Executive compensation and non-financial risk: An empirical examination

Journal of Accounting and Public Policy 26 (2007) 436–462

www.elsevier.com/locate/jaccpubpol

Journal of Accounting and Public Policy 26 (2007) 436–462

www.elsevier.com/locate/jaccpubpol

Executive compensation and non-financialrisk: An empirical examination

Katherine Campbell a, Derek Johnston b,Stephan E. Sefcik c, Naomi S. Soderstrom d,*

a University of North Dakota, Department of Accountancy, College of Business and Public

Administration, P.O. Box 8097, Centennial Drive, Grand Forks, ND 58202-8097, United Statesb Colorado State University, College of Business, Rockwell Hall, Room 256, Fort Collins,

CO 80524, United Statesc University of Washington, Business School, Department of Accounting, Box 353200,

Seattle, WA 98195-3200, United Statesd University of Colorado at Boulder, 419 UCB, Boulder, CO 80309-0419, United States

Abstract

Executives face potentially severe (non-financial) personal risks if firm environmentalperformance is below industry best practice. We examine the relation between CEOcompensation and the non-financial risk associated with environmental exposure, andhow use of environmental performance as an explicit determinant of compensationaffects this relation. We find evidence that CEOs are compensated for exposure to envi-ronmental risk, even after controlling for financial risk. We also find that this premiumis reduced when the CEO has greater opportunities to improve the firm’s environmentalperformance.� 2007 Elsevier Inc. All rights reserved.

Keywords: Executive compensation; Non-financial risk; Environmental accounting

0278-4254/$ - see front matter � 2007 Elsevier Inc. All rights reserved.doi:10.1016/j.jaccpubpol.2007.05.001

* Corresponding author. Tel.: +1 3037356620; fax: +1 3034925962.E-mail address: [email protected] (N.S. Soderstrom).

mailto:[email protected]

K. Campbell et al. / Journal of Accounting and Public Policy 26 (2007) 436–462 437

1. Introduction

Environmental performance is frequently used in determining executivecompensation.1 Such a policy gives managers direct financial incentives tomanage firm environmental performance. An important aspect of environmen-tal performance is that there are significant risks associated with poor environ-mental performance – risks which extend to corporate officers as well as tofirms. In addition to indirectly bearing environmentally-related firm risks(e.g., fines, penalties, increasing regulation, mandated capital expenditures),executives may be subject to severe personal risk. For example, several environ-mental laws impose both civil and criminal liability on executives. Therefore,the risk of personal penalties for CEOs will be greater when their firm’senvironmental performance is poor. This context allows us to empiricallyinvestigate the role of a specific non-financial risk factor in executive compen-sation. Specifically, CEOs may demand a wage premium to compensate themfor their exposure to environmental risk. Consistent with this conjecture, theempirical results reveal a positive relation between a firm’s environmental riskand CEO compensation, after controlling for other determinants of compensa-tion such as financial risk. Moreover, we find that this premium is reducedwhen the CEO has greater opportunities to improve the firm’s environmentalperformance.

Both stock market and accounting-based measures of performance are fre-quently incorporated into compensation contracts and the relative use of thesemeasures has been extensively investigated. For example, Murphy (1985) con-cludes that executive compensation is positively related to stock market rate ofreturn. In addition, Antle and Smith (1986) demonstrate that executive com-pensation is more sensitive to return on equity relative to industry than to grossreturn on equity. Lambert and Larcker (1987) examine the cross-sectional dif-ferences in the relative importance of market returns and return on equity incompensation, while Ely (1991) provides evidence of inter-industry differencesin the relative weight placed on alternative performance measures. Finally,Sloan (1993) shows that earnings-based incentives help shield executives fromrisk associated with fluctuations in firm value that are beyond their control.

Other research examines characteristics of firms that use non-financial per-formance measures and the association of those measures with executive com-pensation and future financial performance. For example, Ittner et al. (1997)use proprietary data to investigate the prevalence of, and choice between,financial versus non-financial performance metrics in executive bonus con-tracts. Bushman et al. (1996) examine factors that are associated with use of

1 A Price Waterhouse (1995) survey of corporate environmental accounting and managementpractices reports that 38% of companies tie senior management compensation to environmentalperformance.

438 K. Campbell et al. / Journal of Accounting and Public Policy 26 (2007) 436–462

individual performance evaluation (i.e., achievement of specified individualperformance goals) in CEO annual incentive plans. Moreover, Davila andVenkatachalam (2004) find that a non-financial performance measure in theairline industry, passenger load factor, is associated with CEO compensation.Finally, Banker et al. (2000) report that the use of non-financial performancemeasures in incentive plans improve future financial performance.

A related research stream in the area of executive compensation investigatesthe causes of executive wage premiums. For instance, Duru and Reeb (2002)find that geographic diversification at the firm level results in a compensationpremium for CEOs. Combs and Skill (2003) document a positive associationbetween pay premiums and executives’ managerial abilities for some firms,and a positive relation between wage premiums and executive entrenchmentfor other firms. In addition, Roulstone (2003) reports that executives receivea compensation premium when firm-level policies impose additional insider-trading restrictions.

Finally, prior research examines the association between risk and compensa-tion. For instance, Garen (1994) and Core and Guay (1999) investigate the rela-tion between elements of financial risk and CEO pay, and document a positiverelation between the two. In addition, Smith and Watts (1992) and Prendergast(2002) note that the association between risk and compensation may be contin-gent upon uncertainty (i.e., information asymmetry). Overall, the theoreticaland empirical literature addressing the role of risk in compensation is extensive.However, previous empirical analyses have been limited to considering elementsof financial risk. To the best of our knowledge, this is one of the first studies toempirically examine the relation between executive compensation premiumsand a non-financial risk element, after controlling for financial risk.

This paper may also have implications for non-academics. How a firm setsthe compensation of its executives has come under much scrutiny and theSecurities and Exchange Commission recently proposed increased disclosuresconcerning compensation figures (Scannell, 2006). Our results suggest thatenvironmental risk is a driver of CEO compensation and, therefore, regulatorsand other stakeholders should consider non-financial risk when evaluating thereasonableness of CEO salaries and bonuses.

The remainder of our study is organized as follows. We develop the hypoth-eses in Section 2 and discuss the research design in Section 3. Section 4 detailsour sample selection process and reports descriptive statistics. Section 5 pre-sents the empirical results and Section 6 concludes.

2. Development of hypotheses

Firms involved with use or release of toxic materials are subject to environ-mentally-related economic risks. Poor environmental performance can result in


mandated remediation expenditures, penalties, litigation costs, and reputationcosts. In addition, firms involved with toxic materials face the risk of increasedregulatory scrutiny and further environmental legislation that may imposecostly compliance efforts. Thus, firms have strong incentives to evaluate andmanage their environmental performance.

However, the risks associated with environmentally sensitive business activ-ities are not limited to the firm. Environmental regulations create explicit lia-bility for executives and corporate officers. In the case of executives, risksare both economic (i.e., civil) and personal (i.e., criminal). For example, theUS Resource Conservation and Recovery Act provides criminal penaltiesincluding fines and imprisonment for any person who ‘‘knowingly’’ transports,treats, stores, or disposes of hazardous materials improperly. The courts haveliberally interpreted this provision so that a CEO can be held responsible for‘‘knowing’’ how her/his firm’s hazardous materials are handled. As such, anexecutive is unlikely to escape personal criminal liability by delegating respon-sibility for environmental compliance.2 Bolstering this notion, Epstein (1996)makes the following observation in his book, which reports the results froma field study on environmental management practices:

2 CriCleanand thdefendone ye

‘‘A significant change has occurred in the criminal prosecution of theenvironmental laws, whereby officers, directors, and employees now arebeing prosecuted in addition to the corporations that employ them. Reg-ulators believe that because corporations can only be assessed fines, asuccessful deterrent for environmental violations is the threat of prisonterms for the responsible corporate employees. Indeed, one of the mostnoticeable changes I observed during this two-year research project wasmanagers’ dramatically rising concern for the potential personal liabilityof environmental violations.’’ (Epstein, 1996, p. 5)

Despite both firm and CEO incentives to improve environmental perfor-mance, there may be a limit to economically advantageous improvements evenwhen technology would allow further enhancements to environmental perfor-mance. For instance, there may be diminishing marginal returns for the reduc-tion of emissions. Moreover, positive net present value capital projects mayincrease a firm’s emissions. As such, the economically optimal solution forthe firm as a whole, which includes consideration of the interests of a variety

minal sanctions, including imprisonment for one to 15 years, are also provided under theAir and Clean Water Acts. For enforcement purposes, the Office of Criminal Investigationse Environmental Crimes Section were established in 1981 at the federal level. In 1991, 82ants were convicted of environmental violations with an average jail term of approximatelyar (Kolluru, 1994).


of stakeholders, may not result in the highest possible level of environmentalperformance. Hence, the environmental risk imposed on the firm (and its exec-utives) may not be minimized.

Due to asymmetry in the costs of poor environmental performance (and thecorresponding increased environmental risk) borne by executives and the firm,tension arises between interests of the firm and those of its executives. Althoughexecutives might contractually be relieved of civil penalties (i.e., firms canindemnify executives for environmental fines and penalties), criminal penaltiesand personal reputation consequences cannot be contracted away. The potentialseverity of personal and reputation costs suggests that managers will be partic-ularly risk averse when facing potential criminal penalties. Thus, when it is notoptimal for the firm to maximize the level of environmental performance, addi-tional risk is imposed on its executives. This environmental performance-relatedrisk for executives is at the heart of our research question and hypotheses.

Regulators recognize that there are constraints in the extent to which environ-mental performance can be improved. To this end, some environmental lawsexplicitly incorporate both cost-benefit analysis and best-practice concepts.For example, the Clean Air Act empowered the EPA to set emission standardsfor certain new sources of pollution. These performance standards were requiredto be based on the ‘‘best technological system of continuous emission reductions’’and had to be affordable by the affected parties (Portney, 1990, p. 38). Similarly,the Clean Water Act directed the EPA to consider ‘‘best conventional pollutioncontrol technology,’’ ‘‘best practicable control technology currently available,’’or ‘‘best available technology economically achievable’’ when setting effluentlimits for industrial wastewater (Stimson et al., 1993).

Although environmental regulations do not explicitly address this issue,environmental regulators and courts are less likely to pursue and impose crim-inal penalties for adverse environmental outcomes when all possible measureshave been taken to minimize the risk of such an adverse outcome. For example,the risk of major oil spills from tankers can be reduced by using double-hulledtankers, monitoring captains for alcohol and drug use, and having establishedcontainment procedures for spills that do occur. While the risk of serious spillswill never be reduced to zero, it is likely to be much more difficult to justify anadverse outcome to the courts when the firm has not at least adopted the risk-reducing measures implemented by its industry peers. As such, we define ‘‘envi-ronmental exposure’’ as the risk stemming from environmental performancethat is below industry best-practice levels.

To the extent that a firm is environmentally exposed (i.e., its environmentalperformance lags behind industry best practice), its executives can reduce theirpersonal risk by implementing appropriate environmental management proce-dures. However, as previously mentioned, other corporate imperatives, includ-ing near-term profitability, may not allow for improvement in environmentalperformance. In these cases, executives do not have the opportunity to improve


environmental performance and, therefore, are likely to demand compensationfor the personal risk imposed by this environmental exposure. Consequently,this paper tests the following hypothesis (stated in alternative form):

H1: CEO compensation is positively associated with environmental exposure.

It is important to note that the first hypothesis hinges on the assumption thatthe firm has already maximized its environmental performance subject to the con-straints imposed by other corporate initiatives. As such, the CEO is not providedwith additional avenues to improve the firm’s environmental performance anddecrease her/his personal environmental risk. However, this is an unlikelyassumption for all firms as some will have yet to maximize their environmentalperformance subject to the aforementioned constraints. We hereafter refer tothese types of firms as NON-MAX firms, whereas firms that have maximizedtheir environmental performance are henceforth called MAX firms.

In our research setting, it is important to differentiate NON-MAX firmsfrom MAX companies since the relation between environmental exposureand CEO compensation is likely to be different for these two types of firms.By definition, a NON-MAX firm still has an opportunity and (arguably) hasa greater incentive to improve its environmental performance relative toMAX firms. Consequently, the CEO of a NON-MAX firm has the ability toincrease the firm’s environmental performance, thereby reducing her/his envi-ronmental exposure. Since the degree of the CEO’s personal environmentalexposure risk is within her/his control, we believe NON-MAX firms will payless of an environmental risk premium than MAX firms.

To investigate this assertion, we partition our sample firms into NON-MAXand MAX firms. We execute this separation by assuming that firms that havechosen to make environmental performance an explicit determinant of execu-tive compensation (as reported in the Investor Responsibility Research Cen-ter’s survey) are NON-MAX firms. Presumably, the existence of anenvironmental performance-based compensation plan implies that opportuni-ties to improve environmental performance are available, since some dimen-sion of environmental performance is being measured, monitored, and usedas a determinant of compensation. Further, the environmental performance-basedcomponent of compensation is designed to reward improved environmentalperformance and shield the executive from negative compensation conse-quences that may arise from other correlated compensation determinants.3

3 Initial expenditures necessary to improve environmental performance (e.g., production methodrevisions, environmental performance monitoring and control systems development, etc.) mightadversely impact accounting metrics of performance. Thus, compensation plans explicitlyincorporating environmental performance metrics might make adjustments for environmentalexpenditures or shift away from accounting performance measures. In our sensitivity analyses,however, the empirical weights on accounting and firm performance were not related tocompensation structure.


As such, we believe the existence of an environmental performance-based com-pensation component is an indication that the CEO has opportunities to improvethe firm’s environmental performance and, hence, lower her/his personal envi-ronmental exposure risk. Since the CEO has control over her/his environmentalrisk, the firm is likely to pay less of an environmental exposure premium.

In sum, when the firm uses environmental performance-based compensa-tion, we posit a reduction in the environmental exposure premium.4 Statedin alternative form, we test the following hypothesis:

H2: Ceteris paribus, use of an environmental performance-based executive

compensation plan reduces the environmental exposure premium component ofexecutive compensation.

3. Research design

To examine the relation between environmental exposure and executivecompensation, we use a regression model similar to that used in the prior com-pensation literature (Clinch, 1991). In particular, we estimate the followingequation in pooled cross-section:5,6

COMPit ¼ b0 þ b1ENVit þ b2EXPit þ b3ðENVit�EXPitÞþ b4ASSETit þ b5ROAit þ b6RTNit þ b7BETAit

þ b8STDit þ b9TENUREit þXY�1

y¼1

b10yYRyit

þXJ�1

j¼1

b11jINDjit þ eit ð1Þ

4 Nevertheless, even if environmental performance is one of several explicit determinants ofcompensation, executives will still need to manage environmental performance in balance withother firm objectives. For example, Polaroid incorporated seven dimensions over which executiveswere evaluated, only one of which was environmental (Barth et al., 1994). In these cases,environmental exposure may still not be reduced to the lowest possible level (i.e., best-practicelevels), and executives may continue to demand some environmental exposure premium. On theother hand, the explicit environmental determinant of compensation may offset or even exceed theeffect of the compensation plan on the risk premium. For instance, if environmental performanceimproves, then CEO compensation will be higher as a function of the environmental performancedeterminant in the plan.

5 Although both levels and changes models have been extensively used in the compensationliterature, the levels specification is conceptually most consistent with our risk-based hypotheses.

6 We use a cross-sectional analysis approach (as opposed to an individual time-series) due to dataconstraints. Both cross-sectional (e.g., Adams, 1987; Ely, 1991; and Sloan, 1993) and time-series(e.g., Antle and Smith, 1986; and Lambert and Larcker, 1987) approaches have been used in priorresearch investigating relations between executive compensation and firm performance.


where i, t denote firm and year, respectively; COMP is the natural log of com-pensation; ENV is a dichotomous variable that equals one if the firm uses envi-ronmental performance in determining executive compensation, zerootherwise; EXP is the environmental exposure proxy, specifically, the firm’sInvestor Responsibility Resource Center efficiency index (Toxic Releases/Sales)divided by the lowest efficiency index in the firm’s industry (as defined bytwo-digit SIC code); ASSET is the natural logarithm of total assets; ROA isthe return on assets, defined as net income divided by end of year total assets;RTN is the stock market returns; TENURE is the number of years CEO is inposition; BETA is the company beta (as reported in CRSP); STD is the stan-dard deviation of monthly stock returns over previous 24–60 months; YRy is adichotomous variable that equals one if the observation is from year y, zerootherwise; and INDj is a dichotomous variable that equals one if the firm oper-ates in industry j (as defined by two-digit SIC code), zero otherwise.

3.1. Measurement of executive compensation

Based on prior research, the range of alternatives for estimating annualcompensation includes short-term compensation (Murphy, 1985, 1986; Lam-bert and Larcker, 1987; Sloan, 1993), short-term compensation plus estimatesof some long-term (equity) components of compensation (Murphy, 1985;Clinch, 1991), and estimates of total compensation (Antle and Smith, 1986;Ely, 1991; Core et al., 2003).

In this paper, we use annual short-term compensation (salary plus bonus) asthe measure of compensation (COMP).7 We hypothesize that environmentalperformance will be reflected in executive compensation as an exposure pre-mium. Such a premium should be a component of short-term compensationsince it is unlikely that a firm would choose to compensate its CEO for takingon environmental exposure risk by introducing additional compensation risk.Including the premium in long-term compensation by, for example, granting

7 It would be useful to explore whether the results from the testing of our hypotheses change if weuse only salary or bonuses as our measure of compensation. However, salary and bonus were notseparately disclosed in the Forbes Compensation Surveys for 1988–1991. As such, using only fixedsalary as our measure of compensation would require changing the sample period. However due tothe nature and evolution of TRI reporting, we cannot do this. When first disclosed, TRI numberswere relatively un-managed and therefore provide a valid measure of environmental exposure. Astime progressed, however, the Environmental Protection Agency began to express concern over the‘‘inaccuracy’’ of these reports (Hess, 1995), possibly because of the dramatic decrease the agencysaw in the first few years of mandatory TRI reporting (EPA Journal, 1995). Consequently, webelieve that TRI releases best capture a CEO’s environmental exposure in the time period currentlyused in our study.


additional stock options, would result in a non-zero probability that the exec-utive would never realize the premium.8

It is important to note that we include only CEO compensation in our anal-yses. Field study data suggest that, in most companies, major efforts in measur-ing and reporting environmental impacts do not begin until the CEO iscommitted to environmental management (Epstein, 1996). Moreover, exami-nation of survey data from the Investor Responsibility Resource Center revealsthat the reporting level for environmental officers is frequently (34%) at theCEO or Board level. In these cases, inclusion of other executives (withoutdirect responsibility for environmental performance) would add noise to theanalyses. When a firm indicates a lower reporting level in the IRRC survey,the level is often sufficiently low that the appropriate manager’s compensationis not included in proxy statements. Nevertheless, in cases where environmentalresponsibility falls below the CEO level, exclusion of more directly responsibleofficers from our analyses may cause us to fail in finding support for ourhypotheses.

3.2. Measurement of environmental exposure and environmental compensation

policy

Our investigation of the association between environmental exposure andexecutive compensation requires data on firm-level environmental perfor-mance, policies and practices. The Investor Responsibility Resource Center(IRRC) releases annual Corporate Environmental Profiles Directories, whichcontain data that allow us to develop proxies for the necessary environmentalvariables. In particular, the IRRC directories provide measures of environmen-tal performance (e.g., toxic releases and chemical spills) for S&P 500 firmsbased on federal environmental data, reviews of SEC filings, and self-reporteddata. In addition, IRRC gathers additional data regarding internal policies andpractices (e.g., whether or not environmental performance is a factor in deter-mining executive compensation) through annual surveys. We use the data inthe IRRC directories to measure the two independent environmental variablesin Eq. (1) (i.e., EXP and ENV).9

8 As a robustness check, we re-estimate Eq. (1) using the Black–Scholes value of the optionsgranted to the CEO as our dependent variable instead of salary and bonus. Consistent with ourabove argument, the untabled results reveal that our environmental exposure proxy is notsignificantly related to this measure of compensation. However, due to additional data constraints,we were only able to estimate the model on approximately one-third of our sample observations,severely reducing the statistical power of the test.

9 Firms that responded to the survey are larger than non-respondents; however, there is nosignificant difference between these two sub-samples along the dimensions of profitability (asmeasured by ROA) and leverage (as measured by debt-to-equity).


First, we develop a proxy for the environmental exposure construct (EXP).We base this proxy on Toxic Release Inventory (TRI) emissions, which repre-sent legal releases of approximately 325 toxic chemicals that firms must reportto the federal government under the Emergency Planning and CommunityRight-to-Know Act.10 TRI emissions reflect the firm’s domestic (US) use oftoxic and hazardous substances, and, hence, the firm’s overall US environmen-tally-related risks. IRRC gathers TRI emissions data from federal governmentdatabases and includes these data in its annual directories.11

To develop our environmental exposure construct, we first scale TRIreleases by domestic sales (this measure constitutes the IRRC ‘‘efficiencyindex’’). Since emissions are likely to fluctuate in response to changes in pro-duction levels, scaling TRI releases by sales produces a meaningful measureof emission efficiency that facilitates comparison across firms and is less sensi-tive to changes in production levels within firms across time. We use domesticsales as the deflator since TRI emissions are reported only for US operations.

Next, we adjust this TRI-based metric for industry best-practice levels ofperformance. Specifically, we divide the firm’s efficiency index (TRI releases/Sales) by the lowest efficiency index in firm’s industry (measured by two-digitSIC code).12 Thus, our environmental exposure proxy in Eq. (1) is a measureof each firm’s environmental performance relative to industry best-practice lev-els. Using an industry-specific best-practice benchmark is important in the con-text of our study since overall environmental performance, as well as use ofchemicals subject to TRI reporting, can vary widely across industries. More-over, as previously mentioned, regulators are likely to evaluate executives’management of environmental matters by comparison with peer institutions.

The design of our environmental exposure construct assumes that compa-nies generating revenues with higher levels of emissions than others in theirindustry have more environmental exposure, and executives of such firms likelyface higher risks associated with environmental exposure, including criminalpenalties. Although TRI emissions are legal, the amount of toxic material pro-

10 The number of chemicals that are included in TRI emissions has increased to 654 since the timeperiod of this study.11 The first TRI data was reported in 1988, and quickly became a highly visible public metric of

environmental performance. The EPA started the ‘‘33/50 Project,’’ a voluntary program designedto encourage companies to reduce emissions of certain chemicals from 1988 levels by 33% by 1992and 50% by 1995. Companies such as Monsanto, BP Chemicals, and Alcoa agreed to participate inthis program. Monsanto’s Chairman, Richard Mahoney, developed a 90% reduction target that farexceeded the voluntary target (Hunter, 1993).12 To avoid dividing by zero, we transform the firm and industry indices by adding one to each.


duced, especially relative to similar enterprises, reflects the environmental riskrelated to the firm’s activities.13 If executives receive compensation for assum-ing this personal environmental exposure risk, then EXP will be positivelyrelated to COMP.

Our analyses also require use of survey data gathered by IRRC and includedin its annual directories. Specifically, we use the firm’s response regarding expli-cit inclusion of environmental performance as a determinant of executive com-pensation to construct the second independent variable, ENV. In particular,ENV is a dichotomous variable that equals one if the firm uses environmentalperformance in determining executive compensation, zero otherwise. Oursecond hypothesis contends that the environmental exposure premium thatexecutives demand will be lower when there is an environmental perfor-mance-based compensation plan in place. If this conjecture is true, then theinteraction term (ENV*EXP) will be negatively related to COMP. We alsoinclude ENV as an independent variable in Eq. (1) to control for the unobserv-able dimensions of environmental performance that are determinants of com-pensation when such a compensation plan is in place.14

Since ENV is a choice variable, endogeneity of this variable is a concern. Tomitigate concerns about this potential problem, we use an instrumental vari-able method by applying three-stage least squares (3SLS) to Eq. (1) and a

13 We would not expect that such a global metric as TRI emissions would be the basis of anenvironmental performance metric incorporated in CEO compensation contracts (i.e., a contract-ing variable). While we have been unable to examine environmental performance-basedcompensation contracts for CEOs, executives at several firms have indicated to us thatenvironmental performance is usually determined relative to very specific, individualized goals.This is consistent with case study evidence for executives below the CEO level reported in Epstein(1996) and Lanen (1995). Empirically, we would expect a contracting variable to behave differentlythan a risk-related variable.14 To the extent there are biases in the survey information reported by IRRC, our analyses related

to the use of environmental performance-based compensation may be compromised. Since IRRCattempts to corroborate self-reported information with other data sources and corrects reportedinformation for prior periods if errors are detected, no significant systematic bias can be identifiedex ante. Federal audits and reporting penalties should constrain such reporting bias for ourenvironmental exposure proxy based on Toxic Release Inventory (TRI) data. The EPA performscompliance inspections of facilities with regard to these regulatory environmental reportingrequirements and penalties for improper reporting are substantial. Financial penalties include finesup to $25,000 per day of reporting violation. We are unaware of an alternative or preferable datasource identifying firms using environmental performance-based compensation that wouldfacilitate large sample analysis. Proxy statements were considered as a means to determineinclusion of environmental factors in performance. However, proxy disclosures are often vague andinconsistently worded even for firms reporting use of an environmental factor in compensation tothe IRRC.


second equation.15 In the second equation, we model the firm’s decision toadopt an environmental performance plan:

ENVit ¼ a0 þ a1ROAit þ a2RTNit þ a3TQMþ a4ENVREPit

þ a5TENUREit þ a6USAUDit þ a7USSTDit þ a8EXPit

þ a9SPILLit þ a10COMPLYit þ cit ð1aÞ

where TQM is a dichotomous variable that assumes the value of one if the firmemploys total quality environmental management, zero otherwise; ENVREP isthe reporting level for highest environmental officer: one if CEO or Board ofDirectors, two if Vice President level, three if lower than Vice President;USAUD is the percent of US facilities audited; USSTD is a dichotomous var-iable that assumes the value of one if US environmental standards are used inoverseas operations, zero otherwise; SPILL is the firm’s IRRC’s spill index(number of chemical and oil spills/sales) divided by the lowest spill index inthe firm’s industry (as defined by two-digit SIC code); and COMPLY is thefirm’s IRRC’s compliance index (dollar amount of penalties/sales) divided bythe lowest compliance index in the firm’s industry (as defined by two-digitSIC code).

All other variables are as previously defined. We include two financial per-formance measures (ROA and RTN) in Eq. (1a) since a firm may be more aptto adopt an environmental performance compensation plan if it is already per-forming well financially. Hence, ROA and RTN may be positively related toENV. We include four environmental variables (TQM, ENVREP, USAUD,USSTD) from the IRRC surveys that, taken together, likely measure the valuethat the company places on its environmental performance. As this valueincreases, firms may be more likely to have an environmental performancecompensation plan in place. Consequently, we posit a positive associationbetween these four variables and ENV.

We also control for firm-level environmental performance by including theindustry-adjusted TRI efficiency index (EXP) and two other indices based ondata from the IRRC directories. In particular, SPILL is the combined numberof chemical and oil spills, whereas COMPLY is the total dollar value of envi-ronmental penalties incurred under nine environmental statutes. Both indicesare deflated by domestic sales and scaled by industry-specific best-practice lev-els in the same manner as the TRI efficiency index. Finally, we control for CEO

15 We choose to implement the instrumental variable approach by applying the 3SLS procedure asopposed to two-stage least squares to correct for any potential correlation among the error termsacross the two equations.


tenure (TENURE) since the number of years a CEO has been in position mayaffect her/his compensation scheme.

We conduct two specification tests to determine the appropriateness of our3SLS estimation procedure. In particular, we test for the existence of an end-ogeneity problem as well as for the appropriateness of our instrumental vari-ables (i.e., that our instruments are exogenous). Concerning the former, theDurbin-Wu-Hausman test rejects the null hypothesis that ENV is exogenous.However, this test assumes that the instruments are appropriate and, hence,Larcker and Rusticus (2005) stress the importance of testing that assumptionbefore proceeding with the two- (or three-) stage least squares estimation. Assuch, we test the validity of our instrumental variables by executing the testdescribed by Larcker and Rusticus (2005). Specifically, we regress the residualsfrom Eq. (1) on all of the exogenous variables (including the instrumental vari-ables from Eq. (1a)). Larcker and Rusticus (2005) note that if the instrumentalvariables are valid, then the R2 from this model will be close to zero. Consistentwith that notion, we are unable to reject the null hypothesis that the R2 fromthe model is significantly different from zero (v2 = 8.70, p-value > 0.10). Takentogether, the results from these specification checks suggest that the endogene-ity of ENV is a potential problem and that our instruments appear to beappropriate.

3.3. Measurement of control variables

We include two measures of financial performance in our model: anaccounting performance measure (return on assets) and a stock market perfor-mance measure (stock market returns). Both measures are commonly used incompensation contracts and have been previously demonstrated to be relatedto compensation. Moreover, following Bushman et al. (1996) and Davilaand Venkatachalam (2004), we control for CEO power and the quality ofthe CEO by including the length of time the CEO has been in position (TEN-URE). In addition, Jin (2002) shows that firm size is a determinant of executivecompensation; hence we include the natural logarithm of total assets (ASSET)in Eq. (1).

In addition, we control for the relation between financial risk and executivecompensation. Specifically, we include two measures of financial risk: the com-pany beta (BETA) and the standard deviation of returns (STD). STD is definedas the standard deviation of past 24–60 (depending on data availability)monthly stock returns.

Since we estimate Eq. (1) on pooled data, we control for differences acrossthe four year time period using a fixed effects approach. Specifically, we intro-duce dummy variables allowing a different intercept for each year. Moreover,we include industry controls to disentangle the environmental exposure effect


from any potential industry effects. In particular, we include an industry indi-cator variable for each two-digit SIC code in which our sample includes morethan one firm.16

4. Sample selection and descriptive statistics

Our sample selection procedure is multi-phased. First, we begin with the 514firms that are covered by the 1992–1994 IRRC Corporate Environmental Pro-files Directories. We obtain environmental performance data (e.g., toxicreleases and the number of chemical and oil spills) for 1988–1991 from the1993 and 1994 IRRC directories.17 We chose this sample period since itincludes initial years of reporting under the Emergency Planning and Commu-nity Right-to-Know Act (EPCRA), which focused not on regulating individualchemicals, but on regulating the information reported about chemical use. As aresult of EPCRA, corporations installed systems to monitor, measure, report,and control toxic chemical use, and both regulators and the public gainedaccess to information regarding corporate chemical use. Thus, during this per-iod, patterns of chemical use varied across firms, internal and external scrutinyof emission reports was high, and firms developed emission reduction targets.As a consequence, we expect EPCRA data from this period to be the mostpowerful in reflecting risk. It has been alleged that after this relatively short ini-tial period, firms began to proactively manage these publicly reported TRIemissions to avoid negative stakeholder attention. Overall TRI emissionsdeclined significantly after this period (USEPA, 1999), thus lowering the powerof this metric to capture risk. While TRI emissions continue to be reported, webelieve that the underlying construct captured by them has changed over time.

Data on environmental compensation policies and practices are taken fromthe survey included in the 1992 IRRC directory.18 By using the survey pub-lished in the 1992 report, we assume that a company incorporating environ-mental performance as a determinant of executive compensation at the endof our sample period (1991) had a similar compensation policy in place overthe entire sample period. Since IRRC survey data are not available prior to1992, we make this assumption due to data constraints, recognizing that it

16 The results that follow are not materially affected when we exclude the year and industryindicator variables.17 The IRRC directories report lagged environmental performance and survey data. For example,

the 1992 report includes environmental performance data (such as emission and spill data) from1987 to 1989 and survey data from 1991.18 Recognizing differences in environmental issues faced by service and industrial firms, IRRC

distributes two different survey instruments. The service firm questionnaire does not include aquestion about environmental factors in compensation plans. Consequently, our sample excludesthese firms.

Table 1Sample selection and industry distribution

Criterion Number of firms

Panel A: Sample selection process

Firms covered by Investor Responsibility Research Center(IRRC)

514

Less: Firms that do not have the necessary IRRCenvironmental data

(266)

Less: Firms that do not have compensation data availablefrom Forbes compensation survey during 1988–1991 and/or

CEO is in first year of office

(105)

Less: Firms that do not have the necessary COMPUSTATdata

(12)

Final sample 131

1-Digit SICcode for industry

Environmentalperformanceincluded incompensation

Environmentalperformance notincluded incompensation

Totalsample

Panel B: Industry distribution of sample firms

1000’s 1 1 22000’s 39 11 503000’s 31 20 514000’s 10 9 195000’s 2 1 36000’s 2 1 37000’s 1 2 3Total 86 45 131


may weaken the power of our tests. Of the 514 firms covered in the IRRCdirectories, 248 have the necessary environmental performance and surveydata.19

Next, we obtain annual salary plus bonus compensation from the Forbes

Compensation Surveys for 1988–1991. Further sample selection criteria includeavailability of CEO short-term compensation from Forbes compensation sur-veys and financial data from COMPUSTAT. After deleting firms for whichcompensation data (105 firms) and COMPUSTAT data (12 firms) are unavail-able, the final sample is comprised of 131 firms. Panel A of Table 1 summarizesour sample selection process, while Panel B reports the industry distribution ofthe 131 sample firms.

19 Hence, our sample consists only of firms that responded to the IRRC survey. Unreportedunivariate analyses reveal that non-respondents are smaller and have less environmental exposureas measured by toxic releases, compliance penalties, and number of spills. Thus, overall variation infirm environmental exposure may be reduced in the sample.

Table 2Descriptive statistics

Variable Firm-years Mean Median Minimum Maximum

Panel A: Financial and environmental exposure variables

All observations

Total assets ($mill.) 364 15,473 7129 453 184,325Sales ($mill.) 364 12,017 6772 478 124,993Net income ($mill.) 364 586 298 �4992 6020Return on assets 364 0.06 0.05 �0.06 0.36Market return 364 1.12 1.13 0.31 2.39Efficiency Indexa 364 2.34 0.52 0.00 56.38Spill Indexb 364 2.07 0.00 0.00 48.29Compliance Indexc 364 59.53 1.22 0.00 9203.78CEO salary and bonus ($thou) 364 1259 1083 355 11,233CEO tenure (years) 364 7.91 6.00 1.00 47.00

Firms that include environmental performance in compensation (ENV = 1)d

Total assets ($mill.) 247 18,075* 8441 453 184,325Sales ($mill.) 247 13,851* 7672 478 124,993Net income ($mill.) 247 660** 365 �4992 5600Return on assets 247 0.06 0.05 �0.06 0.36Market return 247 1.11 1.14 0.31 1.96Efficiency Indexa 247 3.18* 0.81 0.00 56.38Spill Indexb 247 2.72* 0.00 0.00 48.29Compliance Indexc 247 82.18 2.82 0.00 9203.78CEO salary and bonus ($thou) 247 1276 1161 355 3888CEO tenure (years) 247 8.26 6.00 1.00 47.00

Firms that do not include environmental performance in compensation (ENV = 0)d

Total assets ($mill.) 117 9981 5675 668 92,473Sales ($mill.) 117 8146 5284 784 69,018Net income ($mill.) 117 432 193 �1086 6020Return on assets 117 0.06 0.05 �0.05 0.19Market return 117 1.13 1.12 0.40 2.39Efficiency Indexa 117 0.58 0.06 0.00 11.15Spill Indexb 117 0.70 0.00 0.00 12.15Compliance Indexc 117 11.69 0.24 0.00 286.03CEO salary and bonus ($thou) 117 1225 807 375 11,233CEO tenure (years) 117 7.18 5.00 1.00 41.00

Mechanism No. of firm-yearswith available data

% of Firms withmechanism or meanvalue

Panel B: Internal environmental mechanismse

All observations

Total Quality Environmental Management 364 68%Percent of US Facilities that are Environmentally

Audited364 64%

US Environmental Standards used Overseas 364 41%Reporting Level for Environmental Officerf 364 1.30

(Continued on next page)


Table 2 (continued)

Mechanism No. of firm-yearswith available data

% of Firms withmechanism or meanvalue

Firms that include environmental performance in compensation (ENV = 1)

Total quality environmentalmanagement

247 72%*

Percent of US facilities that areenvironmentally audited

247 69%**

US environmental standards usedoverseas

247 46%**

Reporting level for environmentalofficerf

247 1.34*

Firms that do not include environmental performance in compensation (ENV = 0)

Total quality environmentalmanagement

117 60%

Percent of US facilities that areenvironmentally audited

117 54%

us environmental standards usedoverseas

117 32%

Reporting level for environmentalofficerf

117 1.22

a Toxic releases normalized by domestic sales.b Number of chemical and oil spills normalized by domestic sales.c Total penalties imposed on the firm under nine environmental statutes normalized by domestic

sales.d We assume the existence of an environmental performance component of executive compen-

sation (i.e., ENV = 1) suggests that the CEO has the opportunity to improve the firm’s environ-mental performance (referred to as a NON-MAX firm in Section 2), whereas the CEO of anENV = 0 firm does not have the opportunity to improve the firm’s environmental performance(referred to as a MAX firm in Section 2).

e As reported in the IRRC Environmental Performance Directory.f We code this variable as one if CEO or board level, two if Vice President level, and three if lower

than Vice President level.**,* Denote that value is significantly different from ENV = 0 sub-sample at the 1% and 5% levels,respectively.*,** Denote that means are significantly different from ENV = 0 sub-sample at the 1% and 5%levels, respectively.


Panel A of Table 2 reports descriptive statistics for financial data and theenvironmental performance variables. For comparison purposes, we providedescriptive statistics for all firms as well as separately for firms that use envi-ronmental performance as a determinant of executive compensation and forfirms that do not. Panel A reveals that the mean total assets and sales forfirms that include environmental performance in executive compensationare $18,075 million and $13,851 million, respectively, whereas the mean total

Table 3Pearson correlations among the independent variables (N = 364)

Variable ENV ROA RTN EXP COMPLY SPILL TENURE BETA STD CEOMV

ENV –ROA .035 –RTN �.025 .061 –EXP .180** .031 �.020 –COMPLY .067 .004 .036 .005 –SPILL .210** �.056 .032 .233** .543** –TENURE .067 �.028 �.001 �.072 .078 �.054 –BETA �.074 .017 �.087 .053 �.028 �.132* .165** –STD .193 �.012 �.104* .067 .031 .048 .119* �.467** –CEOMV .033 .015 �.053 �.016 �.008 �.028 .012 �.049 �.054 –CEOIND .143** .117* �.019 .030 �.077 .061 .101 .161** .001 .022

The variables are: ENV is a dichotomous variable that equals one if the firm uses environmental performance in determining executive compensation,zero otherwise; ROA is return on assets, defined as net income divided by end of year total assets; RTN is stock market returns; EXP is transformedratio of IRRC Efficiency Index (Toxic Releases/Sales) relative to best in the industry; COMPLY is transformed ratio of IRRC Compliance Index ($Penalties/Sales) relative to best in the industry; SPILL is transformed ratio of IRRC Spill Index (Number of Chemical and Oil Spills/Sales) relative tobest in the industry; TENURE is the number of years CEO is in position; BETA is the company beta from CRSP; STD is the standard deviation ofmarket returns based upon prior 24–60 months of data; CEOMV is the market value of CEO stock holdings; and CEOIND is the industry averageCEO salary and bonus.**,* Denote that correlation is significant at the 1% and 5% levels, respectively.

K.

Ca

mp

bell

eta

l./

Jo

urn

al

of

Acco

un

ting

an

dP

ub

licP

olicy

26

(2

00

7)

43

6–

46

2453


assets and sales for firms that do not are $9981 million and $8146 million,respectively. Difference-in-means tests reveal that the average total assetsand sales are significantly different for these two sub-samples. Moreover,the two sub-samples also have significantly different environmental perfor-mance measures. Specifically, firms that explicitly consider environmental per-formance when setting executive compensation have higher average toxicreleases (p < .01) and more chemical and oil spills (p < .01); however, thereis no significant difference between the groups with respect to compliance withenvironmental laws.

Panel B of Table 2 reports descriptive statistics about internal environmen-tal mechanisms. Firms that compensate based on environmental performanceare more likely to employ total quality management (p < .01), conduct environ-mental audits of their facilities (p < .05), and employ US-based environmentalstandards overseas (p < .05). These firms also have a higher reporting level fortheir top environmental officer (p < .01).

Table 3 reports correlation coefficients for independent variables used in theregression analyses. None of the variables exhibit high correlation, althoughthere is a significantly positive correlation (p < 0.01) between the ENV indica-tor variable and the toxic release and spill environmental metrics (EXP andSPILL). In addition, SPILL is significantly correlated with the two otherenvironmental performance measures, EXP and COMPLY.

5. Results

5.1. Tests of hypotheses

We report summary regression statistics from the 3SLS estimation of Eqs.(1a) and (1) in the Model 1 column of Table 4. The Eq. (1a) results suggestas the reporting level of the firm’s environmental officer becomes closer tothe CEO level, the more likely it is to include environmental performance asa determinant of CEO compensation. Moreover, the percentage of US facilitiesthat are environmentally audited and CEO tenure are positively related to theprobability that the firm implements an environmental performance-basedcompensation plan. Finally, two of our environmental performance metrics(EXP and SPILL) are positively related to ENV. Recall that larger values ofEXP and SPILL imply that the firm’s environmental performance lags behindindustry best practices. Hence, the results suggest that as environmental perfor-mance decreases, the firm is more apt to base part of its CEO’s compensationon environmental performance.

The results from Eq. (1) show that, after controlling for financial risk andother determinants of CEO compensation, the coefficient on EXP is positiveand significant (p < .01), supporting the existence of an environmental riskpremium. This finding provides support for our conjecture that CEOs are

Table 4Association between CEO compensation and environmental exposure (based on toxic releases)using three-stage least squares estimation procedure

Logit:

ENVit ¼ a0 þ a1ROAit þ a2RTNit þ a3TQMþ a4ENVREPit þ a5TENUREit

þ a6USAUDit þ a7USSTDit þ a8EXPit þ a9SPILLit þ a10COMPLYit þ cit ð1aÞ

Regression of interest:

COMPit ¼ b0 þ b1ENVit þ b2EXPit þ b3ðENVit � EXPitÞ þ b4ASSETit þ b5ROAit

þ b6RTNit þ b7BETAit þ b8STDit þ b9TENUREit þXY�1

y¼1

b10yYRyit

þXJ�1

j¼1


Variable Prediction Model 1 Model 2

Coefficient t-Statistic Coefficient t-Statistic

Logit (Eq. (1a)):

Intercept 0.158 1.24 0.157 1.23ROA 0.161 0.35 0.165 0.36RTN �0.023 �0.30 �0.025 �0.32TQM 0.091 1.83* 0.077 1.54ENVREP 0.097 2.34** 0.106 2.54**

TENURE 0.008 2.36** 0.008 2.35**

USAUD 0.003 4.04*** 0.003 4.20***

USSTD 0.077 1.60 0.069 1.44EXP 0.011 2.42** 0.011 2.50**

SPILL 0.018 2.91*** 0.018 2.90***

COMPLY �0.000 �1.01 �0.000 �0.99

Regression of interest (Eq. (1)):

Intercept 7.700 31.55*** �3.954 �1.97**

ENV �0.104 �0.67 �0.89 0.55EXP + 0.078 4.14*** 0.074 4.07***

ENV*EXP � �0.070 �3.69*** �0.66 �3.60***

ASSET + 0.181 8.12*** 0.192 8.93***

ROA + 3.079 7.00*** 2.786 6.50***

RTN + 0.066 0.95 0.063 0.94BETA 0.444 3.13*** 0.169 1.17STD �0.458 �3.82*** �0.292 �2.46**

TENURE + 0.012 4.13*** 0.014 4.87***

CEOMV – – 0.000 0.25CEOIND + – – 1.436 5.84***

N 364 364System R2 .49 .52

(Continued on next page)


Summary regression statistics for industry and year indicator variables have been omitted from thetable.The variables are: COMP is the natural logarithm of compensation; ENV is a dichotomous variablethat equals one if the firm uses environmental performance in determining executive compensation,zero otherwise; EXP is the transformed ratio of IRRC Efficiency Index (Toxic Releases/Sales)relative to best in the industry; ROA is the return on assets, defined as net income divided by end ofyear total assets; RTN is the stock market returns; TENURE is the number of years CEO is inposition; ASSET is the natural logarithm of total assets; BETA is the company beta from CRSP;STD is the standard deviation of market returns based upon prior 24–60 months of data; CEOMVis the market value of CEO stock holding; CEOIND is the industry average salary and bonus; YRy

is a dichotomous variable that equals one if the observation is from year y, zero otherwise; andINDj is a dichotomous variable that equals one if the firm operates in industry j (as defined by two-digit SIC code), zero otherwise.***,**,* Coefficient is significantly different from zero at the 1%, 5%, and 10% levels (one-tailed testsfor predicted signs), respectively.

Table 4 (continued)


compensated for taking on personal risk associated with the failure to achievethe best-practice level of environmental performance.20

Consistent with our prediction set forth in hypothesis 2, when environmen-tal performance is used as a determinant of executive compensation, thecoefficient on the incremental proxy for environmental exposure is significantlynegative (p < 0.01). This finding is consistent with our assertion that the exis-tence of an environmental performance component of CEO compensationindicates that the CEO’s personal environmental exposure risk is within her/his control and, therefore, the firm is able to pay less of an environmental expo-sure premium.

Although the signs are as predicted, the absolute values of the coefficients onthe two environmental exposure variables (EXP and ENV*EXP) are not statis-tically different from each other. This finding suggests that, on average, theenvironmental exposure premium is completely eliminated when environmen-tal performance is a determinant of compensation.

Turning to our control variables, firm size (ASSET) and profitability (ROA)are positively associated with compensation. Consistent with prior research(e.g., Davila and Venkatachalam, 2004), the coefficient on our proxy forCEO power and the quality of the CEO (TENURE) is positive and significant.Finally, our two control variables for the financial risk component are

20 Given the findings of Smith and Watts (1992), we also consider the impact of leverage andgrowth options on our results. While the results are no longer consistent with expectations when weincorporate proxies for growth opportunities (market-to-book) and leverage (debt-to-equity) in ouranalyses, it is important to note that our sample size decreases by approximately one-third.Moreover, the correlations for the market-to-book and debt/equity ratios with our environmentalexposure proxy, EXP, are not significant at conventional levels. Therefore, reduced power mayexplain the lack of robustness to this sensitivity check.


significantly related to CEO compensation. Specifically, the coefficient on thefirm’s beta (standard deviation of returns) is positive (negative).

Due to data constraints, our measure of compensation does not includestock-based compensation. In addition, our proxy for environmental exposure(EXP) is industry-adjusted, while our compensation variable is not. As such,we expand Eq. (1) to include two additional control variables: CEOMV andCEOIND. CEOMV is the market value of the CEO’s stockholdings and CEO-IND is the average salary and bonus of the other CEOs employed by firms inthe same two-digit SIC code. Model 2 in Table 4 presents the results from the3SLS estimation of this augmented system of equations. Of most importance,the coefficients on our variables of interest still have the predicted signs andremain statistically significant; namely EXP is positively related to COMP,while ENV*EXP is negatively associated with COMP.21

5.2. Alternative environmental performance-based exposure proxies

The internal validity of our results rests on the assumption that our TRI-based measure reflects the CEO’s environmental risk. As a robustness check,we investigate an alternative environmental performance-based metric. In par-ticular, we replace EXP in Eq. (1) with the industry best practice-adjustedIRRC environmental compliance index (COMPLY). In contrast to TRI emis-sions, this metric represents negative environmental outcomes with immediateeconomic consequences, and may not represent a reasonable metric of ex-anteenvironmental risk. For example, regardless of the quality of environmentalmanagement and control systems, fines may still occur. Similarly, the fact thata fine was incurred this year may not suggest a greater probability that a finewill be incurred next year. Nevertheless, this metric is based on objectivelymeasurable environmental outcomes, and, hence, may reflect some aspects ofenvironmental performance-related risk.

Table 5 reports the 3SLS regression results from this alternative specifica-tion. Consistent with the results reported in Table 4, COMPLY is positivelyrelated to compensation, but its coefficient is incrementally negative when envi-ronmentally-based compensation is used. This provides further support for anenvironmental exposure premium that is reduced when an environmental per-formance-based compensation plan is used.

21 Since levels regressions may be more susceptible to correlated omitted variables than theirchanges counterparts, we re-estimate Eq. (1) in changes. Although the unreported results do notprovide support for our hypotheses, it is important to note that our sample is reduced byapproximately one-third which, in turn, reduces the statistical power of our tests.

Table 5Association between CEO compensation and environmental exposure (based on environmentalpenalities) using three-stage least squares estimation procedure

Logit:

ENVit ¼ a0 þ a1ROAit þ a2RTNit þ a3TQMþ a4ENVREPit þ a5TENUREit

þ a6USAUDit þ a7USSTDit þ a8EXPit þ a9SPILLit þ a10COMPLYit þ cit ð1aÞ

Regression of interest:

COMPit ¼ b0 þ b1ENVit þ b2EXPit þ b3ðENVit�EXPitÞ þ b4ASSETit þ b5ROAit

þ b6RTNit þ b7BETAit þ b8STDit þ b9TENUREit þXY�1

y¼1

b10yYRyit

þXJ�1

j¼1


Variable Prediction Model 1 Model 2

Coefficient t-Statistic Coefficient t-Statistic

Logit (Eq. (1a)):

Intercept 0.160 1.25 0.159 1.24ROA 0.187 0.40 0.187 0.40RTN �0.025 �0.33 �0.026 �0.34TQM 0.084 1.62 0.078 1.50ENVREP 0.103 2.40** 0.106 2.48**

TENURE 0.008 2.36** 0.008 2.36**

USAUD 0.003 3.79*** 0.003 3.89***

USSTD 0.080 1.61 0.077 1.56EXP 0.010 2.16** 0.010 2.15**

SPILL 0.021 3.21*** 0.021 3.22***

COMPLY �0.000 �0.21 �0.000 �1.22

Regression of interest (Eq. (1)):

Intercept 7.488 31.54*** �3.012 �1.48ENV 0.001 0.01 0.012 0.09COMPLY + 0.003 2.58*** 0.002 1.68**

ENV*COMPLY – �0.003 �2.62*** �0.002 �1.72**

ASSET + 0.192 8.75*** 0.201 9.44***

ROA + 2.725 6.57*** 2.454 6.05***

RTN + 0.121 1.86* 0.113 1.79**

BETA 0.376 2.67*** 0.120 0.83STD �0.414 �3.52*** �0.261 �2.22**

TENURE + 0.011 3.86*** 0.012 4.53***

CEOMV – – �0.000 �0.11CEOIND + – – 1.298 5.17***

N 364 364System R2 .49 .52


Summary regression statistics for industry and year indicator variables have been omitted from thetable.The variables are: COMP is the natural logarithm of compensation; ENV is a dichotomous variablethat equals one if the firm uses environmental performance in determining executive compensation,zero otherwise; COMPLY is the transformed ratio of IRRC Compliance Index ($ Penalties/Sales)relative to best in the industry; ROA is the return on assets, defined as net income divided by end ofyear total assets; RTN is the stock market returns; TENURE is the number of years CEO is inposition; ASSET is the natural logarithm of total assets; BETA is the company beta from CRSP;STD is the standard deviation of market returns based upon prior 24–60 months of data; CEOMVis the market value of CEO stock holding; CEOIND is the industry average salary and bonus; YRy

is a dichotomous variable that equals one if the observation is from year y, zero otherwise; andINDj is a dichotomous variable that equals one if the firm operates in industry j (as defined by two-digit SIC code), zero otherwise.***,**,* Coefficient is significantly different from zero at the 1%, 5%, and 10% levels (one-tailed testsfor predicted signs), respectively.

Table 5 (continued)


6. Discussion and conclusion

We extend prior accounting research by empirically investigating the rela-tion between CEO compensation and a specific non-financial personal risk fac-tor. Our environmental performance context is unique because the costs andbenefits borne by firms and their executives are asymmetric, thereby allowingus to examine risk-induced tradeoffs in compensation. To execute this researchobjective, we first examine the relation between CEO compensation and thenon-financial risk imposed on CEOs due to environmental exposure. We defineenvironmental exposure as the risk of adverse consequences arising from afirm’s relatively poor environmental performance. To carry out our analyses,we develop a proxy for environmental exposure by comparing the firm’s envi-ronmental performance to an industry-specific best-practice benchmark. Sinceregulators are more likely to pursue CEOs under criminal penalty provisionswhen they have not exercised proper due diligence in managing environmentalperformance, CEOs bear more risk of personal liability when their firm is per-forming poorly relative to industry best-practice standards. Due to incrementalpersonal risk, we hypothesize CEOs will demand higher compensation whenother interests of the firm constrain their ability to reduce environmental expo-sure to the lowest feasible level (i.e., industry best practice or economic cost/benefit considerations), all other things being equal. Consistent with this con-jecture, we document a positive relation between environmental risk and CEOcompensation, after controlling for other determinants of compensation suchas financial risk.

Next, we investigate the impact of explicitly incorporating environmentalperformance into executive compensation. When environmental performanceis explicitly incorporated as a determinant of CEO compensation, we believethe existence of an environmental performance-based compensation compo-


nent is an indication that the CEO has opportunities to improve the firm’s envi-ronmental performance and, hence, lower her/his personal environmentalexposure risk. Since the CEO has control over her/his environmental risk,the firm is likely to pay less of an environmental exposure premium. After con-trolling for the explicit environmental component of compensation, we findsupport for our hypothesis that when an environmental performance-basedcompensation plan is in place, the environmental exposure premium is reduced.

This study is subject to limitations. First, our measure of executive compen-sation includes fixed salary as well as contingent bonuses. If CEOs are compen-sated for environmental risk via salary rather than bonuses (as standard agencytheory suggests), then our dependent variable is measured with error. However,as noted above, we are unable to separate the fixed and contingent componentsdue to data restrictions, effectively preventing us from conducting additionalrobustness tests. Second, we test our hypotheses by employing levels regres-sions, which may be susceptible to correlated omitted variables. Finally, webase our study on data from the sample period 1988–1991. Consequently, weare unable to comment on whether there still exists a compensation premiumfor environmental exposure. Notwithstanding these caveats, this paper pro-vides evidence that, for the period we studied, CEO compensation appearsto include a premium for environmental risk. Our results set the stage forresearchers to investigate other elements of non-financial risk which may bepart of executive compensation.

Acknowledgments

The fourth author gratefully acknowledges the financial support of the Ful-bright Commission of Germany. We thank Paul Bahnson, David Balkin, WaltBlacconiere, Bob Bowen, John Core, Ilia Dichev, Michael Eames, Joe Fischer,Jim Frederickson, Herb Hunt, Jon Karpoff, Terry Mitchell, Joe Paperman,Jamie Pratt, Mark Peecher, Frank Selto, Terry Shevlin, D. Shores and workshopparticipants at Instituto de Empresa, Indiana University, Carnegie Mellon,University of Michigan, University of Oregon, University of Pittsburgh, Uni-versity of Washington, University of California Davis, University of Coloradoat Denver, Arizona State University, University of Maryland, and NorwegianSchool of Management for their helpful comments. We also acknowledge con-tributions of the two anonymous reviewers and the Journal Editor.

References

Adams, H., 1987. Factors Affecting the Use of Performance Variables in Executive CompensationContracts. Northwestern University Graduate School of Management, Evanston, IL.

Antle, R., Smith, A., 1986. An empirical investigation of the relative performance evaluation ofcorporate executives. Journal of Accounting Research 24 (1), 1–39.


Banker, R.D., Potter, G., Srinivasan, D., 2000. An empirical investigation of an incentive plan thatincludes nonfinancial performance measures. The Accounting Review 75 (1), 65–92.

Barth, M., Epstein, M., Stark, R., 1994. Polaroid: Managing Environmental Responsibilities andtheir Costs. Harvard Business School Case #9-194-052. Harvard Business School, PublishingDivision, Boston, MA.

Bushman, R., Indejikian, R., Smith, A., 1996. CEO compensation: The role of individualperformance evaluation. Journal of Accounting and Economics 21 (2), 161–193.

Clinch, G., 1991. Employee compensation and firms’ research and development activity. Journal ofAccounting Research 29 (1), 59–78.

Combs, J.G., Skill, M.S., 2003. Managerialist and human capital explanations for key executivepay premiums: A contingency perspective. Academy of Management Journal 46 (1),63–73.

Core, J., Guay, W., 1999. The use of equity grants to manage optimal equity incentive levels.Journal of Accounting and Economics 28 (2), 151–184.

Core, J., Guay, W., Verrecchia, R., 2003. Price vs. non-price performance measures in optimalCEO compensation contracts. The Accounting Review 78 (4), 957–981.

Davila, A., Venkatachalam, M., 2004. The relevance of non-financial performance measures forCEO compensation: Evidence from the airline industry. Review of Accounting Studies 9 (4),443–464.

Duru, A., Reeb, D.M., 2002. Geographic and industrial corporate diversification: The level andstructure of executive compensation. Journal of Accounting, Auditing and Finance 17 (1), 1–24.

Ely, K., 1991. Interindustry differences in the relation between compensation and firm performancevariables. Journal of Accounting Research 29 (1), 37–58.

EPA Journal, 1995. Release of toxic chemicals down in’93. 21 (2), 4–5.Epstein, M., 1996. Measuring Corporate Environmental Performance: Best Practices for Costing

and Managing an Effective Environmental Strategy, first ed. Irwin, Chicago, IL.Garen, J.E., 1994. Executive compensation and principal-agent theory. Journal of Political

Economy 102 (6), 1175–1199.Hess, G., 1995. TRI estimates are focus of agency scrutiny. Chemical Marketing Reporter 247 (18),

24.Hunter, D., 1993. Responsible care: Monsanto. Chemical Week 153 (1), 106–107.Ittner, C., Larcker, D., Rajan, M., 1997. The choice of performance measures in annual bonus

contracts. The Accounting Review 72 (2), 231–255.Jin, L., 2002. CEO compensation, diversification, and incentives. Journal of Financial Economics

66 (2), 29–63.Kolluru, R. (Ed.), 1994. Environmental Strategies Handbook: A Guide to Effective Policies and

Practices. Prentice Hall, New York, NY.Lambert, R., Larcker, D., 1987. An analysis of the use of accounting and market measures of

performance in executive compensation contracts. Journal of Accounting Research 25 (3), 85–125.

Lanen, W., 1995. Waste Minimization at 3M Corporation: A Case Study in PerformanceMeasurement. University of Michigan Business School, Ann Arbor, MI.

Larcker, D.F., Rusticus, T.O., 2005. On the use of instrumental variables in accounting research.Working paper, University of Pennsylvania.

Murphy, K., 1985. Corporate performance and managerial remuneration: An empirical analysis.Journal of Accounting and Economics 11 (1–3), 11–42.

Murphy, K., 1986. Incentives, learning, and compensation: A theoretical and empirical investi-gation of managerial labor contracts. The Rand Journal of Economics 17 (1), 59–76.

Portney, P. (Ed.), 1990. Public Policies for Environmental Protection. Resources for the Future,Washington, DC.


Prendergast, C., 2002. The tenuous trade-off between risk and incentives. Journal of PoliticalEconomy 110 (5), 1071–1102.

Price Waterhouse, 1995. Progress on the Environmental Challenge.Roulstone, D.T., 2003. The relation between insider-trading restrictions and executive compen-

sation. Journal of Accounting Research 41 (3), 525–551.Scannell, K., 2006. SEC Rules on Fuller Disclosure for Corporate Pay Clear Hurdle. Wall Street

Journal, January 18.Sloan, R., 1993. Accounting earnings and top executive compensation. Journal of Accounting and

Economics 16 (1–3), 55–100.Smith, C.W., Watts, R.L., 1992. The investment opportunity set and corporate financing, dividend,

and compensation policies. Journal of Financial Economics 32 (3), 263–292.Stimson, J., Kimmel, J., Rollin, S., 1993. Guide to Environmental Laws from Premanufacture to

Disposal. Bureau of National Affairs, Washington, DC.Unites States Environmental Protection Agency (USEPA), 1999. 1999 Toxic Release Inventory

Public Data Release Data Report. EPA Document No. 260-R-01-001. Office of EnvironmentalInformation, Washington, DC.

Documents

Executive compensation and non-financial risk: An empirical examination