Description
Previous empirical evidence provides mixed results on the relationship between corporate environmental performance
and the level of environmental disclosures. We revisit this relation by testing competing predictions from economics
based and socio-political theories of voluntary disclosure using a more rigorous research design. In
particular, we improve on the prior literature by focusing on purely discretionary environmental disclosures and by
developing a content analysis index based on the Global Reporting Initiative sustainability reporting guidelines to
assess the extent of discretionary disclosures in environmental and social responsibility reports.
Revisiting the relation between environmental
performance and environmental disclosure:
An empirical analysis
Peter M. Clarkson
a,b
, Yue Li
c
, Gordon D. Richardson
c,
*
, Florin P. Vasvari
d
a
UQ Business School, The University of Queensland, Australia
b
Faculty of Business Administration, Simon Fraser University, Canada
c
Joseph Rotman School of Management, University of Toronto, 105 St. George Street, Toronto, Ontario, Canada M5S 3E6
d
London Business School, University of London, London, United Kingdom
Abstract
Previous empirical evidence provides mixed results on the relationship between corporate environmental perfor-
mance and the level of environmental disclosures. We revisit this relation by testing competing predictions from eco-
nomics based and socio-political theories of voluntary disclosure using a more rigorous research design. In
particular, we improve on the prior literature by focusing on purely discretionary environmental disclosures and by
developing a content analysis index based on the Global Reporting Initiative sustainability reporting guidelines to
assess the extent of discretionary disclosures in environmental and social responsibility reports. This index better cap-
tures ?rm disclosures related to its commitment to protect the environment than the indices employed by prior studies.
Using a sample of 191 ?rms from the ?ve most polluting industries in the US, we ?nd a positive association between
environmental performance and the level of discretionary environmental disclosures. The result is consistent with the
predictions of the economics disclosure theory but inconsistent with the negative association predicted by socio-political
theories. Nevertheless, we show that socio-political theories explain patterns in the data (‘‘legitimization’’) that cannot
be explained by economics disclosure theories.
Ó 2007 Elsevier Ltd. All rights reserved.
Introduction
An unresolved research issue in environmental
accounting is the empirical association between
the level (i.e., amount) of corporate environmental
disclosures and corporate environmental perfor-
mance (Al-Tuwaijri, Christensen, & Hughes,
0361-3682/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aos.2007.05.003
*
Corresponding author. Tel.: +1 416 946 8601.
E-mail addresses: [email protected] (P.M.
Clarkson), [email protected] (Y. Li), gordon.richard-
[email protected] (G.D. Richardson), fvasvari@london.
edu (F.P. Vasvari).
www.elsevier.com/locate/aos
Available online at www.sciencedirect.com
Accounting, Organizations and Society 33 (2008) 303–327
2004; Hughes, Anderson, & Golden, 2001; Patten,
2002). Accounting standard setters and securities
regulators are increasingly being made aware of
de?ciencies in corporate environmental disclosures
(Beets & Souther, 1999; Chan-Fishel, 2002;
Franco, 2001). The results of previous studies on
the relation between corporate environmental per-
formance and environmental disclosure in ?nan-
cial reports have been mixed. Patten (2002)
attributes the failure to ?nd a signi?cant and con-
sistent relation between environmental perfor-
mance and environmental disclosure to problems
in the research designs of existing research. These
problems include failure to control for other fac-
tors associated with the level of environmental
disclosure, inadequate sample selection, and inad-
equate measures of environmental performance
and disclosure.
This study seeks to revisit the relation between
environmental performance and the level of envi-
ronmental disclosure using a more rigorous
research design. We test two competing predic-
tions about the level of voluntary environmental
disclosures. Voluntary disclosure theory (Dye,
1985; Verrecchia, 1983) predicts a positive associa-
tion between environmental performance and the
level of discretionary environmental disclosure.
The notion is that superior environmental per-
formers will convey their ‘‘type’’ by pointing to
objective environmental performance indicators
which are di?cult to mimic by inferior type ?rms.
Inferior performers will choose to disclose less or
to be ‘‘silent’’ on their environmental performance,
thus being placed in a pool of ?rms where inves-
tors and other users ascribe the ‘‘average type’’
to that pool. What sustains this partial disclosure
equilibrium is proprietary costs associated with
disclosure about environmental performance (Ver-
recchia, 1983) and uncertainty as to whether the
?rm is informed regarding its type (Dye, 1985).
Socio-political theories including political econ-
omy, legitimacy theory, and stakeholder theory
(Patten, 2002), on the other hand, predict a negative
association between environmental performance
and the level of discretionary environmental disclo-
sures. These overlapping theories suggest that
social disclosure is a function of social and political
pressures facing the corporation. To the extent that
poor environmental performers face more political
and social pressures and threatened legitimacy,
they will attempt to increase discretionary environ-
mental disclosures to change stakeholder percep-
tions about their actual performance. Thus, we
have competing directional predictions from alter-
native theories, and the observed direction of asso-
ciation between environmental performance and
the level of discretionary disclosures will eliminate
one of the two predictions.
The predictions of the above theories relate to
discretionary, not mandatory, environmental dis-
closures. Previous studies assessed environmental
disclosures mainly from annual reports and other
regulatory ?lings such as 10 Ks and many of those
studies rely on a Wiseman (1982) based content
analysis index to measure the extent of environ-
mental disclosures. The Wiseman index focuses
on the ?nancial consequences of corporate envi-
ronmental activities and puts more weight on
quantitative disclosures. Using this measure, poor
environmental performers may actually have
higher disclosure scores than good performers
because they have greater exposures and must dis-
cuss any material ?nancial information in their
regulatory ?lings such as annual reports and
10 Ks. This may partially explain the inconclusive
?ndings in the previous literature and why Patten
(2002) ?nds a negative relation between environ-
mental disclosure and a toxics release inventory
(TRI) based environmental performance
indicator.
1
In collaboration with an environmental disclo-
sure expert, we develop a content analysis index
based on the global reporting initiative guidelines
(GRI) to assess the level of discretionary environ-
mental disclosures in environmental and social
responsibility reports or similar disclosures pro-
vided on the ?rm’s web site. This index di?ers from
Wiseman (1982) index, previously used in the liter-
ature, because we focus on ?rm disclosures related
to its commitment to protect the environment. Our
index potentially allows investors, regulators, and
1
Patten (2002) was aware of the problem of non-discretion-
ary disclosures in annual reports and dropped litigation
disclosures as a partial attempt to deal with this (see p. 768).
304 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
environmental stakeholder groups to infer envi-
ronmental performance ‘‘type’’ from the disclosure
score. This is valuable to users who seek to assess
the ?rms’ true environmental commitment and
related environmental exposures.
We focus on the 2003 environmental disclosures
of 191 ?rms drawn from ?ve industries: Pulp and
Paper, Chemicals, Oil and Gas, Metals and Min-
ing, and Utilities. These ?ve industries are consid-
ered to have a high pollution propensity and have
collectively been the subject of a whole range of
environmental regulations in the US in the past
30 or more years. The magnitude of the environ-
mental spending by these industries to comply
with the environmental regulations and the impact
of their operations on the natural environment
should be a major concern to investors and other
environmental stakeholder groups. Thus, ?rms in
these industries collectively form an ideal sample
to test the competing predictions of voluntary dis-
closure and socio-political theories.
In brief, our results are as follows. We ?nd a
positive association between environmental per-
formance and the level of discretionary disclosures
in environmental and social reports or related web
disclosures. In other words, superior environmen-
tal performers are more forthcoming in truly dis-
cretionary disclosure channels, as predicted by
the economics based voluntary disclosure theory.
Our result is inconsistent with the prediction of a
negative association arising from socio-political
theories such as legitimacy theory and stakeholder
theory. Further, using the Janis–Fadner coe?cient
of imbalance as a direct measure of perceived legit-
imacy, we fail to observe the negative association
between legitimacy and the level of disclosures
implied by socio-political theories.
2
Thus, our
results suggest that socio-political theories are
not robust in predicting the level of discretionary
environmental disclosures.
We do, however, ?nd that socio-political theo-
ries are helpful in predicting what is being said,
which moves the focus of enquiry beyond the sim-
ple level of discretionary disclosure. Using the
ratio of soft disclosure scores to total awarded
scores as a proxy for ‘‘legitimization’’, we show
that ?rms with unfavorable prior year media cov-
erage are more likely to make soft claims to be
committed to the environment which are not read-
ily veri?able. This behavior is not predicted by
economic disclosure theories, which assume
truth-telling. Thus, socio-political theories do
indeed explain additional patterns in the data.
The paper is organized as follows. Following lit-
erature review and hypothesis development, we
describes our content analysis disclosure index
and the measures of environmental performance
which we use in the study. We then present our
econometric model and preliminary empirical evi-
dence. The sections ‘Empirical results involving
the level of disclosure’ and ‘A revised role for
socio-political theories’ contain the main results
followed by sensitivity analysis. The ?nal section
summarizes the main ?ndings of the study with a
discussion of implications for future research.
Literature review and hypothesis development
Literature review
The existing literature in environmental
accounting research can be categorized into three
broad groups. The ?rst group of studies examines
the valuation relevance of corporate environmen-
tal performance information and has found that
such information is valuable to investors seeking
to assess environmental liabilities in di?erent set-
tings.
3
The second line of literature examines fac-
tors a?ecting managerial decisions to disclose
potential environmental liabilities. This group of
studies ?nds that there are strategic factors a?ect-
ing ?rms’ decisions to disclose environmental
liability information, especially when disclosures
2
The Janis–Fadner coe?cient of imbalance measures the
propensity of a ?rm’s prior year press articles pertaining to the
environment to be unfavorable. See section ‘Other control
variables speci?c to environmental disclosures’ for a detailed
de?nition.
3
See Cormier, Magnan, and Morard (1993), Blacconiere
and Patten (1994), Barth and McNichols (1994), Cormier and
Magnan (1997), Li and McConomy (1999), Richardson and
Welker (2001), & Clarkson, Li, and Richardson (2004).
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 305
are discretionary.
4
A third line of studies, one
which we discuss in the following paragraphs since
it is most relevant to this study, explores the rela-
tion between environmental disclosures and envi-
ronmental performance.
Ingram and Frazier (1980) examine the associa-
tion between the content of corporate environmen-
tal disclosure and corporate environmental
performance. The study was concerned with a lack
of corporate social responsibility disclosures in
annual reports due to their voluntary nature. The
authors scored environmental disclosures in 20
pre-selected content categories along four dimen-
sions; evidence, time, speci?city, and theme.
Ingram and Frazier (1980) proxied environmental
performance by a performance index devised by
the Council on Economic Priorities (CEP), a
non-pro?t organization specializing in the analy-
sis of corporate social activities. Forty ?rms
were selected from the 50 ?rms that were moni-
tored by the CEP. Regression results indicated
no association between environmental disclosure
and environmental performance, consistent with
authors’ prior expectation about an overall poor
quality of environmental disclosures in annual
reports.
Wiseman (1982) examines the extent of volun-
tary environmental disclosures made by corpora-
tions in their annual reports using a research
design almost identical to Ingram and Frazier
(1980). The study focuses on the 26 largest US
companies that were monitored by the CEP for
the 1972–1976 period. Wiseman designed an envi-
ronmental disclosure index covering 18 items in
four categories: economic factors (5 items), envi-
ronmental litigation (2 items), pollution abatement
activities (5 items), and environmental disclosures
that do not fall into the other three (6 items). In
addition, Wiseman assigned a score to each item
based on whether the disclosure is quantitative or
qualitative (3 for quantitative disclosure, 2 for
non-quantitative disclosure, 1 for mentioning in
general terms, 0 for no disclosure).
5
The CEP
rankings were used as a proxy for environmental
performance. Spearman rank order correlation
indicates that there is no signi?cant association
between the CEP environmental performance
rankings and the Wiseman environmental disclo-
sure index rankings.
Freedman and Wasley (1990) examine the rela-
tionship between corporate pollution performance
and pollution disclosures made in annual reports
and 10 K reports ?led with the SEC. Their sample
consists of 50 US companies in four industries
(Steel, Oil, Pulp and Paper, Electric Utilities).
Again, the CEP rankings are used as a proxy for
environmental performance. The authors measure
environmental disclosures in both annual and
10 K reports using the same indexing procedure
developed by Wiseman (1982). Spearman rank
order correlation tests are conducted to examine
the associations both between annual report dis-
closure indices and the CEP indices, and between
10 K disclosure indices and the CEP indices. The
results indicate that neither annual report environ-
mental disclosures nor the 10 K environmental
disclosures are indicative of ?rms’ actual environ-
mental performance.
Bewley and Li (2000) examine factors associ-
ated with the environmental disclosures in Canada
from a voluntary disclosure theory perspective.
The authors measure environmental disclosures
by 188 Canadian manufacturing ?rms in their
1993 annual reports using the Wiseman index. A
?rm’s pollution propensity (i.e., environmental
performance) is proxied by their industry member-
ship and by whether they report to the Ministry of
Environment under the National Pollution
Release Inventory program. The study ?nds that
?rms with more news media coverage of their envi-
ronmental exposure, higher pollution propensity,
and more political exposure are more likely to dis-
4
See Patten (1992), Li, Richardson, and Thornton (1997),
Barth, McNichols, and Wilson (1997), Li and McConomy
(1999), & Aerts, Cormier, and Magnan (2006).
5
Many environmental disclosure studies since then rely on
the Wiseman index in order to measure the extent of corporate
environmental disclosures. Few recognize the fact that the
Wiseman index places a heavy weight on the ?nancial conse-
quences of corporate environmental activities, most of which
are required disclosures in 10 Ks for public companies regis-
tered with the SEC.
306 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
close general environmental information, suggest-
ing a negative association between environmental
disclosures and environmental performance.
Hughes et al. (2001) examine environmental dis-
closures made by 51 US manufacturing ?rms for
1992 and 1993. Again, the authors use a slightly
modi?ed Wiseman index to measure environmen-
tal disclosures made within the President’s letter,
MD&A, and note section, and then assess whether
environmental disclosures are consistent with envi-
ronmental performance ratings (good, mixed, and
poor) by the CEP. Although the study ?nds no dif-
ference in environmental disclosures between good
and mixed groups, ?rms rated as poor environ-
mental performers by the CEP tend to make sub-
stantially more environmental disclosures under
the Wiseman disclosure index. The authors attri-
bute this ?nding to increased scrutiny in 1992
and 1993 by the FASB and SEC with respect to
environmental disclosures, which forces poor per-
formers to make more disclosure as they are sub-
ject to more remediation activities.
Patten (2002) identi?es three issues in the previ-
ous studies in this area; (1) failure to control for
other factors, (2) inadequate sample selection;
and (3) inadequate measures of environmental per-
formance. Since the CEP only followed a small
group of ?rms in only four industries, reliance on
the CEP for sample selection may be problematic.
In addition, the CEP did not use the same criteria
and consistent methodology to assess corporate
environmental performance in di?erent industries.
To overcome this issue, Patten uses TRI data, nor-
malized by sales, to proxy for environmental per-
formance. Using a sample of 131 US ?rms from
24 di?erent industries, and a modi?ed Wiseman
index measure and line count of environmental
disclosures in 1990 annual reports, Patten ?nds
that TRI/sales are positively associated with both
measures of environmental disclosures, suggesting
a negative relation between environmental perfor-
mance and environmental disclosures.
More recently, Al-Tuwaijri et al. (2004) explore
the relations among environmental disclosure,
environmental performance and economic perfor-
mance using a simultaneous equations approach.
Similar to Patten (2002) and Al-Tuwaijri et al.
(2004) use TRI based data to assess environmental
performance. Speci?cally, they assess environmen-
tal performance as the percentage of total waste
generated that is recycled. The authors measure
environmental disclosure using a content analysis
in four categories (potential responsible parties’
designation, toxic waste, oil and chemical spills,
and environmental ?nes and penalties). These dis-
closures are largely non-discretionary, in contrast
to the discretionary disclosures which we examine.
They ?nd a positive association between environ-
mental performance and environmental disclosure.
In summary, the existing studies ?nd mixed
results on the relation between environmental dis-
closure and environmental performance. One rea-
son for the inconclusive ?ndings is due to the
choice of non-discretionary disclosure channels
and use of the Wiseman (1982) index. It is self evi-
dent that, as environmental problems and expo-
sures increase, non-discretionary disclosures in
regulated channels such as annual reports and
10 Ks should increase. Thus, a negative associa-
tion between environmental performance and con-
tent analysis scores in annual reports and 10 Ks
may be driven by non-discretionary disclosures.
The disclosure theories reviewed in the next sec-
tion, on the other hand, apply to discretionary dis-
closures. Thus, inferences about the robustness of
these theories are confounded when the disclosure
media are formal channels like the annual report
and the 10 K. Adding to this problem is the heavy
weight the Wiseman index places on disclosures
about the ?nancial consequences of environmental
activities, whereas our index places more weight on
disclosures that reveal true (but unobservable)
environmental performance. By focusing exclu-
sively on environmental and social responsibility
reports or similar disclosures on ?rms’ web sites,
locations where disclosures are purely discretion-
ary, and with an index that aims at revealing per-
formance ‘‘type’’, we enhance the reliability of
inferences about the true direction of association
between environmental performance and discre-
tionary disclosure.
Hypothesis development
The voluntary disclosure literature suggests that
companies have incentives to disclose ‘‘good
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 307
news’’ to di?erentiate themselves from companies
with ‘‘bad news’’ in order to avoid the adverse
selection problem (Dye, 1985; Verrecchia, 1983).
While these theories do not pertain to environmen-
tal performance per se, they are applied to this set-
ting by Bewley and Li (2000) and Li et al. (1997).
Companies with superior environmental perfor-
mance due to their proactive environmental strat-
egy have incentives to inform investors and other
stakeholders of their strategy by voluntarily dis-
closing more environmental information. Put sim-
ply, they seek to reveal their performance type,
something not directly observable to investors
and other stakeholders, through direct voluntary
disclosures that cannot be easily mimicked by poor
environmental performers. In doing so, they
potentially increase ?rm valuation since knowl-
edgeable investors will infer that exposures and
latent environmental liabilities are lower for good
as opposed to poor environmental performers.
6
Thus, voluntary disclosure theory predicts a
positive association between environmental per-
formance and the level of discretional environmen-
tal disclosure.
Turning to the predictions of socio-political the-
ories, Gray, Kouhy, and Lavers (1995) and Lindb-
lom (1994) argue that companies whose social
legitimacy is threatened have incentives to increase
environmental disclosures to: (1) educate and
inform relevant publics about (actual) changes in
their performance, (2) change perceptions about
their performance, (3) de?ect attention from the
issue of concern by highlighting other accomplish-
ments, and (4) seek to change public expectations
of their performance.
7
According to Patten (2000,
2002), socio-political theories predict a negative
association between corporate environmental per-
formance and level of discretional environmental
disclosure. Thus, the two competing theories pro-
vide opposite predictions on how environmental
performance may a?ect discretionary environmen-
tal disclosure strategies. Our hypotheses now fol-
low (stated in the alternate form):
H1a: Environmental performance and the level of
discretionary environmental disclosures are
positively associated, as implied by econom-
ics based voluntary disclosure theories.
H1b: Environmental performance and the level of
discretionary environmental disclosures are
negatively associated, as implied by the
socio-political theories.
Research design
Environmental performance indicators
A key research design issue in this study is to
develop a reliable proxy for a ?rm’s environmental
performance. The di?culty in assessing environ-
mental performance is well documented in the lit-
erature (see, for example, Ilinitch et al., 1998).
Since we seek to assess relative environmental per-
formance in this study, we follow the existing liter-
ature and develop our relative environmental
performance proxy using the actual pollution dis-
charge data from the US Environmental Protec-
tion Agency’s (EPA) TRI database (King &
Lenox, 2001). Speci?cally, we ?rst aggregate the
total toxic releases (in pounds) and the toxic waste
treated or processed for each of our sample ?rms
in 2003, as reported by the EPA in 2005 (EPA
annually reports the data at the plant level, with
a two year lag). To verify the accuracy of our
aggregation procedure, we compared the TRI
measures obtained at the ?rm level with those pro-
vided by the Investor Responsibility Research
Center in their proprietary database. We found
that the di?erences were minor.
Our ?rst measure is the total toxic waste that is
treated, recycled or processed as a percentage of
the total toxic waste generated by each ?rm (%
recycled). This measure is similar to the one used
6
The assumption that environmental performance ‘‘type’’ is
not readily observable to investors and other stakeholders is
supported by the di?culties environmental researchers have
had obtaining reliable measures of environmental performance
that are comparable across companies in the same industry and
across industries (see Al-Tuwaijri et al., 2004 & Ilinitch,
Soderstrom, & Thomas, 1998).
7
Socio-political theories include political economy, legiti-
macy theory and stakeholder theory (see Patten, 2002). We do
not di?erentiate them in this study as they have the same
prediction with respect to the relation between environmental
performance and environmental disclosure.
308 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
by Al-Tuwaijri et al. (2004). We supplement our
analysis with an alternative measure, the ratio of
TRI to total ?rm sales. This gives us pounds of
toxic releases per thousand dollars of sales for each
of our sample ?rms (TRI/sales). However, if ?rms
are not homogenous in terms of production pro-
cesses within one of our ?ve industries (Pulp and
Paper, Chemicals, Oil and Gas, Metals and Min-
ing, and Utilities), such that TRI/Sales is not
directly comparable across ?rms in that industry,
the recycling measure reduces this noise and facil-
itates pooling.
Since industry pollution propensity di?ers sig-
ni?cantly, we alternatively rank the above two
measures within each industry. Thus, the %Recy-
cled ranks and TRI/sales ranks are our proxies
for a ?rm’s relative performance within its indus-
try. Similar relative performance measures are also
used in Clarkson et al. (2004).
Environmental disclosure index
As discussed in section ‘Literature review and
hypothesis development’, from the economic the-
ory perspective, superior environmental perfor-
mance (EP) types are hypothesized to seek
credible direct disclosures to reveal their (unob-
servable) performance type. A crucial property of
these disclosures is that they focus on objective,
‘‘hard’’ measures that cannot be easily mimicked
by poor environmental performers. Thus, reliable
inferences about theories like Verrecchia (1983)
and Dye (1985) require a content analysis disclo-
sure index that puts a heavy emphasis on objective
measures of performance as opposed to soft (i.e.,
not easily veri?able) claims to be committed to
the environment.
8
As a simple example, consider
a good and a poor EP type ?rm in the same indus-
try. The good EP ?rm will voluntarily disclose
objective measures of environmental impact (e.g.,
quantitative environmental performance indica-
tors) and will benchmark its performance relative
to the industry, something the poor EP ?rm will
not want to do. Thus, the good EP ?rm will
emphasize discretionary disclosures that are hard
to mimic. We assume these hard disclosures are
truthful, in that a ?rm would face litigation expo-
sure if caught lying by informed stakeholders in
social responsibility reports or web related
disclosures.
9
What is striking in all this is that there is a
demand by environmental stakeholders for pre-
cisely the same thing: hard, objective measures of
environmental performance in social responsibility
reports, so that poor EP performers cannot mimic
good EP performers by soft, unveri?able claims to
be committed to the environment. Indeed, good
EP performers and environmental stakeholders
have joined forces to develop standards for ?rms
preparing social responsibility reports that put a
premium on hard, objective measures.
The Global Reporting Initiative (GRI) was
launched in 1997 as a joint initiative of Coalition
for Environmentally Responsible Economies, a
US non-government organization and the United
Nations Environmental Program. The overall goal
of the initiative is to develop a globally accepted
reporting framework to enhance the quality, rigor,
and utility of sustainability reporting (Global
Reporting Initiative, 2002). The GRI Guidelines
follow 11 principles (transparency, inclusiveness,
auditability, completeness, relevance, sustainabil-
ity context, accuracy, neutrality, comparability,
clarity, and timeliness) to ensure that sustainability
reports (1) present a reasonable and balanced
account of economic, environmental, and social
performance, (2) facilitate comparison over time
and across organizations, and (3) credibly address
issues of concerns to stakeholders. The ?rst set of
GRI Guidelines was published in 1999 as an Expo-
sure Draft and several revisions have followed
since then. For the purpose of this study, we rely
on the GRI Sustainability Reporting Guidelines
published in 2002.
8
For a related discussion on the need for objective and
veri?able disclosures to achieve the separation predicted by
Verrecchia (1983) see Hutton, Miller, and Skinner (2003).
9
As anecdotal evidence in support of this argument, Green-
peace issued a press release on October 14, 1994 accusing
MacMillan Bloedel of deliberately lying to the public by
claiming that, in 25 years, the company had been convicted of
only 15 environmental o?enses. Greenpeace identi?ed 26
convictions in the last four years.
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 309
We engaged an expert in the ?eld of environ-
mental reporting to help us develop a content anal-
ysis index suitable for ?rms’ sustainability reports
or the corresponding sections of a broad social
responsibility report or equivalent discussions on
the web.
10
From the outset, we agreed that the
construct we seek to measure is the extent of a
?rm’s disclosure in their sustainability report.
With this aim established, the expert convinced
us that the GRI reporting guidelines are consistent
with that purpose. Thus, the expert helped us
develop a scoring model containing 95 line items
that re?ect the spirit of the GRI guidelines. Table
1 contains the scoring model with a reference to
the corresponding section in the GRI guidelines.
Our disclosure index follows closely the report-
ing requirements of the GRI guidelines for the fol-
lowing reasons. Firms do not have to prepare
social responsibility reports or related web based
disclosures discussing their environmental impacts,
and if they voluntarily do so, they do not have to
adopt the GRI guidelines. The voluntary decision
by a ?rm to both prepare a social responsibility
report and use the GRI guidelines means that the
?rm has opted for a format (the GRI format) that,
by the intent of the GRI guidelines, will result in
hard disclosures not easily mimicked by the poor
EP types. Thus, a ?rm making a sincere attempt
to use the GRI guidelines will score high using
our content analysis index, which is precisely the
result we seek (i.e., the poor EP types will not want
to conform to GRI guidelines that place a pre-
mium, for example, on objective environmental
performance indicators). Of the 95 equally
weighted items in our disclosure index, 79 relate
to ‘‘hard’’ disclosure measures compared to only
16 for ‘‘soft’’ disclosure items, a proportion which
(according to the expert who helped us developing
the index) re?ects the spirit of the GRI guidelines.
We now turn to discuss our disclosure index in
more detail. Our disclosure index consists of seven
broad categories, A1–A7, of environmental disclo-
sures (see Table 1). We consider A1–A4 and A5–
A7 to represent ‘‘hard’’ and ‘‘soft’’ environmental
disclosures, respectively.
To score environmental disclosures in discre-
tionary channels and web related disclosures, we
accessed the internet web site of each sample ?rm
and identi?ed its environmental report, if any,
and any web based environmental disclosures.
We saved all such disclosures as of September
2004. This arbitrary choice of timing worked out
well for us as all ?rms discussed their ?scal 2003
environmental performance in the environmental
reports and related web based disclosures which
we obtained.
Hard disclosure items
Category A1 focuses on disclosures pertaining
to a ?rm’s governance structure and management
systems put in place with respect to environmental
protection. For instance, ?rms whose Board of
Directors have an environmental committee or
have implemented ISO 14001 will inform their
stakeholders of such commitments. A2 focuses
on the credibility of a ?rm’s disclosures in its envi-
ronmental report. Firms that obtained indepen-
dent veri?cation of their environmental reports,
and ?rms with their products and environmental
programs certi?ed by independent agencies and
third parties will receive higher scores in this cate-
gory. In A3, we assess the extent to which ?rms
disclose speci?c environmental performance indi-
cators, both about their actual pollution emissions
and their conservation and recycling e?orts. These
are the ‘‘hard’’ data that ?rms can disclose to con-
vince stakeholders about their environmental com-
mitments. In addition, we also award scores when
?rms disclose performance indicators with respect
to historical trends, the ?rms’ own emission reduc-
tion targets, and the industry average. Disclosing
actual performance indicators in the above context
can convey critical information for stakeholders to
assess the ?rm’s long-term environmental perfor-
mance (and commitments).
The ?nal category in the ‘‘hard’’ disclosure
group is A4, which re?ects a ?rm’s environmental
10
Alan Willis, CA, Project Director – Performance Reporting
Initiatives, The Canadian Institute of Chartered Accountants.
He was a member of the GRI Steering Committee since its
inception, and has been a member of the GRI Guidelines
development and revision working groups from 1998 to date.
He is also a judge for the Canadian Institute for Chartered
Accountants’ Corporate Reporting Awards.
310 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
Table 1
Index assessing the quality of discretionary disclosures about environmental policies, performance and inputs
Hard disclosure items Map to
GRI
Percentage of
?rms attaining
the item (%)
Average score
Good EP
?rms
(N = 61)
Poor EP
?rms
(N = 61)
(A1) Governance structure and management systems (max score is 6) 71.31 2.08 1.27
***
1. Existence of a Department for pollution control and/or management
positions for env. management (0–1)
3.1 37.70 0.44 0.31
2. Existence of an environmental and/or a public issues committee in
the board (0–1)
3.1 31.15 0.38 0.25
3. Existence of terms and conditions applicable to suppliers and/or
customers regarding env. practices (0–1)
3.16 21.31 0.31 0.11
***
4. Stakeholder involvement in setting corporate environmental policies
(0–1)
1.1, 3.10 27.05 0.36 0.18
**
5. Implementation of ISO14001 at the plant and/or ?rm level (0–1) 3.14, 3.20 45.90 0.51 0.41
6. Executive compensation is linked to environmental performance
(0–1)
3.5 4.92 0.08 0.02
*
(A2) Credibility (max score is 10) 83.61 2.88 1.95
**
1. Adoption of GRI sustainability reporting guidelines or provision of
a CERES report (0–1)
3.14 12.30 0.21 0.03
**
2. Independent veri?cation/assurance about environmental
information disclosed in the EP report/web (0–1)
2.20, 2.21 1.64 0.03 0.00
3. Periodic independent veri?cations/audits on environmental
performance and/or systems (0–1)
3.19 17.21 0.18 0.16
4. Certi?cation of environmental programs by independent agencies
(0–1)
3.20 15.57 0.23 0.08
**
5. Product Certi?cation with respect to environmental impact (0–1) 3.16 9.84 0.08 0.11
6. External environmental performance awards and/or inclusion in a
sustainability index (0–1)
51.64 0.52 0.51
7. Stakeholder involvement in the environmental disclosure process
(0–1)
1.1, 3.10 4.92 0.08 0.02
*
8. Participation in voluntary environmental initiatives endorsed by
EPA or Department of Energy (0–1)
3.15 33.61 0.31 0.36
9. Participation in industry speci?c associations/initiatives to improve
environmental practices (0–1)
3.15 54.10 0.71 0.38
***
10. Participation in other environmental organizations/assoc. to
improve. environmental practices (if not awarded under 8 or 9
above) (0–1)
3.15 40.98 0.52 0.30
***
(A3) Environmental performance indicators (EPI) (max score is 60)
a
73.77 10.19 6.00
***
1. EPI on energy use and/or energy e?ciency (0–6) EN3, 4, 17 41.80 1.46 0.75
***
2. EPI on water use and/or water use e?ciency (0–6) EN5, 17 30.33 1.07 0.49
**
3. EPI on green house gas emissions (0–6) EN8 31.97 1.10 0.59
**
4. EPI on other air emissions (0–6) EN9,10 43.44 1.45 1.08
5. EPI on TRI (land, water, air) (0–6) EN11 33.61% 1.05 0.65
*
6. EPI on other discharges, releases and/or spills (not TRI) (0–6) EN12, 13 28.69 1.15 0.43
***
7. EPI on waste generation and/or management (recycling, re-use,
reducing, treatment and disposal) (0–6)
EN11 50.00 1.44 1.04
8. EPI on land and resources use, biodiversity and conservation (0–6) EN6, 7 36.89 0.71 0.47
9. EPI on environmental impacts of products and services (0–6) EN14 4.10 0.13 0.00
*
10. EPI on compliance performance (e.g., exceedances, reportable
incidents) (0–6)
EN16 25.41 0.64 0.48
(continued on next page)
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 311
Table 1 (continued)
Hard disclosure items Map to
GRI
Percentage of
?rms attaining
the item (%)
Average score
Good EP
?rms
(N = 61)
Poor EP
?rms
(N = 61)
(A4) Environmental spending (max score is 3) 44.26 0.84 0.45
**
1. Summary of dollar savings arising from environment initiatives to
the company (0-1)
23.77 0.30 0.18
*
2. Amount spent on technologies, R& D and/or innovations to enhance
environ. perf. and/or e?ciency (0–1)
EN35 20.49 0.21 0.19
3. Amount spent on ?nes related to environmental issues (0–1) EN16 25.41 0.33 0.18
**
Soft disclosure items Map to
GRI
Percentage of
?rms attaining
the item (%)
Average score
Good EP
?rms
(N = 61)
Poor EP
?rms
(N = 61)
(A5) Vision and strategy claims (max score is 6) 95.90 3.48 3.04
1. CEO statement on environmental performance in letter to
shareholders and/or stakeholders (0–1)
1.1, 1.2 61.48 0.69 0.54
*
2. A statement of corporate environmental policy, values and
principles, environ. codes of conduct (0–1)
1.1, 1.2,
3.7
87.70 0.85 0.90
3. A statement about formal management systems regarding
environmental risk and performance (0–1)
3.19 58.20 0.57 0.59
4. A statement that the ?rm undertakes periodic reviews and evaluations
of its environ. performance (0–1)
3.19 37.70 0.47 0.27
**
5. A statement of measurable goals in terms of future env. performance
(if not awarded under A3) (0–1)
1.1, 1.2 27.05 0.31 0.22
6. A statement about speci?c environmental innovations and/or new
technologies (0–1)
1.1, 1.2 54.92 0.58 0.52
(A6) Environmental pro?le (max score is 4) 70.49 1.49 1.23
1. A statement about the ?rm’s compliance (or lack thereof) with
speci?c environmental standards (0–1)
GN 8 32.79 0.38 0.28
2. An overview of environmental impact of the industry (0–1) GN 8 22.13 0.26 0.20
3. An overview of how the business operations and/or products and
services impact the environment. (0–1)
GN 8 56.56 0.61 0.52
4. An overview of corporate environmental performance relative to
industry peers (0–1)
GN 8 24.59 0.26 0.23
(A7) Environmental initiatives (max score is 6) 72.95 1.93 1.34
**
1. A substantive description of employee training in environmental
management and operations (0–1)
3.19 30.33 0.39 0.21
**
2. Existence of response plans in case of environmental accidents (0–1) 22.95 0.30 0.16
*
3. Internal environmental awards (0–1) 13.11 0.18 0.08
4. Internal environmental audits (0–1) 3.19 3.20 34.43 0.38 0.31
5. Internal certi?cation of environmental programs (0–1) 3.19 9.84 0.15 0.05
*
6. Community involvement and/or donations related to environ. (if not
awarded under A1.4 or A2.7) (0–1)
SO1,
EC10
53.28 0.54 0.52
This table presents the index used toassess the discretionary disclosures about environmental policies, performance andinputs. Index items
are classi?ed in two categories: ‘‘hard’’ and ‘‘soft’’ disclosures. The second column presents the mapping of items in the index to the Global
Initiative Reporting (GRI) guidelines. The thirdcolumnpresents the percentage of ?rms whichmade disclosures onthat item(discretionary
channels considered are Environmental and/or Social Responsibility Reports, or similar disclosures in ?rms’ web site). Good environ-
mental performance (EP) ?rms are ?rms that have the environmental performance measure (% recycled) above the industry median. The
last twocolumns present the average score oneachitemfor each groupof ?rms. The signi?cance levels presented inthe last columnare from
two-sample t-statistics that test the di?erence between the good and the poor group.
***
,
**
,
*
represent signi?cance levels (two-tailed) at 1%,
5% and 10%, respectively. Wilcoxon Rank tests and t-tests with Bootstrap Resampling provided similar results. Sample size is 122 ?rms.
312 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
spending. We do not score disclosures pertaining
to environmental spending as a result of comply-
ing with the existing environmental regulations,
as such disclosures are largely non-discretionary
and appear in mandatory disclosure channels such
as 10 Ks and annual reports. Rather, we focus on
disclosures of dollar savings from existing environ-
mental programs and e?orts and discretionary
spending to further enhance future environmental
performance such as investing in new environmen-
tal technologies or environmentally related R&D
and innovations. We also include disclosures of
?nes related to environmental issues. Such penal-
ties are usually immaterial thus their reporting is
not mandatory. Nevertheless, the dollar amount
of ?nes is important to environmental stakehold-
ers to assess the level of true commitment to the
environment. In summary, our index design in
the A1–A4 ‘‘hard’’ disclosure categories makes it
relatively di?cult for poor environmental per-
formers to mimic the environmental disclosures
of good environmental performers.
Soft disclosure items
We measure a ?rm’s disclosures of vision and
environmental strategy claims in A5. For instance,
?rms often disclose broadly that they have an envi-
ronmental policy, that management is committed
to protecting the environment, etc. Such disclo-
sures can be genuine when put in the speci?c con-
text but they can also be deceiving as they lack
credibility and substantiation, and can be easily
mimicked.
11
A6 assesses the disclosure of a ?rm’s
environmental pro?le given the existing and forth-
coming environmental regulations. Finally, we
code a ?rm’s disclosures of its environmental ini-
tiatives in A7. Items coded here include employee
training in environmental management, existence
of response plans for environmental accidents,
internal environmental awards and audit, and
community involvement through scholarship and
donations. Again, these kinds of initiatives can
represent true commitment but they can also be
imitated by companies with no real commitments
to protecting the environment.
Econometric model
Model and variable descriptions
In order to test our hypotheses, we employ the
following econometric model:
VED ¼ b
0
þb
1
EP þb
2
J–F coefficient þb
3
FIN
þb
4
TOBIN Q þb
5
VOLAT þb
6
ROA
þb
7
LEV þb
8
SIZE þb
9
NEW
þb
10
CAPINþe
The variables in the regression above are de?ned
as follows:
VED – is a score of voluntary environmental
disclosures using web based disclosures as of Sep-
tember, 2004. We perform a content analysis using
our disclosure index (see Table 1).
EP – is an environmental performance proxy.
We use two alternative proxies to capture the envi-
ronmental performance of each ?rm. The ?rst is the
TRI emission scaled by total sales revenue. To
facilitate the interpretation of the results, we
reverse the sign of this variable. In other words,
the larger this measure is, the better the environ-
mental performance of the ?rm. The second mea-
sure is the percentage of toxic waste treated,
Table 1 (continued)
a
The scoring scale of environmental performance data is from 0 to 6. A point is awarded for each of the following items: (1)
Performance data is presented; (2) Performance data is presented relative to peers/rivals or industry; (3) Performance data is presented
relative to previous periods (trend analysis); (4) Performance data is presented relative to targets; (5) Performance data is presented
both in absolute and normalized form; (6) Performance data is presented at disaggregate level (i.e., plant, business unit, geographic
segment).
11
Consider, for example, items A5-2, 3 and 5. We classify
these statements as soft because they involve claims about
environmental management control systems without details or
substantiation. In contrast, the corresponding items in A1-1
and 2 are hard because they provide speci?c information about
the existence of the department, management positions or
board committees responsible for monitoring pollution control.
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 313
recycled or processed in the production. Large
recycling percentages imply environmentally pro-
active ?rms. Both measures are computed using
the public database made available by the US Envi-
ronmental and Protection Agency (EPA). We
aggregate the plant speci?c data at the ?rm level.
J–F coe?cient – the Janis–Fadner coe?cient of
imbalance, measured for the ?rm’s 2002 ?scal year
and de?ned in greater detail below.
FIN – is the amount of debt or equity capital
raised by the ?rm in the ?scal year 2004. It is the
sale of common stock and preferred shares minus
the purchase of common stock and preferred
shares (#108–#115) plus long term debt issuance
minus the long term debt reduction (#111–
#114).
12
The amount is scaled by the size of total
assets at the end of the ?scal year 2002.
TOBIN Q – is Tobin’s Q, measured as market
value of common equity (#25
*
#199) plus book
value of preferred stock (#10), book value of long
term debt (#9) and current liabilities (#5), divided
by book value of total assets (#6).
VOLAT – is stock price volatility, measured as
standard deviation of market adjusted monthly
stock return during ?scal year 2003.
ROA – is total return on assets measured as the
ratio of income before extraordinary items (#18)
at the end of ?scal year 2004 and total assets
(#6) at the end of ?scal year 2003.
LEV – is the leverage ratio, measured as the
ratio of total debt (#9 + #34) divided by total
assets (#6) at the end of ?scal year 2003.
SIZE – is the natural logarithm of the total
asset value measured as of the end of ?scal year
2003.
NEW – is the asset newness, measured as a ratio
of net properties, plant and equipment (#8)
divided by the gross properties, plant and equip-
ment (#7) at the end of ?scal year 2003.
CAPIN – is the capital intensity, measured as a
ratio of capital spending (#128) divided by total
sales revenues (#12) at the end of ?scal year 2003.
The control variables included in the multivari-
ate regression model besides our environmental
performance (EP) measures have been docu-
mented to be causes of voluntary disclosures in
the disclosure literature. We include them to avoid
a correlated omitted variables threat that EP is
standing in for some other well known determi-
nants of disclosure in other disclosure channels.
We divide the control variables in three groups:
variables that proxy for the bene?ts of voluntary
disclosures, variables that measure costs of volun-
tary disclosures and other control variables.
Bene?ts of voluntary disclosure
Financing: It is well known that ?rms that raise
?nancing in debt and equity markets have a higher
propensity for disclosures in voluntary channels
(Frankel, McNichols, & Wilson, 1995) to lower
their cost of capital. We use the amount of debt
and equity ?nancing raised by the ?rm in the ?scal
year following the measurement of the environ-
mental performance (FIN).
Information asymmetry: It is generally asserted
in the voluntary disclosure literature that manag-
ers seek to lower information asymmetry through
voluntary disclosures in order to lower the cost
of capital (Healy & Palepu, 2001). Our chosen
proxies for information asymmetry are: monthly
stock return volatility measured (VOLAT) over
the 12 month period represented by ?scal 2003
(Lim, 2001) and Tobin’s Q, based on the argument
that ?rms with greater unbooked intangibles and a
positive NPV investment opportunity set enjoy
larger Tobin’s Q (Barth & Kasznik, 1999; Smith
& Watts, 1992).
Firm performance: Lang and Lundholm (1993)
and others have shown that ?rms with superior
upcoming earnings performance have a higher dis-
closure propensity to reveal their ‘‘good news’’ to
?nancial markets. At the time of observing web
disclosures (September 2004), markets would
know ?scal year 2003 ROA so earnings for the
upcoming year would be ?scal 2004 ROA.
Leverage: A number of disclosure studies (e.g.,
Leftwich, Watts, & Zimmerman, 1981) have
argued that the monitoring demand for informa-
tion increases as ?rm debt increases, and empirical
evidence is consistent with managers being more
forthcoming, generally to facilitate the contracting
12
Numbers in brackets represent data items in the Compustat
Annual File.
314 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
demand for information. Agency costs of debt are
higher for ?rms with relatively more debt in their
capital structure (Jensen & Meckling, 1976), thus
voluntary disclosures are expected to increase with
leverage.
Costs of voluntary disclosures
Firm Size: Most voluntary disclosure studies
control for ?rm size (see, for example, Lang &
Lundholm, 1993) based on the assumption of
economies of scale with respect to information
production costs.
13
Proprietary costs: In Verrecchia (1983), a key
friction sustaining a partial disclosure equilibrium
is the existence of proprietary costs associated with
being forthcoming. In our setting, proprietary
costs pertain to the manager revealing information
to environmental regulators and other environ-
mental activist groups that increase the probability
of criticism, sanction or attack (see Li et al., 1997).
We assume that industry serves as a measure of
proprietary costs, since pollution propensity and
related monitoring by opponents is well known
to vary by industry. In our inter-industry regres-
sions, we control for industry ?xed e?ects in order
to control for di?ering proprietary costs and other
unidenti?ed factors that might vary by industry.
Other control variables speci?c to environmental
disclosures
Equipment age and annual capital spending:
Healy and Palepu (2001) describe a common criti-
cism in the voluntary disclosure literature involv-
ing endogeneity. EP and sustainability disclosures
might be joint endogenous variables driven by
some underlying exogenous variables such as the
level of investments in clean technologies, thus rep-
resenting a threat to causal inferences. Firms with
newer, cleaner technologies are likely to have a
superior environmental performance measure and
it is reasonable to assume that they will want
stakeholders to know about this superior environ-
mental performance in discretionary disclosure
channels. To address this threat, we control for
the average age of a ?rm’s equipment (NEW)
based on the argument that newer equipment is
expected to employ newer and less polluting tech-
nologies. For similar reasons, ?rms with higher
sustaining capital expenditures, as proxied by
CAPIN, are expected to have newer equipment
and may want to signal their environmental type
through more discretionary disclosures regarding
their environmental performance.
Favorable media coverage: Following Aerts and
Cormier (2006), Bansal and Clelland (2004) and
Janis et al. (1965), we measure lagged environmen-
tal legitimacy as the propensity for unfavorable
press articles using the Janis–Fadner coe?cient
of imbalance. This coe?cient ranges from À1
(unfavorable) to +1 (favorable), with zero imply-
ing neutral perceptions about the ?rm’s environ-
mental legitimacy. Following Aerts and Cormier
(2006) and Patten (2002), legitimacy theory pre-
dicts a negative association between lagged envi-
ronmental legitimacy and the level of voluntary
environmental disclosures.
14
The Janis–Fadner
coe?cient is measured as follows:
Janis–Fadner coefficient
¼
ðe
2
ÀecÞ
t
2
if e > c;
¼
ðec Àe
2
Þ
t
2
if c > e;
¼ 0 if e ¼ c;
13
Firm size has been regularly established as a determinant of
voluntary disclosure in the literature, so its e?ects must be
controlled for. Firm size is marginally signi?cant and positively
correlated with % recycled in the Chemical and Pulp and Paper
industries, but is insigni?cant for our other three industries. In
Table 4, we control for ?rm size as a covariate to ensure that
our EP measures are not standing in for size. In Table 5, ?rm
size is once again a covariate in our model. Thus, the role
played by ?rm size in environmental disclosure is controlled for
in both tables when we isolate e?ects due to EP.
14
In our sample of 191 ?rms, 126 have at least one article in
Factiva database related to the environment in ?scal 2002. The
126 ?rms generate 770 total articles related to the environment
during ?scal 2002. Of these, 393, 207, and 170 are coded by us
as unfavorable, neutral and favorable, respectively. For the 65
?rms with no environmental articles, the Janis–Fadner coe?-
cient is set to a default measure of zero. Thus, silence in the
media is interpreted to imply neutrality of perceptions about
environmental legitimacy.
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 315
where e is the number of favorable environmental
articles, c is the number of unfavorable articles,
and t is e + c.
Sample selection, summary statistics and univariate
empirical results
As mentioned in section ‘Introduction’, the
sample of this study consists of public US compa-
nies from ?ve polluting industries that report toxic
release data to the US Environmental Protection
Agency and have ?nancial and stock price data
in the Compustat and CRSP database.
15
The ?nal
sample contains 191 ?rms covered by the EPA-
TRI database for 2003 with adequate variables
available in Compustat and CRSP. The distribu-
tion across the ?ve industries is as follows: 27 ?rms
in the Pulp and Paper industry (14.14% of the sam-
ple), 63 ?rms in the Chemical industry (32.98% of
the sample), 18 ?rms in the Oil and Gas industry
(9.42% of the sample), 42 ?rms in the Metals and
Mining industry (21.99% of the sample) and 41
?rms in the Utilities industry (21.47% of the sam-
ple). Among these ?rms, 122 ?rms (63.87% of
the sample) chose to provide discretionary disclo-
sures about the environment in ?scal 2003: 55 of
those ?rms had stand alone environmental reports
and invariably also had supplementary web disclo-
sures, while the remaining 67 ?rms had discretion-
ary web disclosures but no stand alone
environmental reports. If a ?rm has no environ-
mental report or related discretionary web disclo-
sures, we classify such a ?rm as being ‘‘silent’’,
with the disclosure score set to zero for these 69
?rms.
16
We allow silent ?rms in the sample since
non-disclosure is a choice in a partial disclosure
equilibrium setting.
For ?rms providing discretionary environmen-
tal disclosures, Table 1 presents our scoring model
along with descriptive statistics as to the percent-
age of ?rms disclosing a particular item and the
di?erence in average disclosure scores across good
and poor EP ?rms. The table also shows the GRI
reference for each line item. A crucial assumption
of our scoring approach is that the disclosures are
discretionary. In support of that assumption, we
use only disclosures that came from ?rm environ-
mental reports (hereafter EP reports) or web based
disclosures other than ?rm annual reports or
10 Ks. We assume that EP reports/web disclosures
are purely discretionary, i.e., silence is always an
option.
In Table 1, A1–A4 summarize the separation in
average awarded scores across good and poor EP
performers, using the %Recycled as the measure
of environmental performance. We use the median
% recycled in a given industry to classify ?rms as
good versus poor EP performers.
17
Recall that,
according to H1a, good EP performers should
have higher scores for hard disclosure items that
are di?cult to mimic by poor EP performers.
The results in Table 1 con?rm that prediction for
the A1 category (Governance Structure and Man-
agement Systems). The average score for good
(poor) EP performers is 2.08 (1.27). Using a two-
tailed t-test the di?erence is signi?cant at the 1%
level. For the A2 category (Credibility), the aver-
age score for good (poor) EP performers is 2.88
(1.95), and the di?erence is again signi?cant at
the 5% level. Not surprisingly, good EP ?rms are
more likely to disclose that they adopt GRI guide-
lines. The di?erence in average scores 0.21 versus
0.03 is statistically signi?cant at the 1% level. It
is apparent from the scores for A2-2 that 2
(61 · 0.03) good EP ?rms obtained independent
assurance for their ER report/web disclosures
compared to zero poor EP ?rms. While this is con-
sistent with an attempt by good EP ?rms to signal
their type, the di?erence between the two types of
?rms is not statistically signi?cant.
15
We identify all ?rms with available TRI data and two-digit
SIC codes: 26 (pulp and papers), 28 (chemicals), 29 (oil and
gas), 33 (metals and mining), and 49 (utilities). For these
companies, we read the ?rm’s business descriptions and
dropped companies that identify material business operations
in industries outside of their primary two-digit SIC code.
16
We contacted the 69 ?rms with zero disclosures ?rst by
email and then by phone. None of them indicated that they
published a stand alone environmental report in 2003.
17
Our results in Table 1 are qualitatively una?ected when we
partitioned the sample using the mean % recycled.
316 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
Of special interest is category A3 (EPI indica-
tors), since this type of hard disclosure is not one
that poor EP ?rms will want to mimic. The aver-
age score for Good EP performers of 10.19 is well
in excess of the corresponding average score for
poor EP performers, 6.00 (the di?erence is signi?-
cant at the 1% level). Each item has a total score of
6 based on the dimensions indicated in Table 1.
The low scores for both good and poor EP per-
formers suggest that disclosures in this area are
less than ideal given the GRI guidelines. In unta-
bulated analyses, we analyzed the contribution of
each of the 6 dimensions to the total scores (i.e.,
‘‘hits’’) awarded in A3, for good versus poor EP
?rms: performance data presented represents
43.50% (48.44%) of total A3 scores awarded for
good (poor) EP ?rms, respectively; performance
data presented relative to peers represents only
1.06% (2.44%) of total A3 scores awarded for good
(poor) EP ?rms, respectively; trend analysis repre-
sents 28.86% (26.33%) of total A3 scores awarded
for good (poor) EP ?rms, respectively; perfor-
mance relative to targets represents 8.76%
(7.31%) of total A3 scores awarded, for good
(poor) EP ?rms, respectively; performance data
presented in both absolute and normalized form
represents 7.87% (8.63%) of total A3 scores
awarded, for good (poor) EP ?rms, respectively;
and, ?nally, performance data presented at the
disaggregated level represents 9.64% (6.68%) of
total A3 scores awarded, for good (poor) EP ?rms,
respectively. Thus, it is apparent that GRI guide-
lines with respect to performance relative to peers
are not being followed by either good or poor EP
performers. This is not surprising, as it is di?cult
for both types of ?rms to decide on appropriate
‘‘peers’’ given di?erences in production processes
across ?rms within a given industry. It is precisely
this dilemma that makes relative environmental
performance ‘‘unobservable’’ to the typical inves-
tor or stakeholder, creating the potential for the
setting of a partial disclosure equilibrium where
some ?rms are more transparent than others in
their report/web disclosures, and some ?rms are
entirely silent.
The ?nal hard category is A4 (disclosures per-
taining to discretionary environmental spending).
For that category, the average score for good
(poor) EP performers is 0.84 (0.45). The di?erence
between the good and poor groups is signi?cant at
5%. In this category we ?nd that, on average, good
EP ?rms disclose signi?cantly more often amounts
spent on ?nes than the poor ?rms. Economics
based voluntary disclosure theories predict that
good EP ?rms will be more forthcoming about
the dollar amount spent on ?nes because this
amount will be lower than the corresponding
amount for poor EP ?rms, consistent with supe-
rior environmental performance within the indus-
try. To validate that disclosing the amount spent
on ?nes signals a ?rm’s commitment to the envi-
ronment, we examine the dollar amount spent on
?nes and the number of environmental violations
in ?scal year 2000 for the 100 of our sample ?rms
(56 good EP ?rms and 44 poor EP ?rms) covered
in the Corporate Environmental Pro?le Database
developed by the Investor Responsibility and
Research Center. In untabulated results, we ?nd
that the ?fty-six good EP ?rms had an average
of 1.13 violations per ?rm and 0.17 cents of ?nes
per thousand of dollars of sales. In comparison,
the forty four poor EP ?rms had an average of
2.63 violations per ?rm and 0.42 cents of ?nes
per thousand dollars of sales. Using a two-tailed
t-test, the di?erence in the number of violations
(scaled dollar ?nes) between the two groups is sta-
tistically signi?cant at 0.02 (0.01) level.
18
Table 1 contains corresponding statistics for the
soft disclosure categories A5–A7. Consistent with
H1a, good EP performers have signi?cantly
greater soft discretionary disclosures for A7 (envi-
ronmental initiatives) relative to poor EP perform-
ers. However, there is no signi?cant di?erence for
A5 (vision and strategy) and A6 (environmental
pro?le). Overall, the results for our soft disclosure
scores suggest less separation between good and
poor EP ?rms, a ?nding which we explore in more
depth in section ‘A revised role for socio-polittical
theories’ below.
18
An alternative interpretation of our results for ?nes disclo-
sures is that the ?rms disclosing such ?nes are seeking to
legitimize the violation of environmental regulation and
requirements. This pattern of disclosure behaviour is predicted
by legitimacy theory (Patten, 2000). We leave the validation of
this alternative explanation to future research.
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 317
In Table 2, Panel A we present descriptive sta-
tistics on environmental disclosure scores by
industry. The lowest average score (out of 95) is
10.08 and is obtained by the Metals and Mining
Industry while the highest score, 25.21, is obtained
by the Pulp and Paper Industry.
19
In Table 2,
Panel B we show that the average ratios of TRI/
Sales for Pulp and Paper and Metals and Mining
are 2.34 (i.e., 2.34 pounds of toxic emissions per
thousand dollar sales) and 1.72, respectively. Thus,
it appears that industries with a high (low) pollu-
tion propensity are more (less) likely to provide
discretionary environmental disclosures to ?rm
stakeholders. This is by no means inconsistent
with H1a which conditions disclosure predictions
by environmental type at the ?rm level relative to
industry peers. Table 2 does con?rm di?erences
across sectors in both environmental disclosure
and pollution propensities suggesting the need
for industry controls.
Untabulated analysis for our 69 ?rms with zero
disclosure scores indicates that these ?rms have an
average TRI/Sales (% recycled) measure of 2.97
(59.01%) compared to 1.82 (65.61%) for the 122
?rms with non-zero disclosure scores. The di?er-
ences in both EP measures across the two groups
are signi?cant at the 5% level. This is generally
consistent with H1a, i.e., poor EP ?rms are more
likely to opt for silence.
In Table 3, Panel A we present descriptive sta-
tistics for independent variables used in the estima-
tion. The average ?rm has a negative ?nancing
variable (FIN) meaning that it reduces debt or
repurchases shares more than it raises new ?nanc-
ing. Also, on average, the ROA is about 5% and
the average leverage (LEV) is 33% of total assets.
The ?rm size measured by the logarithm of total
assets (SIZE) is 8.01 implying average total assets
in dollar terms of $3.01 bn, thus our sample con-
sists of relatively larger ?rms. The mean (and med-
ian) ?rm size is comparable with the median ?rm
size reported by Patten (2002) for his sample.
The average J–F coe?cient is À0.08.
We present in Panel B of Table 3 Pearson cor-
relations between the independent variables used
in the regressions. The Pearson correlation
between the negative of TRI/Sales and % recycled
is equal to 0.29 and is statistically signi?cant at 1%,
implying that (minus) TRI/Sales and % recycled
both measure environmental performance but the
overlap is modest.
Empirical results involving the level of disclosure
Inter-industry analysis
We present in Table 4 the results of inter-indus-
try multivariate regressions of disclosure scores on
the environmental performance measures and the
control variables. We estimate the regressions
using a Tobit analysis to account for the censoring
of the dependent variable at zero.
20
We estimate
the Tobit regressions by maximum likelihood
using a Newton–Raphson algorithm.
21
We run
three sets of equations based on the di?erent dis-
closure scores (total, hard and soft) used on the
dependent side. All regressions are inter-industry
analyses using dummy variables to control for
industry ?xed e?ects. We estimate the regressions
using each environmental performance variable
separately and then include both variables at the
same time.
The ?rst three columns present the results for
the total disclosure scores. As predicted by H1a,
the estimated coe?cients for our environmental
19
While our scoring scheme has 95 available points to allow
for rich variety of disclosures from one sustainability report to
the next, and from one industry to another, the attainable score
is lower than this, even for an excellent sustainability report.
For example, our top scoring ?rm, Weyerhaeuser, obtained 68
of the 95 available points, suggesting an e?ective maximum of
71%. This is comparable to the e?ective maximum in similar
scoring scheme used in ‘‘Risk and Opportunity: the Best
Practice in Non-?nancial Reporting’’ by Standard and Poor’s
(2004). Nonetheless, the low average scores (out of 95) for each
of the four industries point to an overall need for improvement
in sustainability reporting in the years beyond 2003.
20
As an alternative, we have re-run our regressions using
simple OLS and the inferences are unchanged.
21
In the Tobit model, the marginal e?ect of a change in an
independent variable on the dependent variable (i.e., disclosure
score) is the estimated coe?cient times the probability that the
?rm provides discretionary environmental disclosures (Verbeek,
2004).
318 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
Table 2
Descriptive Statistics for variables of interest
Panel A: Environmental disclosures (N = 122 ?rms)
Overall Pulp and paper Chemicals Metals and mining Oil and gas Utilities
(n = 122) (n = 24) (n = 41) (n = 13) (n = 13) (n = 31)
Hard disclosures (0–79) 12.88 (10.00) 16.95 (13.00) 12.19 (10.00) 5.76 (2.00) 14.31 (11.00) 13.00 (11.00)
0–53 0–53 0–39 0–42 1–33 0–31
(A1) Governance structure and
manage. systems (0–6)
1.68 (1.00) 1.95 (2.00) 1.95 (2.00) 1.31 (1.00) 2.00 (2.00) 1.10 (1.00)
0–5 0–5 0–5 0–5 0–5 0–5
(A2) Credibility (0–10) 2.42 (2.00) 3.71 (3.00) 2.14 (2.00) 1.23 (1.00) 2.77 (3.00) 2.13 (2.00)
0–9 0–9 0–6 0–6 0–5 0–6
(A3) Environmental performance 8.09 (6.50) 10.13 (6.00) 7.49 (5.00) 3.00 (0.00) 8.69 (4.00) 9.23 (9.00)
Indicators (0–60) 0–36 0–36 0–27 0–29 0–21 0–23
(A4) Environmental spending (0–3) 0.70 (0.00) 1.17 (1.00) 0.58 (0.00) 0.23 (0.00) 0.85 (1.00) 0.61 (0.00)
0–3 0–3 0–3 0–2 0–3 0–3
Soft Disclosures (0–16) 6.26 (6.00) 7.91 (8.50) 6.39 (6.00) 4.31 (3.00) 6.69 (7.00) 5.45 (5.00)
0–15 2–15 0–13 0–13 0–12 1–11
(A5) Vision and strategy (0–6) 3.26 (3.00) 3.71 (4.00) 3.27 (3.00) 2.53 (3.00) 3.61 (4.00) 3.06 (3.00)
0–6 1–6 0–6 0–6 1–6 1–6
(A6) Environmental pro?le (0–4) 1.36 (1.00) 2.08 (2.00) 1.19 (1.00) 1.08 (1.00) 1.38 (1.00) 1.12 (1.00)
0–4 0–4 0–3 0–3 0–3 0–3
(A7) Environmental initiatives (0–6) 1.64 (1.00) 2.13 (2.00) 1.90 (1.00) 0.69 (0.00) 1.77 (2.00) 1.26 (1.00)
0–6 0–6 0–4 0–4 0–4 0–5
Total (0–95) 19.13 (15.00) 25.21 (19.00) 18.59 (15.00) 10.08 (4.00) 21.00 (22.00) 18.45 (15.00)
1–68 3–68 1–50 1–55 2–43 1–42
Panel B: Environmental performance (N = 191 ?rms)
Overall Pulp and paper Chemicals Metals and mining Oil and gas Utilities
% Recycled 63.12% (81.67%) 68.41% (83.22%) 76.72% (93.92%) 70.47% (84.52%) 77.70% (87.46%) 25.32% (24.69%)
0–99.85% 1.80–99.01% 0.01–99.50% 0–99.85% 0–99.28% 0–68.54%
TRI/Sales 2.24 (0.62) 2.34 (1.49) 2.16 (0.58) 1.72 (0.27) 0.89 (0.13) 3.45 (2.07)
0.01–23.19 0.09–12.02 0.01–18.99 0.01–21.19 0.01–8.02 0.01–14.92
This table presents descriptive statistics on environmental disclosure scores and environmental performance measures by industry. Descriptive statistics present means
(medians) and ranges (min–max) below. Panel A presents disclosures scores for ?rms that chose to provide discretionary disclosures (i.e., ‘‘disclosing’’ ?rms) on their
environmental performance (N = 122 ?rms). The scale for each category of disclosure items is presented in brackets. Panel B presents environmental performance
measures for the full sample, i.e., ‘‘disclosing’’ ?rms and ‘‘silent’’ ?rms (N = 191 ?rms). TRI/Sales is toxics release inventory data (in pounds) divided by total sales (in
thousands). Percent recycled is toxic waste treated or recycled divided by total waste generated by ?rm.
P
.
M
.
C
l
a
r
k
s
o
n
e
t
a
l
.
/
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
)
3
0
3
–
3
2
7
3
1
9
performance proxies are all positive and signi?-
cant (recall, we reverse the sign of TRI/Sales to
facilitate the interpretation). This result is consis-
tent with the discretionary disclosure theories of
Verrecchia (1983) and Dye (1985), but inconsis-
tent with the negative association (H1b) pre-
dicted by socio-political theories. Firms with
better environmental performance have more vol-
untary disclosures about their environmental
impact. Furthermore, when we introduce both
scaled TRI and % recycled in the regression we
?nd that they are signi?cant and positive suggest-
ing that each provides incremental information
to the other with respect to voluntary environmen-
tal disclosures.
Table 4 indicates the predicted sign of associa-
tion for our control variables given prior general
disclosure literature and intuition. For the total
disclosures as well as hard versus soft disclosures,
the forward looking ?nancing proxy is positive
(consistent with the prior literature) but is statisti-
cally signi?cant only for the soft disclosure cate-
Table 3
Descriptive and correlation statistics for variables used in the estimation
Panel A: Descriptive statistics
Variable Mean Median Q1 Q3 Std dev
% Recycled 0.63 0.82 0.30 0.95 0.35
ÀTRI/Sales À2.24 À0.62 À2.50 À0.13 3.68
J–F coe?cient À0.08 0.00 0.00 0.37 0.81
SIZE 8.01 7.89 6.70 9.28 1.70
FIN À0.02 À0.01 À0.03 0.01 0.31
TOBIN Q 1.13 1.04 0.81 1.35 0.75
VOLAT 0.09 0.07 0.05 0.10 0.08
ROA 0.05 0.03 0.00 0.08 0.07
LEV 0.33 0.31 0.23 0.41 0.16
NEW 0.54 0.54 0.45 0.62 0.14
CAPIN 0.07 0.04 0.02 0.09 0.08
Panel B: Pearson correlation statistics
TRI/Sales Size FIN TOBIN Q VOLAT ROA LEV AGE CAPIN Total
disclosure
J–F
coe?cient
% Recycled 0.29
***
À0.06 0.15
**
0.28
***
0.07 0.14
**
0.02 À0.26
***
À0.30
***
0.23
***
À0.05
ÀTRI/Sales – À0.04 À0.01 0.11 À0.16
**
0.08 0.01 À0.03 À0.01 0.17
***
0.13
*
SIZE – 0.07 0.17
**
À0.09 0.01 0.01 0.33
***
0.21
***
0.46
***
0.05
FIN – À0.03 0.03 0.05 0.02 0.05 0.03 0.03 0.01
TOBIN Q – À0.01 0.47
***
À0.04 À0.15
**
À0.06 0.14 0.03
VOLAT – À0.11 0.03 À0.15
**
À0.06 À0.16
**
À0.15
**
ROA – À0.11 À0.10 À0.07 0.04 0.03
LEV – 0.14
*
0.01 0.19
***
0.10
NEW – 0.18
**
À0.12
**
À0.02
CAPIN 0.34
***
0.09
Total disclosure – 0.09
*
This table presents descriptive and correlation statistics for independent variables used in multivariate tests. Statistics are presented for
the full sample of 191 ?rms. Percent recycled is toxic waste treated or recycled divided by total waste generated by ?rm. TRI/Sales is the
negative of Toxics Release Inventory data (in pounds) divided by total sales in thousands. SIZE is the logarithm of market value. FIN
is the amount of debt or equity capital raised in ?scal year 2004 divided by total assets. TOBIN Q is the sum of market value of equity,
book value of preferred stock and book value of debt divided by total assets. VOLAT is stock price volatility (standard deviation of
monthly returns during 2003). ROA is return on assets. LEV is the leverage ratio. NEW is asset newness measured as the ratio of net
PPE to gross PPE. CAPIN is capital intensity measured as the ratio of capital spending to total sales. Total Disclosure is total
environmental disclosure score achieved using the disclosure index presented in Table 1. J–F coe?cient is the Janis–Fadner Coe?cient
(see Bansal and Clelland, 2004 for details). Spearman Correlation statistics provide similar results.
***
,
**
,
*
represent signi?cance levels
(two-tailed) at 1%, 5% and 10%, respectively.
320 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
gory. The information asymmetry proxies (Tobin’s
Q and stock volatility) are insigni?cant. Similarly,
?rm pro?tability (ROA) in the following year is
insigni?cant. One interpretation of this result is
that ?rms resort to other disclosure channels to
reduce the information asymmetry and convey
their good news about future ROA, and do not
use ER reports/web disclosures for this purpose.
Turning to our agency proxy, we ?nd that the
leverage variable is signi?cantly positive, suggest-
ing that debtholders exercise pressure on ?rms to
disclose environmental related matters to assess
potential future liabilities. Also, large ?rms dis-
close more, consistent with their lower information
production costs. Similarly, ?rms with greater cap-
ital expenditures disclose more: the coe?cient of
the capital intensity control variable is signi?cantly
positive. In contrast to our intuition, ?rms with
newer equipment (NEW) are less likely to provide
discretionary environmental disclosures. The
lagged Janis–Fadner coe?cient is not associated
with the level of disclosure, which is consistent
with the results of Aerts and Cormier (2006).
We also split our disclosure scores in two parts:
hard disclosures and soft disclosures. Hard disclo-
sures are closest in spirit to the assumed truthful
disclosures in the voluntary disclosure theories dis-
cussed in section ‘Literature review and hypothesis
Table 4
Inter-industry regressions with industry ?xed e?ects (Tobit analysis)
Dependent variables
Total disclosures Hard disclosures Soft disclosures
(1) (2) (3) (1) (2) (3) (1) (2) (3)
Intercept À81.2
***
À74.72
***
À78.67
***
À70.65
***
À64.00
***
À68.01
***
À20.65
***
À19.03
***
À19.85
***
(122.45) (103.76) (115.87) (115.51) (105.64) (112.24) (75.46) (63.87) (69.60)
% Recycled (±) 11.48
***
9.81
***
8.83
***
8.02
***
2.40
**
2.11
**
(10.31) (7.41) (7.62) (6.53) (5.68) (4.35)
ÀTRI/Sales (±) 0.84
***
0.65
**
0.77
***
0.72
***
0.21
**
0.18
*
(6.67) (4.07) (7.90) (6.99) (4.43) (3.20)
J–F coe?cient (À) 2.21 1.76 1.83 2.00 1.59 1.56 0.60 0.47 0.49
(2.02) (1.23) (1.40) (2.43) (1.52) (1.52) (1.33) (0.79) (0.90)
FIN (+) 18.82
*
21.20
*
18.05
*
13.17 14.41
*
11.73 9.42
***
9.94
***
9.13
***
(2.79) (3.47) (2.65) (2.02) (2.48) (1.70) (6.32) (6.99) (6.07)
TOBIN Q (+) À1.54 À1.88 À1.83 À2.18 À2.05 À2.45 À0.44 À0.48 À0.51
(0.86) (1.22) (1.24) (2.49) (2.24) (3.25) (0.64) (0.75) (0.85)
VOLAT (+) À4.69 4.05 3.56 À1.44 7.45 8.39 À1.93 À0.05 À0.08
(0.17) (0.11) (0.09) (0.02) (0.55) (0.72) (0.26) (0.01) (0.01)
ROA (+) À2.56 À1.31 0.85 9.49 11.32 13.73 À5.01 À5.57 À4.54
(0.02) (0.00) (0.01) (0.36) (0.51) (0.81) (0.62) (0.74) (0.52)
LEV (+) 27.27
***
26.58
***
27.63
***
23.80
***
23.01
***
22.67
***
7.29
***
7.13
***
7.16
***
(21.38) (19.77) (20.98) (21.76) (21.04) (20.89) (13.99) (13.20) (13.71)
SIZE (+) 9.70
***
10.06
***
9.67
***
7.90
***
8.05
***
7.81
***
2.61
***
2.68
***
2.57
***
(152.17) (160.43) (155.71) (140.36) (148.80) (145.05) (101.38) (106.75) (100.57)
NEW (+) À17.57
**
À17.63
**
À18.22
**
À12.16
*
À13.09
**
À12.73
**
À5.62
**
À5.64
**
À5.73
**
(5.00) (4.81) (5.40) (3.38) (3.96) (3.86) (4.55) (4.45) (4.76)
CAPIN (+) 33.38
***
26.96
**
30.32
***
28.22
***
21.26
**
24.96
***
9.67
**
8.04
**
8.82
**
(8.33) (5.26) (6.97) (8.62) (5.06) (7.06) (5.18) (3.54) (4.34)
Industry e?ects Yes Yes Yes Yes Yes Yes Yes Yes Yes
Log-likelihood À497.98 À499.55 À495.88 À435.27 À434.96 À431.62 À359.50 À360.01 À357.84
N = no of ?rms 191 191 191 191 191 191 191 191 191
Dependent variables are disclosures scores as indicated by the columns. The expected signs for the control variables are presented in
brackets. Coe?cients are estimated by maximum likelihood (Tobit regressions). The signi?cance levels are based on Chi-squared
statistics (presented in parentheses). All control variables are de?ned in Table 3.
***
,
**
,
*
represent signi?cance levels (two-tailed) at 1%,
5% and 10%, respectively.
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 321
development’. However, we ?nd similar results for
both hard and soft disclosure scores: the estimated
coe?cients for our two EP measures are positive
and signi?cant for both hard and soft disclosures
as predicted by H1a.
Intra-industry analysis
Following Lang and Lundholm (1993, 1996)
and Healy et al. (1999) we supplement the above
tests with an intra-industry approach which seeks
to explain the within-industry variation in the
ranks of disclosure scores using the within-indus-
try variation in the ranks of our independent
variables. There are two reasons justifying this
approach: (1) some of our variables (i.e., TRI/
Sales) can be compared within an industry but
cannot be meaningfully compared across indus-
tries due to di?erences in production processes;
(2) the distribution of our disclosure scores for
the Metal and Mining sector are quite low, so tak-
ing intra-industry ranks and then pooling facili-
tates the i.i.d assumption requirement of our
OLS regression model.
We rank the dependent and the independent
variables within industry and then we pool the cor-
responding percentiles across the ?ve sectors. The
regression results are presented in Table 5. They
are generally the same as in Table 4 and indicate
a positive association between the environmental
Table 5
Intra-industry rank regressions (OLS analysis)
Dependent variables
Total disclosures Hard disclosures Soft disclosures
(1) (2) (3) (1) (2) (3) (1) (2) (3)
Intercept 13.91
*
15.30
**
11.08 20.03
***
19.03
***
16.20
***
15.38
**
16.42
***
13.65
***
(1.84) (2.03) (1.48) (2.65) (2.53) (2.18) (1.96) (2.09) (3.41)
% Recycled (±) 0.17
***
0.15
***
0.15
***
0.14
***
0.11
**
0.11
**
(3.56) (3.19) (3.05) (2.98) (2.34) (2.24)
ÀTRI/Sales (±) 0.16
***
0.14
***
0.17
***
0.16
***
0.09
*
0.09
*
(3.32) (2.93) (3.48) (3.42) (1.93) (1.73)
J–F coe?cient (À) 0.07 0.05 0.05 0.05 0.04 0.03 0.18
***
0.17
***
0.17
***
(1.57) (1.07) (1.19) (1.09) (0.84) (0.66) (3.74) (3.51) (3.46)
FIN (+) 0.06 0.07 0.06 0.06 0.07 0.07 0.03 0.04 0.04
(1.24) (1.45) (1.38) (1.35) (1.50) (1.51) (0.69) (0.79) (0.75)
TOBIN Q (+) À0.04 À0.05 À0.06 À0.04 À0.06 À0.07 À0.02 À0.03 À0.04
(À0.67) (À0.89) (À1.07) (À0.72) (À0.98) (À1.21) (À0.36) (À0.50) (À0.62)
VOLAT (+) À0.06 À0.04 À0.05 À0.06 À0.03 À0.05 À0.04 À0.03 À0.04
(À1.15) (À0.83) (À0.99) (À1.15) (À0.54) (À0.97) (À0.84) (À0.60) (À0.76)
ROA (+) 0.06 0.06 0.06 0.06 0.08 0.06 0.03 0.02 0.02
(1.09) (1.01) (0.97) (1.09) (1.33) (0.94) (0.44) (0.37) (0.36)
LEV (+) 0.17
***
0.17
***
0.17
***
0.15
***
0.15
***
0.15
***
0.16
***
0.15
***
0.15
***
(3.51) (3.40) (3.44) (3.15) (3.02) (3.07) (3.12) (2.98) (3.04)
SIZE (+) 0.43
***
0.44
***
0.43
***
0.40
***
0.41
***
0.39
***
0.36
***
0.37
***
0.36
***
(8.39) (8.55) (8.53) (7.67) (8.03) (7.84) (6.77) (6.94) (6.81)
NEW (+) À0.11
**
À0.11
**
À0.11
**
À0.09
*
À0.09
**
À0.09
**
À0.08
*
À0.08
*
À0.08
*
(À2.34) (À2.26) (À2.42) (À1.92) (À1.99) (À2.03) (À1.65) (À1.64) (À1.68)
CAPIN (+) 0.11
**
0.10
**
0.10
**
0.10
**
0.08
*
0.09
*
0.13
**
0.13
**
0.12
**
(2.20) (2.13) (1.97) (2.07) (1.75) (1.79) (2.54) (2.54) (2.39)
Adj. R
2
40.83% 40.33% 43.22% 36.38% 37.32% 39.94% 34.98% 34.36% 35.69%
N = no of ?rms 191 191 191 191 191 191 191 191 191
Dependent variables are disclosures scores as indicated by the columns. The expected signs for the control variables are presented in
brackets. All variables are ranked within industry. Coe?cients are estimated by OLS regressions using the ranked variables. The
signi?cance levels are based on t statistics (presented in parentheses). All control variables are de?ned in Table 3.
***
,
**
,
*
represent
signi?cance levels (two-tailed) at 1%, 5% and 10%, respectively.
322 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
performance measures and the discretionary envi-
ronmental disclosures (total scores). Both (minus)
TRI/Sales and % recycled complement each other
and neither is redundant for the total sample or for
the hard and soft groups considered separately.
We ?nd similar results when we split the total
scores into hard and soft disclosure scores: the esti-
mated coe?cients for our two environmental per-
formance measures are positive and signi?cant as
predicted by H1a. The results for the control vari-
ables are also generally comparable to Table 4.
The Janis–Fadner coe?cient is now positively
associated with the level of soft disclosures, which
is opposite to the negative association predicted by
the socio-political theories.
A revised role for socio-political theories
Overall, the results in the previous sections
point to economic disclosure theory and not
socio-political theory as being robust in predicting
the level of discretionary environmental disclosure.
If the focus of enquiry is switched to disclosure
strategies, it may be that socio-political theories
explain interesting patterns in the data. While we
leave this possibility to future research, we o?er
in this section some preliminary evidence on the
question.
Legitimacy theory predicts that ?rms with
threatened legitimacy are likely to make self-serv-
ing disclosures referred to as ‘‘legitimization’’
(see, for example, Adams, 2004; Gray et al.,
1995, p. 54 & Hughes et al., 2001, p. 219). One
example of legitimization is for a poor environ-
mental performer to make soft claims to be com-
mitted to the environment which are not readily
veri?able. In fact, it is apparent from an inspection
of Table 1 that many poor EP ?rms do make soft
claims to be committed to the environment. Focus-
ing on A5, Vision and Strategy Claims, over 90%
of our poor EP ?rms make a claim which is
awarded at least one mark in category A5. For
example, 55 of 61 poor EP ?rms (i.e., 0.90
*
61)
make a claim that is awarded a score of 1 for item
A5-2 which is a statement of corporate environ-
mental policy and/or commitment. Further, 33 of
our 61 poor EP ?rms are awarded a score under
A5-1 which involves a CEO statement on environ-
mental performance to stakeholders.
22
More formally, using the ratio of soft disclosure
scores to total awarded scores as a proxy for legit-
imization, we would expect to ?nd a negative rela-
tion between prior perceived legitimacy and this
ratio. Our tests based on the conditional sample
of ?rms with scored sustainability/web disclosures
appear in Table 6. We measure threats to legiti-
macy in two ways: group EP membership based
on the median % recycled and group media cover-
age based on the median Janis–Fadner Coe?cient.
There are 122 (95) ?rms with scored sustainability/
web disclosures (at least one environmental press
article in 2002) and available EP. As indicated in
Table 6, the ratio of soft/total scores is 50.95%
for poor EP ?rms, compared to 34.23% for good
EP ?rms. Similarly, the ratio of soft/total scores
is 47.54% for unfavorable media coverage ?rms,
compared to 32.58% for ?rms with favorable
media coverage. In both instances, the di?erence
is statistically signi?cant at the 1% level. Both
results are consistent with ?rms whose environ-
mental legitimacy is threatened to make soft claims
to be committed to the environment. This result is
predicted by legitimacy theory but cannot be
explained by economic disclosure theories, which
assume truth-telling and thus o?er no predictions
about biased disclosures.
These results hold when we repeat the tests
(untabulated) in a multivariate fashion using all
other control variables in Tables 4 and 5 and
employ the ratio of soft/total as the dependent var-
iable. There is a signi?cant negative association
between the lagged Janis–Fadner Coe?cient and
the ratio of soft/total scores, after including all
control variables. Similar results hold for our
two EP measures. This result is robust to the Tobit
analysis approach (Table 4) and the rank regres-
sion analyses approach (Table 5) and implies a
greater propensity for ‘‘legitimization’’ behavior
22
As anecdotal evidence, consider the following soft environ-
mental commitment claim made by one of our sample chemical
?rms in the bottom quartile of our environmental performance
ranking: ‘‘Senior management leads the industry with respect to
responsible care.’’
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 323
for ?rms whose environmental legitimacy is
threatened.
Sensitivity analysis
For each disclosure item in the index presented
in Table 1, we have replicated the di?erence tests
across the two EP groups using Wilcoxon rank
tests. Non-parametric Wilcoxon rank tests are
robust to the possibility that the data does not fol-
low a normal distribution. In addition, we have
bootstrapped the p-values of the two-sample para-
metric t-tests by re-sampling the data with replace-
ment.
23
Both approaches result in the same levels
of signi?cance as the ones reported for the para-
metric t-tests.
We verify whether the di?erence tests in disclo-
sure scores reported in Table 1 are robust to the
classi?cation of sample ?rms as good and poor
environmental performers, using the % recycled
measure. We split the ?rms in the two groups
based on the level of the alternative environmental
performance measure, TRI/sales. We still ?nd sig-
ni?cant di?erences between the good and the poor
groups across all main categories in the index.
Finally, we run a Logit model to investigate
whether the decision to provide any discretionary
environmental disclosures is a?ected by the envi-
ronmental performance of the sample ?rms. We
perform an inter-industry analysis using the same
control variables as discussed in section ‘Econo-
metric model’. Results (unreported) are consistent
with the main results from Table 4. Environmen-
tal performance is found to be one of the main
drivers of the probability that the ?rm provides
environmental disclosures in discretionary chan-
nels. Precent recycled and (minus) TRI/Sales are
signi?cantly positive at the 1% level when intro-
duced separately. When both are in the model,
the signi?cance level decreases to 10%. The J–F
coe?cient is not signi?cant. All models show a
good ?t (signi?cant likelihood ratios and almost
90% concordant observations) suggesting a good
model speci?cation.
Table 6
Comparisons of soft to total disclosure scores
Panel A: Group membership is based on median % recycled (N = 122 ?rms)
Disclosure categories Average score Di?erence (t-stat)
Good EP ?rms (N = 61) Poor EP ?rms (N = 61)
Soft/total (%) 34.23% 50.95% À16.72%
***
(3.99)
Panel B: Group membership is based on median Janis–Fadner coe?cient (N = 95 ?rms)
*
Disclosure categories Average score Di?erence (t-stat)
Favorable media coverage (N = 48) Unfavourable media coverage (N = 47)
Soft/total (%) 32.58% 47.54% À14.96%
***
(3.22)
This table presents average ratios of soft to total disclosure scores. Panel A presents di?erences in soft to total disclosure ratios across
good environmental performance (EP) ?rms and poor environmental performance ?rms. Good environmental performance (EP) ?rms
are ?rms that have the environmental performance measure (% recycled) above the industry median. Panel B presents di?erences in
soft to total disclosure ratios across favorable media coverage ?rms and unfavorable media coverage ?rms. Firms with favorable media
coverage are ?rms that have a Janis–Fadner Coe?cient above the sample median (see Bansal and Clelland, 2004 for details). Only ?rms
that have at least one article in Factiva are included. The signi?cance levels presented in the last column are from two-sample t-
statistics that test the di?erence between the groups.
***
,
**
,
*
represent signi?cance levels (two-tailed) at 1%, 5% and 10%, respectively.
Wilcoxon Rank tests and t-tests with Bootstrap Resampling provided similar results. Groupings based on mean values also provide
similar results.
23
We implement the bootstrap procedure by drawing with
replacement 20000 samples from each EP group. Prior to
resampling, the procedure mean-centers the data within each
group (for details, see Westfall & Young, 1993).
324 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
Conclusions
Previous empirical evidence provides mixed
results on the relationship between corporate envi-
ronmental performance and the level of environ-
mental disclosures. We revisit this relation by
testing competing predictions from economics
based and socio-political theories of voluntary dis-
closure using a more rigorous research design. In
particular, we improve the prior literature in two
important ways. First, the predictions of voluntary
disclosure theory relate to discretionary, not man-
datory, environmental disclosures. This study
focuses on purely voluntary disclosure media such
as corporate Internet web sites and stand-alone
environmental reports. Previous studies assessed
environmental disclosures mainly from annual
reports and other regulatory ?lings such as 10 Ks
and many of those studies rely on a Wiseman
(1982) based content analysis index to measure
the extent of environmental disclosures.
Second, in collaboration with an environmental
reporting expert, we develop a content analysis
index to assess the level of environmental disclo-
sure in environmental and social responsibility
reports or similar disclosures in the ?rm’s web site.
The index, which follows closely the Global
Reporting Initiative (2002) sustainability reporting
guidelines, di?ers from the Wiseman index in that
we focus on ?rm disclosures related to its commit-
ment to protect the environment.
Our results are as follows. We ?nd a positive
association between environmental performance
and the level of discretionary disclosures in envi-
ronmental and social reports or related web disclo-
sures. In other words, superior environmental
performers are more forthcoming in truly discre-
tionary disclosure channels, as predicted by eco-
nomics based voluntary disclosure theories. Our
?ndings are robust to two reliable environmental
performance measures that use actual toxic emis-
sion and waste management data. The ?rst one is
based on Toxics Release Inventory scaled by sales
data at ?rm level (i.e., TRI normalized by ?rm’s
operational scale) and the second one is percentage
of total toxic wastes that were treated or processed
by each ?rm. In addition, our ?ndings are not
a?ected when we assess the relative environment
performance within each industry in order to con-
trol for industry di?erences in pollution propensity.
Our results are inconsistent with the prediction of a
negative association from socio-political theories,
suggesting that these theories are not robust in pre-
dicting the level of discretionary disclosure.
Although there is separation in scores and good
EP ?rms disclose more, the scores of good EP
?rms are low relative to the expectation implied
by the 2002 GRI reporting guidelines and point
to the need for improvement in the years beyond
2003. Speci?c areas where improvement is required
include obtaining independent assurance of sus-
tainability reports and the disclosure of environ-
mental performance indicators.
Finally, our results suggest important directions
for future research. Speci?cally, we provide preli-
minary evidence that socio-political theories are
robust in predicting what is being said. In particu-
lar, we ?nd that ?rms whose environmental legiti-
macy is threatened make soft claims to be
committed to the environment. This behavior is
predicted by legitimacy theory but cannot be
explained by economics disclosure theory. Thus,
we argue that future environmental disclosure
research should move the focus of enquiry beyond
the level of disclosure.
Acknowledgements
We are grateful for comments and suggestions
from the Editor (Anthony G. Hopwood), two
anonymous reviewers and from Kate Bewley, Walt
Blacconiere, Denis Cormier, Gus De Franco,
Kathy Herbohn, Ole-Kristian Hope, Hai Lu, Mi-
chel Magnan, Naomi Soderstrom and seminar
participants at Chinese University of Hong Kong,
University of Queensland and University of Tor-
onto. We acknowledge the ?nancial support of
the Canadian Academic Accounting Association,
Canadian Institute of Chartered Accountants,
and the AIC Institute for Corporate Citizenship
at Rotman School of Management, University of
Toronto. We also thank concurrent session partic-
ipants and discussants at the June 2006, July 2006
and August 2006 meetings of the CAAA,
AFAANZ (a Best Paper Award) and AAA,
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 325
respectively. We thank Rod Lohin, Bill Swirsky,
and Alan Willis for their support and encourage-
ment for this project and Gauri Bhat, Grace Jin
and Dushyantkumar Vyas for help in data collec-
tion. Gordon Richardson thanks KPMG for their
generous ?nancial support.
References
Adams, C. A. (2004). The ethical, social and environmental
reporting-performance portrayal gap. Accounting, Auditing
and Accountability Journal, 17(5), 731–757.
Aerts, W., & Cormier, D. (2006). The association between
media legitimacy and corporate environmental communi-
cation. Working paper.
Aerts, W., Cormier, D., & Magnan, M. (2006). The interface
between print- and web-based corporate environmental
disclosure, ?nancial markets and the media. Working Paper.
Al-Tuwaijri, S. A., 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(5-6), 447–471.
Bansal, P., & Clelland, I. (2004). Talking trash: Legitimacy,
impression management, and unsystematic risk in the
context of the natural environment. Academy of Manage-
ment Journal, 47(1), 93–102.
Barth, M., & Kasznik, R. (1999). Share Repurchases and
Intangible Assets. Journal of Accounting and Economics,
28(2), 211–241.
Barth, M., & McNichols, M. (1994). Estimation and market
valuation of environmental liabilities relating to superfund
sites. Journal of Accounting Research Supplement, 32,
177–209.
Barth, M., McNichols, M., & Wilson, P. (1997). Factors
in?uencing ?rms’ disclosures about environmental liabili-
ties. Review of Accounting Studies, 2, 35–64.
Beets, S. D., & Souther, C. (1999). Corporate environmental
reports: the need for standards and an environmental
assurance service. Accounting Horizons, 13(2), 129–145.
Bewley, K., & Li, Y. (2000). Disclosure of environmental
information by Canadian manufacturing companies: A
Voluntary Disclosure Perspective. Advances in Environmen-
tal Accounting and Management, 1, 201–226.
Blacconiere, W. G., & Patten, D. M. (1994). Environmental
disclosures, regulatory costs, and changes in ?rm value.
Journal of Accounting and Economics, 18, 357–377.
Chan-Fishel, Michelle. (2002). Survey of climate change
disclosure in SEC ?lings of automobile, insurance, oil &
gas, petrochemical, and utilities companies. Friends of the
Earth – US, September.
Clarkson, P., Li, Y., & Richardson, G. (2004). The market
valuation of environmental expenditures by pulp and paper
companies. The Accounting Review, 79, 329–353.
Cormier, D., & Magnan, M. (1997). Investors’ assessment of
implicit environmental liabilities: An empirical investiga-
tion. Journal of Accounting and Public Policy, 16, 215–241.
Cormier, D., Magnan, M., & Morard, B. (1993). The impact of
corporate pollution on market valuation: Some empirical
evidence. Ecological Economics, 8, 135–155.
Dye, R. A. (1985). Disclosure of non-proprietary information.
Journal of Accounting Research, 123–145, Spring.
Franco, N.C. (2001). Corporate environmental disclosure:
opportunities to harness market forces to improve corporate
environmental performance. American Bar Association
Conference on Environmental Law, Keystone, Colorado,
March 8–11.
Frankel, R., McNichols, M., & Wilson, P. (1995). Discretionary
disclosure and external ?nancing. The Accounting Review,
70, 135–150.
Freedman, M., & Wasley, C. (1990). ‘The association between
environmental performance and environmental disclosure in
annual reports and 10 Ks’. Advances in Public Interest
Accounting, 3, 183–193.
Global Reporting Initiative (2002). Sustainable Reporting
Guidelines, www.globalreporting.org.
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.
Healy, P. M., Hutton, A. P., & Palepu, K. G. (1999). A Stock
Performance and Intermediation Changes Surrounding
Sustained Increases in Disclosure. Contemporary Accounting
Research, 16, 485–520.
Healy, P. M., & Palepu, K. G. (2001). Information asymmetry,
corporate disclosure, and the capital markets: A review of
the empirical disclosure literature. Journal of Accounting and
Economics, 31(1–3), 405–440.
Hughes, S. B., Anderson, A., & Golden, S. (2001). Corporate
environmental disclosures: Are they useful in determining
environmental performance? Journal of Accounting and
Public Policy, 20, 217–240.
Hutton, A., Miller, G., & Skinner, D. (2003). The Role of
Supplementary Statements with Management Earnings
Forecasts. Journal of Accounting Research, 41, 867–890.
Ilinitch, A., Soderstrom, N., & Thomas, T. (1998). Measuring
corporate environmental performance. Journal of Account-
ing and Public Policy, 17, 387–408.
Ingram, R. W., & Frazier, K. (1980). Environmental Perfor-
mance and Corporate Disclosure. Journal of Accounting
Research, 18(2), 614–622. Autumn.
Janis, I. L., & Fadner, R. (1965). The coe?cient of imbalance.
In H. D. Lasswell & N. Leites, et al. (Eds.), Language of
Politics (pp. 153–169). Cambridge MA: MIT Press.
Jensen, M., & Meckling, W. (1976). Theory of the Firm:
Managerial Behavior, Agency Costs, and Ownership Struc-
ture. Journal of Financial Economics, 3(4), 305–360.
King, Andrew, & Lenox, Michael. (2001). Does it really pay to
be green? An empirical study of ?rm environmental and
?nancial performance. The Journal of Industrial Ecology,
5(1), 105–116.
326 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
Lang, Mark, & Lundholm, Russell. (1993). Cross-sectional
determinants of analyst ratings of corporate disclosures.
Journal of Accounting Research, 31(2), 246–247. Autumn.
Lang, M., & Lundholm, R. (1996). Corporate disclosure policy
and analyst behavior. The Accounting Review, 71(4),
467–492.
Leftwich, R. W., Watts, R. L., & Zimmerman, J. L. (1981).
Voluntary corporate disclosure: The case of interim report-
ing. Journal of Accounting Research, 18, 50–77.
Li, Y., & McConomy, B. (1999). An empirical examination of
factors a?ecting the timing of environmental accounting
standard adoption and the impact on corporate valuation.
Journal of Accounting, Auditing and Finance, 14(3), 279–313.
Li, Y., Richardson, G. D., & Thornton, D. (1997). Corporate
disclosure of environmental information; theory and evi-
dence. Contemporary Accounting Research, 14(3), 435–474.
Lim, T. (2001). Rationality and Analysts’ Forecast Bias.
Journal of Finance, LVI(1), 369–385.
Lindblom, C. (1994). The implications of organizational
legitimacy for corporate social performance disclosure.
Paper presented at the Critical Perspectives on Accounting
Conference, New York.
Patten, Dennis (1992). Intra-industry Environmental Disclo-
sures in Response to the Alaskan Oil Spill: A Note on
Legitimacy Theory. Accounting, Organizations, and Society,
17, 471–475.
Patten, Dennis (2000). Changing Superfund disclosure and its
relation to the provision of other environmental informa-
tion. Advances in Environmental Accounting and Manage-
ment, 1, 101–121.
Patten, Dennis (2002). The relation between environmental
performance and environmental disclosure: A research note.
Accounting, Organizations, and Society, 27, 763–773.
Richardson, A., & Welker, M. (2001). Social disclosure,
?nancial disclosure and the cost of equity capital. Account-
ing, Organizations and Society, 26(7/8), 597–616.
Smith, C., & Watts, R. (1992). The Investment Opportunity Set
and Corporate Financing, Dividend and Compensation
Policies. Journal of Financial Economics, 32(1), 263–293.
Standard & Poor’s (2004). Risk and Opportunity: the best
practice in non-?nancial reporting, ?rst edition, ISBN 1-
903168-12-0.
Verbeek, M. (2004). A Guide to Modern Econometrics. John
Wiley and Sons.
Verrecchia, R. (1983). Discretionary disclosure. Journal of
Accounting and Economics, 5, 179–194.
Westfall, P., & Young, S. (1993). Resampling-Based Multiple
Testing: Examples and Methods for P-Value Adjustment.
John Wiley and Sons.
Wiseman, J. (1982). An evaluation of environmental disclosures
made in corporate annual reports. Accounting, Organiza-
tions and Society, 7(1), 553–563.
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 327
doc_617911647.pdf
Previous empirical evidence provides mixed results on the relationship between corporate environmental performance
and the level of environmental disclosures. We revisit this relation by testing competing predictions from economics
based and socio-political theories of voluntary disclosure using a more rigorous research design. In
particular, we improve on the prior literature by focusing on purely discretionary environmental disclosures and by
developing a content analysis index based on the Global Reporting Initiative sustainability reporting guidelines to
assess the extent of discretionary disclosures in environmental and social responsibility reports.
Revisiting the relation between environmental
performance and environmental disclosure:
An empirical analysis
Peter M. Clarkson
a,b
, Yue Li
c
, Gordon D. Richardson
c,
*
, Florin P. Vasvari
d
a
UQ Business School, The University of Queensland, Australia
b
Faculty of Business Administration, Simon Fraser University, Canada
c
Joseph Rotman School of Management, University of Toronto, 105 St. George Street, Toronto, Ontario, Canada M5S 3E6
d
London Business School, University of London, London, United Kingdom
Abstract
Previous empirical evidence provides mixed results on the relationship between corporate environmental perfor-
mance and the level of environmental disclosures. We revisit this relation by testing competing predictions from eco-
nomics based and socio-political theories of voluntary disclosure using a more rigorous research design. In
particular, we improve on the prior literature by focusing on purely discretionary environmental disclosures and by
developing a content analysis index based on the Global Reporting Initiative sustainability reporting guidelines to
assess the extent of discretionary disclosures in environmental and social responsibility reports. This index better cap-
tures ?rm disclosures related to its commitment to protect the environment than the indices employed by prior studies.
Using a sample of 191 ?rms from the ?ve most polluting industries in the US, we ?nd a positive association between
environmental performance and the level of discretionary environmental disclosures. The result is consistent with the
predictions of the economics disclosure theory but inconsistent with the negative association predicted by socio-political
theories. Nevertheless, we show that socio-political theories explain patterns in the data (‘‘legitimization’’) that cannot
be explained by economics disclosure theories.
Ó 2007 Elsevier Ltd. All rights reserved.
Introduction
An unresolved research issue in environmental
accounting is the empirical association between
the level (i.e., amount) of corporate environmental
disclosures and corporate environmental perfor-
mance (Al-Tuwaijri, Christensen, & Hughes,
0361-3682/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aos.2007.05.003
*
Corresponding author. Tel.: +1 416 946 8601.
E-mail addresses: [email protected] (P.M.
Clarkson), [email protected] (Y. Li), gordon.richard-
[email protected] (G.D. Richardson), fvasvari@london.
edu (F.P. Vasvari).
www.elsevier.com/locate/aos
Available online at www.sciencedirect.com
Accounting, Organizations and Society 33 (2008) 303–327
2004; Hughes, Anderson, & Golden, 2001; Patten,
2002). Accounting standard setters and securities
regulators are increasingly being made aware of
de?ciencies in corporate environmental disclosures
(Beets & Souther, 1999; Chan-Fishel, 2002;
Franco, 2001). The results of previous studies on
the relation between corporate environmental per-
formance and environmental disclosure in ?nan-
cial reports have been mixed. Patten (2002)
attributes the failure to ?nd a signi?cant and con-
sistent relation between environmental perfor-
mance and environmental disclosure to problems
in the research designs of existing research. These
problems include failure to control for other fac-
tors associated with the level of environmental
disclosure, inadequate sample selection, and inad-
equate measures of environmental performance
and disclosure.
This study seeks to revisit the relation between
environmental performance and the level of envi-
ronmental disclosure using a more rigorous
research design. We test two competing predic-
tions about the level of voluntary environmental
disclosures. Voluntary disclosure theory (Dye,
1985; Verrecchia, 1983) predicts a positive associa-
tion between environmental performance and the
level of discretionary environmental disclosure.
The notion is that superior environmental per-
formers will convey their ‘‘type’’ by pointing to
objective environmental performance indicators
which are di?cult to mimic by inferior type ?rms.
Inferior performers will choose to disclose less or
to be ‘‘silent’’ on their environmental performance,
thus being placed in a pool of ?rms where inves-
tors and other users ascribe the ‘‘average type’’
to that pool. What sustains this partial disclosure
equilibrium is proprietary costs associated with
disclosure about environmental performance (Ver-
recchia, 1983) and uncertainty as to whether the
?rm is informed regarding its type (Dye, 1985).
Socio-political theories including political econ-
omy, legitimacy theory, and stakeholder theory
(Patten, 2002), on the other hand, predict a negative
association between environmental performance
and the level of discretionary environmental disclo-
sures. These overlapping theories suggest that
social disclosure is a function of social and political
pressures facing the corporation. To the extent that
poor environmental performers face more political
and social pressures and threatened legitimacy,
they will attempt to increase discretionary environ-
mental disclosures to change stakeholder percep-
tions about their actual performance. Thus, we
have competing directional predictions from alter-
native theories, and the observed direction of asso-
ciation between environmental performance and
the level of discretionary disclosures will eliminate
one of the two predictions.
The predictions of the above theories relate to
discretionary, not mandatory, environmental dis-
closures. Previous studies assessed environmental
disclosures mainly from annual reports and other
regulatory ?lings such as 10 Ks and many of those
studies rely on a Wiseman (1982) based content
analysis index to measure the extent of environ-
mental disclosures. The Wiseman index focuses
on the ?nancial consequences of corporate envi-
ronmental activities and puts more weight on
quantitative disclosures. Using this measure, poor
environmental performers may actually have
higher disclosure scores than good performers
because they have greater exposures and must dis-
cuss any material ?nancial information in their
regulatory ?lings such as annual reports and
10 Ks. This may partially explain the inconclusive
?ndings in the previous literature and why Patten
(2002) ?nds a negative relation between environ-
mental disclosure and a toxics release inventory
(TRI) based environmental performance
indicator.
1
In collaboration with an environmental disclo-
sure expert, we develop a content analysis index
based on the global reporting initiative guidelines
(GRI) to assess the level of discretionary environ-
mental disclosures in environmental and social
responsibility reports or similar disclosures pro-
vided on the ?rm’s web site. This index di?ers from
Wiseman (1982) index, previously used in the liter-
ature, because we focus on ?rm disclosures related
to its commitment to protect the environment. Our
index potentially allows investors, regulators, and
1
Patten (2002) was aware of the problem of non-discretion-
ary disclosures in annual reports and dropped litigation
disclosures as a partial attempt to deal with this (see p. 768).
304 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
environmental stakeholder groups to infer envi-
ronmental performance ‘‘type’’ from the disclosure
score. This is valuable to users who seek to assess
the ?rms’ true environmental commitment and
related environmental exposures.
We focus on the 2003 environmental disclosures
of 191 ?rms drawn from ?ve industries: Pulp and
Paper, Chemicals, Oil and Gas, Metals and Min-
ing, and Utilities. These ?ve industries are consid-
ered to have a high pollution propensity and have
collectively been the subject of a whole range of
environmental regulations in the US in the past
30 or more years. The magnitude of the environ-
mental spending by these industries to comply
with the environmental regulations and the impact
of their operations on the natural environment
should be a major concern to investors and other
environmental stakeholder groups. Thus, ?rms in
these industries collectively form an ideal sample
to test the competing predictions of voluntary dis-
closure and socio-political theories.
In brief, our results are as follows. We ?nd a
positive association between environmental per-
formance and the level of discretionary disclosures
in environmental and social reports or related web
disclosures. In other words, superior environmen-
tal performers are more forthcoming in truly dis-
cretionary disclosure channels, as predicted by
the economics based voluntary disclosure theory.
Our result is inconsistent with the prediction of a
negative association arising from socio-political
theories such as legitimacy theory and stakeholder
theory. Further, using the Janis–Fadner coe?cient
of imbalance as a direct measure of perceived legit-
imacy, we fail to observe the negative association
between legitimacy and the level of disclosures
implied by socio-political theories.
2
Thus, our
results suggest that socio-political theories are
not robust in predicting the level of discretionary
environmental disclosures.
We do, however, ?nd that socio-political theo-
ries are helpful in predicting what is being said,
which moves the focus of enquiry beyond the sim-
ple level of discretionary disclosure. Using the
ratio of soft disclosure scores to total awarded
scores as a proxy for ‘‘legitimization’’, we show
that ?rms with unfavorable prior year media cov-
erage are more likely to make soft claims to be
committed to the environment which are not read-
ily veri?able. This behavior is not predicted by
economic disclosure theories, which assume
truth-telling. Thus, socio-political theories do
indeed explain additional patterns in the data.
The paper is organized as follows. Following lit-
erature review and hypothesis development, we
describes our content analysis disclosure index
and the measures of environmental performance
which we use in the study. We then present our
econometric model and preliminary empirical evi-
dence. The sections ‘Empirical results involving
the level of disclosure’ and ‘A revised role for
socio-political theories’ contain the main results
followed by sensitivity analysis. The ?nal section
summarizes the main ?ndings of the study with a
discussion of implications for future research.
Literature review and hypothesis development
Literature review
The existing literature in environmental
accounting research can be categorized into three
broad groups. The ?rst group of studies examines
the valuation relevance of corporate environmen-
tal performance information and has found that
such information is valuable to investors seeking
to assess environmental liabilities in di?erent set-
tings.
3
The second line of literature examines fac-
tors a?ecting managerial decisions to disclose
potential environmental liabilities. This group of
studies ?nds that there are strategic factors a?ect-
ing ?rms’ decisions to disclose environmental
liability information, especially when disclosures
2
The Janis–Fadner coe?cient of imbalance measures the
propensity of a ?rm’s prior year press articles pertaining to the
environment to be unfavorable. See section ‘Other control
variables speci?c to environmental disclosures’ for a detailed
de?nition.
3
See Cormier, Magnan, and Morard (1993), Blacconiere
and Patten (1994), Barth and McNichols (1994), Cormier and
Magnan (1997), Li and McConomy (1999), Richardson and
Welker (2001), & Clarkson, Li, and Richardson (2004).
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 305
are discretionary.
4
A third line of studies, one
which we discuss in the following paragraphs since
it is most relevant to this study, explores the rela-
tion between environmental disclosures and envi-
ronmental performance.
Ingram and Frazier (1980) examine the associa-
tion between the content of corporate environmen-
tal disclosure and corporate environmental
performance. The study was concerned with a lack
of corporate social responsibility disclosures in
annual reports due to their voluntary nature. The
authors scored environmental disclosures in 20
pre-selected content categories along four dimen-
sions; evidence, time, speci?city, and theme.
Ingram and Frazier (1980) proxied environmental
performance by a performance index devised by
the Council on Economic Priorities (CEP), a
non-pro?t organization specializing in the analy-
sis of corporate social activities. Forty ?rms
were selected from the 50 ?rms that were moni-
tored by the CEP. Regression results indicated
no association between environmental disclosure
and environmental performance, consistent with
authors’ prior expectation about an overall poor
quality of environmental disclosures in annual
reports.
Wiseman (1982) examines the extent of volun-
tary environmental disclosures made by corpora-
tions in their annual reports using a research
design almost identical to Ingram and Frazier
(1980). The study focuses on the 26 largest US
companies that were monitored by the CEP for
the 1972–1976 period. Wiseman designed an envi-
ronmental disclosure index covering 18 items in
four categories: economic factors (5 items), envi-
ronmental litigation (2 items), pollution abatement
activities (5 items), and environmental disclosures
that do not fall into the other three (6 items). In
addition, Wiseman assigned a score to each item
based on whether the disclosure is quantitative or
qualitative (3 for quantitative disclosure, 2 for
non-quantitative disclosure, 1 for mentioning in
general terms, 0 for no disclosure).
5
The CEP
rankings were used as a proxy for environmental
performance. Spearman rank order correlation
indicates that there is no signi?cant association
between the CEP environmental performance
rankings and the Wiseman environmental disclo-
sure index rankings.
Freedman and Wasley (1990) examine the rela-
tionship between corporate pollution performance
and pollution disclosures made in annual reports
and 10 K reports ?led with the SEC. Their sample
consists of 50 US companies in four industries
(Steel, Oil, Pulp and Paper, Electric Utilities).
Again, the CEP rankings are used as a proxy for
environmental performance. The authors measure
environmental disclosures in both annual and
10 K reports using the same indexing procedure
developed by Wiseman (1982). Spearman rank
order correlation tests are conducted to examine
the associations both between annual report dis-
closure indices and the CEP indices, and between
10 K disclosure indices and the CEP indices. The
results indicate that neither annual report environ-
mental disclosures nor the 10 K environmental
disclosures are indicative of ?rms’ actual environ-
mental performance.
Bewley and Li (2000) examine factors associ-
ated with the environmental disclosures in Canada
from a voluntary disclosure theory perspective.
The authors measure environmental disclosures
by 188 Canadian manufacturing ?rms in their
1993 annual reports using the Wiseman index. A
?rm’s pollution propensity (i.e., environmental
performance) is proxied by their industry member-
ship and by whether they report to the Ministry of
Environment under the National Pollution
Release Inventory program. The study ?nds that
?rms with more news media coverage of their envi-
ronmental exposure, higher pollution propensity,
and more political exposure are more likely to dis-
4
See Patten (1992), Li, Richardson, and Thornton (1997),
Barth, McNichols, and Wilson (1997), Li and McConomy
(1999), & Aerts, Cormier, and Magnan (2006).
5
Many environmental disclosure studies since then rely on
the Wiseman index in order to measure the extent of corporate
environmental disclosures. Few recognize the fact that the
Wiseman index places a heavy weight on the ?nancial conse-
quences of corporate environmental activities, most of which
are required disclosures in 10 Ks for public companies regis-
tered with the SEC.
306 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
close general environmental information, suggest-
ing a negative association between environmental
disclosures and environmental performance.
Hughes et al. (2001) examine environmental dis-
closures made by 51 US manufacturing ?rms for
1992 and 1993. Again, the authors use a slightly
modi?ed Wiseman index to measure environmen-
tal disclosures made within the President’s letter,
MD&A, and note section, and then assess whether
environmental disclosures are consistent with envi-
ronmental performance ratings (good, mixed, and
poor) by the CEP. Although the study ?nds no dif-
ference in environmental disclosures between good
and mixed groups, ?rms rated as poor environ-
mental performers by the CEP tend to make sub-
stantially more environmental disclosures under
the Wiseman disclosure index. The authors attri-
bute this ?nding to increased scrutiny in 1992
and 1993 by the FASB and SEC with respect to
environmental disclosures, which forces poor per-
formers to make more disclosure as they are sub-
ject to more remediation activities.
Patten (2002) identi?es three issues in the previ-
ous studies in this area; (1) failure to control for
other factors, (2) inadequate sample selection;
and (3) inadequate measures of environmental per-
formance. Since the CEP only followed a small
group of ?rms in only four industries, reliance on
the CEP for sample selection may be problematic.
In addition, the CEP did not use the same criteria
and consistent methodology to assess corporate
environmental performance in di?erent industries.
To overcome this issue, Patten uses TRI data, nor-
malized by sales, to proxy for environmental per-
formance. Using a sample of 131 US ?rms from
24 di?erent industries, and a modi?ed Wiseman
index measure and line count of environmental
disclosures in 1990 annual reports, Patten ?nds
that TRI/sales are positively associated with both
measures of environmental disclosures, suggesting
a negative relation between environmental perfor-
mance and environmental disclosures.
More recently, Al-Tuwaijri et al. (2004) explore
the relations among environmental disclosure,
environmental performance and economic perfor-
mance using a simultaneous equations approach.
Similar to Patten (2002) and Al-Tuwaijri et al.
(2004) use TRI based data to assess environmental
performance. Speci?cally, they assess environmen-
tal performance as the percentage of total waste
generated that is recycled. The authors measure
environmental disclosure using a content analysis
in four categories (potential responsible parties’
designation, toxic waste, oil and chemical spills,
and environmental ?nes and penalties). These dis-
closures are largely non-discretionary, in contrast
to the discretionary disclosures which we examine.
They ?nd a positive association between environ-
mental performance and environmental disclosure.
In summary, the existing studies ?nd mixed
results on the relation between environmental dis-
closure and environmental performance. One rea-
son for the inconclusive ?ndings is due to the
choice of non-discretionary disclosure channels
and use of the Wiseman (1982) index. It is self evi-
dent that, as environmental problems and expo-
sures increase, non-discretionary disclosures in
regulated channels such as annual reports and
10 Ks should increase. Thus, a negative associa-
tion between environmental performance and con-
tent analysis scores in annual reports and 10 Ks
may be driven by non-discretionary disclosures.
The disclosure theories reviewed in the next sec-
tion, on the other hand, apply to discretionary dis-
closures. Thus, inferences about the robustness of
these theories are confounded when the disclosure
media are formal channels like the annual report
and the 10 K. Adding to this problem is the heavy
weight the Wiseman index places on disclosures
about the ?nancial consequences of environmental
activities, whereas our index places more weight on
disclosures that reveal true (but unobservable)
environmental performance. By focusing exclu-
sively on environmental and social responsibility
reports or similar disclosures on ?rms’ web sites,
locations where disclosures are purely discretion-
ary, and with an index that aims at revealing per-
formance ‘‘type’’, we enhance the reliability of
inferences about the true direction of association
between environmental performance and discre-
tionary disclosure.
Hypothesis development
The voluntary disclosure literature suggests that
companies have incentives to disclose ‘‘good
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 307
news’’ to di?erentiate themselves from companies
with ‘‘bad news’’ in order to avoid the adverse
selection problem (Dye, 1985; Verrecchia, 1983).
While these theories do not pertain to environmen-
tal performance per se, they are applied to this set-
ting by Bewley and Li (2000) and Li et al. (1997).
Companies with superior environmental perfor-
mance due to their proactive environmental strat-
egy have incentives to inform investors and other
stakeholders of their strategy by voluntarily dis-
closing more environmental information. Put sim-
ply, they seek to reveal their performance type,
something not directly observable to investors
and other stakeholders, through direct voluntary
disclosures that cannot be easily mimicked by poor
environmental performers. In doing so, they
potentially increase ?rm valuation since knowl-
edgeable investors will infer that exposures and
latent environmental liabilities are lower for good
as opposed to poor environmental performers.
6
Thus, voluntary disclosure theory predicts a
positive association between environmental per-
formance and the level of discretional environmen-
tal disclosure.
Turning to the predictions of socio-political the-
ories, Gray, Kouhy, and Lavers (1995) and Lindb-
lom (1994) argue that companies whose social
legitimacy is threatened have incentives to increase
environmental disclosures to: (1) educate and
inform relevant publics about (actual) changes in
their performance, (2) change perceptions about
their performance, (3) de?ect attention from the
issue of concern by highlighting other accomplish-
ments, and (4) seek to change public expectations
of their performance.
7
According to Patten (2000,
2002), socio-political theories predict a negative
association between corporate environmental per-
formance and level of discretional environmental
disclosure. Thus, the two competing theories pro-
vide opposite predictions on how environmental
performance may a?ect discretionary environmen-
tal disclosure strategies. Our hypotheses now fol-
low (stated in the alternate form):
H1a: Environmental performance and the level of
discretionary environmental disclosures are
positively associated, as implied by econom-
ics based voluntary disclosure theories.
H1b: Environmental performance and the level of
discretionary environmental disclosures are
negatively associated, as implied by the
socio-political theories.
Research design
Environmental performance indicators
A key research design issue in this study is to
develop a reliable proxy for a ?rm’s environmental
performance. The di?culty in assessing environ-
mental performance is well documented in the lit-
erature (see, for example, Ilinitch et al., 1998).
Since we seek to assess relative environmental per-
formance in this study, we follow the existing liter-
ature and develop our relative environmental
performance proxy using the actual pollution dis-
charge data from the US Environmental Protec-
tion Agency’s (EPA) TRI database (King &
Lenox, 2001). Speci?cally, we ?rst aggregate the
total toxic releases (in pounds) and the toxic waste
treated or processed for each of our sample ?rms
in 2003, as reported by the EPA in 2005 (EPA
annually reports the data at the plant level, with
a two year lag). To verify the accuracy of our
aggregation procedure, we compared the TRI
measures obtained at the ?rm level with those pro-
vided by the Investor Responsibility Research
Center in their proprietary database. We found
that the di?erences were minor.
Our ?rst measure is the total toxic waste that is
treated, recycled or processed as a percentage of
the total toxic waste generated by each ?rm (%
recycled). This measure is similar to the one used
6
The assumption that environmental performance ‘‘type’’ is
not readily observable to investors and other stakeholders is
supported by the di?culties environmental researchers have
had obtaining reliable measures of environmental performance
that are comparable across companies in the same industry and
across industries (see Al-Tuwaijri et al., 2004 & Ilinitch,
Soderstrom, & Thomas, 1998).
7
Socio-political theories include political economy, legiti-
macy theory and stakeholder theory (see Patten, 2002). We do
not di?erentiate them in this study as they have the same
prediction with respect to the relation between environmental
performance and environmental disclosure.
308 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
by Al-Tuwaijri et al. (2004). We supplement our
analysis with an alternative measure, the ratio of
TRI to total ?rm sales. This gives us pounds of
toxic releases per thousand dollars of sales for each
of our sample ?rms (TRI/sales). However, if ?rms
are not homogenous in terms of production pro-
cesses within one of our ?ve industries (Pulp and
Paper, Chemicals, Oil and Gas, Metals and Min-
ing, and Utilities), such that TRI/Sales is not
directly comparable across ?rms in that industry,
the recycling measure reduces this noise and facil-
itates pooling.
Since industry pollution propensity di?ers sig-
ni?cantly, we alternatively rank the above two
measures within each industry. Thus, the %Recy-
cled ranks and TRI/sales ranks are our proxies
for a ?rm’s relative performance within its indus-
try. Similar relative performance measures are also
used in Clarkson et al. (2004).
Environmental disclosure index
As discussed in section ‘Literature review and
hypothesis development’, from the economic the-
ory perspective, superior environmental perfor-
mance (EP) types are hypothesized to seek
credible direct disclosures to reveal their (unob-
servable) performance type. A crucial property of
these disclosures is that they focus on objective,
‘‘hard’’ measures that cannot be easily mimicked
by poor environmental performers. Thus, reliable
inferences about theories like Verrecchia (1983)
and Dye (1985) require a content analysis disclo-
sure index that puts a heavy emphasis on objective
measures of performance as opposed to soft (i.e.,
not easily veri?able) claims to be committed to
the environment.
8
As a simple example, consider
a good and a poor EP type ?rm in the same indus-
try. The good EP ?rm will voluntarily disclose
objective measures of environmental impact (e.g.,
quantitative environmental performance indica-
tors) and will benchmark its performance relative
to the industry, something the poor EP ?rm will
not want to do. Thus, the good EP ?rm will
emphasize discretionary disclosures that are hard
to mimic. We assume these hard disclosures are
truthful, in that a ?rm would face litigation expo-
sure if caught lying by informed stakeholders in
social responsibility reports or web related
disclosures.
9
What is striking in all this is that there is a
demand by environmental stakeholders for pre-
cisely the same thing: hard, objective measures of
environmental performance in social responsibility
reports, so that poor EP performers cannot mimic
good EP performers by soft, unveri?able claims to
be committed to the environment. Indeed, good
EP performers and environmental stakeholders
have joined forces to develop standards for ?rms
preparing social responsibility reports that put a
premium on hard, objective measures.
The Global Reporting Initiative (GRI) was
launched in 1997 as a joint initiative of Coalition
for Environmentally Responsible Economies, a
US non-government organization and the United
Nations Environmental Program. The overall goal
of the initiative is to develop a globally accepted
reporting framework to enhance the quality, rigor,
and utility of sustainability reporting (Global
Reporting Initiative, 2002). The GRI Guidelines
follow 11 principles (transparency, inclusiveness,
auditability, completeness, relevance, sustainabil-
ity context, accuracy, neutrality, comparability,
clarity, and timeliness) to ensure that sustainability
reports (1) present a reasonable and balanced
account of economic, environmental, and social
performance, (2) facilitate comparison over time
and across organizations, and (3) credibly address
issues of concerns to stakeholders. The ?rst set of
GRI Guidelines was published in 1999 as an Expo-
sure Draft and several revisions have followed
since then. For the purpose of this study, we rely
on the GRI Sustainability Reporting Guidelines
published in 2002.
8
For a related discussion on the need for objective and
veri?able disclosures to achieve the separation predicted by
Verrecchia (1983) see Hutton, Miller, and Skinner (2003).
9
As anecdotal evidence in support of this argument, Green-
peace issued a press release on October 14, 1994 accusing
MacMillan Bloedel of deliberately lying to the public by
claiming that, in 25 years, the company had been convicted of
only 15 environmental o?enses. Greenpeace identi?ed 26
convictions in the last four years.
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 309
We engaged an expert in the ?eld of environ-
mental reporting to help us develop a content anal-
ysis index suitable for ?rms’ sustainability reports
or the corresponding sections of a broad social
responsibility report or equivalent discussions on
the web.
10
From the outset, we agreed that the
construct we seek to measure is the extent of a
?rm’s disclosure in their sustainability report.
With this aim established, the expert convinced
us that the GRI reporting guidelines are consistent
with that purpose. Thus, the expert helped us
develop a scoring model containing 95 line items
that re?ect the spirit of the GRI guidelines. Table
1 contains the scoring model with a reference to
the corresponding section in the GRI guidelines.
Our disclosure index follows closely the report-
ing requirements of the GRI guidelines for the fol-
lowing reasons. Firms do not have to prepare
social responsibility reports or related web based
disclosures discussing their environmental impacts,
and if they voluntarily do so, they do not have to
adopt the GRI guidelines. The voluntary decision
by a ?rm to both prepare a social responsibility
report and use the GRI guidelines means that the
?rm has opted for a format (the GRI format) that,
by the intent of the GRI guidelines, will result in
hard disclosures not easily mimicked by the poor
EP types. Thus, a ?rm making a sincere attempt
to use the GRI guidelines will score high using
our content analysis index, which is precisely the
result we seek (i.e., the poor EP types will not want
to conform to GRI guidelines that place a pre-
mium, for example, on objective environmental
performance indicators). Of the 95 equally
weighted items in our disclosure index, 79 relate
to ‘‘hard’’ disclosure measures compared to only
16 for ‘‘soft’’ disclosure items, a proportion which
(according to the expert who helped us developing
the index) re?ects the spirit of the GRI guidelines.
We now turn to discuss our disclosure index in
more detail. Our disclosure index consists of seven
broad categories, A1–A7, of environmental disclo-
sures (see Table 1). We consider A1–A4 and A5–
A7 to represent ‘‘hard’’ and ‘‘soft’’ environmental
disclosures, respectively.
To score environmental disclosures in discre-
tionary channels and web related disclosures, we
accessed the internet web site of each sample ?rm
and identi?ed its environmental report, if any,
and any web based environmental disclosures.
We saved all such disclosures as of September
2004. This arbitrary choice of timing worked out
well for us as all ?rms discussed their ?scal 2003
environmental performance in the environmental
reports and related web based disclosures which
we obtained.
Hard disclosure items
Category A1 focuses on disclosures pertaining
to a ?rm’s governance structure and management
systems put in place with respect to environmental
protection. For instance, ?rms whose Board of
Directors have an environmental committee or
have implemented ISO 14001 will inform their
stakeholders of such commitments. A2 focuses
on the credibility of a ?rm’s disclosures in its envi-
ronmental report. Firms that obtained indepen-
dent veri?cation of their environmental reports,
and ?rms with their products and environmental
programs certi?ed by independent agencies and
third parties will receive higher scores in this cate-
gory. In A3, we assess the extent to which ?rms
disclose speci?c environmental performance indi-
cators, both about their actual pollution emissions
and their conservation and recycling e?orts. These
are the ‘‘hard’’ data that ?rms can disclose to con-
vince stakeholders about their environmental com-
mitments. In addition, we also award scores when
?rms disclose performance indicators with respect
to historical trends, the ?rms’ own emission reduc-
tion targets, and the industry average. Disclosing
actual performance indicators in the above context
can convey critical information for stakeholders to
assess the ?rm’s long-term environmental perfor-
mance (and commitments).
The ?nal category in the ‘‘hard’’ disclosure
group is A4, which re?ects a ?rm’s environmental
10
Alan Willis, CA, Project Director – Performance Reporting
Initiatives, The Canadian Institute of Chartered Accountants.
He was a member of the GRI Steering Committee since its
inception, and has been a member of the GRI Guidelines
development and revision working groups from 1998 to date.
He is also a judge for the Canadian Institute for Chartered
Accountants’ Corporate Reporting Awards.
310 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
Table 1
Index assessing the quality of discretionary disclosures about environmental policies, performance and inputs
Hard disclosure items Map to
GRI
Percentage of
?rms attaining
the item (%)
Average score
Good EP
?rms
(N = 61)
Poor EP
?rms
(N = 61)
(A1) Governance structure and management systems (max score is 6) 71.31 2.08 1.27
***
1. Existence of a Department for pollution control and/or management
positions for env. management (0–1)
3.1 37.70 0.44 0.31
2. Existence of an environmental and/or a public issues committee in
the board (0–1)
3.1 31.15 0.38 0.25
3. Existence of terms and conditions applicable to suppliers and/or
customers regarding env. practices (0–1)
3.16 21.31 0.31 0.11
***
4. Stakeholder involvement in setting corporate environmental policies
(0–1)
1.1, 3.10 27.05 0.36 0.18
**
5. Implementation of ISO14001 at the plant and/or ?rm level (0–1) 3.14, 3.20 45.90 0.51 0.41
6. Executive compensation is linked to environmental performance
(0–1)
3.5 4.92 0.08 0.02
*
(A2) Credibility (max score is 10) 83.61 2.88 1.95
**
1. Adoption of GRI sustainability reporting guidelines or provision of
a CERES report (0–1)
3.14 12.30 0.21 0.03
**
2. Independent veri?cation/assurance about environmental
information disclosed in the EP report/web (0–1)
2.20, 2.21 1.64 0.03 0.00
3. Periodic independent veri?cations/audits on environmental
performance and/or systems (0–1)
3.19 17.21 0.18 0.16
4. Certi?cation of environmental programs by independent agencies
(0–1)
3.20 15.57 0.23 0.08
**
5. Product Certi?cation with respect to environmental impact (0–1) 3.16 9.84 0.08 0.11
6. External environmental performance awards and/or inclusion in a
sustainability index (0–1)
51.64 0.52 0.51
7. Stakeholder involvement in the environmental disclosure process
(0–1)
1.1, 3.10 4.92 0.08 0.02
*
8. Participation in voluntary environmental initiatives endorsed by
EPA or Department of Energy (0–1)
3.15 33.61 0.31 0.36
9. Participation in industry speci?c associations/initiatives to improve
environmental practices (0–1)
3.15 54.10 0.71 0.38
***
10. Participation in other environmental organizations/assoc. to
improve. environmental practices (if not awarded under 8 or 9
above) (0–1)
3.15 40.98 0.52 0.30
***
(A3) Environmental performance indicators (EPI) (max score is 60)
a
73.77 10.19 6.00
***
1. EPI on energy use and/or energy e?ciency (0–6) EN3, 4, 17 41.80 1.46 0.75
***
2. EPI on water use and/or water use e?ciency (0–6) EN5, 17 30.33 1.07 0.49
**
3. EPI on green house gas emissions (0–6) EN8 31.97 1.10 0.59
**
4. EPI on other air emissions (0–6) EN9,10 43.44 1.45 1.08
5. EPI on TRI (land, water, air) (0–6) EN11 33.61% 1.05 0.65
*
6. EPI on other discharges, releases and/or spills (not TRI) (0–6) EN12, 13 28.69 1.15 0.43
***
7. EPI on waste generation and/or management (recycling, re-use,
reducing, treatment and disposal) (0–6)
EN11 50.00 1.44 1.04
8. EPI on land and resources use, biodiversity and conservation (0–6) EN6, 7 36.89 0.71 0.47
9. EPI on environmental impacts of products and services (0–6) EN14 4.10 0.13 0.00
*
10. EPI on compliance performance (e.g., exceedances, reportable
incidents) (0–6)
EN16 25.41 0.64 0.48
(continued on next page)
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 311
Table 1 (continued)
Hard disclosure items Map to
GRI
Percentage of
?rms attaining
the item (%)
Average score
Good EP
?rms
(N = 61)
Poor EP
?rms
(N = 61)
(A4) Environmental spending (max score is 3) 44.26 0.84 0.45
**
1. Summary of dollar savings arising from environment initiatives to
the company (0-1)
23.77 0.30 0.18
*
2. Amount spent on technologies, R& D and/or innovations to enhance
environ. perf. and/or e?ciency (0–1)
EN35 20.49 0.21 0.19
3. Amount spent on ?nes related to environmental issues (0–1) EN16 25.41 0.33 0.18
**
Soft disclosure items Map to
GRI
Percentage of
?rms attaining
the item (%)
Average score
Good EP
?rms
(N = 61)
Poor EP
?rms
(N = 61)
(A5) Vision and strategy claims (max score is 6) 95.90 3.48 3.04
1. CEO statement on environmental performance in letter to
shareholders and/or stakeholders (0–1)
1.1, 1.2 61.48 0.69 0.54
*
2. A statement of corporate environmental policy, values and
principles, environ. codes of conduct (0–1)
1.1, 1.2,
3.7
87.70 0.85 0.90
3. A statement about formal management systems regarding
environmental risk and performance (0–1)
3.19 58.20 0.57 0.59
4. A statement that the ?rm undertakes periodic reviews and evaluations
of its environ. performance (0–1)
3.19 37.70 0.47 0.27
**
5. A statement of measurable goals in terms of future env. performance
(if not awarded under A3) (0–1)
1.1, 1.2 27.05 0.31 0.22
6. A statement about speci?c environmental innovations and/or new
technologies (0–1)
1.1, 1.2 54.92 0.58 0.52
(A6) Environmental pro?le (max score is 4) 70.49 1.49 1.23
1. A statement about the ?rm’s compliance (or lack thereof) with
speci?c environmental standards (0–1)
GN 8 32.79 0.38 0.28
2. An overview of environmental impact of the industry (0–1) GN 8 22.13 0.26 0.20
3. An overview of how the business operations and/or products and
services impact the environment. (0–1)
GN 8 56.56 0.61 0.52
4. An overview of corporate environmental performance relative to
industry peers (0–1)
GN 8 24.59 0.26 0.23
(A7) Environmental initiatives (max score is 6) 72.95 1.93 1.34
**
1. A substantive description of employee training in environmental
management and operations (0–1)
3.19 30.33 0.39 0.21
**
2. Existence of response plans in case of environmental accidents (0–1) 22.95 0.30 0.16
*
3. Internal environmental awards (0–1) 13.11 0.18 0.08
4. Internal environmental audits (0–1) 3.19 3.20 34.43 0.38 0.31
5. Internal certi?cation of environmental programs (0–1) 3.19 9.84 0.15 0.05
*
6. Community involvement and/or donations related to environ. (if not
awarded under A1.4 or A2.7) (0–1)
SO1,
EC10
53.28 0.54 0.52
This table presents the index used toassess the discretionary disclosures about environmental policies, performance andinputs. Index items
are classi?ed in two categories: ‘‘hard’’ and ‘‘soft’’ disclosures. The second column presents the mapping of items in the index to the Global
Initiative Reporting (GRI) guidelines. The thirdcolumnpresents the percentage of ?rms whichmade disclosures onthat item(discretionary
channels considered are Environmental and/or Social Responsibility Reports, or similar disclosures in ?rms’ web site). Good environ-
mental performance (EP) ?rms are ?rms that have the environmental performance measure (% recycled) above the industry median. The
last twocolumns present the average score oneachitemfor each groupof ?rms. The signi?cance levels presented inthe last columnare from
two-sample t-statistics that test the di?erence between the good and the poor group.
***
,
**
,
*
represent signi?cance levels (two-tailed) at 1%,
5% and 10%, respectively. Wilcoxon Rank tests and t-tests with Bootstrap Resampling provided similar results. Sample size is 122 ?rms.
312 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
spending. We do not score disclosures pertaining
to environmental spending as a result of comply-
ing with the existing environmental regulations,
as such disclosures are largely non-discretionary
and appear in mandatory disclosure channels such
as 10 Ks and annual reports. Rather, we focus on
disclosures of dollar savings from existing environ-
mental programs and e?orts and discretionary
spending to further enhance future environmental
performance such as investing in new environmen-
tal technologies or environmentally related R&D
and innovations. We also include disclosures of
?nes related to environmental issues. Such penal-
ties are usually immaterial thus their reporting is
not mandatory. Nevertheless, the dollar amount
of ?nes is important to environmental stakehold-
ers to assess the level of true commitment to the
environment. In summary, our index design in
the A1–A4 ‘‘hard’’ disclosure categories makes it
relatively di?cult for poor environmental per-
formers to mimic the environmental disclosures
of good environmental performers.
Soft disclosure items
We measure a ?rm’s disclosures of vision and
environmental strategy claims in A5. For instance,
?rms often disclose broadly that they have an envi-
ronmental policy, that management is committed
to protecting the environment, etc. Such disclo-
sures can be genuine when put in the speci?c con-
text but they can also be deceiving as they lack
credibility and substantiation, and can be easily
mimicked.
11
A6 assesses the disclosure of a ?rm’s
environmental pro?le given the existing and forth-
coming environmental regulations. Finally, we
code a ?rm’s disclosures of its environmental ini-
tiatives in A7. Items coded here include employee
training in environmental management, existence
of response plans for environmental accidents,
internal environmental awards and audit, and
community involvement through scholarship and
donations. Again, these kinds of initiatives can
represent true commitment but they can also be
imitated by companies with no real commitments
to protecting the environment.
Econometric model
Model and variable descriptions
In order to test our hypotheses, we employ the
following econometric model:
VED ¼ b
0
þb
1
EP þb
2
J–F coefficient þb
3
FIN
þb
4
TOBIN Q þb
5
VOLAT þb
6
ROA
þb
7
LEV þb
8
SIZE þb
9
NEW
þb
10
CAPINþe
The variables in the regression above are de?ned
as follows:
VED – is a score of voluntary environmental
disclosures using web based disclosures as of Sep-
tember, 2004. We perform a content analysis using
our disclosure index (see Table 1).
EP – is an environmental performance proxy.
We use two alternative proxies to capture the envi-
ronmental performance of each ?rm. The ?rst is the
TRI emission scaled by total sales revenue. To
facilitate the interpretation of the results, we
reverse the sign of this variable. In other words,
the larger this measure is, the better the environ-
mental performance of the ?rm. The second mea-
sure is the percentage of toxic waste treated,
Table 1 (continued)
a
The scoring scale of environmental performance data is from 0 to 6. A point is awarded for each of the following items: (1)
Performance data is presented; (2) Performance data is presented relative to peers/rivals or industry; (3) Performance data is presented
relative to previous periods (trend analysis); (4) Performance data is presented relative to targets; (5) Performance data is presented
both in absolute and normalized form; (6) Performance data is presented at disaggregate level (i.e., plant, business unit, geographic
segment).
11
Consider, for example, items A5-2, 3 and 5. We classify
these statements as soft because they involve claims about
environmental management control systems without details or
substantiation. In contrast, the corresponding items in A1-1
and 2 are hard because they provide speci?c information about
the existence of the department, management positions or
board committees responsible for monitoring pollution control.
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 313
recycled or processed in the production. Large
recycling percentages imply environmentally pro-
active ?rms. Both measures are computed using
the public database made available by the US Envi-
ronmental and Protection Agency (EPA). We
aggregate the plant speci?c data at the ?rm level.
J–F coe?cient – the Janis–Fadner coe?cient of
imbalance, measured for the ?rm’s 2002 ?scal year
and de?ned in greater detail below.
FIN – is the amount of debt or equity capital
raised by the ?rm in the ?scal year 2004. It is the
sale of common stock and preferred shares minus
the purchase of common stock and preferred
shares (#108–#115) plus long term debt issuance
minus the long term debt reduction (#111–
#114).
12
The amount is scaled by the size of total
assets at the end of the ?scal year 2002.
TOBIN Q – is Tobin’s Q, measured as market
value of common equity (#25
*
#199) plus book
value of preferred stock (#10), book value of long
term debt (#9) and current liabilities (#5), divided
by book value of total assets (#6).
VOLAT – is stock price volatility, measured as
standard deviation of market adjusted monthly
stock return during ?scal year 2003.
ROA – is total return on assets measured as the
ratio of income before extraordinary items (#18)
at the end of ?scal year 2004 and total assets
(#6) at the end of ?scal year 2003.
LEV – is the leverage ratio, measured as the
ratio of total debt (#9 + #34) divided by total
assets (#6) at the end of ?scal year 2003.
SIZE – is the natural logarithm of the total
asset value measured as of the end of ?scal year
2003.
NEW – is the asset newness, measured as a ratio
of net properties, plant and equipment (#8)
divided by the gross properties, plant and equip-
ment (#7) at the end of ?scal year 2003.
CAPIN – is the capital intensity, measured as a
ratio of capital spending (#128) divided by total
sales revenues (#12) at the end of ?scal year 2003.
The control variables included in the multivari-
ate regression model besides our environmental
performance (EP) measures have been docu-
mented to be causes of voluntary disclosures in
the disclosure literature. We include them to avoid
a correlated omitted variables threat that EP is
standing in for some other well known determi-
nants of disclosure in other disclosure channels.
We divide the control variables in three groups:
variables that proxy for the bene?ts of voluntary
disclosures, variables that measure costs of volun-
tary disclosures and other control variables.
Bene?ts of voluntary disclosure
Financing: It is well known that ?rms that raise
?nancing in debt and equity markets have a higher
propensity for disclosures in voluntary channels
(Frankel, McNichols, & Wilson, 1995) to lower
their cost of capital. We use the amount of debt
and equity ?nancing raised by the ?rm in the ?scal
year following the measurement of the environ-
mental performance (FIN).
Information asymmetry: It is generally asserted
in the voluntary disclosure literature that manag-
ers seek to lower information asymmetry through
voluntary disclosures in order to lower the cost
of capital (Healy & Palepu, 2001). Our chosen
proxies for information asymmetry are: monthly
stock return volatility measured (VOLAT) over
the 12 month period represented by ?scal 2003
(Lim, 2001) and Tobin’s Q, based on the argument
that ?rms with greater unbooked intangibles and a
positive NPV investment opportunity set enjoy
larger Tobin’s Q (Barth & Kasznik, 1999; Smith
& Watts, 1992).
Firm performance: Lang and Lundholm (1993)
and others have shown that ?rms with superior
upcoming earnings performance have a higher dis-
closure propensity to reveal their ‘‘good news’’ to
?nancial markets. At the time of observing web
disclosures (September 2004), markets would
know ?scal year 2003 ROA so earnings for the
upcoming year would be ?scal 2004 ROA.
Leverage: A number of disclosure studies (e.g.,
Leftwich, Watts, & Zimmerman, 1981) have
argued that the monitoring demand for informa-
tion increases as ?rm debt increases, and empirical
evidence is consistent with managers being more
forthcoming, generally to facilitate the contracting
12
Numbers in brackets represent data items in the Compustat
Annual File.
314 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
demand for information. Agency costs of debt are
higher for ?rms with relatively more debt in their
capital structure (Jensen & Meckling, 1976), thus
voluntary disclosures are expected to increase with
leverage.
Costs of voluntary disclosures
Firm Size: Most voluntary disclosure studies
control for ?rm size (see, for example, Lang &
Lundholm, 1993) based on the assumption of
economies of scale with respect to information
production costs.
13
Proprietary costs: In Verrecchia (1983), a key
friction sustaining a partial disclosure equilibrium
is the existence of proprietary costs associated with
being forthcoming. In our setting, proprietary
costs pertain to the manager revealing information
to environmental regulators and other environ-
mental activist groups that increase the probability
of criticism, sanction or attack (see Li et al., 1997).
We assume that industry serves as a measure of
proprietary costs, since pollution propensity and
related monitoring by opponents is well known
to vary by industry. In our inter-industry regres-
sions, we control for industry ?xed e?ects in order
to control for di?ering proprietary costs and other
unidenti?ed factors that might vary by industry.
Other control variables speci?c to environmental
disclosures
Equipment age and annual capital spending:
Healy and Palepu (2001) describe a common criti-
cism in the voluntary disclosure literature involv-
ing endogeneity. EP and sustainability disclosures
might be joint endogenous variables driven by
some underlying exogenous variables such as the
level of investments in clean technologies, thus rep-
resenting a threat to causal inferences. Firms with
newer, cleaner technologies are likely to have a
superior environmental performance measure and
it is reasonable to assume that they will want
stakeholders to know about this superior environ-
mental performance in discretionary disclosure
channels. To address this threat, we control for
the average age of a ?rm’s equipment (NEW)
based on the argument that newer equipment is
expected to employ newer and less polluting tech-
nologies. For similar reasons, ?rms with higher
sustaining capital expenditures, as proxied by
CAPIN, are expected to have newer equipment
and may want to signal their environmental type
through more discretionary disclosures regarding
their environmental performance.
Favorable media coverage: Following Aerts and
Cormier (2006), Bansal and Clelland (2004) and
Janis et al. (1965), we measure lagged environmen-
tal legitimacy as the propensity for unfavorable
press articles using the Janis–Fadner coe?cient
of imbalance. This coe?cient ranges from À1
(unfavorable) to +1 (favorable), with zero imply-
ing neutral perceptions about the ?rm’s environ-
mental legitimacy. Following Aerts and Cormier
(2006) and Patten (2002), legitimacy theory pre-
dicts a negative association between lagged envi-
ronmental legitimacy and the level of voluntary
environmental disclosures.
14
The Janis–Fadner
coe?cient is measured as follows:
Janis–Fadner coefficient
¼
ðe
2
ÀecÞ
t
2
if e > c;
¼
ðec Àe
2
Þ
t
2
if c > e;
¼ 0 if e ¼ c;
13
Firm size has been regularly established as a determinant of
voluntary disclosure in the literature, so its e?ects must be
controlled for. Firm size is marginally signi?cant and positively
correlated with % recycled in the Chemical and Pulp and Paper
industries, but is insigni?cant for our other three industries. In
Table 4, we control for ?rm size as a covariate to ensure that
our EP measures are not standing in for size. In Table 5, ?rm
size is once again a covariate in our model. Thus, the role
played by ?rm size in environmental disclosure is controlled for
in both tables when we isolate e?ects due to EP.
14
In our sample of 191 ?rms, 126 have at least one article in
Factiva database related to the environment in ?scal 2002. The
126 ?rms generate 770 total articles related to the environment
during ?scal 2002. Of these, 393, 207, and 170 are coded by us
as unfavorable, neutral and favorable, respectively. For the 65
?rms with no environmental articles, the Janis–Fadner coe?-
cient is set to a default measure of zero. Thus, silence in the
media is interpreted to imply neutrality of perceptions about
environmental legitimacy.
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 315
where e is the number of favorable environmental
articles, c is the number of unfavorable articles,
and t is e + c.
Sample selection, summary statistics and univariate
empirical results
As mentioned in section ‘Introduction’, the
sample of this study consists of public US compa-
nies from ?ve polluting industries that report toxic
release data to the US Environmental Protection
Agency and have ?nancial and stock price data
in the Compustat and CRSP database.
15
The ?nal
sample contains 191 ?rms covered by the EPA-
TRI database for 2003 with adequate variables
available in Compustat and CRSP. The distribu-
tion across the ?ve industries is as follows: 27 ?rms
in the Pulp and Paper industry (14.14% of the sam-
ple), 63 ?rms in the Chemical industry (32.98% of
the sample), 18 ?rms in the Oil and Gas industry
(9.42% of the sample), 42 ?rms in the Metals and
Mining industry (21.99% of the sample) and 41
?rms in the Utilities industry (21.47% of the sam-
ple). Among these ?rms, 122 ?rms (63.87% of
the sample) chose to provide discretionary disclo-
sures about the environment in ?scal 2003: 55 of
those ?rms had stand alone environmental reports
and invariably also had supplementary web disclo-
sures, while the remaining 67 ?rms had discretion-
ary web disclosures but no stand alone
environmental reports. If a ?rm has no environ-
mental report or related discretionary web disclo-
sures, we classify such a ?rm as being ‘‘silent’’,
with the disclosure score set to zero for these 69
?rms.
16
We allow silent ?rms in the sample since
non-disclosure is a choice in a partial disclosure
equilibrium setting.
For ?rms providing discretionary environmen-
tal disclosures, Table 1 presents our scoring model
along with descriptive statistics as to the percent-
age of ?rms disclosing a particular item and the
di?erence in average disclosure scores across good
and poor EP ?rms. The table also shows the GRI
reference for each line item. A crucial assumption
of our scoring approach is that the disclosures are
discretionary. In support of that assumption, we
use only disclosures that came from ?rm environ-
mental reports (hereafter EP reports) or web based
disclosures other than ?rm annual reports or
10 Ks. We assume that EP reports/web disclosures
are purely discretionary, i.e., silence is always an
option.
In Table 1, A1–A4 summarize the separation in
average awarded scores across good and poor EP
performers, using the %Recycled as the measure
of environmental performance. We use the median
% recycled in a given industry to classify ?rms as
good versus poor EP performers.
17
Recall that,
according to H1a, good EP performers should
have higher scores for hard disclosure items that
are di?cult to mimic by poor EP performers.
The results in Table 1 con?rm that prediction for
the A1 category (Governance Structure and Man-
agement Systems). The average score for good
(poor) EP performers is 2.08 (1.27). Using a two-
tailed t-test the di?erence is signi?cant at the 1%
level. For the A2 category (Credibility), the aver-
age score for good (poor) EP performers is 2.88
(1.95), and the di?erence is again signi?cant at
the 5% level. Not surprisingly, good EP ?rms are
more likely to disclose that they adopt GRI guide-
lines. The di?erence in average scores 0.21 versus
0.03 is statistically signi?cant at the 1% level. It
is apparent from the scores for A2-2 that 2
(61 · 0.03) good EP ?rms obtained independent
assurance for their ER report/web disclosures
compared to zero poor EP ?rms. While this is con-
sistent with an attempt by good EP ?rms to signal
their type, the di?erence between the two types of
?rms is not statistically signi?cant.
15
We identify all ?rms with available TRI data and two-digit
SIC codes: 26 (pulp and papers), 28 (chemicals), 29 (oil and
gas), 33 (metals and mining), and 49 (utilities). For these
companies, we read the ?rm’s business descriptions and
dropped companies that identify material business operations
in industries outside of their primary two-digit SIC code.
16
We contacted the 69 ?rms with zero disclosures ?rst by
email and then by phone. None of them indicated that they
published a stand alone environmental report in 2003.
17
Our results in Table 1 are qualitatively una?ected when we
partitioned the sample using the mean % recycled.
316 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
Of special interest is category A3 (EPI indica-
tors), since this type of hard disclosure is not one
that poor EP ?rms will want to mimic. The aver-
age score for Good EP performers of 10.19 is well
in excess of the corresponding average score for
poor EP performers, 6.00 (the di?erence is signi?-
cant at the 1% level). Each item has a total score of
6 based on the dimensions indicated in Table 1.
The low scores for both good and poor EP per-
formers suggest that disclosures in this area are
less than ideal given the GRI guidelines. In unta-
bulated analyses, we analyzed the contribution of
each of the 6 dimensions to the total scores (i.e.,
‘‘hits’’) awarded in A3, for good versus poor EP
?rms: performance data presented represents
43.50% (48.44%) of total A3 scores awarded for
good (poor) EP ?rms, respectively; performance
data presented relative to peers represents only
1.06% (2.44%) of total A3 scores awarded for good
(poor) EP ?rms, respectively; trend analysis repre-
sents 28.86% (26.33%) of total A3 scores awarded
for good (poor) EP ?rms, respectively; perfor-
mance relative to targets represents 8.76%
(7.31%) of total A3 scores awarded, for good
(poor) EP ?rms, respectively; performance data
presented in both absolute and normalized form
represents 7.87% (8.63%) of total A3 scores
awarded, for good (poor) EP ?rms, respectively;
and, ?nally, performance data presented at the
disaggregated level represents 9.64% (6.68%) of
total A3 scores awarded, for good (poor) EP ?rms,
respectively. Thus, it is apparent that GRI guide-
lines with respect to performance relative to peers
are not being followed by either good or poor EP
performers. This is not surprising, as it is di?cult
for both types of ?rms to decide on appropriate
‘‘peers’’ given di?erences in production processes
across ?rms within a given industry. It is precisely
this dilemma that makes relative environmental
performance ‘‘unobservable’’ to the typical inves-
tor or stakeholder, creating the potential for the
setting of a partial disclosure equilibrium where
some ?rms are more transparent than others in
their report/web disclosures, and some ?rms are
entirely silent.
The ?nal hard category is A4 (disclosures per-
taining to discretionary environmental spending).
For that category, the average score for good
(poor) EP performers is 0.84 (0.45). The di?erence
between the good and poor groups is signi?cant at
5%. In this category we ?nd that, on average, good
EP ?rms disclose signi?cantly more often amounts
spent on ?nes than the poor ?rms. Economics
based voluntary disclosure theories predict that
good EP ?rms will be more forthcoming about
the dollar amount spent on ?nes because this
amount will be lower than the corresponding
amount for poor EP ?rms, consistent with supe-
rior environmental performance within the indus-
try. To validate that disclosing the amount spent
on ?nes signals a ?rm’s commitment to the envi-
ronment, we examine the dollar amount spent on
?nes and the number of environmental violations
in ?scal year 2000 for the 100 of our sample ?rms
(56 good EP ?rms and 44 poor EP ?rms) covered
in the Corporate Environmental Pro?le Database
developed by the Investor Responsibility and
Research Center. In untabulated results, we ?nd
that the ?fty-six good EP ?rms had an average
of 1.13 violations per ?rm and 0.17 cents of ?nes
per thousand of dollars of sales. In comparison,
the forty four poor EP ?rms had an average of
2.63 violations per ?rm and 0.42 cents of ?nes
per thousand dollars of sales. Using a two-tailed
t-test, the di?erence in the number of violations
(scaled dollar ?nes) between the two groups is sta-
tistically signi?cant at 0.02 (0.01) level.
18
Table 1 contains corresponding statistics for the
soft disclosure categories A5–A7. Consistent with
H1a, good EP performers have signi?cantly
greater soft discretionary disclosures for A7 (envi-
ronmental initiatives) relative to poor EP perform-
ers. However, there is no signi?cant di?erence for
A5 (vision and strategy) and A6 (environmental
pro?le). Overall, the results for our soft disclosure
scores suggest less separation between good and
poor EP ?rms, a ?nding which we explore in more
depth in section ‘A revised role for socio-polittical
theories’ below.
18
An alternative interpretation of our results for ?nes disclo-
sures is that the ?rms disclosing such ?nes are seeking to
legitimize the violation of environmental regulation and
requirements. This pattern of disclosure behaviour is predicted
by legitimacy theory (Patten, 2000). We leave the validation of
this alternative explanation to future research.
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 317
In Table 2, Panel A we present descriptive sta-
tistics on environmental disclosure scores by
industry. The lowest average score (out of 95) is
10.08 and is obtained by the Metals and Mining
Industry while the highest score, 25.21, is obtained
by the Pulp and Paper Industry.
19
In Table 2,
Panel B we show that the average ratios of TRI/
Sales for Pulp and Paper and Metals and Mining
are 2.34 (i.e., 2.34 pounds of toxic emissions per
thousand dollar sales) and 1.72, respectively. Thus,
it appears that industries with a high (low) pollu-
tion propensity are more (less) likely to provide
discretionary environmental disclosures to ?rm
stakeholders. This is by no means inconsistent
with H1a which conditions disclosure predictions
by environmental type at the ?rm level relative to
industry peers. Table 2 does con?rm di?erences
across sectors in both environmental disclosure
and pollution propensities suggesting the need
for industry controls.
Untabulated analysis for our 69 ?rms with zero
disclosure scores indicates that these ?rms have an
average TRI/Sales (% recycled) measure of 2.97
(59.01%) compared to 1.82 (65.61%) for the 122
?rms with non-zero disclosure scores. The di?er-
ences in both EP measures across the two groups
are signi?cant at the 5% level. This is generally
consistent with H1a, i.e., poor EP ?rms are more
likely to opt for silence.
In Table 3, Panel A we present descriptive sta-
tistics for independent variables used in the estima-
tion. The average ?rm has a negative ?nancing
variable (FIN) meaning that it reduces debt or
repurchases shares more than it raises new ?nanc-
ing. Also, on average, the ROA is about 5% and
the average leverage (LEV) is 33% of total assets.
The ?rm size measured by the logarithm of total
assets (SIZE) is 8.01 implying average total assets
in dollar terms of $3.01 bn, thus our sample con-
sists of relatively larger ?rms. The mean (and med-
ian) ?rm size is comparable with the median ?rm
size reported by Patten (2002) for his sample.
The average J–F coe?cient is À0.08.
We present in Panel B of Table 3 Pearson cor-
relations between the independent variables used
in the regressions. The Pearson correlation
between the negative of TRI/Sales and % recycled
is equal to 0.29 and is statistically signi?cant at 1%,
implying that (minus) TRI/Sales and % recycled
both measure environmental performance but the
overlap is modest.
Empirical results involving the level of disclosure
Inter-industry analysis
We present in Table 4 the results of inter-indus-
try multivariate regressions of disclosure scores on
the environmental performance measures and the
control variables. We estimate the regressions
using a Tobit analysis to account for the censoring
of the dependent variable at zero.
20
We estimate
the Tobit regressions by maximum likelihood
using a Newton–Raphson algorithm.
21
We run
three sets of equations based on the di?erent dis-
closure scores (total, hard and soft) used on the
dependent side. All regressions are inter-industry
analyses using dummy variables to control for
industry ?xed e?ects. We estimate the regressions
using each environmental performance variable
separately and then include both variables at the
same time.
The ?rst three columns present the results for
the total disclosure scores. As predicted by H1a,
the estimated coe?cients for our environmental
19
While our scoring scheme has 95 available points to allow
for rich variety of disclosures from one sustainability report to
the next, and from one industry to another, the attainable score
is lower than this, even for an excellent sustainability report.
For example, our top scoring ?rm, Weyerhaeuser, obtained 68
of the 95 available points, suggesting an e?ective maximum of
71%. This is comparable to the e?ective maximum in similar
scoring scheme used in ‘‘Risk and Opportunity: the Best
Practice in Non-?nancial Reporting’’ by Standard and Poor’s
(2004). Nonetheless, the low average scores (out of 95) for each
of the four industries point to an overall need for improvement
in sustainability reporting in the years beyond 2003.
20
As an alternative, we have re-run our regressions using
simple OLS and the inferences are unchanged.
21
In the Tobit model, the marginal e?ect of a change in an
independent variable on the dependent variable (i.e., disclosure
score) is the estimated coe?cient times the probability that the
?rm provides discretionary environmental disclosures (Verbeek,
2004).
318 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
Table 2
Descriptive Statistics for variables of interest
Panel A: Environmental disclosures (N = 122 ?rms)
Overall Pulp and paper Chemicals Metals and mining Oil and gas Utilities
(n = 122) (n = 24) (n = 41) (n = 13) (n = 13) (n = 31)
Hard disclosures (0–79) 12.88 (10.00) 16.95 (13.00) 12.19 (10.00) 5.76 (2.00) 14.31 (11.00) 13.00 (11.00)
0–53 0–53 0–39 0–42 1–33 0–31
(A1) Governance structure and
manage. systems (0–6)
1.68 (1.00) 1.95 (2.00) 1.95 (2.00) 1.31 (1.00) 2.00 (2.00) 1.10 (1.00)
0–5 0–5 0–5 0–5 0–5 0–5
(A2) Credibility (0–10) 2.42 (2.00) 3.71 (3.00) 2.14 (2.00) 1.23 (1.00) 2.77 (3.00) 2.13 (2.00)
0–9 0–9 0–6 0–6 0–5 0–6
(A3) Environmental performance 8.09 (6.50) 10.13 (6.00) 7.49 (5.00) 3.00 (0.00) 8.69 (4.00) 9.23 (9.00)
Indicators (0–60) 0–36 0–36 0–27 0–29 0–21 0–23
(A4) Environmental spending (0–3) 0.70 (0.00) 1.17 (1.00) 0.58 (0.00) 0.23 (0.00) 0.85 (1.00) 0.61 (0.00)
0–3 0–3 0–3 0–2 0–3 0–3
Soft Disclosures (0–16) 6.26 (6.00) 7.91 (8.50) 6.39 (6.00) 4.31 (3.00) 6.69 (7.00) 5.45 (5.00)
0–15 2–15 0–13 0–13 0–12 1–11
(A5) Vision and strategy (0–6) 3.26 (3.00) 3.71 (4.00) 3.27 (3.00) 2.53 (3.00) 3.61 (4.00) 3.06 (3.00)
0–6 1–6 0–6 0–6 1–6 1–6
(A6) Environmental pro?le (0–4) 1.36 (1.00) 2.08 (2.00) 1.19 (1.00) 1.08 (1.00) 1.38 (1.00) 1.12 (1.00)
0–4 0–4 0–3 0–3 0–3 0–3
(A7) Environmental initiatives (0–6) 1.64 (1.00) 2.13 (2.00) 1.90 (1.00) 0.69 (0.00) 1.77 (2.00) 1.26 (1.00)
0–6 0–6 0–4 0–4 0–4 0–5
Total (0–95) 19.13 (15.00) 25.21 (19.00) 18.59 (15.00) 10.08 (4.00) 21.00 (22.00) 18.45 (15.00)
1–68 3–68 1–50 1–55 2–43 1–42
Panel B: Environmental performance (N = 191 ?rms)
Overall Pulp and paper Chemicals Metals and mining Oil and gas Utilities
% Recycled 63.12% (81.67%) 68.41% (83.22%) 76.72% (93.92%) 70.47% (84.52%) 77.70% (87.46%) 25.32% (24.69%)
0–99.85% 1.80–99.01% 0.01–99.50% 0–99.85% 0–99.28% 0–68.54%
TRI/Sales 2.24 (0.62) 2.34 (1.49) 2.16 (0.58) 1.72 (0.27) 0.89 (0.13) 3.45 (2.07)
0.01–23.19 0.09–12.02 0.01–18.99 0.01–21.19 0.01–8.02 0.01–14.92
This table presents descriptive statistics on environmental disclosure scores and environmental performance measures by industry. Descriptive statistics present means
(medians) and ranges (min–max) below. Panel A presents disclosures scores for ?rms that chose to provide discretionary disclosures (i.e., ‘‘disclosing’’ ?rms) on their
environmental performance (N = 122 ?rms). The scale for each category of disclosure items is presented in brackets. Panel B presents environmental performance
measures for the full sample, i.e., ‘‘disclosing’’ ?rms and ‘‘silent’’ ?rms (N = 191 ?rms). TRI/Sales is toxics release inventory data (in pounds) divided by total sales (in
thousands). Percent recycled is toxic waste treated or recycled divided by total waste generated by ?rm.
P
.
M
.
C
l
a
r
k
s
o
n
e
t
a
l
.
/
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
)
3
0
3
–
3
2
7
3
1
9
performance proxies are all positive and signi?-
cant (recall, we reverse the sign of TRI/Sales to
facilitate the interpretation). This result is consis-
tent with the discretionary disclosure theories of
Verrecchia (1983) and Dye (1985), but inconsis-
tent with the negative association (H1b) pre-
dicted by socio-political theories. Firms with
better environmental performance have more vol-
untary disclosures about their environmental
impact. Furthermore, when we introduce both
scaled TRI and % recycled in the regression we
?nd that they are signi?cant and positive suggest-
ing that each provides incremental information
to the other with respect to voluntary environmen-
tal disclosures.
Table 4 indicates the predicted sign of associa-
tion for our control variables given prior general
disclosure literature and intuition. For the total
disclosures as well as hard versus soft disclosures,
the forward looking ?nancing proxy is positive
(consistent with the prior literature) but is statisti-
cally signi?cant only for the soft disclosure cate-
Table 3
Descriptive and correlation statistics for variables used in the estimation
Panel A: Descriptive statistics
Variable Mean Median Q1 Q3 Std dev
% Recycled 0.63 0.82 0.30 0.95 0.35
ÀTRI/Sales À2.24 À0.62 À2.50 À0.13 3.68
J–F coe?cient À0.08 0.00 0.00 0.37 0.81
SIZE 8.01 7.89 6.70 9.28 1.70
FIN À0.02 À0.01 À0.03 0.01 0.31
TOBIN Q 1.13 1.04 0.81 1.35 0.75
VOLAT 0.09 0.07 0.05 0.10 0.08
ROA 0.05 0.03 0.00 0.08 0.07
LEV 0.33 0.31 0.23 0.41 0.16
NEW 0.54 0.54 0.45 0.62 0.14
CAPIN 0.07 0.04 0.02 0.09 0.08
Panel B: Pearson correlation statistics
TRI/Sales Size FIN TOBIN Q VOLAT ROA LEV AGE CAPIN Total
disclosure
J–F
coe?cient
% Recycled 0.29
***
À0.06 0.15
**
0.28
***
0.07 0.14
**
0.02 À0.26
***
À0.30
***
0.23
***
À0.05
ÀTRI/Sales – À0.04 À0.01 0.11 À0.16
**
0.08 0.01 À0.03 À0.01 0.17
***
0.13
*
SIZE – 0.07 0.17
**
À0.09 0.01 0.01 0.33
***
0.21
***
0.46
***
0.05
FIN – À0.03 0.03 0.05 0.02 0.05 0.03 0.03 0.01
TOBIN Q – À0.01 0.47
***
À0.04 À0.15
**
À0.06 0.14 0.03
VOLAT – À0.11 0.03 À0.15
**
À0.06 À0.16
**
À0.15
**
ROA – À0.11 À0.10 À0.07 0.04 0.03
LEV – 0.14
*
0.01 0.19
***
0.10
NEW – 0.18
**
À0.12
**
À0.02
CAPIN 0.34
***
0.09
Total disclosure – 0.09
*
This table presents descriptive and correlation statistics for independent variables used in multivariate tests. Statistics are presented for
the full sample of 191 ?rms. Percent recycled is toxic waste treated or recycled divided by total waste generated by ?rm. TRI/Sales is the
negative of Toxics Release Inventory data (in pounds) divided by total sales in thousands. SIZE is the logarithm of market value. FIN
is the amount of debt or equity capital raised in ?scal year 2004 divided by total assets. TOBIN Q is the sum of market value of equity,
book value of preferred stock and book value of debt divided by total assets. VOLAT is stock price volatility (standard deviation of
monthly returns during 2003). ROA is return on assets. LEV is the leverage ratio. NEW is asset newness measured as the ratio of net
PPE to gross PPE. CAPIN is capital intensity measured as the ratio of capital spending to total sales. Total Disclosure is total
environmental disclosure score achieved using the disclosure index presented in Table 1. J–F coe?cient is the Janis–Fadner Coe?cient
(see Bansal and Clelland, 2004 for details). Spearman Correlation statistics provide similar results.
***
,
**
,
*
represent signi?cance levels
(two-tailed) at 1%, 5% and 10%, respectively.
320 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
gory. The information asymmetry proxies (Tobin’s
Q and stock volatility) are insigni?cant. Similarly,
?rm pro?tability (ROA) in the following year is
insigni?cant. One interpretation of this result is
that ?rms resort to other disclosure channels to
reduce the information asymmetry and convey
their good news about future ROA, and do not
use ER reports/web disclosures for this purpose.
Turning to our agency proxy, we ?nd that the
leverage variable is signi?cantly positive, suggest-
ing that debtholders exercise pressure on ?rms to
disclose environmental related matters to assess
potential future liabilities. Also, large ?rms dis-
close more, consistent with their lower information
production costs. Similarly, ?rms with greater cap-
ital expenditures disclose more: the coe?cient of
the capital intensity control variable is signi?cantly
positive. In contrast to our intuition, ?rms with
newer equipment (NEW) are less likely to provide
discretionary environmental disclosures. The
lagged Janis–Fadner coe?cient is not associated
with the level of disclosure, which is consistent
with the results of Aerts and Cormier (2006).
We also split our disclosure scores in two parts:
hard disclosures and soft disclosures. Hard disclo-
sures are closest in spirit to the assumed truthful
disclosures in the voluntary disclosure theories dis-
cussed in section ‘Literature review and hypothesis
Table 4
Inter-industry regressions with industry ?xed e?ects (Tobit analysis)
Dependent variables
Total disclosures Hard disclosures Soft disclosures
(1) (2) (3) (1) (2) (3) (1) (2) (3)
Intercept À81.2
***
À74.72
***
À78.67
***
À70.65
***
À64.00
***
À68.01
***
À20.65
***
À19.03
***
À19.85
***
(122.45) (103.76) (115.87) (115.51) (105.64) (112.24) (75.46) (63.87) (69.60)
% Recycled (±) 11.48
***
9.81
***
8.83
***
8.02
***
2.40
**
2.11
**
(10.31) (7.41) (7.62) (6.53) (5.68) (4.35)
ÀTRI/Sales (±) 0.84
***
0.65
**
0.77
***
0.72
***
0.21
**
0.18
*
(6.67) (4.07) (7.90) (6.99) (4.43) (3.20)
J–F coe?cient (À) 2.21 1.76 1.83 2.00 1.59 1.56 0.60 0.47 0.49
(2.02) (1.23) (1.40) (2.43) (1.52) (1.52) (1.33) (0.79) (0.90)
FIN (+) 18.82
*
21.20
*
18.05
*
13.17 14.41
*
11.73 9.42
***
9.94
***
9.13
***
(2.79) (3.47) (2.65) (2.02) (2.48) (1.70) (6.32) (6.99) (6.07)
TOBIN Q (+) À1.54 À1.88 À1.83 À2.18 À2.05 À2.45 À0.44 À0.48 À0.51
(0.86) (1.22) (1.24) (2.49) (2.24) (3.25) (0.64) (0.75) (0.85)
VOLAT (+) À4.69 4.05 3.56 À1.44 7.45 8.39 À1.93 À0.05 À0.08
(0.17) (0.11) (0.09) (0.02) (0.55) (0.72) (0.26) (0.01) (0.01)
ROA (+) À2.56 À1.31 0.85 9.49 11.32 13.73 À5.01 À5.57 À4.54
(0.02) (0.00) (0.01) (0.36) (0.51) (0.81) (0.62) (0.74) (0.52)
LEV (+) 27.27
***
26.58
***
27.63
***
23.80
***
23.01
***
22.67
***
7.29
***
7.13
***
7.16
***
(21.38) (19.77) (20.98) (21.76) (21.04) (20.89) (13.99) (13.20) (13.71)
SIZE (+) 9.70
***
10.06
***
9.67
***
7.90
***
8.05
***
7.81
***
2.61
***
2.68
***
2.57
***
(152.17) (160.43) (155.71) (140.36) (148.80) (145.05) (101.38) (106.75) (100.57)
NEW (+) À17.57
**
À17.63
**
À18.22
**
À12.16
*
À13.09
**
À12.73
**
À5.62
**
À5.64
**
À5.73
**
(5.00) (4.81) (5.40) (3.38) (3.96) (3.86) (4.55) (4.45) (4.76)
CAPIN (+) 33.38
***
26.96
**
30.32
***
28.22
***
21.26
**
24.96
***
9.67
**
8.04
**
8.82
**
(8.33) (5.26) (6.97) (8.62) (5.06) (7.06) (5.18) (3.54) (4.34)
Industry e?ects Yes Yes Yes Yes Yes Yes Yes Yes Yes
Log-likelihood À497.98 À499.55 À495.88 À435.27 À434.96 À431.62 À359.50 À360.01 À357.84
N = no of ?rms 191 191 191 191 191 191 191 191 191
Dependent variables are disclosures scores as indicated by the columns. The expected signs for the control variables are presented in
brackets. Coe?cients are estimated by maximum likelihood (Tobit regressions). The signi?cance levels are based on Chi-squared
statistics (presented in parentheses). All control variables are de?ned in Table 3.
***
,
**
,
*
represent signi?cance levels (two-tailed) at 1%,
5% and 10%, respectively.
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 321
development’. However, we ?nd similar results for
both hard and soft disclosure scores: the estimated
coe?cients for our two EP measures are positive
and signi?cant for both hard and soft disclosures
as predicted by H1a.
Intra-industry analysis
Following Lang and Lundholm (1993, 1996)
and Healy et al. (1999) we supplement the above
tests with an intra-industry approach which seeks
to explain the within-industry variation in the
ranks of disclosure scores using the within-indus-
try variation in the ranks of our independent
variables. There are two reasons justifying this
approach: (1) some of our variables (i.e., TRI/
Sales) can be compared within an industry but
cannot be meaningfully compared across indus-
tries due to di?erences in production processes;
(2) the distribution of our disclosure scores for
the Metal and Mining sector are quite low, so tak-
ing intra-industry ranks and then pooling facili-
tates the i.i.d assumption requirement of our
OLS regression model.
We rank the dependent and the independent
variables within industry and then we pool the cor-
responding percentiles across the ?ve sectors. The
regression results are presented in Table 5. They
are generally the same as in Table 4 and indicate
a positive association between the environmental
Table 5
Intra-industry rank regressions (OLS analysis)
Dependent variables
Total disclosures Hard disclosures Soft disclosures
(1) (2) (3) (1) (2) (3) (1) (2) (3)
Intercept 13.91
*
15.30
**
11.08 20.03
***
19.03
***
16.20
***
15.38
**
16.42
***
13.65
***
(1.84) (2.03) (1.48) (2.65) (2.53) (2.18) (1.96) (2.09) (3.41)
% Recycled (±) 0.17
***
0.15
***
0.15
***
0.14
***
0.11
**
0.11
**
(3.56) (3.19) (3.05) (2.98) (2.34) (2.24)
ÀTRI/Sales (±) 0.16
***
0.14
***
0.17
***
0.16
***
0.09
*
0.09
*
(3.32) (2.93) (3.48) (3.42) (1.93) (1.73)
J–F coe?cient (À) 0.07 0.05 0.05 0.05 0.04 0.03 0.18
***
0.17
***
0.17
***
(1.57) (1.07) (1.19) (1.09) (0.84) (0.66) (3.74) (3.51) (3.46)
FIN (+) 0.06 0.07 0.06 0.06 0.07 0.07 0.03 0.04 0.04
(1.24) (1.45) (1.38) (1.35) (1.50) (1.51) (0.69) (0.79) (0.75)
TOBIN Q (+) À0.04 À0.05 À0.06 À0.04 À0.06 À0.07 À0.02 À0.03 À0.04
(À0.67) (À0.89) (À1.07) (À0.72) (À0.98) (À1.21) (À0.36) (À0.50) (À0.62)
VOLAT (+) À0.06 À0.04 À0.05 À0.06 À0.03 À0.05 À0.04 À0.03 À0.04
(À1.15) (À0.83) (À0.99) (À1.15) (À0.54) (À0.97) (À0.84) (À0.60) (À0.76)
ROA (+) 0.06 0.06 0.06 0.06 0.08 0.06 0.03 0.02 0.02
(1.09) (1.01) (0.97) (1.09) (1.33) (0.94) (0.44) (0.37) (0.36)
LEV (+) 0.17
***
0.17
***
0.17
***
0.15
***
0.15
***
0.15
***
0.16
***
0.15
***
0.15
***
(3.51) (3.40) (3.44) (3.15) (3.02) (3.07) (3.12) (2.98) (3.04)
SIZE (+) 0.43
***
0.44
***
0.43
***
0.40
***
0.41
***
0.39
***
0.36
***
0.37
***
0.36
***
(8.39) (8.55) (8.53) (7.67) (8.03) (7.84) (6.77) (6.94) (6.81)
NEW (+) À0.11
**
À0.11
**
À0.11
**
À0.09
*
À0.09
**
À0.09
**
À0.08
*
À0.08
*
À0.08
*
(À2.34) (À2.26) (À2.42) (À1.92) (À1.99) (À2.03) (À1.65) (À1.64) (À1.68)
CAPIN (+) 0.11
**
0.10
**
0.10
**
0.10
**
0.08
*
0.09
*
0.13
**
0.13
**
0.12
**
(2.20) (2.13) (1.97) (2.07) (1.75) (1.79) (2.54) (2.54) (2.39)
Adj. R
2
40.83% 40.33% 43.22% 36.38% 37.32% 39.94% 34.98% 34.36% 35.69%
N = no of ?rms 191 191 191 191 191 191 191 191 191
Dependent variables are disclosures scores as indicated by the columns. The expected signs for the control variables are presented in
brackets. All variables are ranked within industry. Coe?cients are estimated by OLS regressions using the ranked variables. The
signi?cance levels are based on t statistics (presented in parentheses). All control variables are de?ned in Table 3.
***
,
**
,
*
represent
signi?cance levels (two-tailed) at 1%, 5% and 10%, respectively.
322 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
performance measures and the discretionary envi-
ronmental disclosures (total scores). Both (minus)
TRI/Sales and % recycled complement each other
and neither is redundant for the total sample or for
the hard and soft groups considered separately.
We ?nd similar results when we split the total
scores into hard and soft disclosure scores: the esti-
mated coe?cients for our two environmental per-
formance measures are positive and signi?cant as
predicted by H1a. The results for the control vari-
ables are also generally comparable to Table 4.
The Janis–Fadner coe?cient is now positively
associated with the level of soft disclosures, which
is opposite to the negative association predicted by
the socio-political theories.
A revised role for socio-political theories
Overall, the results in the previous sections
point to economic disclosure theory and not
socio-political theory as being robust in predicting
the level of discretionary environmental disclosure.
If the focus of enquiry is switched to disclosure
strategies, it may be that socio-political theories
explain interesting patterns in the data. While we
leave this possibility to future research, we o?er
in this section some preliminary evidence on the
question.
Legitimacy theory predicts that ?rms with
threatened legitimacy are likely to make self-serv-
ing disclosures referred to as ‘‘legitimization’’
(see, for example, Adams, 2004; Gray et al.,
1995, p. 54 & Hughes et al., 2001, p. 219). One
example of legitimization is for a poor environ-
mental performer to make soft claims to be com-
mitted to the environment which are not readily
veri?able. In fact, it is apparent from an inspection
of Table 1 that many poor EP ?rms do make soft
claims to be committed to the environment. Focus-
ing on A5, Vision and Strategy Claims, over 90%
of our poor EP ?rms make a claim which is
awarded at least one mark in category A5. For
example, 55 of 61 poor EP ?rms (i.e., 0.90
*
61)
make a claim that is awarded a score of 1 for item
A5-2 which is a statement of corporate environ-
mental policy and/or commitment. Further, 33 of
our 61 poor EP ?rms are awarded a score under
A5-1 which involves a CEO statement on environ-
mental performance to stakeholders.
22
More formally, using the ratio of soft disclosure
scores to total awarded scores as a proxy for legit-
imization, we would expect to ?nd a negative rela-
tion between prior perceived legitimacy and this
ratio. Our tests based on the conditional sample
of ?rms with scored sustainability/web disclosures
appear in Table 6. We measure threats to legiti-
macy in two ways: group EP membership based
on the median % recycled and group media cover-
age based on the median Janis–Fadner Coe?cient.
There are 122 (95) ?rms with scored sustainability/
web disclosures (at least one environmental press
article in 2002) and available EP. As indicated in
Table 6, the ratio of soft/total scores is 50.95%
for poor EP ?rms, compared to 34.23% for good
EP ?rms. Similarly, the ratio of soft/total scores
is 47.54% for unfavorable media coverage ?rms,
compared to 32.58% for ?rms with favorable
media coverage. In both instances, the di?erence
is statistically signi?cant at the 1% level. Both
results are consistent with ?rms whose environ-
mental legitimacy is threatened to make soft claims
to be committed to the environment. This result is
predicted by legitimacy theory but cannot be
explained by economic disclosure theories, which
assume truth-telling and thus o?er no predictions
about biased disclosures.
These results hold when we repeat the tests
(untabulated) in a multivariate fashion using all
other control variables in Tables 4 and 5 and
employ the ratio of soft/total as the dependent var-
iable. There is a signi?cant negative association
between the lagged Janis–Fadner Coe?cient and
the ratio of soft/total scores, after including all
control variables. Similar results hold for our
two EP measures. This result is robust to the Tobit
analysis approach (Table 4) and the rank regres-
sion analyses approach (Table 5) and implies a
greater propensity for ‘‘legitimization’’ behavior
22
As anecdotal evidence, consider the following soft environ-
mental commitment claim made by one of our sample chemical
?rms in the bottom quartile of our environmental performance
ranking: ‘‘Senior management leads the industry with respect to
responsible care.’’
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 323
for ?rms whose environmental legitimacy is
threatened.
Sensitivity analysis
For each disclosure item in the index presented
in Table 1, we have replicated the di?erence tests
across the two EP groups using Wilcoxon rank
tests. Non-parametric Wilcoxon rank tests are
robust to the possibility that the data does not fol-
low a normal distribution. In addition, we have
bootstrapped the p-values of the two-sample para-
metric t-tests by re-sampling the data with replace-
ment.
23
Both approaches result in the same levels
of signi?cance as the ones reported for the para-
metric t-tests.
We verify whether the di?erence tests in disclo-
sure scores reported in Table 1 are robust to the
classi?cation of sample ?rms as good and poor
environmental performers, using the % recycled
measure. We split the ?rms in the two groups
based on the level of the alternative environmental
performance measure, TRI/sales. We still ?nd sig-
ni?cant di?erences between the good and the poor
groups across all main categories in the index.
Finally, we run a Logit model to investigate
whether the decision to provide any discretionary
environmental disclosures is a?ected by the envi-
ronmental performance of the sample ?rms. We
perform an inter-industry analysis using the same
control variables as discussed in section ‘Econo-
metric model’. Results (unreported) are consistent
with the main results from Table 4. Environmen-
tal performance is found to be one of the main
drivers of the probability that the ?rm provides
environmental disclosures in discretionary chan-
nels. Precent recycled and (minus) TRI/Sales are
signi?cantly positive at the 1% level when intro-
duced separately. When both are in the model,
the signi?cance level decreases to 10%. The J–F
coe?cient is not signi?cant. All models show a
good ?t (signi?cant likelihood ratios and almost
90% concordant observations) suggesting a good
model speci?cation.
Table 6
Comparisons of soft to total disclosure scores
Panel A: Group membership is based on median % recycled (N = 122 ?rms)
Disclosure categories Average score Di?erence (t-stat)
Good EP ?rms (N = 61) Poor EP ?rms (N = 61)
Soft/total (%) 34.23% 50.95% À16.72%
***
(3.99)
Panel B: Group membership is based on median Janis–Fadner coe?cient (N = 95 ?rms)
*
Disclosure categories Average score Di?erence (t-stat)
Favorable media coverage (N = 48) Unfavourable media coverage (N = 47)
Soft/total (%) 32.58% 47.54% À14.96%
***
(3.22)
This table presents average ratios of soft to total disclosure scores. Panel A presents di?erences in soft to total disclosure ratios across
good environmental performance (EP) ?rms and poor environmental performance ?rms. Good environmental performance (EP) ?rms
are ?rms that have the environmental performance measure (% recycled) above the industry median. Panel B presents di?erences in
soft to total disclosure ratios across favorable media coverage ?rms and unfavorable media coverage ?rms. Firms with favorable media
coverage are ?rms that have a Janis–Fadner Coe?cient above the sample median (see Bansal and Clelland, 2004 for details). Only ?rms
that have at least one article in Factiva are included. The signi?cance levels presented in the last column are from two-sample t-
statistics that test the di?erence between the groups.
***
,
**
,
*
represent signi?cance levels (two-tailed) at 1%, 5% and 10%, respectively.
Wilcoxon Rank tests and t-tests with Bootstrap Resampling provided similar results. Groupings based on mean values also provide
similar results.
23
We implement the bootstrap procedure by drawing with
replacement 20000 samples from each EP group. Prior to
resampling, the procedure mean-centers the data within each
group (for details, see Westfall & Young, 1993).
324 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
Conclusions
Previous empirical evidence provides mixed
results on the relationship between corporate envi-
ronmental performance and the level of environ-
mental disclosures. We revisit this relation by
testing competing predictions from economics
based and socio-political theories of voluntary dis-
closure using a more rigorous research design. In
particular, we improve the prior literature in two
important ways. First, the predictions of voluntary
disclosure theory relate to discretionary, not man-
datory, environmental disclosures. This study
focuses on purely voluntary disclosure media such
as corporate Internet web sites and stand-alone
environmental reports. Previous studies assessed
environmental disclosures mainly from annual
reports and other regulatory ?lings such as 10 Ks
and many of those studies rely on a Wiseman
(1982) based content analysis index to measure
the extent of environmental disclosures.
Second, in collaboration with an environmental
reporting expert, we develop a content analysis
index to assess the level of environmental disclo-
sure in environmental and social responsibility
reports or similar disclosures in the ?rm’s web site.
The index, which follows closely the Global
Reporting Initiative (2002) sustainability reporting
guidelines, di?ers from the Wiseman index in that
we focus on ?rm disclosures related to its commit-
ment to protect the environment.
Our results are as follows. We ?nd a positive
association between environmental performance
and the level of discretionary disclosures in envi-
ronmental and social reports or related web disclo-
sures. In other words, superior environmental
performers are more forthcoming in truly discre-
tionary disclosure channels, as predicted by eco-
nomics based voluntary disclosure theories. Our
?ndings are robust to two reliable environmental
performance measures that use actual toxic emis-
sion and waste management data. The ?rst one is
based on Toxics Release Inventory scaled by sales
data at ?rm level (i.e., TRI normalized by ?rm’s
operational scale) and the second one is percentage
of total toxic wastes that were treated or processed
by each ?rm. In addition, our ?ndings are not
a?ected when we assess the relative environment
performance within each industry in order to con-
trol for industry di?erences in pollution propensity.
Our results are inconsistent with the prediction of a
negative association from socio-political theories,
suggesting that these theories are not robust in pre-
dicting the level of discretionary disclosure.
Although there is separation in scores and good
EP ?rms disclose more, the scores of good EP
?rms are low relative to the expectation implied
by the 2002 GRI reporting guidelines and point
to the need for improvement in the years beyond
2003. Speci?c areas where improvement is required
include obtaining independent assurance of sus-
tainability reports and the disclosure of environ-
mental performance indicators.
Finally, our results suggest important directions
for future research. Speci?cally, we provide preli-
minary evidence that socio-political theories are
robust in predicting what is being said. In particu-
lar, we ?nd that ?rms whose environmental legiti-
macy is threatened make soft claims to be
committed to the environment. This behavior is
predicted by legitimacy theory but cannot be
explained by economics disclosure theory. Thus,
we argue that future environmental disclosure
research should move the focus of enquiry beyond
the level of disclosure.
Acknowledgements
We are grateful for comments and suggestions
from the Editor (Anthony G. Hopwood), two
anonymous reviewers and from Kate Bewley, Walt
Blacconiere, Denis Cormier, Gus De Franco,
Kathy Herbohn, Ole-Kristian Hope, Hai Lu, Mi-
chel Magnan, Naomi Soderstrom and seminar
participants at Chinese University of Hong Kong,
University of Queensland and University of Tor-
onto. We acknowledge the ?nancial support of
the Canadian Academic Accounting Association,
Canadian Institute of Chartered Accountants,
and the AIC Institute for Corporate Citizenship
at Rotman School of Management, University of
Toronto. We also thank concurrent session partic-
ipants and discussants at the June 2006, July 2006
and August 2006 meetings of the CAAA,
AFAANZ (a Best Paper Award) and AAA,
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 325
respectively. We thank Rod Lohin, Bill Swirsky,
and Alan Willis for their support and encourage-
ment for this project and Gauri Bhat, Grace Jin
and Dushyantkumar Vyas for help in data collec-
tion. Gordon Richardson thanks KPMG for their
generous ?nancial support.
References
Adams, C. A. (2004). The ethical, social and environmental
reporting-performance portrayal gap. Accounting, Auditing
and Accountability Journal, 17(5), 731–757.
Aerts, W., & Cormier, D. (2006). The association between
media legitimacy and corporate environmental communi-
cation. Working paper.
Aerts, W., Cormier, D., & Magnan, M. (2006). The interface
between print- and web-based corporate environmental
disclosure, ?nancial markets and the media. Working Paper.
Al-Tuwaijri, S. A., 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(5-6), 447–471.
Bansal, P., & Clelland, I. (2004). Talking trash: Legitimacy,
impression management, and unsystematic risk in the
context of the natural environment. Academy of Manage-
ment Journal, 47(1), 93–102.
Barth, M., & Kasznik, R. (1999). Share Repurchases and
Intangible Assets. Journal of Accounting and Economics,
28(2), 211–241.
Barth, M., & McNichols, M. (1994). Estimation and market
valuation of environmental liabilities relating to superfund
sites. Journal of Accounting Research Supplement, 32,
177–209.
Barth, M., McNichols, M., & Wilson, P. (1997). Factors
in?uencing ?rms’ disclosures about environmental liabili-
ties. Review of Accounting Studies, 2, 35–64.
Beets, S. D., & Souther, C. (1999). Corporate environmental
reports: the need for standards and an environmental
assurance service. Accounting Horizons, 13(2), 129–145.
Bewley, K., & Li, Y. (2000). Disclosure of environmental
information by Canadian manufacturing companies: A
Voluntary Disclosure Perspective. Advances in Environmen-
tal Accounting and Management, 1, 201–226.
Blacconiere, W. G., & Patten, D. M. (1994). Environmental
disclosures, regulatory costs, and changes in ?rm value.
Journal of Accounting and Economics, 18, 357–377.
Chan-Fishel, Michelle. (2002). Survey of climate change
disclosure in SEC ?lings of automobile, insurance, oil &
gas, petrochemical, and utilities companies. Friends of the
Earth – US, September.
Clarkson, P., Li, Y., & Richardson, G. (2004). The market
valuation of environmental expenditures by pulp and paper
companies. The Accounting Review, 79, 329–353.
Cormier, D., & Magnan, M. (1997). Investors’ assessment of
implicit environmental liabilities: An empirical investiga-
tion. Journal of Accounting and Public Policy, 16, 215–241.
Cormier, D., Magnan, M., & Morard, B. (1993). The impact of
corporate pollution on market valuation: Some empirical
evidence. Ecological Economics, 8, 135–155.
Dye, R. A. (1985). Disclosure of non-proprietary information.
Journal of Accounting Research, 123–145, Spring.
Franco, N.C. (2001). Corporate environmental disclosure:
opportunities to harness market forces to improve corporate
environmental performance. American Bar Association
Conference on Environmental Law, Keystone, Colorado,
March 8–11.
Frankel, R., McNichols, M., & Wilson, P. (1995). Discretionary
disclosure and external ?nancing. The Accounting Review,
70, 135–150.
Freedman, M., & Wasley, C. (1990). ‘The association between
environmental performance and environmental disclosure in
annual reports and 10 Ks’. Advances in Public Interest
Accounting, 3, 183–193.
Global Reporting Initiative (2002). Sustainable Reporting
Guidelines, www.globalreporting.org.
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.
Healy, P. M., Hutton, A. P., & Palepu, K. G. (1999). A Stock
Performance and Intermediation Changes Surrounding
Sustained Increases in Disclosure. Contemporary Accounting
Research, 16, 485–520.
Healy, P. M., & Palepu, K. G. (2001). Information asymmetry,
corporate disclosure, and the capital markets: A review of
the empirical disclosure literature. Journal of Accounting and
Economics, 31(1–3), 405–440.
Hughes, S. B., Anderson, A., & Golden, S. (2001). Corporate
environmental disclosures: Are they useful in determining
environmental performance? Journal of Accounting and
Public Policy, 20, 217–240.
Hutton, A., Miller, G., & Skinner, D. (2003). The Role of
Supplementary Statements with Management Earnings
Forecasts. Journal of Accounting Research, 41, 867–890.
Ilinitch, A., Soderstrom, N., & Thomas, T. (1998). Measuring
corporate environmental performance. Journal of Account-
ing and Public Policy, 17, 387–408.
Ingram, R. W., & Frazier, K. (1980). Environmental Perfor-
mance and Corporate Disclosure. Journal of Accounting
Research, 18(2), 614–622. Autumn.
Janis, I. L., & Fadner, R. (1965). The coe?cient of imbalance.
In H. D. Lasswell & N. Leites, et al. (Eds.), Language of
Politics (pp. 153–169). Cambridge MA: MIT Press.
Jensen, M., & Meckling, W. (1976). Theory of the Firm:
Managerial Behavior, Agency Costs, and Ownership Struc-
ture. Journal of Financial Economics, 3(4), 305–360.
King, Andrew, & Lenox, Michael. (2001). Does it really pay to
be green? An empirical study of ?rm environmental and
?nancial performance. The Journal of Industrial Ecology,
5(1), 105–116.
326 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
Lang, Mark, & Lundholm, Russell. (1993). Cross-sectional
determinants of analyst ratings of corporate disclosures.
Journal of Accounting Research, 31(2), 246–247. Autumn.
Lang, M., & Lundholm, R. (1996). Corporate disclosure policy
and analyst behavior. The Accounting Review, 71(4),
467–492.
Leftwich, R. W., Watts, R. L., & Zimmerman, J. L. (1981).
Voluntary corporate disclosure: The case of interim report-
ing. Journal of Accounting Research, 18, 50–77.
Li, Y., & McConomy, B. (1999). An empirical examination of
factors a?ecting the timing of environmental accounting
standard adoption and the impact on corporate valuation.
Journal of Accounting, Auditing and Finance, 14(3), 279–313.
Li, Y., Richardson, G. D., & Thornton, D. (1997). Corporate
disclosure of environmental information; theory and evi-
dence. Contemporary Accounting Research, 14(3), 435–474.
Lim, T. (2001). Rationality and Analysts’ Forecast Bias.
Journal of Finance, LVI(1), 369–385.
Lindblom, C. (1994). The implications of organizational
legitimacy for corporate social performance disclosure.
Paper presented at the Critical Perspectives on Accounting
Conference, New York.
Patten, Dennis (1992). Intra-industry Environmental Disclo-
sures in Response to the Alaskan Oil Spill: A Note on
Legitimacy Theory. Accounting, Organizations, and Society,
17, 471–475.
Patten, Dennis (2000). Changing Superfund disclosure and its
relation to the provision of other environmental informa-
tion. Advances in Environmental Accounting and Manage-
ment, 1, 101–121.
Patten, Dennis (2002). The relation between environmental
performance and environmental disclosure: A research note.
Accounting, Organizations, and Society, 27, 763–773.
Richardson, A., & Welker, M. (2001). Social disclosure,
?nancial disclosure and the cost of equity capital. Account-
ing, Organizations and Society, 26(7/8), 597–616.
Smith, C., & Watts, R. (1992). The Investment Opportunity Set
and Corporate Financing, Dividend and Compensation
Policies. Journal of Financial Economics, 32(1), 263–293.
Standard & Poor’s (2004). Risk and Opportunity: the best
practice in non-?nancial reporting, ?rst edition, ISBN 1-
903168-12-0.
Verbeek, M. (2004). A Guide to Modern Econometrics. John
Wiley and Sons.
Verrecchia, R. (1983). Discretionary disclosure. Journal of
Accounting and Economics, 5, 179–194.
Westfall, P., & Young, S. (1993). Resampling-Based Multiple
Testing: Examples and Methods for P-Value Adjustment.
John Wiley and Sons.
Wiseman, J. (1982). An evaluation of environmental disclosures
made in corporate annual reports. Accounting, Organiza-
tions and Society, 7(1), 553–563.
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 327
doc_617911647.pdf