The relations among environmental disclosure, environmental performance, and economic performance: a simultaneous equations approach

Accounting, Organizations and Society 29 (2004) 447–471 www.elsevier.com/locate/aos

The relations among environmental disclosure, environmental performance, and economic performance: a simultaneous equations approach Sulaiman A. Al-Tuwaijria, Theodore E. Christensenb,*, K.E. Hughes IIc a

Department of Accounting and Management Information Systems, College of Industrial Management, King Fahd University of Petroleum and Minerals, KFUPM Box 1996, Dhahran 31261, Saudi Arabia b School of Accountancy and Information Systems, Marriott School of Management, Brigham Young University, 540 TNRB, Provo, UT 84602-3100, USA c Department of Accounting, E. J. Ourso College of Business Administration, Louisiana State University, Baton Rouge, LA 70803-6304, USA

Abstract This study provides an integrated analysis of the interrelations among (1) environmental disclosure, (2) environmental performance, and (3) economic performance. Based on the argument that management’s (unobservable) overall strategy affects each of these corporate responsibilities, we conjecture that prior literature’s mixed results describing their interrelations may be attributable to the fact that researchers have not considered these functions to be jointly determined. After endogenizing these corporate functions in simultaneous equations models, we obtain results that suggest ‘‘good’’ environmental performance is significantly associated with ‘‘good’’ economic performance, and also with more extensive quantifiable environmental disclosures of specific pollution measures and occurrences. # 2003 Elsevier Ltd. All rights reserved.

Introduction As managers scramble to compete in the global economy, they must do so within societal constraints characterized by ever-increasing environmental accountability. This accountability includes heightened public scrutiny of both the firm’s environmental performance and its public disclosure of that performance. These elements of corporate environmental accountability jointly impact the firm’s profitability and the value of its * Corresponding author. Tel.: +1-801-422-1768; fax: +801422-0621. E-mail address: [email protected] (T.E. Christensen).

common equity. This study provides an integrated analysis of how management’s overall strategy jointly affects (1) environmental disclosure, (2) environmental performance, and (3) economic performance. Understanding these interrelations is of increasing interest to both internal and external stakeholders in an era in which corporate environmental costs have become a significant business expense.1 1

Whereas using a landfill to dump hazardous waste cost only $2.50 per ton in 1978, this charge rose to over $200 per ton by 1987 (Buchholtz, Marcus, & Post, 1992). Between 1972 and 1992, total annualized environmental protection costs for US firms tripled as a percentage of Gross Domestic Product (GDP). Senior executives anticipate this trend to continue (Walley & Whitehead, 1994).

0361-3682/03/$ - see front matter # 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0361-3682(03)00032-1

448

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

Prior empirical and social-responsibility research on the relations among environmental performance, environmental disclosure, and economic performance has, in general, considered the strength of pair-wise associations between two of these three factors, while not addressing the third. Academic researchers from management, finance, and economic disciplines have focused on the environmental-performance–economic-performance relation and the root question: Is going green good for profits? Meanwhile, accounting researchers have concentrated on the adequacy of environmental disclosure in financial reporting and its value relevance to investors: Do green disclosures adequately represent the firm’s exposure to future green regulation? And because these disclosures are largely voluntary, what factors determine their shade of green? While this research has advanced our knowledge in specific settings,2 no study has attempted to examine environmental performance, environmental disclosure, and economic performance within a single inclusive model. We propose a holistic approach to examine collectively the relations among the firm’s (1) environmental performance, (2) environmental disclosure, and (3) economic performance, using a conceptual framework first suggested by Ullmann (1985). Ullmann presents a descriptive analysis of prior social-responsibility studies that, in aggregate, report mixed empirical results of pair-wise associations between environmental performance and economic performance, between environmental performance and environmental disclosure, and between environmental disclosure and economic performance.3 Ullmann posits that the inconsistent findings characteristic of these pair-wise studies reflect a common omitted variable—an inclusive management strategy. In executing the corporation’s strategic business plan, management implements policies and initiates decisions that simultaneously affect the firm’s

2 For example, Barth, McNichols, and Wilson (1997) examine the determinants of environmental disclosure for firms with Superfund liabilities. 3 Although Ullmann examined social-responsibility studies in general, approximately half (14 of 31) of these studies focused on environmental concerns.

environmental performance, environmental disclosure, and economic performance. If these corporate functions are endogenously determined, then piecemeal Ordinary Least Squares (OLS) estimation of pair-wise relations among these three functions will produce biased and inconsistent results. While accepting Ullmann’s premise that earlier research models used to explore these relations may have been mis-specified, we consider incorporating an unobservable managerial strategy into an empirical model to be problematic. Instead, we implement Ullmann’s conceptual framework by explicitly treating environmental performance, environmental disclosure, and economic performance as endogenous variables, jointly determined by the firm’s strategic management process. In doing so, we advance the following research questions: First, how are the firm’s environmental performance, environmental disclosure, and economic performance interrelated after the endogeneity of these three corporate functions is explicitly considered? Second, does joint estimation of these relations significantly differ from independent OLS estimation? If so, significant methodological differences in estimating the coefficients of the endogenous variables may be due to bias in the OLS estimator. Documentation of such bias has implications for both interpreting prior research and planning future research designs. To address these questions, we first specify environmental performance, environmental disclosure, and economic performance in three multivariate equations in which at least one of these functions is an explanatory variable of another. While our empirical proxies for economic performance are market-based and our measure for environmental performance is a nonfinancial ratio based on the relative quantity of hazardous waste recycled, we feel that it is important to qualify our measure of environmental disclosure and distinguish it from its more generic connotation. Within the context of this study, environmental disclosure is the disclosure of specific pollution measures and occurrences (toxic waste emissions, oil spills, Superfund sites, etc.) that an investor might find useful in estimating future cash flows. This definitional constraint focuses on the disclosure of cost

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

drivers of future environmental costs and intentionally excludes the ‘‘greenwash’’ commonly found in annual financial reports. Using a crosssectional sample of 198 US ‘‘Standard & Poors 500’’ firms, we first ensure that all sample firms exceed a minimum threshold for exposure to future environmental costs. We then compare independent OLS estimation of the relations among the three corporate functions under investigation, with joint estimation using two-stage least squares (2SLS) and three-stage least squares (3SLS) simultaneous equations models. After controlling for endogeneity, we observe that ‘‘good’’ environmental performance is positively associated with ‘‘good’’ economic performance, and also with more extensive quantifiable environmental disclosures of specific pollution measures and occurrences. This research contributes to our understanding of how societal concerns for the environment affect corporate strategy and, ultimately, firm value. First, we recognize the endogeneity of the firm’s environmental performance, environmental disclosure, and economic performance, and find that this research-design consideration significantly affects the statistical significance of estimated interrelations. Second, the significantly positive relation observed between environmental performance and economic performance suggests that managers should change their strategic outlook regarding a firm’s environmental performance, from fixating on the deadweight costs of ex post regulatory compliance, to focusing on the ex ante opportunity costs represented by environmental pollution. Third, we find that good environmental performers disclose (within the context of our definition of environmental disclosure) more pollution-related environmental information than do poor performers, which is consistent with discretionary disclosure theory’s ‘‘good news’’ explanation. Fourth, while providing additional evidence regarding the determinants of environmental performance, environmental disclosure, and economic performance, this research introduces new empirical proxies for environmental performance and environmental disclosure. Finally, this study spans the research agendas of multiple academic disciplines and contributes to a

449

growing body of interdisciplinary environmental knowledge. The remainder of this study proceeds as follows: the following reviews the pertinent literature and frames the testable hypotheses; the next introduces the structural equations in the simultaneous equations models and describes the variables along with the data sources; next, provides the empirical results; and the final section summarizes the study’s conclusions, implications, and limitations.

Literature review and hypotheses development The relation between environmental performance and economic performance Prior empirical research on the relation between environmental performance and economic performance has reported mixed results. Bragdon and Marlin (1972) argued that pollution abatement and profitability are compatible, and found a positive relation between profitability (earnings per share and return on equity) and the Counsel on Economic Priorities’ (CEP’s) environmental performance ratings for pulp-and-paper firms.4 Spicer (1978) used firms in the pulp-and-paper industry to measure the association between five firm-specific variables—profitability, size, total risk, systematic risk, and the price-earnings ratio—and the CEP’s pollution performance ratings. His results indicated that all signs were in the directions hypothesized; however, only the correlation coefficients for size, systematic risk, and the price-earnings ratio were statistically significant. Although both of these studies suffered from relatively low power due to small sample sizes and

4 Early US environmental studies have almost universally relied on environmental performance data compiled by the Council on Economic Priorities (CEP), a nonprofit corporation organized in 1970. Its purpose was to foster socially responsible business policies and practices. Environmental reports, of varying quality and depth, were produced by the CEP between 1970 and 1977. Indices were formulated to measure the environmental performance of firms from four highly polluting industries: steel, oil, electric utilities, and paper and pulp (Abbott & Monsen, 1979).

450

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

measurement error, their findings were consistent with the idea that good environmental and economic performance are complements. Adherents to this concept of complementary association believe that acting on a firm’s social responsibilities (or externalities) reduces risk to which the capital markets are increasingly sensitive (Narver, 1971). Additionally, if environmental pollution represents resources that have been inefficiently or incompletely used by the firm, the elimination of such waste and inefficiencies benefits both the environment and the bottom line (Porter & van der Linde, 1995a, 1995b). Although these early empirical studies suggested a positive environmental-performance–economicperformance relation, later researchers have generally found the association to be statistically insignificant. Rockness, Schlachter, and Rockness (1986) examined hazardous waste disposal in the chemical industry using environmental-performance data from a special site survey submitted to the US Congress in 1979. Testing the association among two waste disposal variables and 12 financial indicators representing economic performance, Rockness et al. failed to document a statistically significant relation. Freedman and Jaggi (1992) examined the long-term relation between environmental performance and economic performance, using the percentage change in three pollution measures and various accounting ratios as empirical proxies for environmental performance and economic performance, respectively. Again, Freedman and Jaggi’s results failed to reject the null hypothesis of no significant association. Perhaps this observed lack of significance resulted from a negative association between the variables of interest that effectively countered any positive association suggested by earlier studies. An inverse relation between environmental and economic performance is consistent with traditional economic thought that depicts this relation as a tradeoff between the firm’s profitability and acting on its social responsibility (Friedman, 1962). In summary, the relation between environmental performance and economic performance is founded on contradictory theoretical support that prior empirical research has failed to clarify.

The relation between environmental disclosure and environmental performance Establishing a relation between environmental performance and environmental disclosure is important from a social responsibility perspective in that a positive relation tends to validate the credibility of the latter. However, empirical research on the environmental-disclosure–environmental-performance relation has generally found no significant association between the two. Ingram and Frazier (1980) compared content analysis ratings of environmental disclosures that appeared in corporate annual reports to CEP environmental-performance ratings. They did not find a significant association between environmental disclosure and environmental performance. Freedman and Jaggi (1982) also reported insignificant results using disclosures made in Forms 10-K. Using a different method to evaluate environmental disclosures in annual reports, Wiseman (1982) again found no significant association between environmental disclosure and environmental performance. Freedman and Wasley (1990) used Wiseman’s method to evaluate environmental disclosures appearing in 10-Ks as well as annual reports. Again they observed no significant association between environmental disclosure and environmental performance as measured by CEP performance ratings. In addition to these archival studies, Rockness (1985) conducted a field experiment in which financial analysts, members of environmental protection organizations, environmental regulators, and MBA students evaluated environmental disclosures contained in annual reports. Rockness, then compared these subjects’ evaluations of environmental disclosures to CEP environmentalperformance ratings, and reported negative correlation coefficients, implying that subjects evaluate the worst environmental performance as best, and vice versa. As a result, Rockness suggested that subjects might have been misled by the disclosures. Li, Richardson, and Thornton (1997) arrived at a somewhat similar result in providing empirical support for their game-theory model of environmental disclosure. Using a sample of Canadian firms, they found a significantly positive relation

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

between a firm’s decision to disclose and its propensity to pollute. This implied a negative relation between environmental disclosure and environmental performance. More recently, Hughes, Anderson, and Golden (2001) observed that poorer US environmental performers tended to make the most disclosures, consistent with their responsibility to report contingent liabilities under SFAS 5 (FASB, 1975). However, although these disclosures differed between groups, Hughes et al. did not find them to be useful in classifying the firms’ actual environmental performance. A negative environmental-performance–environmental-disclosure relation appears to be inconsistent with Verrecchia’s (1983) discretionary disclosure model.5 If we assume that good environmental performance reduces the firm’s exposure to future environmental costs, then disclosure of this information should be perceived as good news by investors. Therefore, firms with good environmental performance should disclose more environmental information (in quantity and quality) than should firms with poorer environmental performance. On the other hand, if greater disclosure provides information that may be used in litigation against the disclosing firm (presumably by third parties with political or social agendas), good environmental performers might elect to minimize such disclosure (Li et al., 1997). A negative environmental-performance–environmentaldisclosure relation is also consistent with increased disclosure from poor environmental performers 5 The lattitude afforded US firms to make environmental disclosures under SFAS 5 (Accounting for Contingencies) led researchers to classify such disclosures as largely voluntary (Barth et al., 1997). 6 We examine environmental disclosure as a function of environmental performance. Gray, Jarad, Power, and Sinclair (2001) examine environmental disclosure as a function of turnover, capital employed, profit, and the number of employees. For their sample of UK firms during an eight-year sample period, they were unable to document any ‘‘unique and/or stable relationship between any measure of disclosure and any corporate characteristic’’ (Gray et al., 2001, p. 349). However, their disclosure variable is significantly different from our disclosure metric. 7 Partitioning their sample by firm size into quartiles, Freedman and Jaggi observed a significantly negative correlation between environmental disclosure and economic indicators for the top quartile.

451

that strictly comply with SFAS 5. In summary, prior research has not found a consistently significant association between environmental performance and environmental disclosure.6 The relation between environmental disclosure and economic performance Prior research on the environmental-disclosure– economic-performance relation has used both market-based and accounting-based measures of economic performance. Freedman and Jaggi (1982) tested the association of their measurements of environmental disclosure against six accounting ratios used to measure economic performance. They found insufficient statistical significance to reject the null hypothesis of no association.7 However, Shane and Spicer (1983) used an event study design and documented a negative market reaction during the two days preceding the release of CEP environmental reports. Similarly, Stevens (1984) reported that a portfolio of firms that disclosed higher estimated future pollution-abatement costs experienced monthly returns consistently lower than did a similar portfolio of firms that disclosed lower estimates of future environmental costs.8 More recently, Richardson and Welker (2001) observed that social disclosure (which subsumes environmental disclosure) behaved differently than general financial disclosure in tests of association with the firm’s cost of capital. These researchers

8 Event studies have also established the informational relevance of significant environmental events to investors while documenting their intra-industry effects. Bowen, Castanias, and Daley (1983) examined the electric utility industry after the Three Mile Island nuclear accident; Blacconiere and Patten (1994) examined the chemical industry in the wake of the Bhopal chemical leak. Both studies documented a significantly negative intra-industry effect. Ironically, Patten and Nance (1998) found a positive intra-industry effect following the 1989 grounding of the Exxon Valdez, as the spill triggered substantial price increases in the wholesale and retail gasoline markets. Additionally, Blacconiere and Patten (1994) observed that firms with more extensive environmental disclosures experienced less of a negative market reaction to the Bhopal disaster.

452

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

reported a significantly negative relation between the level of financial disclosure and the cost of capital (consistent with prior research; Botosan, 1997), an association primarily driven by firms in less informationally rich environments. Conversely, Richardson and Welker (2001) found the relation between social disclosure and cost of capital to be significantly positive, with more profitable firms being penalized more for their social disclosures. Assuming an inverse relation between cost of capital and share price as suggested by the dividend discount model, Richardson et al.’s findings imply that increased social disclosure is associated with lower share prices. This evidence is not consistent with the notion that discretionary disclosure reduces asymmetrical information costs or that increased social disclosure triggers a significantly favorable investor preference effect.9 Hypotheses Prior empirical research examining pair-wise associations among environmental performance, environmental disclosure, and economic performance has often been based on contradictory theoretical support. Additionally, prior empirical studies have often drawn on relatively small samples because of the limited availability of environmental performance information. All of the variables of interest are measured with error, especially environmental performance and environmental disclosure. The resultant low-powered tests have produced mixed results. Given the absence of unambiguous theoretical or empirical support for predicting the interrelations among these three corporate functions, we test the following nondirectional null hypotheses:

9 An example of this effect is that as of 1997, some 13 US electric utilities had adopted some form of green pricing under which the customer is asked to pay a premium of up to 15% of the normal bill. In return, the utility acquires renewable energy sources according to a set formula. Additionally, surveys consistently reveal that from 56 to 80% of respondents are willing to pay more for environmentally friendly energy sources (Tietenberg, 1998).

H1. Economic performance is not associated with environmental performance. H2. Environmental disclosure is not associated with environmental performance. H3. Economic performance is not associated with environmental disclosure.

The empirical model and sample selection Management’s grand strategy Ullmann (1985) reviewed prior empirical studies that investigated the interrelations among social performance, social disclosure, and economic performance. He noted that the ambiguous results reported in aggregate by this body of research might be due to incomplete specification of the empirical models measuring the statistical significance of pair-wise associations. Ullmann conjectured that the ubiquitous omitted variable was management’s grand strategy, and reasoned that the firm’s social performance, social disclosure, and economic performance were jointly determined by this common missing element. While recognizing that there have been several environmental studies examining one or more of these constructs since 1985 (e.g., Barth et al., 1997; Hughes, 2000; Hughes et al., 2001; Richardson & Welker, 2001), we believe that the problem described in Ullmann’s meta-analysis nevertheless persists. This body of empirical environmental research may have produced mixed results by failing to recognize the potential for endogenous relations among these three constructs. We follow Ullmann in suggesting that management’s overall strategy affects economic performance, environmental performance, and environmental disclosure. While we cannot directly represent a firm’s unobservable strategy, we can accommodate the joint determination process by estimating the relations among these constructs using a system of simultaneous equations defined in the following structural form:

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

453

Endogenous variables

used measure is the Council on Economic Priorities’ (CEP) company rating charts, which are published quarterly. Corporate environmental performance ranking provides insight into thirteen environmental issues that affect US corporations. Although prior researchers have used these ratings to measure corporate environmental performance, we elect not to use this measure, for the following reasons: (1) the unavailability of the CEP’s formulae to determine their environmental-performance rating limits its interpretation; (2) weights assigned to different environmental factors are not constant across industries, and we use an interindustry, cross-sectional sample; and (3) environmental reporting (disclosure) is one of the factors used to determine the CEP rating of a corporation’s environmental performance. In contrast to studies that rely on qualitative rankings to measure environmental performance, this study employs a quantitative measure: the ratio of toxic waste recycled to total toxic waste generated (ENVPERF). For example, if a firm introduces a pollution-abatement process, decreasing the total amount of toxic waste generated, the denominator decreases and thus ENVPERF increases. Or if the firm adopts processes that recycle toxic waste (such as closed-loop cooling systems), the numerator increases, again increasing ENVPERF. Therefore as a rule, the higher the ratio of recycled waste to total waste, the better the firm’s environmental performance. This environmental performance measure also incorporates the first three principles of good environmental performance as promulgated by the Coalition for Environmentally Responsible Economies (CERES): minimize pollutants, conserve resources, and reduce waste. Also, this particular measure is sufficiently generic to be used by virtually all polluting industries, avoiding the problem of relying on industry-specific pollution measures in our cross-sectional, inter-industry, research design. We obtain recycling ratio data from the Corporate Environmental Profiles Directory, which is published annually (since

Environmental performance (ENVPERF) Prior research has measured environmental performance in several ways.10 The most frequently

10 See Ilinitch, Soderstrom, and Thomas (1998) for a discussion of current problems in measuring and reporting environmental performance information.

Economic Performance ¼ f Environmental Performance and
predetermined variables

ð1:1Þ

Environmental Performance ¼ f Economic Performance and
predetermined variables

ð1:2Þ

Environmental Disclosure ¼ f Environmental Performance and
predetermined variables

ð1:3Þ

The empirical model Having defined the theoretical model, we propose the following structural equations as an empirical model to test the study’s hypotheses: ECONPERF ¼ 0 þ 1 ENVPERF þ 2 UE þ 3 PREDISC þ 4 GROWTH þ 5 MARGIN þ 6 ENVEXP þ 1 ð2:1Þ

ENVPERF ¼ 0 þ 1 ECONPERF þ 2 PREDISC þ 3 GROWTH þ 4 ENVEXP þ 5 ENVCON þ 6 VISIBILITY þ 2 ð2:2Þ

ENVDISCL ¼ 0 þ 1 ENVPERF þ 2 ENVEXP þ 3 ENVCON þ 4 SIZE þ 3 ð2:3Þ We define the variables included in these models below.

454

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

1992) by the Investor Responsibility Research Center (IRRC).11 Environmental disclosure (ENVDISCL) Environmental disclosure measurement techniques can be classified into two general groups. The first group includes measures that quantify the level of environmental disclosure in the annual report, such as the number of pages (Gray, Kouhy, & Lavers, 1995; Guthrie & Parker, 1989; Patten, 1992, 1995), sentences (Frazier, 1982; Ingram & Wiseman, 1980), and words (Deegan & Gordon, 1996; Zeghal & Ahmed, 1990). Each of these measures has its limitations. While pages may include pictures that have no information on environmental or social activities, sentences and words may ignore necessary graphs and tables. This form of measurement is also susceptible to ‘‘greenwashing,’’ in which management puts its best ‘‘spin’’ on what otherwise might be a lackluster environmental performance. The second measurement technique uses a disclosure-scoring measure derived from content analysis. Using this technique, researchers first identify certain environmental issues, then analyze the environmental disclosure of each issue using a ‘‘yes/no’’ (or 1, 0) scoring methodology. After individual issues are quantified, researchers determine the aggregate score for each firm.12 We adopt a similar disclosure-scoring methodology based on content analysis that incorporates disclosures of four key environmental indicators: (1) the total amount of toxic waste generated and transferred or recycled; (2) financial penalties resulting from violations of 10 federal environmental laws; (3) Potential Responsible Party

11 The IRRC publishes various types of information on S&P 500 firms that are deemed to be useful to institutional investors, corporations, and other users. The IRRC uses three major environmental information sources: government agencies, corporations, and the media. ‘‘Much of this information was obtained by filing Freedom of Information Act requests, often after extensive consultation with the numerous agencies involved’’ (Corporate Environmental Profiles Directory, 1995, p. 3). 12 For example, Barth et al. (1997) used a variation of this scoring technique.

(PRP) designation for the cleanup responsibility of hazardous-waste sites; and (4) the occurrence of reported oil and chemical spills. Thus our environmental disclosure measure is based on information reported in SEC Forms 10-K and focuses on pollution-related information in these four areas. This informational restriction allows a structured evaluation while capturing significant indicators of the firm’s environmental exposure to future environmental costs. We believe that quantitative disclosures are more objective and informative to stakeholders than qualitative information. Prior studies (e.g., Hughes et al., 2001) using quantitative disclosure measures have assigned weights to different disclosure items based on the perceived importance of each item to various user groups. Similarly, we assign the greatest weight (+3) to quantitative disclosures related to the four environmental indicators described above. We assign the next highest weight (+2) to non-quantitative but specific information related to these indicators. Finally, general qualitative disclosures receive the lowest weight (+1). Firms that do not disclose information for a given indicator receive a score of zero for that indicator. In constructing our disclosure measure, we realize that firms using production processes that are not associated with one of our polluting activities should not be penalized for failing to make environmental disclosures for that activity. We accommodate this adjustment by factoring the firm’s participation in a polluting process (as reported to regulators rather than to shareholders) into the denominator of our disclosure score. These data are obtained from the IRRC database (i.e., the firm’s pollution data reported to the EPA or some other agency). For example, if a firm is designated a PRP but does not disclose this information in its annual report, the numerator of the disclosure score would not be affected (0), while the denominator of the score would be increased (+1), thereby decreasing the firm’s overall disclosure score. This metric is the first environmental-disclosure measure, to our knowledge, that attempts to capture the ‘‘truth-telling’’ or transparency property of the firm’s environmental disclosure by measuring the firm’s disclosure to investors conditioned

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

455

Fig. 1. Example of disclosure scoring technique—ENVDISCL=quality score/occurrence score. Examples of scored disclosures: Qualitative: In addition, the Company has been identified as a potentially responsible party for investigation and cleanup costs at a number of locations in the United States and Puerto Rico under federal remediation laws and is voluntarily investigating potential contamination at a number of Company-owned locations (Abbot Laboratories, 1993). Qualitative specific: The Company had also been named in several groundwater contamination suits, nearly all related to its former agricultural pesticide, ethylene dibromide (EDB), which was banned by the US government in 1983. These cases, primarily in Florida, California, and the State of Washington, stem from uses of EDB approved by all governmental regulatory agencies prior to its ban. All suits remaining in Florida and the State of Washington were settled by 1991 in amounts not material to the Company’s financial results. The only significant remaining suit is in California, where the Company is one of many defendants named. The Company is actively defending itself in this suit and does not now believe that this case will result in material financial consequences for the Company (Great Lakes Chemical, 1991). Quantitative: Alcoa’s remediation reserve balance at the end of 1994 was $329 million and reflects the most probable costs to remediate identified environmental conditions for which costs can be reasonably estimated. About 28% of this balance relates to Alcoa’s Massena, New York plant site. Remediation costs charged to the reserve were $79 million in 1994, $71 million in 1993, and $102 million in 1992. They include expenditures currently mandated, as well as those not required by any regulatory authority or third parties (Aluminum Co. of America, 1994).

on its polluting activity reported to environmental regulators.13 The total quality score (minimum=0, maximum=+12) is summed for the four activities and then scaled by the total number of polluting activities associated with the firm’s production processes, the occurrence score (minimum=0, maximum =+4). Thus environmentaldisclosure scores (ENVDISCL) range from 0 to +3. Fig. 1 provides an illustrative example of an environmental-disclosure score calculation. Economic performance (ECONPERF) Prior environmental studies have used both accounting-based and market-based measures to represent economic performance. For example, Bragdon and Marlin (1972) used accountingbased measures (earnings per share and return on equity), while Spicer (1978) used both accountingbased and market-based measures (profitability and the price-earnings ratio). One limitation in

using various economic performance metrics is that they tend to focus narrowly on one aspect of a firms’ economic performance. Net income measures a firm’s profitability without considering firm size. This limitation can be addressed by using measures such as return on assets (ROA), and scaling profitability by the firm’s investment in their asset base. However, this measure may be biased if the sample (such as this study’s sample) includes firms from different industries with different industry-driven levels of fixed assets, and where there may be systematic differences across industries in the age of these assets. Given these limitations, and assuming (a semistrong form of) 13 Measuring environmental disclosure conditioned on the firm’s polluting activities as reported to the EPA is consistent with the SEC and EPA’s consideration of formalizing their currently informal exchange of information programs (Slavich, 1994).

456

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

market efficiency, we elect to use a market-based metric to represent economic performance. We measure the firm’s economic performance using an industry-adjusted annual return. We calculate this metric as the change in stock price during the year (adjusted for dividends), scaled by the beginning-of-year stock price minus the industry median return (based on two-digit SIC codes). While other market-derived measures, such as debt ratings and the cost of capital, may appear to be viable candidates for representing economic performance, they are also associated with increased subjectivity and measurement error. We believe that annual stock returns represent a more objective and comprehensive measure of economic performance. The latter is due to the proposition that stock price should impound information about the firm’s future prospects from a vast array of both financial and nonfinancial measures, such as net income, ROA, operational data, etc. The annual industry-adjusted stock return, ECONPERF, represents a comprehensive measure of the firm’s current-period economic performance relative to other firms in the same industry.14 We complement this ‘‘changes’’ (return) specification of ECONPERF in subsequent sensitivity testing in which a ‘‘levels’’ (share price) specification is used to represent economic performance. Predetermined variables Unexpected earnings (UE) Given our industry-adjusted-return measure of economic performance, we control for the unexpected portion of earnings as the annual change in earnings per share scaled by stock price at the beginning of the period (Christie, 1987). The association between returns and earnings is well documented in the earnings-response-coefficient (ERC) literature (e.g., Collins & Kothari, 1989). Predisclosure environment (PREDISC) A firm’s prior environmental disclosures may represent a lower bound for current environmental performance. Investors’ expectations of environ14

Using industry-adjusted returns also permits us to hold the firm’s investment opportunity set (IOS) constant.

mental performance are conditioned on information provided by prior environmental disclosures. If a firm’s current environmental performance deteriorates (without disclosure) and reduces stock price, then shareholders may have grounds for litigation. Also, management’s reputation for providing credible disclosures would suffer. We average ENVDISCL (the environmental-disclosure score) over the three most recent years to proxy for past environmental disclosure (PREDISC), and we predict a positive relation between PREDISC and ENVPERF. We also use PREDISC rather than ENVDISCL in examining the relation between environmental disclosure and economic performance. This is necessary because our proxy for economic performance—annual industry-adjusted return—is measured at the end of the fiscal year, while the ENVDISCL variable is derived from Forms 10-K financial disclosures that are released approximately three months later. Therefore it is not possible to observe a contemporaneous relation between these variables. PREDISC approximates the average level of disclosure in prior years (ENVDISCL). Growth opportunities (GROWTH) We use the ratio of market value of equity to book value of equity as a proxy for future growth opportunities (Gaver & Gaver, 1993; Smith & Watts, 1992) in the economic-performance equation. This ratio measures the difference between the market’s appraisal of firm value and the estimate of value aggregated from GAAP-mandated accounting transactions. For example, US firms conducting research and development (R&D) must expense this cost, whereas the market views (at least some) R&D costs as an investment yielding future benefits. Growth opportunities should be positively related to economic performance (ECONPERF). We also include GROWTH in the environmentalperformance equation as a proxy for intangible assets associated with innovation (Porter & van der Linde, 1995a). We expect GROWTH to be positively related to environmental performance. Profit margin (MARGIN) Profit margin, the ratio of net income to net sales, captures both profitability and the presence

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

of competitive markets. Firms may obtain a higher profit margin by increasing sales prices, decreasing costs, or doing both. In competitive markets, because firms have limited potential for increasing prices, cost control becomes of primary importance. We assume that the relatively large firms that comprise our sample operate in globally competitive markets. Therefore, higher profit margins signal better cost control, and this indictor should be positively associated with economic performance (ECONPERF).15

457

positively correlated with future environmental costs. We expect ENVEXP to be negatively related to economic performance, since future environmental costs should reduce future cash flows and hence firm value. We posit that environmental exposure is positively related to both environmental performance and environmental disclosure. Firms with greater exposures to future costs have greater incentives to perform well environmentally in order to assure stakeholders that their investments will not bear undue risk. Similarly, firms with greater ENVEXP are likely to disclose more environmental information, both to fulfill the disclosure requirements of SFAS 5 and to reduce information asymmetry that adversely affects their cost of capital (Richardson & Welker, 2001).

Environmental exposure (ENVEXP) Environmental exposure is defined as the firm’s exposure to future environmental costs. Since we predict that a firm’s environmental exposure will affect both environmental performance and environmental disclosure, we use it as a control variable in the environmental performance and environmental disclosure equations. In addition, environmental performance and environmental disclosure may affect economic performance if a firm has a significant level of exposure to environmental costs. Consequently we also include our environmental exposure measure as a control variable in the economic performance equation. Because the firm’s production processes determine environmental pollution, several researchers have employed industry classification (SIC code) as a proxy for environmental exposure (i.e., Brockhoff, 1979; Deegan & Gordon, 1996; Little, Muoghlu, & Robison, 1995; Niskala & Pretes, 1995; Wiseman, 1982). Given that SIC codes may inadequately describe a firm’s unique portfolio of operations, we again use a quantitative measure. We measure environmental exposure (ENVEXP) as the amount of toxic waste generated by the firm scaled by total revenues.16 The amount of toxic waste generated per dollar of sales provides an indication of the pollution intensity of the firm’s production processes, a measure we propose to be

Environmental concern (ENVCON) Including the firm’s environmental concern as a predetermined variable in the environmental performance and environmental disclosure equations is consistent with Ullmann’s (1985) conceptual emphasis on including management’s strategy in models examining the firm’s social responsibility. We operationalize this variable by using a factor analysis of three firm characteristics that capture its concern for the environment, thereby extracting a single composite measure representing the firm’s environmental concern (ENVCON).17 The first firm characteristic we consider, COMMITTEE, is an indicator variable that is coded one (1) if the firm has a corporate environmental committee or department charged with monitoring the environmental impact of the firm’s actions; otherwise it is coded as zero (0). The second characteristic, REPORT, measures the frequency with which the firm publishes a stand-alone environmental report (separate from the annual report). REPORT is coded 1.0 if the report is published annually, 0.5 if

15 However, because environmental issues drive some costs, we further isolate the profit-margin metric to capture the efficiency of the firm’s cost control efforts, given a specific level of environmental exposure. To do this, we condition profit margin on environmental exposure (ENVEXP). 16 We also use SIC codes in the sensitivity analysis.

17 Using factor analysis to construct the ENVCON variable is consistent with the concept of a latent variable that cannot be adequately described by a single metric. Also, this methodology reduces the potential for encountering multicollinearity problems among ENVCON’s correlated component variables.

458

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

the report is published biennially, 0.33 if the report is published triennially, and zero if no report is published. The third measure, PROGRAM, is a discrete variable that measures the number of EPA voluntary programs in which the firm participates.18 Because management’s decision to engage in these three activities is completely voluntary, these variables provide evidence of management’s awareness and concern for the environment. We predict that ENVCON, the primary factor derived from factor analysis of these variables, is positively related to both environmental performance and environmental disclosure.19 Public visibility (VISIBLTY) Firms exposed to greater public scrutiny are more likely to incur political costs associated with poor environmental performance. Consequently, we control for a firm’s public visibility by including the number of Wall Street Journal news announcements about the firm during the year (VISIBLTY) in the environmental-performance equation. We expect higher-visibility firms to have higher standards of environmental performance because of increased public scrutiny. Firm size (SIZE) We use market value of common equity (SIZE) as a control variable for firm size in the environmental-disclosure equation. Prior research (Atiase, 18 One example is the EPA’s ‘‘Green Lights’’ program, in which firms voluntarily scrutinize every path of electrical energy consumption. In return they receive EPA guidance on efficient lighting, heating, and cooling operations (Porter & van der Linde, 1995a). 19 The eigenvalue for the first principle component (2.0351) is more than three times larger than that of the second principal component (0.5256) and more than four times larger than that of the third principal component (0.4393). Furthermore, the first principal component explains 67.8% of the variation of the three environmental concern proxies. Finally, the communality estimates for the three environmental concern variables are REPORT 0.70, PROGRAM 0.69, COMMITTEE 0.64 and the correlations between the first principle component and the three environmental concern variables are as follows: REPORT 0.83, PROGRAM 0.83, COMMITTEE 0.80. Therefore, we conclude that the first principle component captures the core construct common to these variables—management’s environmental concern.

1985) has found that firm size proxies for the firm’s information environment, where informationally rich environments are associated with larger firms. Although we expect that larger firms have greater incentives to disclose environmental information, our restrictive measure for environmental disclosure (specific pollution information) may distort the measurement of this relation. Therefore, we do not predict the sign of the coefficient on the SIZE variable. Sample selection and data collection This study uses a cross-sectional research design and firm data for the year 1994. In order to be included in our sample, a firm must  be listed in the IRRC’s 1994 Environmental Profiles Directory, which is limited to Standard and Poor’s 500 companies;  generate at least one pound of toxic waste per $10,000 of revenue (in order to ensure some minimum level of environmental exposure);  have complete financial data reported in Compustat;  have its annual reports accessable using the LexisNexis database; and  appear in the Wall Street Journal Index. Of the 531 firms included in the 1994 IRRC Environmental Profiles Directory, 313 do not have sufficient environmental exposure to meet our second criterion. Four firms do not have complete data in the IRRC Directory, and 16 firms do not have complete Compustat data. The final sample includes 198 firms that meet all of the selection criteria.

Results Descriptive statistics Panel A of Table 1 provides descriptive statistics for the study’s dependent variables. The mean (median) industry-adjusted stock return, ECONPERF, for sample firms is 0.0076 (0.0064). The

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

459

Table 1 Descriptive statistics—cross-sectional data for 198 environmentally exposed firms Variables

Mean

Standard deviation

25th Percentile

Median

75th Percentile

Panel A: Dependent variables ECONPERF 0.0076 ENVPERF 0.74 ENVDISCL 0.67

0.3227 0.23 0.66

0.1166 0.62 0.00

0.0064 0.83 0.50

0.1091 0.93 1.00

Panel B: Predetermined variables PREDISC 0.47 ENVEXP 9.95 ENVCON 0.01 REPORT 0.26 PROGRAM 1.20 COMMITTEE 0.62 UE 0.0112 GROWTH 2.66 MARGIN 0.07 VISIBLTY 26.58 SIZE $8,534.31

0.54 24.61 1.00 0.41 1.12 0.49 0.0461 1.69 0.19 34.27 $13,568.52

0.00 0.53 1.19 0.00 0.00 0.00 0.0022 1.69 0.003 11.00 $1,781.56

0.17 2.06 0.02 0.00 1.00 1.00 0.0092 2.25 0.02 16.00 $3,374.25

0.67 9.24 0.86 0.50 2.00 1.00 0.0220 3.08 0.09 27.00 $7,842.94

ECONPERF=Industry-adjusted annual stock return. ENVPERF=Environmental performance measured as the percentage of total waste generated that is recycled. ENVDISCL=Environmental disclosure score obtained from content analysis of the firm’s annual report. PREDISC=Past environmental disclosure measured as the average ENVDISCL over the preceding three years. ENVEXP =Environmental exposure measured as toxic waste generated scaled by total revenues. ENVCON=Environmental concern measured as the primary factor obtained from factor analysis of REPORT, PROGRAM, AND COMMITTEE. REPORT=Indicator variable coded one if the firm publishes an annual environmental report separate from its annual report, 0.5 if the report is biennial, 0.33 if the report is triennial, and 0 if no report is published. PROGRAM=The number of voluntary EPA programs in which the firm participates. COMMITTEE=Indicator variable coded one if the firm has an environmental committee and zero otherwise. UE= Unexpected earnings measured as the annual change in earnings per share scaled by beginning of year stock price. MARGIN= Profit margin (net income/net sales). GROWTH=Market-to-book ratio of common equity. VISIBLTY=The number of Wall Street Journal news announcements about the firm. SIZE=Market value of common equity (in thousands of US dollars).

mean (median) percentage of waste recycled, ENVPERF, 74% (83%) implies that sample firms recycle most of their generated waste. Finally, the mean (median) disclosure score, ENVDISCL, 0.67 (0.50) suggests that on a scale of zero to three (where three represents quantitative disclosures of all significant environmental activities), sample firms, on average, disclose only qualitative information at best. Panel B of Table 1 reports descriptive statistics for the predetermined variables used in our system of equations. The environmental-disclosure score that represents the predisclosure environment averaged over the three prior years, PREDISC, (0.47) is significantly less (p < 0.01) than the current year’s (1994) environmental-disclosure score, ENVDISCL, (0.67).20 This result suggests that the higher 1994 environmental-disclosure level may have been triggered by the SEC’s release of Staff

Accounting Bulletin No. 92 in June 1993, and is also consistent with the findings of Barth et al. (1997).21 Environmental exposure, ENVEXP, is positively skewed as the mean, 9.95, and is higher than the 75th percentile, 9.24. This suggests that, on average, sample firms generate approximately ten pounds of waste for every one thousand dollars of sales generated. Because this variable is 20 Supplemental analysis reveals that, on average, the current level of environmental disclosure is 55% higher than the average of the past 3 years. 21 The SEC signaled a continuing interest in improving the accounting for and disclosure of environmental liabilities with the 1993 release of Staff Accounting Bulletin No. 92 (SAB 92). This bulletin (1) limited offsetting against insurance consistent with GAAP, (2) limited the discounting of future environmental liabilities while imposing a ceiling on the allowable discount rate, and (3) attempted to elicit more meaningful information concerning environmental matters.

460

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

highly skewed, we use a log-transformation of ENVEXP in the empirical model. The firm’s environmental concern, ENVCON, is the primary factor from a factor analysis of three different measures of management’s environmental concern. Examining these three measures, we first find that the frequency of separate environmental reports, REPORT, (0.26) implies that, on average, sample firms publish a distinct environmental report approximately every fourth year. The mean PROGRAM shows that the average sample firm voluntarily subscribes to 1.20 environmental EPA programs. Finally, the mean and median COMMITTEE (0.62 and 1.00, respectively) suggest that more than half of the sample firms have environmental committees charged with monitoring their environmental responsibilities. Panel B of Table 1 also reports that the financial determinants of a firm’s economic performance— unexpected earnings (UE), the market-to-book ratio (GROWTH), and profit margin (MARGIN)—are all slightly positively skewed, which is consistent with prior research. The variables representing public visibility (VISIBLTY) and firm size (SIZE) are also positively skewed, suggesting that many of the sample’s S&P 500 firms with significant environmental exposure are among the largest and most profitable public firms. Because VISIBLTY and SIZE are so highly skewed, we again employ log-transformations for these two variables in the empirical model. Table 2 depicts both the parametric and nonparametric, pair-wise correlation coefficients for all variables included in the system of equations. The three variables of interest (ECONPERF, ENVPERF, and ENVDISCL) are all positively correlated, but the strengths of these pair-wise linear relations are relatively weak (coefficients < 0.4). This observation is consistent with the mixed results experienced by prior researchers in evaluating these pair-wise relations. Of note, the relation between the two environmental disclosure variables, ENVDISCL and PREDISC approach perfect correlation, consistent with the proposition that a firm’s disclosure policy is relatively ‘‘sticky.’’ As these variables do not appear in the same structural equation, this high correlation

does not present an econometric problem. Similarly, the proxies for SIZE (market value of equity) and VISIBILITY (number of WSJ news announcements) are highly correlated but again do not appear in the same structural equation. Positive correlations of medium strength are noted between environmental performance (ENVPERF) and environmental exposure (ENVEXP), MARGIN and SIZE, and between MARGIN and GROWTH. Other pair-wise relations depicted in Table 2 are relatively weak. None of the depicted relations represent obvious anomalies to theoretical explanation. The endogeneity problem We posit that the mixed results of prior research may be attributable to the fact that managers’ overall strategies likely affect economic performance, environmental performance, and environmental disclosure simultaneously. As depicted in Table 2, our proxies for these constructs are all positively correlated. We then examine the appropriateness of using an OLS regression analysis to estimate this system of equations by employing a Hausman (1978) test. This statistical test should detect the presence of any endogenous relations among our three dependent variables—ECONPERF, ENVPERF, and ENVDISCL.22 Using this procedure, we reject the null hypothesis of no endogeneity with respect to ENVPERF in the third equation (t=2.910, p< 0.004). We therefore conclude that OLS estimators are potentially 22

This particular Hausman test involves a two-stage procedure. In the first stage, each dependent variable is regressed on all of the predetermined variables in the system, and predicted values for the dependent variable are calculated using the estimated coefficients from the first stage regressions. In the second stage, each dependent variable is regressed on the righthand-side dependent variables, the predicted values of the right-hand-side dependent variables, and the respective predetermined variables for that equation. The significance of each predicted right-hand-side dependent variable is then tested against zero using a T-test or an F-test with the null hypothesis of no endogeneity with respect to that variable. If a predicted dependent variable is determined to have significant explanatory power, the dependent variable is presumed to be endogenous. See Kennedy (1994, p. 169) for an intuitive explanation of this procedure.

Table 2 Correlation coefficients (P-values) for all variables included in the simultaneous equation model (Pearson coefficients in upper right and Spearman coefficients in lower left) ENVPERF

ENVDISCL

PREDISC

ENVEXP

ENVCON

UE

GROWTH

MARGIN

VISIBLTY

SIZE

ECONPERF

1.0000

ENVPERF

0.1850 (0.0057) 0.1771 (0.0082) 0.1771 (0.0082) 0.1879 (0.0050) 0.2283 (0.0006) 0.2592 (0.0001) 0.3426 (0.0001) 0.2502 (0.0002) 0.1944 (0.0036) 0.2184 (0.0011)

0.3166 (0.0001) 1.0000

0.2196 (0.0010) 0.3418 (0.0001) 1.0000

0.1868 (0.0052) 0.2965 (0.0001) 0.9861 (0.0001) 1.0000

0.2401 (0.0003) 0.5402 (0.0001) 0.3909 (0.0001) 0.3444 (0.0001) 1.0000

0.2392 (0.0003) 0.3897 (0.0001) 0.4589 (0.0001) 0.4285 (0.0001) 0.2525 (0.0001) 1.0000

0.4550 (0.0001) 0.1983 (0.0030) 0.1915 (0.0042) 0.1649 (0.0139) 0.3081 (0.0001) 0.1384 (0.0394) 1.0000

0.3658 (0.0001) 0.1571 (0.0192) 0.0128 (0.8494) 0.0003 (0.9970) 0.0678 (0.3146) 0.0381 (0.5720) 0.04476 (0.4801) 1.0000

0.4132 (0.0001) 0.3874 (0.0001) 0.0853 (0.2054) 0.0720 (0.2857) 0.2941 (0.0001) 0.1494 (0.0260) 0.2139 (0.0013) 0.5535 (0.0001) 1.0000

0.2290 (0.0006) 0.3476 (0.0001) 0.2209 (0.0009) 0.1794 (0.0074) 0.0577 (0.3926) 0.5256 (0.0001) 0.0616 (0.3607) 0.2860 (0.0001) 0.4079 (0.0001) 1.0000

0.2789 (0.0001) 0.4259 (0.0001) 0.2483 (0.0002) 0.2098 (0.0017) 0.2261 (0.0007) 0.5263 (0.0001) 0.0491 (0.4665) 0.4735 (0.0001) 0.5570 (0.0001) 0.8258 (0.0001) 1.0000

ENVDISCL PREDISC ENVEXP ENVCON UE GROWTH MARGIN VISIBLTY SIZE

0.3264 (0.0001) 0.3264 (0.0001) 0.4990 (0.0001) 0.3870 (0.0001) 0.0831 (0.2175) 0.2028 (0.0024) 0.1824 (0.0064) 0.2738 (0.0001) 0.3340 (0.0001)

1.0000 (0.0001) 0.4399 (0.0001) 0.5063 (0.0001) 0.1606 (0.0166) 0.0863 (0.2004) 0.0152 (0.8218) 0.2993 (0.0001) 0.2794 (0.0001)

0.4399 (0.0001) 0.5063 (0.0001) 0.1606 (0.0166) 0.0863 (0.2004) 0.0152 (0.8218) 0.2993 (0.0001) 0.2794 (0.0001)

0.3276 (0.0001) 0.2833 (0.0001) 0.1734 (0.0096) 0.1827 (0.0063) 0.1252 (0.0627) 0.2351 (0.0004)

0.1278 (0.0573) 0.2033 (0.0023) 0.1892 (0.0047) 0.5935 (0.0001) 0.5655 (0.0001)

0.0450 (0.5051) 0.0329 (0.6259) 0.0117 (0.8622) 0.0349 (0.6051)

0.6243 (0.0001) 0.3491 (0.0001) 0.5434 (0.0001)

0.3090 (0.0001) 0.5173 (0.0001)

0.8039 (0.0001)

ECONPERF=Industry-adjusted annual stock return. ENVPERF=Environmental performance measured as the percentage of total waste generated that is recycled. ENVDISCL=Environmental disclosure score obtained from content analysis of the firm’s annual report. PREDISC=Past environmental disclosure measured as the average ENVDISCL over the preceding three years. ENVEXP=Environmental exposure measured as toxic waste generated scaled by total revenues. ENVCON =Environmental concern measured as the primary factor obtained from factor analysis of REPORT, PROGRAM, AND COMMITTEE REPORT=Indicator variable coded one if the firm publishes an annual environmental report separate from its annual report, 0.5 if the report is biennial, 0.33 if the report is triennial and 0 if no report is published. PROGRAM=The number of voluntary EPA programs in which the firm participates. COMMITTEE=Indicator variable coded one if the firm has an environmental committee and zero otherwise. MARGIN=Profit margin (net income/net sales). GROWTH=Market-to-book ratio of common equity. VISIBLTY=The number of Wall Street Journal news announcements about the firm. SIZE=Market value of common equity (in thousands of US dollars).

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

ECONPERF

461

462

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

biased and inconsistent. After ensuring that each structural equation passes a White (1980) test for homogeneity and correct specification,23 we continue our analysis using two-stage least squares (2SLS) and three-stage least squares (3SLS) simultaneous equation models in order to control for endogeneity while obtaining asymptotically unbiased results. Because the results from the 2SLS and 3SLS models are largely similar, and because 3SLS parameter estimates are the more efficient of the two, we report only 3SLS results. Our examination of this system using the 3SLS model incorporates all the available information from all the equations in simultaneously estimating the parameters. Regression analysis (three-stage least squares) The results of our 3SLS simultaneous equation model examining the relations among economic performance, environmental performance, and environmental disclosure, as specified in structural Eqs. (2.1)–(2.3), appear in Table 3. The coefficients for the explanatory variables in Eq. (2.1), which specifies the determinants of economic performance, suggest a positive relation between economic performance, ECONPERF, and environmental performance, ENVPERF. This positive relation (P=0.0739) is consistent with Porter and van der Linde’s (1995a) win–win scenario and the proposition that good environmental performance is rewarded in the market. This result is also consistent with investors who view good environmental performance as an intangible asset. In further examining the results of Eq. (2.1), we find that both UE and GROWTH are significantly positively associated with economic performance, as predicted. We also find some evidence that MARGIN is positively associated with ECONPERF, and that our proxy for environmental exposure (ENVEXP) is insignificant. The results presented in Table 3 for Eq. (2.2) suggest that economic performance is not a sig23 White’s (1980) test fails to reject the null hypothesis of homoskedasticity and correct model specification for all equations at the 0.05 level. Furthermore, variance-inflation factors and collinearity diagnostics (Belsley, Kuh, & Welsch, 1980) are well within acceptable ranges.

nificant determinant of environmental performance. This result is not consistent with the economic argument that profitability drives good environmental performance and that environmental accountability is strictly a matter of affordability. Furthermore, the firm’s predisclosure environment (PREDISC) is positively related to ENVPERF, consistent with the level of disclosure in the prior periods serving as a lower bound for performance in the current period.24 In addition, the coefficients for both ENVEXP and VISIBLTY are significantly positive, as predicted. This implies that firms with greater environmental exposure and greater public visibility respond with higher environmental performance standards than other comparable firms. Coefficients for variables representing GROWTH and environmental concern, ENVCON, are statistically insignificant. The 3SLS results for Eq. (2.3) suggest that ENVPERF is significantly and positively associated with ENVDISCL. This is consistent with discretionary disclosure theory’s explanation that good environmental performers believe that disclosure of their performance represents ‘‘good news’’ to market participants. These firms can therefore be more forthright in disclosing indicators of environmental pollution. The positive relation between environmental performance and environmental disclosure is counter to that suggested by Li et al.’s (1997) gaming model. Additionally, this result is inconsistent with poor environmental performers that have potentially greater contingent environmental liabilities, disclosing these contingent liabilities as required by SFAS 5. This implies that poor performers either do not strictly follow the requirements of SFAS 5, or that SFAS 5’s requirements provide sufficient latitude (e.g., materiality judgments) for poor 24 Because of the high correlation between PREDISC and ENVDISCL as depicted in Table 2, an alternative explanation for this result is that the predisclosure environment variable is a proxy for contemporaneous environmental disclosure. If we substitute ENVDISCL for PREDISC in Eq. (2.2), we observe similar results to those reported in Table 3. We acknowledge the research design’s inability to disentangle these variables with regard to Eq. (2.2). Having said this, we elect to retain PREDISC, rather than ENVDISCL, as a determinant of environmental performance because we believe the theoretical support for PREDISC’s inclusion to be more persuasive.

463

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471 Table 3 Three-stage least squares regression results (economic performance=industry-adjusted returns) coefficients (t-statistics) Independent variable

Predicted sign

Dependent variable ECONPERF [Eq. (2.1)] 0.44b (3.24)

INTERCEPT ECONPERF ENVPERF UE

+

PREDISC

+

GROWTH

+

MARGIN

+

ENVEXP

/+/+

ENVCON

+

VISIBLTY

+

0.40a (1.80) 2.42b (6.53) 0.01 (0.22) 0.05b (3.47) 0.01a (1.54) 0.02 (1.24)

0.37b (6.05) 0.04 (0.42)

ENVDISCL [Eq. (2.3)] 0.32 (0.81)

2.70b (3.01)

0.26b (10.44) 0.01 (0.78)

0.06b (6.27) 0.01 (0.35) 0.06b (3.17)

0.09 (1.31) 0.10a (1.39)

0.11a (1.86)

SIZE System weighted R2

ENVPERF [Eq. (2.2)]

0.5483

ECONPERF=Industry-adjusted annual stock return. ENVPERF=Environmental performance measured as the percentage of total waste generated that is recycled. ENVDISCL=Environmental disclosure score obtained from content analysis of the firm’s annual report. UE=Annual change in earnings per share scaled by stock price at fiscal year end. PREDISC=Past environmental disclosure measured as the average ENVDISCL over the preceding three years. GROWTH=Market-to-book ratio of common equity. ENVEXP=Environmental exposure measured as the natural log of toxic waste generated scaled by total revenues. ENVCON =Environmental concern measured as the primary factor obtained from factor analysis of REPORT, PROGRAM, and COMMITTEE. REPORT=Indicator Variable coded one if the firm publishes an annual environmental report separate from its annual report, 0.5 if the report is biennial, 0.33 if the report is triennial, and 0 if no report is published. PROGRAM=The number of voluntary EPA programs in which the firm participates. COMMITTEE=Indicator variable coded one if the firm has an environmental committee and zero otherwise. MARGIN=Profit margin (net income / net sales) conditioned on environmental exposure. VISIBLTY=The natural log of the number of Wall Street Journal news announcements about the firm. SIZE=The natural log of market value of common equity. a Statistically significant at the 0.10 level (one-tailed test if the sign is predicted, otherwise two-tailed). b Statistically significant at the 0.01 level (one-tailed test if the sign is predicted, otherwise two-tailed).

performers to justify nondisclosure. On the surface, it appears inconsistent that good environmental performers would disclose more quantifiable environmental information regarding their polluting activities, which the market would then use to punish them by discounting their stock prices. However, such disclosure may be viewed as a signal that the firm is in fact a good environmental performer, and may be used by the market to help differentiate between the good and the

not-so-good. We also find marginal evidence that management’s level of environmental concern, ENVCON, is positively related to environmental disclosure. This suggests that managers who are concerned about the environment are more likely to disclose their environmental activities publicly. In sum, the 3SLS results suggest two significant relations among our dependent variables. First, environmental performance is positively related to economic performance, suggesting that the market

464

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

rewards firms with higher levels of environmental performance. Second, environmental performance is positively related to environmental disclosure, suggesting that good environmental performers are more forthright and disclose more with respect to quantifiable pollution information. We observe this last relation to be significant only after controlling for endogeneity (3SLS) and to be statistically insignificant using OLS estimation (P=0.37). This contrast in results addresses one of our research questions by providing evidence of the importance for controlling for endogeneity in this research setting, and validates our adoption of a joint-estimation research design. Sensitivity analysis Model specification using stock price as a proxy for economic performance In specifying our system of equations, we elected to use annual industry-adjusted stock return as our dependent variable representing economic performance. While this decision is justifiable, assuming efficient markets, researchers have often preferred to use a valuation approach by measuring the economic performance construct using market share price. We therefore provide a robustness test of our model using market price instead of annual industry-adjusted returns as our dependent variable in Eq. (2.1). We elect to make two minor changes in the control variables in order to reflect the change from returns to market price as a proxy for economic performance. First, since this new proxy for economic performance measures the level of performance as opposed to the change in performance, we substitute book value of common equity, BOOKVAL, for UE in Eq. (2.1). This acknowledges the theoretical linkage between market value and book value of equity, on which accounting valuation researchers have heavily focused (e.g., Barth & McNichols, 1994; Hughes, 2000) in attempting to measure environmental liabilities. We also add operating income, OPINC, which is intended to capture the present value of expected future abnormal earnings, as per Ohlson (1995). Finally, we substitute net sales for market value of equity as a control variable for firm size (SIZE) in Eq. (2.3), since we

could not include the same variable as both an endogenous and predetermined variable in the same system of equations. These are the only changes incorporated in the respecified system of equations. Table 4 presents results for the respecified system of equations using stock price to represent economic performance. The results are consistent with those observed using annual stock return as our economic performance proxy. In Eq. (2.1), we observe a significantly positive relation between environmental performance and economic performance, providing some assurance for inferences made based on weaker statistical results in our original returns-specified model. All other explanatory variables are significant and correctly signed, with two exceptions: our operating income variable (OPINC) is insignificant, and the coefficient on the variable representing the firm’s predisclosure environment is significantly negative. Because the latter variable was insignificant in our original returns model, we make no inferences regarding the ECONPERF–PREDISC relation. In Eq. (2.2), we observe results identical to those presented in our main analysis, except that the coefficient on VISIBLTY is no longer statistically significant. Finally, the results for Eq. (2.3) are very similar to those presented in Table 3, with one exception. While we still observe a highly significantly positive relation between ENVPERF and ENVDISCL, the coefficient on ENVCON is no longer statistically significant. In sum, the results of the respecified system of equations with regard to our variables of interest are similar, regardless of whether the economic performance variable is specified in a ‘‘levels’’ (share price) or in a ‘‘changes’’ (returns) format. This provides some assurance that our primary results are quite robust to economic performance measurement. Industry effects The study’s sample is composed of firms from 31 two-digit SIC codes. Given this sample size, it would be very inefficient to include an indicator variable for every industry to completely control for all industry differences. However, we isolate six two-digit SIC codes containing at least 14 firms each and comprising a total of 117 firms (SIC codes 26, 28, 35, 36, 37, and 38). The remaining

465

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471 Table 4 Three-stage least squares regression results (economic performance=market price per share) coefficients (t-statistics) Independent variable

Predicted sign

INTERCEPT

Dependent variable ECONPERF [Eq. (2.1)]

ENVPERF [Eq. (2.2)]

ENVDISCL [Eq. (2.3)]

103.79b (6.44)

0.58b (10.94) 0.00 (0.48)

2.95b (2.65)

ECONPERF ENVPERF BOOKVAL

+

OPINC

+

PREDISC

+

GROWTH

+

MARGIN

+

ENVEXP

/+/+

ENVCON

+

VISIBLTY

+

135.07b (6.49) 1.51b (15.91) 5.78 (0.58) 14.12 (7.41) 9.23b (10.25) 0.30a (1.99) 4.64b (4.01)

SIZE System weighted R2

5.72a (2.53)

0.22b (11.44) 0.00 (0.12)

0.03b (3.51) 0.00 (0.39) 0.01 (0.87)

0.17 (1.49) 0.03 (0.39)

0.06 (0.79) 0.4968

ECONPERF=Market stock price per share. ENVPERF=Environmental performance measured as the percentage of total waste generated that is recycled. ENVDISCL=Environmental-disclosure score obtained from content analysis of the firm’s annual report. BOOKVAL=Book value of common equity per share. OPINC=Operating income per share. PREDISC=Past environmental disclosure measured as the average ENVDISCL over the preceding three years. GROWTH=Market-to-book ratio of common equity. ENVEXP=Environmental exposure—the natural log of toxic waste generated scaled by total revenues. ENVCON=Environmental concern measured as the primary factor obtained from factor analysis of REPORT, PROGRAM, and COMMITTEE. REPORT =Indicator variable coded 1.0 if the firm publishes an annual environmental report separate from its annual report, 0.5 if the report is biennial, 0.33 if the report is triennial, and 0 otherwise. PROGRAM=The number of voluntary EPA programs in which the firm participates. COMMITTEE=Indicator variable coded one if the firm has an environmental committee and zero otherwise. MARGIN=Profit margin (net income/net sales) conditioned on environmental exposure. VISIBLTY=The natural log of the number of Wall Street Journal news announcements about the firm. SIZE=The natural log of net sales. a Statistically significant at the 0.05 level (one-tailed test if the sign is predicted, otherwise two-tailed). b Statistically significant at the 0.01 level (one-tailed test if the sign is predicted, otherwise two-tailed).

40% of our firms are distributed broadly throughout the other 25 two-digit SIC codes represented in our sample. Repeating our analysis, including indicator variables to control for differences across industries, yields virtually identical results. Only one of the industry control variables is significant in Eq. (2.1), while none is significant in the other two equations. This analysis suggests

that there may be systematic differences in economic performance across industries, but that there are no systematic differences in environmental performance or environmental disclosure related to industry-specific firm characteristics. This result may also be driven by our selection of a broad-based, generic metric for environmental performance.

466

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

Changes in environmental disclosure To further explore the environmental-performance–environmental-disclosure relation, we note that the SEC actively focused on increasing corporate environmental disclosure during the sample period (see note 21). The effect of this increased regulatory interest on environmental disclosure is noted in Table 1, as our measure for environmental disclosure significantly increased ( p < 0.01) between the prior periods represented by PREDISC (the disclosure score averaged over the 1991, 1992, and 1993 period) and the current period in which ENVDISCL is measured (1994). If the level of ENVDISCL of good performers is greater than that of poor performers for the current period, as the results suggest, then we assume that the same conditions existed in prior periods. If good performers disclose more than poor performers do, then we believe that the increase in environmental disclosure depicted in Table 1 must have been predominantly made by good environmental performers. We test this proposition by ranking firms by their environmental performance and measuring the pair-wise correlation between their rankings and the corresponding changes in environmental disclosure (ENVDISCLPREPREDISC). The correlation coefficient is significantly positive in both parametric (0.1985, p< 0.005) and nonparametric (0.2262, p < 0.001) tests. This result is consistent with good performers, on average, increasing their 1994 environmental disclosure levels over their average disclosure levels for the prior three years. If the increased SEC oversight of environmental disclosures during this period was intended to increase disclosures from poor environmental performers, these results suggest that it was not very effective.

Conclusions, limitations, and implications This study investigates the relations among economic performance, environmental performance, and environmental disclosure, after explicitly considering that these three corporate functions are jointly determined. This specification is consistent with Ullmann’s (1985) argument that the

execution of each of these corporate responsibilities is determined by management’s (unobservable) overall strategy. We, like Ullmann, conjecture that the mixed results reported by prior environmental empirical research may have arisen because researchers did not allow for these constructs to be endogeneous. By explicitly controlling for endogeneity in the study’s research design, this research contributes to our understanding of how societal concerns for the environment affect corporate strategy and, ultimately, firm value. A more detailed description of how this study contributes to the current stream of environmental research follows. First, we find that allowing for the potential endogeneity associated with specifying our three corporate functions—economic performance, environmental performance, and environmental disclosure—makes a statistically significant difference in estimating their interrelations. Finding our proxy for environmental performance to be endogeneous, we provide evidence of the bias associated with OLS estimation under such conditions by using a system of simultaneous equations. The OLS results suggest that only the economic-performance–environmental-performance relation of the potential interrelations among economic performance, environmental performance, and environmental disclosure is statistically significant in our cross-sectional sample. However, by using a joint-estimation research design, we also observe a significantly positive relation between good environmental performance and more extensive quantifiable disclosure of environmental information. The contrast between independently estimated OLS results and those obtained through joint determination highlight the importance of controlling for endogeneity in the research design. Second, the significantly positive relation observed between environmental performance and economic performance is consistent with Michael Porter’s theoretical argument that innovative solutions to reduce the inefficiencies associated with pollution promote both environmentalism and industrial competitiveness simultaneously. Porter and van der Linde (1995a, 1995b) reject the economic-ecological tradeoff paradigm because it assumes that everything except regulation

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

(i.e., technology, products, processes, and customer needs) is fixed.25 They argue that environmental pollution represents resources that have been used incompletely, inefficiently, or ineffectively. Our results, consistent with this argument, suggest that managers should change their strategic outlook regarding a firm’s environmental performance, from fixating on the deadweight costs of ex post regulatory compliance to focusing on the ex ante opportunity costs represented by environmental pollution.26 Our finding that good environmental performance and economic profitability go hand-inhand, in addition to supporting Porter’s theoretical advocacy of this relation, is also consistent with the view that economic performance and environmental performance are both related to the quality of management. Good managers, acting in the firm’s long-term interest, accept the firm’s social responsibility and adopt pro-active strategies for controlling environmental pollution. Finally, because we use market-based proxies to represent economic performance, the observed positive economic-performance–environmentalperformance relation is also consistent with investors’ preferences for equities of environmentally responsible firms. Socially responsible investing is becoming more popular, with over $1.5 trillion worldwide currently invested according to social or ethical criteria (Vogel, 2002).27

467

Third, we find that good environmental performers disclose (within the context of our definition of environmental disclosure) more pollutionrelated environmental information than do poor performers. This finding is consistent with discretionary disclosure theory’s ‘‘good news’’ explanation. This result is also consistent with firms using voluntary disclosure to project a proactive environmental image by providing candid information regarding their environmental performance, even though that information may be viewed as ‘‘negative’’ on a situational basis. This disclosure policy appears to be at odds with strategies that minimize environmental disclosures because market participants may perceive such information as ‘‘bad news.’’ These results suggest that firms with records of good environmental performance can be more forthright in disclosing that performance. The study also documents a positive relation between past environmental disclosure and current environmental performance. This association is consistent with the notion that prior disclosure establishes a lower bound for management’s environmental performance. Not achieving this lower bound might adversely challenge the expectations of market participants, and potentially trigger shareholder litigation. Fourth, this research introduces new empirical proxies for environmental performance and environmental disclosure. In contrast to studies that

25

Porter and van der Linde (1995a) cite the ‘‘false’’ tradeoff between quality and cost as a similar misconception, induced by the static manner in which the relation is framed. 26 This result is also consistent with a recent report (MCA, 2000) released by the Management Consultancies Association that examines the role of business in environmental issues. The report describes a ‘‘triple bottom line approach’’ that benefits those firms that adopt proactive environmental strategies. First, these firms see competitive advantage in attaining environmental regulations that tend to crowd out competitors that are accustomed to less stringent regulation. Second, these firms have exploited alternative technologies to fit in with society’s desire for a clean environment. Finally, these firms have accepted the responsibility to minimize and, they hope, neutralize the negative impacts of their polluting activities (MCA, 2000, pp. 56–57).

27 ‘‘Good’’ corporate social responsibility does not neccesssarily carry-over to ‘‘good’’ accounting practices, as evidenced by the scandals appearing in today’s business press. Infamous Enron lobbied the Bush administration to accept the Kyoto Protocol on global warming, no doubt with the expectation of profiting from trading in a market for carbon dioxide emissions. Merck, recently cited for misrepresenting revenues, received the prestegious Business Enterprise Trust award in 1991 for developing and distributing Mectizan, a drug effective against river blindness, which treatens millions of the Third World’s poor. And while Xerox has been lauded as an international leader in enviro-management for its program of recycling copy cartridges, the firm recently paid a $10 million fine to settle a civil suit filed by the US SEC for overstating profits (Vogel, 2002).

468

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

rely on qualitative rankings to measure environmental performance, this study adopts a quantitative measure: the ratio of toxic waste recycled to total toxic waste generated. Whereas this measure may have certain limitations as discussed below, this ratio also is sufficiently general to be a useful environmental performance measure across industries—a property that is essential in an interindustry, cross-sectional research design, such as ours. This recycling ratio measure might be viewed as a summary statistic, much like the earnings number. We also construct a new measurement of environmental disclosure reported to investors that is uniquely preconditioned on the firm’s polluting activities as reported to regulators. This disclosure metric not only controls for the degree of disclosure made by firms with limited polluting activities, but also extracts a ‘‘transparency’’ dimension for polluters that file mandated environmental information with regulators, yet do not disclose this information in financial reports. Finally, this study spans the research agendas of multiple academic disciplines and contributes to a growing body of interdisciplinary environmental knowledge. In motivating our research design, we call upon Ullmann’s (1985) meta-analysis of empirical studies that investigate the interrelations among social performance, social disclosure, and economic performance, which was published in the Academy of Management Review. Our results are also consistent with management guru Michael Porter’s win-win argument regarding the positive relation between environmental and economic performance, which is widely cited in both management and economic literature. The marketbased proxies used to represent the firm’s economic performance are common to both accounting and finance research. While environmental research questions may extend into the domains of several academic disciplines, researchers are reluctant to violate these academic boundaries. However, without such interdisciplinary trespassing, a holistic approach to investigating these interdisciplinary interrelations of interest would not be possible. Like all cross-sectional studies, limitations to interpreting our results apply regarding whether the time period examined is representative and the

observed relations among the variables of interest are relatively stable over time. We know of no external event limiting 1994 as a sample period, and believe two of our three dependent variables—environmental performance and environmental disclosure—do not change dramatically around this temporal datum. The stability of economic performance is more problematic. We address this limitation by using two specifications for this variable: a changes specification (industryadjusted returns) and a levels specification (share price). The cross-sectional design presents another limitation in our choice of a suitable proxy for environmental performance. Because pollution is determined by the production process, pollutionrelated measures of environmental performance tend to be industry-specific. For example, measures of air pollution (e.g., sulfur dioxide emissions) may be a relevant performance measure for the electric utility industry (Hughes, 2000), whereas measures of water pollution may be more relevant in measuring the environmental performance of firms in the pulp-and-paper industry (Cormier & Magnan, 1997). Because our measure of environmental performance (the ratio of toxic waste recycled to total toxic waste generated) is one of the few metrics suitable for inter-industry comparisons (National Academy of Engineering, 1999), it is probably less representative of environmental performance for some firms than industry-specific measures. Additionally, our measure for environmental performance does not consider the relative toxicity of the waste being recycled, and aggregates all waste into one medium (i.e., air pollution is combined with water pollution). Although this summary statistic (recycling ratio) is suitable for inter-industry comparisons, it likely to be noisy relative to industry-specific metrics. As explained above, one of the primary challenges in conducting this research was measuring each of the three dependent variables. Whereas we used two measures of economic performance and a recycling ratio common across industries for our measure of environmental performance, our choice of a measure for environmental disclosure was, by its very nature, subjective. Although we restrict our definition of environmental disclosure

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

to include only specific pollution-related environmental measures, and clearly highlight this restriction throughout the discussion related to variable selection, our method of scoring disclosure quality retains a subjective element.28 As a final limitation, we recognize that our sample, drawn from S&P 500 firms, induces a size bias. Whereas our results may be generalized for large firms, inferring that small firms may behave similarly is overreaching. Although we acknowledge these limitations, we also believe that we have used those econometric tests most relevant in assuring that our models of simultaneous equations are correctly specified. Assuming correct specification, the simultaneousequation research design provides a significant improvement in estimating the relations among environmental performance, environmental disclosure, and economic performance over that obtained by independently estimated OLS models. We conclude that the simultaneous-equation approach used in this study provides a more coherent explanation regarding the relations among environmental disclosure, environmental performance, and economic performance than do those reported by prior studies using pair-wise tests of association. Further research might incorporate different variables for these constructs over different sample periods. As the world becomes more environmentally conscious (e.g., the Kyoto Protocol on climate change), additional examination of the relations among economic performance, environmental performance, and environmental disclosure is increasingly warranted. The study’s primary results that good environmental performance is associated with good economic performance and also with more 28 Our measure of environmental disclosure appears to be consistent with information sought by philanthropic foundations and investment managers who recently called upon the SEC to enforce regulations requiring firms to more fully disclose environmental liabilities. These investor groups appealed to the SEC that hidden environmental costs threaten their portfolios similar to other corporate accounting ‘‘tricks.’’ SEC guidelines require publicly held companies to disclose any ‘‘material’’ events and specifically require companies to disclose legal proceding that might result in sanctions of $100,000 or more (Bank, 2002).

469

forthcoming and factual environmental disclosure should be good news for those questioning the compatibility of corporate social responsibility and economic profitability. Non-financial measures of environmental performance, such as the recycling ratio presented in this study, may also be leading indicators of future financial performance. These measures therefore may be suitable candidates for incorporation in the firm’s ‘‘balanced scorecard’’ used to evaluate managerial and firm performance. Managers evaluated in this manner should be increasingly proactive in introducing new processes that improve both production efficiency and environmental sustainability.

Acknowledgements We appreciate the helpful insights and suggestions provided by Julia D’Souza, Robin Dubin, Tim Fogarty, Charlene Henderson, Dennis Oswald, Larry Parker, Kenny Reynolds, Kay Stice, and Kristina Zvinakis. This paper has also benefited from comments made by participants at the Brigham Young University, Case Western Reserve University, and Louisiana State University accounting workshops, the Tenth Annual Financial Economics and Accounting Conference, and the 2000 AAA Annual Meeting. The first author gratefully acknowledges the financial support of King Fahd University of Petroleum and Minerals.

References Abbott, W., & Monsen, R. (1979). On the measurement of corporate social responsibility: self reported disclosures as a method of measuring corporate social involvement. Academy of Management Journal, 22, 501–515. Atiase, R. K. (1985). Predisclosure information, firm capitalization, and security price behavior around earnings announcements. Journal of Accounting Research, 23(1), 21– 36. Bank, D. (2002). Group urges enforcing rules of environmental disclosure. The Wall Street Journal, 22 August 2002, B2. Barth, M., & McNichols, M. (1994). Estimation and market valuation of environmental liabilities relating to superfund sites. Journal of Accounting Research, 32(Supplement), 177– 209.

470

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471

Barth, M., McNichols, M., & Wilson, P. (1997). Factors influencing firms’ disclosures about environmental liabilities. Review of Accounting Studies, 2, 35–64. Belsley, D., Kuh, E., & Welsch, R. (1980). Regression diagnostics: identifying influential data and sources of collinearity. New York: John Wiley & Sons. Blacconiere, W., & Patten, D. (1994). Environmental disclosures, regulatory costs, and changes in firm value. Journal of Accounting and Economics, 18(3), 357–377. Botosan, C. (1997). Disclosure level and the cost of equity capital. The Accounting Review, 72(3), 323–349. Bowen, R., Castanias, R., & Daley, L. (1983). Intra-industry effects of the accident at three mile island. Journal of Financial and Quantitative Analysis, 18, 87–111. Bragdon, J., & Marlin, J. (1972). Is pollution profitable? Risk Management, 19, 9–18. Brockhoff, K. (1979). A note on external social reporting by German companies. Accounting, Organizations and Society, 4(1/2), 77–85. Buchholtz, R., Marcus, A., & Post, J. (1992). Managing environmental issues: a casebook. Englewood Cliffs, NJ: PrenticeHall. Christie, A. (1987). On cross-sectional analysis in accounting research. Journal of Accounting and Economics, 9(3), 231– 258. Collins, D., & Kothari, S. P. (1989). An analysis of intertemporal and cross-sectional determinants of earnings response coefficients. Journal of Accounting and Economics, 11, 143–182. Cormier, D., & Magnan, M. (1997). Research note: investors’ assessment of implicit environmental liabilities: an empirical investigation. Journal of Accounting and Public Policy, 16, 215–241. Corporate Environmental Profiles Directory. (1994, 1995). Investment. Responsibility Research Center: Washington DC. Deegan, C., & Gordon, B. (1996). A study of the environmental disclosure practices of Australian corporations. Accounting and Business Research, 26(3), 187–199. Financial Accounting Standards Board. (1975). Statement of financial accounting standards No. 5 Accounting for Contingencies. Norwalk, CT: FASB. Freedman, M., & Jaggi, B. (1982). Pollution disclosure, pollution performance and economic performance. Omega, 10, 167–176. Freedman, M., & Jaggi, B. (1992). An investigation of the long-run relationship between pollution performance and economic performance: the case of pulp-and-paper firms. Critical Perspectives on Accounting, 3(4), 315–336. Freedman, M., & Wasley, C. (1990). The association between environmental performance and environmental disclosure in annual reports and 10-Ks. Advances in Public Interest Accounting, 3, 183–193. Friedman, M. (1962). Capitalism and freedom. Chicago: The University of Chicago Press, Phoenix Books. Gaver, J., & Gaver, K. (1993). Additional evidence on the association between the investment opportunity set and corporate

financing, dividend and compensation policies. The Journal of Accounting and Economics, January/April/July, 125–160. Gray, R., Jarad, M., Power, D., & Sinclair, D. (2001). Social and environmental disclosure and corporate characteristics: a research note and extension. Journal of Business Finance and Accounting, April/May, 327–356. Gray, R., Kouhy, R., & Lavers, S. (1995). Corporate social and environmental reporting. Accounting, Auditing, and Accountability Journal, 8(2), 47–77. Guthrie, J., & Parker, L. (1989). Corporate social reporting: a rebuttal of legitimacy theory. Accounting and Business Research, 19(76), 343–352. Hausman, J. (1978). Specification tests in econometrics. Econometrica, 46, 1251–1271. Hughes, K. (2000). The value relevance of nonfinancial measures of air pollution in the electric utility industry. The Accounting Review, 75(2), 209–228. Hughes, S. B., Anderson, A., & Golden, S. (2001). Corporate environmental disclosures: are they useful in determining environmental performance? Journal of Accounting and Public Policy, 3(20), 217–240. Ilinitch, A., Soderstrom, N., & Thomas, T. (1998). Measuring corporate environmental performance. Journal of Accounting and Public Policy, 17, 383–408. Ingram, R., & Frazier, K. (1980). Environmental performance and corporate disclosure. Journal of Accounting Research, 18, 612–622. Kennedy, P. (1994). A guide to econometrics. Cambridge, MA: The MIT Press. Li, Y., Richardson, G., & Thornton, D. (1997). Corporate disclosure of environmental liability information: theory and evidence. Contemporary Accounting Research, 14(3), 435– 474. Little, P., Muoghlu, M., & Robison, D. (1995). An empirical analysis of the comparability of disclosure tendencies within and across industries. Research in Accounting Regulation, 105–117. Management Consultancies Association (MCA). (2000). Business and the environment. London, UK: MCA Available: www.mca.org.uk. Narver, J. (1971). Rational management responses to external effects. Academy of Management Journal, March, 99–115. National Academy of Engineering. (1999). Industrial environmental performance metrics: challenges and opportunities. Washington DC: National Academy Press Chapter 8.. Niskala, M., & Pretes, M. (1995). Environmental reporting in Finland. Accounting, Organizations and Society, 457–466. Ohlson, J. (1995). Earnings, book values, and dividends in equity valuation. Contemporary Accounting Research, 11(2), 661–687. Patten, D. (1992). Intra-industry environmental disclosures in response to the Alaskan oil spill. Accounting, Organizations and Society, 471–475. Patten, D. (1995). Variability in social disclosure: a legitimacybased analysis. Advances in Public Interest Accounting, 6, 273–285. Patten, D., & Nance, J. (1998). Regulatory cost effects in a

S.A. Al-Tuwaijri et al. / Accounting, Organizations and Society 29 (2004) 447–471 good news environment: the intra-industry reaction to the Alaskan oil spill. Journal of Accounting and Public Policy, 17, 409–429. Porter, M., & van der Linde, C. (1995a). Green and competitive: Ending the stalemate. Harvard Business Review, 73(5), 120–134. Porter, M., & van der Linde, C. (1995b). Toward a new conception of the environment-competitiveness relationship. Journal of Economic Perspectives, 9(4), 97–118. Richardson, A., & Welker, M. (2001). Social disclosure, financial disclosure and the cost of equity capital. Accounting, Organizations and Society, 26(7/8), 597–616. Rockness, J. (1985). An assessment of the relationship between U.S. corporate environmental performance and disclosure. Journal of Business Finance and Accounting, 12, 339–354. Rockness, J., Schlachter, P., & Rockness, H. (1986). Hazardous waste disposal, corporate disclosure and financial performance in the chemical industry. Advances in Public Interest Accounting, 1, 167–191. Smith, C., & Watts, R. (1992). The investment opportunity set and corporate financing, dividend, and compensation policies. Journal of Financial Economics, 32, 263–292. Shane, P., & Spicer, B. (1983). Market response to environmental information produced outside the firm. The Accounting Review, 58, 521–538. Slavich, J. (1994). Show and tell: new developments impacting the disclosure of environmental liabilities. Environmental Permitting, Summer, 327–343.

471

Spicer, B. (1978). Investors, corporate social performance and information disclosure: an empirical study. The Accounting Review, 53, 94–111. Stevens, W. (1984). Market reaction to corporate environmental performance. Advances in Accounting, 1, 41–61. Tietenberg, T. (1998). Disclosure strategies for pollution control. Environmental and Resource Economics, 11(3–4), 587–602. Ullmann, A. (1985). Data in search of a theory: a critical examination of relationships among social performance, social responsibility, and economic performance of U.S. firms. Academy of Management Review, 10, 540–557. Verrecchia, R. (1983). Discretionary disclosure. Journal of Accounting and Economics, 5(3), 179–194. Vogel, D. (2002). Recycling corporate responsibility. The Wall Street Journal, 20 August 2002, B2. Walley, N., & Whitehead, B. (1994). It’s not easy being green. Harvard Business Review, May–June, 46–52. White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica, 48, 817–838. Wiseman, J. (1982). An evaluation of environmental disclosure made in corporate annual reports. Accounting, Organizations and Society, 7, 53–63. Zeghal, D., & Ahmed, S. (1990). Comparison of social responsibility information disclosure media by Canadian firms. Accounting, Auditing, and Accountability Journal, 3(1), 38–53.