Socially Responsible Investment Styles: Equity Risk, Return and Valuation. Working Title: Socially Responsible Investment Styles

Transcription

1 Socially Responsible Investment Styles: Equity Risk, Return and Valuation Working Title: Socially Responsible Investment Styles Conference Theme: Investment Strategy by Swee-Sum Lam 1 Gabriel Henry Jacob Andy Toh Song Yee 1 Swee-Sum Lam, CPA, CFA, is with the National University of Singapore, Department of Finance. She is also the Director, Asia Centre for Social Entrepreneurship and Philanthropy, NUS Business School; swee.sum@nus.edu.sg. Gabriel H. Jacob is with the Asia Centre for Social Entrepreneurship and Philanthropy, NUS Business School; bizhvj@nus.edu.sg. Toh is with the NUS Business School; tohsongyee@gmail.com. Lam would like to acknowledge NUS research grant for this project.

2 Socially Responsible Investment Styles: Equity Risk, Return and Valuation Abstract This study examines whether corporate ESG (Environmental, Social and Governance) performance affects risk, return and stock valuation globally. We analyze constituents of the MSCI All Country World Index, in a sample of 253,929 firm-month observations from 2002 to We find that while there is no financial benefit or cost in the form of abnormal returns when adopting positive ESG investment screens in a global context, investing socially does lead to reductions in systematic risk globally and idiosyncratic risk in developed markets. Globally, ESG performance has been priced in market to book valuation ratios, which may explain the absence of four-factor abnormal returns. Interestingly, specific countries have country-specific ESG concerns. These findings have implications for investment strategies at both the global and country levels through security selection decisions.

3 1. Introduction Socially responsible investments (SRI) have been a steadily growing market segment within the global asset management industry. According to the US Social Investment, the trade membership association for social investors, social investing enjoyed annual growth rates greater than 13% where its assets under management (AUM) grew from US$2.71 trillion in 2007 to US$3.07 trillion in 2010 (US SIF, 2012). Similar trends are also reported in Europe by the European Sustainable Investments Forum where total SRI AUM grew from 2.7 trillion in 2007 to 5 trillion in 2009 (Eurosif, 2010). According to the Association for Sustainable & Responsible Investment in Asia, SRI AUM in Asia is expected to grow at a compounded annual growth rate of 26% to reach US$3 trillion in 2015 (Sheil, 2011). Despite these rapid developments, we note that there is still a lack of literature exploring the risk, return and the impact on equity valuation of a global SRI investment strategy implemented beyond the US market (Derwall, Keodijk, & Horst, 2011). Thus, we aim to address this gap in the literature by using constituents of the MSCI All Country World Index and Asset4 ESG scores to examine whether ESG performance plays a role in influencing risk, return and stock valuation globally. In addition, we investigate if crosscountry variation in ESG characteristics affects the choice of ESG investment screens and therefore investment strategies in constructing socially responsible portfolios. Our findings suggest that while there is no significant financial cost or benefit in the form of abnormal returns by investing in socially responsible equities, we find evidence of portfolio risk reduction benefits in socially responsible portfolios. For instance, investing in companies with good ESG performance do reduce portfolio systematic risk globally. This systematic risk reduction is delivered by investing in companies of higher market capitalization, which on average has better ESG performance. Idiosyncratic risk reductions

4 are however observed in socially responsible assets in developed markets only and not emerging markets. We also find that ESG performance has been priced into market to book valuation ratios globally. This explains the lack of statistical significance of abnormal returns attributable to ESG performance since the HML factor is included as part of the Carhart fourfactor model in controlling for risk. In addition, we find that specific countries have countryspecific ESG concerns, which influence the appropriate choice of ESG investment screens. The practical implications of these findings are as follows. Investors can be socially responsible and invest with investment managers that operate with a SRI mandate without incurring any significant financial cost on their investment performance. Furthermore, investment managers that operate without a SRI mandate can reduce their portfolio idiosyncratic risk by implementing ESG screens when performing security selection. In addition, it is unwise to impose the same ESG screens across all countries as local circumstances and culture influence the investment performance of the various ESG screens. These findings are also relevant for managers who may find it beneficial to improve certain aspects of their ESG performance based on their location in order to improve firm valuations. These findings also suggest that a global ESG standard, such as the global reporting standards proposed by the Global Reporting Initiative (GRI), might not be adequate as a measure in the benchmarking of ESG performance globally due to the role that local circumstances play in the perception of moral values.

5 2. Background 2.1. Performance of SRI Portfolio Following the rapid development of the SRI industry, there are an increasing number of studies exploring the performance of SRI that could be categorized into two general strands of research. The first strand studies the performance of SRI mutual funds and the second strand examines the performance of socially screened stock portfolios. The first strand of literature studies the performance of SRI mutual funds by comparing the performance of SRI funds and conventional mutual funds. In a review of the empirical performance of SRI funds by Renneboog, Horst and Zhang (2008), SRI funds perform differently in different markets. While the average performance of SRI mutual funds in the US and UK is not different from that of conventional funds, SRI funds in Continental Europe and Asia-Pacific underperform that of the conventional funds. This result implies that the impact of SRI on portfolio returns is not homogeneous across different markets. However, we note that the studies of mutual funds data are affected by the drawback of not being able to separately attribute the performance to its focus on SRI or because of non-sri related characteristics such as the skills of the fund manager, fees, and transaction costs (Lee & Faff, 2009). To overcome this limitation, our paper focuses on the performance of socially screened portfolio formed using ESG rating that disentangles ESG from non-sri factors. The second strand of literature studies the performance of socially screened stock portfolios, by examining the performance of hypothetical portfolios formed by either employing negative screens which exclude socially controversial stocks or positive screens to select stocks that excel in ESG performance. Several studies show that socially controversial stocks produce abnormal positive returns after controlling for conventional factor models (Derwall et al., 2011). One such study

6 is Hong and Kacperczyk (2009) that studies the effect of social norms on markets by using sin stocks, defined as companies involved in tobacco, alcohol and gaming. They find that a portfolio of sin stocks significantly outperform comparable stocks by more than 4.5% per year while international sin stocks outperform by 2.5% per year. Outperformance of socially controversial stocks is also found in markets outside US such as China (Visaltanachoti, Zou, & Zheng, 2009) and Europe (Salaber, 2007). This implies that social factors impact portfolio returns globally. Nonetheless, the exclusion of sin stocks alone may be too narrow a definition of social investment screens as the Social Investment Forum reported that SRI managers employ multiple screens, such as community, environment, diversity and human rights, at the same time (Kempf and Osthoff, 2007). The literature on the impact of positive screens on investment performance find that portfolios formed using SRI factors produce positive abnormal returns in the US market. Statman and Glushkov (2009) analyze the relation between corporate social and stock performance of the US market on the portfolio level using KLD data as the proxy for corporate social performance. They find significant advantage in tilting towards stocks with high KLD social responsibility scores after taking into account the Carhart four factors. Kempf and Osthoff (2007) find that buying high-rated ESG portfolio and selling low-rated ESG portfolio yields a positive four-factor alpha of up to 8.7% per year. This abnormal return remains significant even after taking into account transaction costs. Derwall, Guenster, Baeur, and Koedijk (2005) show that equity portfolio with high eco-efficiency score provides a four-factor alpha of 3.98% per year and outperforms the low-ranked portfolio by 5.06% per year. Edmans (2009) analyzes the relationship between employee satisfaction and long-stock returns and find that a value-weighted portfolio of the 100 Best Companies to Work For in America produces a four-factor alpha of 3.5% per year. These results imply intangibles are not fully valued in the stock market and SRI screens can improve investment returns.

7 Other studies show that although ESG does not significantly improve investment returns, it is positively related with market valuation. Peiris and Evans (2010) find a lack of consistency in the relationship between ESG ratings and stock returns but a clearly positive relationship between ESG ratings, firms operating performance and market valuation. Meanwhile, Galema, Plantinga and Scholtens (2008) find that SRI had a significant positive impact on stock returns. This positive impact on stock returns is delivered by a statistically significant influence on market to book ratios instead of positive alpha generation in a linear regression model. Studies on the performance of screened SRI portfolio outside the US market are rare. The few studies that focus on the European market do not show significant relationship between ESG screens and portfolio returns. Brammer, Brooks and Pavelin (2006) study the link between ESG and performance of UK firms using corporate social performance data from FTSE s Ethical Investment Research Service (EIRIS). Although they find that firms with low social responsibility exhibit higher financial returns, it is not statistically significant. Another study that investigates the effect of ESG factors on the financial performance of UK firms also shows a similar result (Humphrey, Lee, & Shen, 2010). They do not find any difference in the financial performance of firms, systematic risks, book-to-market ratios and momentum exposures between firms with high or low ESG ranking. A study by Velde, Vermeir and Corten (2005) on SRI portfolios in the European Monetary Union also do not show that portfolio with high CSR rating significantly perform better than portfolio with low CSR rating. In summary, our review shows that despite the global growth in SRI investments, there is little research on SRI beyond the US market. The limited studies outside the US market show that ESG factors yield different impact in different markets and imply that the empirical findings in the US market should not be generalized globally.

8 2.2. Risk of SRI Portfolio Besides return, employing ESG screens for portfolio construction also affects portfolio risk. According to Boutin-Dufresne and Savaria (2004), eliminating lesser socially responsible stocks leads to smaller investment universe, and thus social investor s portfolio diversifiable risk will not be completely eliminated. Moreover, social screens can create uncompensated risk even in large portfolios as the impact of social screens is decidedly nonrandom (Kurtz, 1997). However, SRI advocate argues that being socially responsible could reduce firm-specific risk. For example, socially responsible firms are less likely to face adverse events such as lawsuits or labor strikes that could substantially affect their profitability. The empirical evidence finds that socially responsible portfolio exhibits lower portfolio risk. For example, Boutin-Dufresne and Savaria (2004) examine the diversifiable risk of a responsible portfolio and non-responsible portfolio comprising of Canadian firms and find that the responsible portfolio exhibits the lowest idiosyncratic risk. In another study, Lee and Faff (2009) found similar result in their studies of the constituents of the Dow Jones Sustainability Index. Comparing the differences in idiosyncratic risk between the leading and lagging socially responsible firms, they find strong evidence that leading firms exhibit lower idiosyncratic risk than lagging firms. In our review, we have not encountered any literature covering this research area using ESG rating databases on a global scale, which will be addressed in this paper. 2.3 Risk and Return of SRI Portfolio Globally Given the above research issues, we raise the following research questions: Does riskadjusted return on ESG investing differ globally? What is the effect of ESG on portfolio risk globally? Has ESG performance been priced into equity valuation in global markets?

9 This study extends current literature by addressing the impact of implementing ethical screens on equity risk, return and valuation in a global context. Currently, literature pertaining to corporate social performance and stock returns is largely limited to the US and there are thus significant gaps with regards to the issue beyond the US. As corporate social performance comprises of multiple criteria, it is possible that social investors from different markets value the various aspects of corporate social performance differently from conventional US investors. Brammer, Pavelin and Porter (2006) conclude that institutional and shareholder pressures seem to be region-specific and the corporate social performance of firms are largely shaped by the preferences of local stakeholders. Therefore, there is a need for firms to shape their social performance strategies to their geographical profile. Brammer, Brooks and Pavelin (2006) also find that it is important for firms to achieve a fit between the type of corporate social performance and the firm s stakeholder environment in order to enhance corporate reputation. As such, our paper aims to identify the components of corporate social performance that is relevant to stakeholders across different geographies and how these factors are incorporated in the market in the form of abnormal returns, portfolio risk and valuation for socially responsible securities across different markets. This study uses ESG ratings from Asset4 for its global coverage to provide consistency in ratings across different markets. As Hedesstrom, Lundqvist and Biel (2011) find low to moderate convergence among ratings provided by ESG rating agencies, we argue that comparing existing studies on socially screened portfolio in the US market and non-us market may give an imperfect picture. This is because most existing studies in the US market use ESG rating from KLD while non-us studies use ESG ratings from other vendors. Thus, using ESG ratings from a single vendor is better to capture the differential impact of ESG in different markets as it is not affected by different rating methodology employed by vendors.

10 The rest of our paper proceeds as follows. In section 3, we describe our data set and the methodology to construct the hypothetical portfolios and measure their performance. In section 4, we report the empirical results from our analyses. In section 5, we discuss the robustness tests. Section 6 concludes this paper. 3. Data and Methodology In this study, we employ the Asset4 data as construct for corporate social performance. Asset4, a database of Thomson Reuters, provides environmental, social and governance information, primarily for use by professional investors and corporate executives. Asset4 collects information from publicly available sources, such as annual reports, CSR reports and NGO websites with no subjective inputs from analysts in the formulation of its ESG ratings. Asset4 ESG ratings are derived by equal-weighting Asset4 key performance indicators and then z-scored and normalized to derive the score from 0 to 100, where 0 reflects the worst score and 100 reflects the best performance. Firms with no Asset4 ratings are excluded from this study. Asset4 is used by institutional investors, such as Blackrock, and NYSE has worked with Asset4 to provide tools to help NYSE-listed companies to benchmark their corporate social performance. As such, using investment screens formulated from Asset4 ESG scores is likely to provide a good overview of the capital market expectations associated with a socially responsible portfolio that is constructed by socially responsible investment managers. The pillars that form the basis of the Asset4 ESG scoring model seems to share much commonalities with the sustainability drivers and outcomes devised by GRI, a non-profit independent organization that devise standards for global sustainability reporting (Weber, Kollner, Habegger, Steffensen & Ohnemus, 2005). Asset4 rates companies based on the following characteristics:

11 Economic Client loyalty (e.g. market share & market leadership) Performance (e.g. inventory performance & cost innovations) Shareholder loyalty (e.g. profit warnings & insider dealings controversy) Social Employment quality (e.g. salaries distribution & strikes) Health & safety (e.g. injuries & health and safety controversies) Training & development (e.g. training costs & supplier ESG training) Diversity (e.g. female to male ratio & work-life balance) Human rights (e.g. child labor and human rights controversies) Community (e.g. cash donations & bribery, corruption and fraud controversies) Product Responsibility (e.g. quality management & product recalls) Environmental Resource reduction (e.g. materials recycled & reuse ratio) Emission reduction (e.g. greenhouse gas emissions) Product innovation (e.g. environmental R&D expenditure & renewable energy use) Governance Board structure (e.g. CEO-chairman separation & independent board members) Compensation policy (e.g. board member compensation & remuneration structure) Board function (e.g. board attendance & independence of audit committee) Shareholder rights (e.g. voting rights & anti-takeover devices) Vision and strategy (e.g. stakeholder engagement & challenges and trade-off) We draw the Asset4 ESG ratings data from Datastream and use the security s ISIN number as the primary identifier for matching with other datasets. While Asset4 updates its data on a continuous basis whenever significant ESG-related information is made available, we note that Asset4 scores are typically stable over a course of a year and major updates usually comes about with releases of company annual and CSR reports. ESG ratings data for developed markets, US, UK and Japan is drawn from 2002 to 2010 while ESG ratings information for emerging markets is drawn from only 2007 to 2010 due to poor data availability in prior years. We present the number of firms with Asset4 ESG ratings by country and year in Table 1 below. According to Asset4, there is a 6 to 9 months lag for the disclosure of CSR information. Thus we assume a 9 month lag in our analysis to prevent look-ahead bias. For instance, data drawn for Jan 2007 is assumed to be made available publicly by the end of Sep 2007.

12 Table1. Sample Size by Geography and Year This table presents the number of MSCI All Country World Index (ACWI) constituents with Asset4 ESG ratings by geography and year. Year Developed Emerging US UK Japan While we acknowledge that the time period of study is relatively short compared to other studies using KLD data that extended back to the 1970s, the KLD dataset covers the US markets only. To conduct a global study of the impact of implementing ESG screens on investment performance, we therefore choose the Asset4 dataset for its global coverage. Furthermore, we argue that 2002 to 2010 is a sufficiently long to cover a market cycle consisting of both a boom and recession. Even though the time period of study for emerging markets is merely three years, we are of the opinion that this is sufficient in attaining our goal of gaining some insights into the value of ESG screens in emerging markets. To our best knowledge, this is the only study that studies the relation between ESG performance and investment performance for emerging markets. Like other earlier studies that do similar analysis on developed markets outside of the US, such as Brammer, Brooks and Pavelin (2006), their studies has only three years of data but it is sufficient to provide early insights on the value of ESG screens in the UK context.

13 Table2. Summary Statistics of Asset4 Scores for Sample This table presents the summary statistics of Asset4 ESG scores. These scores range from 0 to 100, where 0 reflects the worst score and 100 reflects the best performance across all dimensions of ESG performance. Asset4 Variable Mean Standard Deviation Overall Pillar Economic Pillar Environment Pillar Social Pillar Governance Pillar Our sample comprises of the constituents of the MSCI All Country World Index (ACWI) as of Dec 2010, which we adopt as the proxy for GIM (Global Investible Market) universe where there is sufficient market depth and stock liquidity for investors of all categories. Furthermore, restricting the analysis to constituents of MSCI ACWI has the added benefit of mitigating any biases that arose from the steep increase in security coverage from 2002 to We obtain data from Datastream for the following variables: total return, market capitalization, market-to-book ratio, risk-free rate and adjusted closing price of the constituents. We conduct pair-wise correlation analysis for all variables employed in the study. Consistent with the correlation table of EIRIS data in Brammer, Brooks and Pavelin (2006), there is a high degree of association within corporate social rating scores of the same category, i.e. economic, environmental, social and governance. In addition, the overall score is highly correlated with the economic, environmental, social and governance aggregate scores. For presentation purposes, we show only the correlation matrix of the key ESG variables in each category. As correlations between the ESG variables appear high on the surface, we performed variance inflation factor checks during the regression process to ensure that there are no multicollinearity concerns.

14 Table3. Correlation Matrix of Asset4 ESG Aggregate Scores Overall Economic Environment Social Governance Overall 1.00 Economic Environment Social Governance Socially Responsible Portfolio Construction As most empirical studies highlight that constructing portfolios based on ESG scores alone typically leads to extreme sector tilts, we adopt a best-in-class screening policy in the portfolio construction process to mitigate this issue. We first group the equity universe based on their GICS sector and construct equally-weighted long and short portfolios by triciles on the basis of their Asset4 ESG scores. This helps to ensure that a sector-neutral portfolio is constructed and that a reasonable portfolio size is examined in each case. Occasionally, when there are insufficient companies within an industry sector group, companies were assigned to both the top and bottom tricile portfolios to maintain a balance in the number of securities within each portfolio. Portfolios are rebalanced on a monthly basis to take into account updates in ESG-related information whenever they are publicly available. In our tests, we conduct our analysis by stage of development (developed markets versus emerging markets), and by country (US, UK and Japan). We follow the MSCI s classification framework, which takes into account each country s economic development, size, liquidity and market accessibility, as the bases for classifying constituent countries of the MSCI ACWI into developed and emerging markets. Ideally, we would prefer to perform country-level analysis. However, sample size constraints lead us to work with the developedemerging markets dichotomy and country-level analysis is restricted to US, UK and Japan only.

15 3.2. Portfolio Time Series Regression The monthly differential abnormal returns on SRI portfolios are measured using the Carhart (1997) four-factor model, which consists of the Fama and French (1993) three factors and a momentum factor:,,,, 1 where R i,t,top represents the return on a socially desirable portfolio with high Asset4 ESG scores in month i and R i,t,bot represents the return on a socially undesirable portfolio with low Asset4 ESG scores in month i. RM t represents the market return in month t on an equalweighted market proxy, R ft is the risk free return in month t, SMB t is the difference in monthly return between a small and large-cap portfolio in month t, HML t is the difference in monthly returns between a value and growth portfolio in month t, MOM t represents the difference in monthly returns long on one-year winners and short on one-year losers and α represents the differential abnormal performance due to difference in corporate social performance of the two portfolios. Fama and French (2011) show that while global four-factor models have sufficient explanatory power in explaining returns of global portfolios, there are short-comings in assuming integrated pricing across geographical regions, i.e. global models should not be used to explain regional portfolio returns and local four-factor models have greater explanatory power. Thus, it is important to use localized estimates of the Carhart four factors so as to avoid the problems associated with assuming integrated pricing across regions. Therefore, we estimate explanatory factor portfolios in the asset pricing tests, i.e. SMB, HML & MOM, using portfolios constructed from 2x3 sorts on size and market to book or size and momentum as per Fama and French (2011) based on our geographical specifications. For instance, we limit the market proxy to stocks from developed countries within the MSCI

16 ACWI for the purpose of estimating SMB, HML & MOM for the developed markets regional model. For the regional models, returns are expressed in USD terms and returns on 3-month US T-bills are used in Equation (1). For local country models, returns are expressed in local currency and returns on domestic short term government bonds are used as the risk free rate in Equation (1). Regression analysis in Equation (1) allows us to analyze if there is a significant difference in the systematic risk involved in investing in socially desirable assets and socially undesirable assets. Ang, Hodrick, Xing and Zhang (2006) define systematic risk as the sensitivity of returns relative to the four factors of the Carhart (1997) model. If β 1, β 2, β 3 or β 4 in Equation (1) are significantly negative, we may conclude that socially desirable assets have indeed less systematic risk relative to socially undesirable stocks. Newey-West (1987) standard errors are used to ensure robustness against autocorrelation and heteroskedasticity that may be present in the regression. Choice of lag for the Newey-West estimator is determined by following the procedure in Newey-West (1994). Following the definition of Ang et al., (2006), we define idiosyncratic risk relative to the four factor model as the standard deviation of the residuals of the regression model. We first obtain the rolling 12-month standard deviation of the residuals of the regression for both the socially desirable and socially undesirable portfolios to estimate their idiosyncratic volatilities. By using the nonparametric Wilcoxon signed rank paired tests to compare the idiosyncratic risk of both portfolios, we assess if there is significantly less idiosyncratic risk associated with socially desirable securities. We also perform Mann-Whitney and Kruskal- Wallis paired tests to ensure robustness of our results which are available on request. The findings are similar to those from the Wilcoxon signed rank paired tests.

17 3.3. Corporate Social Performance and Equity Valuation To evaluate the impact of corporate social performance on equity valuation, we estimate the following regression equation:,,, log, & & 2 where MTB i,t represents the market to book ratio for stock i in month t and the vector ESG again represents the four Asset4 aggregate indicators in key categories. In order to avoid repetition of ESG variables, we only include the individual Asset4 performance indicators and not the composite measures. To account for the time lag taken for Asset4 ESG information to be made available to public, we lag the ESG variable by nine months. The remaining control variables are drawn from extant literature and they are also employed by Hong and Kacperczyk (2009) and Galema et al., (2008). ROE i,t represents the trailing 12 months return on equity for the stock i in month t, R&D t is the fraction of sales spent on R&D in year t, R&Dmissing is a dummy for stocks that do not disclose R&D spending and finally age i,t is the age of the firm in year t. Following the advice provided by Petersen (2009) and approach used by Galema (2008), pooled OLS with robust standard errors is used instead of the Fama Macbeth approach. The latter is useful in producing unbiased standard errors when there is crosssectional correlation and little serial correlation, such as asset pricing models where serial correlation in stock returns are low. Book to market ratios are however correlated both across time and the cross-section and thus Fama Macbeth regressions are not appropriate. Therefore, we use a normal OLS model and adjust for time effects by including time dummies (Tdum) and cross-sectional effects by using sector and country dummies where applicable. From an investment practitioner s perspective, investment screening is typically performed on a

18 country and industry level rather than on a firm level. Therefore, we argue that our approach of estimation is both academically robust and an accurate reflection of the screening practices performed by the investment industry. 4. Findings 4.1. Returns Attributable to ESG Performance in a Portfolio Context In this section, we investigate if there are any differential abnormal returns that are attributable to making investment decisions based on Asset4 ESG scores. Assuming that there is abnormal investment performance attributable to certain corporate social performance attributes, we also investigate if this outperformance is actually consistent across geographical regions. Using our analysis of US and UK social portfolios formed using Asset4 ESG scores, our findings conform with other empirical literature. From our findings in Table 3, monthly abnormal returns, attributable to ESG performance, are generally negative. While these results seem to support that socially responsible stocks have lower expected returns than socially irresponsible stocks, these differences are not statistically significant at the 10% confidence level. This finding is consistent with earlier US study (Galema et al., 2008) and UK study (Brammer, Brooks and Pavelin, 2006), where they find negligible differences in investment performance from investing in SRI portfolios.

19 Table 4. Regression of Differenced Portfolio Returns against Carhart Four Factors Portfolios are constructed taking a long position in stocks of companies in the top third by Asset4 ESG characteristics and a short position in stocks of companies in the bottom third by Asset4 ESG characteristics. The dependent variables in these regressions are monthly differenced portfolio returns. The independent variables are the standard Carhart four factors. Standard errors are adjusted for heteroskedasticity and serial correlation using Newey West (1987) standard errors. Lag selection for the Newey West adjustments is determined using the procedure by Newey West (1994). p-values are reported in the parentheses. ***1% significance **5% significance *10% significance Portfolio Sorts on MSCI ACWI Constituents in Developed Markets Monthly Abnormal Returns Monthly Market Excess Returns SMB HML MOM Adjusted R 2 Economic -0.18% *** -0.11* (0.26) (0.29) (0.01) (0.10) (0.32) Environment -0.16% 0.09*** -0.32*** 0.32*** (0.12) (0.00) (0.00) (0.00) (0.54) Social -0.22% 0.15*** -0.64*** 0.18*** 0.10*** 0.52 (0.12) (0.00) (0.00) (0.01) (0.01) Governance -0.36% 0.16** -0.57** -0.32** 0.28*** 0.36 (0.19) (0.02) (0.04) (0.03) (0.00) Portfolio Sorts on MSCI Emerging Markets Constituents Monthly Abnormal Returns Monthly Market Excess Returns SMB HML MOM Adjusted R 2 Economic -0.15% ** (0.68) (0.39) (0.19) (0.02) (0.30) Environment 0.14% -0.07** *** 0.06 (0.66) (0.03) (0.55) (0.54) (0.00) Social -0.04% ** ** 0.10 (0.93) (0.27) (0.05) (0.38) (0.01) Governance -0.56% ** (0.11) (0.76) (1.00) (0.01) (0.75)

20 Monthly Abnormal Returns Portfolio Sorts on MSCI US Constituents Monthly Market Excess Returns SMB HML MOM Adjusted R 2 Economic -0.13% -0.11*** -0.30*** -0.09** (0.25) (0.00) (0.00) (0.04) (0.54) Environment -0.10% 0.06* -0.24*** 0.21*** 0.11*** 0.39 (0.31) (0.05) (0.01) (0.00) (0.00) Social -0.12% *** 0.25*** 0.08*** 0.40 (0.17) (0.46) (0.00) (0.00) (0.00) Governance -0.02% 0.05** -0.19*** 0.16*** 0.10*** 0.33 (0.84) (0.03) (0.00) (0.00) (0.00) Monthly Abnormal Returns Portfolio Sorts on MSCI UK Constituents Monthly Market Excess Returns SMB HML MOM Adjusted R 2 Economic -0.38% *** (0.19) (0.63) (0.00) (0.36) (0.84) Environment 0.08% -0.18*** *** (0.79) (0.00) (0.11) (0.00) (0.19) Social -0.03% *** 0.16*** (0.93) (0.96) (0.00) (0.01) (0.35) Governance -0.14% *** 0.15*** (0.52) (0.72) (0.00) (0.00) (0.52) Monthly Abnormal Returns Portfolio Sorts on MSCI Japan Constituents Monthly Market Excess Returns SMB HML MOM Adjusted R 2 Economic 0.35% *** * 0.15 (0.23) (0.67) (0.00) (0.31) (0.05) Environment -0.02% 0.25*** -0.35*** (0.96) (0.00) (0.00) (0.71) (0.85) Social -0.10% 0.13*** -0.32*** (0.81) (0.00) (0.00) (0.82) (0.69) Governance -0.31% 0.18*** -0.39*** (0.29) (0.00) (0.00) (0.92) (0.57)

21 4.2. Risk Reduction Attributable to ESG Performance in a Portfolio Context In this section, we analyze whether globally there is any significant reduction in both systematic and idiosyncratic risk by investing in socially responsible portfolios. In order to evaluate the sensitivity of the socially responsible portfolios to systematic risk, we analyze the coefficients of monthly market excess returns, SMB, HML and MOM from the ESG long-short portfolios in our earlier Carhart regressions in Table 3. Across all geographical regions, the long-short portfolio, formed based on the various ESG characteristics have consistent and statistically significant negative sensitivities to the SMB risk factor, which suggests a plausible reduction in systematic risk attributable to the size factor. This finding is within our expectations as bigger firms have more resources to carry out activities to enhance their ESG performance and hence a socially desirable portfolio is likely to have a bias to firms that are of larger market capitalization. We also note that the ESG ratings of firms and their market capitalizations are positively correlated, which confirmed our intuition that firm size and ESG performance are positively related. We observed no consistent relation pertaining to reduced systematic risk to the monthly excess market returns (RM-RF), value (HML) or momentum (MOM) factors across the various ESG factors and geographical region. As such, we conclude that there is likely to be systematic risk reduction benefits from investing in socially desirable equities and this risk reduction is primarily delivered through investing in stocks of larger market capitalizations. These findings support that socially responsible stocks have less systematic risk.

22 Table 5. Nonparametric Paired Tests of Portfolio Idiosyncratic Risk Socially desirable portfolios in this table are constructed taking stocks in the top third by Asset4 ESG characteristics and socially undesirable portfolios are constructed by taking stocks in the bottom third by Asset4 ESG characteristics. We adopt Ang et al. (2006) definition in computing idiosyncratic risk relative to the Carhart four factors, where idiosyncractic risk is defined as the rolling 12-months standard deviation of the residuals from the Carhart regression model. Nonparametric paired tests were performed to test the alternative hypoethesis of whether socially desirable portfolios have significantly less idiosyncratic risk than a socially undesirable one. All values are p-values from the Wilcoxon signed rank test. ***1% significance **5% significance *10% significance Developed Emerging US UK Japan Overall 0*** *** 0*** 1 Economic 0*** 0.8 0*** Environment *** 0*** 0*** Social *** 0*** Governance 0*** *** 0.99 We next look at whether there is significant idiosyncratic risk reduction from holding socially responsible securities by comparing the idiosyncratic risk of a socially responsible and socially irresponsible portfolio. Table 5 suggests that there are indeed benefits to investing in portfolios consisting firms of good ESG performance as the socially responsible portfolios have significantly less idiosyncratic risk in general. Assuming CAPM does hold, investors are not rewarded additional returns for bearing additional idiosyncratic risk in their portfolios. While portfolio diversification does help to reduce idiosyncratic volatility, investors in practice are unlikely to be able to hold a large number of securities in practice to eliminate idiosyncratic risk completely and thus holding a socially responsible portfolio in this sense might be beneficial. Comparing the results of US and UK in Table 5, we note that there are significant differences in the ESG factors that offer idiosyncratic risk reduction even though both countries are of relatively close geographical proximities. For instance, social ESG factors that worked in reducing portfolio idiosyncratic risk in the UK does not seem to provide similar benefits in the US. This supports Aguilera, Williams, Conley and Rupp

23 (2006) s findings that the social aspects of ESG concerns are more likely to be incorporated into core corporate governance in the UK than in the US. Interestingly, our finding in Table 5 suggests that the reduction of idiosyncratic risk from holding a socially responsible portfolio is non-existent in emerging markets. A plausible reason is the regulatory framework and law enforcement dealing with environmental and social issues in emerging economies are generally less established than developed economies due to their stage of economic development. Firms in these countries are thus less prone to litigation risk that is attributable to ESG issues. Our result suggests that local circumstances play an important role in determining the type of ESG characteristics that are rewarded by capital markets. Considering that the Asset4 ESG framework is probably adopted from the GRI, a global sustainability reporting standard for companies, these results seem to imply that current indicators of corporate social performance might not be comprehensive enough as a measure to be used on a global scale. Similarly, Welford (2005) also observes that many CSR policies are based on localized issues and cultural traditions at the country level. We conclude that the total risk associated with investing in socially desirable equities relative to poor social performers are lower due to the lower idiosyncratic risks associated with the former and systematic risk reduction from investing in such stocks that are typically of larger market capitalizations Impact of ESG Performance on Stock Valuation In this section, we extend Hong and Kacperczyk (2009) and Galema et al. s (2008) study to investigate if ESG performance is priced into firm s valuations globally. If ESG performance is already priced into share prices globally, this will rationalize earlier results

24 where we are unable to establish a statistically significant link between ESG performance and portfolio returns. In order to study the impact of ESG performance on market to book ratios, we analyze the coefficients of the Environmental, Social and Governance factors in Table 6. We find that ESG performances do influence firm valuations globally. Corporate governance seems to be the only ESG factor that has the ability to improve firm valuations in both developed and emerging markets. In addition, we note that improvement in governance has the greatest positive impact on valuations in emerging markets where every one point of improvement in the Asset4 governance aggregate score leads to an increase of times in market to book ratios. In the developed markets, a similar improvement only increases market to book ratios by times. Klapper and Love (2004) find that better governance matters more in countries with weak shareholder protection and poor judicial efficiency as firms with better governance will are less likely to require the legal system to resolve governance conflicts. In US and UK however, improving performance in corporate governance has a negative effect on firm valuation but this effect is not statistically significant at 10% levels. Klapper and Love (2004) also find that firms cannot compensate fully for the absence of strong laws and enforcement through better governance alone. As judicial and corporate governance standards are already high in the US and UK, investors may not be willing to pay a premium for the marginal benefits from further improvements in governance. We find that excellence in environmental issues have consistent effect of lowering firm valuations in all the developed markets studied. For instance, every one point improvement in Asset4 environment aggregate score lowers firms market to book ratios by on average. This implies that enhancing environmental performance beyond the minimum standards required by regulators can be costly and the benefits of doing so may not necessarily be commensurate with the costs involved. In emerging markets however, we fail

25 to find any statistically or economically significant impact on market to book ratios that is attributable to better environmental performance. This may imply a possible lack of concern on environmental issues by investors in emerging markets. Excellence in social issues is find to have led to improvements in market to book valuations for firms in the developed region, US and UK in the order of to times for every one point improvement in Asset4 social score. In Japan and emerging markets however, performance in social issues seem to have the opposite effect on firms market to book valuations. This suggests that investors in the latter markets may have little interests in social concerns as defined by Asset4.

26 Table 6. Regression of Market to Book Ratios against Asset4 ESG Scores,,, log, & & The dependent variable in these regressions is market to book ratios of firms from the MSCI ACWI constituents, grouped by geographical specifications. ESG represents the scores of the 4 Asset4 aggregate indicators, ROE is trailing 12 months return on equity, age is the age of the firm at the end of the year, R&D is fraction of sales spent on R&D in the year and R&Dmissing is a dummy for firms that do not disclose R&D spending. Sector and time dummies are added for all the models while country dummies are only added for the regional specifications. Parentheses consists White's (1980) heteroscedasticity robust p-values. Variance inflation factor checks were conducted to ensure that there are no multicollinearity issues. Ratio items are winsorized at 1% levels. ***1% significance **5% significance *10% significance Developed Emerging (1) (2) (3) (1) (2) (3) Intercept 3.47*** 4.007*** 4.163*** 0.577** 0.927*** 0.703*** (0.00) (0.00) (0.00) (0.02) (0.00) (0.01) Economic *** *** *** *** *** *** (0.00) (0.00) (0.00) (0) (0.00) (0.00) Environment *** *** *** (0.00) (0.00) (0.00) (0.33) (0.65) (0.98) Social 0.006*** 0.006*** 0.006*** *** *** *** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Governance 0.001** 0.002*** 0.002*** 0.007*** 0.007*** 0.007*** (0.04) (0.00) (0.00) (0.00) (0.00) (0.00) ROE 5.816*** 5.772*** 5.806*** *** *** *** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) log(age) -0.26*** *** *** -0.19*** (0.00) (0.00) (0.00) (0.00) R&D 2.751*** 4.632*** (0.00) (0.00) R&Dmissing *** 0.224*** (0) (0.00) Sector Dummies Y Y Country Dummies Y Y N Adjusted R

27 Table 6. Regression of Market to Book Ratios against Asset4 ESG Scores (Cont d) US UK Japan (1) (2) (3) (1) (2) (3) (1) (2) (3) Intercept 3.829*** 4.443*** 4.946*** 4.488*** 4.756*** 4.305*** 1.742*** 2.197*** 2.077*** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Economic *** *** *** 0.009*** 0.009*** 0.01*** 0.001*** 0.001*** 0.001*** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Environment *** *** *** *** -0.02*** *** *** *** *** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Social 0.008*** 0.008*** 0.009*** 0.004** 0.006*** 0.006*** *** *** *** (0.00) (0.00) (0.00) (0.029) (0.00) (0.002) (0.00) (0.00) (0.00) Governance ** ** ** * 0.003*** 0.004*** 0.003*** (0.04) (0.37) (0.40) (0.024) (0.013) (0.05) (0.00) (0.00) (0.00) ROE 5.008*** 5.012*** 5.015*** 4.197*** 4.142*** 4.164*** 2.256*** 2.234*** 2.226*** (0.00) (0.00) (0.00) (0.00) (0.00) (0) (0.00) (0.00) (0.00) log(age) *** *** *** *** *** *** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) R&D 1.605*** 5.133*** 0.916*** (0.00) (0.00) (0.00) R&Dmissing *** 0.314*** 0.219*** (0.00) (0.00) (0.00) Sector Dummies Y Y Y Country Dummies N N N N Adjusted R

28 5. Robustness To check the robustness of our results, we perform additional analyses that are available on request. We re-construct socially responsible portfolios by quartiles and quintiles instead of triciles and repeat our regression analysis and non-parametric paired tests. The portfolio regression results remain fundamentally similar and there is no evidence to suggest a financial cost or benefit in the form of abnormal returns. Systematic risk reductions in the form of exposure to the size factor are similarly observed in a global context when performing these robust tests. Our findings pertaining to idiosyncratic risk reductions by investing in socially responsible portfolios remain consistent with the exception of the social factor which no longer reduces idiosyncratic risk in Japanese equities portfolios. 6. Conclusion One key limitation of this paper is the relatively short period of study from 2002 to 2010 due to constraints in the data availability of ESG ratings as compared to similar papers of US markets where the period of study starts from the 1970s. The data is even more limited for emerging markets where the period of study is only from 2007 to This made it difficult to perform robustness tests of inter-temporal stability due to the infeasibility of splitting up the period of study into shorter sub-periods and at the time, maintain a reasonable sample size for each sub-period. As such, we are unable to provide insights on how ESG performance has impacted investment performance over time. Nonetheless, the number of firms disclosing ESG information has been growing steadily and future research can help address these limitations as more data becomes available. In addition, the findings of this study suggest that it is necessary to do country-level analysis. Future research may possibly deal with the development of country- 28

29 specific ESG standards to take into account local circumstances and country-specific culture and values. To our best knowledge, this study is the first to compare the difference in investment performance and equity valuation attributable to ESG performance at a global level, which includes emerging markets. This extends the current literature in a few dimensions. These findings suggest that investment managers can reduce portfolio risk by taking into account ESG information and underweighting socially irresponsible firms during the portfolio construction process. Our findings also suggest that managers may improve the valuation of their firms by enhancing certain aspects of their firms ESG performance. 29

30 References Aguilera, R., William, C., Conley, J., & Rupp, D. (2006). Corporate governance and social responsibility: A comparative analysis of the UK and the US. Corporate Governance and Social Responsibility, 14(3), Alder, T., & Kritzman, M. (2008). The cost of socially responsible investing. The Journal of Portfolio Management, 35(1), Ang, A., Hodrick, R., Xing, Y., & Zhang, X. (2006). The cross-section of volatility and expected returns. Journal of Finance, 13, Boutin-Dufresne, F., & Savaria, P. (2004). Corporate social responsibility and financial risk. The Journal of Investing, 13, Brammer, S., Brooks, C., & Pavelin, S. (2006). Corporate social performance and stock returns: UK evidence from disaggregate measures. Financial Management, 35, Brammer, S., Pavelin, S., & Porter, L. (2006). Corporate social performance and geographical diversification. Journal of Business Research, 59, Carhart, M. (1997). On Persistence in Mutual Fund Performance. Journal of Finance, 52(1), Derwall, J., Koedijk, K., & Horst, J. T. (2011). A tale of values-driven and profit-seeking social investors. Journal of Banking & Finance, 35, Derwall, J., Guenster, N., Bauer, R., & Koedijk, K. (2005). The eco-efficiency premium puzzle. Financial Analysts Journal, 61(2),