2008 - PeterM - Clarkson - Revisiting The Relation Between Environmental Performance and Environmental Disclosure An Empirical Analysis

Available online at www.sciencedirect.
com
Accounting, Organizations and Society 33 (2008) 303–327

www.elsevier.com/locate/aos
Revisiting the relation between environmental

performance and environmental disclosure:
An empirical analysis
a,b
Peter M. Clarkson , 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 firm disclosures related to its commitment to protect the environment than the indices employed by prior studies.
Using a sample of 191 firms from the five most polluting industries in the US, we find 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.
2007 Elsevier Ltd. All rights reserved.
Introduction
*
An unresolved research issue in environmental
Corresponding author. Tel.: +1 416 946 8601. accounting is the empirical association between
E-mail addresses: [email protected] (P.M.
Clarkson), [email protected] (Y. Li), gordon.richard- the level (i.e., amount) of corporate environmental
[email protected] (G.D. Richardson), fvasvari@london. disclosures and corporate environmental perfor-
edu (F.P. Vasvari). mance (Al-Tuwaijri, Christensen, & Hughes,
0361-3682/$ - see front matter 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aos.2007.05.003
304 P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327
2004; Hughes, Anderson, & Golden, 2001; Patten, poor environmental performers face more political
2002). Accounting standard setters and securities and social pressures and threatened legitimacy,
regulators are increasingly being made aware of they will attempt to increase discretionary environ-
deficiencies in corporate environmental disclosures mental disclosures to change stakeholder percep-
(Beets & Souther, 1999; Chan-Fishel, 2002; tions about their actual performance. Thus, we
Franco, 2001). The results of previous studies on have competing directional predictions from alter-
the relation between corporate environmental per- native theories, and the observed direction of asso-
formance and environmental disclosure in finan- ciation between environmental performance and
cial reports have been mixed. Patten (2002) the level of discretionary disclosures will eliminate
attributes the failure to find a significant and con- one of the two predictions.
sistent relation between environmental perfor- The predictions of the above theories relate to
mance and environmental disclosure to problems discretionary, not mandatory, environmental dis-
in the research designs of existing research. These closures. Previous studies assessed environmental
problems include failure to control for other fac- disclosures mainly from annual reports and other
tors associated with the level of environmental regulatory filings such as 10 Ks and many of those
disclosure, inadequate sample selection, and inad- studies rely on a Wiseman (1982) based content
equate measures of environmental performance analysis index to measure the extent of environ-
and disclosure. mental disclosures. The Wiseman index focuses
This study seeks to revisit the relation between on the financial consequences of corporate envi-
environmental performance and the level of environmental activities and puts more weight on
ronmental disclosure using a more rigorous quantitative disclosures. Using this measure, poor
research design. We test two competing predic- environmental performers may actually have
tions about the level of voluntary environmental higher disclosure scores than good performers
disclosures. Voluntary disclosure theory (Dye, because they have greater exposures and must dis-
1985; Verrecchia, 1983) predicts a positive associa- cuss any material financial information in their
tion between environmental performance and the regulatory filings such as annual reports and
level of discretionary environmental disclosure. 10 Ks. This may partially explain the inconclusive
The notion is that superior environmental per- findings in the previous literature and why Patten
formers will convey their ‘‘type’’ by pointing to (2002) finds a negative relation between environ-
objective environmental performance indicators mental disclosure and a toxics release inventory
which are difficult to mimic by inferior type firms. (TRI) based environmental performance
Inferior performers will choose to disclose less or indicator.1
to be ‘‘silent’’ on their environmental performance, In collaboration with an environmental disclo-
thus being placed in a pool of firms where inves- sure expert, we develop a content analysis index
tors and other users ascribe the ‘‘average type’’ based on the global reporting initiative guidelines
to that pool. What sustains this partial disclosure (GRI) to assess the level of discretionary environ-
equilibrium is proprietary costs associated with mental disclosures in environmental and social
disclosure about environmental performance (Ver- responsibility reports or similar disclosures pro-
recchia, 1983) and uncertainty as to whether the vided on the firm’s web site. This index differs from
firm is informed regarding its type (Dye, 1985). Wiseman (1982) index, previously used in the liter-
Socio-political theories including political econ- ature, because we focus on firm disclosures related
omy, legitimacy theory, and stakeholder theory to its commitment to protect the environment. Our
(Patten, 2002), on the other hand, predict a negative index potentially allows investors, regulators, and
association between environmental performance
and the level of discretionary environmental disclo-
sures. These overlapping theories suggest that 1
Patten (2002) was aware of the problem of non-discretion-
social disclosure is a function of social and political ary disclosures in annual reports and dropped litigation
pressures facing the corporation. To the extent that disclosures as a partial attempt to deal with this (see p. 768).
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327 305
environmental stakeholder groups to infer envi- which moves the focus of enquiry beyond the sim-
ronmental performance ‘‘type’’ from the disclosure ple level of discretionary disclosure. Using the
score. This is valuable to users who seek to assess ratio of soft disclosure scores to total awarded
the firms’ true environmental commitment and scores as a proxy for ‘‘legitimization’’, we show
related environmental exposures. that firms with unfavorable prior year media cov-
We focus on the 2003 environmental disclosures erage are more likely to make soft claims to be
of 191 firms drawn from five industries: Pulp and committed to the environment which are not read-
Paper, Chemicals, Oil and Gas, Metals and Min- ily verifiable. This behavior is not predicted by
ing, and Utilities. These five industries are consid- economic disclosure theories, which assume
ered to have a high pollution propensity and have truth-telling. Thus, socio-political theories do
collectively been the subject of a whole range of indeed explain additional patterns in the data.
environmental regulations in the US in the past The paper is organized as follows. Following lit-
30 or more years. The magnitude of the environ- erature review and hypothesis development, we
mental spending by these industries to comply describes our content analysis disclosure index
with the environmental regulations and the impact and the measures of environmental performance
of their operations on the natural environment which we use in the study. We then present our
should be a major concern to investors and other econometric model and preliminary empirical evi-
environmental stakeholder groups. Thus, firms in dence. The sections ‘Empirical results involving
these industries collectively form an ideal sample the level of disclosure’ and ‘A revised role for
to test the competing predictions of voluntary dis- socio-political theories’ contain the main results
closure and socio-political theories. followed by sensitivity analysis. The final section
In brief, our results are as follows. We find a summarizes the main findings of the study with a
positive association between environmental per- discussion of implications for future research.
formance and the level of discretionary disclosures
in environmental and social reports or related web
disclosures. In other words, superior environmen- Literature review and hypothesis development
tal performers are more forthcoming in truly dis-
cretionary disclosure channels, as predicted by Literature review
the economics based voluntary disclosure theory.
Our result is inconsistent with the prediction of a The existing literature in environmental
negative association arising from socio-political accounting research can be categorized into three
theories such as legitimacy theory and stakeholder broad groups. The first group of studies examines
theory. Further, using the Janis–Fadner coefficient the valuation relevance of corporate environmen-
of imbalance as a direct measure of perceived legit- tal performance information and has found that
imacy, we fail to observe the negative association such information is valuable to investors seeking
between legitimacy and the level of disclosures to assess environmental liabilities in different set-
implied by socio-political theories.2 Thus, our tings.3 The second line of literature examines fac-
results suggest that socio-political theories are tors affecting managerial decisions to disclose
not robust in predicting the level of discretionary potential environmental liabilities. This group of
environmental disclosures. studies finds that there are strategic factors affect-
We do, however, find that socio-political theo- ing firms’ decisions to disclose environmental
ries are helpful in predicting what is being said, liability information, especially when disclosures
2
The Janis–Fadner coefficient of imbalance measures the
3
propensity of a firm’s prior year press articles pertaining to the See Cormier, Magnan, and Morard (1993), Blacconiere
environment to be unfavorable. See section ‘Other control and Patten (1994), Barth and McNichols (1994), Cormier and
variables specific to environmental disclosures’ for a detailed Magnan (1997), Li and McConomy (1999), Richardson and
definition. Welker (2001), & Clarkson, Li, and Richardson (2004).
are discretionary.4 A third line of studies, one general terms, 0 for no disclosure).5 The CEP
which we discuss in the following paragraphs since rankings were used as a proxy for environmental
it is most relevant to this study, explores the rela- performance. Spearman rank order correlation
tion between environmental disclosures and envi- indicates that there is no significant association
ronmental performance. between the CEP environmental performance
Ingram and Frazier (1980) examine the associa- rankings and the Wiseman environmental disclo-
tion between the content of corporate environmen- sure index rankings.
tal disclosure and corporate environmental Freedman and Wasley (1990) examine the rela-
performance. The study was concerned with a lack tionship between corporate pollution performance
of corporate social responsibility disclosures in and pollution disclosures made in annual reports
annual reports due to their voluntary nature. The and 10 K reports filed with the SEC. Their sample
authors scored environmental disclosures in 20 consists of 50 US companies in four industries
pre-selected content categories along four dimen- (Steel, Oil, Pulp and Paper, Electric Utilities).
sions; evidence, time, specificity, and theme. Again, the CEP rankings are used as a proxy for
Ingram and Frazier (1980) proxied environmental environmental performance. The authors measure
performance by a performance index devised by environmental disclosures in both annual and
the Council on Economic Priorities (CEP), a 10 K reports using the same indexing procedure
non-profit organization specializing in the analy- developed by Wiseman (1982). Spearman rank
sis of corporate social activities. Forty firms order correlation tests are conducted to examine
were selected from the 50 firms that were moni- the associations both between annual report dis-
tored by the CEP. Regression results indicated closure indices and the CEP indices, and between
no association between environmental disclosure 10 K disclosure indices and the CEP indices. The
and environmental performance, consistent with results indicate that neither annual report environ-
authors’ prior expectation about an overall poor mental disclosures nor the 10 K environmental
quality of environmental disclosures in annual disclosures are indicative of firms’ actual environ-
reports. mental performance.
Wiseman (1982) examines the extent of volun- Bewley and Li (2000) examine factors associ-
tary environmental disclosures made by corpora- ated with the environmental disclosures in Canada
tions in their annual reports using a research from a voluntary disclosure theory perspective.
design almost identical to Ingram and Frazier The authors measure environmental disclosures
(1980). The study focuses on the 26 largest US by 188 Canadian manufacturing firms in their
companies that were monitored by the CEP for 1993 annual reports using the Wiseman index. A
the 1972–1976 period. Wiseman designed an envi- firm’s pollution propensity (i.e., environmental
ronmental disclosure index covering 18 items in performance) is proxied by their industry member-
four categories: economic factors (5 items), envi- ship and by whether they report to the Ministry of
ronmental litigation (2 items), pollution abatement Environment under the National Pollution
activities (5 items), and environmental disclosures Release Inventory program. The study finds that
that do not fall into the other three (6 items). In firms with more news media coverage of their envi-
addition, Wiseman assigned a score to each item ronmental exposure, higher pollution propensity,
based on whether the disclosure is quantitative or and more political exposure are more likely to dis-
qualitative (3 for quantitative disclosure, 2 for
non-quantitative disclosure, 1 for mentioning in
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 financial conse-
4
See Patten (1992), Li, Richardson, and Thornton (1997), quences of corporate environmental activities, most of which
Barth, McNichols, and Wilson (1997), Li and McConomy are required disclosures in 10 Ks for public companies regis-
(1999), & Aerts, Cormier, and Magnan (2006). tered with the SEC.
close general environmental information, suggest- performance. Specifically, they assess environmen-
ing a negative association between environmental tal performance as the percentage of total waste
disclosures and environmental performance. generated that is recycled. The authors measure
Hughes et al. (2001) examine environmental dis- environmental disclosure using a content analysis
closures made by 51 US manufacturing firms for in four categories (potential responsible parties’
1992 and 1993. Again, the authors use a slightly designation, toxic waste, oil and chemical spills,
modified Wiseman index to measure environmen- and environmental fines and penalties). These dis-
tal disclosures made within the President’s letter, closures are largely non-discretionary, in contrast
MD&A, and note section, and then assess whether to the discretionary disclosures which we examine.
environmental disclosures are consistent with envi- They find a positive association between environ-
ronmental performance ratings (good, mixed, and mental performance and environmental disclosure.
poor) by the CEP. Although the study finds no dif- In summary, the existing studies find mixed
ference in environmental disclosures between good results on the relation between environmental dis-
and mixed groups, firms rated as poor environ- closure and environmental performance. One rea-
mental performers by the CEP tend to make sub- son for the inconclusive findings is due to the
stantially more environmental disclosures under choice of non-discretionary disclosure channels
the Wiseman disclosure index. The authors attri- and use of the Wiseman (1982) index. It is self evi-
bute this finding to increased scrutiny in 1992 dent that, as environmental problems and expo-
and 1993 by the FASB and SEC with respect to sures increase, non-discretionary disclosures in
environmental disclosures, which forces poor per- regulated channels such as annual reports and
formers to make more disclosure as they are sub- 10 Ks should increase. Thus, a negative associa-
ject to more remediation activities. tion between environmental performance and con-
Patten (2002) identifies three issues in the previ- tent analysis scores in annual reports and 10 Ks
ous studies in this area; (1) failure to control for may be driven by non-discretionary disclosures.
other factors, (2) inadequate sample selection; The disclosure theories reviewed in the next sec-
and (3) inadequate measures of environmental per- tion, on the other hand, apply to discretionary dis-
formance. Since the CEP only followed a small closures. Thus, inferences about the robustness of
group of firms in only four industries, reliance on these theories are confounded when the disclosure
the CEP for sample selection may be problematic. media are formal channels like the annual report
In addition, the CEP did not use the same criteria and the 10 K. Adding to this problem is the heavy
and consistent methodology to assess corporate weight the Wiseman index places on disclosures
environmental performance in different industries. about the financial consequences of environmental
To overcome this issue, Patten uses TRI data, nor- activities, whereas our index places more weight on
malized by sales, to proxy for environmental per- disclosures that reveal true (but unobservable)
formance. Using a sample of 131 US firms from environmental performance. By focusing exclu-
24 different industries, and a modified Wiseman sively on environmental and social responsibility
index measure and line count of environmental reports or similar disclosures on firms’ web sites,
disclosures in 1990 annual reports, Patten finds locations where disclosures are purely discretion-
that TRI/sales are positively associated with both ary, and with an index that aims at revealing per-
measures of environmental disclosures, suggesting formance ‘‘type’’, we enhance the reliability of
a negative relation between environmental perfor- inferences about the true direction of association
mance and environmental disclosures. between environmental performance and discre-
More recently, Al-Tuwaijri et al. (2004) explore tionary disclosure.
the relations among environmental disclosure,
environmental performance and economic perfor- Hypothesis development
mance using a simultaneous equations approach.
Similar to Patten (2002) and Al-Tuwaijri et al. The voluntary disclosure literature suggests that
(2004) use TRI based data to assess environmental companies have incentives to disclose ‘‘good
news’’ to differentiate themselves from companies association between corporate environmental per-
with ‘‘bad news’’ in order to avoid the adverse formance and level of discretional environmental
selection problem (Dye, 1985; Verrecchia, 1983). disclosure. Thus, the two competing theories pro-
While these theories do not pertain to environmen- vide opposite predictions on how environmental
tal performance per se, they are applied to this set- performance may affect discretionary environmen-
ting by Bewley and Li (2000) and Li et al. (1997). tal disclosure strategies. Our hypotheses now fol-
Companies with superior environmental perfor- low (stated in the alternate form):
mance due to their proactive environmental strat-
egy have incentives to inform investors and other H1a: Environmental performance and the level of
stakeholders of their strategy by voluntarily dis- discretionary environmental disclosures are
closing more environmental information. Put sim- positively associated, as implied by econom-
ply, they seek to reveal their performance type, ics based voluntary disclosure theories.
something not directly observable to investors H1b: Environmental performance and the level of
and other stakeholders, through direct voluntary discretionary environmental disclosures are
disclosures that cannot be easily mimicked by poor negatively associated, as implied by the
environmental performers. In doing so, they socio-political theories.
potentially increase firm valuation since knowl-
edgeable investors will infer that exposures and
latent environmental liabilities are lower for good Research design
as opposed to poor environmental performers.6
Thus, voluntary disclosure theory predicts a Environmental performance indicators
positive association between environmental per-
formance and the level of discretional environmen- A key research design issue in this study is to
tal disclosure. develop a reliable proxy for a firm’s environmental
Turning to the predictions of socio-political the- performance. The difficulty in assessing environ-
ories, Gray, Kouhy, and Lavers (1995) and Lindb- mental performance is well documented in the lit-
lom (1994) argue that companies whose social erature (see, for example, Ilinitch et al., 1998).
legitimacy is threatened have incentives to increase Since we seek to assess relative environmental per-
environmental disclosures to: (1) educate and formance in this study, we follow the existing liter-
inform relevant publics about (actual) changes in ature and develop our relative environmental
their performance, (2) change perceptions about performance proxy using the actual pollution dis-
their performance, (3) deflect attention from the charge data from the US Environmental Protec-
issue of concern by highlighting other accomplish- tion Agency’s (EPA) TRI database (King &
ments, and (4) seek to change public expectations Lenox, 2001). Specifically, we first aggregate the
of their performance.7 According to Patten (2000, total toxic releases (in pounds) and the toxic waste
2002), socio-political theories predict a negative treated or processed for each of our sample firms
in 2003, as reported by the EPA in 2005 (EPA
annually reports the data at the plant level, with
6
The assumption that environmental performance ‘‘type’’ is a two year lag). To verify the accuracy of our
not readily observable to investors and other stakeholders is aggregation procedure, we compared the TRI
supported by the difficulties environmental researchers have
had obtaining reliable measures of environmental performance
measures obtained at the firm level with those pro-
that are comparable across companies in the same industry and vided by the Investor Responsibility Research
across industries (see Al-Tuwaijri et al., 2004 & Ilinitch, Center in their proprietary database. We found
Soderstrom, & Thomas, 1998). that the differences were minor.
7
Socio-political theories include political economy, legiti- Our first measure is the total toxic waste that is
macy theory and stakeholder theory (see Patten, 2002). We do
not differentiate them in this study as they have the same
treated, recycled or processed as a percentage of
prediction with respect to the relation between environmental the total toxic waste generated by each firm (%
performance and environmental disclosure. recycled). This measure is similar to the one used
by Al-Tuwaijri et al. (2004). We supplement our emphasize discretionary disclosures that are hard
analysis with an alternative measure, the ratio of to mimic. We assume these hard disclosures are
TRI to total firm sales. This gives us pounds of truthful, in that a firm would face litigation expo-
toxic releases per thousand dollars of sales for each sure if caught lying by informed stakeholders in
of our sample firms (TRI/sales). However, if firms social responsibility reports or web related
are not homogenous in terms of production pro- disclosures.9
cesses within one of our five industries (Pulp and What is striking in all this is that there is a
Paper, Chemicals, Oil and Gas, Metals and Min- demand by environmental stakeholders for pre-
ing, and Utilities), such that TRI/Sales is not cisely the same thing: hard, objective measures of
directly comparable across firms in that industry, environmental performance in social responsibility
the recycling measure reduces this noise and facil- reports, so that poor EP performers cannot mimic
itates pooling. good EP performers by soft, unverifiable claims to
Since industry pollution propensity differs sig- be committed to the environment. Indeed, good
nificantly, we alternatively rank the above two EP performers and environmental stakeholders
measures within each industry. Thus, the %Recy- have joined forces to develop standards for firms
cled ranks and TRI/sales ranks are our proxies preparing social responsibility reports that put a
for a firm’s relative performance within its indus- premium on hard, objective measures.
try. Similar relative performance measures are also The Global Reporting Initiative (GRI) was
used in Clarkson et al. (2004). launched in 1997 as a joint initiative of Coalition
for Environmentally Responsible Economies, a
Environmental disclosure index US non-government organization and the United
Nations Environmental Program. The overall goal
As discussed in section ‘Literature review and of the initiative is to develop a globally accepted
hypothesis development’, from the economic the- reporting framework to enhance the quality, rigor,
ory perspective, superior environmental perfor- and utility of sustainability reporting (Global
mance (EP) types are hypothesized to seek Reporting Initiative, 2002). The GRI Guidelines
credible direct disclosures to reveal their (unob- follow 11 principles (transparency, inclusiveness,
servable) performance type. A crucial property of auditability, completeness, relevance, sustainabil-
these disclosures is that they focus on objective, ity context, accuracy, neutrality, comparability,
‘‘hard’’ measures that cannot be easily mimicked clarity, and timeliness) to ensure that sustainability
by poor environmental performers. Thus, reliable reports (1) present a reasonable and balanced
inferences about theories like Verrecchia (1983) account of economic, environmental, and social
and Dye (1985) require a content analysis disclo- performance, (2) facilitate comparison over time
sure index that puts a heavy emphasis on objective and across organizations, and (3) credibly address
measures of performance as opposed to soft (i.e., issues of concerns to stakeholders. The first set of
not easily verifiable) claims to be committed to GRI Guidelines was published in 1999 as an Expo-
the environment.8 As a simple example, consider sure Draft and several revisions have followed
a good and a poor EP type firm in the same indus- since then. For the purpose of this study, we rely
try. The good EP firm will voluntarily disclose on the GRI Sustainability Reporting Guidelines
objective measures of environmental impact (e.g., published in 2002.
quantitative environmental performance indica-
tors) and will benchmark its performance relative
to the industry, something the poor EP firm will
9
not want to do. Thus, the good EP firm will 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
8
For a related discussion on the need for objective and claiming that, in 25 years, the company had been convicted of
verifiable disclosures to achieve the separation predicted by only 15 environmental offenses. Greenpeace identified 26
Verrecchia (1983) see Hutton, Miller, and Skinner (2003). convictions in the last four years.
We engaged an expert in the field of environ- We now turn to discuss our disclosure index in
mental reporting to help us develop a content anal- more detail. Our disclosure index consists of seven
ysis index suitable for firms’ sustainability reports broad categories, A1–A7, of environmental disclo-
or the corresponding sections of a broad social sures (see Table 1). We consider A1–A4 and A5–
responsibility report or equivalent discussions on A7 to represent ‘‘hard’’ and ‘‘soft’’ environmental
the web.10 From the outset, we agreed that the disclosures, respectively.
construct we seek to measure is the extent of a To score environmental disclosures in discre-
firm’s disclosure in their sustainability report. tionary channels and web related disclosures, we
With this aim established, the expert convinced accessed the internet web site of each sample firm
us that the GRI reporting guidelines are consistent and identified its environmental report, if any,
with that purpose. Thus, the expert helped us and any web based environmental disclosures.
develop a scoring model containing 95 line items We saved all such disclosures as of September
that reflect the spirit of the GRI guidelines. Table 2004. This arbitrary choice of timing worked out
1 contains the scoring model with a reference to well for us as all firms discussed their fiscal 2003
the corresponding section in the GRI guidelines. environmental performance in the environmental
Our disclosure index follows closely the report- reports and related web based disclosures which
ing requirements of the GRI guidelines for the fol- we obtained.
lowing reasons. Firms do not have to prepare
social responsibility reports or related web based Hard disclosure items
disclosures discussing their environmental impacts, Category A1 focuses on disclosures pertaining
and if they voluntarily do so, they do not have to to a firm’s governance structure and management
adopt the GRI guidelines. The voluntary decision systems put in place with respect to environmental
by a firm to both prepare a social responsibility protection. For instance, firms whose Board of
report and use the GRI guidelines means that the Directors have an environmental committee or
firm has opted for a format (the GRI format) that, have implemented ISO 14001 will inform their
by the intent of the GRI guidelines, will result in stakeholders of such commitments. A2 focuses
hard disclosures not easily mimicked by the poor on the credibility of a firm’s disclosures in its envi-
EP types. Thus, a firm making a sincere attempt ronmental report. Firms that obtained indepen-
to use the GRI guidelines will score high using dent verification of their environmental reports,
our content analysis index, which is precisely the and firms with their products and environmental
result we seek (i.e., the poor EP types will not want programs certified by independent agencies and
to conform to GRI guidelines that place a pre- third parties will receive higher scores in this cate-
mium, for example, on objective environmental gory. In A3, we assess the extent to which firms
performance indicators). Of the 95 equally disclose specific environmental performance indi-
weighted items in our disclosure index, 79 relate cators, both about their actual pollution emissions
to ‘‘hard’’ disclosure measures compared to only and their conservation and recycling efforts. These
16 for ‘‘soft’’ disclosure items, a proportion which are the ‘‘hard’’ data that firms can disclose to con-
(according to the expert who helped us developing vince stakeholders about their environmental com-
the index) reflects the spirit of the GRI guidelines. mitments. In addition, we also award scores when
firms disclose performance indicators with respect
to historical trends, the firms’ own emission reduc-
tion targets, and the industry average. Disclosing
10
Alan Willis, CA, Project Director – Performance Reporting actual performance indicators in the above context
Initiatives, The Canadian Institute of Chartered Accountants. can convey critical information for stakeholders to
He was a member of the GRI Steering Committee since its assess the firm’s long-term environmental perfor-
inception, and has been a member of the GRI Guidelines
development and revision working groups from 1998 to date.
mance (and commitments).
He is also a judge for the Canadian Institute for Chartered The final category in the ‘‘hard’’ disclosure
Accountants’ Corporate Reporting Awards. group is A4, which reflects a firm’s environmental
Table 1
Index assessing the quality of discretionary disclosures about environmental policies, performance and inputs
Hard disclosure items Map to Percentage of Average score
GRI firms attaining
Good EP Poor EP
the item (%) firms firms
(N = 61) (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 3.1 37.70 0.44 0.31
positions for env. management (0–1)
2. Existence of an environmental and/or a public issues committee in 3.1 31.15 0.38 0.25
the board (0–1)
3. Existence of terms and conditions applicable to suppliers and/or 3.16 21.31 0.31 0.11***
customers regarding env. practices (0–1)
4. Stakeholder involvement in setting corporate environmental policies 1.1, 3.10 27.05 0.36 0.18**
(0–1)
5. Implementation of ISO14001 at the plant and/or firm level (0–1) 3.14, 3.20 45.90 0.51 0.41
6. Executive compensation is linked to environmental performance 3.5 4.92 0.08 0.02*
(0–1)
(A2) Credibility (max score is 10) 83.61 2.88 1.95**
1. Adoption of GRI sustainability reporting guidelines or provision of 3.14 12.30 0.21 0.03**
a CERES report (0–1)
2. Independent verification/assurance about environmental 2.20, 2.21 1.64 0.03 0.00
information disclosed in the EP report/web (0–1)
3. Periodic independent verifications/audits on environmental 3.19 17.21 0.18 0.16
performance and/or systems (0–1)
4. Certification of environmental programs by independent agencies 3.20 15.57 0.23 0.08**
(0–1)
5. Product Certification 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 51.64 0.52 0.51
sustainability index (0–1)
7. Stakeholder involvement in the environmental disclosure process 1.1, 3.10 4.92 0.08 0.02*
(0–1)
8. Participation in voluntary environmental initiatives endorsed by 3.15 33.61 0.31 0.36
EPA or Department of Energy (0–1)
9. Participation in industry specific associations/initiatives to improve 3.15 54.10 0.71 0.38***
environmental practices (0–1)
10. Participation in other environmental organizations/assoc. to 3.15 40.98 0.52 0.30***
improve. environmental practices (if not awarded under 8 or 9
above) (0–1)
(A3) Environmental performance indicators (EPI) (max score is 60)a 73.77 10.19 6.00***
1. EPI on energy use and/or energy efficiency (0–6) EN3, 4, 17 41.80 1.46 0.75***
2. EPI on water use and/or water use efficiency (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, EN11 50.00 1.44 1.04
reducing, treatment and disposal) (0–6)
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 EN16 25.41 0.64 0.48
incidents) (0–6)
(continued on next page)
Table 1 (continued)
Hard disclosure items Map to Percentage of Average score
Good EP Poor EP
the item (%)
firms firms
(N = 61) (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 23.77 0.30 0.18*
the company (0-1)
2. Amount spent on technologies, R& D and/or innovations to enhance EN35 20.49 0.21 0.19
environ. perf. and/or efficiency (0–1)
3. Amount spent on fines related to environmental issues (0–1) EN16 25.41 0.33 0.18**
Soft disclosure items Map to Percentage of Average score

Good EP Poor EP
the item (%)
firms firms
(N = 61) (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 1.1, 1.2 61.48 0.69 0.54*
shareholders and/or stakeholders (0–1)
2. A statement of corporate environmental policy, values and 1.1, 1.2, 87.70 0.85 0.90
principles, environ. codes of conduct (0–1) 3.7
3. A statement about formal management systems regarding 3.19 58.20 0.57 0.59
environmental risk and performance (0–1)
4. A statement that the firm undertakes periodic reviews and evaluations 3.19 37.70 0.47 0.27**
of its environ. performance (0–1)
5. A statement of measurable goals in terms of future env. performance 1.1, 1.2 27.05 0.31 0.22
(if not awarded under A3) (0–1)
6. A statement about specific environmental innovations and/or new 1.1, 1.2 54.92 0.58 0.52
technologies (0–1)
(A6) Environmental profile (max score is 4) 70.49 1.49 1.23
1. A statement about the firm’s compliance (or lack thereof) with GN 8 32.79 0.38 0.28
specific environmental standards (0–1)
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 GN 8 56.56 0.61 0.52
services impact the environment. (0–1)
4. An overview of corporate environmental performance relative to GN 8 24.59 0.26 0.23
industry peers (0–1)
(A7) Environmental initiatives (max score is 6) 72.95 1.93 1.34**
1. A substantive description of employee training in environmental 3.19 30.33 0.39 0.21**
management and operations (0–1)
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 certification of environmental programs (0–1) 3.19 9.84 0.15 0.05*
6. Community involvement and/or donations related to environ. (if not SO1, 53.28 0.54 0.52
awarded under A1.4 or A2.7) (0–1) EC10
This table presents the index used to assess the discretionary disclosures about environmental policies, performance and inputs. Index items
are classified 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 third column presents the percentage of firms which made disclosures on that item (discretionary
channels considered are Environmental and/or Social Responsibility Reports, or similar disclosures in firms’ web site). Good environ-
mental performance (EP) firms are firms that have the environmental performance measure (% recycled) above the industry median. The
last two columns present the average score on each item for each group of firms. The significance levels presented in the last column are from
two-sample t-statistics that test the difference between the good and the poor group. ***, **, * represent significance 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 firms.
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).
spending. We do not score disclosures pertaining environmental profile given the existing and forth-
to environmental spending as a result of comply- coming environmental regulations. Finally, we
ing with the existing environmental regulations, code a firm’s disclosures of its environmental ini-
as such disclosures are largely non-discretionary tiatives in A7. Items coded here include employee
and appear in mandatory disclosure channels such training in environmental management, existence
as 10 Ks and annual reports. Rather, we focus on of response plans for environmental accidents,
disclosures of dollar savings from existing environ- internal environmental awards and audit, and
mental programs and efforts and discretionary community involvement through scholarship and
spending to further enhance future environmental donations. Again, these kinds of initiatives can
performance such as investing in new environmen- represent true commitment but they can also be
tal technologies or environmentally related R&D imitated by companies with no real commitments
and innovations. We also include disclosures of to protecting the environment.
fines related to environmental issues. Such penal-
ties are usually immaterial thus their reporting is Econometric model
not mandatory. Nevertheless, the dollar amount
of fines is important to environmental stakehold- Model and variable descriptions
ers to assess the level of true commitment to the
environment. In summary, our index design in In order to test our hypotheses, we employ the
the A1–A4 ‘‘hard’’ disclosure categories makes it following econometric model:
relatively difficult for poor environmental per-
formers to mimic the environmental disclosures VED ¼ b0 þ b1 EP þ b2 J–F coefficient þ b3 FIN
of good environmental performers. þ b4 TOBIN Q þ b5 VOLAT þ b6 ROA
þ b7 LEV þ b8 SIZE þ b9 NEW
Soft disclosure items
We measure a firm’s disclosures of vision and þ b10 CAPIN þ e
environmental strategy claims in A5. For instance,
The variables in the regression above are defined
firms often disclose broadly that they have an envi-
as follows:
ronmental policy, that management is committed
VED – is a score of voluntary environmental
to protecting the environment, etc. Such disclo-
disclosures using web based disclosures as of Sep-
sures can be genuine when put in the specific con-
tember, 2004. We perform a content analysis using
text but they can also be deceiving as they lack
our disclosure index (see Table 1).
credibility and substantiation, and can be easily
EP – is an environmental performance proxy.
mimicked.11 A6 assesses the disclosure of a firm’s
We use two alternative proxies to capture the envi-
ronmental performance of each firm. The first is the
11
Consider, for example, items A5-2, 3 and 5. We classify TRI emission scaled by total sales revenue. To
these statements as soft because they involve claims about facilitate the interpretation of the results, we
environmental management control systems without details or reverse the sign of this variable. In other words,
substantiation. In contrast, the corresponding items in A1-1
and 2 are hard because they provide specific information about
the larger this measure is, the better the environ-
the existence of the department, management positions or mental performance of the firm. The second mea-
board committees responsible for monitoring pollution control. sure is the percentage of toxic waste treated,
recycled or processed in the production. Large performance (EP) measures have been docu-
recycling percentages imply environmentally pro- mented to be causes of voluntary disclosures in
active firms. Both measures are computed using the disclosure literature. We include them to avoid
the public database made available by the US Envi- a correlated omitted variables threat that EP is
ronmental and Protection Agency (EPA). We standing in for some other well known determi-
aggregate the plant specific data at the firm level. nants of disclosure in other disclosure channels.
J–F coefficient – the Janis–Fadner coefficient of We divide the control variables in three groups:
imbalance, measured for the firm’s 2002 fiscal year variables that proxy for the benefits of voluntary
and defined in greater detail below. disclosures, variables that measure costs of volun-
FIN – is the amount of debt or equity capital tary disclosures and other control variables.
raised by the firm in the fiscal year 2004. It is the
sale of common stock and preferred shares minus Benefits of voluntary disclosure
the purchase of common stock and preferred
shares (#108–#115) plus long term debt issuance Financing: It is well known that firms that raise
minus the long term debt reduction (#111– financing in debt and equity markets have a higher
#114).12 The amount is scaled by the size of total propensity for disclosures in voluntary channels
assets at the end of the fiscal year 2002. (Frankel, McNichols, & Wilson, 1995) to lower
TOBIN Q – is Tobin’s Q, measured as market their cost of capital. We use the amount of debt
value of common equity (#25 * #199) plus book and equity financing raised by the firm in the fiscal
value of preferred stock (#10), book value of long year following the measurement of the environ-
term debt (#9) and current liabilities (#5), divided mental performance (FIN).
by book value of total assets (#6). Information asymmetry: It is generally asserted
VOLAT – is stock price volatility, measured as in the voluntary disclosure literature that manag-
standard deviation of market adjusted monthly ers seek to lower information asymmetry through
stock return during fiscal year 2003. voluntary disclosures in order to lower the cost
ROA – is total return on assets measured as the of capital (Healy & Palepu, 2001). Our chosen
ratio of income before extraordinary items (#18) proxies for information asymmetry are: monthly
at the end of fiscal year 2004 and total assets stock return volatility measured (VOLAT) over
(#6) at the end of fiscal year 2003. the 12 month period represented by fiscal 2003
LEV – is the leverage ratio, measured as the (Lim, 2001) and Tobin’s Q, based on the argument
ratio of total debt (#9 + #34) divided by total that firms with greater unbooked intangibles and a
assets (#6) at the end of fiscal year 2003. positive NPV investment opportunity set enjoy
SIZE – is the natural logarithm of the total larger Tobin’s Q (Barth & Kasznik, 1999; Smith
asset value measured as of the end of fiscal year & Watts, 1992).
2003. Firm performance: Lang and Lundholm (1993)
NEW – is the asset newness, measured as a ratio and others have shown that firms with superior
of net properties, plant and equipment (#8) upcoming earnings performance have a higher dis-
divided by the gross properties, plant and equip- closure propensity to reveal their ‘‘good news’’ to
ment (#7) at the end of fiscal year 2003. financial markets. At the time of observing web
CAPIN – is the capital intensity, measured as a disclosures (September 2004), markets would
ratio of capital spending (#128) divided by total know fiscal year 2003 ROA so earnings for the
sales revenues (#12) at the end of fiscal year 2003. upcoming year would be fiscal 2004 ROA.
The control variables included in the multivari- Leverage: A number of disclosure studies (e.g.,
ate regression model besides our environmental Leftwich, Watts, & Zimmerman, 1981) have
argued that the monitoring demand for informa-
tion increases as firm debt increases, and empirical
12
Numbers in brackets represent data items in the Compustat evidence is consistent with managers being more
Annual File. forthcoming, generally to facilitate the contracting
demand for information. Agency costs of debt are level of investments in clean technologies, thus rep-
higher for firms with relatively more debt in their resenting a threat to causal inferences. Firms with
capital structure (Jensen & Meckling, 1976), thus newer, cleaner technologies are likely to have a
voluntary disclosures are expected to increase with superior environmental performance measure and
leverage. it is reasonable to assume that they will want
stakeholders to know about this superior environ-
Costs of voluntary disclosures mental performance in discretionary disclosure
channels. To address this threat, we control for
Firm Size: Most voluntary disclosure studies the average age of a firm’s equipment (NEW)
control for firm size (see, for example, Lang & based on the argument that newer equipment is
Lundholm, 1993) based on the assumption of expected to employ newer and less polluting tech-
economies of scale with respect to information nologies. For similar reasons, firms with higher
production costs.13 sustaining capital expenditures, as proxied by
Proprietary costs: In Verrecchia (1983), a key CAPIN, are expected to have newer equipment
friction sustaining a partial disclosure equilibrium and may want to signal their environmental type
is the existence of proprietary costs associated with through more discretionary disclosures regarding
being forthcoming. In our setting, proprietary their environmental performance.
costs pertain to the manager revealing information Favorable media coverage: Following Aerts and
to environmental regulators and other environ- Cormier (2006), Bansal and Clelland (2004) and
mental activist groups that increase the probability Janis et al. (1965), we measure lagged environmen-
of criticism, sanction or attack (see Li et al., 1997). tal legitimacy as the propensity for unfavorable
We assume that industry serves as a measure of press articles using the Janis–Fadner coefficient
proprietary costs, since pollution propensity and of imbalance. This coefficient ranges from 1
related monitoring by opponents is well known (unfavorable) to +1 (favorable), with zero imply-
to vary by industry. In our inter-industry regres- ing neutral perceptions about the firm’s environ-
sions, we control for industry fixed effects in order mental legitimacy. Following Aerts and Cormier
to control for differing proprietary costs and other (2006) and Patten (2002), legitimacy theory pre-
unidentified factors that might vary by industry. dicts a negative association between lagged envi-
ronmental legitimacy and the level of voluntary
Other control variables specific to environmental environmental disclosures.14 The Janis–Fadner
disclosures coefficient is measured as follows:
Janis–Fadner coefficient
Equipment age and annual capital spending:
Healy and Palepu (2001) describe a common criti- ðe2 ecÞ
¼ if e > c;
cism in the voluntary disclosure literature involv- t2
ing endogeneity. EP and sustainability disclosures ðec e2 Þ
¼ if c > e;
might be joint endogenous variables driven by t2
some underlying exogenous variables such as the ¼ 0 if e ¼ c;
13
Firm size has been regularly established as a determinant of
14
voluntary disclosure in the literature, so its effects must be In our sample of 191 firms, 126 have at least one article in
controlled for. Firm size is marginally significant and positively Factiva database related to the environment in fiscal 2002. The
correlated with % recycled in the Chemical and Pulp and Paper 126 firms generate 770 total articles related to the environment
industries, but is insignificant for our other three industries. In during fiscal 2002. Of these, 393, 207, and 170 are coded by us
Table 4, we control for firm size as a covariate to ensure that as unfavorable, neutral and favorable, respectively. For the 65
our EP measures are not standing in for size. In Table 5, firm firms with no environmental articles, the Janis–Fadner coeffi-
size is once again a covariate in our model. Thus, the role cient is set to a default measure of zero. Thus, silence in the
played by firm size in environmental disclosure is controlled for media is interpreted to imply neutrality of perceptions about
in both tables when we isolate effects due to EP. environmental legitimacy.
where e is the number of favorable environmental For firms providing discretionary environmen-
articles, c is the number of unfavorable articles, tal disclosures, Table 1 presents our scoring model
and t is e + c. along with descriptive statistics as to the percent-
age of firms disclosing a particular item and the
difference in average disclosure scores across good
Sample selection, summary statistics and univariate and poor EP firms. The table also shows the GRI
empirical results reference for each line item. A crucial assumption
of our scoring approach is that the disclosures are
As mentioned in section ‘Introduction’, the discretionary. In support of that assumption, we
sample of this study consists of public US compa- use only disclosures that came from firm environ-
nies from five polluting industries that report toxic mental reports (hereafter EP reports) or web based
release data to the US Environmental Protection disclosures other than firm annual reports or
Agency and have financial and stock price data 10 Ks. We assume that EP reports/web disclosures
in the Compustat and CRSP database.15 The final are purely discretionary, i.e., silence is always an
sample contains 191 firms covered by the EPA- option.
TRI database for 2003 with adequate variables In Table 1, A1–A4 summarize the separation in
available in Compustat and CRSP. The distribu- average awarded scores across good and poor EP
tion across the five industries is as follows: 27 firms performers, using the %Recycled as the measure
in the Pulp and Paper industry (14.14% of the sam- of environmental performance. We use the median
ple), 63 firms in the Chemical industry (32.98% of % recycled in a given industry to classify firms as
the sample), 18 firms in the Oil and Gas industry good versus poor EP performers.17 Recall that,
(9.42% of the sample), 42 firms in the Metals and according to H1a, good EP performers should
Mining industry (21.99% of the sample) and 41 have higher scores for hard disclosure items that
firms in the Utilities industry (21.47% of the sam- are difficult to mimic by poor EP performers.
ple). Among these firms, 122 firms (63.87% of The results in Table 1 confirm that prediction for
the sample) chose to provide discretionary disclo- the A1 category (Governance Structure and Man-
sures about the environment in fiscal 2003: 55 of agement Systems). The average score for good
those firms had stand alone environmental reports (poor) EP performers is 2.08 (1.27). Using a two-
and invariably also had supplementary web disclo- tailed t-test the difference is significant at the 1%
sures, while the remaining 67 firms had discretion- level. For the A2 category (Credibility), the aver-
ary web disclosures but no stand alone age score for good (poor) EP performers is 2.88
environmental reports. If a firm has no environ- (1.95), and the difference is again significant at
mental report or related discretionary web disclo- the 5% level. Not surprisingly, good EP firms are
sures, we classify such a firm as being ‘‘silent’’, more likely to disclose that they adopt GRI guide-
with the disclosure score set to zero for these 69 lines. The difference in average scores 0.21 versus
firms.16 We allow silent firms in the sample since 0.03 is statistically significant at the 1% level. It
non-disclosure is a choice in a partial disclosure is apparent from the scores for A2-2 that 2
equilibrium setting. (61 · 0.03) good EP firms obtained independent
assurance for their ER report/web disclosures
compared to zero poor EP firms. While this is con-
15
We identify all firms with available TRI data and two-digit
sistent with an attempt by good EP firms to signal
SIC codes: 26 (pulp and papers), 28 (chemicals), 29 (oil and their type, the difference between the two types of
gas), 33 (metals and mining), and 49 (utilities). For these firms is not statistically significant.
companies, we read the firm’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 firms with zero disclosures first by
17
email and then by phone. None of them indicated that they Our results in Table 1 are qualitatively unaffected when we
published a stand alone environmental report in 2003. partitioned the sample using the mean % recycled.
Of special interest is category A3 (EPI indica- (poor) EP performers is 0.84 (0.45). The difference
tors), since this type of hard disclosure is not one between the good and poor groups is significant at
that poor EP firms will want to mimic. The aver- 5%. In this category we find that, on average, good
age score for Good EP performers of 10.19 is well EP firms disclose significantly more often amounts
in excess of the corresponding average score for spent on fines than the poor firms. Economics
poor EP performers, 6.00 (the difference is signifi- based voluntary disclosure theories predict that
cant at the 1% level). Each item has a total score of good EP firms will be more forthcoming about
6 based on the dimensions indicated in Table 1. the dollar amount spent on fines because this
The low scores for both good and poor EP per- amount will be lower than the corresponding
formers suggest that disclosures in this area are amount for poor EP firms, consistent with supe-
less than ideal given the GRI guidelines. In unta- rior environmental performance within the indus-
bulated analyses, we analyzed the contribution of try. To validate that disclosing the amount spent
each of the 6 dimensions to the total scores (i.e., on fines signals a firm’s commitment to the envi-
‘‘hits’’) awarded in A3, for good versus poor EP ronment, we examine the dollar amount spent on
firms: performance data presented represents fines and the number of environmental violations
43.50% (48.44%) of total A3 scores awarded for in fiscal year 2000 for the 100 of our sample firms
good (poor) EP firms, respectively; performance (56 good EP firms and 44 poor EP firms) covered
data presented relative to peers represents only in the Corporate Environmental Profile Database
1.06% (2.44%) of total A3 scores awarded for good developed by the Investor Responsibility and
(poor) EP firms, respectively; trend analysis repre- Research Center. In untabulated results, we find
sents 28.86% (26.33%) of total A3 scores awarded that the fifty-six good EP firms had an average
for good (poor) EP firms, respectively; perfor- of 1.13 violations per firm and 0.17 cents of fines
mance relative to targets represents 8.76% per thousand of dollars of sales. In comparison,
(7.31%) of total A3 scores awarded, for good the forty four poor EP firms had an average of
(poor) EP firms, respectively; performance data 2.63 violations per firm and 0.42 cents of fines
presented in both absolute and normalized form per thousand dollars of sales. Using a two-tailed
represents 7.87% (8.63%) of total A3 scores t-test, the difference in the number of violations
awarded, for good (poor) EP firms, respectively; (scaled dollar fines) between the two groups is sta-
and, finally, performance data presented at the tistically significant at 0.02 (0.01) level.18
disaggregated level represents 9.64% (6.68%) of Table 1 contains corresponding statistics for the
total A3 scores awarded, for good (poor) EP firms, soft disclosure categories A5–A7. Consistent with
respectively. Thus, it is apparent that GRI guide- H1a, good EP performers have significantly
lines with respect to performance relative to peers greater soft discretionary disclosures for A7 (envi-
are not being followed by either good or poor EP ronmental initiatives) relative to poor EP perform-
performers. This is not surprising, as it is difficult ers. However, there is no significant difference for
for both types of firms to decide on appropriate A5 (vision and strategy) and A6 (environmental
‘‘peers’’ given differences in production processes profile). Overall, the results for our soft disclosure
across firms within a given industry. It is precisely scores suggest less separation between good and
this dilemma that makes relative environmental poor EP firms, a finding which we explore in more
performance ‘‘unobservable’’ to the typical inves- depth in section ‘A revised role for socio-polittical
tor or stakeholder, creating the potential for the theories’ below.
setting of a partial disclosure equilibrium where
some firms are more transparent than others in
18
their report/web disclosures, and some firms are An alternative interpretation of our results for fines disclo-
entirely silent. sures is that the firms disclosing such fines are seeking to
legitimize the violation of environmental regulation and
The final hard category is A4 (disclosures per- requirements. This pattern of disclosure behaviour is predicted
taining to discretionary environmental spending). by legitimacy theory (Patten, 2000). We leave the validation of
For that category, the average score for good this alternative explanation to future research.
In Table 2, Panel A we present descriptive sta- The firm size measured by the logarithm of total
tistics on environmental disclosure scores by assets (SIZE) is 8.01 implying average total assets
industry. The lowest average score (out of 95) is in dollar terms of $3.01 bn, thus our sample con-
10.08 and is obtained by the Metals and Mining sists of relatively larger firms. The mean (and med-
Industry while the highest score, 25.21, is obtained ian) firm size is comparable with the median firm
by the Pulp and Paper Industry.19 In Table 2, size reported by Patten (2002) for his sample.
Panel B we show that the average ratios of TRI/ The average J–F coefficient is 0.08.
Sales for Pulp and Paper and Metals and Mining We present in Panel B of Table 3 Pearson cor-
are 2.34 (i.e., 2.34 pounds of toxic emissions per relations between the independent variables used
thousand dollar sales) and 1.72, respectively. Thus, in the regressions. The Pearson correlation
it appears that industries with a high (low) pollu- between the negative of TRI/Sales and % recycled
tion propensity are more (less) likely to provide is equal to 0.29 and is statistically significant at 1%,
discretionary environmental disclosures to firm implying that (minus) TRI/Sales and % recycled
stakeholders. This is by no means inconsistent both measure environmental performance but the
with H1a which conditions disclosure predictions overlap is modest.
by environmental type at the firm level relative to
industry peers. Table 2 does confirm differences
across sectors in both environmental disclosure Empirical results involving the level of disclosure
and pollution propensities suggesting the need
for industry controls. Inter-industry analysis
Untabulated analysis for our 69 firms with zero
disclosure scores indicates that these firms have an We present in Table 4 the results of inter-indus-
average TRI/Sales (% recycled) measure of 2.97 try multivariate regressions of disclosure scores on
(59.01%) compared to 1.82 (65.61%) for the 122 the environmental performance measures and the
firms with non-zero disclosure scores. The differ- control variables. We estimate the regressions
ences in both EP measures across the two groups using a Tobit analysis to account for the censoring
are significant at the 5% level. This is generally of the dependent variable at zero.20 We estimate
consistent with H1a, i.e., poor EP firms are more the Tobit regressions by maximum likelihood
likely to opt for silence. using a Newton–Raphson algorithm.21 We run
In Table 3, Panel A we present descriptive sta- three sets of equations based on the different dis-
tistics for independent variables used in the estima- closure scores (total, hard and soft) used on the
tion. The average firm has a negative financing dependent side. All regressions are inter-industry
variable (FIN) meaning that it reduces debt or analyses using dummy variables to control for
repurchases shares more than it raises new financ- industry fixed effects. We estimate the regressions
ing. Also, on average, the ROA is about 5% and using each environmental performance variable
the average leverage (LEV) is 33% of total assets. separately and then include both variables at the
same time.
The first three columns present the results for
19
While our scoring scheme has 95 available points to allow the total disclosure scores. As predicted by H1a,
for rich variety of disclosures from one sustainability report to the estimated coefficients for our environmental
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 firm, Weyerhaeuser, obtained 68
20
of the 95 available points, suggesting an effective maximum of As an alternative, we have re-run our regressions using
71%. This is comparable to the effective maximum in similar simple OLS and the inferences are unchanged.
21
scoring scheme used in ‘‘Risk and Opportunity: the Best In the Tobit model, the marginal effect of a change in an
Practice in Non-financial Reporting’’ by Standard and Poor’s independent variable on the dependent variable (i.e., disclosure
(2004). Nonetheless, the low average scores (out of 95) for each score) is the estimated coefficient times the probability that the
of the four industries point to an overall need for improvement firm provides discretionary environmental disclosures (Verbeek,
in sustainability reporting in the years beyond 2003. 2004).
Table 2
Descriptive Statistics for variables of interest
Panel A: Environmental disclosures (N = 122 firms)
Overall Pulp and paper Chemicals Metals and mining Oil and gas Utilities
(n = 122) (n = 24) (n = 41) (n = 13) (n = 13) (n = 31)
P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327

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 1.68 (1.00) 1.95 (2.00) 1.95 (2.00) 1.31 (1.00) 2.00 (2.00) 1.10 (1.00)
manage. systems (0–6) 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 profile (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 firms)
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 firms that chose to provide discretionary disclosures (i.e., ‘‘disclosing’’ firms) on their
environmental performance (N = 122 firms). 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’’ firms and ‘‘silent’’ firms (N = 191 firms). 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 firm.
319
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
TRI/Sales 2.24 0.62 2.50 0.13 3.68
J–F coefficient 0.08 0.00 0.00 0.37 0.81
SIZE 8.01 7.89 6.70 9.28 1.70
FIN 0.02 0.01 0.03 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 J–F
disclosure coefficient
% Recycled 0.29*** 0.06 0.15** 0.28*** 0.07 0.14** 0.02 0.26*** 0.30*** 0.23*** 0.05
TRI/Sales – 0.04 0.01 0.11 0.16** 0.08 0.01 0.03 0.01 0.17*** 0.13*
SIZE – 0.07 0.17** 0.09 0.01 0.01 0.33*** 0.21*** 0.46*** 0.05
FIN – 0.03 0.03 0.05 0.02 0.05 0.03 0.03 0.01
TOBIN Q – 0.01 0.47*** 0.04 0.15** 0.06 0.14 0.03
VOLAT – 0.11 0.03 0.15** 0.06 0.16** 0.15**
ROA – 0.11 0.10 0.07 0.04 0.03
LEV – 0.14* 0.01 0.19*** 0.10
NEW – 0.18** 0.12** 0.02
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 firms. Percent recycled is toxic waste treated or recycled divided by total waste generated by firm. 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 fiscal 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 coefficient is the Janis–Fadner Coefficient
(see Bansal and Clelland, 2004 for details). Spearman Correlation statistics provide similar results. ***, **, * represent significance levels
(two-tailed) at 1%, 5% and 10%, respectively.
performance proxies are all positive and signifi- find that they are significant and positive suggest-
cant (recall, we reverse the sign of TRI/Sales to ing that each provides incremental information
facilitate the interpretation). This result is consis- to the other with respect to voluntary environmen-
tent with the discretionary disclosure theories of tal disclosures.
Verrecchia (1983) and Dye (1985), but inconsis- Table 4 indicates the predicted sign of associa-
tent with the negative association (H1b) pre- tion for our control variables given prior general
dicted by socio-political theories. Firms with disclosure literature and intuition. For the total
better environmental performance have more vol- disclosures as well as hard versus soft disclosures,
untary disclosures about their environmental the forward looking financing proxy is positive
impact. Furthermore, when we introduce both (consistent with the prior literature) but is statisti-
scaled TRI and % recycled in the regression we cally significant only for the soft disclosure cate-
Table 4
Inter-industry regressions with industry fixed effects (Tobit analysis)
Dependent variables
Total disclosures Hard disclosures Soft disclosures
(1) (2) (3) (1) (2) (3) (1) (2) (3)
Intercept 81.2*** 74.72*** 78.67*** 70.65*** 64.00*** 68.01*** 20.65*** 19.03*** 19.85***
(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)
TRI/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 coefficient () 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 (+) 1.54 1.88 1.83 2.18 2.05 2.45 0.44 0.48 0.51
(0.86) (1.22) (1.24) (2.49) (2.24) (3.25) (0.64) (0.75) (0.85)
VOLAT (+) 4.69 4.05 3.56 1.44 7.45 8.39 1.93 0.05 0.08
(0.17) (0.11) (0.09) (0.02) (0.55) (0.72) (0.26) (0.01) (0.01)
ROA (+) 2.56 1.31 0.85 9.49 11.32 13.73 5.01 5.57 4.54
(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 (+) 17.57** 17.63** 18.22** 12.16* 13.09** 12.73** 5.62** 5.64** 5.73**
(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 effects Yes Yes Yes Yes Yes Yes Yes Yes Yes
Log-likelihood 497.98 499.55 495.88 435.27 434.96 431.62 359.50 360.01 357.84
N = no of firms 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. Coefficients are estimated by maximum likelihood (Tobit regressions). The significance levels are based on Chi-squared
statistics (presented in parentheses). All control variables are defined in Table 3. ***, **, * represent significance levels (two-tailed) at 1%,
5% and 10%, respectively.
gory. The information asymmetry proxies (Tobin’s production costs. Similarly, firms with greater cap-
Q and stock volatility) are insignificant. Similarly, ital expenditures disclose more: the coefficient of
firm profitability (ROA) in the following year is the capital intensity control variable is significantly
insignificant. One interpretation of this result is positive. In contrast to our intuition, firms with
that firms resort to other disclosure channels to newer equipment (NEW) are less likely to provide
reduce the information asymmetry and convey discretionary environmental disclosures. The
their good news about future ROA, and do not lagged Janis–Fadner coefficient is not associated
use ER reports/web disclosures for this purpose. with the level of disclosure, which is consistent
Turning to our agency proxy, we find that the with the results of Aerts and Cormier (2006).
leverage variable is significantly positive, suggest- We also split our disclosure scores in two parts:
ing that debtholders exercise pressure on firms to hard disclosures and soft disclosures. Hard disclo-
disclose environmental related matters to assess sures are closest in spirit to the assumed truthful
potential future liabilities. Also, large firms dis- disclosures in the voluntary disclosure theories dis-
close more, consistent with their lower information cussed in section ‘Literature review and hypothesis
development’. However, we find similar results for approach: (1) some of our variables (i.e., TRI/
both hard and soft disclosure scores: the estimated Sales) can be compared within an industry but
coefficients for our two EP measures are positive cannot be meaningfully compared across indus-
and significant for both hard and soft disclosures tries due to differences in production processes;
as predicted by H1a. (2) the distribution of our disclosure scores for
the Metal and Mining sector are quite low, so tak-
Intra-industry analysis ing intra-industry ranks and then pooling facili-
tates the i.i.d assumption requirement of our
Following Lang and Lundholm (1993, 1996) OLS regression model.
and Healy et al. (1999) we supplement the above We rank the dependent and the independent
tests with an intra-industry approach which seeks variables within industry and then we pool the cor-
to explain the within-industry variation in the responding percentiles across the five sectors. The
ranks of disclosure scores using the within-indus- regression results are presented in Table 5. They
try variation in the ranks of our independent are generally the same as in Table 4 and indicate
variables. There are two reasons justifying this 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)
TRI/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 coefficient () 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 (+) 0.04 0.05 0.06 0.04 0.06 0.07 0.02 0.03 0.04
(0.67) (0.89) (1.07) (0.72) (0.98) (1.21) (0.36) (0.50) (0.62)
VOLAT (+) 0.06 0.04 0.05 0.06 0.03 0.05 0.04 0.03 0.04
(1.15) (0.83) (0.99) (1.15) (0.54) (0.97) (0.84) (0.60) (0.76)
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 (+) 0.11** 0.11** 0.11** 0.09* 0.09** 0.09** 0.08* 0.08* 0.08*
(2.34) (2.26) (2.42) (1.92) (1.99) (2.03) (1.65) (1.64) (1.68)
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. R2 40.83% 40.33% 43.22% 36.38% 37.32% 39.94% 34.98% 34.36% 35.69%
N = no of firms 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. Coefficients are estimated by OLS regressions using the ranked variables. The
significance levels are based on t statistics (presented in parentheses). All control variables are defined in Table 3. ***, **, * represent
significance levels (two-tailed) at 1%, 5% and 10%, respectively.
performance measures and the discretionary envi- A5-1 which involves a CEO statement on environ-
ronmental disclosures (total scores). Both (minus) mental performance to stakeholders.22
TRI/Sales and % recycled complement each other More formally, using the ratio of soft disclosure
and neither is redundant for the total sample or for scores to total awarded scores as a proxy for legit-
the hard and soft groups considered separately. imization, we would expect to find a negative rela-
We find similar results when we split the total tion between prior perceived legitimacy and this
scores into hard and soft disclosure scores: the esti- ratio. Our tests based on the conditional sample
mated coefficients for our two environmental per- of firms with scored sustainability/web disclosures
formance measures are positive and significant as appear in Table 6. We measure threats to legiti-
predicted by H1a. The results for the control vari- macy in two ways: group EP membership based
ables are also generally comparable to Table 4. on the median % recycled and group media cover-
The Janis–Fadner coefficient is now positively age based on the median Janis–Fadner Coefficient.
associated with the level of soft disclosures, which There are 122 (95) firms with scored sustainability/
is opposite to the negative association predicted by web disclosures (at least one environmental press
the socio-political theories. article in 2002) and available EP. As indicated in
Table 6, the ratio of soft/total scores is 50.95%
for poor EP firms, compared to 34.23% for good
A revised role for socio-political theories EP firms. Similarly, the ratio of soft/total scores
is 47.54% for unfavorable media coverage firms,
Overall, the results in the previous sections compared to 32.58% for firms with favorable
point to economic disclosure theory and not media coverage. In both instances, the difference
socio-political theory as being robust in predicting is statistically significant at the 1% level. Both
the level of discretionary environmental disclosure. results are consistent with firms whose environ-
If the focus of enquiry is switched to disclosure mental legitimacy is threatened to make soft claims
strategies, it may be that socio-political theories to be committed to the environment. This result is
explain interesting patterns in the data. While we predicted by legitimacy theory but cannot be
leave this possibility to future research, we offer explained by economic disclosure theories, which
in this section some preliminary evidence on the assume truth-telling and thus offer no predictions
question. about biased disclosures.
Legitimacy theory predicts that firms with These results hold when we repeat the tests
threatened legitimacy are likely to make self-serv- (untabulated) in a multivariate fashion using all
ing disclosures referred to as ‘‘legitimization’’ other control variables in Tables 4 and 5 and
(see, for example, Adams, 2004; Gray et al., employ the ratio of soft/total as the dependent var-
1995, p. 54 & Hughes et al., 2001, p. 219). One iable. There is a significant negative association
example of legitimization is for a poor environ- between the lagged Janis–Fadner Coefficient and
mental performer to make soft claims to be com- the ratio of soft/total scores, after including all
mitted to the environment which are not readily control variables. Similar results hold for our
verifiable. In fact, it is apparent from an inspection two EP measures. This result is robust to the Tobit
of Table 1 that many poor EP firms do make soft analysis approach (Table 4) and the rank regres-
claims to be committed to the environment. Focus- sion analyses approach (Table 5) and implies a
ing on A5, Vision and Strategy Claims, over 90% greater propensity for ‘‘legitimization’’ behavior
of our poor EP firms make a claim which is
awarded at least one mark in category A5. For
example, 55 of 61 poor EP firms (i.e., 0.90 * 61)
22
make a claim that is awarded a score of 1 for item As anecdotal evidence, consider the following soft environ-
mental commitment claim made by one of our sample chemical
A5-2 which is a statement of corporate environ- firms in the bottom quartile of our environmental performance
mental policy and/or commitment. Further, 33 of ranking: ‘‘Senior management leads the industry with respect to
our 61 poor EP firms are awarded a score under responsible care.’’
Table 6
Comparisons of soft to total disclosure scores
Panel A: Group membership is based on median % recycled (N = 122 firms)
Disclosure categories Average score Difference (t-stat)
Good EP firms (N = 61) Poor EP firms (N = 61)
Soft/total (%) 34.23% 50.95% 16.72%***
(3.99)
Panel B: Group membership is based on median Janis–Fadner coefficient (N = 95 firms)*
Disclosure categories Average score Difference (t-stat)
Favorable media coverage (N = 48) Unfavourable media coverage (N = 47)
Soft/total (%) 32.58% 47.54% 14.96%***
(3.22)
This table presents average ratios of soft to total disclosure scores. Panel A presents differences in soft to total disclosure ratios across
good environmental performance (EP) firms and poor environmental performance firms. Good environmental performance (EP) firms
are firms that have the environmental performance measure (% recycled) above the industry median. Panel B presents differences in
soft to total disclosure ratios across favorable media coverage firms and unfavorable media coverage firms. Firms with favorable media
coverage are firms that have a Janis–Fadner Coefficient above the sample median (see Bansal and Clelland, 2004 for details). Only firms
that have at least one article in Factiva are included. The significance levels presented in the last column are from two-sample t-
statistics that test the difference between the groups. ***, **, * represent significance 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.
for firms whose environmental legitimacy is environmental performers, using the % recycled
threatened. measure. We split the firms in the two groups
based on the level of the alternative environmental
performance measure, TRI/sales. We still find sig-
Sensitivity analysis nificant differences between the good and the poor
groups across all main categories in the index.
For each disclosure item in the index presented Finally, we run a Logit model to investigate
in Table 1, we have replicated the difference tests whether the decision to provide any discretionary
across the two EP groups using Wilcoxon rank environmental disclosures is affected by the envi-
tests. Non-parametric Wilcoxon rank tests are ronmental performance of the sample firms. We
robust to the possibility that the data does not fol- perform an inter-industry analysis using the same
low a normal distribution. In addition, we have control variables as discussed in section ‘Econo-
bootstrapped the p-values of the two-sample parametric model’. Results (unreported) are consistent
metric t-tests by re-sampling the data with replace- with the main results from Table 4. Environmen-
ment.23 Both approaches result in the same levels tal performance is found to be one of the main
of significance as the ones reported for the para- drivers of the probability that the firm provides
metric t-tests. environmental disclosures in discretionary chan-
We verify whether the difference tests in disclo- nels. Precent recycled and (minus) TRI/Sales are
sure scores reported in Table 1 are robust to the significantly positive at the 1% level when intro-
classification of sample firms as good and poor duced separately. When both are in the model,
the significance level decreases to 10%. The J–F
23 coefficient is not significant. All models show a
We implement the bootstrap procedure by drawing with
replacement 20 000 samples from each EP group. Prior to good fit (significant likelihood ratios and almost
resampling, the procedure mean-centers the data within each 90% concordant observations) suggesting a good
group (for details, see Westfall & Young, 1993). model specification.
Conclusions performance within each industry in order to con-

trol for industry differences in pollution propensity.
Previous empirical evidence provides mixed Our results are inconsistent with the prediction of a
results on the relationship between corporate envi- negative association from socio-political theories,
ronmental performance and the level of environ- suggesting that these theories are not robust in pre-
mental disclosures. We revisit this relation by dicting the level of discretionary disclosure.
testing competing predictions from economics Although there is separation in scores and good
based and socio-political theories of voluntary dis- EP firms disclose more, the scores of good EP
closure using a more rigorous research design. In firms are low relative to the expectation implied
particular, we improve the prior literature in two by the 2002 GRI reporting guidelines and point
important ways. First, the predictions of voluntary to the need for improvement in the years beyond
disclosure theory relate to discretionary, not man- 2003. Specific areas where improvement is required
datory, environmental disclosures. This study include obtaining independent assurance of sus-
focuses on purely voluntary disclosure media such tainability reports and the disclosure of environ-
as corporate Internet web sites and stand-alone mental performance indicators.
environmental reports. Previous studies assessed Finally, our results suggest important directions
environmental disclosures mainly from annual for future research. Specifically, we provide preli-
reports and other regulatory filings such as 10 Ks minary evidence that socio-political theories are
and many of those studies rely on a Wiseman robust in predicting what is being said. In particu-
(1982) based content analysis index to measure lar, we find that firms whose environmental legiti-
the extent of environmental disclosures. macy is threatened make soft claims to be
Second, in collaboration with an environmental committed to the environment. This behavior is
reporting expert, we develop a content analysis predicted by legitimacy theory but cannot be
index to assess the level of environmental disclo- explained by economics disclosure theory. Thus,
sure in environmental and social responsibility we argue that future environmental disclosure
reports or similar disclosures in the firm’s web site. research should move the focus of enquiry beyond
The index, which follows closely the Global the level of disclosure.
Reporting Initiative (2002) sustainability reporting
guidelines, differs from the Wiseman index in that
we focus on firm disclosures related to its commit- Acknowledgements
ment to protect the environment.
Our results are as follows. We find a positive We are grateful for comments and suggestions
association between environmental performance from the Editor (Anthony G. Hopwood), two
and the level of discretionary disclosures in envi- anonymous reviewers and from Kate Bewley, Walt
ronmental and social reports or related web disclo- Blacconiere, Denis Cormier, Gus De Franco,
sures. In other words, superior environmental Kathy Herbohn, Ole-Kristian Hope, Hai Lu, Mi-
performers are more forthcoming in truly discre- chel Magnan, Naomi Soderstrom and seminar
tionary disclosure channels, as predicted by eco- participants at Chinese University of Hong Kong,
nomics based voluntary disclosure theories. Our University of Queensland and University of Tor-
findings are robust to two reliable environmental onto. We acknowledge the financial support of
performance measures that use actual toxic emis- the Canadian Academic Accounting Association,
sion and waste management data. The first one is Canadian Institute of Chartered Accountants,
based on Toxics Release Inventory scaled by sales and the AIC Institute for Corporate Citizenship
data at firm level (i.e., TRI normalized by firm’s at Rotman School of Management, University of
operational scale) and the second one is percentage Toronto. We also thank concurrent session partic-
of total toxic wastes that were treated or processed ipants and discussants at the June 2006, July 2006
by each firm. In addition, our findings are not and August 2006 meetings of the CAAA,
affected when we assess the relative environment AFAANZ (a Best Paper Award) and AAA,
respectively. We thank Rod Lohin, Bill Swirsky, Cormier, D., & Magnan, M. (1997). Investors’ assessment of
and Alan Willis for their support and encourage- implicit environmental liabilities: An empirical investiga-
tion. Journal of Accounting and Public Policy, 16, 215–241.
ment for this project and Gauri Bhat, Grace Jin Cormier, D., Magnan, M., & Morard, B. (1993). The impact of
and Dushyantkumar Vyas for help in data collec- corporate pollution on market valuation: Some empirical
tion. Gordon Richardson thanks KPMG for their evidence. Ecological Economics, 8, 135–155.
generous financial support. 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
References environmental performance. American Bar Association
Conference on Environmental Law, Keystone, Colorado,
Adams, C. A. (2004). The ethical, social and environmental March 8–11.
reporting-performance portrayal gap. Accounting, Auditing Frankel, R., McNichols, M., & Wilson, P. (1995). Discretionary
and Accountability Journal, 17(5), 731–757. disclosure and external financing. The Accounting Review,
Aerts, W., & Cormier, D. (2006). The association between 70, 135–150.
media legitimacy and corporate environmental communi- Freedman, M., & Wasley, C. (1990). ‘The association between
cation. Working paper. environmental performance and environmental disclosure in
Aerts, W., Cormier, D., & Magnan, M. (2006). The interface annual reports and 10 Ks’. Advances in Public Interest
between print- and web-based corporate environmental Accounting, 3, 183–193.
disclosure, financial markets and the media. Working Paper. Global Reporting Initiative (2002). Sustainable Reporting
Al-Tuwaijri, S. A., Christensen, T. E., & Hughes, K. E. (2004). Guidelines, www.globalreporting.org.
The relations among environmental disclosure, environ- Gray, R., Kouhy, R., & Lavers, S. (1995). Corporate social and
mental performance, and economic performance: A simul- environmental reporting: A review of the literature and a
taneous equations approach. Accounting, Organizations and longitudinal study of UK disclosure. Accounting, Auditing
Society, 29(5-6), 447–471. and Accountability Journal, 8(2), 47–77.
Bansal, P., & Clelland, I. (2004). Talking trash: Legitimacy, Healy, P. M., Hutton, A. P., & Palepu, K. G. (1999). A Stock
impression management, and unsystematic risk in the Performance and Intermediation Changes Surrounding
context of the natural environment. Academy of Manage- Sustained Increases in Disclosure. Contemporary Accounting
ment Journal, 47(1), 93–102. Research, 16, 485–520.
Barth, M., & Kasznik, R. (1999). Share Repurchases and Healy, P. M., & Palepu, K. G. (2001). Information asymmetry,
Intangible Assets. Journal of Accounting and Economics, corporate disclosure, and the capital markets: A review of
28(2), 211–241. the empirical disclosure literature. Journal of Accounting and
Barth, M., & McNichols, M. (1994). Estimation and market Economics, 31(1–3), 405–440.
valuation of environmental liabilities relating to superfund Hughes, S. B., Anderson, A., & Golden, S. (2001). Corporate
sites. Journal of Accounting Research Supplement, 32, environmental disclosures: Are they useful in determining
177–209. environmental performance? Journal of Accounting and
Barth, M., McNichols, M., & Wilson, P. (1997). Factors Public Policy, 20, 217–240.
influencing firms’ disclosures about environmental liabili- Hutton, A., Miller, G., & Skinner, D. (2003). The Role of
ties. Review of Accounting Studies, 2, 35–64. Supplementary Statements with Management Earnings
Beets, S. D., & Souther, C. (1999). Corporate environmental Forecasts. Journal of Accounting Research, 41, 867–890.
reports: the need for standards and an environmental Ilinitch, A., Soderstrom, N., & Thomas, T. (1998). Measuring
assurance service. Accounting Horizons, 13(2), 129–145. corporate environmental performance. Journal of Account-
Bewley, K., & Li, Y. (2000). Disclosure of environmental ing and Public Policy, 17, 387–408.
information by Canadian manufacturing companies: A Ingram, R. W., & Frazier, K. (1980). Environmental Perfor-
Voluntary Disclosure Perspective. Advances in Environmen- mance and Corporate Disclosure. Journal of Accounting
tal Accounting and Management, 1, 201–226. Research, 18(2), 614–622. Autumn.
Blacconiere, W. G., & Patten, D. M. (1994). Environmental Janis, I. L., & Fadner, R. (1965). The coefficient of imbalance.
disclosures, regulatory costs, and changes in firm value. In H. D. Lasswell & N. Leites, et al. (Eds.), Language of
Journal of Accounting and Economics, 18, 357–377. Politics (pp. 153–169). Cambridge MA: MIT Press.
Chan-Fishel, Michelle. (2002). Survey of climate change Jensen, M., & Meckling, W. (1976). Theory of the Firm:
disclosure in SEC filings of automobile, insurance, oil & Managerial Behavior, Agency Costs, and Ownership Struc-
gas, petrochemical, and utilities companies. Friends of the ture. Journal of Financial Economics, 3(4), 305–360.
Earth – US, September. King, Andrew, & Lenox, Michael. (2001). Does it really pay to
Clarkson, P., Li, Y., & Richardson, G. (2004). The market be green? An empirical study of firm environmental and
valuation of environmental expenditures by pulp and paper financial performance. The Journal of Industrial Ecology,
companies. The Accounting Review, 79, 329–353. 5(1), 105–116.
Lang, Mark, & Lundholm, Russell. (1993). Cross-sectional Patten, Dennis (2000). Changing Superfund disclosure and its
determinants of analyst ratings of corporate disclosures. relation to the provision of other environmental informa-
Journal of Accounting Research, 31(2), 246–247. Autumn. tion. Advances in Environmental Accounting and Manage-
Lang, M., & Lundholm, R. (1996). Corporate disclosure policy ment, 1, 101–121.
and analyst behavior. The Accounting Review, 71(4), Patten, Dennis (2002). The relation between environmental
467–492. performance and environmental disclosure: A research note.
Leftwich, R. W., Watts, R. L., & Zimmerman, J. L. (1981). Accounting, Organizations, and Society, 27, 763–773.
Voluntary corporate disclosure: The case of interim report- Richardson, A., & Welker, M. (2001). Social disclosure,
ing. Journal of Accounting Research, 18, 50–77. financial disclosure and the cost of equity capital. Account-
Li, Y., & McConomy, B. (1999). An empirical examination of ing, Organizations and Society, 26(7/8), 597–616.
factors affecting the timing of environmental accounting Smith, C., & Watts, R. (1992). The Investment Opportunity Set
standard adoption and the impact on corporate valuation. and Corporate Financing, Dividend and Compensation
Journal of Accounting, Auditing and Finance, 14(3), 279–313. Policies. Journal of Financial Economics, 32(1), 263–293.
Li, Y., Richardson, G. D., & Thornton, D. (1997). Corporate Standard & Poor’s (2004). Risk and Opportunity: the best
disclosure of environmental information; theory and evi- practice in non-financial reporting, first edition, ISBN 1-
dence. Contemporary Accounting Research, 14(3), 435–474. 903168-12-0.
Lim, T. (2001). Rationality and Analysts’ Forecast Bias. Verbeek, M. (2004). A Guide to Modern Econometrics. John
Journal of Finance, LVI(1), 369–385. Wiley and Sons.
Lindblom, C. (1994). The implications of organizational Verrecchia, R. (1983). Discretionary disclosure. Journal of
legitimacy for corporate social performance disclosure. Accounting and Economics, 5, 179–194.
Paper presented at the Critical Perspectives on Accounting Westfall, P., & Young, S. (1993). Resampling-Based Multiple
Conference, New York. Testing: Examples and Methods for P-Value Adjustment.
Patten, Dennis (1992). Intra-industry Environmental Disclo- John Wiley and Sons.
sures in Response to the Alaskan Oil Spill: A Note on Wiseman, J. (1982). An evaluation of environmental disclosures
Legitimacy Theory. Accounting, Organizations, and Society, made in corporate annual reports. Accounting, Organiza-
17, 471–475. tions and Society, 7(1), 553–563.

2008 - PeterM - Clarkson - Revisiting The Relation Between Environmental Performance and Environmental Disclosure An Empirical Analysis

Uploaded by

Copyright:

Available Formats

2008 - PeterM - Clarkson - Revisiting The Relation Between Environmental Performance and Environmental Disclosure An Empirical Analysis

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2008 - PeterM - Clarkson - Revisiting The Relation Between Environmental Performance and Environmental Disclosure An Empirical Analysis

Uploaded by

Copyright:

Available Formats

Available online at www.sciencedirect.

Accounting, Organizations and Society 33 (2008) 303–327

Revisiting the relation between environmental

Soft disclosure items Map to Percentage of Average score

P.M. Clarkson et al. / Accounting, Organizations and Society 33 (2008) 303–327

Conclusions performance within each industry in order to con-

You might also like