The Effect of Environmental Firm Performance
on Investors’ Expectations: Value-Adding or Value-Destroying?
WENDY CHANTAL PLAKKE (1611747)
University of Groningen
Thesis MSc. Business Administration, Finance
Supervisor: prof. dr. L.J.R. Scholtens Second supervisor: L. Dam
June 2012
ABSTRACT
This paper examines the effect of a company’s environmental performance on the required rate of return demanded by investors, measured by the cost of equity. I hypothesize that a firm with superior environmental performance has a lower cost of equity than a firm with inferior environmental performance, since it leads to a competitive advantage that improves financial results. I use panel data consisting of 930 European listed companies from 21 countries for the period 2006 to 2010. I find that a higher amount of pollution results in a lower cost of equity, suggesting that superior environmental firms have a higher required rate of return, thus I reject the hypothesis. This suggests that investors perceive investing in corporate responsible firms as risky, and demand a higher rate of return for their investment.
2
1. INTRODUCTION
This paper investigates what the consequences of the growing market segment of socially responsibility investing (SRI) are for the investment climate of organizations. In recent years, the concept sustainable development has received increasing attention from governments and policy makers of countries from all over the world. Consequently, an increasing importance of corporate social responsibility has developed, and since recently investors show a preference for investing in projects that are socially responsible. In this paper, I examine the effect of a firm’s environmental performance on the required rate of return demanded by investors. I hypothesize that firms with superior environmental performance have a lower required rate of return than firms with inferior environmental performance. Several studies have found evidence that a positive relation exists between environmental and financial performance based on resource-based theory (Waddock and Graves, 1997; Clemens, 2006; Nakao et al, 2007; Clarkson et al, 2011). However, other authors question these results and built a more comprehensive empirical model to find a neutral effect of environmental performance (McWilliams and Siegel, 2000, Surroca, Tribó and Waddock, 2010). Another stream of researchers is more interested in the effect of environmental performance from investors’ perspective, and they examine the effect on firm market value or the cost of capital. While most authors find a positive influence of environmental performance on firm value (Derwall et al, 2005; Jo and Harjota, 2011; Ziegler, Busch and Hoffman, 2011; Berthelot, Coulmont and Serret, 2012), other authors find a negative effect (Hassel, Nilsson, and Nyquist, 2005). Most authors find that environmental performance leads to a lower cost of equity (Feldman, Soyka, and Ameer, 1997; Sharfman and Fernando, 2008; Ghoul et al, 2010), although Hamilton, Jo and Statman (1993) find no effect on stock returns or the cost of equity. A third body of research has created theoretical models to predict the reaction of investors to the pollution of firms, and concludes that the relationship between environmental performance and the cost of equity depends on the supply and demand for socially responsible firms, thus the percentage of green investors in the market currently present. In all, results are mixed and the influence of green performance on both firm performance and the cost of equity must still be determined.
3
consumption, water withdrawal and waste production to measure environmental performance. To compare results across firms, I create a relative measure of the indicators, by scaling them to common equity and total revenues. The research methodology is based on Ohlson’s valuation accounting model (Ohlson, 1995) and the cost of equity is determined using the Capital Asset Pricing Model developed by Sharpe (1964).
The results show that a higher amount of pollution leads to a lower cost of equity, thus the required rate of return is higher for firms with superior environmental performance. This implies that firms with higher amounts of pollution exhibit cheaper financing for investors. The remainder of the paper is structured as follows. In section two, I discuss the relevant literature. In section three the data is described and the methodology is explained. Next the results are analyzed. The final section draws conclusions and discusses limitations.
2. LITERATURE REVIEW
4
developed; 95 percent of the world’s 250 biggest companies disclose sustainability performance information in 2011, almost an 80 percent increase compared to 2008 (www.globalreporting.org). GRI promotes transparent reporting for the economic, environmental and social performance areas. The GRI G3 Guidelines are the latest guidelines, and the environmental reporting section includes thirty indicators divided into eight categories, including among others the total set of greenhouse gas emissions, energy consumption, water withdrawal and the percentage of recycled materials. The complete guidelines can be found in Appendix A.
Socially responsibility investing
The increasing importance of corporate social responsibility among all stakeholder groups caused a rapid growth of socially responsibility investing (SRI). SRI integrates environmental, social and governance factors into investment decisions. It is a growing market segment, stimulated by investors who show a preference for investing in firms that pursue socially responsible activities (uksif.org). According to The Forum for Sustainable and Responsibility Investment (USSIF), more than one out of every ten dollar of total investment in the US is invested according to socially responsible principles (ussif.org), and this trend is still visible: “over the past twenty years, the total dollars invested in SRI has grown exponentially, as has the number of institutional, professional, and individual investors involved in the field.” As a result of this new investment strategy, substantial number of studies have started to examine the economic consequences of socially responsible investing, while at the same time the impact of CSR on firm value has become of great interest for shareholders, causing “a growing number of firms worldwide have undertaken serious efforts to integrate CSR into various business aspects” (Jo and Harjoto, 2011). Hence, currently organizations attempt to reduce the size of their environmental impact. In this paper I attempt to explain the consequences of the growing interest in SRI for organizations. I focus on investors’ perception of environmental firm performance, and examine the effect of environmental performance on a stock’s required rate of return demanded by investors.
The cost of equity
5
through the cost of capital, existing of the cost of equity and the cost of debt. The cost of equity is the discount rate that investors apply to a firm’s future cash flows to determine its current market value (Ghoul et al, 2011). It is important for firm’s valuation for two reasons. First, the higher the cost of equity, the lower the present value of a firm’s future cash flows. Therefore, firms with a low cost of capital are more highly valued thus are more attractive for investors than firms with a high cost of capital, and the relationship between cost of equity and firm value is negative (Sharfman and Fernando, 2008). Secondly, the cost of equity is “the required rate of return given the market’s perception of a firm’s riskiness” (Sharfman and Fernando, 2008; Ghoul et al, 2011). Consequently, the higher the required rate of return demanded by investors for the capital they provide to the firm, the more costly it is for a firm to finance itself, consequently decreasing firm value. The cost of debt is the interest rate a firm pays on its current debt, but this is beyond the scope of this paper.
The model of Heinkel, Kraus and Zechner (2001) demonstrates that corporate responsible firms have a lower cost of capital and higher firm value than polluting firms; what they explain according to risk diversification opportunities. “The stocks of polluting firms are held by fewer investors since green investors eschew polluting firms’ stock, causing a lack of risk sharing among non-green investors”. As a result less arbitrage opportunities exist, which leads to an increase of the cost of capital and a decrease of polluting firms’. Hong and Kacperczyk (2009) support the argumentation behind this model with empirical evidence regarding ‘sin’ stocks, stocks related to the tobacco, alcohol and gaming industries. They show that these have outperformed market averages by about 3,5 percent per year for decades. In addition to the argument of limited risk sharing, they argue that an increased litigation risk exists for sin stocks, causing the expected return to increase even more.
6
Prior research
To understand the influence of CSR and SRI on companies, I briefly examine prior research published on this subject. The rise of CSR practices has recently fuelled research on the relationship between CSR and financial performance, although no consistent findings have resulted (Ullman, 1985; Murphy, 2002). As pointed out by Ullman (1985) and the more recent work of Ghoul et al (2011), the main reasons for the inconsistencies are contrasting theoretical views, inappropriate definition of key terms, and deficiencies in the empirical data bases available. The theoretical foundation of previous research is mainly based on two different perspectives; first, the perspective of the firm where researchers focus on the strengths of a company’s resources, and second investors’ perspective, where the two focal points are market value and the cost of capital (see table I).
Firm’s perspective
The first group of scholars examines the influence of CSR from firm’s perspective (see table I, panel A). A theory commonly used to explain the relationship between CSR and firm performance is the resource-based theory, explaining that “a firm’s sustainable competitive advantage stems from the value and imitability of a firm’s resources or management capabilities” (Wernerfelt, 1984). The theory explains the mixed results of the relation between environmental and firm performance found by prior researches. Waddock and Graves (1997), Clemens (2006), Nakao et al (2007) and Clarkson et al (2011) find a positive correlation between environmental and firm performance, while McWilliams and Siegel (2001) and Surroca, Tribó and Waddock (2010) find no relationship between CSR and firm performance. The latter build a more extensive regression model with extra variables; McWilliams and Siegel (2001) add the rate of investment in R&D and the advertising intensity of an industry and Surroca, Tribó and Waddock (2010) use intangible resources as an extra independent variable. They explain that the positive relation found by the other researchers exists due to the quality of a firm’s resources and management’s capabilities, instead of an involvement in CSR practices. Thus “the relation between environmental and financial performance is merely indirect” (McWilliams and Siegel, 2001).
Table I. Results previous research
Panel A. Firm’s perspective
Study Subject Independent variable Database Control variables Sample Method Result
Waddock and Graves, 1997
Corporate social performance (CSP) and financial performance
CSP KLD database Size, risk, industry 469 S&P500 firms, 1989
Regression analysis Positive correlation CSP – financial performance McWilliams and Siegel, 2001 CSP and financial performance CSP, R&D intensity, advertising intensity
KLD database Size, risk, industry 524 S&P500 and DS400 firms, 1991-1996
Regression analysis No correlation
Bansal, 2005 Determinants of corporate sustainable development International experience, management capabilities, organizational slack, fines and penalties, mimicry, media attention
Annual reports Size, return on equity
45 Canadian firms in the oil and gas, mining and forestry industries, 1986-1995
Panel; time series fixed-effects regression analysis Positive correlation international experience, media pressure, mimicry, size. Negative correlation return on equity
Clemens, 2006 Green performance and financial performance Environmental performance, green economic incentives Survey based on the Likert Scale Size, respondents' confidence 76 US firms in the steel industry, 2003 Hierarchical regression analysis Positive correlation green – financial performance Nakao et al, 2007 Environmental and financial performance Environmental management score, advertising expense, R&D expense Nikkei Environmental Management Survey Reports Earnings per share, ROA, Tobin's q, financial leverage 121 Japanese firms in the manufacturing sector, 1999-2003 Multiple regression equation Positive correlation in a two-way interaction environmental – financial performance Surroca, Tribó, Waddock, 2010 Corporate responsibility (CRP) and financial performance CRP, innovation, human capital, reputation, culture. Sustainalytics Platform
Size, risk, industry, country, year 599 industrial firms included in the Sustainalytics Platform, 2002-2004
Panel; times series fixed-effects regression analysis. No correlation; mediating effect through intangible resources Clarkson et al, 2011 Determinants and consequences of proactive environmental strategies
Panel B. Investors’ perspective Firm value
Study Subject Independent variable Database Control variables Sample Method Result
Derwall et al, 2005
Socially responsible investing (SRI) and stock returns
Eco-efficiency ratings Innovest Strategic Value Advisors’ scores
Size, values versus growth, momentum effects, industry 180-450 US companies, 1995-2003
CAPM, Fama and French three-factor model
Positive correlation SRI – stock returns
Hassel, Nilsson, Nyquist, 2005 Environmental information and market value Quarterly environmental performance CaringCompany (CC) Research, quarterly financial statements Book value, abnormal earnings, industry, accounting legislation 71 Stockholmsbörsen firms, 1998-2000 Ohlson's valuation model Negative correlation environmental performance – market value Kemp and Osthoff, 2007 Effect of SRI on portfolio performance Factor portfolio returns
KLD Market risk, size, book-to-market, momentum factor 650 S&P500 and DS400 firms, 1991-2003 Carhart four-factor model Positive correlation SRI – portfolio performance Godfrey, Merrill and Hansen, 2009 CSR and shareholder value Negative legal or regulatory actions Socrates KLD ratings Size, market-to-book 160 US firms, 1993-2003 Cumulative abnormal rating (CAR) CSR creates insurance-like protection in the face of negative events Schadewitz, Niskala, 2010 Environmental reporting and market value Sustainability report publication dummy Annual and sustainability reports Book value, abnormal earnings, 276 OMX-Helsinki firms, 2002-2005 Ohlson's valuation model Environmental reporting explains market value Jo and Harjoto, 2011 Determinants of corporate sustainable development Equity ownership by institutional holders, number of analysts, board characteristics
KLD Size, leverage 2952 US firms, 1993-2004
Univariate tests and bivariate correlations Positive correlation independent variables – sustainable development Ziegler, Busch and Hoffman, 2011 Effect of disclosed corporate responses to stock performance
Portfolio returns Swiss company ASSET4
Market risk, size, book-to-market, momentum factor. 1014 US and European firms, 2001-2006
CAPM, Carhart four-factor model
Positive correlation disclosure – market value for energy firms in the USA Berthelot,
Coulmont, Serret, 2012
9 Cost of capital
Study Subject Independent variable Database Control variables Sample Method Result
Hamilton, Jo and Statman, 1993
Socially responsible mutual funds and stock returns Excess returns of socially responsible mutual funds Lipper Analytical Services NA 320 US mutual funds, 1981-1990
Jensen’s alpha, No correlation socially responsible mutual funds and both expected stock returns and cost of capital
Feldman et al, 1997
Environmental management systems (EMS) and capital costs
Beta TRI NA 330 S&P500
firms, 1980-1994
NA Negative correlation
adoption EMS and cost of capital. Positive correlation cost of capital and firm value
Sharfman and Fernando, 2008
Environmental risk management (ERM) and the cost of capital
Cost of capital, cost of debt, cost of equity capital
KLD, TRI Size, leverage, industry
267 S&P500 firms, 2002
Capital Asset Pricing Model (CAPM) Negative correlation ERM - cost of capital Hong and Kacperczyk, 2009
The effects of social norms on stock returns
Stock returns of tobacco, alcohol and gaming industry CRSP Institutional ownership, size, risk, market-to-book ratio, 193 NYSE, Amex and Nasdaq stocks, 1962-2006
CAPM, Fama French four-factor model, panel; cross-section and time-series return regression Negative correlation CSR and stock returns Ghoul et al, 2011 Corporate social responsibility and the cost of capital
have superior financial resources and capabilities, therefore better opportunities to accomplish a sustainable competitive advantage that is hard to mimic by other firms in the industry. This is consistent with the findings of McWilliams and Siegel (2001) and Surroca, Tribó and Waddock (2010), where innovation is positive and significantly related to both environmental and financial performance. In conclusion, there is no direct positive relation between environmental and financial performance, only an indirect relation because responsible and sustainable investments can only offer a competitive advantage if adequate resources are present.
Investors’ perspective
The second group of scholars investigates the impact of CSR from investors’ perspective (see table I, panel B). Different scholars examine the effects of environmental performance on market value, cost of capital, or both. Current debate on how environmental performance impacts firm value is basically divided into two schools (The Assabet Group, 2000). The cost-concerned school argues that “environmental issues represent primarily increased costs and offer little positive potential for shareholders”. As a result firms that engage in CSR practices experience an increased cost of capital, which in turn results in decreased market value. The value creation school regards environmental performance as a competitive advantage that improves financial results, therefore a positive relation between CSR and market value is expected. Previous research finds support for both views; the majority confirms the value creation school (Feldman, Soyka, and Ameer, 1997; Derwall et al, 2005; Kemp and Osthoff, 2007; Sharfman and Fernando, 2008; Ghoul et al, 2010; Ziegler, Busch and Hoffman, 2011; Berthelot, Coulmont and Serret, 2012), Hassel, Nilsson, and Nyquist (2005) support the cost-concerned school, while Hamilton, Jo and Statman (1993) find no correlation at all.
11
positive correlation consistent with the value creation view. The findings of Hassel, Nilsson and Nyquist (2005) have to be interpreted with caution, since the sample is small and the time period relatively short. Nevertheless, the authors argue that their result is valid and reliable, what they explain by three arguments; “first, investors perceive that environmental performance is used for the window dressing of book values and financial performance; second, investors perceive that environmental responsibility activities are made at the expense of increased profits… and third, the market is short-term oriented, and investors do not consider longer-term environmental information when making investment decisions.” Sharfman and Fernando (2008) and Ghoul et al (2010) examine the perceived riskiness linked to CSR, building upon risk management theory. This theory states that superior environmental risk management improves the market’s risk perception of the firm. Polluting firms face a higher risk than other firms for two reasons. First, there is a higher litigation risk (Waddock and Graves, 1997; Hong and Kacperczyk, 2009). Second, since socially responsible investors abstain from polluting stocks, demand for shares is lower and the opportunities for risk diversification are reduced (Heinkel, Kraus and Zechner, 2001). This is supported by the evidence of Feldman, Soyka, and Ameer (1997) that demonstrates that socially responsible firms indeed have a lower beta than polluting firms.
Although Sharfman and Fernando (2008) and Ghoul et al (2010) both use beta as a measure of systematic risk, the methodologies to calculate the cost of equity differ. Sharfman and Fernando (2008) apply the Capital Asset Pricing Model (CAPM) developed by Sharpe (1964), where market behavior is predicted with risk as a factor in the analysis. The dependent variable is the cost of equity, and the model suggests a positive relation between a stock’s market beta and its cost of equity. Ghoul et al (2010) calculate the cost of equity premium based on the average cost of equity obtained from four different models developed by other authors and which are widely used in recent accounting and finance research. Moreover, they argue that their regression model better estimates the effect of CSR than Sharfman and Fernando (2008), since they “control for various firm-level corporate governance characteristics that have been shown to affect the cost of equity capital.”
Contradicting results
12
A first explanation is the difference in underlying assumptions that are used to build the regression model. This appears to be true for the mixed results of Waddock and Graves (1997) on the one hand, and McWilliams and Siegel (2002) and Surroca, Tribó, Waddock (2010) on the other hand. Waddock and Graves (1997) build upon the argument of slack resource theorists that better financial performance potentially results in the availability of slack resources, providing an opportunity for companies to invest in social performance activities. However, they do not include slack resources but only include financial performance in their regression. This is remarkable, since they acknowledge that financial performance is indirectly linked to social performance through slack resources: “if slack resources are available, then better social performance would result from the allocation of these resources into the social domains, and thus better financial performance would be predictor linkage between its values and strategy”. The same is true for the variable management capabilities; they agree with good management theorists that there is a high correlation between good management practice and corporate social performance, but they lack to include a measure of it in their regression. McWilliams and Siegel (2002) and Surroca, Tribó, Waddock (2010) add a measure for R&D resources, and demonstrate that once R&D intensity is included the correlation between corporate social responsibility and financial performance is no longer significant.
A second explanation is the choice of dataset. A sample with a small number of observations, a relatively short time period, or a sample that is limited to certain types of industries or countries may be biased and not appropriate to come to a valid conclusion. This is probably the case for the contradicting findings of Hassel, Nilsson and Nyquist (2005) and Berthelot, Schadewitz and Niskala (2010). Both authors use environmental information published in companies’ annual reports and both apply the Ohlson mode but Hassel, Nilsson and Nyquist (2005) find a negative correlation while Berthelot, Schadewitz and Niskala (2010) find a positive correlation. A possible cause is the size of the sample; the sample of Hassel, Nilsson and Nyquist (2005) contains 71 firms, and the sample of Berthelot, Schadewitz and Niskala (2010) consists of 146 firms and the target group exists of firms that publish a sustainability report, which exists of only 28 firms.
13
Fama and French investigate cross-sectional variation in stock returns associated with market beta. They use month-to-month regressions of stock returns of more than thousand stocks for the period July 1963 to December 1990, and find “a strong negative relation between size and average return”.
Secondly, Fama en French use the book and market value of financial leverage, defined as the ratio of book assets to book equity and book assets to market equity, respectively. They find a positive relation between stock returns and market leverage and a negative one with book leverage that they explain by book-to-market equity.
The results of other authors relating to firm size support Fama and French’s; Gebhardt, Lee and Swaminathan (2001) and Surroca, Tribó and Waddock (2010) find size to be negatively related to returns. Moreover, Sharfman and Fernando (2008) and Ghoul et al (2010) relate size to a firm’s risk premium and both find a negative correlation.
Secondly, also with regard to leverage results are consistent among authors; Gebhardt, Lee, and Swaminathan (2001) who use both book and market value of debt, determine a positive relation for both measures with the firm’s risk premium similar to Sharfman and Fernando (2008), although not significant, and Ghoul et al (2010).
Another control variable that most authors use is industry membership. Some authors already control for this in the construction of their sample; Clemens (2006) and Nakao et al (2007) examine firms in only one industry, and Surroca, Tribó and Waddock (2010), Clarkson et al (2011) focus on the most polluting industries. Waddock and Graves (1997) and McWilliams and Siegel (2001) provide evidence that CSR levels vary according to industry characteristics, thus industry effects have to be taken into account for valid results.
14
reduce the present value of a firm’s cash flows will be positively or negatively related to that firm’s market value”. This is the case when the demand for socially responsible investment opportunities is equal to the supply of these opportunities, because then an excess supply is created. Based on this argumentation, I state that the time period under consideration is an important determinant for the results of the research. USSIF provides evidence that SRI has experienced a rapid growth in recent years: “total US dollars under professional management in SRI in from $639 billion to $3.07 trillion, outpacing the overall market between 1995 and 2010” (ussif.org). This shows that the demand for socially responsible investment opportunities is growing, and on the basis of the model of Mackey, Mackey and Barney (2007) this suggests that studies performed in more recent periods will lead to higher market value than studies examining an earlier time period.
Hypothesis
In this paper, the main question to be answered is: Have firms with superior environmental performance a lower cost of equity than firms with inferior environmental performance? I hypothesize that better environmental will lead to a lower cost of equity, since it increases organization effectiveness (Waddock and Craves, 1997; Nakao et al, 2007; Clarkson et al, 2011), receives higher valuations (Derwall et al, 2005; Jo and Harjoto, 2011; Berthelot, Coulmont and Serret, 2012), and reduces systematic risk (Feldman, Soyka, and Ameer, 1997; Heinkel, Kraus and Zechner, 2001; Sharfman and Fernando, 2008, Hong and Kacperczyk, 2009; Ghoul et al, 2011).
Table I summarizes three researches that use the cost of equity as dependent variable, including Gebhardt, Lee, and Swaminathan (2001), Sharfman and Fernando (2008), and Ghoul et al (2010), and all find a negative relationship between CSR and the cost of equity. Furthermore, they control for firm size, financial leverage and industry effects. All present significant and consistent findings relating to the effect of size and leverage; larger firms have a lower cost of equity, while higher debt results in a higher cost of equity. I argue that size, financial leverage and industry effects are explanatory variables and must be controlled for in the regression model to obtain valid results.
15
in groups according to pollution propensity (Sharfman and Fernando, 2008; Clarkson et al, 2011), this is the first paper that includes the company data in the regression model. Secondly, this is the first study that examines the effect on the entire European stock market. I argue that this is an interesting market since the number of GRI certified sustainability reports is much higher; Europe accounts for 45 percent of total reports, while the US for merely 14 percent (globalreporting.org). Secondly, the European Union already agreed to reduce greenhouse gas emissions (GHG) established under the Kyoto Protocol in 1997, and since 2009 it adopted the target for a 20 percent reduction in GHG in 2020 compared to 1990 levels, while the US has not set any legally binding targets to reduce its GHG level (unfccc.int). In total 930 firms of 21 countries during a four-year are examined, resulting in a total of 4231 numerical observations.
3. DATA AND METHODOLOGY
In this section, I explain the data and methodology used for testing the hypothesis. First I explain the models that I will use to construct my regression. Next I discuss the dataset. Thirdly I elaborate on the methodology of the research. Finally the sensitivity test is explained.
The models
I follow Sharfman and Fernando (2008) and Ghoul et al (2011) who define the cost of equity as “the required rate of return given the market’s perception of a firm’s riskiness”. Thus, the required rate of return is measured using the cost of equity. To compute a firms’ cost of equity, I use the Capital Asset Pricing Model (CAPM) developed by Sharpe (1964), similar to Gebhardt, Lee and Swaminathan (2001) and Sharfman and Fernando (2008). The CAPM predicts market behavior with risk as a factor in the analysis; it suggests that a stock’s market beta is positively related with its cost of equity.
16
Ghoul et al (2011) propose two more arguments for using ex ante values: first, “unlike traditional measures of firm value, it allows one to control for differences in growth rates and expected future cash flows”, and second, “it circumvents the use of noisy realized returns and the failure of traditional asset pricing models to deliver accurate estimates of firm-level cost of equity”.
The CAPM model can be specified as follows:
(1)
where
= the expected investor return from holding security , or the cost of equity; = the equity risk premium, where
= a measure of firm’s systematic risk; = the expected return on the market portfolio, and the risk-free rate.
Next, I link environmental performance to the market value of a firm based on the accounting-based valuation model developed by Ohlson (1995), consistent with the methodology of Hassel, Nilsson and Nyquist (2005), Schadewitz and Niskala (2010) and Berthelot, Coulmont and Serret (2012). The assumption of the Ohlson model is that market capitalization depends on book value of equity plus abnormal earnings. Abnormal earnings are a proxy for a company’s goodwill. Next to this, Ohlson adds an additional variable to explain firm’s market value, i.e. other value-relevant information. He explains that this variable is important to add because “value-relevant events may affect future expected earnings as opposed to current earnings…. and incorporate some value-relevant events only after a time delay. The behavior does not depend on current or future dividends” (Ohlson, 1995). The model is specified as follows:
17 Where
= the market value, or price, of equity date t; = the (net) book value of common equity at date t;
= , where abnormal earnings equal current earnings minus the cost of equity capital times the beginning of period book value, and
= other non-accounting value-relevant information.
Book value and abnormal earnings are accounting data that is publicly available, book value through databases and abnormal earnings through markets’ forecast, since “analysts’ forecasts are meant to capture and reflect all value-relevant information in addition to accounting data” (Schadewitz and Niskala, 2010). I use company’s environmental performance as a proxy for other non-accounting value-relevant information, consequently the value of the coefficient captures the incremental value-relevance of environmental performance. In my regression model, I replace market value with the cost of equity derived from CAPM, since this allows me to study the incremental value-relevance of environmental performance on the cost of equity.
Data
The dataset consists of panel data containing 930 unique firms listed on European stock markets during the period 2006 to 2010. Since studies based on US data are extensive, and no other study has investigated the European market as a whole, I choose to investigate the Europe stock market. This allows me to determine if previous results that are dominated by US and Canadian firms are generalizable to other geographic areas. Since emission targets for climate control are legally bounded in Europe in contrast to the US where no formal commitments are set (unfccc.int), it is interesting to see if this has any consequences for the effect of CSR performance on the cost of equity.
18
Table II. Industries
Number of firms Percentage
Industrial 630 68% Utility 82 9% Transportation 26 3% Bank/Savings&Loan 81 9% Insurance 36 4% Other financial 75 8% Total 930 100%
Number of firms is the total number of unique firms per industry. Percentage is the relative amount of firms included per industry.
The dependent variable
I am interested to find out what the effect of environmental performance is on the cost of equity; do investors value superior environmental performance according to the value creation school, or is the cost-concerned school correct that high environmental performance is evaluated as a cost. If investors attach additional value to environmental firm performance, demand for shares will increase and consequently the cost of equity will decrease. On the other hand, a negative correlation between a company’s carbon footprint and the cost of equity implies that investors require more return for the stocks of environmental responsible firms than for polluting firms, supporting the cost-concerned school.
Data on the expected investor return is derived from the public database Datastream and is based on monthly stock returns of firm’s individual stocks, for the period January 1983 to December 2011. Since environmental performance data becomes available for the market at the next fiscal year, the influence on cost of equity is noticeable with a time lag of at least one year, therefore the dependent variable is based on data of 2007-2011. I further explain the use of time lags in the paragraph model specification in the methodology section.
19
Table III. Market Return and the risk-free rate
2007 2008 2009 2010 2011 1-5-2012
Average Market Return 12,72% 9,97% 11,17% 11,06% 10,74% 10,95%
Risk-free rate 4,32% 4,95% 1,73% 1,24% 2,14% 1,30%
The average market return is based on monthly stock returns since 1983 derived from the MSCI World index and calculated with the formula (
. The risk-free rate is based on the 12-month Euribor and is based on the average daily values per year.
The independent variable
The independent variable is company’s environmental performance, and is measured according to four GRI G3 indicators, specifically greenhouse gas emissions (GHG), energy consumption, water withdrawal and waste discharge. Since the GRI guidelines for reporting are not mandatory, organizations are to a certain extent free to choose which indicators to include in their environmental report. I choose these four indicators because they are the most frequently reported in annual reports.
20
Table IV summarizes the total number of observations available per environmental performance measure. First of all, I checked all outliers for validity and I deleted invalid numbers. For example, I removed the 2008 value for RHI and the 2010 value for Ryanair, which are almost ten times higher than the other years. Table IV shows that only a small amount of data is available regarding energy consumption, water withdrawal and waste production. Therefore, I run the regressions with only one independent variable at the time, to avoid a decrease in the large number of observations because not all four variables are available. In addition, the number of observations in 2006 is much lower than in other years, thus I have to interpret the results for changes in 2007 compared to 2006 with caution.
Table IV. Environmental performance data – total number of number of observations
2006 2007 2008 2009 2010 Total
Greenhouse gas emissions 177 773 837 855 716 3358
Energy consumption 28 44 68 86 85 311
Water withdrawal 24 36 59 76 77 272
Waste production 28 42 62 77 81 290
Total 257 895 1026 1094 959 4231
Greenhouse gas emissions exist of total direct and indirect greenhouse gas emissions by weight and are measured in metric tons. Energy consumption is measured in megawatt hour (MwH). Water withdrawal is expressed per cubic meters (m³) and waste involves the total weight of waste in metric tons.
Important to take into notice is the different role the four variables play in the production process of organizations. Energy and water are input variables; firms use them for the production process. Greenhouse gas emissions and waste production are output variables; they are released as consequence of firms’ production processes. Entities classified as inputs or outputs are differently related to an organization; energy and water could be substituted at the input of the production process, for example with solar and wind energy, while greenhouse gas emissions and waste cannot be substituted and should be restricted as far as possible. The final dataset forms an unbalanced panel, since data is not available for company and for every year.
Control variables
21
cross-sectional variation in average stock returns (Fama and French, 1992). Fama and French find in their size portfolios that size and beta are almost perfectly negative correlated (0,98), and according to CAPM the cost of equity is an increasing function of systematic risk, hence I expect cost of equity to be negatively related to firm size. Multiple authors elaborate on these findings and confirm this negative relation (Gebhardt, Lee, and Swaminathan, 2001; Sharfman and Fernando, 2008; Ghoul et al, 2010). To eliminate non-normality, size is measured as the natural logarithm of market capitalization.
I control for financial leverage since I expect a positive relation between a firm’s cost of equity and the amount of debt in its capital structure is expected, consistent with prior studies (Modigliani and Miller, 1958; Gebhardt, Lee, and Swaminathan, 2001; Ghoul et al, 2010). Following Fama and French (1992) and Gebhardt, Lee and Swaminathan (2001), I use a company’s market and book leverage, defined as the total long term debt to market equity, and the ratio of long term debt to total book value of equity, respectively. The natural logarithms are used, because “preliminary tests indicate that logs are a good functional for capturing leverage effects in average returns” (Fama and French, 1992). Lastly I control for industry effects. Several studies choose to use only the most polluting industries (Bansal, 2005; Clarkson et al, 2011) or only one specific industry (Clemens, 2006), but I choose to follow Sharfman and Fernando (2008) and include all industries, while controlling for industry effects. The advantage is that no firms have to be eliminated from the sample thus I do not decrease the sample size.
Firms are categorized in six categories based on the general industry classification (see table I). According to Sharfman and Fernando (2008), controlling for industry effects is important to rule out a spurious correlation between the cost of equity and environmental performance that would stem from differences between industries. They run an analysis of variance (ANOVA) test on the different industry groups and find a significant effect, indicating that industry effects are present. With Dunnett’s T³ test they determine differences among groups, and eventually they create a dummy variable with the homogeneous groups in one category and the nonhomogeneous groups in the other.
22
leverage is higher than the mean for market value of leverage. The explanation is that market equity is higher than book equity, and investors pay more for shares than actual book value.
Since I use panel data, Table VI presents the amount of GHG per year so that changes over years are clearly visible. Because of limited data availability for energy consumption, water withdrawal and waste production, I choose to present only the yearly values for GHG. I included both the relative change for 2006 to 2010, and for 2007 to 2011, because of the small number of observations in 2006, which may bias the results. Overall, the absolute and scaled amount of GHG has decreased over years. This decrease is consistent with the general trend that socially responsibility investing is a growing market segment, and that firms started to integrate SR into various business aspects.
Descriptive statistics
Table VII presents the correlations for carbon footprint based on equity and revenues. Panel A presents the correlations for GHG including the control variables. Panel B presents the correlations between the four independent variables. I only included GHG in panel A, because if energy consumption, water withdrawal and waste production are included, the number of observations reduces to 126 instead of 2601 observations.
First of all, Panel A shows that the correlations between carbon footprint values based on equity and revenues are, as expected, extremely high and significant. If all would be included in one regression, then multicollinearity exists (Brooks, 2008). This is also true for book and market value of debt. Therefore I run the regressions including carbon footprint based on either equity or revenues, and either market or book value of debt.
Secondly, the correlations between cost of equity and GHG scaled on both equity and revenues are negative and significant at a 1% level. This is against my expectations; it implies that higher carbon footprint results in a lower cost of equity, vice versa green firms have a higher cost of equity. This result is inconsistent with the findings of Feldman, Soyka, and Ameer (1997), Sharfman and Fernando (2008) and Ghoul et al (2010).
Next, the relation between cost of equity and market capitalization is highly significant and positive, which means that a larger firm has a higher cost of equity. This is a surprising result, and inconsistent with earlier literature that finds a negative correlation (Gebhardt, Lee and Swamithan, 2001; Sharfman and Fernando, 2008; Ghoul et al, 2010).
23
Table V. Descriptive statistics
Variable Nr of firms Mean SD Median Minimum Maximum Skewness Kurtosis JB
Cost of equity (%) 929 10,47% 0,036 10% 2 % 39,5% 1,199 6,661 3542,74
GHG (x000 metric tons) 923 3790,34 15282294 112,98 0,13 2,39E+05 7,377 77,42 806766,2
Energy (x000 MwH) 92 56957,6 2,59E+08 881,13 6,252 2,33E+06 6,094 44,544 24133,65
Water (x000 cubic meters) 82 81785,03 3,75E+08 1494,37 8,31 4,02E+06 8,443 79,91 70791,67
Waste (x000 metric tons) 84 6542,15 45920312 78,617 0,001 5,83E+05 10,188 115,635 155038,4
Log market capitalization 930 6,592 0,821 6,504 2,973 9,909 0,484 3,653 250,68
Log book value leverage 908 1,831 0,676 1,86 -1,699 4 -0,97 6,577 2906,18
Log market value leverage 912 1,515 0,465 1,617 -1,699 2,777 -2,726 14,057 27208,18
Table IV presents descriptive statistics for the variables. The mean and standard deviation are based on firm’s average values for 2006 to 2010. Cost of equity is expressed as percentage, GHG is greenhouse gas emissions expressed in metric tons, energy is consumption expressed in MwH, water is withdrawal expressed in cubic meters, and waste is expressed in metric tons.
Table V. Descriptive statistics – average per year
Variable 2006 (N= 177) 2007 (N=773) 2008 (N=837) 2009 (N=855) 2010 (N=716) 2006-2010 2007-2010
CO2 (x000 metric tons) 4224 4414,3 2943,6 4124,7 3224,9
+5% -33% +40% -22% -24% -27%
CO2 as % of revenues 27 31 43 61 14
+13% +38% +44% -76% -47% -60%
CO2 as % of equity 59 56 77 90 24
-7% 38% 17% -74% -60% -57%
predicted direction and confirms earlier literature (Fama and French, 1992; Gebhardt, Lee and Swaminathan, 2001; Sharfman and Fernando, 2008; Ghoul et al, 2010).
The correlations with regard to GHG scaled to equity and revenues have identical signs, though different significance levels. First, the relationship between GHG and market capitalization is negative, hence the larger the firm, the lower the relative carbon footprint. It supports the resource based view: “larger firms are more likely to commit to sustainable development because they attract more shareholder attention and have generally more resources” (Bansal, 2005). Also the results of Clemens (2006), Sharfman and Fernando (2008), Ghoul et al (2010) and Surroca, Tribó and Waddock (2010) indicate a negative correlation. On the other hand, Waddock and Graves (1997) and Clarkson et al (2011) find a negative relation between size and CSR, although the result of Waddock and Graves is not significant. The opposing results may be explained by the choice of measurement; the first group uses market capitalization to measure firm size, and the latter uses book value, specifically Waddock and Graves (1997) use total sales, total assets and the number of employees, and Clarkson et al (2011) use total assets.
Furthermore, I find a negative relation between financial leverage and GHG, though only GHG scaled to equity is significant at a 10% level. Thus the higher the amount of debt, the better a company’s CSR performance. However, results of previous research are mixed. Sharfman and Fernando (2008) find similar results, while Clarkson et al (2011) find a negative relation, although not significant. Additionally, Ghoul et al (2010) find no relation at all, and Surroca, Tribó and Waddock (2010) find both positive and negative relations dependent on the inclusion of variables in the regression, but none significant. Sharfman and Fernando (2008) explain their result based on risk management theory: “as a firm increases its level of risk management, it can correspondingly shift its financing from equity to debt capital (i.e., increase its leverage) because the firm is perceived as being less risky”. Appendix D presents the correlations for energy consumption, water withdrawal and waste production. The correlation between cost of equity and these three independent variables is similar to the negative correlation between cost of equity and GHG, although less significant. Furthermore, all correlations between cost of equity and the control variables are positive, consistent with the GHG correlations. Only the correlation between cost of equity and market capitalization is opposite, although not significant.
25
market and book value of debt correlate positive with waste scaled to equity, although not significant.
Altogether, with regard to the correlations of water withdrawal and waste production I find some surprising results that are the opposite effect of the correlations of GHG and energy consumption. I propose two reasons for this result; first, the sample size. Table IV shows that the number of yearly observations is 24 or less (see table III), which is probably too small for valid results. Secondly, data is collected via annual reports, and although GRI provides guidelines for reporting, companies are free to choose what information to report and how to report it; for example, the metric tons of waste may be total weight, or merely hazardous or non-hazardous waste. Since no standard format exists, all companies publish different numbers and since it is not straightforward what criteria they apply, a comparison among companies may be invalid.
Panel B presents the correlations between the independent variables. All correlations are positive, thus every variable that proxies for environmental performance correlates with the other variables.
GHG and energy consumption scaled to both equity and revenues are significantly and positively correlated; suggesting that if higher emissions are present, energy consumption is higher as well. In addition, GHG scaled to revenues is positively and significantly related to waste and water scaled to both equity and revenues. The same is true for energy scaled to both variables and all other environmental performance measures; all correlations are positive and significant.
26
Table VI. Correlations
Panel A. Correlations (n=2601)
Cost of equity CO2 EQ CO2 REV Market Cap BV debt MV debt
Cost of equity 1 CO2 EQ -0,1*** 1 CO2 REV -0,09*** 0,94*** 1 Market Cap 0,1053*** -0,508*** -0,478*** 1 BV debt 0,088*** -0,035* -0,058*** 0,003 1 MV debt 0,077*** -0,027 -0,052*** 0,019 0,943*** 1
***, ** and * denote significant levels at 1, 5 and 10%, respectively
Panel B. Correlations independent variables (n=125)
CO2 EQ CO2 REV Energy EQ Energy REV Waste EQ Waste REV Water EQ Water REV
CO2 EQ 1 CO2 REV 0,9619*** 1 Energy EQ 0,1769** 0,1266 1 Energy REV 0,0879 0,1536* 0,9114*** 1 Waste EQ 0,1286 0,167* 0,2665*** 0,3182*** 1 Waste REV 0,1098 0,1557* 0,2594*** 0,3228*** 0,9837*** 1 Water EQ 0,1327 0,1696** 0,3115*** 0,3603*** 0,1276 0,0868 1 Water REV 0,0708 0,15527* 0,3508*** 0,4725*** 0,0497 0,0479 0,7444*** 1
27
Methodology
The aim of this study is to investigate if there is a relationship between the required rate of return demanded by investors for their shares in a company, and the environmental performance of that company. This is investigated with panel data regressions, and because the number of years available per companies differs, I use an unbalanced panel, i.e. a panel with some cross-sectional elements with fewer observations and observations at different times to others (Brooks, 2008).
Several examiners structured their sample as panel data (Bansal, 1995; Surroca, Tribó and Waddock, 2010; Ghoul et al, 2011). Panel data has several advantages (Brooks, 2008). First, it is possible to address a broader range of issues and tackle more complex problems. Second, by combining cross-sectional and time series data, one can increase the number of degrees of freedom, and thus the power of the test, by employing information on the dynamic behavior of a large number of entities at the same time.
Brooks makes clear that it is necessary to test whether a model with random or fixed effects is in order, or if a simple pooled regression is valid. Fixed effects models allow the intercept to differ only on cross-sectional basis and not over time (time-fixed effects), or only on time basis and not cross-sectional (entity-fixed effects). Random effect models propose different intercept terms for each entity, assumed to arise from a common intercept plus a random variable. The Likelihood Ratio test determines if a fixed effects specification is appropriate, the Hausman test accounts for random effects models. The Likelihood Ratio test is significant both for time and cross-sectional effects; therefore a fixed effects model is inappropriate. Similarly, both random effects tests are significant and rejected. Hence, the pooled least square regression is sufficient; this means estimating a single equation on all the data together, and estimating the equation using ordinary least squares estimation (OLS).
Model specification
To determine the relationship between cost of equity and environmental performance, the following regression is employed:
(2)
28
investor return from holding security at time Secondly, Ohlson’s model is based on the assumption that a firm’s equity price is based on the current book value of equity , abnormal earnings and other non-accounting value-relevant information . In this regression model, environmental performance serves as a proxy for non-accounting value-relevant information ( ), similar to Hassel, Nilsson and Nyquist (2005), Schadewitz and Niskala (2010) and Berthelot, Coulmont and Serret (2012), who determine that the market value is related to environmental performance.
Similar to Sharfman and Fernando (2008), Godfrey, Merrill and Hansen (2009) and Clarkson et al (2008), I include environmental performance with a time lag, since data becomes publicly available at the next fiscal year. Sharfman and Fernando (2008) run regressions with one, two and three year lags, and determine that only a one-year lag of environmental risk management data is necessary for meaningful results, thus I include environmental performance data for year
29
Sensitivity analysis
I estimate the sensitivity of environmental performance to cost of equity in two ways; first, I use a different measure of the cost of equity. Second, I run the regressions with a sample excluding British companies.
Sharfman and Fernando (2008) use different measures of the independent variable beta, leading to different dependent variables, i.e. the cost of equity. Their first beta is a yearly average of weekly values derived from Bloomberg, and the second is based on yearly values derived from COMPUSTAT. I create two different cost of equity based on the expected market return in the CAPM. The results presented in table III are based upon the expected market return calculated with average values of monthly returns till May 2012, since the market return is the expected return on the market portfolio (Sharpe, 1964). However, the average market return decreased significantly in 2008 compared to 2007 from 12,72 to 9,97 percent, and never reached a value above 12 percent again in following years (see table II). Therefore, I believe it is interesting to run regressions using the realized market return based on average values of monthly returns till the year under consideration (i.e. 2007, 2008, 2009, 2010 and 2011 resp.), where investors’ expectations of market returns have changed. I think it is valuable to determine if the effect of carbon dioxide emissions on the cost of equity changes if investor expectations about market return change, because this result demonstrates if investors’ perception of responsibility performance changes if market return expectations change.
30
4. RESULTS
This section presents the results for the relationship between environmental performance and the cost of equity. First I present the estimations results of equation 3 based on the panel least squares method. Next I perform the sensitivity tests.
Panel least square analysis
The panel least square analysis determines the effect of environmental performance on the cost of equity both across firms and over time, thus I find results for the differences between companies, and for the different environmental performance per year. Table VII presents the results of the equation based on GHG as percentage of common equity, and with the inclusion of market value and book value of debt in panel A and B, respectively. The main result is that the coefficients of GHG are negative and significant at a 1% level, independent of the measurement of financial leverage according to book or market value. This result is against my expectations; it implies that the higher the pollution, the lower the cost of equity. Thus the results reject the hypothesis. Furthermore, it is in contrast with the previous work of Feldman, Soyka, and Ameer (1997), Sharfman and Fernando (2008) and Ghoul et al (2011), who find that investments in CSR practices that result in lower pollution or better environmental performance lead to a lower cost of equity.
Secondly, the sign for market capitalization is positive and significant, implying that a larger firm has a higher cost of equity. This is not what I expected and differs from earlier studies; Fama and French (1992) find a significant and negative relation between size and systematic risk ( ), and since according to CAPM the cost of equity is an increasing function of systematic risk, confirmed by the work of Unfortunately, I do not find an empirical explanation for this remarkable result.
The relationship between leverage and cost of equity positive and significant at a 1% level for both market and book leverage and confirms my expectations. It suggests that a firm with higher amounts of debt is more risky, hypothesized and confirmed by among others Gebhardt, Lee and Swaminathan (2001), Sharfman and Fernando (2008), and Ghoul et al (2011) although Sharfman and Fernando (2008) do not find a significant effect.
31
Table VIII presents the results based on the regression where GHG is measured as a percentage of total revenues. The results are similar to the results in table VII. I find a significant and negative relation between GHG and cost of equity, regardless of the inclusion of market or book value of leverage. Thus for the second time the results suggest that a higher carbon footprint leads the cost of equity to decrease, rejecting the hypothesis and contrasting with Gebhardt, Lee and Swaminathan (2001), Sharfman and Fernando (2008), and Ghoul et al (2011).
Next, I find a positive and significant coefficient for market capitalization, similar to the regression results where GHG is scaled to equity. As stated before, this result is not what I expected and is different from other authors.
Thirdly, leverage is positively and significant related to cost of equity as expected, confirming the results found in table VII. At last, the industry effect is in significant, thus industry effects are controlled.
In all, the result suggests that superior environmental performance is linked to a higher cost of equity, thus I reject the hypothesis that firms with superior environmental performance experience a lower cost of equity than firms with inferior environmental performance. In the literature review section I stated that the cost of equity is the discount rate applied by investors to determine the current market value of future cash flows (Sharfman and Fernando, 2008; Ghoul et al, 2010), thus a higher cost of equity results in a lower present value of future cash flows, which in turn results in a lower market value. From this view, firms with superior environmental performance have a lower market value than polluting firms. This is congruent with the cost-concerned school that investing in CSR is expensive and reduces market value, supported by the work of Hassel, Nilsson and Nyquist (2005).
Next I discuss the results based on the other three independent variables, including energy consumption, water withdrawal and waste production. The results are presented in appendix E. First I discuss the results where the independent variables are scaled to equity, next the results based on revenues.
32
results (see table III). Moreover, the effect of market capitalization on the cost of equity is still positive but no longer significant, probably also caused by the small number of observations. This suggests that a large firm has a higher cost of equity than a small firm. However, the coefficient of market capitalization is negative once water scaled to equity is the independent variable. This is an interesting result, since it differs from the results of the other three measures of environmental performance. I argue that there is a second explanation in addition to the sample size, i.e., because companies can choose if they report on environmental performance and how they measure the different aspects, there is a high possibility that companies use the measure most favorable for their results, resulting in biased data. Thirdly, the relationship between leverage and cost of equity is positive and significant for book and market value of debt, which is similar to the GHG regressions. Finally, no industry effects exist.
The regression results regarding energy consumption, water withdrawal and waste production scaled to revenues are not significantly different from the results where the independent variables are scaled to equity. Also here the regression results for water withdrawal and waste production are not included, since the F-test is rejected. The key result is that the coefficient of environmental performance is negative, although not significant. This implies a lower cost of equity for a firm with a higher carbon footprint, thus the hypothesis is again rejected.
The sign for market capitalization is positive, but not significant. This is similar to the other regression results, and against my expectations. Financial leverage has the expected sign in all regressions; positive and significant, thus a highly levered firm is perceived as more risky, which in turn causes a higher cost of equity. Any industry effects are eliminated, thus a false correlation between cost of equity and carbon footprint is ruled out.
Lastly, I conclude that there are no differences in the effect of the input measures energy and water, and the output measures GHG and waste. All independent variable coefficients are negatively related to the cost of equity.
cost-33
concerned school that argues that a firm’s involvement in CSR practices results in little positive potential for investors, merely in increased costs, because investments in socially responsible activities are expensive that are not by definition profitable. The argument that CSR is costly can have negative consequences for the cost of equity; if investors are afraid that companies attach more value to sustainable projects instead of profit-generating projects, they may perceive investing in this firm as risky, and require I higher rate of return. In addition, I argue there are three possible explanations for my contradicting findings compared to previous research. The first is my definition and choice of measurement of the independent variable. Most authors measure a firm’s CSR score according to KLD Research Analytics (KLD) ratings (Waddock and Graves, 1997; McWilliams and Siegel, 2000; Sharfman and Fernando, 2008; Surroca, Tribó and Waddock, 2010) or base a company’s performance on interviews (Bansal, 2005; Clemens, 2006). I use a quantitative valuation, i.e. the account of pollution scaled to both common equity and total revenues. Although both Sharfman and Fernando (2008) and Clarkson et al (2011) use quantitative numbers as well, the first uses only the percentage of total waste generation that is treated on site, or that is reused or recycled, and the latter uses the amount of pollution propensity to categorize firms in five categories, and uses only the categorization in the regression.
A second explanation is related to the control variables used in the regression. I already explained in the literature review section that the inclusion or omission of relevant variables can cause inconsistency in results, demonstrated by McWilliams and Siegel (2000) and Surroca, Tribó and Waddock (2010). McWilliams and Siegel (2000) state that R&D intensity is important to include next to firm riskiness, size and industry, while Surroca, Tribó and Waddock (2010) use intangible resources to explain the relationship between environmental and firm performance. Sharfman and Fernando (2008) include next to firm size, leverage and industry a variable that accounts for the percentage of share ownership concentration and institutional shareholders, and they confirm that “firms with more dispersion in the number of shareholders experienced a lower cost of equity capital”.
34
Table VII. Cost of equity regression results based on % equity
Panel A. Results based on market value leverage
This table presents the results of the regression: E(Rit) = β0 + β1(GHGit-1) + β2(Market Capitalizationit) + β3 Leverageit) + β4(Industry Dummiesit) + εit. Variable definitions: E(Rit) is the cost of equity, GHG is greenhouse gas emissions divided by common equity, market capitalization is the natural logarithm, market value leverage is the natural logarithm of the ratio book assets to market equity, and industry is a dummy variable for capital-intensive and non-capital-intensive industries. The figures reported are coefficient estimates and p-values (in parentheses). ***, **, * denote significant levels at 1, 5, and 10% respectively.
Variable Cost of equity
GHG -0,0022***
(-3,6)
Market capitalization 0,0024*
(2,52)
Market value leverage 0,0057***
(-3,65) Industry -0,0025 (-1,48) Constant 0,0781*** (11,38) Adj. R² 0,0206 F-statistic 14,78*** N 2621
Panel B. Results based on book value leverage
This table presents the results of the regression: E(Rit) = β0 + β1(GHGit-1) + β2(Market Capitalizationit) + β3 Leverageit) + β4(Industry Dummiesit) + εit. Variable definitions: E(Rit) is the cost of equity, GHG is greenhouse gas emissions divided by common equity, market capitalization is the natural logarithm, book value leverage is the natural logarithm of the ratio book assets to book equity, and industry is a dummy variable for capital-intensive and non-capital-intensive industries. The figures reported are coefficient estimates and p-values (in parentheses). ***, **, * denote significant levels at 1, 5, and 10% respectively.
Variable Cost of equity
GHG -0,00121***
(-3,35)
Market capitalization 0,0029***
(2,87)
Book value leverage 0,0044***
35
Table VIII. Cost of equity regression results based on % revenues
Panel A. Results based on market value leverage
This table presents the results of the regression: E(Rit) = β0 + β1(GHGit-1) + β2(Market Capitalizationit) + β3 Leverageit) + β4(Industry Dummiesit) + εit. Variable definitions: E(Rit) is the cost of equity, GHG is greenhouse gas emissions divided by total revenues, market capitalization is the natural logarithm, market value leverage is the natural logarithm of the ratio book assets to market equity, and industry is a dummy variable for capital-intensive and non-capital-intensive industries. The figures reported are coefficient estimates and p-values (in parentheses). ***, **, * denote significant levels at 1, 5, and 10% respectively.
Variable Cost of equity
GHG -0,0021***
(-3,37)
Market capitalization 0,0026***
(2,71)
Market value leverage 0,0049***
(3,2) Industry -0,0035* (-2,13) Constant 0,079*** (11,59) Adj. R² 0,0185 F-statistic 13.55*** N 2659
Panel B. Results based on book value leverage
This table presents the results of the regression: E(Rit) = β0 + β1(GHGit-1) + β2(Market Capitalizationit) + β3 Leverageit) + β4(Industry Dummiesit) + εit. Variable definitions: E(Rit) is the cost of equity, GHG is greenhouse gas emissions divided by total revenues, market capitalization is the natural logarithm, book value leverage is the natural logarithm of the ratio book assets to book equity, and industry is a dummy variable for capital-intensive and non-intensive industries. The figures reported are coefficient estimates and p-values (in parentheses). ***, **, * denote significant levels at 1, 5, and 10% respectively.
Variable Cost of equity
GHG -0,0019***
(-3,05)
Market capitalization 0,0032***
(-3,25)
Book value leverage 0,0038***
