Does it Really Pay to be Green?
An Empirical Study of Firm Environmental and Financial Performance
Stern School of Business
New York University
40 West 4th St. Suite 707
New York, NY 10012
Tel: (212) 998-0288
Fax: (212) 995-4227
Stern School of Business
New York University
40 West 4th St. Suite 717
New York, NY 10012
Tel: (212) 998-0261
Fax: (212) 995-4235
Forthcoming in The Journal of Industrial Ecology
* Corresponding author. (http://www.stern.nyu.edu/~mlenox/)
Does it Really Pay to be Green?
An Empirical Study of Firm Environmental and Financial Performance
Previous empirical work suggests that profitable firms tend to have high
environmental performance, but questions persist about the nature of the relationship.
Does stronger environmental performance really lead to better financial performance or
is the observed relationship the outcome of some other underlying firm attribute? Does
it pay to have clean running facilities or to have facilities in relatively clean industries?
To explore these questions, we analyze 652 U.S. manufacturing firms over the time
period 1987 to 1996. While we find evidence of an association between lower pollution
and higher financial valuation, we find that a firm's fixed characteristics and strategic
position might cause this association.
Keywords: beyond compliance, corporate strategy, environmental performance,
Porter hypothesis, win-win
Does it Really Pay to be Green?
An Empirical Study of Firm Environmental and Financial Performance
Scholars had long assumed that investments to protect the natural environment
provided few financial benefits to firms. In the last 20 years, however, a growing
number of researchers have challenged this assumption. In the field of industry
ecology, scholars argue that there are situations where beyond compliance behavior by
firms is a win-win for both the environment and the firm (Nelson 1994; Panayotou and
Zinnes 1994; Esty and Porter 1998; Reinhardt 1999). Scholars now suggest that firms
may be both "green and competitive" (Porter and van der Linde 1995; Reinhardt 1999).
Qualitative research has identified numerous examples of profitable pollution prevention
opportunities (Denton 1994; Deutsch 1998; Graedel and Allenby 1995; Porter and van
der Linde 1995; King 1995). Many scholars now argue that discretionary improvements
in environmental performance often provide financial benefit (e.g. Hart 1997).
In response, a growing empirical literature has applied econometric techniques
to test the "pays to be green" hypothesis. Several studies provide evidence that higher
environmental performance is associated with better financial performance, but these
early studies often lacked the longitudinal data needed to fully test the relationship.
Several years of data are needed if one wants to rule out rival explanations for the
apparent association or show that environmental improvement causes financial gain.
Furthermore, the empirical literature does not clarify whether the apparent association
is generated by a firm’s choice to operate in cleaner industries or to operate cleaner
facilities. Existing research cannot answer whether it pays to be green or whether it
pays to operate in green industries.
In this paper, we review and comment on the empirical "pays to be green"
literature. We discuss how a firm's stable attributes and strategic position may jointly
cause both lower pollution levels and better financial performance, and thereby create
the appearance of a direct relationship between the two. For example, innovative firms
may have both lower emissions levels and greater profits. Alternatively, managers may
choose to improve their firm’s environmental performance when they have an especially
To help distinguish the effect of pollution reduction from other underlying factors,
we adopt empirical methods that account for unmeasured firm attributes. Furthermore,
to differentiate between pollution reduction and divestiture of operations in dirtier
industries, we separate environmental performance into two constructs: (1) relative
performance within one's industries, and (2) the average performance of the industries
in which one chooses to operate. We analyze 652 U.S. manufacturing firms over the
time period 1987 to 1996. We find evidence of a real association between lower
pollution and higher financial performance. We also show that a firm's environmental
performance relative to its industry is associated with higher financial performance.
However, we cannot show conclusively that a firm's choice to operate in cleaner
industries is associated with better financial performance, nor can we prove the
direction of the observed relationships. Thus, our research provides support for a
connection between some means of pollution reduction and financial performance, but
it also suggests that the connection should be viewed with caution.
Evidence to Date
Proponents of a causal link between environmental and financial performance
argue that pollution reduction provides future cost savings by increasing efficiency,
reducing compliance costs, and minimizing future liabilities (Porter and van der Linde
1995; Reinhardt 1999). Porter and van der Linde (1995) theorize that opportunities for
profitable pollution reduction exist because managers often lack the experience and
skill to understand the full cost of pollution (Jaffe, Peterson et al. 1995). Hart (1997)
proposes that excess returns (i.e., profits above the industry average) result from
differences in the underlying fixed characteristics of firms and industries. Managers
may possess unique resources or capabilities that allow them to employ profitable
strategies that are difficult to imitate.
Using a variety of measures (See Tables 1 & 2), much of the empirical "pays to
be green" literature has supported the proposed positive relationship between pollution
reduction and financial gain by relying on correlative studies of environmental and
financial performance. A series of studies conducted by the Council on Economic
Priorities (CEP) in the 1970s found that expenditures on pollution control were
significantly correlated with financial performance among a sample of pulp and paper
firms (Bragdon and Marlin 1972; Spicer 1978).i More recently, Russo and Fouts (1997)
found a significant positive correlation between various financial returns and an index of
environmental performance developed by the Council on Economic Priorities. Dowell
and colleagues (2000) found that firms that adopt a single, stringent environmental
standard worldwide have higher market valuation (Tobin’s q) than firms that do not
adopt such standards.
Insert Table 1 about here
Insert Table 2 about here
In the finance literature, a number of studies have examined the market returns
of portfolios of environmentally friendly firms. Cohen and colleagues (1995) used
several measures of environmental performance derived from U.S. EPA databases to
construct two industry-balanced portfolios of firms. They found no penalty for investing
in the green portfolio and a positive return to green investing. Similarly, White (1996)
found a significantly higher risk-adjusted return for a portfolio of green firms using the
CEP ratings of environmental performance.ii
To the extent that one cares merely about correlation and little about causation,
these correlative studies are informative. Market analysts, for example, increasingly
gather environmental performance data as an indicator of future capital market returns
(Kiernan 1998). For their purposes, it matters little whether environmental performance
leads to financial performance or simply provides an indicator of firms that have high
From the perspective of corporate managers and policy analysts, however, the
distinction is critical. The prescription that often follows from the "pays to be green"
literature is that managers should make investments to lower their firm's environmental
impact (Hart and Ahuja 1996). To fully demonstrate that it “pays to be green”, research
must demonstrate that environmental improvements produce financial gain.
Event studies are one means of demonstrating that greening indeed causes
financial gain. Such studies look at the relative changes in stock price following some
environmental event. By isolating a single environmental event within a narrow time
frame, event studies control for important differences among firms that cannot be
observed. The limitation with event studies is that they often study the effect of events
that are only partially environmental in nature. Klassen and McLaughlin (1996), White
(1996), Karpoff and colleagues (1998), and Jones and Rubin (1999) study the effect of
published reports of events and awards on firm valuation and find a relationship
between the valence of the event (positive or negative) and the resulting change in
market valuation. Blacconiere and Patten (1994) estimate that Union Carbide lost $1
billion in market capitalization, or 28%, following the Bhopal chemical accident in 1984.
Muoghalu and colleagues (1990) found that firms named in lawsuits concerning
improper disposal of hazardous waste suffered significant losses in capital market
value. Each of these events has environmental elements, but they also are affected by
other firm attributes. King and Baerwald (1998) argue the size, market power, and
unique firm characteristics influence how events are reported and interpreted. A firm
with good public relations may be able to put a positive spin on negative news. A firm
that possesses good legal resources may better forestall lawsuits.
Some event studies seek to avoid these problems by using the annual release of
toxic emission data through the U.S. EPA’s Toxic Release Inventory (TRI) program as
the "event". Hamilton (1995), Konar and Cohen (1997), and Khanna and colleagues
(1998) all find that polluting firms lost market value in a one-day window following the
release of TRI information. These important studies still may suffer from construct
validity however. Given the complexity of analyzing TRI data, it seems possible that
same day stock price movements probably reflect contemporaneously reported
pollution rankings. These rankings are strongly affected by firm size and industry
choice and thus the stock market effect may be the result of temporary bad press rather
than a real change in perception of a firm’s long-term value. Perhaps for this reason,
these TRI event studies showed only weak and inconsistent evidence in a 5-day
window following the TRI data release.
Another way to account for unobserved firm differences is to use standard
regression techniques to evaluate the effect of changes in pollution on changes in
financial performance. This in essence is the approach used in a widely cited study by
Hart and Ahuja (1996). They show that changes in pollution (emission per sales dollar)
predate changes in financial performance. While an important advance in the literature,
their measure of environmental performance conflates reduction of pollution at current
operations and divestiture of dirty operations, making it difficult to interpret the meaning
of their study. Is it that it pays to be green or does it pay to operate in clean industries?
This issue underscores a larger debate within the strategy literature on the
source of returns in excess of investments of similar risk (Rumelt 1991; McGahan &
Porter 1997). The industrial organization literature out of economics suggests that
excess returns result from differences in the underlying structure of industries.
According to this logic, greener industries may have higher returns than dirtier industries
because of lower compliance and regulatory costs. In contrast, the resource-based
view of strategic management suggests that individual firm capabilities may lead to
excess returns when they are difficult to imitate, not substitutable, rare, and valuable
(Barney 1986; Wernerfelt 1984). According to this view, superior ability to manage
environmental problems relative to others in your industry may lead to higher returns.
Much of the empirical "pays to be green" literature uses strategy resource-based logic
to justify a relationship between environmental and financial performance.
Unfortunately, they fail to disentangle the effects of industry choice from the effects of
variance in environmental strategies amongst firms in the same industry.
An Empirical Approach
In the following sections, we analyze whether it really "pays to be green" using a
methodology that allows us to explore whether unmeasured firm and industry
characteristics may explain the observed link between environmental and financial
performance. We also use a measure of environmental performance that untangles the
effect of a firm’s relative performance within its industries and the average performance
of the industries in which it chooses to be.
We create a sample of publicly traded U.S. manufacturing firms during the period
1987-1996 by combining the U.S. EPA's Toxic Release Inventory (TRI) with facility data
from Dun & Bradstreet and corporate data from Standard & Poor's Compustat
database. The U.S. EPA started the TRI in 1987 to track emissions of over two
hundred toxic chemicals from U.S. manufacturing firms. Facilities must complete
annual TRI reports if they manufacture or process 25,000 pounds (or about 11,340
kilograms), use more than 10,000 pounds of any listed chemical during a calendar year,
and employ ten or more full-time people. To be in our sample, a firm must have at least
one facility that meets these requirements and be among the public corporations listed
in the Compustat database. Matching the two sets, we created an unbalanced sample
of 652 firms constituting 4483 firm-year observations for the years 1987 to 1996. iii
Financial Performance. The dependent variable for our analysis is financial
performance as reflected by Tobin's q. Tobin’s q measures the market valuation of a
firm relative to the replacement costs of tangible assets (Lindenberg & Ross, 1981).
Essentially, it reflects what cash flows the market thinks a firm will provide per dollar
invested in assets. It should be higher if future cash flows are expected to be greater or
if they are expected to be less risky. In accordance with more recent "pays to be
green" studies, we use a simplified measure of Tobin's q (Dowell, Hart, & Yeung 2000).
We calculate Tobin's q by dividing the sum of firm equity value, book value of long-term
debt, and net current liabilities by the book value of total assets.iv All financial data
were obtained from the Compustat database. .
Environmental Performance. Previous research has measured the
environmental performance of a firm as the degree to which that firm emits toxic
pollution given its size (Hart & Ahuja 1996). We create a similar measure (Total
Emissions) by calculating the log of total facility emissions of toxic chemicals .
Unfortunately, the meaning of this variable is ambiguous because it confounds pollution
that results from industry positioning with pollution that results from poor environmental
management. Consequently, we form two additional variables to separate the effect of
environmental management from the effect of industry positioning. Relative Emissions
measures the firm's ability to manage and reduce its pollution by comparing the degree
to which a firm's facilities are more or less polluting than other facilities in the same
industry (measured by 4 digit SIC code and adjusted for differences in size). Industry
Emissions measures the degree to which a firm tends to operate in industries where
production entails pollution. If a firm operates in industries where the average facility
has higher emissions, this variable will have a larger value. Please refer to the
Appendix for a detailed description of the construction of these variables.
Controls. We include a number of measures commonly used in the analysis of
financial performance as controls. These measures include 1) the company's size
(Firm Size) calculated as the log of the company’s assets, 2) the capital intensity of a
firm (Capital Intensity) calculated by dividing capital expenditures by sales, 3) the
annual growth of the firm (Growth) calculated as the percentage change in sales, 4) the
degree to which the firm is leveraged (Leverage) divided as the ratio of its debt to
assets, and 5) the R&D intensity (R&D Intensity) calculated by dividing research &
development expenses by total assets.
In addition, we control for the stringency of the regulatory environment in which
the firm operates (Regulatory Stringency). Environmental regulation varies across
regions and imposes greater (or lesser) penalties for pollution from facilities operating in
those regions. We measure a state's regulatory stringency by calculating the inverse of
the log of toxic emissions divided by total employees in four main polluting industries –
chemicals, petroleum, pulp & paper, and materials processing (Meyer 1995). The logic
for this measure is that higher regulation leads to lower emissions per employee (for
these industries) and thus increases the inverse of this ratio (Regulatory Stringency).
For each firm, we create a measure of the average regulation it faces by calculating the
weighted-average of the regulatory stringency for all the states in which the firm
To create an alternative measure of the degree to which the different facilities in
our sample are regulated, we count the number of performance criteria with which each
facility must comply (i.e., the number of permits issued to a facility). Under the U.S.
Clean Water Act, regulators may impose limits on water flow, suspended solids, and
chemical concentration. Although guidelines exist for administering the law, substantial
discretionary power remains. We created an alternative measure of regulatory
stringency, Permits, by summing the number of federal permits and then dividing by
Insert Table 3 about here
Previous studies have found that pollution precedes poor financial performance
by one or more years (Hart & Ahuja, 1996). To test these findings, we use least-
squares regression analysis to find a linear relationship between our independent
variables and the firm’s future Tobin's q (See Table 4).v Because firms may differ in
ways that we do not capture with our independent variables, we include dummy
variables that allow each firm to have a different constant value. This is often called a
"fixed effects" analysis because it reduces the possibility that a firm's fixed attributes
confounds the analysis. In essence, this fixed-effect regression requires that changes in
independent variables (rather than their baseline level) be associated with changes in
Consistent with much of the "pays to be green" literature, we find that Total
Emissions are associated with superior financial performance even when controlling for
firm fixed effects (Model 1). Thus, we provide evidence that environmental
performance is associated with financial performance rather than the observed
relationship being the outcome of some other underlying firm attribute.
Insert Table 4 about here
As discussed earlier, evidence of such a relationship still leaves many
unanswered questions. Does it pay to have clean running facilities, or to have facilities
in relatively clean industries? To better account for these differences, we separate
Total Emissions into two parts that reflect a firm's tendency to operate in polluting
industries (Industry Emissions) and its tendency to operate dirtier facilities within these
industries (Relative Emissions). In Model 2, the significant and negative coefficient for
Relative Emissions indicates that firms with lower emissions in their industries tend to
experience higher financial performance in the subsequent year. The lack of
significance for the coefficient for Industry Emissions means that we cannot conclude
that firms that operate in cleaner industries have higher financial performance.
One problem with fixed effects analysis is that it can do its job too well. By
eliminating the effect of all firm attributes that are relatively constant, the fixed effect
may obscure evidence that a fixed attribute is actually important. If firms do not
frequently change industry, and thus industry position is relatively constant, we might
miss the financial effect of industry choice. To check this, we use an alternative
specification called random effects. While this method continues to reduce the effect of
fixed firm attributes, it assumes that these are normally distributed. This method
suggests that firms that operate in cleaner industries (Industry Emissions) have higher
What might explain the difference between Model 3 and Model 2? One
possibility is that few firms in our sample actually move across industries and thus the
fixed effects analysis removes the effect of industry position. Another possibility is that
firms benefit from being in cleaner industries, but not from moving to cleaner industries.
Perhaps such movement entails costs that reduce a firm’s valuation or signals some
difficulty or problem. It is important to note that in our particular case, statistical tests
suggest that the fixed effects and not random effects analysis should carry more
Finally, we still have not considered the effect of causality. Which way does the
relationship run? Do more profitable firms invest more in environmental performance or
does environmental performance lead to profit? In Model 4, we present one method for
answering this question. To reduce the effect of a previous profitable year, we include
the previous year's Tobin's q in the regression.vii Unfortunately, this analysis does not
provide reliable evidence that firms with lower emissions in their industries (Relative
Emissions) tend to experience higher financial performance. Thus, while we find
evidence of an association between reduced emissions and profit, we can not say with
confidence which way the relationship runs. We again find no evidence that the
cleanliness of the industries in which the firm has facilities (Industry Emissions) is
associated with higher market valuation when we control for firm fixed effects.
The above analysis is illustrative of what we found throughout our analysis.
Using different forms of models and different methods for measuring our variables, we
often found an association between environmental and financial performance.
However, we also found that variations in model specification, sample, and
measurement method could reduce the significance of this effect below accepted
thresholds (although it never reversed in sign). We have presented the most careful
and complete specification of our analyses.
In this paper, we further explore whether it "pays to be green". We use
longitudinal data and statistical methods that reduce the potential for unobserved
differences among firms to create a misleading association between environmental and
financial performance. We also test to see whether pollution reduction causes financial
gain. Table 5 presents a summary of these results. We find evidence of an association
between pollution reduction and financial gain, but we cannot prove the direction of
causality. We also show that firms in cleaner industries have higher Tobin's q, but we
are unable to rule out possible confounding effects from fixed firm attributes. Moreover,
we cannot show that firms that move to cleaner industries improve their financial
Our research provides both additional support for the "pays to be green"
hypothesis and suggests caution in interpreting its implications. Much of the variance in
our study is attributed to firm level differences. Better understanding of these
differences might provide a richer understanding of profitable environmental
improvement. It may be that it pays to reduce pollution be certain means and not
others. Alternatively, it may be that only firms with certain attributes can profitably
reduce their pollution.
Additional research is needed to explore how underlying firm characteristics
affect the relationship between relative environmental performance and financial
performance. The relationship between underlying capabilities and environmental
management is likely to be complex and contingent. Environmental management and
other capabilities may prove to be complementarities. Depending on industrial
conditions, different bundles of capabilities may be important. Our research suggests
that firm attributes and different strategies for environmental improvement may
moderate the apparent link. It suggests that "When does it pay to be green" may be a
more important question than "does it pay to be green".
Barney, J. 1986. Strategic factor markets: Expectations, luck, and business strategy.
Management Science. 32(10): 1231-1241.
Blacconiere, W. G. and D. M. Patten. 1994. Environmental disclosures, regulatory
costs, and changes in firm value. Journal of Accounting and Economics. 8: 357-
Bragdon, J.H. and J. Marlin. 1972. Is pollution profitable? Risk Management. 19(4).
Chen, K. and R.W. Metcalf. 1980. The relationship between pollution control records
and financial indicators revisited. Accounting Review. 55:168-180.
Cohen, M., S. Fenn, and J. Naimon. 1995. Environmental and financial performance:
Are they related? Working paper, Vanderbilt University.
Denton, K. 1994. Enviro-Management: How Smart Companies Turn Environmental
Costs into Profits. Boston: Prentice Hall.
Deutsch, C. H. 1998. For Wall Street, increasing evidence that green begets green. The
New York Times. July 19, 1998.
Dowell, G., S. Hart, and B. Yeung. 2000. Do corporate global environmental standards
create or destroy value? Management Science. 46(8):1059-1074.
Esty, D. and M. Porter. 1998. Industrial ecology and competitiveness: strategic
implications for the firm. Journal of Industrial Ecology. 2(1):35-43.
Graedel, T. E., and Allenby, B. R. 1995. Industrial Ecology. Englewood, NJ: Prentice
Hamilton, J. 1995. Pollution as news: Media and stock market reactions to the toxic
release inventory data. Journal of Environmental Economics and Management.
Hart, S. and G. Ahuja. 1996. Does it pay to be green? An empirical examination of the
relationship between emission reduction and firm performance. Business
Strategy and the Environment. 5:30-37.
Hart, S. 1997. Beyond greening: Strategies for a sustainable world. Harvard Business
Review. 75(1): 66.
Jaffe, A. B., S. R. Peterson, et al. (1995). "Environmental Regulation and the
Competitiveness of U.S. Manufacturing: What Does the Evidence Tell Us?"
Journal of Economic Literature 33: 132-163.
Jones, K. and P. Rubin. 1999. Effects of harmful environmental events on reputations
of firms. Working paper, Emory University.
Karpoff, J., J. Lott, and G. Rankine. 1998. Environmental violations, legal penalties,
and reputation costs. Working paper, Social Science Research Network.
King, A. 1995. Innovation from differentiation: Pollution control departments and
innovation in the printed circuit industry. IEEE Transactions on Engineering
Management. 42(3): 270-277.
King, A. and Baerwald, S. 1998. Greening arguments: Opportunities for the strategic
management of public opinion. In Better Environmental Decisions: Strategies for
Governments, Businesses and Communities edited by K. Sexton et. al.
Washington, D.C.: Island Press.
King, A. and M. Lenox. 2000. Industry self-regulation without sanctions: The chemical
industry's Responsible Care Program. Academy of Management Journal, 43(4):
Khanna, M., W. R. Quimio, et al. (1998). "Toxic Release Information: A Policy Tool for
Environmental Protection." Journal of Environmental Economics and
Management 36: 243-266.
Kiernan, M. 1998. The eco-efficiency revolution. Investment Horizon. April: 68-70.
Klassen, R. and C. McLaughlin. 1996. The impact of environmental management on
firm performance. Management Science. 42:1199-1214.
Konar, S. and M. Cohen (1997). "Information as Regulation: the Effect of Community
Right to Know Laws on Toxic Emissions." Journal of Environmental Economics
and Management 32: 109-124.
Lindenberg, E. and S. Ross. 1981. Tobin's q ratio and industrial organization. Journal of
McGahan, A. and M. Porter. 1997. How much does industry matter, really? Strategic
Management Journal. July: 5-30.
Meyer, S. 1995. The economic impact of environmental regulation. Journal of
Environmental Law & Practice. 3(2): 4-15.
Muoghalu, M., H. D. Robinson, and J. Glascock. 1990. Hazardous waste lawsuits,
stockholder returns, and deterrence. Southern Economic Journal. 57:357-370.
Nehrt, C. 1996. Timing and intensity of environmental investments. Strategic
Management Journal. 17: 535-547.
Nelson, K. 1994. Finding and implementing projects that reduce waste. In Industrial
Ecology and Global Change edited by Socolow et al.. New York: Cambridge
Panayotou, T. and C. Zinnes. 1994. Free-lunch economics for industrial ecologists. In
Industrial Ecology and Global Change edited by Socolow et al.. New York:
Cambridge University Press.
Porter, M. and C. van der Linde. 1995. Green and competitive. Harvard Business
Review. September-October 1995: 121-134.
Reinhardt, F. 1999. Market failure and the environmental policies of firms. Journal of
Industrial Ecology. 3(1): 9-21.
Rumelt, R. 1991. How much does industry matter? Strategic Management Journal.12:
Russo, M. and P. Fouts. 1997. A resource-based perspective on corporate
environmental performance and profitability. Academy of Management Journal.
Spicer, B.H. 1978. Investors, corporate social performance, and informational
disclosure: An empirical study. Accounting Review. 53: 94-103.
Wernerfelt, B. 1984. A resource-based view of the firm. Strategic Management Journal.
White, M. 1995. The performance of environmental mutual funds in the United States
and Germany: Is there economic hope for ‘green’ investors? Research in
Corporate Social Performance and Policy Supplement. 1:325-346.
White, M. 1996. Corporate environmental performance and shareholder value.
Working paper, University of Virginia.
Appendix - Environmental Performance Measures
To correct for differences in toxicity between emitted chemicals, we follow King &
Lenox (2000) and weight each chemical by its toxicity using the "reportable quantities"
(RQ) database in the CERCLA statute. We construct aggregate releases for a given
facility in a given year (Eit) by summing the weighted releases of the 246 chemicals that
have been consistently a part of the TRI database.
Eit = Σ∀c wcecit (1)
where Eit is aggregate emissions for facility i in year t, wc is the toxicity weight for
chemical c in year t, and eci is the pounds of emissions of chemical c.
Following King & Lenox (2000), we measure relative environmental performance
at the facility level by estimating production function relationship between facility size
and aggregate toxic emissions for each 4-digit Standard Industrial Classification (SIC)
code within each year using standard OLS regression. The relative environmental
performance of a facility (REit) is given by the standardized residual, or deviation,
between observed and predicted emissions given the facility’s size and industry sector.
We must use employees to measure facility size because we have no measure of
production units or sales at the facility level. Thus, if a facility emits more than predicted
given its size and SIC code, it will have a positive residual and a positive score for
environmental impact. We estimate a production function for each industry.
Eit = e
ln Eit =
1jt (ln sit) +
2jt (ln sit)2 +
REit = e(ln Eit – ln E*it) (4)
where E*it is predicted emissions for facility i in year t, sit is facility size, and
2jt are the estimated coefficients for sector j in year t.
To create a firm-level measure of relative environmental performance, we
calculate the weighted-average of the facility-level scores. We weighted the scores by
the percentage of total production that each facility represented for the company.
Relative Emissions nt = Log Σ∀i in n (sit/snt)REit (5)
where sit is facility i size in year t, and snt is firm size.
With the above data in hand, we can differentiate performance within an industry
(Relative Emissions) from the degree to which a firm chooses to operate in dirty or
clean industries (Industry Emissions). We calculated the dirtiness of the sector as the
total emissions for the sector divided by the total number of employees in the sector,
i.e., emissions per employee. We create our firm-level measure (Industry Emissions) of
the firm's tendency to operate in dirty or clean industry sectors by aggregating the
dirtiness of the sectors in which a company owns a facility. In performing this
aggregation, we use a weighted-average, using the percentage of the company’s total
production in each sector for weights.
IEnt = ln(Σ∀i in n (sit/snt)Ejt) (6)
Ejt = Σ∀i in j Eit
where IEnt is weighted industry emissions for firm n in year t, and Ejt is total toxicity-
weighted emissions per employee for industry j in year t.
Measures of Corporate Financial Performance Used
in ‘Pays to be Green’ Scholarship
Measure Description Examples
Tobin’s q Firm market valuation over replacement value
of assets Dowell et. al. (2000)
Return on Assets The ratio of income to total assets Hart & Ahuja (1996),
Russo & Fouts (1997)
Return on Equity The ratio of income to firm equity Hart & Ahuja (1996),
Russo & Fouts (1997)
Return on Investment The ratio of operating income to book value of
assets. Hart & Ahuja (1996),
Russo & Fouts (1997)
Measures of Corporate Environmental Performance Used
in ‘Pays to be Green’ Scholarship
Capital Expenditures on pollution control technology Spicer (1978),
Emissions of toxic chemicals (typical source: TRI) Hamilton (1995),
Hart & Ahuja (1996)
Spills and other plant accidents Karpoff et. al. (1998)
Lawsuits concerning improper disposal of hazardous waste Muoghalu et. al. (1990)
Rewards or other recognition for superior environmental performance Klassen & McLaughlin (1996)
Participation in Environmental Management Standards White (1996),
Dowell et. al. (2000)
Rankings of superior environmental performers (e.g. CEP) White (1996),
Russo & Fouts (1997)
Variable Description Mean Standard
Deviation Minimum Maximum
Tobin’s q Firm market valuation over replacement
value of assets 1.58 0.94 0.28 12.67
Total Emissions Log of total emissions of facilities 5.82 3.28 0.00 13.76
Emissions Average relative emissions of facilities
based on sector and size (in employees) 0.21 0.77 -7.08 9.41
Emissions Average total emissions per employee of
sectors in which the firm operates 0.22 0.50 -1.80 1.62
Firm Size Natural log of firm assets 6.27 1.94 0.76 12.52
Capital Intensity Capital expenditures over sales 0.07 0.06 0.00 1.19
Growth Percent change in sales 0.12 0.44 -0.91 13.36
R&D Intensity Research and development outlays over
firm assets 0.04 0.04 0.00 1.03
Leverage The ratio of debt to firm assets 0.18 0.16 0.00 1.93
Stringency The regulatory stringency of the states the
firm operates 0.51 0.84 0.00 7.01
Permits The number of firm CWA and RCRA
permits over firm size 0.49 0.74 0.00 5.95
Note: n = 4483
1 2 3 4 5 6 7 8 9 10 11
1. Tobin’s q 1.00
2. Total Emissions -0.12 * 1.00
3. Relative Emissions -0.04 0.46 * 1.00
4. Industry Emissions -0.09 * 0.38 * -0.08 * 1.00
5. Firm Size -0.02 0.49 * 0.09 * 0.05 1.00
6. Capital Intensity 0.17 * -0.01 -0.03 0.03 0.11 * 1.00
7. Growth 0.14 * -0.05 * -0.03 -0.01 -0.06 * 0.05 * 1.00
8. R&D Intensity 0.28 * -0.15 * -0.07 * 0.00 -0.05 0.17 * 0.04 1.00
9. Leverage -0.19 * 0.09 * 0.06 * -0.01 0.07 * -0.02 0.01 -0.23 * 1.00
10.Regulatory Stringency 0.00 0.30 * 0.05 * 0.13 * 0.22 * 0.13 * -0.03 -0.09 * 0.09 * 1.00
11. Permits -0.11 * 0.55 * 0.10 * 0.07 * 0.48 * -0.01 -0.06 * -0.15 * 0.06 * 0.27 * 1.00
Note: n = 4483, * p < 0.001
Estimates of Future Financial Performance (Tobin's q t+1)
IV & Fixed
Total Emissions -0.021 **
Relative Emissions -0.036 *
(0.018) -0.029 +
Industry Emissions -0.027
(0.049) -0.076 *
Firm Size -0.219 ***
(0.030) -0.219 ***
(0.030) -0.034 *
(0.014) -0.238 ***
Capital Intensity -0.420 *
(0.198) -0.416 *
(0.187) -1.645 ***
Growth 0.053 *
(0.022) 0.053 *
(0.023) 0.068 **
R&D Intensity 3.429 ***
(0.535) 3.377 ***
(0.535) 5.062 ***
(0.101) -0.330 ***
(0.071) 0.080 *
Tobin's q -0.321 ***
n 4483 4483 4483 3130
Number of Firms 652 652 652 544
F Stat 24.36 *** 22.80 ***
χ2 Stat 505.30 *** 255.09 ***
Adj. R2 0.667 0.667 0.714 0.756
a The sample is slightly smaller due to the inclusion of lagged instruments.
Firm and year dummies are included but not presented in all models.
Standard errors are in parenthesis.
+ p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001 (two-tailed test)
Summary of Findings
Variable Description Result
Total Emissions Log of total emissions of
facilities Associated with financial performance
but direction of the relationship
Relative Emissions Emissions relative to
other facilities of similar
sector and size
Associated with financial performance
but direction of the relationship
Industry Emissions Emissions per employee
for the sectors in which
the firm operates
Apparent but possibly spurious
association with financial
performance. Direction of effect
i Interestingly, a follow-up study by Chen and Metcalf (1980) found that the effect
disappeared when the analysis corrected for differences in size.
ii In contrast, White (1995) found that a group of six mutual funds that employed
environmentally responsible screens performed worse than the S&P 500 in both
nominal and risk-adjusted terms. White resolved the contradiction between the two
findings by concluding that environmental performance and financial performance
are indeed correlated but managers of environmentally oriented mutual funds are
less skilled than managers of other funds.
iii Such a sample is often referred to as a panel or longitudinal data set since we have
multiple observations of the same entity over time.
iv We did not use the more complicated measure of Tobin's q as proposed by
Lindenberg & Ross (1981) because past research in this domain has found little
qualitative difference between this measure and the simplified version used in this
analysis (Dowell, Hart, & Yeung, 1998). We chose to use Tobin's q rather than
accounting measures of financial performance, such as return on assets (ROA) or
return on sales (ROS), because Tobin's q reflects expected future gains.
v OLS is a technique for estimating the parameters of a mathematical model by
minimizing the square of the difference between actual data and the predicted
vi Performing a Hausman test on the random effects model suggests that a random-
effects specification is recommended over a fixed-effects specification.
vii Estimating the model with a lagged dependent variable increases the likelihood of
serial correlation. We use an instrumental variables approach to correct for this
potential problem. The lagged values of the exogenous regressors are used as
instruments. These regressors have the desirable property that they will not be
correlated with the error but will be correlated with the lagged value of the
dependent variable (Kennedy 1993).