more pollutants. Therefore,
to account for this e?ect, we include the percent
of coal used to create a BTU of heat at a plant
(BTUC) as a proxy for fuel mix di?erences.
Olatubi and Dismukes (2000) suggest that the
presence of a scrubber may have an impact on
564 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
the e?ciency of a plant and certainly ought to
a?ect the level of pollution. Thus, the presence of
scrubbers (SCR) seems like a reasonable way to
measure one approach used to reduce pollution.
Finally, one of the key components of Title IV is
allowing plants to use permits to emit di?erent lev-
els of pollution. Information about the initial allo-
cation of permits, the use of permits held (initial
allocation plus those bought, sold, or otherwise
distributed), and the amount of permits remaining
after reconciliation could provide some insight
into how the plant managed its pollution, and
the extent to which permits were used to achieve
compliance. For 1990, (PERMIT) is computed
as: the total allowances on hand at the start of
Phase 1 divided by the total number of allowances
that were initially awarded. For 1995, (PERMIT)
is computed as: the total amount of allowances
actually used divided by the actual amount on
hand at the start of Phase 1. This proxy a?ects
only Phase One plants and is used only for the lon-
gitudinal analysis.
Regression models
The regressions for each category of variables
will be run for cross-sectional and longitudinal set-
tings, using the following basic models:
EI ¼ b
0
þ b
1
REG þ b
2
STATE þ b
3
AGE
þ b
4
MKTSHR þ b
5
DUMMY þ e
1
ð15Þ
EI ¼ b
0
þ b
1
AVC þ b
2
AVP þ b
3
BTUC
þ b
4
SCR þ b
5
PERMIT þ b
6
DUMMY þ e
2
ð16Þ
The dependent variable, EI, corresponds to EI
p
for
pooled samples and to EI
w
for within-group sam-
ples. In addition, an indicator variable, DUMMY,
is included to test whether relative e?ciencies di?er
across the current groups being investigated (either
plant type or year). Speci?cally, DUMMY is used
only for pooled regressions and not for within-
group regressions. For cross-sectional pooled sam-
ple regressions, DUMMY is labeled TYPE and is
equal to one if the plant is a Phase-One plant and
zero if it is a Non-Phase-One plant. For longitudi-
nal pooled sample regressions, DUMMY is la-
beled as YEAR and is equal to one if the period
is 1995 and zero if it is 1990. Finally, the variable,
PERMIT, is included only for the sample longitu-
dinal regressions exclusively involving Phase-One
plants.
Sample and data collection
Data were collected for Phase-One plants for the
periods 1990 and 1995 and for Non-Phase-One
plants for the same periods. Non-Phase-One plants
were identi?ed using the Inventory of Power Plants
in the United States for 1990. Non-Phase-One
plants were drawn from the same North American
Electric Reliability Council (NERC) region as
Phase One plants. NERC region is used in this
manner because it is assumed that plants located
within the same region face a common set of pro-
duction, operating, and regulatory characteristics.
In addition, prior studies have used NERC regions
as a basis for cost comparisons within the electric
utility industry (Blacconiere et al., 1997; Knutson,
1995). Further, to the extent possible, nameplate
capacity is also used to identify Non-Phase One
plants. This approach was adopted based on simi-
lar methodologies present in existing studies that
make comparisons within the electric utility indus-
try (Goto & TsuTsui, 1998; Whiteman, 1995;
Yaisawarng & Klein, 1994).
The Federal Energy Regulatory Commission
(FERC) Form 1 reports were used to gather data
related to the inputs, good (kilowatt hour) output,
and the age of the plants. Only major investor-
owned electric utilities are required to submit a
FERC Form 1. Investor owned utilities are those
electric utilities owned by investors. In contrast,
publicly owned utilities are operated by the federal
government, a state or local municipality, or an
electric cooperative. Data incorporated in a FERC
Form 1 include income and earnings, operating
and maintenance expenses, and various produc-
tion statistics. Plant level ?nancial and operating
data are located in the ‘‘Steam Generating Plant
Statistics for Large Plants’’ section of the FERC
Form 1. The relative e?ciencies (i.e., DEA scores)
for this study were produced using DEA-Solver-
Pro 3.0, a windows based linear programming
model developed by Saitech, Inc.
The Energy Information Administration of the
Department of Energy was the source for SO
2
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 565
emissions. Like Hughes (2000), assessments of the
regulatory environment were drawn from the
Value Line Investment Surveys. The market share
of the operators within the NERC region was
compiled using the Energy Information Adminis-
tration of the Department of Energy Form 906
Database. The average sulfur content of coal and
petroleum, as well as the use of coal to produce
electricity were all obtained from the 1990 and
1995 Cost and Quality of Fuels For Electric Utility
Plants. Information about scrubber activity was
found at the Energy Information Administration
Clean Air Act Database Browser. Finally, informa-
tion about the permits came from the Environmen-
tal Protection Agency’s 1995 Acid Rain Compliance
Report.
Empirical results
Descriptive statistics
Table 1 details and summarizes the sample
selection process used. Phase-One of the 1990
CAAA a?ected 164 plants. Of these 164 plants,
80 were eliminated from the sample for one of sev-
eral reasons: (1) they did not ?le a FERC Form 1;
(2) plant level data were not included in the FERC
Form 1 submission; (3) a plant had missing or
incomplete operating or ?nancial information;
and, (4) a suitable Non-Phase One plant could
not be located. Ultimately 84 Phase-One and
Non-Phase-One plants were included in the study.
Table 2 provides a summary of plant level data
and the descriptive statistics for Phase One and
Non-Phase-One plants for 1990 and 1995. Panels
IA and IB of Table 2 re?ect cross-sectional com-
parisons of Phase One and Non-Phase One plants
at 1990 and 1995, respectively. Panels IIA and IIB
provide a longitudinal comparison of plant types
across 1990 and 1995. Statistical tests of the means
of the data presented in the Panels are also
included to provide a preliminary assessment of
the e?ects of the 1990 CAAA on the plants.
Panel IA of Table 2 reveals that there is no sig-
ni?cant di?erence in size (as measured by name
plate generating capacity), kilowatt hours, or costs
between the Phase-One and Non-Phase-One
plants in 1990. However, there is a signi?cant dif-
ference in SO
2
production between the Phase-One
and Non-Phase-One plants (the higher polluting
plants produce nearly four times as much total
SO
2
). Thus, in 1990, Non-Phase-One plants on
average produce the same good output as Phase-
One plants, but do so with a fraction of the bad
output and no greater cost. The traditional
hypothesis predicts that reducing bad output is
costly while ecoe?ciency maintains that reducing
bad output will increase e?ciency. Thus, the
descriptive evidence provides some weak or mild
support for the empirical prediction that the lower
Table 1
Sample of phase one plants with available FERC Form 1 data
a
Total Phase One Plants
b
164
Plants Not Required to File a FERC Form 1
c
(25)
Initial Set of Plants with FERC Form 1 Data 139
Data Not Located in the FERC Form 1
d
(15)
Data in FERC Form 1 was Unusable
e
(26)
Unable to Locate as Suitable Non-Phase One Plant
f
(14)
Final Phase One Plants Used
g
84
a
The data for the study are obtained from the Federal Energy
Regulatory Commission Form 1 (FERC Form 1). Only major
investor-owned electric utilities are required to submit a FERC
Form 1. FERC Form 1 submissions are required annually.
Data presented in a FERC Form 1 include income and earn-
ings, operating and maintenance expenses, and various pro-
duction statistics. Plant level ?nancial and operating data are
located in the ‘‘Steam Generating Plant Statistics for Large
Plants’’ section of the FERC Form 1.
b
Includes all of the electric utility plants a?ected by Phase
One of Title IV of the 1990 CAAA. These plants are referred to
as Phase One plants. Starting in 1995, the SO
2
emissions of
Phase One plants were evaluated by the Environmental Pro-
tection Agency (EPA) to ensure that they are in compliance
with the 1990 CAAA.
c
Includes those plants that are publicly-owned and thus, are
not required to submit a FERC Form 1. Publicly-owned utili-
ties are operated by (1) the federal government; (2) a state or
local municipality; or (3) an electric cooperative.
d
Includes those Phase One plants whose ‘‘Steam Generating
Plant Statistics For Large Plants’’ data were not included with
the FERC Form 1 ?led with the Commission.
e
Includes those Phase One plants whose ‘‘Steam Generating
Plant Statistics For Large Plants’’ data ?led in the FERC
FORM 1 were incomplete.
f
Includes those Phase One plants that had no similar Non-
Phase-One plant.
g
Re?ects the total number of Phase One Plants: (1) whose
data is complete; (2) can be successfully associated with Non-
Phase One plants. Thus, the grand total of the plants included
in the study is 168 (i.e., 84 Phase One and 84 Non-Phase One).
566 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
Table 2
Summary and descriptive statistics for phase one and non-phase one plants
a
Variable Phase
One
Non-Phase
One
Net
Di?
% Di? Phase One
Mean
Phase One
Std. Dev
Non-Phase
One Mean
Non-Phase
One Std. Dev
T Stat
Part I: Cross-sectional analyses (Hypothesized mean di?erences: Non-Phase-One mean less Phase-One mean)
Panel A: 1990 Comparisons (n = 84 for each group)
Inputs
NPC 72407 73064 657 0.91 861.98 607.60 869.81 613.93 0.08
PC 6995 6837 (158) (2.26) 83 63 81 70 À0.18
FC 5467 5370 (97) (1.78) 65 54 63 56 À0.13
OC 1527 1467 (60) (3.96) 18 11 17 20 À0.28
Outputs
KWH 332428 264437 (67991) (20.45) 3957 3307 3148 3084 À1.64
POLL 5153797 1378377 (3775420) (72.26) 61354 68062 16409 29040 À5.57
*
Panel B: 1995 Comparisons (n = 84 for each group)
Inputs
NPC 71816 72557 741 1.03 854.96 614.16 863.77 614.14 0.09
PC 5331 5008 (323) (6.06) 63 54 59 48 À0.48
FC 3907 3835 (72) (1.84) 46 42 45 36 À0.14
OC 1424 1172 (252) (17.70) 16 14 13 16 À1.27
Outputs
KWH 317387 266246 (51141) (16.11) 3778 3573 3169 2984 À1.20
POLL 2810175 1397005 (1413170) (50.29) 33454 40824 16631 27285 À3.14
*
Operations
NDECREASE 76 44 (32) (42.1) 0.904 0.294 0.524 .499 À5.47
*
Variable 1990 1995 Net Di? % Di? 1990 Mean 1990 Std. Dev 1990 Mean 1995 Std. Dev T Stat
Part II: Longitudinal Analyses (Hypothesized mean di?erences: 1995 mean less 1990 mean)
Panel A: Phase One Plants (n = 84)
Inputs
NPC 72407 71816 (591) (0.82) 861.98 607.60 854.96 614.16 À0.07
PC 6995 5331 (1664) (23.79) 83 63 63 54 À2.16
*
FC 5467 3907 (1560) (28.53) 65 54 46 42 À2.45
*
OC 1527 1424 (103) (6.75) 18 11 16 14 À0.61
Outputs
KWH 332428 317387 (15041) (4.52) 3957 3307 3778 3573 À0.34
POLL 5153797 2810175 (2343622) (45.47) 61354 68062 33454 40824 À3.22
*
Panel B: Non-Phase One Plants (n = 84)
Inputs
NPC 73064 72557 (507) (0.69) 869.81 613.93 863.77 614.14 0.06
PC 6837 5008 (1829) (26.75) 81 70 59 48 2.33
*
FC 5370 3835 (1535) (28.58) 63 56 45 36 2.48
*
(continued on next page)
R
.
D
.
B
u
r
n
e
t
t
,
D
.
R
.
H
a
n
s
e
n
/
A
c
c
o
u
n
t
i
n
g
,
O
r
g
a
n
i
z
a
t
i
o
n
s
a
n
d
S
o
c
i
e
t
y
3
3
(
2
0
0
8
)
5
5
1
–
5
8
1
5
6
7
polluting 1990 Non-Phase-One plants are at least
as e?cient as the 1990 Phase-One plants. Panel
IB of Table 2 compares the Phase-One and Non-
Phase-One plants in 1995 after intervention. The
Phase-One plants continue to have signi?cantly
higher pollution than Non-Phase-One plants with
no signi?cant di?erences in cost and good output.
Thus, the lower polluting plants continue to pro-
duce (on average) the same good output with no
greater cost, an outcome weakly consistent with
ecoe?ciency.
However, Panel IB of Table 2 also reveals that
the di?erence in the levels of pollution has nar-
rowed, with the Phase-One plants producing only
about double the pollution of Non-Phase-One
plants in 1995. Also, as Panels IIA and IIB of
Table 2 indicate, from 1990 to 1995 the Phase-
One plants reduced total pollution by 45.47 per-
cent (a statistically signi?cant decrease) while the
Non-Phase-One plants increased their total pollu-
tion by 1.35 percent (not a signi?cant change).
Furthermore, the pollution reductions of the
Phase-One plants were wide-spread with 76 or
about 90% of the plants reducing pollution from
1990 to 1995 (Panel IB) with an average reduction
of 27,900 tons. Thus, not only did the Phase-One
plants reduce pollution by a large amount but
there were also signi?cant decreases in production
and fuel costs from 1990 to 1995. Reducing pollu-
tion while simultaneously decreasing cost is consis-
tent with ecoe?ciency. The outcome tends to
support the Porter Hypothesis and thus, provides
evidence to support the validity of ecoe?ciency.
Thus, the descriptive data suggest that the
Phase-One plants have no signi?cant di?erence in
output from 1990 to 1995, post signi?cant reduc-
tions in fuel costs and total production costs
and, yet, manage to simultaneously reduce pollu-
tion by a signi?cant amount. The Non-Phase-
One plants experience the same outcomes with
one exception: there is no signi?cant change in pol-
lution. Interestingly, both groups experience rela-
tively the same percentage reduction in fuel costs
(28 percent). Collectively, these results provide sev-
eral additional insights. First, since fuel costs are
an environmental cost (re?ecting the use of nonre-
newable resources), the descriptive evidence por-
trays an environmental management model T
a
b
l
e
2
(
c
o
n
t
i
n
u
e
d
)
V
a
r
i
a
b
l
e
1
9
9
0
1
9
9
5
N
e
t
D
i
?
%
D
i
?
1
9
9
0
M
e
a
n
1
9
9
0
S
t
d
.
D
e
v
1
9
9
0
M
e
a
n
1
9
9
5
S
t
d
.
D
e
v
T
S
t
a
t
O
C
1
4
6
7
1
1
7
2
(
2
9
5
)
(
2
0
.
1
1
)
1
7
2
0
1
3
1
6
1
.
2
4
O
u
t
p
u
t
s
K
W
H
2
6
4
4
3
7
2
6
6
2
4
6
1
8
0
9
0
.
6
8
3
1
4
8
3
0
8
4
3
1
6
9
2
9
8
4
À
0
.
0
5
P
O
L
L
1
3
7
8
3
7
7
1
3
9
7
0
0
5
1
8
6
2
8
1
.
3
5
1
6
4
0
9
2
9
0
4
0
1
6
6
3
1
2
7
2
8
5
À
0
.
0
5
a
T
h
i
s
t
a
b
l
e
r
e
p
o
r
t
s
s
e
l
e
c
t
e
d
d
e
s
c
r
i
p
t
i
v
e
a
n
d
s
t
a
t
i
s
t
i
c
a
l
p
l
a
n
t
l
e
v
e
l
d
a
t
a
o
n
:
(
1
)
a
c
r
o
s
s
-
s
e
c
t
i
o
n
a
l
(
b
e
t
w
e
e
n
)
p
l
a
n
t
t
y
p
e
b
a
s
i
s
;
a
n
d
(
2
)
a
l
o
n
g
i
t
u
d
i
n
a
l
(
w
i
t
h
i
n
)
p
l
a
n
t
t
y
p
e
b
a
s
i
s
.
A
l
l
p
-
v
a
l
u
e
s
r
e
l
a
t
e
t
o
a
t
w
o
-
s
i
d
e
d
t
-
t
e
s
t
w
h
e
r
e
*
r
e
?
e
c
t
s
p
<
0
.
1
0
.
T
h
e
v
a
r
i
a
b
l
e
d
e
?
n
i
t
i
o
n
s
:
N
P
C
T
h
e
n
a
m
e
p
l
a
t
e
(
f
u
l
l
l
o
a
d
)
g
e
n
e
r
a
t
i
n
g
c
a
p
a
c
i
t
y
e
x
p
r
e
s
s
e
d
a
s
m
i
l
l
i
o
n
s
o
f
w
a
t
t
s
c
a
p
a
c
i
t
y
.
K
W
H
T
h
e
k
i
l
o
w
a
t
t
p
e
r
h
o
u
r
o
f
e
l
e
c
t
r
i
c
i
t
y
g
e
n
e
r
a
t
e
d
e
x
p
r
e
s
s
e
d
a
s
m
i
l
l
i
o
n
s
o
f
K
W
H
.
P
C
T
h
e
t
o
t
a
l
p
r
o
d
u
c
t
i
o
n
c
o
s
t
s
(
f
u
e
l
p
l
u
s
a
l
l
o
t
h
e
r
c
o
s
t
s
)
e
x
p
r
e
s
s
e
d
i
n
m
i
l
l
i
o
n
s
o
f
c
o
n
s
t
a
n
t
(
1
9
9
0
)
d
o
l
l
a
r
s
.
F
C
F
u
e
l
c
o
s
t
s
e
x
p
r
e
s
s
e
d
i
n
m
i
l
l
i
o
n
s
o
f
c
o
n
s
t
a
n
t
(
1
9
9
0
)
d
o
l
l
a
r
s
.
O
C
O
p
e
r
a
t
i
n
g
c
o
s
t
s
e
x
p
r
e
s
s
e
d
i
n
m
i
l
l
i
o
n
s
o
f
c
o
n
s
t
a
n
t
(
1
9
9
0
)
d
o
l
l
a
r
s
.
P
O
L
L
T
h
e
a
m
o
u
n
t
o
f
S
O
2
p
o
l
l
u
t
i
o
n
i
n
a
c
t
u
a
l
t
o
n
s
.
N
D
E
C
R
E
A
S
E
N
u
m
b
e
r
o
f
p
l
a
n
t
s
s
h
o
w
i
n
g
a
d
e
c
r
e
a
s
e
i
n
p
o
l
l
u
t
i
o
n
f
r
o
m
1
9
9
0
t
o
1
9
9
5
.
568 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
where environmental costs are simultaneously
reduced with improvements in environmental
quality. Second, this evidence is consistent with
the claims of process innovations within the elec-
tric utility industry and spillover innovations pro-
duced by suppliers. Finally, it also appears that the
Non-Phase-One plants may be capturing bene?ts
produced by the Phase-One plants, an outcome
consistent with the derived-bene?ts assumption.
Results of DEA analyses with unadjusted e?ciency
indices
The distribution of the unadjusted DEA scores
and the summary statistics are provided in Tables
3 and 4 for each of the three e?ciency indices
(pooled, within-group and between-group) that
are needed to test the various hypotheses of inter-
est. Table 5 provides the appropriate statistical
tests using these three indices.
Cross-sectional analysis
The cross-sectional hypotheses predict that the
lower polluting, Non-Phase-One plants will be rel-
atively more e?cient than the higher polluting,
Phase-One plants. To ensure that any di?erences
in e?ciency are not caused by di?erences that arise
because the two types face di?erent frontiers, we
?rst test to see whether the Phase-One and Non-
Phase-One groups share a common production
Table 3
Cross-sectional frequencies and summary statistics unadjusted e?ciency indices
a
E?ciency score ranges EI
p
= pooled index EI
w
= within-group index EI
b
= between–group index
Total Phase One Non-Phase One Phase One Non-Phase-One Phase One Non-Phase-One
Panel A: 1990 Statistics for Phase-One and Non-Phase-One Plants
1.00 26 10 16 19 27 38 24
0.90–0.99 2 0 2 0 1 20 18
0.80–0.89 1 0 1 1 3 10 30
0.70–0.79 10 2 8 3 2 6 9
0.60–0.69 15 8 7 10 9 3 2
0.50–0.59 29 17 12 15 14 2 1
0.40–0.49 47 31 16 25 12 3 0
0.30–0.39 21 12 9 9 6 1 0
0.20–0.29 15 3 12 1 10 1 0
0.10–0.19 2 1 1 1 0 0 0
N 168 84 84 84 84 84 84
Mean 0.5624 0.5350 0.5897 0.6193 0.6612 0.8922 0.8948
Standard deviation 0.2338 0.2021 0.2600 0.2327 0.2770 0.1724 0.0974
Panel B: 1995 Statistics for Phase-One and Non-Phase-One Plants
1.00 25 10 15 13 21 14 59
0.90–0.99 1 1 0 0 0 41 12
0.80–0.89 2 1 1 1 3 25 4
0.70–0.79 8 1 7 5 6 1 7
0.60–0.69 14 9 5 13 4 2 0
0.50–0.59 28 14 14 19 10 1 2
0.40–0.49 43 26 17 18 16 0 0
0.30–0.39 37 16 21 11 20 0 0
0.20–0.29 9 5 4 3 4 0 0
0.10–0.19 1 1 0 1 0 0 0
N 168 84 84 84 84 84 84
Mean 0.5544 0.5345 0.5743 0.5815 0.6105 0.9149 0.9578
Standard deviation 0.2283 0.2158 0.2398 0.2159 0.2655 0.0787 0.0970
a
This table re?ects the distribution and descriptive statistics for the three types of e?ciency indices for Phase One and Non-Phase
One plants within each year before adjusting for confounding variables. Unadjusted indices are bounded between zero and one.
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 569
function (Panel A of Table 5). This question is
addressed by comparing the pooled indices for
1990 and 1995 with the within-group indices for
the two types of plants. Wilcoxon Rank Sum
z-scores are used to compare the distributions of
the two groups.
For each year, at least one frontier of the two
subgroups is statistically di?erent from the pooled
frontier. Thus, the two groups do not share a com-
mon production function in either year. For 1990,
both frontiers are di?erent from the pooled fron-
tier, leaving the e?ciency relationships unclear.
However, for 1995, only the Phase-One plants have
a frontier that is shifted out from the pooled fron-
tier, which immediately supports the hypothesis
that the Non-Phase-One plants are relatively more
e?cient. Based on the absence of a common fron-
tier, between-group tests are used to assess di?er-
ences in relative e?ciency between the two
groups (Table 5, Panel B). For 1990, the evidence
indicates that the Phase-One plants are relatively
more e?cient than the Non-Phase-One plants, an
outcome inconsistent with the 1990 cross-sectional
hypothesis. However, the evidence is strongly sup-
portive of the 1995 cross-sectional hypothesis.
Longitudinal analysis
The longitudinal hypotheses predict that the
1995 plants of each type will be relatively more e?-
cient than their 1990 counterparts. Table 5, Panel
Table 4
Longitudinal frequencies and summary statistics unadjusted e?ciency indices
a
E?ciency score ranges EI
p
= pooled index EI
w
= within-group index EI
b
= between-group index
Total 1990 1995 1990 1995 1990 1995
Panel A: Phase-One Statistics for 1990 and 1995
1.00 23 10 13 19 13 10 81
0.90–0.99 0 0 0 0 0 15 3
0.80–0.89 1 0 1 1 1 38 0
0.70–0.79 7 2 5 3 5 9 0
0.60–0.69 16 3 13 10 13 3 0
0.50–0.59 30 11 19 15 19 4 0
0.40–0.49 50 32 18 25 18 2 0
0.30–0.39 29 18 11 9 11 2 0
0.20–0.29 10 7 3 1 3 1 0
0.10–0.19 2 1 1 1 1 0 0
N 168 84 84 84 84 84 84
Mean 0.5429 0.5044 0.5813 0.6193 0.5815 0.8301 0.9997
Standard deviation 0.2155 0.2094 0.2159 0.2327 0.2159 0.1533 0.0020
Panel B: Non-Phase-One Statistics for 1990 and 1995
1.00 33 15 18 27 21 16 32
0.90–0.99 1 1 0 1 0 6 45
0.80–0.89 3 1 2 3 3 17 4
0.70–0.79 11 6 5 2 6 23 1
0.60–0.69 4 2 2 9 4 15 0
0.50–0.59 4 11 13 14 10 5 1
0.40–0.49 29 13 16 12 16 2 1
0.30–0.39 39 17 22 6 20 0 0
0.20–0.29 20 14 6 10 4 0 1
0.10–0.19 4 4 0 0 0 0 0
N 168 84 84 84 84 84 84
Mean 0.5562 0.2665 0.2896 0.6612 0.6105 0.7955 0.9596
Standard deviation 0.2660 0.2730 0.2585 0.2770 0.2655 0.1433 0.1141
a
This table re?ects the distribution and descriptive statistics for the three types of e?ciency indices for Phase-One and Non-Phase-
One plants across years before adjusting for confounding variables. Unadjusted indices are bounded between zero and one.
570 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
A, indicates that the 1990 frontiers are statistically
di?erent from the pooled frontier, while the 1995
frontiers are not statistically di?erent. Thus, the
1990 frontiers are shifted out from the pooled
frontier, while the 1995 frontiers are not statisti-
cally di?erent from the pooled frontiers. This out-
come supports the hypotheses that the 1995
plants are relatively more e?cient than the 1990
plants (see Fig. 2). The between-group compari-
sons (Table 5, Panel B) strongly support the
hypotheses that the 1995 plants are relatively more
e?cient than the 1990 plants. Thus, the longitudi-
nal hypotheses are consistent with the Porter
Hypothesis.
The evidence of the unadjusted DEA scores is
largely supportive of the Porter Hypothesis and
of ecoe?ciency, with only one cross-sectional
hypothesis failing the test. However, while the
unadjusted indices provide useful evidence, they
may not be completely de?nitive because of the
possibility that the indices are a?ected by contam-
inating variables.
DEA analysis with adjusted indices
Descriptive data for the confounding and com-
pliance variables are given in Table 6 for both
pooled and within-group samples. Tobit regression
Table 5
Statistical analyses of unadjusted productive e?ciency
a
Cross-sectional analysis Distribution
b
Longitudinal analysis Distribution
b
Panel A: EI
p
versus EI
w
1990 EI
p
6308 Phase-One EI
p
5844
1990 Phase-One EI
w
7888 1990 Phase-One EI
w
8352
Z-score 2.51
**
Z-score À3.98
***
1990 EI
p
6574 Phase-One EI
p
7091
1990 Non-Phase-One EI
w
7622 1995 Phase-One EI
w
7104
Z-score 1.67
*
z-Score À0.02
1995 EI
p
6536 Non-Phase-One EI
p
6099
1995 Phase-One EI
w
7660 1990 Non-Phase-One EI
w
8097
Z-score 1.78* z-score À3.19
***
1995 EI
p
6902 Non-Phase-One EI
p
6857
1995 Non-Phase-One EI
w
7293 1995 Non-Phase-One EI
w
7339
Z-score 0.62 Z-score À0.76
Cross-sectional analysis Distribution
c
Longitudinal analysis Distribution
c
Panel B: EI
b
versus EI
b
1990 Phase One EI
b
7691 1990 Phase-One EI
b
4005
1990 Non-Phase One EI
b
6505 1995 Phase-One EI
b
10191
Z-score À1.92
**
Z-score 10.69
***
1995 Phase One EI
b
5291 1990 Non-Phase One EI
b
5005
1995 Non-Phase One EI
b
8905 1995 Non-Phase One EI
b
9191
Z-score 5.98
***
Z-score 6.71
***
a
This table reports the results of tests of comparative relative e?ciency based on the unadjusted productive e?ciency indices. For
this group of tests, Panel A tests for the presence of a common production function. Panel B tests for di?erences between groups when
a common production function is unavailable. Signi?cance levels of p < 0.10 and p < 0.05 and p < 0.001 are denoted as * , **, and ***
respectively.
b
For all tests, the null hypothesis assumes that the expected rank sums (distributions) of the groups being compared is the same.
Results re?ect the cumulative expected rank sum of each group and the z-score when a Wilcoxon Rank Sum test is employed. All
z-scores are two sided.
c
For the cross-sectional hypotheses, the null hypotheses assume that the Phase-One plants are at least as e?cient as the Non-Phase-
One plants. For the longitudinal analyses, the null hypotheses assume that the 1990 plants are at least as e?cient as the 1995 plants.
Results re?ect the cumulative expected rank sum of each group and the Z-score when a Wilcoxon Rank Sum test is employed. All
Z-scores are one sided.
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 571
Table 6
Summary and descriptive statistics
a
Variable A: 1990 Pooled B: 1995 Pooled C: Phase One D: Non-Phase-One
Mean Std Dev Mean Std Dev Mean Std Dev Mean Std Dev
Part I: Overall comparisons (N = 168)
Confounding
REG À0.0952 0.5411 À0.7142 0.5745 À0.3571 0.4756 À.1309 0.6525
STATE 0.3214 0.7285 0.3214 0.7285 0.2619 0.6403 0.3809 0.8027
AGE 32.92 15.5761 37.928 15.5761 35.2380 13.7737 35.61 17.5505
MKTSHR 0.0829 0.0855 0.0821 0.0858 0.0732 0.0671 0.0918 0.1000
Compliance
AVC 0.0133 0.0108 0.0106 0.0093 0.0159 0.0096 0.0081 0.0092
AVP 0.003 0.0051 0.0026 0.0037 0.0026 0.0034 0.0032 0.0053
BTUC 0.7967 0.3779 0.781 0.3843 0.9322 0.2260 0.6516 0.4471
SCR 0.1607 0.5721 0.2380 0.6678 0.2083 0.5979 0.1904 0.6469
PERMIT 0.8958 0.4938
Outcome
INDEX 0.5624 0.2338 0.5544 0.2283 0.5429 0.2155 0.5562 0.2660
E: 1990 Phase One F: 1995 Phase One G: 1990 Non-Phase One H: 1995 Non-Phase
One
Part II: Yearly comparisons (N = 168)
Confounding
REG À0.4761 0.4885 À0.2380 0.4650 À0.1428 0.6427 À0.1190 0.6659
STATE 0.2619 0.6422 0.2619 0.6422 0.3809 0.8051 0.3809 0.8051
AGE 32.73 13.5842 37.73 13.58 33.11 17.4227 38.1190 17.4227
MKTSHR 0.7433 0.0688 0.7220 0.0658 0.9155 0.0992 0.9219 0.1014
Compliance
AVC 0.0184 0.0096 0.0133 0.0089 0.0083 0.0094 0.0079 0.0090
AVP 0.0029 0.0040 0.0022 0.0027 0.0036 0.0061 0.0029 0.0045
BTUC 0.9403 0.140 0.9240 0.2384 0.6532 0.4470 0.6501 0.4498
SCR 0.1309 0.4854 0.2857 0.6867 0.1904 0.6489 0.1904 0.6489
PERMIT 0.5666 0.5471 0.3292 0.2780
Outcome
INDEX 0.6193 0.2370 0.5815 0.2159 0.6612 0.2770 0.6105 0.2655
a
This table re?ects the summary and descriptive data relative to the confounding variables that will be used to adjust the initial
(unadjusted) e?ciency indices. Panels A and B of Part I re?ect data when Phase One and Non-Phase One plants are pooled in 1990 and
1995. Panel C provides results when 1990 and 1995 Phase One are pooled. Panel D re?ects the same information for Non-Phase One
plants. Panels E and F of Part II provide data related to Phase One plants when 1990 and 1995 are considered separately. Panels G and
H re?ect the same information for the Non-Phase One plants. The variable de?nitions are:
REG Measures how favorably a state regulatory commission addresses rate increase requests.
STATE Measures the stringency of state environmentalism.
AGE The age of the plant.
MKTSHR The percentage of kilowatt hours of electricity produced by the operator of the plant in the NERC region in which the plant
operates.
AVC The average sulfur content of the coal used by the plant.
AVP The average sulfur content of the petroleum used by the plant.
BTUC The percent of heat (in BTUs) produced by the use of coal at the plant to produce kilowatt hours of electricity.
SCR The number of scrubbers at the plant.
INDEX The unadjusted data envelopment e?ciency index of the electricity plants. These scores lie between zero and one.
PERMIT Measures the e?ciency of the allocation and use of pollution permits. In 1990 and 1995, permits were allocated only to
Phase One.
572 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
is used to control for these potential contaminat-
ing variables; the residuals of the regressions are
used as the adjusted e?ciency indices. The pooled
and within-group regressions for cross-sectional
and longitudinal settings are presented in Tables
7 and 8, respectively. Table 9 presents the statisti-
cal analyses using the adjusted indices.
Pooled regressions
Pooled regressions are of two types: cross-sec-
tional and longitudinal. The pooled cross-sectional
regressions for 1990 reveal that all confounding
variables are signi?cant with negative coe?cients
(Table 7, Panel A). For 1995, only stringency and
market share are signi?cant. Thus, regulatory
climate, the stringency of state environmental reg-
ulations, plant age and market share have a nega-
tive e?ect on e?ciency levels. This may indicate,
for example, that the wrong kind of regulations
(climate and stringency) can have a depressive
e?ect on e?ciency (a common complaint of ?rms
facing environmental restrictions). It is reasonable
to presume that as a plant ages, e?ciency
decreases. The negative relationship between mar-
ket share and e?ciency is somewhat puzzling –
unless the operator has gone beyond the optimal
size. Finally, it is most interesting to note that when
plant type is added in the equation (Model 2), it is
found to be signi?cant with a negative coe?cient
(for both 1990 and 1995). This outcome provides
Table 7
Pooled regression analysis controlling for confounding variables
a
Model Intercept REG STATE AGE MKTSHR TYPE Log R
2
Model 1 (M1): EI
p
= b0 + b1REG + b2 STATE+b3AGE + b4MKTSHR + e
Model 2 (M2): EI
p
= b0 + b1REG + b2 STATE + b3AGE + b4MKTSHR + b5TYPE + e
Panel A: Cross-sectional pooled sample results, N = 168
M1: 1990 0.7574 À0.0823 À0.0517 À0.0032 À0.7579
15.33
***
À2.42
**
À1.83
*
À2.57
***
À3.35
***
À33.53 0.45
M2: 1990 0.8020 À0.0761 À0.0564 À0.0032 À0.7961 À0.0787
15.00
***
À2.25
**
À2.04
**
À2.60
***
À3.55
***
À2.07
**
À31.42 0.46
M1: 1995 0.6903 0.0142 À0.0849 À0.0010 À0.6199
12.60
***
0.417 À3.018
***
À0.850 À2.57
***
À33.36 0.45
M2: 1995 0.7315 0.0192 À0.0886 À0.0010 À0.6528 À0.0726
12.52
***
0.54 À.3174
***
À0.875 À2.73
***
À1.90
*
À31.57 0.46
Panel B: Longitudinal Pooled Sample Results, N = 168
M1: Phase-One 0.7332 0.0211 À0.0524 À0.0027 À0.9074
13.70
***
0.549 À1.79
*
À2.02
**
À3.37
***
À22.84 0.46
M2: Phase-One 0.7062 0.0204 À0.0496 À0.0033 À0.8931 0.0963
13.31
***
0.54 À1.74
*
À2.53
***
À3.40
***
2.71
***
À19.26 0.46
M1: Non-Phase-One 0.7489 À0.0698 À0.0583 À0.0021 À0.8164
12.75
***
À1.87
*
À1.85
*
À1.57 À3.33
***
À65.80 0.43
M2: Non-Phase-One 0.7254 À0.0704 À0.0573 À0.0024 À0.8235 0.0691
12.02
***
À1.90
*
À1.83
*
À1.78
*
À3.39
***
1.47 À64.71 0.44
a
This table provides the regression results for the regression results using pooled data. These regressions assess the e?ects of
confounding variables. Amounts in the tables are parameter and t-statistic values. Levels of signi?cance of p < 0.10, 0.05, and 0.01 are
denoted as *, **, and ***, respectively. The variable de?nitions are:
EI
p
The unadjusted data envelopment e?ciency score of the electricity plants (pooled indices). These indices lie between zero
and one.
REG Measures how favorably a state regulatory commission addresses rate increase requests.
STATE Measures the stringency of state environmentalism.
AGE The age of the plant.
MKTSHR The percentage of kilowatt hours of electricity produced by the operator of the plant in the NERC region in which the
plant operates.
TYPE Dummy variable equal to one if Phase One plant and zero if a Non-Phase One plant (for Panel A). Dummy variable equal to
one if 1995 and zero if 1990 (for Panel B).
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 573
evidence that the Non-Phase-One plants are more
e?cient than Phase-One plants, an outcome consis-
tent with the two cross-sectional hypotheses.
The pooled longitudinal regressions (Table 7,
Panel B) indicate that both Phase-One and Non-
Phase-One plants are a?ected by the confounding
variables over time, although the e?ects di?er.
The Phase-One plants have signi?cant negative
coe?cients for stringency, age, and market share,
while the Non-Phase-One plants have signi?cant
negative coe?cients for regulatory climate, strin-
gency, age, and market share. The presence of these
variables tends to decrease the relative e?ciency of
the plants, similar to the e?ects noted for cross-
sectional analysis. When year is added as a vari-
able, the 1995 Phase-One plants emerge with a
signi?cant positive coe?cient, supporting the pre-
diction that relative e?ciency is greater for 1995
plants than 1990 plants. However, the absence of
a signi?cant relationship for the Non-Phase-One
plants is inconsistent with the prediction that the
1995 Non-Phase-One plants are relatively more
e?cient than their 1990 counterparts.
Within-group regressions
Table 8 displays the within-group regressions
that allow the assessment of how the confounding
variables a?ect the within-group indices. The
regressions also provide the within-group adjusted
indices (residuals) that will be used in Table 9 for
testing to see if a common production function is
now possible (by comparing the within-group
adjusted indices with the corresponding adjusted
pooled indices). The Non-Phase-One plants have
two confounding variables in each year that are
signi?cant and negative (climate and age in 1990
and stringency and market share in 1995), while
the Phase-One plants only have market share with
a signi?cant negative coe?cient for 1995.
Statistical analyses with adjusted e?ciency indices
In Table 9, Panel A, initial tests results provide
evidence that is in favor of a common production
function for both cross-sectional and longitudinal
pooled data. Thus, the statistical comparisons of
the subsamples by type and by year are made using
the pooled indices (shown in Panel B of Table 9).
The statistical results in Panel B of Table 9 are
fully consistent with the predictions that the
Non-Phase-One plants, which are lower polluting,
are relatively more e?cient than the higher pollut-
ing Phase-One plants. Moreover, the predictions
that the 1995 plants are more e?cient than the
1990 plants are also fully supported. Thus, both
Table 8
Within-group regression analyses controlling for confounding variables
a
Model Intercept REG STATE AGE MKTSHR Log R
2
Model: EI
w
= b0 + b1REG + b2 STATE + b3AGE + b4MKTSHR + e
1990 0.6897 0.0419 À0.0568 0.0005 À0.5514
Phase-One 7.50
***
0.62 À1.09 0.23 À1.18 À32.46 0.45
1995 0.7808 0.0555 À0.0639 À0.0025 À0.9742
Phase-One 9.55
***
0.98 À1.54 À1.28 À2.45
***
À12.65 0.46
1990 0.9965 À0.1528 À0.0382 À0.0055 À1.045
Non-Phase One 10.61
***
À2.52
***
À0.75 À2.43
***
À2.73 À42.15 0.48
1995 0.8098 À0.0322 À0.1302 À0.0012 À0.7039
Non-Phase-One 8.79
***
À0.56 À2.72
***
À0.59 À1.86
*
À36.30 0.46
a
This table provides the regression results for the longitudinal pool (1990 and 1995) for Phase One and Non-Phase One plants.
Amounts in the tables are parameter and t-statistic values. Levels of signi?cance of p < 0.10, 0.05, and 0.01are denoted as *, **, and
***, respectively. The variable de?nitions are:
EI
w
The unadjusted data envelopment e?ciency score (within–group index) of the electricity plants. These indices lie between
zero and one.
REG Measures how favorably a state regulatory commission addresses rate increase requests.
STATE Measures the stringency of state environmentalism.
AGE The age of the plant.
MKTSHR The percentage of kilowatt hours of electricity produced by the operator of the plant in the NERC region in which the
plant operates.
574 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
ecoe?ciency and the Porter Hypothesis are sup-
ported by the statistical results.
DEA analysis using compliance variables
Pooled regressions
Table 10 gives the pooled regression results with
compliance variables. Regressing the DEA scores
on the compliance variables yields insights regard-
ing the measures taken to reduce pollution and
increase e?ciency. For the cross-sectional regres-
sions (Panel A) there are two variables that have
signi?cant negative coe?cients – for both 1990
and 1995: the average sulfur content of the petro-
leum used and the percentage of heat produced by
coal. The fact that these compliance variables
(which are associated with SO
2
releases) decrease
relative e?ciency as they increase in value is inter-
esting as it provides evidence that pollution is
indeed related to economic ine?ciency as claimed
by the Porter Hypothesis.
For the longitudinal regressions (Table 10,
Panel B), the same two variables (AVP and BTUC)
also emerge as negative and signi?cant for the Non-
Phase-One plants. These variables, however, are
not signi?cant for the Phase-One plants.
However, for the Phase-One plants scrubbers are
a signi?cant and positive variable. Thus, as scrub-
bers increase, relative e?ciency increases. Since
scrubbers reduce pollution, this again supports
the claim that as pollution decreases, e?ciency
increases.
Within-group regressions
Regressing the within-group indices on the
compliance variables also supplies some revealing
outcomes. Table 11 displays the results of these
regressions. In 1990, the Non-Phase-One plants
Table 9
Statistical analyses of adjusted e?ciency indices confounding variables
a
Cross-sectional analysis Distribution
b
Longitudinal analysis Distribution
b
Panel A: Adjusted EI
p
versus adjusted EI
w
1990 EI
p
7049 Phase-One EI
p
7591
1990 Phase-One EI
w
7147 1990 Phase-One EI
w
6605
Z-score À0.15 Z-score 1.56
1990 EI
p
7505 Phase-One EI
p
6899
1990 Non-Phase-One EI
w
6691 1995 Phase-One EI
w
7297
Z-score 1.28 Z-score À0.62
1995 EI
p
6701 Non-Phase-One EI
p
6787
1995 Phase-One EI
w
7495 1990 Non-Phase-One EI
w
7409
Z-score À1.25 Z-score À0.98
1995 EI
p
7524 Non-Phase-One EI
p
7409
1995 Non-Phase-One EI
w
6672 1995 Non-Phase-One EI
w
6787
Z-score 1.34 Z-score 0.98
Panel B: Adjusted EI
p
versus adjusted EI
p
1990 Phase One EI
p
6577 1990 Phase-One EI
p
6039
1990 Non-Phase One EI
p
7619 1995 Phase-One EI
p
8157
Z-score 1.65
*
Z-score 3.35
***
1995 Phase One EI
p
6464 1990 Non-Phase One EI
p
6453
1995 Non-Phase One EI
p
7732 1995 Non-Phase One EI
p
7743
Z-score 2.00
**
Z-score 2.04
**
a
This table provides the tests of relative e?ciency using the adjusted productive e?ciency scores when only confounding variables
are considered. Panel A reveals the presence of a common production function. As a consequence, the adjusted EI
p
indices are used to
conduct these tests and the results are reported in Panel B. Signi?cance levels of p < 0.10, p < 0.05, and p < 0.01 and are denoted as
*, **, and *** respectively.
b
For the cross-sectional hypotheses, the null hypotheses assume that the Phase-One plants are at least as e?cient as the Non-Phase-
One plants. For the longitudinal analyses, the null hypotheses assume that the 1990 plants are at least as e?cient as the 1995 plants.
Results re?ect the cumulative expected rank sum of each group and the Z-score when a Wilcoxon Rank Sum test is employed. All
Z-scores are one-sided.
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 575
continue to show signi?cant negative coe?cients
for petroleum sulfur and coal BTUs. Interestingly,
for Non-Phase-One plants the BTUC variable is
no longer signi?cant in 1995, perhaps indicating
a change in fuel mix for these plants, something
that was made possible by new blending fuel tech-
nology. For the Phase-One plants, the sulfur con-
tent of coal is a signi?cant negative coe?cient in
1990 and scrubbers produce a signi?cant positive
coe?cient. For 1995, scrubbers continue as a posi-
tive signi?cant coe?cient but the sulfur content of
coal is no longer signi?cant. The disappearance of
coal sulfur as a negative signi?cant factor from
1990 to 1995 is consistent with the new fuel blend-
ing technologies and other actions such as switch-
ing to lower sulfur coal. The persistence of
scrubbers as a signi?cant variable also signals their
role in achieving pollution reduction. Overall,
these results are consistent with the claims of the
Porter Hypothesis and ecoe?ciency.
Table 10
Pooled sample regression: compliance variables
a
Model Intercept AVC AVP BTUC SCR TYPE Log R
2
Model 1 (M1): EI
p
= b
0
+ b
1
AVC + b
2
AVP + b
3
BTUC + b
4
SCR + e
Model 2 (M2): EI
p
= b
0
+ b
1
AVC + b
2
AVP + b
3
BTUC + b
4
SCR + b
5
TYPE + e
Panel A: Cross-sectional pooled sample results (EI
p
), N = 168 (pooled by plant type)
M1: 1990 0.8123 0.2241 À11.9728 À0.2522 0.0393
14.82
***
0.095 À3.07
***
À3.63
***
1.08 À38.00 046
M2: 1990 0.8126 0.2913 À11.9672 À0.2516 0.0387 À0.0033
14.77
***
0.11 À3.00
***
À3.59
***
1.04 À0.07 À37.99 0.46
M1: 1995 0.7338 1.6398 À20.075 À0.1711 0.0280
13.18
***
0.651 À3.56
***
À2.61
***
0.35 À36.90 0.46
M2: 1995 0.7368 1.8609 À19.9024 À0.1591 0.0280 À0.0305
13.21
***
0.73 À3.54
***
À2.36
**
0.92 À0.72 À36.64 0.46
Model Intercept AVC AVP BTUC SCR PERMIT YEAR Log R
2
Model 1 (M1): EI
p
= b
0
+ b
1
AVC + b
2
AVP + b
3
BTUC + b
4
SCR + e
Model 2 (M2): EI
p
= b
0
+ b
1
AVC + b
2
AVP + b
3
BTUC + b
4
SCR + b
5
YEAR + e
Panel B: Longitudinal pooled sample results (EI
p
), N = 168 (pooled over years)
M1: 0.6192 À1.9423 À6.7058 À0.011 0.1092 À0.0289
Phase-One 5.85
***
À0.85 À1.04 À0.11 3.24
***
À0.70 À26.54 0.46
M2: 0.5647 À1.480 À5.413 À0.0127 0.1041 À0.0066 0.0511
Phase-One 4.92
***
À0.64 À0.83 À0.12 3.09
***
À0.14 1.18 À25.84 0.46
M1: 0.7451 4.992 À15.0425 À0.2182 À0.0421
Non-Phase-One 15.07
***
1.48 À3.28
***
À3.13
***
À1.09 À69.17 0.43
M2: 0.7217 5.0287 À14.7772 À0.2174 À0.0421 0.0434
Non-Phase-One 12.99
***
1.49 À3.22
***
À3.13
***
À1.09 0.90 À68.77 0.43
a
This table provides the regression results for the pooled regressions. Amounts in the tables are parameter and t-statistic values.
Levels of signi?cance of p < 0.10, 0.05, and 0.01 are denoted as *, **, and ***, respectively. The variable de?nitions are:
EI
p
The unadjusted data envelopment e?ciency score of the electricity plants (pooled indices). These indices lie between zero
and one.
AVC The average sulfur content of the coal used by the plant.
AVP The average sulfur content of the petroleum used by the plant.
BTUC The percent of heat (in BTUs) produced by the use of coal at the plant to produce kilowatt hours of electricity.
SCR The number of scrubbers at the plant.
TYPE Dummy variable equal to one if Phase One plant and zero if a Non-Phase One plant.
PERMIT Measures the e?ciency of the allocation and use of pollution permits (for Phase-One plants only).
YEAR Dummy variable equal to one if 1995 and zero if 1990.
576 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
Conclusions, limitations, and implications
This study investigates the relationships between
environmental performance and economic perfor-
mance in the electric utility industry in the United
States, where environmental performance is mea-
sured by levels of SO
2
and economic performance
is measured by relative e?ciency. The setting is of
particular interest because it is one where reduc-
tions were mandated without specifying how the
reductions were to be achieved, thus meeting the
requirements of the Porter Hypothesis. Under these
conditions, the Porter Hypothesis predicts that the
reductions can be achieved while maintaining or
improving economic e?ciency. Furthermore, given
the fact that a set of lower polluting plants was not
subject to the mandated reductions, it a?ords the
opportunity to see if ecoe?ciency is valid by com-
paring the lower polluting plants at a point in time
with the higher polluting plants (subject to the man-
date). The traditional view o?ers a competing
hypothesis which claims that any reductions in
SO
2
necessarily decrease economic e?ciency.
Adjusting for confounding variables, the evidence
indicates that lower polluting plants are relatively
more e?cient, both cross-sectionally and longitudi-
nally. Thus, the evidence is consistent with ecoe?-
ciency and the Porter Hypothesis. The support for
the Porter Hypothesis is reinforced by evidence con-
sistent with the use of new and innovative fuel
blending technologies to reduce SO
2
.
There are two very important limitations that
need to be mentioned. First, the study is for a
single industry and the results may not extend
beyond that industry. Second, we are measuring a
single aspect of environmental performance, albeit
a very important aspect. In this sense we duplicate
other studies that have attempted to assess the rela-
tionship between environmental performance and
economic performance. In our study, only a partial
environmental performance measure is used. Of
course, other studies look at di?erent measures of
environmental performance and it can be argued
that, collectively, the evidence may have more to
Table 11
Within-group sample regression: compliance variables
a
Model Intercept AVC AVP BTUC SCR Permit Log R
2
Within-group sample results (EI
w
) N = 168
Model 1 (M1): E
w
= b
0
+ b
1
AVC + b
2
AVP + b
3
BTUC + b
4
SCR + b
5
PERMIT + e
Model 2 (M2): E
w
= b
0
+ b
1
AVC + b
2
AVP + b
3
BTUC + b
4
SCR + e
Ml:1990 0.7541 0.7093 À7.525 À0.0909 0.1718 À0.0490
Phase-One 4.05
***
À1.83
*
À0.83 0.50 2.24
**
À0.80 À28.15 0.48
M2:1990 0.9863 7.7618 À16.70 À0.3623 À0.0890
Non-Phase-One 11.23
***
1.19 À2.46
***
À2.91
***
À1.22 À45.91 0.46
M1:1995 0.6483 À0.4049 À13.3757 À0.0131 0.0897 À0.0421
Phase-One 4.04
***
À0.11 À1.06 À0.08 2.08
**
À0.41 À15.96 0.46
M2:1995 0.7864 4.2128 À22.4140 À0.1455 À0.0236
Non-Phase-One 10.11
***
0.84 À2.70
***
À1.39 À0.42 À40.33 0.43
a
This table provides the regression results for the within-group regressions. Amounts in the tables are parameter and t-statistic
values. Levels of signi?cance of p < 0.10, 0.05, and 0.01 are denoted as *, **, and ***, respectively. The variable de?nitions are:
EI
w
The unadjusted data envelopment e?ciency score (within–group index) of the electricity plants. These indices lie between
zero and one.
MKTSHR The percentage of kilowatt hours of electricity produced by the operator of the plant in the NERC region in which the plant
operates.
AVC The average sulfur content of the coal used by the plant.
AVP The average sulfur content of the petroleum used by the plant.
BTUC The percent of heat (in BTUs) produced by the use of coal at the plant to produce kilowatt hours of electricity.
SCR The number of scrubbers at the plant.
PERMIT Measures the e?ciency of the allocation and use of pollution permits. In 1990 and 1995, permits were allocated only to
Phase One plants.
YEAR Dummy variable equal to one if 1995 and zero if 1990.
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 577
say than any individual study. Thus, we add to
this collective body further evidence of a positive
relation between environmental performance and
economic performance. Moreover, there is some
merit to arguing that a single very important mea-
sure of environmental degradation may be a good
proxy for the environmental performance of an
industry. Nonetheless, developing a more compre-
hensive measure of environmental performance
would be a signi?cant contribution, something per-
haps future research will produce.
As the evidence builds for ecoe?ciency, the
incentive increases for management to obtain
and use more information about environmental
activities and costs. This implies an increase in the
likelihood that management will demand better
environmental accounting, including the imple-
mentation of formal environmental management
accounting systems. After all, if the possibility exists
that paying more attention to environmental per-
formance will create more opportunities to increase
pro?ts, then a demand for environmental manage-
ment accounting systems seems logical. Indeed,
these systems are needed to facilitate management
decisions and corporate accountability (Bartolo-
meo et al., 2000; Epstein, Flamholtz, & McDon-
ough, 1976) and to aid in both external and
internal decision making (Bouma & Kamp-Roe-
lands, 2000; Everett & Neu, 2000; Gray, 2002).
However, while this study, other empirical studies,
and a number of case studies are supportive of eco-
e?ciency (voluntary or guided), the empirical evi-
dence is in its infancy and more studies are needed.
Like previous endeavors that delve into emerg-
ing cost accounting research agendas, investiga-
tions of the behavior and impact of environmental
costs have a long way to go. For instance, it is pos-
sible that resistance to the Porter view that pollu-
tion is a form of economic ine?ciency is similar in
nature to the much earlier Deming view that poor
product quality was a form of economic ine?-
ciency. While it is nowwidely accepted that it is pos-
sible to simultaneously improve product quality
and reduce quality costs, this view was not readily
accepted in the beginning. Evidence had to accumu-
late. Furthermore, more theoretical understanding
of the ecoe?ciency paradigm would also be useful.
For example, it would be interesting to model and
then empirically investigate the predictions of a bet-
ter-speci?ed environmental cost function – one that
speci?cally recognizes the presence of ecoe?cient
behavior. In addition, there may be value in explor-
ing how and when management decides to make a
commitment to improve the environment. For
instance, it may be that the bounded rationality
assumed by Porter is attributable mostly to many
managers being committed to the traditional envi-
ronmental cost paradigm. If ecoe?ciency is valid,
as evidence is beginning to suggest, a paradigmshift
may be needed to bring managers to a point where
they actively engage in actions that improve the
environment.
Acknowledgements
The authors would like to thank William F.
Bowlin, Marc Epstein, Kevin Hughes, Michael
Stein, and Annie McGowan for their informative
comments on previous drafts. The authors would
like to express a special note of thanks to the En-
ergy Information Administration of the Depart-
ment of Energy for its assistance procuring data
for this project. Finally, Royce D. Burnett is espe-
cially grateful for the ?nancial support provided
by the AIPCA Minority Doctoral Fellowship Pro-
gram and the PhD project.
References
Al-Tuwaijri, A. S., Christensen, T. E., & Hughes, K. E. (2004).
The relations among environmental disclosure, environ-
mental performance, and economic performance: a simul-
taneous equations approach. Accounting, Organizations and
Society, 29, 447–471.
Athanassopoulos, A. D., Lambroukos, N., & Sieford, L.
(1999). Data envelopment scenario analysis for setting
targets to electricity generating plants. European Journal of
Operations Research, 115, 413–428.
Banker, R. D., Charnes, A., & Cooper, W. W. (1984). Some
models for estimating technical and scale e?ciencies in data
envelopment analysis. Management Science, 30, 1078–1092.
Bebbington, J. (2001). Sustainable development: a review of the
international development, business and accounting litera-
ture. Accounting Forum, 25, 128–157.
Bartolomeo, M., Bennett, M., Bouma, J., Heydkamp, P.,
James, P., & Wolters, T. (2000). Environmental manage-
ment accounting in europe: current practice and future
potential. The European Accounting Review, 9(1), 31–52.
578 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
Bebbington, J., Gray, R., & Owen, D. (1999). Seeing the wood
for the trees: taking the pulse of social and environmental
accounting. Accounting, Auditing, and Accountability Jour-
nal, 12(1), 47–51.
Beets, S. D., & Souther, C. C. (1999). Corporate environmental
reports: the need for standards and environmental assurance
service. Accounting Horizons, 13(June), 129–145.
Berthelot, S., Cormier, D., & Magnan, M. (2003). Environ-
mental disclosure research: review and synthesis. Journal of
Accounting Literature, 22, 1–44.
Bhat, V. N. (1996). The green corporation: The next competitive
advantage. Westport, CT: Quorum Books.
Birkin, F., & Woodward, D. (1997a). Management accounting
for sustainable development Part 5: accounting for sustain-
able development. Management Accounting (UK), 75(10),
52–54.
Birkin, F., & Woodward, D. (1997b). Management accounting
for sustainable development Part 2: from economic to
ecological e?ciency. Management Accounting (UK), 75(7),
42–45.
Blacconiere, W. G., Johnson, M. F., & Johnson, M. S. (1997).
Stranded costs and competitive forces in the electric utility
industry. Journal of Accounting, Auditing, and Finance,
12(Summer), 199–221.
Bossong, E. M. (1999). Electric Utilities. Standard and Poors
Industry Survey, 167(32), 1–38.
Bouma, J., & Kamp-Roelands, N. (2000). Stakeholders’ expec-
tations of an environmental management system: some
exploratory research. The European Accounting Review,
9(1), 131–144.
Callen, J. L. (1991). Data envelopment analysis: partial survey
and applications for management accounting. Journal of
Management Accounting Research, 3(Fall), 35–55.
Charnes, A., Cooper, WW., & Rhodes, E. (1978). Measuring
the e?ciency of decision making units. European Journal of
Operation Research(2), 249–444.
Christensen, L. R., & Greene, W. H. (1976). Economies of scale
in U.S. electric power generation. Journal of Political
Economy, 84, 655–676.
Claggert, E. T., & Ferrier, G. D. (1998). The e?ciency of TVA
power distributors. Managerial and Decision Economics, 19,
365–378.
Cooper, W. W., Seiford, L., & Tone, K. (2000). Data
envelopment analysis: A comprehensive text with models,
applications, references and DEA-solver software. Kluwer
Boston MA: Academic Press.
Cowing, T. G., & Stevenson, R. E. (1981). Introduction:
productivity measurement in regulated industries. In T. G.
Cowing & R. E. Stevenson (Eds.), Productivity measurement
in regulated industries (pp. 3–15). New York, NY: Academic
Press.
De Villiers, C., & Van Staden, C. J. (2006). Can less
environmental disclosure have a legitimizing e?ect? Evi-
dence from Africa. Accounting, Organizations and Society,
31, 763–781.
Deegan, Craig (2002). Introduction: the legitimizing e?ect of
social and environmental disclosures – a theoretical foun-
dation. Accounting, Auditing, and Accountability Journal,
15(3), 282–311.
DeSimone, L. D., & Popo?, F. (1997). Eco-e?ciency: The
business link to sustainable development. MIT press.
Dierkes, M., & Preston, L. (1977). Corporate social accounting
reporting for the physical environment: A critical review
and implementation proposal. Accounting, Organizations
and Society, 2(1), 3–22.
Ditz, D., Ranganathan, J., & Banks, R. D. (1995). Green
ledgers: Case studies in corporate environmental accounting.
World Resources Institute.
D’Souza, J., Jacob, J., & Soderstrom, N. S. (2000). Nuclear
decommissioning costs: The impact of recoverability risk on
valuation. Journal of Accounting and Economics, 29, 207–
230.
Environmental Protection Agency. (1996). 1995 Acid Rain
Compliance Report: Overview. Washington, DC.
Environmental Protection Agency. (2000). Progress Report on
the EPA Acid Rain Program. Washington, DC.
Epstein, M., Flamholtz, E., & McDonough, J. (1976). Corpo-
rate social accounting in the United States of America: state
of the art and future prospects. Accounting, Organizations
and Society, 1(1), 23–42.
Epstein, M. J. (1996). Measuring corporate environmental
performance: Best practices for costing and managing an
e?ective environmental strategy. Chicago, IL: Irwin.
Everett, J., & Neu, D. (2000). Ecological modernization and the
limits of environmental accounting. Accounting Forum,
24(1), 5–29.
Fare, R. Grosskopf, & Tyteca, D. (1996). An activity analysis
model of the environmental performance of ?rms – appli-
cation to fossil-fuel-?red electric utilities. Ecological Eco-
nomics, 18, 161–175.
Farrell, M. J. (1957). The measurement of productive e?ciency.
Journal of Royal Statistical Society, 120(Series A, Part 3),
253–281.
Feldman, S. J., Soyka, P., & Ameer, P. (1997). Does improving
a ?rm’s environmental management system and environ-
mental performance result in a higher stock price? The
Journal of Investing(Winter), 87–97.
Fekrat, M., Inclan, I., & Petroni, D. (1996). Corporate
environmental disclosures: Comprehensive disclosure
hypothesis using 1991 annual report data. The International
Journal of Accounting, 175–195.
Fizel, J. L., & Nunnikhoven, T. S. (1992). Technical e?ciency
of for-pro?t and non-pro?t nursing homes. Managerial and
Decision Economics, 13(5), 429–439.
Freedman, M., & Wasley, C. (1990). The association between
environmental performance and environmental disclosure in
annual reports and 10ks. Advances in Public Interest
Accounting, 183–193.
Freedman, M., & Jaggi, B. (1992). An investigation of the long-
run relationship between pollution performance and eco-
nomic performance: The case of pulp and paper ?rms.
Critical Perspectives on Accounting, 3, 315–336.
Freedman, M., & Jaggi, B. (1994). analysis of the association
between pollution performance and input cost factors: the
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 579
case of electric utility plants. Journal of Accounting and
Public Policy, 13, 31–48.
Golany, B., Roll, Y., & Rybak, D. (1994). Measuring e?ciency
of power plants in Israel by data envelopment analysis.
IEEE Transactions on Engineering Management, 41(3),
291–301.
Goto, M., & TsuTsui, M. (1998). Comparison of productive
and cost e?ciencies among Japanese and US electric
utilities. Omega, International Journal of Management Sci-
ence, 26, 177–194.
Gray, R. H. (1992). Accounting and environmentalism: an
exploration of the challenge of gently accounting for
accountability, transparency, and sustainability. Account-
ing, Organizations and Society, 17, 399–425.
Gray, R., Kouhy, R., & Lavers, S. (1995). Corporate social and
environmental reporting: a review of the literature and a
longitudinal study of UK disclosure. Accounting, Auditing,
and Accountability Journal, 8(2), 47–77.
Gray, R., & Bebbington, J. (2000). Environmental accounting
managerialism and sustainability: is the planet safe in the
hands of business and accounting? Advances in Environ-
mental Accounting and Management(1), 1–44.
Gray, R. (2002). The social accounting project and accounting,
organizations and society privileging engagement, imagin-
ings, new accountings and pragmatism over critique?
Accounting, Organizations and Society, 27, 687–708.
Grosskopf, S., & Valdmanis, V. (1987). Measuring hospital
performance: a non-parametric approach. Journal of Health
Economics, 6, 89–107.
Hackston, D., & Milne, M. J. (1996). Some determinants of
social and environmental disclosures in New Zealnad
companies. Accounting, Auditing, and Accountability Jour-
nal, 77–108.
Hagg, I., & Hedlund, G. (1979). Case studies in accounting
research. Accounting, Organizations, and Society, 4(1/2),
135–143.
Haeri, H., Khawaia, M. S., & Perussi, M. (1997). Competitive
e?ciency: do mergers and ‘‘critical mass’’ really make a
di?erence? Public Utilities Forthnightly(June15), 26–33.
Haeri, H., Perussi, M., & Khawaja, M. S. (1999). The 100
fortnightly revisited: do utility stock prices re?ect operational
e?ciency? Public Utilities Fortnightly(February 15), 36–41.
Herbohn, K. (2005). Afull cost environmental accounting exper-
iment. Accounting, Organizations and Society, 30, 519–536.
Hughes, K. E. (2000). The value relevance of non?nancial
measures of air pollution in the electric utility industry. The
Accounting Review, 75(April), 209–228.
Hughes, S., Anderson, S., & Golden, S. (2001). Corporate
environmental disclosures: are they useful in determining
environmental performance? Journal of Accounting and
Public Policy, 3(20), 217–240.
Ilinitch, A. Y., Soderstrom, N. S., & Thomas, T. E. (1998).
Measuring corporate environmental performance. Journal
of Accounting and Public Policy, 17, 383–408.
Ingram, R., & Frazier, K. (1980). Environmental performance
and corporate disclosure. Journal of Accounting Research,
113–144.
Joshi, S., Krishnan, R., & Lave, L. (2001). Estimating the
hidden costs of environmental regulation. The Accounting
Review, 26(2), 171–198.
Jonson, L., Jonson, B., & Svensson, G. (1978). The applica-
tion of social accounting to absentism and personnel
turnover. Accounting, Organizations, and Society, 3(3/4),
261–268.
Juskow, P. L., & Schmalaensee, R. (1987). The performance of
coal-burning electric generating units in the United States,
1960 – 1980. Journal of Applied Economics, 2(April), 85–109.
Kaplan, R. S. (1986). The role for empirical testing in
management accounting. Accounting, Organizations, and
Society, 11(4/5), 429–452.
King, A., & Lenox, M. (2002). Exploring the locus of pro?table
pollution reduction. Management Science, 48(2 February),
289–299.
Knutson, K. (1995). Trends in utility fuel cost. Public Utility
Fortnightly(April 15), 12–13.
KPMG International International Survey of Corporate Sus-
tainability. (2005). , as of 20.1.2007.
Lanen, W. (1999). Waste minimization at 3M company: a ?led
study of non?nancial performance measurement. Journal of
Management Accounting Research, 11, 29–43.
Lehman, G. (2002). Global accountability and sustainability:
research prospects. Accounting Forum, 26(3), 219–232.
Lillis, A., & Mundy, J. (2005). Cross-sectional ?eld studies in
management accounting research – closing the gaps between
surveys and case studies. Journal of Management Accounting
Research, 17, 119–141.
Loudder, M., Khurana, I., & Boatsman, J. (1996). Market
valuation of regulatory assets in public utility ?rms. The
Accounting Review, 71, 161–182.
Majumadar, S. K. (1998). On the utilization of resources:
perspectives from the U.S. telecommunications industry.
Strategic Management Journal, 19, 809–831.
Margolis, J., & Walsh, J. (2003). Miersy loves companies:
rethinking social initiatives by business. Administrative
Science Quarterly, 48, 268–305.
Mathews, M. (1997). Twenty-?ve years of social and environ-
mental research: is there a silver jubilee to celebrate?
Accounting, Auditing, & Accountability Journal, 10(4),
481–531.
Meyer, S. (1992). The economic impact of environmental
regulation. Working Paper. Massachusetts Institute of
Technology.
Moneva, J., Archel, P., & Correa, C. (2006). GRI and the
camou?aging of corporate unsustainability. Accounting
Forum, 30, 121–137.
Murty, M. N., & Kumar, S. (2003). Win–win opportunities and
environmental regulation: testing of the porter hypothesis
for Indian manufacturing industries. Journal of Environ-
mental Management, 67, 139–144.
Neu, D., Warsame, H., & Pedwell., K. (1998). Managing public
impressions: environmental disclosures in annual reports.
Accounting, Organizations and Society, 23(3), 265–282.
580 R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581
Olatubi, W. O., & Dismukes, D. E. (2000). A data envelopment
analysis of the levels and determinants of coal ?red electric
power generating performance. Utilities Policy, 9, 47–59.
Patten, D. (1992). Intra-industry environmental disclosures in
response to the Alaskan oil spill: a note on legitimacy
theory. Accounting, Organizations, and Society, 471–475.
Patten, D. (1991). Exposure, legitimacy, and social disclosure.
Journal of Accounting and Public Policy, 10, 297–308.
Patten, D. (2002). The relation between environmental perfor-
mance and environmental disclosure. Accounting, Organi-
zations, and Society, 27, 763–773.
Pelz, E., & Fitzgerald, T. J. (2001). From market failure to
market-based solution: policy lessons from clean air legis-
lation. Economic Commentary.
Porter, M. E. (1991). America’s green strategy. Scienti?c
American(April), 168.
Porter, M. E., & van der Linde, C. (1995a). Toward a new
conception of the environment-competitiveness relation-
ship. Journal of Economic Perspectives, 9(Fall), 97–118.
Porter, M. E., & van der Linde, C. (1995b). Green and
competitive: ending the stalemate. Harvard Business
Review(September–October), 120–134.
Puig-Junoy, J. (1998). Technical e?ciency in the clinical
management of critically ill patients. Health Economics, 7,
263–277.
Rockness, J. (1985). An assessment of the relationship between
U.S. corporate environmental performance and disclosure.
Journal of Business Finance and Accounting, 12, 339–354.
Rubenstein, D. (1992). Bridging the gap between green
accounting and black ink. Accounting, Organizations and
Society, 17(5), 501–508.
Ruch, D., & Roper, J. J. (1992). The greening of corporate
Canada. CMA Magazine(December–January), 15–18.
Savage, D. E., Jasch, C., & International Federation of
Accountants, 2005. International guidance document: envi-
ronmental management accounting. 28.2.2007.
Schmidheiny, S., & Zorraquin, F. J. L. (1996). Financing
change. MIT Press.
Shank, J. K., & Govindarajan, V. (1994). Measuring the cost of
quality: a strategic cost management perspective. Journal of
Cost Management(Summer), 5–17.
Solomons, D. (1991). Accounting and social change: a neutralist
view. Accounting, Organizations and Society, 16(3), 287–295.
Stone, D. (1995). No longer at the end of the pipe, but still a long
way fromsustainability: a look at management accounting for
the environment and sustainable development in the United
States. Accounting Forum, 19(2/3), 95–110.
Swift, B. (2001). How environmental laws work: an analysis of
the utility sector’s response to regulation of nitrogen oxides
and sulfur dioxides under the clean air act. Tulane Environ-
mental Law Journal, 14(Summer), 309–424.
Tyteca, D. (1996). On the measurement of the environmental
performance of ?rms – a literature review and a productive
e?ciency perspective. Journal of Environmental Manage-
ment, 46, 281–308.
Tyteca, D. (1997). Linear programming models for the
measurement of environmental performance of ?rms –
concepts and empirical results. Journal of Productivity
Analysis, 8, 183–197.
Tyteca, D. (1999). Sustainability indicators at the ?rm level.
Journal of Industrial Ecology, 2, 61–77.
Ullman, A. (1976). The corporate environmental accounting
system: a management tool for ?ghting environmental
degradation. Accounting, Organizations and Society, 1(1),
71–79.
United Nations Division for Sustainable Development. (2001).
Environmental management accounting, procedures and prin-
ciples. accessed 9.3.2007.
United States Department of Energy, Energy Information
Administration (EIA). (1994). Electric utility phase I acid
rain compliance strategies for the clean air act amendments
of 1990. DOE/EIA-0578(94), Washington, DC.
United States Department of Energy, Energy Information
Administration (EIA). (1997). The e?ects of title IV of the
clean air act amendments of 1990 on electric utilities: an
update, DOE/EIA-0582(97). Washington, DC.
Walden, W. D., & Schwartz, B. N. (1997). Environmental
disclosures and public policy pressure. Journal of Accounting
and Public Policy, 16(2), 125–1554.
White, G. (2005). How to report a company’s sustainability
activities. Management Accounting Quarterly, 7(1), 36–43.
Whiteman, J. L. (1995). Benchmarking developing country
electricity systems using data envelopment analysis. Asia-
Paci?c Economic Review, 1(3), 71–78.
Wildavsky, Aaron (1994). Accounting for the environment.
Accounting, Organizations and Society, 19(4/5), 461–481.
Wiseman, Joanne (1982). An evaluation of environmental
disclosures made in corporate annual reports. Accounting,
Organizations and Society, 7(1), 53–63.
Wong, E. A. (2001). How to develop a database for forecasting
environmental expenditures. Strategic Finance(March),
52–56.
World Business Council for Sustainable Development. (2000a).
Eco-e?ciency: creating more value with less input. North
Yorkshire, UK.
World Business Council for Sustainable Development. (2000b).
Measuring eco-e?ciency: a guide to reporting company
performance. North Yorkshire, UK.
Yaisawarng, S., & Klein, J. D. (1994). The e?ects of sulfur
dioxide controls on productivity changes in the U.S. electric
power industry. The Review of Economics and Statistics,
76(August), 447–460.
R.D. Burnett, D.R. Hansen / Accounting, Organizations and Society 33 (2008) 551–581 581
doc_137172020.pdf