THE IMPACT OF ENVIRONMENTAL RISK ON THE COST OF EQUITY CAPITAL: EVIDENCE FROM THE TOXIC RELEASE INVENTORY

Transcription

1 THE IMPACT OF ENVIRONMENTAL RISK ON THE COST OF EQUITY CAPITAL: EVIDENCE FROM THE TOXIC RELEASE INVENTORY Elizabeth Connors College of Management University of Massachusetts Boston 100 Morrissey Boulevard Boston, MA Lucia Silva-Gao College of Management University of Massachusetts Boston Keywords: cost of capital; environmental risk; environmental performance

2 ENVIRONMENTAL PERFORMANCE AND FIRM RISK Abstract: This study investigates the effects of environmental performance, measured as chemical emissions, on cost of equity. Previous literature has focused on the effects of pollution reduction on firm performance relating to accounting measures or market price. The principal motive for inquiry has been to determine the cash flow effects of levels or changes in emissions. Our study measures the effects of environmental performance on risk. We find that firms with high levels of chemical emissions have high cost of equity capital in the electric utility industry. We also show that the relationship is more complex in the chemical industry, in part because chemical companies have a high percentage of international operations and are often diversified. Our results suggest that environmental management of chemical emissions affect the cash flow risk component of firm market value. 2

3 1. Introduction There has been a long-standing disagreement among economists regarding the effects of environmental regulation and environmental performance on firm cash flows. For example, Porter and van der Linde (1995b) argued that firm level (private) costs of pollution are too high and that properly designed regulation could induce innovation that would reduce such costs. They urged companies and regulators to frame their environmental investment decision making in terms of resource productivity and encouraged the efficient use of natural resources in order to improve the economic value of products and competitiveness. Palmer, et al (1995) responded with the classical economic argument that firms do not systematically ignore profitable opportunities and regulators are not capable of correcting market failure. While this particular argument regarding the role of regulation in environmental performance may never come to a satisfactory resolution, voluntary over-compliance with environmental regulations exists. For example, companies that are required to disclose their Toxic Release Inventory (TRI) to the Environmental Protection Agency (EPA) reduced their legal chemical releases by 59% from 1988 to 2006 and by 10% over the period 2001 to 2006 on a larger set of chemicals ( 1 1 The Emergency Planning and Right-to-Know Act (EPCRA) was enacted in response to the Union Carbide chemical release accident in Bhopal, India. Section 313 of EPCRA established the Toxics Release Inventory (TRI) program. TRI is a national database that identifies facilities and chemicals manufactured and used, released and otherwise managed at each reporting facility. The TRI Program has collected and made this information public since A facility must report releases of listed chemicals if it belongs to certain SIC codes industry sectors, has 10 or more full-time equivalent employees and manufactures or processes more than 25,000 pounds or uses more than 10,000 pounds of any listed chemical per year. The EPA s listed chemicals are considered to be the most toxic chemicals in use, not all chemicals. The listed chemicals change from year to year across all industries covered. The database is maintained and data is reported at the facility level. 3

4 The debate over the merits of environmental regulation has provided a platform for a stream of empirical literature that has examined whether, and under what circumstances, improved environmental performance can be win-win and if investors value good environmental performance (eg. Connors and Gao 2008; Clarkson et al ). The reasons for voluntary over-compliance with environmental regulations are varied. In some cases it is unintentional. Pollution reduction may be a secondary consequence of investment in new technologies, or companies may over-invest in regulatory compliance measures due to the discrete nature of capital investments (Arora and Cason 1995). Intentional over-compliance may provide a strategic competitive advantage. First-mover firms may be able to reap the benefits of selling pollution technology and may intentionally over-comply in order to convince regulatory authorities to set tighter standards for the industry, thereby raising the cost of compliance for other firms and restricting competition (Cairncross 1990; Klassen and McLaughlin 1996). Compliance in anticipation of future regulations gives managers flexibility to make costeffective emission reductions without the threat of non-compliance fines and penalties (Boyd 1998; Cairncross 1990). Alternatively, firms may pursue a pollution reduction strategy because the marketing and production related economic benefits are expected to outweigh the costs. Possible benefits are increased market demand through product differentiation and green consumerism, the reduction of the risk of future environmental liabilities and lawsuits, increased productivity, and cost reduction due to the elimination of waste (Porter and van der Linde 1995b; Epstein 1996; Reinhart 1999; Fisher-Vanden and Thorburn 2008). 4

5 The focus of this particular economic debate and ensuing empirical literature has been on the future cash flow implications of current investments to meet or exceed current or future regulatory standards. The purpose of our study is to expand this discussion by examining the effects of pollution performance on firm specific risk. We examine whether improved environmental performance, measured as reduced toxic chemical emissions, reduces cost of equity capital. By disentangling the effects of environmental performance on cost of capital from the effects on expected future cash flows, we can determine if better performance reduces uncertainty for investors and provides a benefit unrelated to operational considerations. Our results are industry dependent. We find that companies in the electric utility industry with high toxic chemical emissions have significantly higher cost of equity capital than those with lower emissions when controlling for beta, leverage, information risk, firm size and growth. These results are robust under several methods of cost of equity estimation and over a seven-year period. Our analysis of a sample of companies in the chemical industry indicates that companies with high chemical emissions have significantly higher cost of equity capital after controlling for sales concentration in pollution intensive products and U.S. sales proportion. These results are important because they provide evidence that environmental performance, a non-financial performance measure that is receiving growing public exposure, is reflected in the cost of capital. Investors are pricing the risk associated with environmental performance because of the uncertainty of the future cash flow effects of the consequences of poor performance such as lawsuits and regulatory exposure. This is particularly interesting because fund managers and financial analysts have not 5

6 enthusiastically supported the notion that environmental performance is a significant determinant of market value. A survey conducted by CSR Europe, Deloitte and Euronext (2003) found that 20% of fund managers and analysts take management of environmental impacts into account when making investment recommendations and 20% of the respondents reported place zero weight on that performance measure. The results also suggest that firms may benefit from environmental risk management through the reduction of toxic chemical emissions. The organization of the remainder of the paper is as follows. In the second section, we present the related literature. In the third section, we discuss implied cost of equity models. In the fourth section, we present our empirical model and variable proxies and detail our sample selection process. We present our results in the fifth section and make our concluding remarks in the sixth section. 2. Related literature 2.1 Environmental performance, firm value, firm performance and risk Several studies provide evidence that environmental performance is priced by the market. Whether the market responds favorably or unfavorably to planned or actual performance improvement depends on the time period covered by the study and how performance is measured. Fisher-Vanden and Thorburn (2008) found significant negative abnormal returns for companies that announce voluntary participation in climate change programs, namely Climate Leaders and Ceres. The authors assumption is that firms that join these programs signal a commitment to reduce carbon emissions to interested parties. The negative returns were lower in carbon-intensive industries, which 6

7 have a higher risk of exposure to future regulatory changes. The authors argue that, overall, their findings support the notion that investors view carbon reduction activities as less profitable than other investment opportunities. Hamilton (1995) investigated whether the first TRI data release in 1989 provided new information to the news media and to the market. He found that companies that reported the highest levels of emissions received the most media attention and experienced the largest negative abnormal returns. Khanna et al. (1998) found negative abnormal returns during the one-day period after the release of TRI emissions data for chemical companies that increased their emissions during the period They found insignificant abnormal returns for companies that decreased their emissions. Connors and Gao (2008) studied the relationship between abnormal returns and TRI information releases in the electric, chemical and pulp and paper industries for the period They found that that investors reward decreases in the levels of pollution emissions, but announcement returns depend on the industry, leverage, and book to market value. The empirical evidence to support the argument that good environmental performance increases financial performance shows mixed results. Hart and Ahuja (1996) found that firms that reduced their TRI emissions from showed increased Return on Assets, Return on Equity and Return on Sales in the three years following the year of reduction. These results applied to companies that initially were relatively poor environmental performers but not for firms that were relatively good performers. The results of this study lend empirical support to the anecdotal evidence that firms can cost effectively reap the benefits of low hanging fruit but face decreasing returns to investment (Cairncross,1990). 7

8 Konar and Cohen (2001) found that good environmental performance, measured as TRI emissions and number of pending environmental lawsuits was positively associated with firm performance measured as Tobin s Q in a sample of S&P 500 firms in eleven industries during King and Lenox (2002) found that companies that had lower than average TRI emissions showed higher Return on Assets and Tobin s Q in the following year during the period Clarkson, et al. (2004) found that environmental capital expenditures in a sample of pulp and paper companies over the period resulted in an increase in firm market value for low-polluting firms, but not for high polluting firms. Clarkson, et al. (2006) found that firms in a sample of four industries that significantly reduced their TRI emissions (moved from the highest two quartiles to the lowest two quartiles) experienced higher subsequent Tobin s Q and Return on Operating Assets in the two years following the reduction. The association between improved pollution performance and concurrent or subsequent improvement in operating performance supports the argument that firms may be able to benefit from good environmental performance under certain conditions. Studies that show a positive relationship between pollution performance and firm value further support the argument that environmental management improvements may have an impact on future cash flows. While the increase in the market value of a company may in part be due to changes increases in the future cash flows from greater revenues and/or lower costs relating to improved environmental performance, it may also be due to a lower cost of capital reflecting a reduction in the perceived riskiness of future cash flows. Heinkel et al. (2001) introduced a theoretical model to show that exclusionary ethical investing leads to 8

9 polluting firms being held by fewer investors because green investors will not invest in polluting firms stock. This lack of risk sharing among non-green investors leads to lower stock prices for polluting firms and to an increase in their cost of capital. Merton (1987) develops a model of capital market equilibrium with incomplete information in which the firm s cost of equity declines as its investor base expands. Empirically, Garber and Hammitt (1998) examined the effect of Superfund liabilities on costs of equity, measured as the capital asset pricing model (CAPM) and beta, for a sample of companies in the chemical industry. They found no relationship between balance sheet liabilities identified to cover Superfund remediation costs and the cost of equity for small firms, but were able to find a robust positive relationship for large firms. It should be noted that Superfund liabilities reflect cleanup costs for past emissions and spills as well as costs to cleanup purchased properties in which the party responsible for cleanup was not the cause of the pollution. As such, Superfund liabilities entail a high level of uncertainty relating to final cleanup costs but may not provide a strong signal of future environmental performance. The findings of Feldman, et al. (1997) support a positive association between qualitative measures of environmental practices and TRI emissions on firm beta and stock price. Sharfman and Fernando (2008) find that TRI emissions reductions and high social performance scores are associated with lower weighted average cost of capital for a sample of S&P 500 firms. This study has several methodological issues that make the results difficult to interpret. For example, their models test the association between 2001 environmental performance measured as TRI emissions, as well as a weighted factor that includes TRI emissions, and 2002 weighted average cost of capital. However, the EPA 9

10 releases TRI data with a significant lag. The 2001 TRI data was not released to the public and to research organizations until June 30, Our models avoid this problem by matching our measures of risk with the TRI data releases by year. We employ models based on ex ante expectations by investors and on analysts forecasts to determine whether environmental risk is priced by the market. Previous studies provide evidence that environmental performance is valued by investors. However, market prices reflect both changes in the expectations for future cash flows and changes in the risk perception of these cash flows. Empirical studies which test the relationship between environmental performance and market price do not disentangle these effects. Our research adds to the existing literature by explicitly considering the value of environmental risk as a complement to the cash flow effect component of the market price. 2.2 Environmental risk and the cost of equity capital Cost of equity accounts for the cost of business risk and the cost of financial risk. We investigate whether environmental risk affects these two classes of risk by measuring if it is priced in the cost of equity capital. To determine the impact of environmental risk on the cost of equity we require a measure of firm-specific cost of equity capital and use the implied cost of equity to proxy for cost-of-equity capital. Several studies on cost of equity use approaches that rely on ex post and realized returns. However, cost of capital estimates derived from realized returns have proved to be disappointing. Fama and French (1997) showed that standard errors of more that three percent per year are typical when the single factor domestic CAPM and the three-factor model of Fama and French (1993) are used to estimate the cost of equity capital for forty-eight industries. Estimates 10

11 of cost of capital for individual firms are likely to be even less accurate due to imprecise estimates of the factor risk premia and risk loadings. Another argument in favor of using approaches based on expected returns rather then realized returns is the fact that environmental performance is likely to have in impact on both future cash flows and on the risk of these cash flows. Blacconiere and Northcutt (1997) documented a negative market reaction to legislative actions relating to the Superfund and Reauthorization Act of 1986 for a sample of chemical companies. In addition to the risk related to enacted legislation, a chemical spill for a single company can affect the perceived risk of an entire industry. For example, Blacconiere and Patten (1994) identified a significant negative market reaction for chemical companies in response to the Union Carbide chemical spill that occurred in Bhopal, India in The threat of future environmental regulation introduces uncertainty in the estimation of future cash flows. During Senate hearings on greenhouse gas regulation in 2006, corporate executives and trade group representatives for the electric utility industry urged Congress to take legislative action so that the industry could make investment and operational decisions and reduce risk to investors (Lohr 2006). The risk of accidental spills and uncontrollable events also affect the uncertainty regarding future cash flows. Recognition of the threat of terrorist attacks on domestic chemical production facilities has intensified since September 11, After a length rulemaking process, the Department of Homeland Security (DHS) published Appendix A to the Chemical Facility Anti-Terrorism Standard on November 20, Chemical companies that produce or 11

12 handle any of 300 hazardous chemicals over threshold levels must submit a Top Screen to the DHS. 2 As such, the impact of environmental risk is not adequately addressed by examining price responses. It is necessary to disentangle the impact of a change in the cost of equity from changes in forecasted cash-flows and to isolate the cost of equity effect (Botosan and Plumlee (2005)). 3. Implied cost of equity models The implied cost of equity is the discount rate that equates current stock prices to expected future payoffs. Most of the models that propose to estimate the implied cost of equity presented in the literature are derived from the dividend discount model, represented as: E(dividend per share t) P (1) 0 = t= 1 + ( 1 r ) e t where: P 0 = price at time 0; r e = estimated cost of equity capital; E(.) = expectations operator; There are several recent papers that examine empirical methods for computing the implied cost of equity capital given stock prices and expectations of future earnings (e.g., Botosan 1997; Gebhardt et al. 2001; Claus and Thomas 2001; Botosan and Plumlee 2002; Easton and Monahan 2003; Gode and Mohanram 2003; Easton 2004; Easton and Monahan 2005; Ohlson and Juettner-Nauroth 2005). These models use forecasts of future 2 For the Final Rule see 12

13 earnings expectations and the current stock price as inputs in a valuation model in order to derive the cost of capital. However, different estimation methods use distinct valuation models and different assumptions regarding terminal value computation. Several studies evaluate alternative empirical measures of implied cost of equity (Guay et al. 2005; Easton and Monahan 2005; Botosan and Plumlee 2005). Botosan and Plumlee (2005) conclude that expected return estimates based on Value Line target prices and dividend forecast and the Easton (2004) approach are superior to alternative measures of expected return, because they are consistently and predictably related to risk proxies, such as beta, size (market value of equity), leverage, residual risk (measured by market-to-book ratio) and growth. Due to the potential error in measuring implied cost of equity, we estimate several alternative proxies that use different underlying models and assumptions and different data. We report results using several alternative estimates to ensure that our results are robust to method choice. We use the following three methods to estimate the implied cost of capital: Gebhardt et al. (2001), labeled GLS, Gode and Mohanram (2003), labeled GM, and Easton (2004), labeled PEG. 3.1 Gebhardt et al. (2001) GLS Gebhardt et al. (2001) introduce a methodology to estimate the implied cost of capital based on the residual income model developed in Ohlson (1995). The residual income (or economic profit) model is algebraically equivalent to the discount growth model, but expresses the values in accounting terms. The model is as follows: P 0 T = BVE + 0 t= 1 1 ( ROEt re ) t ( 1+ r ) e BVE t 1 + TV (2) 13

14 where P 0 is the price at time zero, BVE t is the book value at time t, ROE t is the return on beginning equity for year t, r e is the cost of equity capital, and TV is the terminal value at the end of the forecasting horizon. For their empirical estimation, they use the current book value of equity, analysts forecasts of future earnings, and long-term growth rates. GLS assume mean reversion to the industry median ROE after year 3 and over a 12-year period. The following equation can be solved by numerical approximation: 11 0 = BVE0 + t = 1 P ( ROEt re ) t ( + r ) BVE + ( ROE12 re ) r ( 1+ r ) 11 t 1 11 (3) 1 e e e BVE where FEPSt ROE t = and FEPS t equal to forecasted earnings per share for year t. BVE t Gode and Mohanram (2003) GM The Gode and Mohanram (2003) estimation method is an implementation of the Ohlson and Juettner-Nauroth (2003) model which equates EPS and EPS growth with the capitalized next-period EPS and future abnormal growth in EPS. They estimate the cost of capital as follows: P 0 = FEPS r e 1 FEPS + 2 FEPS 1 r re ( FEPS1 k. FEPS1 ) ( r g) e e (4) where FEPS 1 and FEPS 2 are analysts forecasts obtained from I/B/E/S, k is the actual dividend payout ratio and the growth rate g is set equal to risk-free interest rate minus 3%. Gode and Mohanram (2003) provide the following analytical solution: FEPS r e (5) ( g ( 1) ) 2 1 = A + A + 2 γ P0 14

15 1 k * FEPS ( ) 1 A = γ 1 + (6) 2 P0 g 2 ( FEPS FEPS ) 2 1 = (7) FEPS Easton (2004) - PEG Easton (2004) shows that under the assumption of zero dividends and no growth in abnormal earnings beyond the forecast horizon (after year 2) the cost of capital is proportional to the inverse of the PEG ratio: r e FEPS FEPS P 2 1 = (8) 0 4. Empirical model, variable proxies and sample selection 4.1 Empirical model There are several risk factors that have been shown in the literature to affect the cost of equity capital, including market beta, leverage, information risk, firm size, and growth. We estimate the following models, based on the risk proxies used by Botosan and Plumlee (2005): r e, it = γ + γ UBETA 0 + γ DISP 5 1 it it + γ LTG 6 + γ BM 2 it it + γ MVE + γ EMISS 7 3 it it + γ LEV + γ YEAR 8 4 i + ε it (9) and r e, it = γ + γ UBETA 0 + γ DISP 5 1 it it + γ LTG 6 + γ BM 2 it it + γ MVE + γ EMISS / MD 7 3 it + γ LEV it 4 + ε it (10) where: 15

16 UBETA = unlevered beta; BM = book to market; MVE = market value of equity; LEV= leverage; DISP = dispersion measure of one-year ahead analysts earnings forecast; LTG = mean analysts forecasted long term growth rate; EMISS = tons of chemicals emissions divided by US sales EMISS/MD = EMISS for the firm divided by the median of EMISS for the industry and for the same year. We consider the following control variables for other risk factors: Unlevered BETA (UBETA): Sharpe (1964), Linter(1965) and Black(1972) formalize the prediction that expected return should be positively related to beta. We estimate unlevered beta as UBETA = BETA D 1+ E and use CRSP data to estimate the market beta at December 31 st of each year via a market model that considers one year of daily returns. Leverage (LEV): According to Modigliani and Miller (1958), risk of equity capital increases with leverage. Thus, we expect that the cost of equity is increasing with leverage. Leverage is the ratio of long-term debt to market value of common equity. Dispersion (DISP): Dispersion is a proxy for information risk. Theory suggests that better information is associated with a lower cost of equity (e.g. Clarkson et al. 1996; Handa and Lin 1993). We measure dispersion as the coefficient of variation of the mean I/B/E/S analyst 1-year-ahead earnings per share forecast, considering the first and third quartiles. We expect DISP to be positively related to cost of capital. Book to market (BM ): BM is the ratio of book value of common equity to market value of common equity. Fama and French (2004) use Ohlson s (1995) residual income framework to formalize the valuation role of the market-to-book ratio (MB) in expected returns and predict a negative relation between MB and expected return. Fama and 16

17 French (1993) develop a three factor asset pricing model that includes beta, size and market-to-book, and show that this asset-pricing model out performs the CAPM. BM should be positively associated with cost of equity. Market Value of Equity (MVE): Berk (1995) demonstrates that size will exhibit a negative relation with expected returns, as a residual risk factor, in any incomplete model of expected returns. We expect that cost of equity is decreasing with MVE, our measure of size. Growth (LTG): LTG is the mean I/B/E/S analyst long-term growth in earnings per share forecast for each year of estimation. La Porta (1996) shows empirically that high expected-growth stocks have higher standard deviations of returns and higher betas then low expected-growth stocks. Hence, we expect the coefficient for this variable to be positive. Environmental Performance (EMISS): Consistent with several prior studies (Clarkson et al. 2006, 2004; King and Lenox 2002; Konar and Cohen 2001) we measure environmental performance as annual TRI emissions in pounds. The EPA TRI reports provide information regarding annual emissions of each listed chemical and a summary of emissions released as air emissions, surface water discharges, released to land, underground injection and transfer to off-site disposal. For the purposes of this study, annual emissions have been aggregated across chemicals and across the various methods of release. The EPA reports TRI at the facility level. We have aggregated the TRI reports to the parent company level. Facility ownership has been determined by the review of SEC filed forms 10-K, corporate and facility websites, and through public announcements of acquisitions and disposals of subsidiaries and facilities. In order to 17

18 account for production volume we have normalized the pounds of emissions reported by U.S. sales. Given distinctive characteristics of the chemical industry, we introduce two additional variables into our analysis: Proportion of revenues generated in the U.S. (US.PROP): Is equal to the ratio of U.S. sales to total sales of the company. Operations realized outside the U.S. are not subject to U.S. regulations or reporting requirements and non-domestic emissions are not reflected in TRI data. However, the equity premium reflects global risk. This variable captures the significance of U.S. environmental performance in the overall risk assessment of the firm. Proportion of revenues related to highly pollutant activities (CH.PROP): Is equal to the proportion of sales derived from products in SIC code 28 (Chemicals and Allies), 29 (Petroleum Refining and Related Industries), 30 (Rubber and Miscellaneous Plastic Products) and 33 (Primary Metal Industries). This information was collected from sales and primary SIC codes by segments, as reported by the company. 4.2 Variable proxies We have estimated the implied cost of equity capital using different methods, as described in the previous section, and different data. GLS is estimated using equation (3) and bases on mean one-year ahead earnings, two-year ahead earnings and long term earnings growth rates analysts forecasts from the I/B/E/S database. We have calculated dividend payouts based on data from Compustat. Following the same procedure as in the original paper from Gebhardt, Lee and Swaminathan (2001), when earnings are negative 18

19 we estimate dividend payout ratios as dividends divided by 6% of total assets. We limit the values for dividends payout to values between 0 and 1. Also following the procedures used by Gebhardt, Lee and Swaminathan (2001), we eliminate all observations in which the book value of equity is negative. We estimate the PEG model following equation (8). We estimate the model using both short-term and long-term forecasts. PEG1 labels the estimates obtained from using one-year ahead and two-year ahead forecasts in the variables FEPS1 and FEPS2. In years for which we cannot obtain two-year ahead earnings forecasts from I/B/E/S, we estimate this value as equal to FEPS1*(1+LTG). Following Botosan and Plumlee (2005) we also estimate PEG (labeled PEG2) using twoyear ahead analysts earnings forecasts and FEPS3=FEPS2*(1+LTG), reflecting in this way long-term forecasts instead of short-term earnings forecasts. Accordingly, we estimated GM based on both FEPS1 and FEPS2 (denoted by GM1) and based on FEPS2 and FEPS3=FEPS2*(1+LTG) (denoted by GM2). As is conventional, we use 10-year Treasury Bonds rates as proxies for the risk free rate when estimating risk premiums. 4.3 Sample Selection Our sample is comprised of companies in the electric (SIC 49) and chemical (SIC 28) industries that file with reportable TRI emissions and have information available both in the I/B/E/S, CRPS and Compustat databases between 2001 and These industries have been chosen for study because, during the time period of interest, they have among the highest total emissions and the highest number of public companies. We have excluded all observations for which data necessary to compute any of the three approaches of cost of capital (GLS, PEG and GM) is missing. Our final sample includes a total of 326 company/year observations in the electric industry, including between 45 19

20 and 48 companies per year, and 262 company/year observations in the chemical industry, representing between 32 and 41 companies. Table 1, Panel A provides evidence of some variability in terms of environmental performance in our electric industry sample, with values of emissions that range from pounds per thousand US sales for the first quartile to pounds per thousand US sales for the third quartile. Table 1, Panel B shows that the mean and median total emissions as well as the variability of environmental performance is lower in our sample of chemical companies. First quartile emissions are.243 pounds per thousand US sales and third quartile emissions are pounds per thousand US sales. Insert Table 1 Table 2, Panel A shows that electric companies included in our sample vary widely in size, as measured by sales and market value of equity. There is little variability in terms of beta (with values of 0.53 and 0.93 for the first and third quartiles), book-tomarket (varying between 0.52 and 0.74 between the first and third quartiles) and leverage (with values of 0.54 and 1.54 for first and third quartiles). Chemical companies exhibit higher levels of systematic risk then electric companies, as shown by the mean and median values of beta (1.04 and 1.01, respectively) in Table 2, Panel B. In our sample, chemical companies are less leveraged then electric companies (mean.36, median.22). They have lower book to market ratios (mean.41, median.36), and higher long-term mean (11.46) and median (10.00) forecasted growth rates. Most of the revenue of electric companies is related to operations within the US while chemical companies have a 20

21 considerable proportion of their operations in other countries, as shown by the mean and median values of 58% and 55% of proportion of sales in the US (US.PROP). The average chemical company in our sample has a proportion of revenues of 85% related to highly polluting business activities, as reported for the variable CH.PROP. Insert Table 2 5. Results The descriptive statistics for the risk premium proxies are presented in Table 3. Panel A shows an average risk premium for electric companies that range from.022 for the first quartile and.05 for the third quartile. Panel B shows an average risk premium for chemical companies of.03 in the first quartile and.06 in the third quartile. Insert Table 3 Table 4 shows correlations between estimations of cost of capital for the sample of electric companies using the three different approaches that are similar to those found in previous literature. Botosan and Plumlee (2005), for example, find a correlation between GLS and PEG of 0.3, between GLS and GM of 0.36 and between PEG and GM of The correlation between our estimates based on long-term forecasts (PEG2 and GM2), comparable to the procedures followed by Botosan and Plumlee (2005), are 0.85 between PEG and GM, 0.47 between GLS and PEG and 0.45 between GLS and GM, as shown in Panel A. We also include in our analysis the average of the five measures of 21

22 risk premium (GLS, PEG1, PEG2, GM1 and GM2), which we label AVERAGE. The use of the average of the estimates of the cost of equity may mitigate some of the problems associated with each of the base models. Correlations between the risk factor variables and estimates of cost of equity also mirror the results in Botosan and Plumlee (2005): UBETA, BM, LEV, DISP and LTG are positively correlated and MVE is negatively correlated with the measures of risk premium. Panel B shows that both variables that proxy for environmental performance, EMISS and EMISS/MD, are positively correlated with all cost of equity estimates, with coefficients that range from 5% to 12%. Emissions are positively correlated with long-term growth rate, market value of equity and ROE and are negatively correlated with book-to-market as shown in Panel C, suggesting that larger companies and companies with larger growth prospects are the heaviest polluters. In our sample of electric companies, larger companies also have higher growth prospects. Insert Table 4 In the sample of chemical companies the correlation between our estimates based on long-term forecasts (PEG2 and GM2), comparable to the procedures followed by Botosan and Plumlee (2005), are 0.77 between PEG and GM, between GLS and PEG and 0.37 between GLS and GM as presented in Table 5, Panel A. In contrast with the values obtained in the electric company sample, the correlations between risk proxies are not in the direction that is expected for all variables; UBETA and LTG are negatively correlated with risk premium for many measures of risk (Panel B). 22

23 Compared to electric companies, chemical companies have large proportions of revenues generated outside the US, as can be seen in Panel C. There also tends to be more diversity in their business processes and products. In order to account for these differences we have introduced a variable to control for the proportion of chemical company revenues realized in the US (US.PROP) and a variable that captures the proportion of revenues related to highly pollutant business activities (CH.PROP). As presented in Table 5, Panel B, risk premium is positively correlated with the proportion of operations in U.S. and with the proportion of sales concentrated in pollution intensive products. The values in Table 5, Panel C suggest that companies with larger proportions of revenues in the U.S. have lower levels of emissions. However, it is difficult to draw conclusions from this relationship. TRI emissions data reflects pollution emissions for operations within the United States only and global emissions data is not collected by any entity. As such, our measure of environmental performance will not reflect the entire level of environmental risk incorporated in the cost of equity capital because we are not capturing the overall environmental risk exposure of each company in the chemical industry. Our use of the measure US.PROP adjusts for this discrepancy. Insert Table 5 Table 6 presents regressions of estimates of cost of equity capital on risk proxies for the electric sample. The results show that the PEG method better predicts cost of equity. This finding is consistent with the findings in Botosan and Plumlee (2005); they elect the PEG method as a dominant alternative when compared with the GLS and GM 23

24 approaches. Our results also suggest that estimates based on long-term earnings forecast are more reliable then estimates based on short-term forecasts. However, the inclusion of LTG as a control variable artificially increases the explanatory power of these models, since the construction of the estimate is based on long-term growth rates of earnings forecasts. The models based on PEG2 show better performance when compared with the remaining models. In the first model that uses PEG2, the variables UBETA(t-stat. =5.513, p < 0.01), BM (t-stat. =5.557, p < 0.01), LEV(t-stat. =6.319, p < 0.01) and LTG (t-stat. =19.792, p < 0.01) are significant and represent the expected relationship between the risk factors and risk premium. Botosan and Plumlee (2005) find that estimates obtained from this model are consistently and predictably related to risk proxies. In the models that use GLS estimates only the book-to-market variable has explanatory power on the variation of the risk premium estimate. In all models, MVE, an explanatory variable that proxies for size, is either insignificant or is significant but presents a sign opposite of the predicted, suggesting that the cost of equity is larger for larger companies. Our measure of dispersion (DISP) is significant only when BETA is used as the dependent variable (t-stat. =1.928, p < 0.10 and t-stat=1.959, p<0.10). DISP is also significant in the model that uses GM2 but the relationship between cost of equity and DISP is not as expected. Cost of equity should increase with variability in analysts earnings forecasts due to information risk. Insert Table 6 24

25 For all models and cost of equity specifications, after controlling for systematic risk, book-to-market, leverage, size, information risk and expected growth in earnings, the variables that account for the levels of chemical emissions (EMISS and EMISS/MD) significantly explain variation in the risk premium. For all estimates of cost of equity, the relationship between risk premium and levels of chemical emissions is positive and statistically significant at 1% level. As such, our results are robust under several methods of cost of equity estimation. One might expect that better environmental performance is related to better economic performance, as shown in several studies. It would then follow that the relationship between cost of equity and levels of emissions is actually being driven by the relationship between cost of equity and economic performance. In table 7 we show the decomposition of the model based in the average of the estimates of cost of equity (AVERAGE) in several stages, including the introduction of ROE as a proxy for economic performance. After controlling for economic performance, the relationship between EMISS/ MD and the cost of equity remains statistically significant (t-stat. = 4.255, p<0.01). Insert Table 7 The results for the sample of chemical companies are not as strong as those we have obtained for the electric sample, Table 8 exhibits the results of the regressions of the different estimates of risk premium on risk proxies and variables that proxy for environmental performance. Our results show that some of the control variables for risk premiums are not significant. For example, BETA and BM are not significant in any of 25

26 the models. The variables that proxy for environmental performance, are significant in the models with PEG2 and GLS, but exhibit a relationship with the risk premium opposite to our predictions. Our results suggest that the risk premium is higher for companies with lower levels of chemical emissions. However, the average company in our sample has a little more than half of their operations in the U.S. Therefore, our TRI measure is not reflecting the full operational and strategic environmental decisions of the companies. When we introduce the variable US.PROP and the interaction variable EMISS*US.PROP in the model, we conclude that the risk premium is indeed higher for companies with higher levels of chemical emissions, but this relationship depends on the proportion of revenues of the company generated in the U.S., as suggested by the positive sign and significance of the interaction variable in the models where we use the risk proxies PEG2 (t-stat. = 1.794, p<0.1) and GLS (t-stat. = 1.764, p<0.1). High U.S. proportion of sales combined with high chemical emissions is associated with high risk. If we further introduce the variables CH.PROP and EMISS*CH.PROP, we find that the relationship between risk premium and levels of chemical emissions also depend on the proportion of sales in segments of the company classified with primary SIC codes in high pollutant industries. Specifically, our PEG2 risk proxy model shows that high concentration of sales of pollution intensive products coupled with high emissions is associated with a high risk premium, (t-stat. = 2.913, p<.01). Insert Table 8 26

27 6. Conclusion The purpose of this study is to investigate the effect of environmental performance on firm-level cost of equity capital. We measure the effects of the levels of toxic chemical emissions and relative emissions on cost of equity within the electric utility and chemical industries over a period of seven years. In our sample of electric companies, we find that both levels of emissions and relative environmental performance are positively associated with risk. Our findings are robust while including control variables hypothesized to affect cost of equity and do not appear to be driven by firm financial performance. In our sample of chemical companies, the relationship between cost of equity and emissions is conditioned on the proportion of U.S. revenues in the company. Nearly half of the revenues of the sample companies are sourced to geographic regions outside of the US. These operations are not subject to US regulations or reporting requirements and are not reflected in TRI data. Firms manage global risk, but global chemical emissions data is not available at the firm level. Our analyses show that it is necessary to account for the proportion of revenues that relate to the TRI emissions generated. It is also possible that risk in the chemical industry is dependent on the toxicity of the chemicals emitted as well as the levels of emissions. In our analyses, we assume that all TRI releases are equally toxic because they are the most toxic chemicals in use. It may be that cost of equity capital relies on a finer measure of toxicity in the chemical industry. Our analyses of the chemical industry also show that the degree of concentration in sales of pollution intensive products is associated with risk. Companies in the chemical industry are often diversified in their production. Investors assessment of firm 27

28 risk associated with chemical emissions grows with the proportion of revenues derived from pollution generating production. Our results suggest that firm level chemical emissions are associated with the uncertainty of firm level future cash flows. The findings are important because they indicate that environmental performance management affects risk in addition to the traditionally debated cost and benefit effects on cash flows. Future research on cost of equity capital should consider the effects of chemical emissions on risk in highly polluting industries. Other non-financial performance measures may affect cost of equity capital. Some environmentally related performance measures include carbon emissions, high profile chemicals, and water use (Makower 2009). While data sources are evolving, these areas of inquiry could provide fruitful in our understanding of the firm level factors that affect risk. 28

29 References Arora, S., and Cason, C. S. (1995.) An experiment in voluntary environmental regulation: Participation in EPA's 33/50 Program. Journal of Environmental Economics and Management 28: Berk, J. (1995.) A critique of size-related anomalies. Review of Financial Studies (Summer): Blacconiere, W. G., and Northcutt, D. (1997.) Environmental information and market reactions to environmental legislation. Journal of Accounting, Auditing and Finance 12(2): , and Patten, D. M. (1994.) Environmental disclosures, regulatory costs and changes in share value. Journal of Accounting and Economics 18: Botosan, C.A. (1997.) Disclosure level and the cost of equity capital. The Accounting Review 72(July): , and Plumlee, M. A. (2002.) A re-examination of disclosure level and expected cost of capital. Journal of Accounting Research 40(1): (2005.) Assessing alternative proxies for the expected risk premium. The Accounting Review 80(1): Boyd, J. (1998.) Searching for the profit in pollution prevention: Case studies in the corporate evaluation of environmental opportunities. Resources for the Future. Clarkson, P., Guedes, J., and Thompson R. (1996.) On the diversification, observability, and measurement of estimation risk. Journal of Financial and Quantitative Analysis (March): , Li, Y., and Richardson, G. D. (2004.) The market valuation of environmental capital expenditures by pulp and paper companies. The Accounting Review 79(2): , and Vasvari, F. P. (2006.) Does it really pay to be green? Determinants and consequences of proactive environmental strategies. Working Paper. Claus, J., and Thomas, J. (2001.) Equity premia as low as three percent? Empirical evidence from analysts earnings forecasts for domestic and international stock markets. Journal of Finance 56: CSR Europe, Deloitte and Euronext. (2003.) Investing in responsible business. Available at 29

30 Derwall, J., Guenster, N., Bauer, R., and Koedijk, K. (2005.) The Eco-Efficiency premium puzzle. Financial Analysts Journal 61: Dooley, R., and Fryxell, G. (1999.) Are comglomerates less environmentally responsible? An empirical examination of diversification strategy and subsidiary pollution in the U.S. chemical industry. Journal of Business Ethics 21: Dowell G., Hart, S. L., and Yeung, B. (2000.) Do corporate global environmental standards create or destroy market value? Management Science 8: Easton, P. (2004.) PE ratios, PEG ratios, and estimating the implied expected rate of return on equity capital. The Accounting Review 79: , and Monahan, S. (2005.) An evaluation of accounting-based measures of expected returns. The Accounting Review 80: Epstein, M. J. (1996.) Measuring Corporate Environmental Performance. Montvale, NJ: Institute of Management Accountants. Fama, E., and French, K. R. (1992.) The cross section of expected returns. Journal of Finance 47: , and. (1993.) Common risk factors in the returns of stocks and Journal of Financial Economics 33: bonds., and. (1997.) Industry costs of equity. Journal of Financial Economics 43: Feldman, S. J., Soyka, P. A., and Ameer, P. G. (1997.) Does improving environmental management systems and performance result in higher stock price? Journal of Investing 4: Fisher-Vanden, K., and Thorburn, K.S. (2008.) Voluntary corporate environmental initiatives and shareholder wealth. Working Paper. Dartmouth College. Garber S, and Hammitt, J. K. (1998.) Risk premiums for environmental liabilities: Superfund and the cost of capital. Journal of Environment and Economic Management 36: Gebhardt, L., Lee, C., and Swaminathan, B. (2001.) Toward an implied cost of capital. Journal of Accounting Research 39: Gode, D., and Mohanram, P. (2003.) Inferring the cost of capital using the Ohlson Juettner Model. Review of Accounting Studies 8:

31 Guay, W., Kothari, S. P., and Shu, S. (2005.) Properties of implied cost of capital using analysts forecasts. Working paper. University of Pennsylvania. Hamilton, J. T. (1995.) Pollution as news: Media and stock market reactions to the Toxics Release Inventory data. Journal of Environmental Economics and Management 28: Handa, P., and Linn, S. (1993.) Arbitrage pricing with estimation risk. Journal of Financial and Quantitative Analysis 29(1): Hart, S. L., and Ahuja, J. (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: Heinkel R., Kraus, A. and Zechner, J. (2001.) The effect of green investment on corporate behavior. Journal of Financial and Quantitative Analysis 36(4): Jaffe, A. B, Petersen, S. R., Portney, P. R. and Stavins, R. N. (1995.) Environmental regulation and the competitiveness of U.S. manufacturing: What the evidence tells us. Journal of Economic Literature 33: Khanna, M., Quimo, W. R. H., and Bojilova, D. (1998.) Toxics Release Information: A Policy Tool for Environmental Protection. Journal of Environmental Economics and Management, (36): King, A. A., and Lenox, M. J. (2002.) Exploring the locus of profitable pollution reduction. Management Science (February): Klassen, R. D., and McLaughlin, C. P. (1996.) The impact of environmental management on firm performance. Management Science (August): Konar, S., and Cohen, M. A. (2001.) Does the market value environmental performance? Review of Economics and Statistics (May): La Porta. R. (1996.) Expectations and the cross-section of stock retums. Journal of Finance 51: Lintner J. (1965.) The valuation of risk assets and the selection of risky investments in stock portfolios and capital budgets. Review of Economics and Statistics 47(1): Lohr, S. (2006.) The cost of an overheated planet. The New York Times (December 12). Mackey, A., Mackey, T. B., and Barney, J. B. (2007.) Corporate social responsibility and firm performance: Investor preferences and corporate strategies. Academy Management Review 3: