Description
Previous studies of the relation between environmental performance and environmental disclosure have consistently
documented a lack of significance. This study examines the relation between 1990 annual report environmental disclosures
for a sample of 131 US companies and their environmental performance as based on toxics release data from
1988 (made available in 1990). In contrast to the previous examinations, results indicate that, controlling for firm size
and industry classification (two factors previously shown to be related to the extent of environmental disclosure), there
is a significant negative relation between performance and disclosure for the sample firms.
The relation between environmental performance and
environmental disclosure: a research note
Dennis M. Patten*
Department of Accounting, College of Business, Illinois State University,
Stevenson Hall, Campus Box 5520, Normal, IL 61761, USA
Abstract
Previous studies of the relation between environmental performance and environmental disclosure have consistently
documented a lack of signi?cance. This study examines the relation between 1990 annual report environmental dis-
closures for a sample of 131 US companies and their environmental performance as based on toxics release data from
1988 (made available in 1990). In contrast to the previous examinations, results indicate that, controlling for ?rm size
and industry classi?cation (two factors previously shown to be related to the extent of environmental disclosure), there
is a signi?cant negative relation between performance and disclosure for the sample ?rms. However, the disclosure level
of ?rms from non-environmentally sensitive industries is more a?ected by toxic release levels than is the disclosure of
?rms from environmentally sensitive industries. # 2002 Elsevier Science Ltd. All rights reserved.
Previous studies of the relation between corpo-
rate environmental performance and environ-
mental disclosure in ?nancial reports consistently
document a lack of signi?cance. This is troubling
in that socio-political theories of social disclosure
(see, e.g. Gray, Kouhy, & Lavers, 1995; Guthrie &
Parker, 1990; Hackston & Milne, 1996; Lindblom,
1994; Patten, 1991, 1992, 1995) suggest that the
extent of this disclosure is a function of exposure
to public pressure in the social/political environ-
ment.
1
According to these theories, companies
facing greater exposures, as companies with
poorer environmental performance could be
assumed to do, would be expected to provide more
extensive environmental disclosures, and as such,
0361-3682/02/$ - see front matter # 2002 Elsevier Science Ltd. All rights reserved.
PI I : S0361- 3682( 02) 00028- 4
Accounting, Organizations and Society 27 (2002) 763–773
www.elsevier.com/locate/aos
* Corresponding author. Tel.: +1-309-438-7857; fax: +1-
309-438-8431.
E-mail address: [email protected] (D.M. Patten).
1
Public pressure in the social/political environment is iden-
ti?ed by Walden and Schwartz (1997, p. 127) as consisting of
both social changes and regulatory e?ects. And while the social
changes are in turn seen as a function of what Boulding (1978,
p. 88) refers to as the cultural and political environments, reg-
ulatory e?ects result from the legal environment. Thus, public
pressure can arise due to concerns of the general population,
political bodies, or regulatory (or legal) agencies (see Boulding,
1978, p. 88).
a negative association between performance and
disclosure is posited. This study suggests that the
lack of evidence to date for a relation between
environmental performance and environmental
disclosure may be a function of problems in the
previous analyses. These problems include failure to
control for other factors in?uencing the extent of
disclosure, inadequate sample selection, and an
inadequate measure of environmental performance.
This study examines the relation between the
1990 environmental disclosures included in a
sample of 131 US companies’ annual reports and
?rm environmental performance as based on data
provided by the Environmental Protection Agen-
cy’s 1988 Toxics Release Inventory (released in
1990).
2
The analysis documents that, controlling
for ?rm size and industry classi?cation, two fac-
tors consistently shown to be related to the extent
of environmental disclosure, there is a signi?cant,
negative relation between performance and dis-
closure. This ?nding is robust to measures of dis-
closure based on both content analysis and
?nancial report line count. However, the impact of
the TRI release level on environmental disclosure
is not consistent across sample ?rms. The level of
disclosure for companies from less environmen-
tally sensitive industries is a?ected more by the
TRI release level than is the extent of disclosure
for ?rms from more environmentally sensitive
industries.
In general, it appears that in contrast to the
?ndings of previous studies, there is some relation
between environmental performance and environ-
mental disclosure. However, environmental per-
formance, as based on toxic release information, is
a more signi?cant explanatory factor for varia-
tions in annual report environmental disclosure
for ?rms not already subject to higher potential
environmental regulatory pressure due to industry
involvement. The paper begins with a summary of
previous analyses of the relation between environ-
mental performance and environmental disclosure.
1. Previous studies
Three of the four published studies examining
the relation between environmental performance
and ?nancial report environmental disclosure
relied on indices of environmental performance
issued by the Council on Economic Priorities
(CEP) in the early to mid 1970s. The CEP, a non-
pro?t organization specializing in the analysis of
corporate social activities, attempted to measure
the environmental performance of 50 ?rms from
four highly polluting industries (steel, oil, paper,
and electric utility). The CEP published reports
rating the companies on a 0 (best) to 10 (worst)
scale, and also published within-industry rankings
based on the organization’s analyses. Unfortu-
nately, evaluation criteria di?ered across indus-
tries resulting in reports that ‘‘were of varying
quality and depth’’ (Freedman & Wasley, 1990,
p. 185). However, they were at that time, accord-
ing to Wiseman (1982, p. 54), ‘‘the best credible
source available for objective environmental per-
formance measures on individual companies.’’
Ingram and Frazier (1980), the ?rst study to
examine the environmental performance–dis-
closure relation, used the CEP scores, standar-
dized using industry averages, as the measure of
environmental performance. Annual reports for
40 of the CEP-rated ?rms were analyzed for
environmental disclosures. The year of exami-
nation for each company was based on the year of
the corresponding CEP report and ranged from
1970 through 1974. Content analysis, a scoring
system awarding points based on the presence or
absence of information items, was used to measure
the extent of environmental disclosure. Neither
product–moment correlation nor partial corre-
lation controlling for the volume of disclosure
showed any signi?cant relations with environ-
mental performance. Similarly, multiple regression
analysis using independent variables obtained
from factor analysis of the disclosure items failed
to document any signi?cant relation to the level of
environmental performance.
2
A fundamental argument for this study is that environ-
mental performance will in?uence environmental disclosure
through increased public policy pressure related to perfor-
mance level. For this to hold, the performance information
must be publicly available. As such, it is necessary to use the
1990 annual report disclosures since this is the year the release
data from 1988 was made publicly available.
764 D.M. Patten / Accounting, Organizations and Society 27 (2002) 763–773
Wiseman (1982), the second study using CEP
data, relied on the within-industry rankings of
performance rather than the individual company
scores. Wiseman’s sample of 26 ?rms was drawn
from the oil, steel, and paper industries only.
Similar to Ingram and Frazier (1980), Wiseman
used content analysis to measure the extent of
environmental disclosure in company annual
reports. Wiseman also included a total disclosure
score and a line count measure. Rather than using
these raw measures, however, Wiseman used the
within-industry ranking based on the disclosure
scores. Industry speci?c comparisons were made
for steel ?rms for both 1972 and 1976 (the CEP
released separate rankings based on each of
these years), oil ?rms for 1974, and paper com-
panies for 1972. Of 24 comparisons computed
only one statistically signi?cant correlation was
found. Wiseman concluded (p. 61) that ‘‘overall,
there appears to be no relationship between the
measurable information companies disclose
about environmental performance in their
annual reports and their actual environmental
performance.’’
The last of the CEP-based analyses, Freedman
and Wasley (1990), operationalized environmental
performance and environmental disclosure in the
same manner as Wiseman (1982), although these
authors did use the raw disclosure scores for ana-
lysis rather than the within-industry rankings of
the scores as Wiseman did. In addition, Freedman
and Wasley examined a larger sample, 49 ?rms,
including companies from the electric utility
industry. Finally, they also analyzed 10K reports
in addition to just annual reports. Similar to
Wiseman, comparisons of performance and dis-
closure were made within industry only. Sig-
ni?cant correlations were reported for only two
of 50 comparisons, leading Freedman and Was-
ley to conclude that neither annual report nor
10K environmental disclosures ‘‘are indicative of
actual ?rm environmental performance’’ (p.
191).
The ?nal study of the relation between
environmental performance and environmental
disclosure, Fekrat, Inclan, and Petroni (1996),
is based on more recent disclosures (1991
annual reports) and on a di?erent measure of
environmental performance. Fekrat et al.
employed the content analysis scheme developed
by Wiseman (1982) to quantify the environmental
disclosures for a sample of 26 of the largest US
?rms, and then used the rankings based on these
scores as the disclosure index. Environmental per-
formance measures, obtained from the CEP, were
based not on CEP-based analyses as the 1970s
indices were, but on ?rm-reported toxic emissions
data required through the Emergency Planning
and Community Right-to-Know Act of 1986
(EPCRA). These ?rm speci?c data were made
available in 1990 for 1988 calendar year emissions.
The CEP reported the data adjusted for ?rm size,
and Fekrat et al. then ranked the companies based
on these adjusted measures. Correlations of the
rankings on performance and disclosure were
negative but not statistically signi?cant.
2. Problems with the previous analyses
The failure to ?nd a signi?cant relation between
environmental performance and environmental
disclosure in any of the studies to date appears to
present rather convincing evidence that no such
relation exists. However, there are problems with
each of the previous studies that are substantial
enough to bring the ?ndings into question. The
problems relate to (1) failure to control for other
factors associated with the level of environmental
disclosure; (2) inadequate sample selection; and (3)
an inadequate measure of environmental perfor-
mance. Each of these problems is discussed in
more detail below.
A number of studies over the past decade
attempted to identify factors associated with the
variation in environmental disclosure across com-
panies (see, e.g. Cowen, Ferreri, & Parker, 1987;
Deegan & Gordon, 1996; Gamble, Hsu, Kite, &
Radtke, 1995; Gray et al., 1995; Hackston &
Milne, 1996; Patten, 1992). These studies provide
a strong body of evidence that both company size
and industry classi?cation are factors that
appear to in?uence the level of environmental
information disclosed in ?nancial reports. Lar-
ger companies, presumably due to visibility
concerns, tend to disclose more information
D.M. Patten / Accounting, Organizations and Society 27 (2002) 763–773 765
than smaller ?rms.
3
Similarly, ?rms from indus-
tries that have high sensitivity to potential envir-
onmental legislation, usually categorized as the
petroleum, chemical, metals, and paper industries
(see, e.g. Cowen et al., 1987; Hackston & Milne,
1996; Patten, 1991), tend to make more extensive
disclosures than companies from less environmen-
tally sensitive industries. None of the previous
studies of the relation between environmental
performance and disclosure controlled for the
impact of company size, and Fekrat et al. (1996),
the only study that included companies from non-
environmentally sensitive industries, also did not
make any adjustment for potential industry
e?ects.
The second problem with the previous studies
centers on the sample selection. The studies based
on the CEP reports of the early 1970s (Ingram &
Frazier, 1980; Wiseman, 1982; Freedman & Was-
ley, 1990) were limited, of course, to companies
that had been evaluated by the CEP. Because
these included ?rms from only four industries the
resulting samples were not very diversi?ed. In
addition, because both Wiseman and Freedman
and Wasley made comparisons only within indus-
try, the sample sizes for the analyses were quite
small (industry samples ranged from seven to 21
companies). Similarly, Fekrat et al.’s (1996) sample
of 26 ?rms may not have been large enough for
meaningful analysis particularly given that the
sample consisted only of very large companies and
that no control for industry variation was included.
The ?nal problem with the previous studies of
the performance–disclosure relation centers on the
measure of environmental performance used. The
major shortcoming of the CEP ratings relied upon
by Ingram and Frazier (1980), Wiseman (1982),
and Freedman and Wasley (1990) is that di?erent
criteria were used to evaluate companies in di?er-
ent industries. Thus, inter-industry comparisons
really were not possible. Ingram and Frazier tried
to adjust for this by standardizing company scores
based on industry averages, but Wiseman, and
Freedman and Wasley performed only within-
industry analyses. The real problem with environ-
mental performance measurement is that until
passage of EPCRA in 1986 there was no large
scale database on individual company environ-
mental activities. EPCRA, however, requires large
manufacturing facilities from companies in the
2000 and 3000 SIC codes to ?le annual reports on
routine releases of more than 300 toxic chemicals.
The Environmental Protection Agency (EPA) is
required to make this information available to the
public. Among the information disclosures avail-
able from the EPA is a listing of the top 500
releasing ?rms which also identi?es the amount of
toxics released. Fekrat et al. (1996) did use a per-
formance measure based on toxics release inven-
tory (TRI) information. However, their ranking of
the data, particularly in light of the small sample
used, may have led to a loss of discriminating
information.
Given the signi?cant shortcomings of the pre-
vious studies of the relation between environ-
mental performance and disclosure, it would
appear that a more meaningful analysis is possi-
ble. Because the toxic release data required
through EPCRA (1) represent releases of hazar-
dous chemicals into the environment reported by
the companies themselves, (2) are based on the
same measures for all reporting companies, and
(3) are available for a large, diversi?ed sample of
?rms, they would appear to be a rich measure of
environmental performance for comparison with
environmental disclosure.
4
3. Development of hypothesis
Although a number of sources cite Financial
Accounting Standards Board (FASB) Statement
No. 5 on contingencies as a basis for requiring
environmental disclosure in the United States (see,
e.g. Barth & McNichols, 1994; Siegel & Surma,
1992; Zuber & Berry, 1992), the FASB has not, in
3
The signi?cant relation between ?rm size and the extent of
environmental disclosure is consistent with ?ndings from the
more general voluntary disclosure literature (see, e.g. King,
Pownall, & Waymire, 1990; Lang & Lundholm, 1993; Skinner,
1994). Lang and Lundholm (1993, pp. 250–251) provide an
overview of the arguments presented in this literature for a size-
voluntary disclosure relation.
4
It should be noted, of course, that toxic emissions represent
only one facet of a company’s overall environmental performance.
766 D.M. Patten / Accounting, Organizations and Society 27 (2002) 763–773
fact, issued any statements of accounting stan-
dards that apply to environmental disclosure.
5
With the exception of environmental-related con-
tingencies, therefore, annual report environmental
disclosures are discretionary in nature. As such, it
is not surprising that numerous studies (e.g. Ernst
& Ernst, 1978; Gamble et al., 1995; Patten, 1991,
1992; Walden & Schwartz, 1997) document that
U.S. annual report environmental disclosure var-
ies substantially across ?rms and across time.
Similar variation in environmental disclosure has
been noted for companies in Australia (Deegan &
Gordon, 1996), New Zealand (Hackston & Milne,
1996), and the United Kingdom (Gray et al.,
1995).
Given the widespread variation in environ-
mental (and other social) disclosure, it is perhaps
not surprising that a number of theories of social
disclosure have arisen. Gray et al. (1995, p. 52)
argue that the social-political theories (political
economy, legitimacy theory, stakeholder theory)
have resulted in the ‘‘most penetrating analyses of
[corporate social disclosure].’’ In general, these
overlapping theories (see Gray et al., 1996) suggest
that social disclosure is a function of social and/or
political pressure. More speci?cally in relation to
legitimacy theory, Patten (1991, p. 299; 1992, p.
472), argues that while the economic legitimacy of
corporations is monitored through the market-
place, social legitimacy is monitored through the
public policy process. If companies suspect their
social legitimacy is, or might be, threatened, they
have an incentive to actively participate in the
policy process. One method of participation is
?nancial report disclosure. Lindblom (1994) sug-
gests that a corporation may use disclosure as a
legitimizing device to (1) educate and inform
relevant publics about (actual) changes in its per-
formance, (2) change perceptions about the per-
formance of the organization, (3) de?ect attention
from the issue of concern by highlighting other
accomplishments related to the social issue, or (4)
seek to change public expectations of its perfor-
mance (see also, Gray et al., 1995, p. 54).
Previous studies (e.g. Cowen et al., 1987; Dee-
gan & Gordon, 1996; Hackston & Milne, 1996;
Patten, 1992) suggest that ?rm size and industry
classi?cation are both factors that can lead to
greater potential public pressure relative to envir-
onmental concerns. These studies also document a
signi?cant relation between these factors and the
extent of environmental disclosure. It is argued
here that companies with relatively poorer envir-
onmental performance, as measured by the TRI
data, would also be subject to greater exposure to
potential public and regulatory scrutiny than ?rms
with relatively better performance. Wolf (1996, p.
286), for example, notes that environmental
organizations have used TRI data to produce
‘‘reports publicizing the names of top polluting
facilities, industries, . . . and states in an e?ort to
invite public and regulatory pressure for toxic
substance reductions.’’ Konar and Cohen (1997,
p. 112) suggest that private parties (including
environmental groups) may rely on the data to
bring law suits against companies, ‘‘green con-
sumers’’ may boycott high polluting ?rms, and
government agencies might use the data to target
?rms for enforcement actions. Thus, according to
the legitimacy-based arguments, if poorer envir-
onmental performance does increase the threat to
a ?rm’s social legitimacy, it is hypothesized that
companies with higher amounts of toxics
released, adjusted for ?rm size, will provide
greater environmental disclosure in their annual
reports.
4. Research method
4.1. Sample selection
Sample selection is based on the following
criteria:
1. Companies must be included on the 1988
TRI listing of the top 500 ?rms in terms of
releases into the environment.
5
The FASB, through its Emerging Issues Task Force, has
issued two publications that relate to environmental issues.
EITF Issue No. 90-8 addresses the expensing or capitalization
of contamination treatment costs and EITF Issue No. 93-5
provides guidelines for recognizing environmental liabilities
and the impact of potential recoveries on related losses. No
statements of accounting standards addressing environmental
reporting have been promulgated by the Board.
D.M. Patten / Accounting, Organizations and Society 27 (2002) 763–773 767
2. Companies must have ?nancial data avail-
able for the test period on COMPUSTAT.
3. Companies must have provided a 1990
annual report for review.
Only 243 of the companies included on the 1988
TRI top 500 releasing ?rms listing have data
available on COMPUSTAT. I mailed a request
for 1990 ?nancial reports to these companies, and
received 131 annual reports.
The sample ?rms range in size (based on 1990
revenue levels) from $71.9 million to $124,705
million. The mean (median) size of the sample
companies is $6847.0 million ($2520.0 million). The
sample includes companies from 24 di?erent two-
digit primary Standard Industrial Classi?cations
(SIC).
6
The highest industry concentration is 19
?rms from the 2911 (petroleum) classi?cation.
Summary data on the sample ?rms are provided in
Table 1.
4.2. Environmental disclosure
Similar to Wiseman (1982), the extent of envir-
onmental disclosure was identi?ed using both
content analysis and a line count. More speci?-
cally, two independent reviewers examined the
annual reports for the presence or absence of
statements related to eight di?erent aspects of
environmental concern. The content analysis
areas, presented in the Appendix, were determined
through review of previous studies using content
analysis to determine the extent of environmental
disclosure (Blacconiere & Patten, 1994; Deegan &
Gordon, 1996; Freedman & Wasley, 1990; Wise-
man, 1982).
7
However, unlike the previous ana-
lyses, litigation-related disclosures were not
included. This is because, as noted by Blacconiere
and Patten (1994, p. 371), litigation disclosures
tend to be less discretionary than other environ-
mental disclosures and, by their nature, also tend
to provide information that re?ects poorly on
environmental stewardship. Since this study is
interested in the use of environmental disclosure as a
legitimizing activity, excluding the less discretionary,
more negative disclosures seems appropriate.
8
For the content analysis score, one point was
awarded for each area of disclosure included in the
?nancial report. Thus, environmental scores could
range from zero to 8. Line counts were determined
by summing the number of lines of print allocated
to the environmental disclosure areas in the annual
reports. All evaluation and coding discrepancies
across reviewers were discussed and reconciled.
9
Table 1
Descriptive statistics
n (sample size) 131
Firm size (1990 revenues)
Mean $6847.0 million
Median 2520.0 million
Standard deviation 15,349.1 million
1988 Toxic releases/1988 revenues
Mean 8.510
Median 3.382
Standard deviation 16.136
Firms from environmentally
sensitive industries (chemical,
metals, paper, petroleum)
63
Environmental disclosure
Content analysis score (max=8)
Mean 2.481
Standard deviation 2.088
Line count
Mean 26.924
Standard deviation 38.643
Toxic release data is from the 1988 Top 500 Releasing Com-
panies listing provided by the Environmental Protection
Agency (EPA, 1990). Environmental disclosure content analy-
sis scores and line counts are based upon review of the sample
companies’ annual reports from 1990. Content analysis areas
are identi?ed in the Appendix.
6
EPCRA requires any facility employing ten or more
workers that manufactures in the 20–39 SIC codes to report
emissions data (see Wolf, 1996, p. 231). Thus, some companies
whose primary SIC code is not in the 20–39 classi?cations are
included in the sample.
7
Several alternative content analysis schemes were also tes-
ted. These included models including as few as ?ve disclosure
areas (see Blacconiere & Patten, 1994) and as many as 21 dis-
closure areas (adapted from Deegan & Gordon, 1996). Results
using all alternative content analysis schemes were consistent
with the results presented in this paper.
8
All tests were repeated using disclosure scores and line
counts including litigation-related disclosures. Results, not
presented here, were consistent with the results for the tests
using the limited disclosure codings. Thus, the anomolous
?ndings reported in this study do not appear to be due to the
exclusion of the litigation disclosures.
768 D.M. Patten / Accounting, Organizations and Society 27 (2002) 763–773
Annual report environmental disclosure scores
ranged from zero (17 companies) to eight (one
company) with a mean score of 2.481. The number
of annual report lines allocated to environmental
disclosure ranged from zero to 181. The mean envir-
onmental disclosure line count was 26.924 lines.
4.3. Environmental performance
Environmental performance is quanti?ed as the
company-speci?c amount of toxics released into the
environment, as reported in the 1988 TRI listing of
the top 500 companies,
10
divided by the company’s
1988 revenue level. The adjustment for company size
is made because, ceteris paribus, larger ?rms would
be expected to have higher releases than smaller
?rms.
11
The size-adjusted release levels ranged from
0.158 to 105.921. The mean (median) environmental
performance measure for the 131 sample ?rms was
8.510 (3.382).
12
4.4. Control variables
As noted earlier, a number of previous studies
(e.g. Cowen et al., 1987; Deegan & Gordon; 1996;
Gray et al., 1995; Hackston & Milne, 1996; Pat-
ten, 1992) document a signi?cant relation between
company size and the extent of environmental
disclosure. Similarly, several studies (e.g. Cowen et
al., 1987; Deegan & Gordon, 1996; Walden &
Schwartz, 1997) show that ?rms from industries
with higher sensitivity to environmental regulation
tend to provide more extensive environmental
disclosure than ?rms from less environmentally
sensitive industries. In examining for a relation
between environmental performance and environ-
mental disclosure, therefore, it is necessary to con-
trol for the impact of size and industry membership.
The ?rm size measure used in this analysis is the
natural log of 1990 revenues.
13
Similar to Cowen
et al. (1987) and Patten (1991), the chemical (SIC
code 28xx, excluding pharmaceutical ?rms, code
283x), metals (33xx), paper (26xx), and petroleum
(2911) industries are de?ned as more environmen-
tally sensitive. A one/zero indicator variable is
used to designate companies from these industries.
Classi?cation was based on primary SIC code. In
total, 63 of the 131 sample companies were from
the environmentally sensitive industries.
Preliminary analyses indicated potential inter-
action e?ects.
14
Accordingly, two interaction vari-
ables are also included in the regression model.
The ?rst interaction variable is designed to capture
potential di?erences in the impact of the TRI vari-
able across larger and smaller ?rms. First, a one/zero
classi?action variable was used to designate those
companies with log revenue levels of 9.5 or higher.
15
The interactive e?ect was calculated by taking this
variable times the TRI level (designated Big*TRI).
The second interactive variable was calculated as
9
Inter-rater product–moment correlations for the environ-
mental disclosure measures were 0.92 for the content analysis
scores and 0.96 for the environmental disclosure line counts. Both
are statistically signi?cant at P
Previous studies of the relation between environmental performance and environmental disclosure have consistently
documented a lack of significance. This study examines the relation between 1990 annual report environmental disclosures
for a sample of 131 US companies and their environmental performance as based on toxics release data from
1988 (made available in 1990). In contrast to the previous examinations, results indicate that, controlling for firm size
and industry classification (two factors previously shown to be related to the extent of environmental disclosure), there
is a significant negative relation between performance and disclosure for the sample firms.
The relation between environmental performance and
environmental disclosure: a research note
Dennis M. Patten*
Department of Accounting, College of Business, Illinois State University,
Stevenson Hall, Campus Box 5520, Normal, IL 61761, USA
Abstract
Previous studies of the relation between environmental performance and environmental disclosure have consistently
documented a lack of signi?cance. This study examines the relation between 1990 annual report environmental dis-
closures for a sample of 131 US companies and their environmental performance as based on toxics release data from
1988 (made available in 1990). In contrast to the previous examinations, results indicate that, controlling for ?rm size
and industry classi?cation (two factors previously shown to be related to the extent of environmental disclosure), there
is a signi?cant negative relation between performance and disclosure for the sample ?rms. However, the disclosure level
of ?rms from non-environmentally sensitive industries is more a?ected by toxic release levels than is the disclosure of
?rms from environmentally sensitive industries. # 2002 Elsevier Science Ltd. All rights reserved.
Previous studies of the relation between corpo-
rate environmental performance and environ-
mental disclosure in ?nancial reports consistently
document a lack of signi?cance. This is troubling
in that socio-political theories of social disclosure
(see, e.g. Gray, Kouhy, & Lavers, 1995; Guthrie &
Parker, 1990; Hackston & Milne, 1996; Lindblom,
1994; Patten, 1991, 1992, 1995) suggest that the
extent of this disclosure is a function of exposure
to public pressure in the social/political environ-
ment.
1
According to these theories, companies
facing greater exposures, as companies with
poorer environmental performance could be
assumed to do, would be expected to provide more
extensive environmental disclosures, and as such,
0361-3682/02/$ - see front matter # 2002 Elsevier Science Ltd. All rights reserved.
PI I : S0361- 3682( 02) 00028- 4
Accounting, Organizations and Society 27 (2002) 763–773
www.elsevier.com/locate/aos
* Corresponding author. Tel.: +1-309-438-7857; fax: +1-
309-438-8431.
E-mail address: [email protected] (D.M. Patten).
1
Public pressure in the social/political environment is iden-
ti?ed by Walden and Schwartz (1997, p. 127) as consisting of
both social changes and regulatory e?ects. And while the social
changes are in turn seen as a function of what Boulding (1978,
p. 88) refers to as the cultural and political environments, reg-
ulatory e?ects result from the legal environment. Thus, public
pressure can arise due to concerns of the general population,
political bodies, or regulatory (or legal) agencies (see Boulding,
1978, p. 88).
a negative association between performance and
disclosure is posited. This study suggests that the
lack of evidence to date for a relation between
environmental performance and environmental
disclosure may be a function of problems in the
previous analyses. These problems include failure to
control for other factors in?uencing the extent of
disclosure, inadequate sample selection, and an
inadequate measure of environmental performance.
This study examines the relation between the
1990 environmental disclosures included in a
sample of 131 US companies’ annual reports and
?rm environmental performance as based on data
provided by the Environmental Protection Agen-
cy’s 1988 Toxics Release Inventory (released in
1990).
2
The analysis documents that, controlling
for ?rm size and industry classi?cation, two fac-
tors consistently shown to be related to the extent
of environmental disclosure, there is a signi?cant,
negative relation between performance and dis-
closure. This ?nding is robust to measures of dis-
closure based on both content analysis and
?nancial report line count. However, the impact of
the TRI release level on environmental disclosure
is not consistent across sample ?rms. The level of
disclosure for companies from less environmen-
tally sensitive industries is a?ected more by the
TRI release level than is the extent of disclosure
for ?rms from more environmentally sensitive
industries.
In general, it appears that in contrast to the
?ndings of previous studies, there is some relation
between environmental performance and environ-
mental disclosure. However, environmental per-
formance, as based on toxic release information, is
a more signi?cant explanatory factor for varia-
tions in annual report environmental disclosure
for ?rms not already subject to higher potential
environmental regulatory pressure due to industry
involvement. The paper begins with a summary of
previous analyses of the relation between environ-
mental performance and environmental disclosure.
1. Previous studies
Three of the four published studies examining
the relation between environmental performance
and ?nancial report environmental disclosure
relied on indices of environmental performance
issued by the Council on Economic Priorities
(CEP) in the early to mid 1970s. The CEP, a non-
pro?t organization specializing in the analysis of
corporate social activities, attempted to measure
the environmental performance of 50 ?rms from
four highly polluting industries (steel, oil, paper,
and electric utility). The CEP published reports
rating the companies on a 0 (best) to 10 (worst)
scale, and also published within-industry rankings
based on the organization’s analyses. Unfortu-
nately, evaluation criteria di?ered across indus-
tries resulting in reports that ‘‘were of varying
quality and depth’’ (Freedman & Wasley, 1990,
p. 185). However, they were at that time, accord-
ing to Wiseman (1982, p. 54), ‘‘the best credible
source available for objective environmental per-
formance measures on individual companies.’’
Ingram and Frazier (1980), the ?rst study to
examine the environmental performance–dis-
closure relation, used the CEP scores, standar-
dized using industry averages, as the measure of
environmental performance. Annual reports for
40 of the CEP-rated ?rms were analyzed for
environmental disclosures. The year of exami-
nation for each company was based on the year of
the corresponding CEP report and ranged from
1970 through 1974. Content analysis, a scoring
system awarding points based on the presence or
absence of information items, was used to measure
the extent of environmental disclosure. Neither
product–moment correlation nor partial corre-
lation controlling for the volume of disclosure
showed any signi?cant relations with environ-
mental performance. Similarly, multiple regression
analysis using independent variables obtained
from factor analysis of the disclosure items failed
to document any signi?cant relation to the level of
environmental performance.
2
A fundamental argument for this study is that environ-
mental performance will in?uence environmental disclosure
through increased public policy pressure related to perfor-
mance level. For this to hold, the performance information
must be publicly available. As such, it is necessary to use the
1990 annual report disclosures since this is the year the release
data from 1988 was made publicly available.
764 D.M. Patten / Accounting, Organizations and Society 27 (2002) 763–773
Wiseman (1982), the second study using CEP
data, relied on the within-industry rankings of
performance rather than the individual company
scores. Wiseman’s sample of 26 ?rms was drawn
from the oil, steel, and paper industries only.
Similar to Ingram and Frazier (1980), Wiseman
used content analysis to measure the extent of
environmental disclosure in company annual
reports. Wiseman also included a total disclosure
score and a line count measure. Rather than using
these raw measures, however, Wiseman used the
within-industry ranking based on the disclosure
scores. Industry speci?c comparisons were made
for steel ?rms for both 1972 and 1976 (the CEP
released separate rankings based on each of
these years), oil ?rms for 1974, and paper com-
panies for 1972. Of 24 comparisons computed
only one statistically signi?cant correlation was
found. Wiseman concluded (p. 61) that ‘‘overall,
there appears to be no relationship between the
measurable information companies disclose
about environmental performance in their
annual reports and their actual environmental
performance.’’
The last of the CEP-based analyses, Freedman
and Wasley (1990), operationalized environmental
performance and environmental disclosure in the
same manner as Wiseman (1982), although these
authors did use the raw disclosure scores for ana-
lysis rather than the within-industry rankings of
the scores as Wiseman did. In addition, Freedman
and Wasley examined a larger sample, 49 ?rms,
including companies from the electric utility
industry. Finally, they also analyzed 10K reports
in addition to just annual reports. Similar to
Wiseman, comparisons of performance and dis-
closure were made within industry only. Sig-
ni?cant correlations were reported for only two
of 50 comparisons, leading Freedman and Was-
ley to conclude that neither annual report nor
10K environmental disclosures ‘‘are indicative of
actual ?rm environmental performance’’ (p.
191).
The ?nal study of the relation between
environmental performance and environmental
disclosure, Fekrat, Inclan, and Petroni (1996),
is based on more recent disclosures (1991
annual reports) and on a di?erent measure of
environmental performance. Fekrat et al.
employed the content analysis scheme developed
by Wiseman (1982) to quantify the environmental
disclosures for a sample of 26 of the largest US
?rms, and then used the rankings based on these
scores as the disclosure index. Environmental per-
formance measures, obtained from the CEP, were
based not on CEP-based analyses as the 1970s
indices were, but on ?rm-reported toxic emissions
data required through the Emergency Planning
and Community Right-to-Know Act of 1986
(EPCRA). These ?rm speci?c data were made
available in 1990 for 1988 calendar year emissions.
The CEP reported the data adjusted for ?rm size,
and Fekrat et al. then ranked the companies based
on these adjusted measures. Correlations of the
rankings on performance and disclosure were
negative but not statistically signi?cant.
2. Problems with the previous analyses
The failure to ?nd a signi?cant relation between
environmental performance and environmental
disclosure in any of the studies to date appears to
present rather convincing evidence that no such
relation exists. However, there are problems with
each of the previous studies that are substantial
enough to bring the ?ndings into question. The
problems relate to (1) failure to control for other
factors associated with the level of environmental
disclosure; (2) inadequate sample selection; and (3)
an inadequate measure of environmental perfor-
mance. Each of these problems is discussed in
more detail below.
A number of studies over the past decade
attempted to identify factors associated with the
variation in environmental disclosure across com-
panies (see, e.g. Cowen, Ferreri, & Parker, 1987;
Deegan & Gordon, 1996; Gamble, Hsu, Kite, &
Radtke, 1995; Gray et al., 1995; Hackston &
Milne, 1996; Patten, 1992). These studies provide
a strong body of evidence that both company size
and industry classi?cation are factors that
appear to in?uence the level of environmental
information disclosed in ?nancial reports. Lar-
ger companies, presumably due to visibility
concerns, tend to disclose more information
D.M. Patten / Accounting, Organizations and Society 27 (2002) 763–773 765
than smaller ?rms.
3
Similarly, ?rms from indus-
tries that have high sensitivity to potential envir-
onmental legislation, usually categorized as the
petroleum, chemical, metals, and paper industries
(see, e.g. Cowen et al., 1987; Hackston & Milne,
1996; Patten, 1991), tend to make more extensive
disclosures than companies from less environmen-
tally sensitive industries. None of the previous
studies of the relation between environmental
performance and disclosure controlled for the
impact of company size, and Fekrat et al. (1996),
the only study that included companies from non-
environmentally sensitive industries, also did not
make any adjustment for potential industry
e?ects.
The second problem with the previous studies
centers on the sample selection. The studies based
on the CEP reports of the early 1970s (Ingram &
Frazier, 1980; Wiseman, 1982; Freedman & Was-
ley, 1990) were limited, of course, to companies
that had been evaluated by the CEP. Because
these included ?rms from only four industries the
resulting samples were not very diversi?ed. In
addition, because both Wiseman and Freedman
and Wasley made comparisons only within indus-
try, the sample sizes for the analyses were quite
small (industry samples ranged from seven to 21
companies). Similarly, Fekrat et al.’s (1996) sample
of 26 ?rms may not have been large enough for
meaningful analysis particularly given that the
sample consisted only of very large companies and
that no control for industry variation was included.
The ?nal problem with the previous studies of
the performance–disclosure relation centers on the
measure of environmental performance used. The
major shortcoming of the CEP ratings relied upon
by Ingram and Frazier (1980), Wiseman (1982),
and Freedman and Wasley (1990) is that di?erent
criteria were used to evaluate companies in di?er-
ent industries. Thus, inter-industry comparisons
really were not possible. Ingram and Frazier tried
to adjust for this by standardizing company scores
based on industry averages, but Wiseman, and
Freedman and Wasley performed only within-
industry analyses. The real problem with environ-
mental performance measurement is that until
passage of EPCRA in 1986 there was no large
scale database on individual company environ-
mental activities. EPCRA, however, requires large
manufacturing facilities from companies in the
2000 and 3000 SIC codes to ?le annual reports on
routine releases of more than 300 toxic chemicals.
The Environmental Protection Agency (EPA) is
required to make this information available to the
public. Among the information disclosures avail-
able from the EPA is a listing of the top 500
releasing ?rms which also identi?es the amount of
toxics released. Fekrat et al. (1996) did use a per-
formance measure based on toxics release inven-
tory (TRI) information. However, their ranking of
the data, particularly in light of the small sample
used, may have led to a loss of discriminating
information.
Given the signi?cant shortcomings of the pre-
vious studies of the relation between environ-
mental performance and disclosure, it would
appear that a more meaningful analysis is possi-
ble. Because the toxic release data required
through EPCRA (1) represent releases of hazar-
dous chemicals into the environment reported by
the companies themselves, (2) are based on the
same measures for all reporting companies, and
(3) are available for a large, diversi?ed sample of
?rms, they would appear to be a rich measure of
environmental performance for comparison with
environmental disclosure.
4
3. Development of hypothesis
Although a number of sources cite Financial
Accounting Standards Board (FASB) Statement
No. 5 on contingencies as a basis for requiring
environmental disclosure in the United States (see,
e.g. Barth & McNichols, 1994; Siegel & Surma,
1992; Zuber & Berry, 1992), the FASB has not, in
3
The signi?cant relation between ?rm size and the extent of
environmental disclosure is consistent with ?ndings from the
more general voluntary disclosure literature (see, e.g. King,
Pownall, & Waymire, 1990; Lang & Lundholm, 1993; Skinner,
1994). Lang and Lundholm (1993, pp. 250–251) provide an
overview of the arguments presented in this literature for a size-
voluntary disclosure relation.
4
It should be noted, of course, that toxic emissions represent
only one facet of a company’s overall environmental performance.
766 D.M. Patten / Accounting, Organizations and Society 27 (2002) 763–773
fact, issued any statements of accounting stan-
dards that apply to environmental disclosure.
5
With the exception of environmental-related con-
tingencies, therefore, annual report environmental
disclosures are discretionary in nature. As such, it
is not surprising that numerous studies (e.g. Ernst
& Ernst, 1978; Gamble et al., 1995; Patten, 1991,
1992; Walden & Schwartz, 1997) document that
U.S. annual report environmental disclosure var-
ies substantially across ?rms and across time.
Similar variation in environmental disclosure has
been noted for companies in Australia (Deegan &
Gordon, 1996), New Zealand (Hackston & Milne,
1996), and the United Kingdom (Gray et al.,
1995).
Given the widespread variation in environ-
mental (and other social) disclosure, it is perhaps
not surprising that a number of theories of social
disclosure have arisen. Gray et al. (1995, p. 52)
argue that the social-political theories (political
economy, legitimacy theory, stakeholder theory)
have resulted in the ‘‘most penetrating analyses of
[corporate social disclosure].’’ In general, these
overlapping theories (see Gray et al., 1996) suggest
that social disclosure is a function of social and/or
political pressure. More speci?cally in relation to
legitimacy theory, Patten (1991, p. 299; 1992, p.
472), argues that while the economic legitimacy of
corporations is monitored through the market-
place, social legitimacy is monitored through the
public policy process. If companies suspect their
social legitimacy is, or might be, threatened, they
have an incentive to actively participate in the
policy process. One method of participation is
?nancial report disclosure. Lindblom (1994) sug-
gests that a corporation may use disclosure as a
legitimizing device to (1) educate and inform
relevant publics about (actual) changes in its per-
formance, (2) change perceptions about the per-
formance of the organization, (3) de?ect attention
from the issue of concern by highlighting other
accomplishments related to the social issue, or (4)
seek to change public expectations of its perfor-
mance (see also, Gray et al., 1995, p. 54).
Previous studies (e.g. Cowen et al., 1987; Dee-
gan & Gordon, 1996; Hackston & Milne, 1996;
Patten, 1992) suggest that ?rm size and industry
classi?cation are both factors that can lead to
greater potential public pressure relative to envir-
onmental concerns. These studies also document a
signi?cant relation between these factors and the
extent of environmental disclosure. It is argued
here that companies with relatively poorer envir-
onmental performance, as measured by the TRI
data, would also be subject to greater exposure to
potential public and regulatory scrutiny than ?rms
with relatively better performance. Wolf (1996, p.
286), for example, notes that environmental
organizations have used TRI data to produce
‘‘reports publicizing the names of top polluting
facilities, industries, . . . and states in an e?ort to
invite public and regulatory pressure for toxic
substance reductions.’’ Konar and Cohen (1997,
p. 112) suggest that private parties (including
environmental groups) may rely on the data to
bring law suits against companies, ‘‘green con-
sumers’’ may boycott high polluting ?rms, and
government agencies might use the data to target
?rms for enforcement actions. Thus, according to
the legitimacy-based arguments, if poorer envir-
onmental performance does increase the threat to
a ?rm’s social legitimacy, it is hypothesized that
companies with higher amounts of toxics
released, adjusted for ?rm size, will provide
greater environmental disclosure in their annual
reports.
4. Research method
4.1. Sample selection
Sample selection is based on the following
criteria:
1. Companies must be included on the 1988
TRI listing of the top 500 ?rms in terms of
releases into the environment.
5
The FASB, through its Emerging Issues Task Force, has
issued two publications that relate to environmental issues.
EITF Issue No. 90-8 addresses the expensing or capitalization
of contamination treatment costs and EITF Issue No. 93-5
provides guidelines for recognizing environmental liabilities
and the impact of potential recoveries on related losses. No
statements of accounting standards addressing environmental
reporting have been promulgated by the Board.
D.M. Patten / Accounting, Organizations and Society 27 (2002) 763–773 767
2. Companies must have ?nancial data avail-
able for the test period on COMPUSTAT.
3. Companies must have provided a 1990
annual report for review.
Only 243 of the companies included on the 1988
TRI top 500 releasing ?rms listing have data
available on COMPUSTAT. I mailed a request
for 1990 ?nancial reports to these companies, and
received 131 annual reports.
The sample ?rms range in size (based on 1990
revenue levels) from $71.9 million to $124,705
million. The mean (median) size of the sample
companies is $6847.0 million ($2520.0 million). The
sample includes companies from 24 di?erent two-
digit primary Standard Industrial Classi?cations
(SIC).
6
The highest industry concentration is 19
?rms from the 2911 (petroleum) classi?cation.
Summary data on the sample ?rms are provided in
Table 1.
4.2. Environmental disclosure
Similar to Wiseman (1982), the extent of envir-
onmental disclosure was identi?ed using both
content analysis and a line count. More speci?-
cally, two independent reviewers examined the
annual reports for the presence or absence of
statements related to eight di?erent aspects of
environmental concern. The content analysis
areas, presented in the Appendix, were determined
through review of previous studies using content
analysis to determine the extent of environmental
disclosure (Blacconiere & Patten, 1994; Deegan &
Gordon, 1996; Freedman & Wasley, 1990; Wise-
man, 1982).
7
However, unlike the previous ana-
lyses, litigation-related disclosures were not
included. This is because, as noted by Blacconiere
and Patten (1994, p. 371), litigation disclosures
tend to be less discretionary than other environ-
mental disclosures and, by their nature, also tend
to provide information that re?ects poorly on
environmental stewardship. Since this study is
interested in the use of environmental disclosure as a
legitimizing activity, excluding the less discretionary,
more negative disclosures seems appropriate.
8
For the content analysis score, one point was
awarded for each area of disclosure included in the
?nancial report. Thus, environmental scores could
range from zero to 8. Line counts were determined
by summing the number of lines of print allocated
to the environmental disclosure areas in the annual
reports. All evaluation and coding discrepancies
across reviewers were discussed and reconciled.
9
Table 1
Descriptive statistics
n (sample size) 131
Firm size (1990 revenues)
Mean $6847.0 million
Median 2520.0 million
Standard deviation 15,349.1 million
1988 Toxic releases/1988 revenues
Mean 8.510
Median 3.382
Standard deviation 16.136
Firms from environmentally
sensitive industries (chemical,
metals, paper, petroleum)
63
Environmental disclosure
Content analysis score (max=8)
Mean 2.481
Standard deviation 2.088
Line count
Mean 26.924
Standard deviation 38.643
Toxic release data is from the 1988 Top 500 Releasing Com-
panies listing provided by the Environmental Protection
Agency (EPA, 1990). Environmental disclosure content analy-
sis scores and line counts are based upon review of the sample
companies’ annual reports from 1990. Content analysis areas
are identi?ed in the Appendix.
6
EPCRA requires any facility employing ten or more
workers that manufactures in the 20–39 SIC codes to report
emissions data (see Wolf, 1996, p. 231). Thus, some companies
whose primary SIC code is not in the 20–39 classi?cations are
included in the sample.
7
Several alternative content analysis schemes were also tes-
ted. These included models including as few as ?ve disclosure
areas (see Blacconiere & Patten, 1994) and as many as 21 dis-
closure areas (adapted from Deegan & Gordon, 1996). Results
using all alternative content analysis schemes were consistent
with the results presented in this paper.
8
All tests were repeated using disclosure scores and line
counts including litigation-related disclosures. Results, not
presented here, were consistent with the results for the tests
using the limited disclosure codings. Thus, the anomolous
?ndings reported in this study do not appear to be due to the
exclusion of the litigation disclosures.
768 D.M. Patten / Accounting, Organizations and Society 27 (2002) 763–773
Annual report environmental disclosure scores
ranged from zero (17 companies) to eight (one
company) with a mean score of 2.481. The number
of annual report lines allocated to environmental
disclosure ranged from zero to 181. The mean envir-
onmental disclosure line count was 26.924 lines.
4.3. Environmental performance
Environmental performance is quanti?ed as the
company-speci?c amount of toxics released into the
environment, as reported in the 1988 TRI listing of
the top 500 companies,
10
divided by the company’s
1988 revenue level. The adjustment for company size
is made because, ceteris paribus, larger ?rms would
be expected to have higher releases than smaller
?rms.
11
The size-adjusted release levels ranged from
0.158 to 105.921. The mean (median) environmental
performance measure for the 131 sample ?rms was
8.510 (3.382).
12
4.4. Control variables
As noted earlier, a number of previous studies
(e.g. Cowen et al., 1987; Deegan & Gordon; 1996;
Gray et al., 1995; Hackston & Milne, 1996; Pat-
ten, 1992) document a signi?cant relation between
company size and the extent of environmental
disclosure. Similarly, several studies (e.g. Cowen et
al., 1987; Deegan & Gordon, 1996; Walden &
Schwartz, 1997) show that ?rms from industries
with higher sensitivity to environmental regulation
tend to provide more extensive environmental
disclosure than ?rms from less environmentally
sensitive industries. In examining for a relation
between environmental performance and environ-
mental disclosure, therefore, it is necessary to con-
trol for the impact of size and industry membership.
The ?rm size measure used in this analysis is the
natural log of 1990 revenues.
13
Similar to Cowen
et al. (1987) and Patten (1991), the chemical (SIC
code 28xx, excluding pharmaceutical ?rms, code
283x), metals (33xx), paper (26xx), and petroleum
(2911) industries are de?ned as more environmen-
tally sensitive. A one/zero indicator variable is
used to designate companies from these industries.
Classi?cation was based on primary SIC code. In
total, 63 of the 131 sample companies were from
the environmentally sensitive industries.
Preliminary analyses indicated potential inter-
action e?ects.
14
Accordingly, two interaction vari-
ables are also included in the regression model.
The ?rst interaction variable is designed to capture
potential di?erences in the impact of the TRI vari-
able across larger and smaller ?rms. First, a one/zero
classi?action variable was used to designate those
companies with log revenue levels of 9.5 or higher.
15
The interactive e?ect was calculated by taking this
variable times the TRI level (designated Big*TRI).
The second interactive variable was calculated as
9
Inter-rater product–moment correlations for the environ-
mental disclosure measures were 0.92 for the content analysis
scores and 0.96 for the environmental disclosure line counts. Both
are statistically signi?cant at P