An investigation of regulatory and voluntary environmental capital expenditures

voluntary components. I then examine the relations of regulatory and voluntary envi-

* Corresponding author. Tel.: +970 491 5102; fax: +970 491 2676.

E-mail address: [email protected]

Journal of Accounting and Public Policy 24 (2005) 175206

www.elsevier.com/locate/jaccpubpolronmental capital outlays with future abnormal earnings, stock prices, and stock

returns. As predicted, the empirical analysis reveals that regulatory environmental cap-

ital expenditures are negatively associated with future abnormal earnings. Moreover,

market-based tests indicate that the regulatory component of environmental capital

expenditures is negatively priced. Finally, the results suggest that voluntary environmen-

tal capital expenditures and regulatory environmental capital expenditures have dier-

ent rm-specic economic consequences.

2005 Elsevier Inc. All rights reserved.

Keywords: Environmental expenditures; Environmental regulation; Voluntary eorts; Market

valuationAn investigation of regulatory andvoluntary environmental capital expenditures

Derek Johnston *

Colorado State University, College of Business, Rockwell Hall, Room 256,

Fort Collins, CO 80523, USA

Abstract

This paper investigates the rm-specic economic consequences of regulatory and

voluntary environmental capital expenditures. Using rm-level environmental data, I

decompose total environmental capital expenditures into estimates of regulatory and0278-4254/$ - see front matter 2005 Elsevier Inc. All rights reserved.doi:10.1016/j.jaccpubpol.2005.03.002

1. Introduction

This paper investigates the rm-specic economic consequences of regula-

tory and voluntary environmental capital expenditures. Regulatory environ-

mental expenditures are costs incurred to comply with federal, state, and

local environmental regulations whereas voluntary environmental expendituresare costs that a rm incurs to exceed compliance (USEPA, 1995a, p. 35). To

assess the economic consequences of these two components of environmental

capital expenditures, I examine the relations of regulatory and voluntary envi-

ronmental capital outlays with future abnormal earnings, stock prices and

stock returns.

Regulatory environmental capital expenditures are made in response to

176 D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206environmental legislation that is typically command-and-control in nature.

Command-and-control based regulations specify the processes that rmsshould adopt in order to attain the goals of the legislation (Jae and Palmer,

1997, p. 610). Consequently, this type of legislation fails to encourage innova-

tion at the rm level and may force rms to construct inecient pollution con-

trol facilities, thereby adversely aecting rm productivity (Barbera and

McConnell, 1990, pp. 5762; Boyd and McClelland, 1999, p. 139). In their

annual reports and 10-K disclosures, companies often acknowledge that regu-

latory environmental capital expenditures fail to produce future economic ben-

ets.1 In other words, rms claim that capital expenditures to comply withenvironmental regulations may represent negative net present value projects.2

If these claims are true, then I should nd that: (a) regulatory environmental

capital expenditures (RECAP) are negatively related to future abnormal earn-

ings; and (b) the capital market values these costs as expenses (assuming mar-

ket eciency).

Conversely, due to their exibility, voluntary environmental initiatives may

lead to the design of innovative pollution control techniques, resulting in pro-

1 For example, in its 1999 10-K ling, Kerr McGee states: . . .most (regulatory) environmental

expenditures provide no signicant increases in production capacity, eciency, or revenue. In its

1990 10-K ling, Temple-Inland acknowledges that capital expenditures made to comply with

environmental legislation may adversely aect earnings because these projects . . .provide

minimal, if any, monetary return on investment, and may divert capital from income producing

activities.2 Regulatory environmental capital expenditures can still represent negative NPV projects even

though rms may be subject to hefty environmental nes if they do not undertake these capital

projects. For instance, if a manager is faced with a new environmental regulation that impose a

capital project on the rm, she/he essentially has two choices: compliance (i.e., the rm undertakes

the regulatory environmental capital project) or non-compliance (i.e., the rm does not undertake

the capital project and possibly incurs environmental nes). In this scenario, the NPV of theregulatory capital project could be negative and still be accepted by the manager because its NPV is

closer to zero than the non-compliance alternative.

cess improvements (Khanna and Damon, 1999, pp. 14). In fact, prior research

suggests that voluntary environmental activities increase future revenues and

decrease future compliance costs (Epstein, 1996, p. 19). Thus, I predict that

voluntary environmental capital expenditures (VECAP) are positively associ-

ated with future abnormal earnings and market value.

To test these predictions, the empirical analysis consists of three stages.First, I decompose total environmental capital expenditures into estimates of

regulatory and voluntary components. Specically, I model total environmen-

tal capital expenditures as a function of rm-level environmental data from

the Investor Responsibility Research Centers Corporate EnvironmentalProles Directories and the Right-to-Know Networks Toxic Release Inven-tory Database. The variables constructed based on these data are the

assumed drivers of the regulatory and voluntary components of total environ-

mental capital expenditures. Accordingly, the sum of the products of the esti-mated coecients from this rst-stage regression multiplied by the rm-level

values for the corresponding regulatory (voluntary) variables provides esti-

mates of RECAP (VECAP). However, data availability severely constrains

the set of voluntary cost drivers. As such, I also consider an alternative proxy

for voluntary environmental expenditures. In particular, I regress total

environmental capital expenditures on the regulatory cost drivers and utilize

the estimated residuals from this regression as proxies for the voluntary

expenditures.In the second stage of the analysis, I provide evidence concerning the future

economic benets of regulatory and voluntary environmental capital expendi-

tures by examining their associations with one-year ahead abnormal earnings.

Finally, I include market-based tests that investigate the equity valuation of

RECAP and VECAP.

The results from the empirical analysis are mixed. As predicted, I nd that

regulatory environmental capital expenditures are negatively associated with

future abnormal earnings, stock prices, and stock returns. Moreover, I nd apositive relation between voluntary environmental capital expenditures and

future abnormal earnings when I use the estimated residuals from the rst-

stage regression as proxies for the voluntary environmental costs. Also, the

results suggest that the association between voluntary environmental capital

outlays and one-year ahead abnormal earnings is signicantly dierent than

the association between regulatory environmental capital costs and future

abnormal earnings. Finally, although the relation between voluntary environ-

mental capital expenditures and market value is not signicant, the capitalmarket appears to value the regulatory and voluntary components of environ-

mental capital expenditures dierently.

This paper may be of interest to regulators and academics. Palmer et al.

D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206 177(1995, pp. 130131) note that annual spending to protect the environment in

the United States is at least $100 billion. However, empirical research analyzing

the benets of environmental expenditures has been scarce. This paper con-

tributes to the literature by providing an increased understanding of the

rm-specic consequences associated with rms regulatory arid voluntaryenvironmental expenditures. In particular, the results appear to bolster the

claims that regulatory environmental capital expenditures fail to produce eco-

nomic benets to corporations, while capital expenditures made in response tovoluntary environmental activities provide greater future benets than regula-

and Spicer, 1983, pp. 530534; Klassen and McLaughlin, 1996, p. 1208; Konar

and Cohen, 2001, pp. 286288) and its exposure to regulatory environmental

178 D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206costs (Barth and McNichols, 1994, pp. 198199; Shane and Spicer, 1995, p.

503; Blacconiere and Northcut, 1997, p. 170; Campbell et al., 1998, pp. 357

358; Hughes, 2000, pp. 219220). In addition, some studies document a signif-

icant association between environmental performance and rm protability(Russo and Fouts, 1997, p. 548).3

Of the previous environmental research, this paper most closely relates to a

study by Clarkson et al. (2004, pp. 341350) who investigate the capital market

pricing of environmental capital expenditures by pulp and paper companies.

3 Other empirical studies (e.g., Chen and Metcalf, 1980, p. 174; Jaggi and Freedman, 1992, p. 707)

fail to document signicant associations between environmental performance and nancial

performance. Konar and Cohen (2001, p. 282) attribute these conicting results to insucient

sample sizes and lack of objective environmental performance measures. Consistent with the latter

explanation, Ilinitch et al. (1998, p. 386) acknowledge that there are no generally accepted

standards for measuring environmental performance. Further, Ilinitch et al. (1998, pp. 403405)tory projects. Moreover, since the results suggest that there are dierent payos

to the two types of environmental expenditures, requiring disclosures by type

may provide more information to market participants.

The remainder of this paper is organized as follows. Section 2 reviews

related literature and describes the nature of rms environmental expenditures.Section 3 develops the hypotheses. Section 4 discusses the research design, sum-

marizes the sample selection process, and reports descriptive statistics. Section5 presents the empirical results and Section 6 concludes.

2. Related research and environmental expenditures

2.1. Related research

Prior research examines the impacts of environmental information on rmvalue and nancial performance. A number of these studies report that a firmsstock price includes an assessment of its environmental performance (Shanewarn that existing environmental performance measures are inconsistent and capture dierent

dimensions of environmental performance and, therefore, may be confusing to stakeholders.

audits, voluntarily reduce emissions, implement recycling programs, produce

environmental annual reports, and participate in voluntary programs spon-

sored by the EPA and various industry groups.

From a nancial reporting perspective, voluntary and regulatory environ-

mental expenditures consist of both operating expenses and capital expendi-

tures. For example, voluntary environmental operating expenses may include

environmental audit fees and the costs incurred to produce an environmental

annual report. Certain Superfund remediation costs, hazardous waste disposal

costs, and any nes/penalties levied under environmental regulations are exam-

ples of regulatory environmental operating expenses. Voluntary environmentalcapital expenditures could include the cost of pollution abatement equipment

needed to meet the goals of voluntary programs. For example, Eastman KodakFor a sample of 29 rms, Clarkson et al. (2004, pp. 342345) report that the

markets valuation of environmental capital expenditures depends on rmsenvironmental performances. Most notably, environmental capital expendi-

tures are positively related to market value for low-polluting rms, but not

for high-polluting rms (Clarkson et al., 2004, pp. 342345). This paper ex-

tends Clarkson et al. (2004) by decomposing total environmental capital expen-ditures (for a larger sample of rms across many industries) into regulatory and

voluntary components and analyzing the associations of these two components

of environmental capital expenditures with future abnormal earnings, stock

prices, and stock returns.

2.2. Environmental expenditures

This paper considers two components of environmental expenditures: a reg-ulatory component and a voluntary component. Regulatory expenditures con-

stitute a signicant portion of total environmental outlays and include the

substantial costs related to compliance with federal environmental legislation

(Hamner and Stinson, 1995, pp. 57). For example, the Superfund Act and

the Resource Conservation and Recovery Act (RCRA) impose signicant

remediation and hazardous waste disposal costs on business entities. Addition-

ally, the Clean Air and Water Acts (CAA and CWA, respectively) require reg-

ulated corporations to install mandated pollution abatement equipment. Theappendix provides more details concerning these and other environmental

regulations.

Voluntary environmental costs are expenditures that a rm incurs to exceed

compliance in an eort to enhance its corporate image and improve its environ-

mental performance (USEPA, 1995a, p. 9). Examples of voluntary environ-

mental expenditures include the costs incurred to conduct environmental

D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206 179installed a closed loop dichloromethane recovery system to reduce its emissions

of the chemical in an eort to meet the goal of EPAs voluntary 33/50 program

(USEPA, 1994, p. 4).4 Finally, the cost of pollution abatement equipment to

reduce sulfur dioxide emissions as required by the CAA amendments consti-

tutes a regulatory capital expenditure.

3. Hypothesis development

pollutants by 33% before the end of 1992 and by 50% before the end of 1995.5 Although environmental regulations rarely force rms to adopt the recommended pollution

180 D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206control technology, companies often do so to ease the obtainment of environmental permits and

reduce regulatory scrutiny (Jae and Palmer, 1997, p. 610).6 Exceptions to command-and-control type legislation are evident in the 1990 Clean Air Act

Amendments. As noted by Hughes (2000, p. 213) the CAA Amendments allowed utilities toSince they are classied as assets on rms balance sheets, both regulatoryand voluntary environmental capital expenditures should generate subsequent

economic benets (FASB, 1985, paragraph 25). Analyzing capital expenditures

in the context of abnormal earnings provides a framework that allows the

researcher to discern whether expenditures that are recorded as assets generate

earnings in excess of those implied by the discount rate. In particular, a positive

(negative) relation between environmental capital expenditures and futureabnormal earnings suggests that the capital expenditures generate earnings

greater than (less than) those implied by the discount rate.

While regulatory capital expenditures are necessary to comply with environ-

mental legislation, such capital projects may not facilitate ecient attainment

of regulatory goals due to the type of legislation that is in place. Typically, US

environmental regulations are command-and-control in nature rather than per-

formance-based, meaning they prescribe both the goals of the legislation as

well as the processes for attaining those goals (Jae and Palmer, 1997, p.610).5 This type of legislation does not encourage innovation at the rm level

and may force rms to redesign their production processes in an inecient

manner. For instance, Barbera and McConnell (1990, pp. 5762) and Boyd

and McClelland (1999, pp. 132139) provide evidence suggesting that pollution

abatement capital projects adversely aect plant productivity. Porter (1991, p.

168) and Palmer et al. (1995, p.120) acknowledge that regulatory standards

that aim at outcomes rather than methods (i.e., performance-based standards)

encourage companies to re-engineer their technology, resulting in processesthat not only pollute less but also lower costs and/or improve quality. How-

ever, the majority of regulatory environmental expenditures are made in

response to command-and-control type regulations.6 Due to the ineciencies

resulting from this type of legislation, capital expenditures made in response

4 The 33/50 program challenged member corporations to reduce the releases of 17 toxicexercise discretion over their strategy to comply with the mandatory reduction of sulfur dioxide

emissions imposed by the amendments.

to environmental regulations may not generate earnings in excess of those

implied by the discount rate. Consequently, I posit that regulatory environ-

mental capital expenditures are negatively related to future abnormal earnings

and that the market values these outlays as expenses:

Hypothesis 1: Regulatory environmental capital expenditures are negativelyassociated with future abnormal earnings.

Hypothesis 2: Regulatory environmental capital expenditures are negatively

associated with market value.

In contrast to regulatory capital projects, voluntary environmental projects

may lead to the design of innovative pollution control and waste reduction

techniques (Khanna and Damon, 1999, pp. 14). Prior research suggests that

rms may derive several benets from their voluntary environmental initia-tives, such as increasing future sales revenue by attracting green consumers,

preempting more costly (and less ecient) government legislation, and reduc-

ing future regulatory costs.

A 1990 study by Abt Associates suggests that many US consumers are will-

ing to pay a premium for environmentally sound products (Cairncross, 1992, p.

192). Voluntary environmental initiatives such as the voluntary reduction of

toxic releases likely result in cleaner and more ecient production processes.

Arora and Gangopadhyay (1995, pp. 293300) use a vertical product dieren-tiation model to show that, when consumers incomes increase, rms shouldvoluntarily reduce emissions in order to attract (or retain) environmentally

conscious consumers. Arora and Cason (1995, p. 283) and Khanna and

Damon (1999, p. 15) test this assertion empirically by analyzing rm participa-

tion in EPAs voluntary 33/50 program. Both studies nd that consumer-ori-ented rms are more likely to participate in the 33/50 program. These results

suggest that rms undertake voluntary environmental activities to enhance

relations with green consumers.Moreover, rms may voluntarily invest in environmental capital projects in

an eort to preempt the enactment of more costly, and less ecient, govern-

ment regulations. Due to their exibility, voluntary environmental initiatives

likely result in lower costs than would be incurred under mandated pollution

abatement legislation. Segerson and Miceli (1998, pp. 111127) develop a

model of the interaction between a regulatory agency and a polluting rm to

examine voluntary environmental programs. They nd that a sucient condi-

tion for participation in a voluntary program occurs when the polluting rm isthreatened by the imposition of legislation. Similarly, Maxwell et al. (2000, pp.

589595) construct a model that shows rms will voluntarily reduce emissions

when faced with the threat of regulation. They provide empirical support for

D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206 181their model by documenting a positive association between the likelihood of

the enactment of state pollution abatement legislation and the voluntary

reduction of statewide TRI emissions. By successfully preempting environmen-

tal legislation, voluntary environmental capital projects may reduce future

costs.7

Lastly, voluntary environmental activities may reduce future environmental

compliance costs. By investing in cleaner pollution control technology, rms

can reduce future operating and waste disposal costs.8 Additionally, participa-tion in voluntary programs may result in reduced nes and penalties due to

improved relations with regulators. Consistent with this assertion, Hemphill

(1993, pp. 145154) reports that courts often are more lenient on environmen-

182 D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206tal oenders that have undertaken voluntary initiatives.

In sum, prior research supports the notion that voluntary environmental

projects may increase future earnings. Although Khanna and Damon (1999,

p. 23) nd that participation in the EPAs voluntary 33/50 program is nega-tively associated with current period ROI, their analyses also reveal that par-ticipation in the program is positively related to the excess of market value

over book value. These results suggest that although voluntary environmental

activities impose costs that may not be fully oset in the current period, the

capital market expects that rms participating in voluntary environmental

programs will be more protable in the long run (Khanna and Damon,

1999, p. 22).

On the other hand, the investment in voluntary pollution control equipment

may adversely aect earnings. For example, the voluntary reduction of air andwaste-borne emissions may reduce a rms baseline emissions. Since futureenvironmental regulations may require a specied reduction from that baseline,

voluntarily reducing emissions before the mandated reduction may result in

more costly production process redesign costs since the most cost-eective

reductions would have already been made (USEPA, 1995c, p. 69). Ironically,

this ratchet eect penalizes companies that voluntarily overcomply with

environmental legislation and rewards companies that do the minimum

required by law (USEPA, 1995c, p. 69).Nevertheless, I assume that rms are aware of these risks and undertake vol-

untary environmental capital projects only if the expected benets of these pro-

7 Given that environmental regulations may target industries, a rm that is not voluntarily

overcomplying may free-ride o of other rms within its industry that are overcomplying and,

thus, still benet from the preemption of environmental legislation. Conversely, a single rm that

does not undertake voluntary initiatives may cause its industry to be targeted for legislation

regardless of the overcompliance by other rms in its industry. In this case, the relatively clean rms

would not derive the benets from the preemption of environmental legislation. Segerson and

Miceli (1998, p. 111) do not consider the complicating eects that free-riding may have on their

model.8 As one of its capital projects undertaken to meet the goal of EPAs voluntary 33/50 program,

Inland Steel installed new aqueous cleaning systems at one of its facilities at a cost of approximately$200,000. This eliminated the emission of tetrachloroethylene at the facility, saving Inland

approximately $37,500 in annual disposal costs (USEPA, 1995b, p. 9).

ECAP a a TRI AIR a TRI WATER a TRI SOLIDit 0 1 it 2 it 3 it a4NUMOILit a5NUMCHEMit a6VALPENit

a7SITESit a8ACTSit a9X3

j033 50itj1 33 50itj

XY1

y1a10yYRyit

XJ1

j1a11jINDjit it 1

i, t denote rm and year, respectively. The variables in Eq. (1) are dened asfollows: ECAP = total environmental capital expenditures (in millions ofjects outweigh their costs. Therefore, I hypothesize a positive relation between

voluntary environmental capital expenditures and future abnormal earnings. I

also examine whether the market impounds the subsequent benets associated

with these capital expenditures into rms share prices. Stated in alternativeform, this study tests the following two hypotheses:

Hypothesis 3: Voluntary environmental capital expenditures are positively asso-

ciated with future abnormal earnings.

Hypothesis 4: Voluntary environmental capital expenditures are positively asso-

ciated with market value.

4. Research design, sample and descriptive statistics

4.1. Research design

4.1.1. Measurement of regulatory and voluntary environmental capital

expenditures

Since rms do not make the distinction between the regulatory and volun-

tary components of environmental capital expenditures in their annual reportsand 10-K disclosures, I develop a model to estimate these costs. Specically, I

model total environmental capital expenditures as a function of rm-level envi-

ronmental data. The variables constructed based on these data are the assumed

drivers of the regulatory and voluntary components of total environmental

capital expenditures. Accordingly, the sum of the products of the estimated

coecients from this rst-stage regression multiplied by the rm-level values

for the corresponding regulatory (voluntary) variables provide estimates of reg-

ulatory (voluntary) environmental capital expenditures.In particular, I estimate the following model in pooled cross-section using

ordinary-least-squares (OLS):

D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206 183dollars); TRI_AIR = toxic chemical air emissions (in millions of pounds);

TRI_WATER = toxic chemical water emissions (in millions of pounds);

TRI_SOLID = toxic chemical solid waste emissions (in millions of pounds);

NUMOIL = number of reported oil spills (larger than 10,000 gallons); NUM-

CHEM = number of reported chemical spills (larger than 10,000 pounds);

VALPEN = sum of environmental penalties assessed under nine major envi-

ronmental laws (in millions of dollars); SITES = current number of Superfund

sites on the National Priority List (NPL) that the rm has been identied as apotentially responsible party; ACTS = current number of treatment, storage,

184 D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206and disposal facilities at which the EPA has required a corrective action;

33_50 = total emissions of the 17 toxic chemicals covered under the EPAs vol-untary 33/50 program, deated by rm sales; YRy = a dichotomous variable

that equals one if the observation is from year y, zero otherwise; and INDj = a

dichotomous variable that equals one if the observation operates in industry j

(as dened by two-digit SIC code), zero otherwise.

The rst eight independent variables represent the regulatory cost drivers(RCD). First, although they generally do not exceed legal constraints, I include

the rms toxic chemical emissions (TRI) in Eq. (1) since rms with higheremission levels are likely to incur higher regulatory pollution control costs

(IRRC, 1998). However, the type of chemical emission (e.g., air, water, or solid

waste) likely plays a role in the required regulatory environmental capital

expenditure.9 Thus, I split TRI into three separate cost drivers: (a) air emis-

sions (TRI_AIR); (b) water emissions (TRI_WATER); and c) solid waste emis-

sions (TRI_SOLID).10 Next, I assume that the number of oil and chemicalspills (NUMOIL and NUMCHEM, respectively) drive the capital expenditures

imposed by the Oil Pollution Act and other environmental regulations.11 The

dollar value of environmental penalties assessed against the rm (VALPEN)

provides a measure of the magnitude of compliance problems that a rm has

experienced under federal environmental regulations. Although nes and pen-

alties are expensed in the period incurred, companies may be required to under-

take supplemental environmental projects (SEPs) as a result of their violations.

Therefore, I include VALPEN to capture any capital costs associated withSEPs.

Moreover, Barth and McNichols (1994, p. 196) report that the number of

sites that a rm is identied as a PRP is an appropriate proxy for costs related

to the Superfund Act. I use the current number of sites (SITES) that are on the

9 For instance, a US Census Bureau publication reveals that of the $5.8 billion spent on pollution

abatement capital expenditures in 1999, $3.45 billion was related to air emissions, $1.8 billion to

water emissions, $0.35 billion to solid waste emissions, and $0.2 billion to multimedia.10 Since TRI emissions include emissions of both regulated and unregulated chemicals, TRI_AIR,

TRI_WATER, and TRI_SOLID may be noisy drivers of regulatory abatement costs.11 Since it is possible that rms incur environmental capital expenditure for oil and chemical spillsoccurring in previous years, I include year t 1 NUMOIL and NUMCHEM in Eq. (1). Thisalternative specication does not alter the tenor of the results that follow.

National Priority List since the Superfund process is lengthy and, therefore, it

is likely that rms are incurring costs for past violations of Superfund. Finally,

D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206 185the current number of facilities that the EPA has required the rm to undertake

a RCRA corrective action (ACTS) proxies for the capital costs associated with

the investigation and remediation of treatment, storage, and disposal facilities.

I assume the RCD variables are the cost drivers for the regulatory componentof environmental capital expenditures; so, I expect these variables to be posi-

tively associated with ECAP. To support the use of these cost drivers, Table

1 provides examples of environmental capital expenditures relating to each

driver.

The voluntary cost driver, 33_50, is based on the voluntary reduction of the

toxic chemicals covered under the EPAs 33/50 program, which began in 1988.Under this program, the EPA challenged rms to voluntarily reduce releases of

17 toxic pollutants by 33% by the end of 1992 and by 50% by the end of 1995.In response to the challenge, corporations such as Eastman Kodak and Inland

Steel initiated environmental capital projects to voluntarily reduce their emis-

sions of the 17 chemicals (USEPA, 1994, p. i; 1995b, p. i).

I obtain the total emissions of the 17 toxic chemicals covered under the 33/

50 program for each rm-year observation using publicly available data

from the Right-to-Know (RTK) networks TRI database. Since these emissionsare likely aected by rm size and production levels, I deate the 33_50 emis-

sions by the corresponding year rm sales. I then calculate the annual changein the 33/50 emissions by subtracting the current years size-adjusted emissionsfrom the previous years size-adjusted emissions. Thus, if a rms 33/50 emis-sions decrease relative to the previous year, then the annual change will be

positive.

I assume that rms incur environmental capital expenditures to voluntarily

reduce their emissions; therefore, I expect the annual change in the size-

adjusted 33/50 emissions to be positively associated with environmental capital

expenditures. However, it is likely that rms incur voluntary environmentalcapital expenditures in year t to reduce 33/50 emissions not only in year t

but also future years. Consequently, I aggregate the annual change in 33_50

across four time periods: the current year and the following three years. I

use the sum of these four annual changes in 33_50 as the voluntary cost driver

in Eq. (1).12

To mitigate heteroskedasticity as well as coecient bias resulting from

scale eects, I deate ECAP as well as the regulatory and voluntary cost

drivers by year t sales. Since Eq. (1) is estimated in pooled cross-section, Iinclude year and industry indicator variables to control for omitted12 The results that follow are not materially aected when I estimate Eq. (1) using only the year t

annual change in 33_50 as the voluntary cost driver.

186 D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206time- and industry-specic eects. I measure regulatory environmental capital

expenditures (RECAP) as the sum of the products of the rm-level values for

Table 1

Examples of regulatory and voluntary environmental capital expenditures

Cost driver Example of related environmental capital expenditure

TRI_AIR In its 1993 annual report, Pennzoil discloses that it spent $26.6 million on a

regulatory pollution control project required under the 1990 Clean Air Act

(CAA) Amendments. Of the 165 CAA-regulated pollutants, 153 are included

in a rms TRI emissionsTRI_WATER A 2000 study funded by the EPA estimates that regulatory capital expenditures

on end-of-pipe controls to comply with the Clean Water Act total $3.9 billion

per year (USEPA, 2000a, p. 8-1)

TRI_SOLID The estimated yearly costs to comply with the solid waste disposal requirements

under one subtitle program of the Resource Conservation and Recovery Act

are estimated at $503 million. Capital compliance expenditures include

retrotting existing disposal facilities (USEPA, 2000b, p. 5-5)

NUMOIL Texaco was required to install a leak detection system at one of its facilities at

a cost of $800,000 following an oil spill in 1993 (USEPA, 1995c, p. 116).

In the same year, Marathon Oil spent $265,000 to re-pipe some if its renery

fuel gas lines in response to an oil spill (USEPA, 1995c, p. 116)

NUMCHEM Texas Instruments was forced to replace its vapor degreaser at a cost of

$170,000 following an accidental release of Freon (USEPA, 1995d, p. 119)

VALPEN As a result of a 1993 violation of the Clean Air Act, Sinclair Oil was required

to not only pay a ne of $105,000 but also to undertake a SEP at a cost of

$270,000. The purpose of the SEP was to increase SO2 removal at one of

Sinclairs facilities (USEPA, 1995c, p. 116)SITES Following a violation of the Superfund Act, Scholle Corporation was required

to retrot underground storage tanks at a cost of $46,200 (USEPA, 1995e,

p. 101)

ACTS In addition to the corrective action remediation costs associated with one of its

treatment, storage, and disposal facilities, American Airlines was forced to

undertake a $385,000 capital project to install a chrome waste recovery system

(USEPA, 1998, p. 81)

D33_50 As one of its capital projects undertaken to meet the goal of EPAs voluntary33/50 program, Inland Steel installed new aqueous cleaning systems at one of

its facilities at a cost of approximately $200,000 (USEPA, 1995b, p. 9)

Variables are dened as follows: TRI_AIR = toxic chemical air emissions (in millions of pounds);

TRI_WATER = toxic chemical water emissions (in millions of pounds); TRI_SOLID = toxic

chemical solid waste emissions (in millions of pounds); NUMOIL = number of reported oil

spills (larger then 10,000 gallons); NUMCHEM = number of reported chemical spills (larger

than 10,000 pounds); VALPEN = the sum of environmental penalties assessed under nine

major environmental laws (in millions of dollars); SITES = the current number of Superfund

sites on the National Priority List (NPL) that the rm has been identied as a potentially

responsible party; ACTS = the current number of treatment, storage, and disposal facilities

at which the EPA has required a corrective action; and D33_50 = the annual change (year t 1minus year t) in the emissions of the 17 toxic chemicals covered under the EPAs voluntary 33/50program.

the RCD variables times their corresponding estimated coecients from

Eq. (1).13 Likewise, the rm-level value for the voluntary cost driver times its

estimated coecient from Eq. (1) proxies for the voluntary environmental

capital expenditures (VECAP).

4.1.2. Associations with future abnormal earnings

I test Hypotheses 1 and 3 by examining the relations between current year

regulatory and voluntary environmental capital expenditures and one-year

ahead abnormal earnings. In particular, I estimate the following model in

D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206 187pooled cross-section using OLS:

AEit1 b0 b1AEit b2CAPXit b3RECAPit b4VECAPit

XY1

y1b6yYRyit

XJ1

j1b7jINDjit mit1 2

The variables in Eq. (2) are dened as follows: AE = abnormal earnings;

CAPX = capital investment in property, plant, and equipment (net of the scalyears environmental capital expenditures); RECAP = estimated regulatoryenvironmental capital expenditures (tted values based on the RCD variables

and their related coecients from Eq. (1)); and VECAP = estimated voluntary

environmental capital expenditures (tted values based on the voluntary cost

driver and its related coecient from Eq. (1)).

All other variables are as previously dened. Eq. (2) is based on Dechow

et al. (1999, p. 15) and Myers (1999, p. 6) who model abnormal earnings as

a linear function of past abnormal earnings and other information. However,I augment the models developed in those research papers to investigate

how dierent types of capital expenditures aect abnormal earnings. Capital

expenditures, net of environmental capital expenditures, are included in Eq.

(2) to control for other capital outlays. I dene abnormal earnings as

AEt = Xt rBVt1, where X is earnings before taxes, interest, and depreciationin year t, r is the discount rate, and BV is book value of equity at the end of

year t 1.14

13 It is unclear whether the year and industry variables capture regulatory or voluntary

environmental costs (or a mixture of both). Therefore, I exclude these variables from the

computations of the regulatory and voluntary components.14 I exclude depreciation from earnings in Eq. (2) since capital expenditures are positively related

to depreciation expense, a negative component of earnings. Moreover, Myers (1999, pp. 911)

argues that the income eect of conservative accounting resulting from the use of accelerated

depreciation methods overstates depreciation expense, thereby reducing future abnormal earnings.

Also, I add back interest to earnings to avoid a negative mechanical relation between capital

expenditures and earnings due to nancing costs. Finally, I exclude tax expense to provide ameasure of earnings that more closely approximates operating income; however, including taxes in

abnormal earnings does not materially aect the results that follow.

Following Frankel and Lee (1998, p. 288) and Myers (1999, p. 15) I assign a

188 D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206discount rate to each observation based on industry membership (as dened by

SIC code) and the year of the observation. Specically, I compute the discount

rate as r(j, t) = rf(t) + rp(j), where r(j, t) is the estimated discount rate for indus-

try j in year t, rf(t) is the yearly t-bill return in year t, and rp(j) is the risk pre-

mium for industry j as estimated by Fama and French (1997, pp. 172173).15

All variables, except the year and industry indicators, are deated by year t

sales. I expect b3 to be negative and b4 to be positive. Moreover, I predict thatb4 will be greater than b3.

4.1.3. Market-based tests

To examine the equity valuation of regulatory and voluntary environmental

capital expenditures, I estimate the following regression in pooled cross-section

using OLS:

PRICEit c0 c1BVXPSit c2EPSit c3CAPXPSit c4RECAPSit

c5VECAPSit XY1

y1c6yYRyit

XJ1

j1c7jINDjit git 3

The variables in Eq. (3) are dened as follows: PRICE = share price three

months after the end of the scal year; BVXPS = book value of equity at the

end of the scal year, per share (net of the scal years capital investment inproperty, plant, and equipment); EPS = earnings before extraordinary items,

per share; CAPXPS = capital investment in property, plant, and equipment

(net of the scal years environmental capital expenditures), per share;RECAPS = estimated regulatory environmental capital expenditures, per

share; and VECAPS = estimated voluntary environmental capital expendi-

tures, per share.

All other variables are as previously dened. Eq. (3) is similar to specica-

tions in Barth and Clinch (1988, p. 207); Aboody et al. (1999, p. 165); and Blac-

coniere et al. (2000, p. 242) where book value and earnings per share are

included as summary measures of accounting information. I expect c4 to benegative and c5 to be positive. Concerning the dierential pricing of the volun-tary and regulatory components, I predict that c5 will be greater than c4.

Easton (1998, pp. 236242) notes that scale eects may aect inferences

from price-level regressions and, therefore, returns specications should be

used whenever possible. Moreover, levels regressions may be susceptible to cor-

related omitted variables (Skinner, 1996, pp. 394395; Aboody et al., 1999, p.

166). To address these concerns, I estimate the following equation:15 The mean (median) discount rate over the sample period is 10.30% (10.64%). The results that

follow do not dier materially when I use constant discount rates of 9%, 10%, 11%, and 12%.

RETXit d0 d1DBVXit d2DEBEIit d3DCAPXit d4DRECAPitXY1 XJ1

D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206 189 d5DVECAPit y1

d6yYRyit j1

d7jINDjit lit 4

The variables in Eq. (4) are dened as follows: RETX = annual stock return

(excluding dividends) as reported in CRSP, measured from nine months before

the end of the scal year through three months after the end of the scal year;

and EBEI = earnings before extraordinary items.

All other variables are as previously dened and D denotes annual change.Generally speaking, the returns equation is obtained by rst-dierencing the re-lated price-level regression (i.e., Eq. (3)). Easton and Harris (1991, pp. 2124),

show that the rst-dierence of book value of equity is current period earnings

(assuming clean surplus and adjusting returns for dividends). However, this

relation between earnings and book value is violated since book value is net

of capital expenditures in Eq. (3). As a result, I use the change in book value

(net of capital expenditures) as an independent variable in Eq. (4) instead of

the level of earnings and I exclude dividends from returns. Following prior

research (Aboody et al., 1999, p. 167), I deate all independent variables,except the year and industry indicators, by beginning market value of equity.

Consistent with the expectations for the price-level regression, I predict that

d4 will be negative and d5 will be positive. Moreover, I expect d5 to be greaterthan d4.

4.2. Sample selection and descriptive statistics

The sample selection process consists of several phases. First, I gather rm-level values for the RCD variables (except for the TRI variables) from the

Investor Responsibility Research Centers (IRRC) Corporate EnvironmentalProles Directories. For S&P 500 rms (and for S&P 1500 rms beginning

in 1999), IRRC collects environmental data from various databases maintained

by governmental agencies and reports a lagged time-series of these data for

each rm. To maximize the number of rm-year observations, I obtained the

1993, 1995, 1998, and 1999 directories, resulting in a sample period of 1988

1996.16

Next, I download chemical-specic TRI emission data from the RTK net-

works TRI database for each rm-year observation in the sample. Based onthis data, I compute the annual change in the yearly size-adjusted emissions

of the 17 toxic pollutants covered under the EPAs voluntary 33/50 program.

16 The sample period does not align with the years of the IRRC directories because the directoriesreport lagged rm-level values for the regulatory cost drivers. For example, the 1999 IRRC

directory contains the necessary environmental data from 1995 and 1996.

I then collect environmental capital expenditure data from rms 10-K lingsand annual reports.17 Environmental capital expenditures are available for

589 observations (107 rms).18 Finally, to be included in the estimation of

Eqs. (1)(4), observations must have the necessary data from Compustat and

CRSP. Sample sizes used in the estimation of these equations are shown in

their respective tables.Table 2 presents descriptive statistics associated with the environmental cost

drivers and various nancial variables.19 Table 2 reports that the average toxic

air, water, and solid waste releases are 5.57, 0.51, and 14.19 million pounds,

respectively. The mean (median) number of oil spills is 0.95 (0) and the mean

(median) number of chemical spills is 2.00 (1.0). Table 2 also indicates that

the mean (median) yearly value of penalties assessed against rms is $440,000

($40,000). On average, rms have been identied as potentially responsible par-

ties on 17.95 Superfund NPL sites and have been required to undertake 2.65RCRA corrective actions. Finally, the mean (median) annual reduction in the

emissions of the 17 33/50 chemicals is 167,000 (7000) pounds. Untabulated

190 D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206Pearson correlations among the voluntary and regulatory cost drivers reveal

that 14 of the 28 correlations among the regulatory cost drivers are positive

and signicant, whereas the voluntary cost driver is negatively correlated with

all of the regulatory cost drivers except air emissions, water emissions, and oil

spills.

Turning to the nancial variables, rms spend, on average, $67.6 million peryear on environmental capital expenditures. Unreported statistics reveal that,

on average, environmental capital expenditures represent 0.66% of sales and

7.95% of the yearly capital investment in property, plant and equipment.20

17 Typically, those sample rms that report environmental capital expenditures do so under a

section labeled Environmental Matters. Other sample rms disclose the amount of environ-

mental capital expenditures in their discussion of all capital expenditures.18 With the exception of the 1999 directory, IRRC collects information regarding the disclosure of

environmental capital expenditures from rms 10-K lings. I used these data to identify the rmsthat disclose environmental capital expenditures. I reviewed the 10-Ks and annual reports for a

subset of rms that, according to IRRC, did not disclose their environmental capital expenditures,

This search did not yield any additional observations.19 The tables report descriptive statistics and regression results based on data that have not been

ination adjusted. However, the empirical results are not materially aected when I adjust the data

for ination.20 These descriptive statistics suggest that environmental capital expenditures may generally be

immaterial. Joshi et al. (2001, pp. 188189) examine the extent to which accounting systems

correctly classify all costs of environmental compliance. Relying on plant-level data from 55 steel

mills, they use a translog cost function to estimate regulatory environmental costs that are

hidden in other accounts. They nd that a $1 increase in the visible cost of environmental

regulation (i.e., costs that are correctly classied as environmental) results in a $10$11 increase intotal cost at the margin, of which $9$10 is hidden in other accounts. Their results suggest that the

disclosed amount of environmental capital expenditures may proxy for much larger costs.

Table 2

Descriptive statistics

Variable N Mean Median Standard deviation

TRI_AIR 589 5.57 2.08 9.51

TRI_WATER 589 0.51 0.11 1.41

TRI_SOLID 589 14.19 2.47 36.68

NUMOIL 589 0.95 0.00 2.60

NUMCHEM 589 2.00 1.00 3.98

VALPEN 589 0.44 0.04 1.62

D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206 191For comparison purposes, I report descriptive statistics for rms environ-mental operating expenses (also collected from 10-Ks and annual reports).

Table 2 indicates that rms spend an average of $191.5 million per year

on environmental operating expenses. The results from the estimation of

the decomposition model (discussed in Section 5) yield a mean (median) regu-

latory environmental capital expenditure of $10.05 ($4.12) million and a

mean (median) voluntary environmental capital expenditure of $1,270,000

($120,000).In terms of the distribution of the sample rms across industries, Table 3

presents the frequency distribution, by two-digit SIC code, of the 107 rms

SITES 589 17.95 14.00 16.13

ACTS 589 2.65 1.00 4.05

D33_50 589 0.167 0.007 1.015ECAP 589 67.61 27.00 105.17

EOP 308 191.49 87.00 292.63

RECAP 589 10.05 4.12 15.19

VECAP 589 1.27 0.12 3.67

MV 562 10,435.37 4,261.24 17,814.23

AE 587 1635.86 563.41 3195.92

Variables are dened as follows: TRI_AIR = toxic chemical air emissions (in millions of pounds);

TRI_WATER = toxic chemical water emissions (in millions of pounds); TRI_SOLID = toxic

chemical solid waste emissions (in millions of pounds); NUMOIL = number of reported oil spills

(larger than 10,000 gallons); NUMCHEM = number of reported chemical spills (larger than 10,000

pounds); VALPEN = the sum of environmental penalties assessed under nine major environmental

laws (in millions of dollars); SITES = the current number of Superfund sites on the National

Priority List (NPL) that the rm has been identied as a potentially responsible party; ACTS = the

current number of treatment, storage, and disposal facilities at which the EPA has required a

corrective action; D33_50 = the annual change (year t 1 minus year t) in the emissions of the 17toxic chemicals covered under the EPAs voluntary 33/50 program (in millions of pounds);ECAP = total environmental capital expenditures (in millions of dollars); EOP = total environ-

mental operating expenses (in millions of dollars); RECAP = estimated regulatory environmental

capital expenditures (tted values based on the regulatory cost drivers and their related coecients

from Eq. (1)), in millions of dollars; VECAP = estimated voluntary environmental capital expen-

ditures (tted values based on the voluntary cost driver and its related coecient from Eq. (1)), in

millions of dollars; MV = market value three months after the end of the scal year (in millions of

dollars); and AE = abnormal earnings (in millions of dollars).

industries. This concentration is not surprising, as Levinson (1996, p. 25) notes

Table 3

Frequency distribution of observations (by two-digit SIC codes)

SIC code and industry Number

of rms

Number of

observations

Percentage of

observations in sample

2800: Chemicals 38 187 31.75

2900: Petroleum Rening 18 119 20.20

192 D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206that rms operating in these three industries typically spend more on environ-

mental capital expenditures (as a percentage of total new capital expenditures)

relative to rms that operate in other manufacturing industries.

5. Results

5.1. Measurement of regulatory and voluntary environmental capital(589 observations) that disclose environmental capital expenditures.21 Most

notably, Table 3 reveals that the majority of the sample observations

(70.0%) operate in the chemicals, petroleum rening, and primary metals

3300: Primary Metals 21 106 18.00

2600: Paper 14 75 12.73

2000: Food Products 2 17 2.89

3700: Transportation 2 12 2.04

3600: Appliances 2 12 2.04

3500: Machinery 3 12 2.04

7 other 2-digit SIC codes 7 49 8.31

Totals 107 589 100.00expenditures

Table 4 presents the regression results from the OLS estimation of Eq. (1).22

Following prior research (Barton, 2001, p. 14; Kothari and Shanken, 1997, p.175), I report standardized coecients in all remaining tables to help assess the

economic signicance and relative importance of each regressor.23 The ad-

justed-R2 is 0.30 and TRI_AIR, NUMOIL, NUMCHEM, ACTS, and 33_50

are positively associated with environmental capital expenditures, while the

21 Although the utility industry is signicantly impacted by environmental costs, rms operating in

this industry are not part of my sample because they are not subject to TRI emissions reporting

requirements.22 In all equations, I control for outliers by deleting observations with an R-student statistic

greater than three in absolute value. The maximum number of observations deleted from any of the

equations based on this decision rule is 11.23 Standardized coecients can be interpreted as the eect of a one-standard-deviation change in

the independent variable on the dependent variable, which is also measured in standard deviations.

Table 4

D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206 193Decomposition of environmental capital expenditures into regulatory and voluntary components

ECAPit a0 a1TRI AIRit a2TRI WATERit a3TRI SOLIDit a4NUMOILit a5NUMCHEMit a6VALPENit a7SITESit a8ACTSit

a9X3

33 50itj1 33 50itj XY1

a10yYRyit XJ1

a11jINDjit it 1coecients on TRI_WATER, TRI_SOLID, VALPEN and SITES are not

signicant.24

The results from the estimation of Eq. (1) suggest that the RCD variables

and the voluntary cost driver capture at least a portion of rms regulatory

j0 y1 j1

Independent variable Prediction Standardized coecient t-statistic

TRI_AIR + 0.108** 2.81

TRI_WATER + 0.056 1.57

TRI_SOLID + 0.006 0.15

NUMOIL + 0.103** 2.49

NUMCHEM + 0.115** 2.95

VALPEN + 0.007 0.19SITES + 0.033 0.80

ACTS + 0.174** 3.89X3

j033 50itj1 33 50itj + 0.111** 2.70

Adjusted R2 0.30

F-statistic 8.87++

N 578

Year and industry indicator variables have been suppressed. All variables, except the year and

industry variables, are deated by year t sales. The sample period covers 19881996. **, * indicate

that coecient is signicantly dierent from zero at the 1% and 5% levels (one-tailed tests for

predicted signs), respectively. ++ indicates that regression model is signicant at the 1% level based

on the F-statistic. i, t denote rm and year, respectively; ECAP = total environmental capital

expenditures (in millions of dollars); TRI_AIR = toxic chemical air emissions (in millions of

pounds); TRI_WATER = toxic chemical water emissions (in millions of pounds); TRI_SOLID =

toxic chemical solid waste emissions (in millions of pounds); NUMOIL = number of reported oil

spills (larger than 10,000 gallons); NUMCHEM = number of reported chemical spills (larger than

10,000 pounds); VALPEN = the sum of environmental penalties assessed under nine major envi-

ronmental laws (in millions of dollars); SITES = the current number of Superfund sites on the

National Priority List (NPL) that the rm has been identied as a potentially responsible party;

ACTS = the current number of treatment, storage, and disposal facilities at which the EPA has

required a corrective action; 33_50 = the total emissions of the 17 toxic chemicals covered under the

EPAs voluntary 33/50 program, deated by rm sales; YRy = a dichotomous variable that equalsone if the observation is from year y, zero otherwise; and INDj = a dichotomous variable that

equals one if observation operates in industry j (as dened by two-digit SIC code), zero otherwise.

24 Following Belsley et al. (1980, p. 93), I assess collinearity in Eq. (1) using variance ination

factors (VIFs). All VIFs were less than two, suggesting no evidence of harmful collinearity.

and voluntary environmental capital expenditures. To the extent that the

model does not capture all of the regulatory and voluntary costs, these environ-

mental expenditures will be measured with error. In particular, due to the lack

of publicly available data on voluntary environmental cost drivers, it is likely

194 D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206that voluntary environmental costs are signicantly understated by the decom-

position model. Thus, I consider an alternative measurement of VECAP in alater section.

5.2. Associations with future abnormal earnings

Table 5 presents summary regression statistics from the OLS estimation

of Eq. (2), which relates regulatory and voluntary environmental capital

expenditures to one-year ahead abnormal earnings. The results reveal that

CAPX is positively associated with one-year ahead abnormal earnings(t = 3.37), suggesting that non-environmental capital expenditures generate

earnings greater than those implied by the discount rate. Concerning Hypoth-

esis 1, RECAP is negatively related to one-year ahead abnormal earnings

(t = 3.01), providing support for the hypothesis that regulatory environmen-tal capital expenditures generate earnings less than those implied by the dis-

count rate.

A possible explanation for the negative association between regulatory envi-

ronmental capital expenditures and one-year ahead abnormal earnings is thatRECAP is a proxy for unrecorded environmental liabilities. If rms with higher

unbooked environmental liabilities in year t recognize these liabilities (and the

related expenses) in year t + 1, then one-year ahead net income and abnormal

earnings will be lower. To address this concern, I rst collect the environmental

operating expense in year t + 1 for each rm-year observation. This data is

available for 308 of the 589 rm-year observations. I then add back the

year t + 1 environmental operating expense to one-year ahead abnormal earn-

ings, thereby reversing the income eect of the year t + 1 environmental ac-crual.25 Finally, I re-estimate Eq. (2) on the sub-sample using this alternative

denition of future abnormal earnings. The results (not reported) indicate that

RECAP is still negatively associated with one-year ahead abnormal earnings

(t = 2.93).With respect to Hypothesis 3, the association between VECAP and future

abnormal earnings is not signicant when I base the estimate of voluntary costs

on the voluntary cost driver and its related coecient from the rst-stage

25 This procedure assumes that when a rm recognizes an environmental liability it will record the

associated expense in environmental operating expenses. This assumption seems plausible since anumber of the sample rms describe their environmental operating expenses as those related to

regulatory remediation and cleanup activities.

regression (t = 0.10).26 Still, a t-test reveals that the coecient on VECAP is

signicantly greater than that on RECAP (t = 2.72). This suggests that the reg-

ulatory and voluntary components of environmental capital expenditures have

dierent rm-specic economic consequences.27

5.3. Market-based tests

Table 6 reports the coecient estimates and t-statistics from the OLS estima-

tion of Eq. (3), which examines the capital market pricing of environmental cap-

ital expenditures. Consistent with expectations, BVXPS and EPS are positively

associated with market value of equity (t = 12.19 and 9.87, respectively). More-

over, CAPXPS is positively related to share prices (t = 7.31). As predicted in

Hypothesis 2, RECAPS is negatively associated with stock prices (t = 4.19).However, inconsistent with expectations, VECAPS is not positively associatedwith market value (t = 0.14). Nevertheless, a t-test indicates that the coecienton VECAPS is signicantly greater than the coecient on RECAPS (t = 3.50),

suggesting that the capital market values voluntary environmental capital

expenditures dierently than regulatory environmental capital expenditures.

Table 7 presents summary regression statistics from the returns regression

D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206 195(i.e., Eq. (4)). The results indicate that the annual change in BVX is not signif-

icantly associated with returns (t = 0.53), while the annual change in EBEI is

positively related to returns (t = 5.14). Further, the change in CAPX is not

26 Firms likely undertake voluntary environmental capital projects only if the projects yield

positive returns. Moreover, an exogenous shock that increases earnings may also result in an

increased investment in voluntary environmental capital projects. Hence, VECAP may be an

endogenous variable in Eq. (2), resulting in inconsistent OLS estimates.

To mitigate concerns about endogeneity bias in Eq. (2), I use an approach that is similar to Lev

and Sougiannis (1996, pp. 113115) who investigate the relation between R&D expenditures and

operating income using a two-stage least squares approach. Specically, I begin by regressing

VECAP for rm i on the average industry level of VECAP (IND_VECAP), where industry is

dened by two-digit SIC code. The unreported results from this estimation reveal that

IND_VECAP is positively related to VECAP (t = 4.10). I obtain the tted values of VECAP

from this regression (VECAP*) and use that variable in Eq. (2) instead of VECAP.

The results from this alternative specication of Eq. (2) are quantitatively similar to those

documented in Table 5. Specically, VECAP* is not signicantly related to abnormal earnings

(t = 0.25), while RECAP remains negatively associated with abnormal earnings (t = 2.62).Consequently, it does not appear that endogeneity bias is driving the results from Eq. (2).27 It is also likely that regulatory and voluntary environmental capital expenditures made in year

t 1 (and previous years) aect abnormal earnings in year t + 1. Unfortunately, this issue isdicult to address due to the strong correlation between RECAP and VECAP in year t and their

respective counterparts in year t 1. Thus, I sum the environmental expenditures in year t withthose in year t 1 (going back more than one year signicantly reduces the sample size and,therefore, the power of the tests). The results using the sum of the environmental expenditures (i.e.,RECAPt + RECAPt1 and VECAPt + VECAPt1) as independent variables in Eq. (2) are notmaterially dierent from the results documented in Table 5.

Table 5

Associations of regulatory and voluntary environmental capital expenditures with one-year ahead

abnormal earnings

AEit1 b0b1AEit b2CAPXitb3RECAPit b4VECAPitXY1

y1b5yYRyit

XJ1

j1b6jINDjit mit1

2

196 D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206Independent variable Prediction Standardized coecient t-statistic

AE + 0.630** 12.55

CAPX + 0.113** 3.37signicantly related to returns (t = 0.51).28 Similar to the price-levels regres-sions, the annual change in RECAP is negatively associated with returns

RECAP 0.065** 3.01VECAP + 0.002 0.10

Adjusted R2 0.62

F-statistic 37.04++

N 569

Coecient test Prediction t-statistic

b4 = b3 b4 > b3 2.72##

Year and industry indicator variables have been suppressed. All variables, except the year and

industry variables, are deated by year t sales, t-statistics are based on White (1980, pp. 818821)

standard errors. The sample period covers 19881996. **, * indicate that coecient is signicantly

dierent from zero at the 1% and 5% levels (one-tailed tests for predicted signs), respectively.##,# indicate that coecients are signicantly dierent from each other at the 1% and 5% levels (one-

tailed tests for predicted dierences), respectively. ++ indicates that regression model is signicant at

the 1% level based on the F-statistic. i, t denote rm and year, respectively; AE = abnormal earnings;

CAPX = capital investment in property, plant, and equipment (net of the scal years environmentalcapital expenditures); RECAP = estimated regulatory environmental capital expenditures (tted

values based on the regulatory cost drivers and their related coecients from Eq. (1)); VECAP =

estimated voluntary environmental capital expenditures (tted values based on the voluntary cost

driver and its related coecient from Eq. (1)); YRy = a dichotomous variable that equals one if the

observation is from year y, zero otherwise; and INDj = a dichotomous variable that equals one if the

observation operates in industry j (as dened by two-digit SIC code), zero otherwise.

I dene abnormal earnings as: AEt = Xt rBVt1; where X is earnings before interest, taxes,depreciation and amortization in year t, r is the discount rate, and BV is book value of equity at the

end of year t 1. I calculate the discount rate as r(j, t) = rf(t) + rp(j), where r(j, t) is the estimateddiscount rate for industry j in year t, rf(t) is the yearly t-bill return in year t, and rp(j) is the risk

premium for industry j as estimated by Fama and French (1997, pp. 172173). Industry mem-

bership is determined by SIC code.

28 An analysis of annual reports and 10-Ks for a subset of the sample rms reveals that some of

these rms release estimated (i.e., planned) gures concerning future capital investment. In these

cases, year t 1 CAPX is not an appropriate proxy for the markets expectation of year t CAPX.This measurement error may be driving the insignicant association between DCAPX and stockreturns. Prior studies examining the market reaction to capital expenditure announcements (e.g.,

McConnell and Muscarella, 1985, pp. 403404) use the planned dollar amount of capital

expenditures (if available) to compute unexpected capital investment.

Table 6

Capital market pricing of regulatory and voluntary environmental capital expenditures

PRICEit c0 c1BVXPSit c2EPSit c3CAPXPSit c4RECAPSit c5VECAPSit

XY1

c6yYRyit XJ1

c7jINDjit git 3

D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206 197y1 j1


BVXPS + 0.422** 12.19

EPS + 0.366** 9.87

CAPXPS + 0.276** 7.31**(t = 1.74), while the change in voluntary environmental costs is not signi-cantly related to stock returns. Moreover, the coecient on VECAP is signif-

icantly greater than the coecient on RECAP (t = 1.66).29 To summarize, the

results from the testing of the market valuation hypotheses are robust to a re-

turns specication.

RECAPS 0.129 4.19VECAPS + 0.003 0.14Adjusted R2 0.55

F-statistic 26.11++

N 549


c5 = c4 c5 > c4 3.50##

Year and industry indicator variables have been suppressed. t-statistics are based on White (1980,

pp. 818821) standard errors. The sample period covers 19881996. **, * indicate that coecient is

signicantly dierent from zero at the 1% and 5% levels (one-tailed tests for predicted signs),

respectively. ##,# indicate that coecients are signicantly dierent from each other at the 1% and

5% levels (one-tailed tests for predicted dierences), respectively. ++ indicates that regression model

is signicant at the 1% level based on the F-statistic. i, t denote rm and year, respectively;

PRICE = share price three months after the end of the scal year; BVXPS = book value of equity

at the end of the scal year (net of the scal years capital expenditures), per share; EPS = earningsbefore extraordinary items, per share; CAPXPS = capital investment in property, plant, and

equipment (net of the scal years environmental capital expenditures), per share; RECAPS =estimated regulatory environmental capital expenditures, per share; VECAPS = estimated volun-

tary environmental capital expenditures, per share; YRy = a dichotomous variable that equals one

if the observation is from year y, zero otherwise; and INDj = a dichotomous variable that equals

one if the observation operates in industry j (as dened by two-digit SIC code) in year t, zero

otherwise.

29 It is important to note that since the coecients on voluntary environmental capital

expenditures in Eqs. (2)(4) are not signicantly dierent from zero (possible due to measurement

error), the results from the statistical comparison of the coecients on the voluntary and regulatory

expenditures should be interpreted with caution.

Table 7

Regression of annual stock returns on annual changes in regulatory and voluntary environmental

capital expenditures

RETXit d0 d1DBVXit d2DEBEIit d3DCAPXit d4DRECAPit d5DVECAPit

XY1

y1d6yYRyit

XJ1

j1d7jINDjit lit 4


198 D. Johnston / Journal of Accounting and Public Policy 24 (2005) 1752065.4. Alternative measurement of voluntary environmental capital expenditures

In the preceding empirical analysis, I estimate voluntary environmental

costs using a voluntary cost driver and its estimated coecient from Eq. (1).The voluntary cost driver attempts to capture a rms capital investment inthe EPAs voluntary 33/50 program. However, voluntary environmental capitalexpenditures include costs related to not only the 33/50 program but also other

voluntary activities such as environmental audits and re-engineering of produc-

tion processes to reduce hazardous waste. Since data availability severely con-

strains the set of voluntary explanatory variables, the decomposition model

DBVX + 0.041 0.53DEBEI + 0.296** 5.14DCAPX + 0.031 0.51DRECAP 0.072* 1.74DVECAP + 0.015 0.40Adjusted R2 0.19

F-statistic 4.82++

N 438


d5 = d4 d5 > d4 1.66#

Year and industry indicator variables have been suppressed. All independent variables, except the

year and industry variables, are deated by beginning market value of equity. t-statistics are based

on White (1980, pp. 818821) standard errors. The sample period covers 19891996. **, * indicate

that coecient is signicantly dierent from zero at the 1% and 5% levels (one-tailed tests for

predicted signs), respectively. ##,# indicate that coecients are signicantly dierent from each

other at the 1% and 5% levels (one-tailed tests for predicted dierences), respectively. ++ indicates

that regression model is signicant at the 1% level based on the F-statistic. i, t, D denote rm, year,and annual change, respectively; RETX = annual stock return (excluding dividends) as reported in

CRSP, measured from nine months before the end of the scal year through three months after the

end of the scal year; BVX = book value of equity (net of the scal years capital expenditures) atthe end of the scal year; EBEI = earnings before extraordinary items; CAPX = capital investment

in property, plant, and equipment (net of the scal years environmental capital expenditures);RECAP = estimated regulatory environmental capital expenditures; VECAP = estimated volun-

tary environmental capital expenditures; YRy = a dichotomous variable that equals one if

the observation is from year y, zero otherwise; and INDj = a dichotomous variable that equals

one if the observation operates in industry j (as dened by two-digit SIC code) in year t, zero

otherwise.

negatively associated with future abnormal earnings. In addition, I nd a

positive relation between voluntary environmental capital expenditures andfuture abnormal earning when I use the residuals from a regression of total

environmental capital expenditures on regulatory cost drivers as proxies for

the voluntary environmental costs. Further, the results suggest that the associ-

ation between voluntary environmental capital outlays and one-year ahead

abnormal earnings is signicantly dierent than the association between regu-latory environmental capital costs and future abnormal earnings.

The results from the market-based tests are mixed. As predicted, regulatory

environmental capital expenditures are negatively related to stock prices and

stock returns. However, inconsistent with predictions, the association between

voluntary environmental capital expenditures and market value is not signi-

cant. Still, statistical tests reveal that the capital market appears to value the

regulatory and voluntary components of environmental capital expenditures

dierently.The results of this study may apply to the stream of research examining the

determinants of asset write-downs (and related impairment losses) of property,

plant and equipment (e.g., Francis et al., 1996, pp. 122128). As shown in thismay not capture a large portion of rms voluntary environmental capitalexpenditures. In fact, Table 2 reports that the mean (median) voluntary envi-

ronmental capital expenditure is only $1,270,000 ($120,000). This suggests that

the voluntary cost driver multiplied by its related coecient from Eq. (1) may

result in a noisy surrogate for voluntary environmental costs.

I obtain an alternative proxy for voluntary environmental capital expendi-tures from a regression of total environmental capital expenditures on the reg-

ulatory cost drivers. To the extent that the regulatory cost drivers capture a

large portion of the rms regulatory capital expenditures, the estimated resid-uals from the OLS estimation of Eq. (1) (excluding the voluntary cost driver

from the regression) may proxy for voluntary environmental capital expendi-

tures. The unreported results based on this alternative measurement of VECAP

are similar to those previously discussed, with the following exception: VECAP

is positively related to one-year-ahead abnormal earnings (t = 1.85, p < 0.05,one-tailed test). Nevertheless, it remains unclear which proxy is a less noisy

measure of voluntary environmental capital expenditures.

6. Conclusion

This paper examines the rm-specic economic consequences of regulatory

and voluntary environmental capital expenditures. As hypothesized, the empir-ical analysis reveals that regulatory environmental capital expenditures are

D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206 199study, regulatory environmental capital expenditures may not generate earnings

in excess of those implied by the discount rate and, therefore, a signicant

capital expenditures explain more of the asset writedown than voluntary envi-

ronmental capital expenditures and non-environmental capital expenditures.Moreover, this paper may have public policy implications. Palmer et al.

(1995, p. 131) note that the attractiveness of environmental regulations shouldbe based on cost-benet analyses. Also, Palmer et al. (1995, p. 120) acknowledge

that a heavier reliance on performance-based environmental regulations instead

of command-and-control legislation may lead to more ecient and cost-eec-

tive production processes at the rm level. This paper sheds some light on the

economic attractiveness of US environmental programs by examining the

rm-specic economic consequences (i.e., the net cost) of regulatory and volun-

tary environmental capital expenditures. Still, it is important to note that the

empirical analyses in this paper ignore the social benets of corporate invest-ment in environmental protection, such as the reduced negative impact on our

natural resources as well as reduced mortality rates. Until these positive exter-

nalities are quantied and incorporated into the analyses, the total benets asso-

ciated with the investment in environmental protection will be underestimated.

This study is subject to limitations. First, I must estimate the regulatory and

voluntary components of environmental capital expenditures. Prior environ-

mental research provides little guidance for carrying out this estimation and

the lack of publicly available data constrains the set of regulatory and volun-tary cost drivers. Consequently, measurement error is unavoidable. Second,

although this studys results suggest that regulatory environmental capitalexpenditures are viewed by the market as current period expenses, Holthausen

and Watts (2001, pp. 6364) note that researchers should be careful when sug-

gesting a change in the nancial reporting treatment of a certain type of expen-

diture based on the results from capital market studies. Still, this study can be

viewed as merely testing rms claims concerning the economic benets associ-ated with these recorded assets (Barth, 2000, pp. 810). Third, this paper onlyexamines the regulatory and voluntary components of environmental capital

expenditures. A more useful empirical analysis would involve decomposing

regulatory and voluntary environmental costs into sub-categories. For exam-

ple, regulatory costs could be further separated into stop-gap expenditures

and other regulatory costs such as preventative compliance costs. However,

the lack of publicly available data makes it dicult to perform that detailed

of a decomposition of total environmental costs.

Acknowledgementsportion of the impairment loss may be related to regulatory environmental xed

assets. As such, future research could explore whether regulatory environmental

200 D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206I thank my dissertation committee members, David Guenther, Robert

McNown, Steve Rock, Phil Shane (co-chair) and Naomi Soderstrom (co-

chair), for their support and valuable comments. This paper also beneted

greatly from the comments of two anonymous reviewers and workshop partic-

ipants at Colorado State University; the University of Colorado at Boulder;

the University of New Hampshire; the University of Oregon; the University

of Utah; and the College of William and Mary. I gratefully acknowledge the

nancial support of Colorado State University, the Naval Postgraduate Schooland the Leeds School of Business, the Accounting Department, and the Grad-

uate School at the University of Colorado at Boulder.

RCRA-regulated hazardous wastes include over 350 dierent commercial

chemical products as well as other solid wastes.33 Moreover, rms that dispose

D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206 20130 Further, Superfund and its related amendments established the Hazardous Substance

Superfund that can be used by the EPA in the remediation of contaminated sites (Cardwell,

1999, p. 352). The Superfund is created by taxes imposed upon chemical and petroleum industries,

an environmental tax on corporations, and general tax revenue (Cardwell, 1999, p. 353).31 PRPs may include current and former owners and operators of the site, generators of the

hazardous waste, parties who arranged for the disposal of the waste, and those who transported the

waste to the site (U.S.C., 1980, 9607 (a)(l)(a)(4)).32 RCRA was amended in 1984 pursuant to the Hazardous and Solid Waste Amendments.Appendix A. Summary of major US federal environmental regulations

A.1. The Superfund Act

The Comprehensive Environmental, Response, Compensation, and Liabil-

ity Act (more commonly known as Superfund) empowers the EPA to regulate

the remediation of inactive and/or abandoned hazardous waste sites.30 Signed

into law in 1980 and amended in 1986, the Superfund Act allows the EPA to

identify corporations that contributed to the contamination of a site as poten-

tially responsible parties (PRPs), thereby holding them liable for the cleanup

costs of the sites.31 Superfund can be retroactively applied and courts have

found that it imposes strict as well as joint and several liability (Cardwell,1999, pp. 370372).

A.2. The Resource Conservation and Recovery Act (RCRA)

The goals of the Resource Conservation and Recovery Act of 1976 are to

reduce or eliminate the generation of hazardous wastes and render land dis-

posal the least favored method for hazardous waste management (Case,

1999, p. 92).32 As such, RCRA imposes substantial disposal costs by regulatingthe management of hazardous wastes from cradle-to-grave. Specically,33 See Code of Federal Regulations (C.F.R.), Title 40, 261.31, 261.32, and 261.33 for a complete

list of RCRA-regulated hazardous wastes.

of their own hazardous wastes must follow a comprehensive set of regulations

governing all aspects of their facilities as well as undertake corrective actions

for the remediation of their hazardous waste disposal sites (Case, 1999, pp.

202 D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206119120).

A.3. The Clean Air Act (CAA)

The objective of the CAA of 1970 and its 1990 amendments is to protect and

enhance the nations air resources by controlling air pollution (USEPA, 1995c,p. 86). To rectify air pollution problems, the Act has imposed signicant pro-

cess redesign costs on certain industries. For example, the CAA requires the

use of reformulated fuels in US cities with the worst ozone problems (Bhat,

1996, p. 35). To comply, petroleum-rening rms were forced to incur signi-

cant process redesign costs (USEPA, 1995c, pp. 8889). In addition, the 1990CAA amendments require the regulation of 189 hazardous air pollutants.34

Facilities identied as major sources of any of the 189 substances are required

to install pollution abatement equipment that results in the maximum degree of

pollution reduction given economic, energy and environmental constraints

(Brownell, 1999, pp. 181182).

A.4. The Clean Water Act (CWA)

The purpose of the CWA is to restore and maintain the chemical, physical,

and biological integrity of the nations waters (33, U.S.C. 1251 (a)). To thisend, the CWA requires regulated facilities to obtain permits that control dis-

charges of pollutants into the nations waters and storm sewers.35 Permit hold-ers must monitor their compliance with euent limitations (using the proper

monitoring equipment and analytical methods) and report the results to the

appropriate permitting authority (Gallagher, 1999, p. 219). Moreover, the

CWA regulates the discharge of pollutants into publicly-owned treatmentworks by requiring adherence to various pretreatment standards (Gallagher,

1999, p. 238).

A.5. The Oil Pollution Act (OPA)

In response to the 1989 Exxon Valdez oil spill, Congress passed the Oil Pol-

lution Act of 1990. OPA requires owners and operators of regulated facilities

to submit facility response plans that describe their policies and procedures

34 See CAA 112 (b) for a complete list of the regulated air pollutants.35 Unlike the CAA and RCRA that specically identify the pollutants to be regulated, the CWAs

denition of a pollutant . . .has been broadly interpreted by the courts to include virtually anymaterial (Gallagher, 1999, p. 209).

Enacted in 1986, EPCRA requires rms to increase their disclosure concern-ing hazardous substances to both the public as well as federal, state, and local

ocials. For example, most manufacturing facilities must submit an annual

toxic release inventory (TRI) report to the EPA that details the transfers and

releases of 654 chemicals (Epstein, 1996, p. 254). Pursuant to the amendments

passed with the 1990 Pollution Prevention Act, these facilities are also required

to provide information on their recycling and pollution prevention techniques

for each of the toxic chemicals (Scagnelli, 1999, p. 578). Moreover, manufac-turing facilities must immediately notify local authorities in the event of a

chemical release, providing information concerning the name of the chemical,

an estimate of the quantity released, and any known or anticipated health risks

associated with exposure to the chemical (40 C.F.R. 355.40 (b) (2)).

References

Aboody, D., Barth, M.E., Kasznik, R., 1999. Revaluations of xed assets and future rm

performance: Evidence from the UK. Journal of Accounting and Economics 26, 149178.

Arora, S., Cason, T.C., 1995. An experiment in voluntary environmental regulation: Participation

in EPAs 33/50 program. Journal of Environmental Economics and Management 28, 271286.Arora, S., Gangopadhyay, S., 1995. Toward a theoretical model of voluntary overcompliance.

Journal of Economic Behavior and Organization 28 (December), 289309.

Barbera, A.J., McConnell, V.D., 1990. The impact of environmental regulations on industry

productivity: Direct and indirect eects. Journal of Environmental Economics and Manage-

ment 18, 5065.

Barth, M.E.., 2000. Valuation-based accounting research: Implications for nancial reporting and

opportunities for future research. Accounting and Finance 40, 731.

Barth, M.E., McNichols, M.F., 1994. Estimation and market valuation of environmental liabilities

relating to superfund sites. Journal of Accounting Research 32 (Supplement), 177209.

Barth, M.E., Clinch, G., 1988. Revalued nancial, tangible, and intangible assets: Associations withfor responding to worst case oil spills (Gallagher, 1999, p. 249). In addition,

OPA holds responsible parties strictly liable for the cleanup costs associated

with oil spills as well as for compensatory damages (Olney, 1999, pp. 275

276). For responsible tank vessel owners, the liability limit is the greater of

$1200 per gross ton or $10 million. Responsible parties at onshore facilities

and deepwater ports are liable for up to $350 million, but the federal govern-ment has the authority to adjust this limit (33, U.S.C. 2704 (a)). Furthermore,

OPA establishes civil and criminal penalties for failing to notify the proper

authorities of oil spills. In particular, the maximum ne (prison term) is

$500,000 (5 years) per violation. Finally, the Act sets forth tank vessel stan-

dards, such as the double hull requirement.

A.6. The Emergency Planning and Community Right-to-Know Act (EPCRA)

D. Johnston / Journal of Accounting and Public Policy 24 (2005) 175206 203share prices and non-market-based value estimates. Journal of Accounting Research (Supple-

ment), 199233.

Barton, J., 2001. Does the use of nancial derivatives aect earnings management decisions? The

An investigation of regulatory and voluntary environmental capital expenditures

Documents

Transcript of An investigation of regulatory and voluntary environmental capital expenditures