Corporate Social Responsibility and Financial Performance. Hui-Ju Tsai and Yangru Wu * This Draft: 12/7/2015

Transcription

1 Corporate Social Responsibility and Financial Performance Hui-Ju Tsai and Yangru Wu * This Draft: 12/7/2015 Abstract We examine the relationship between corporate social responsibility (CSR) and financial performance by comparing the portfolio returns of firms with improvement and with deterioration in corporate goodness. Overall, the portfolio of firms with improvement in CSR outperforms the market on a risk-adjusted basis. We show that the relation between CSR and financial performance changes with CSR dimensions and depends on the market condition. The up and down portfolios of employee relations have superior performance during both the noncrisis and crisis periods. Portfolios of firms improving in environment and community have superior returns during the non-crisis periods, while portfolio of firms improving in human rights outperforms the market regardless of the market condition. For corporate governance and product characteristics, the relation between corporate goodness and financial performance during the non-crisis and crisis periods is reversed. JEL Classification: G10, G11, M14 Keywords: SRI; Corporate social responsibility; Financial performance The strategy of sustainable and responsible investing (SRI) has become more and more popular among the investment society. In 2012, 11.3 percent of the $33.3 trillion assets under * Tsai is the corresponding author at the Washington College, 300 Washington Avenue, Chestertown, MD 21620, htsai2@washcoll.edu. Wu is at the Rutgers Business School-Newark and New Brunswick, Rutgers University, 1 Washington Park, Newark, NJ 07102, yangruwu@business.rutgers.edu. Wu thanks the Whitcomb Center for Financial Services at the Rutgers Business School for data and financial support. Part of this work was completed while Wu visited the Central University of Finance and Economics. We are responsible for any remaining errors. 1

2 professional management in the United States are invested according to the concept of SRI. The assets engaged in sustainable and responsible investing practice has increased by 486 percent from 1995 to 2012, while the growth rate of total assets under professional management is 376 percent during the same period. 1 Additionally, more and more corporate managers incorporate corporate social responsibility (CSR) into their management agenda and provide corporate responsibility report to the public. In a study conducted on more than 1,000 top executives across countries in 2013, 93% of CEOs believe that sustainability will be important to the future success of their companies, and 81% of CEOs agree that the sustainability reputation of their company is important in consumers' purchasing decisions. 2 Additionally, according to a survey conducted by KPMG in 2013, about 71 percent of the 4,100 surveyed global companies undertake the practice of corporate responsibility reporting. 3 There are two opposite theoretical views about whether corporate managers should take into account of corporate social responsibility while making managerial decisions. The stakeholder theory argues that managers should maximize the welfare of stakeholders of the business, which not only include its stockholders, but also its employees, customers, suppliers, communities, and so on. The stakeholder theory believes that a friendly relationship between firms and their stakeholders increases firm value which ultimately benefits their stockholders. However, Jensen (2001) argues that the stakeholder theory lacks a precise objective function. Due to the conflicts of interests among different stakeholders, stakeholder theory empowers 1 See 2012 Report on Sustainable and Responsible Investing Trends in the United States. 2 See The UN Global Compact-Accenture CEO Study on Sustainability Sustainability-2013.PDF. 3 See The KPMG Survey of Corporate Responsibility Reporting

3 managers to allocate firm resources based on their own preferences, which may not maximize firm value. According to agency theory, managers may use valuable firm resources to engage in CSR that helps to build up their reputation instead of increasing firm profits (see, e.g., Charness and Rabin, 2002; Cheng, Hong, and Shue, 2014; Fehr and Schimdt, 1999; Jensen and Meckling, 1976). This view is supported by Friedman's (1970) argument that the only goal that corporate managers should pursue is to maximize stockholders' interests and activities for corporate goodness are essentially costing stockholders instead of benefiting them (see also Aupperle et al., 1985; McWilliams and Siegel, 1997). Studies that examine the relationship between corporate goodness and financial performance have found mixed evidence. Derwall, Guenster, Bauer, and Koedijk (2005), Kempf and Osthoff (2007), and Statman and Glushkov (2009) show that environmental performance is positively related to stock returns. Similarly, Statman and Glushkov (2009), Edmans (2011), and Derwall, Koedijk, and Ter Horst (2011) indicate that investors are rewarded by investing in companies with high employee satisfaction, and Jiao (2010) finds firm's performance is positively related to its employee relationship and environmental performance. Bebchuk, Cohen, and Ferrell (2009) show a positive impact of corporate governance on financial performance. Several studies also find that firms with higher CSR performance face lower capital cost and have less financial constraints (see, e.g., Attig, Ghoul, Guedhami, Suh, 2013; Cheng, Ioannou, and Serafeim, 2014; Dhaliwal, Li, Tsang, and Yang, 2011; Ghoul, Guedhami, Kwok, and Mishra, 2011; Goss and Roberts, 2009). By contrast, consistent with agency theory, Cheng, Hong and Shue (2014) find that the investment in corporate goodness declines as managerial ownership or monitoring in the 3

4 company increases. Additionally, Hong, Kubik, and Scheinkman (2012) argue that instead of doing well by doing good, firms do good only when they do well. They show that firms that are less financially constrained are more likely to engage CSR activities. Giuli and Kostovetsky (2014) find that Democratic-leaning firms spend more money on CSR than Republican-leaning companies and the investment in CSR is negatively related to future stock returns and ROA. Similarly, Hong and Kacperczyk (2009) and Kim and Venkatachalam (2011) examine the performance of sin stocks and find that sin stocks offer positive abnormal returns. Another strand of research that studies the effect of corporate social responsibility on financial performance is to compare the performance of portfolios that adopt sustainable and responsible investing (SRI) strategies with that of conventional portfolios. Hamilton, Jo, and Statman (1993), Goldreyer Ahmed, and Diltz (1999), Statman (2000), and Bauer, Koedijk, and Otten (2005) find no statistically significant difference between the performance of socially responsible mutual funds and conventional funds. Girard, Rahman, and Stone (2007) show that socially responsible mutual funds have lower returns than conventional funds. Interestingly, Nofsinger and Varma (2014) indicate that during the period of market crises, socially responsible mutual funds perform much better than conventional funds, suggesting that firms with good social performance face a lower downside risk during economic downturn. If firms tend to maintain a similar level of corporate goodness, the market should have incorporated the impact of corporate goodness into their stock prices. Thus, current evidence obtained by comparing the performance of firms with good social performance with that of firms with bad performance may not fully reflect the effect of corporate goodness on financial performance. Different from current studies that compare the financial return of firms with good 4

5 social performance with that of firms with bad social performance, this paper examines the relationship between corporate goodness and financial performance by comparing the portfolio returns of firms in the following year whose social performance has changed significantly within the past year. If corporate goodness is positively related to financial performance, we expect that the portfolio of firms with improvement in corporate goodness will have higher risk-adjusted returns. On the contrary, if that managers engage in CSR reflects an agency problem between managers and stockholders, we anticipate that portfolio of firms with deterioration in corporate goodness has superior performance. The relation between corporate goodness and financial returns may depend on the condition of the financial market. Nofsinger and Varma (2014) find that socially responsible mutual funds have better performance during economic downturn. They argue that investors of socially responsible mutual funds are willing to accept lower returns during the non-crisis periods in exchange for higher returns during the crisis periods. In the second part of this study, we relax model restrictions by allowing a different relationship between corporate goodness and financial returns during the non-crisis and crisis periods. Unlike Nofsinger and Varma (2014) that study the performance of mutual funds, we compare the portfolio returns of firms with improvement or deterioration in corporate goodness. Because the performance of mutual funds in part depends on fund managers' security selection abilities, directly examining the portfolio performance of firms with changes in corporate goodness helps to better identify the possible impact of investment in CSR on firms' financial performance. The portfolio of firms with improvement CSR outperforms the market on a risk-adjusted basis. We show that the relations between CSR and financial performance change with the CSR 5

6 dimensions and depend on the market condition. The up portfolios of environment, community, and human rights have significantly positive abnormal returns that are higher than the abnormal returns of the down portfolios. Both the up and down portfolios of employee relations and product characteristics outperform the market on a risk adjusted basis. By contrast, the portfolio of firms with deterioration in corporate governance shows significantly positive abnormal returns that are significantly higher than the abnormal returns of the down portfolio. The relationship between corporate goodness and financial performance can change and sometimes is reversed during the non-crisis and crisis periods. The up and down portfolios of employee relations outperform the market during the non-crisis and crisis periods, while there is no significant difference between the returns of the two portfolios. Firms with improvement in environment, community, and diversity have positive risk-adjusted returns only during the crisis periods, while firm with improvement in human rights show superior performance regardless of the market condition. For corporate governance and product characteristics, the relation between corporate goodness and financial performance is reversed during the non-crisis and crisis periods. During the non-crisis period, the portfolios of firms declining in corporate governance and product characteristics perform better than the portfolios of firms with improvement, whereas during the crisis periods, firms with improvement perform better. The remainder of the paper is organized as follows. In Section I, we describe the data used in the study. Section II presents the methodology. In Section III, we present the empirical results of the relation between corporate social responsibility and financial performance. Section IV conducts robustness checks. The conclusions are provided in Section V. I. Data 6

7 We use MSCI ESG (formally known as KLD Research & Analytics) to determine firms' social performance. MSCI ESG measures the performance of corporate social responsibility in thirteen dimensions. The first seven dimensions are community, corporate governance, diversity, employee relations, environment, human rights, and product characteristics. Within each dimension, MSCI ESG establishes a list of strengths and concerns (see Appendix A for the list) and uses them as criteria to measure a firm's social performance. For each strength (concern), MSCI ESG assigns a score of one for the presence of strength (concern) in the evaluated firm and zero otherwise. For instance, employee involvement is one of the strength criteria listed in employee relations. The evaluated firm will get a score of one in employee involvement if the firm fulfills the criterion and zero otherwise. Similarly, employment health & safety is one of the concerns listed in employee relations and the evaluated firm gets a score of one if MSCI ESG sees the presence of that concern in the company and zero otherwise. The last six dimensions indicate if firms are involved in controversial businesses, including alcohol, gambling, tobacco, firearms, military, and nuclear power. For each dimension, a firm gets a score of one in that dimension if its operation is involved and zero otherwise. The number of firms covered in the MSCI ESG data has increased over time. From 1991 to 2000, MSCI ESG covered all firms of the S&P 500 and Domini Social index. In 2001 MSCI ESG expanded to cover firms of the Russell 1000 index and in 2002 added firms of the Large Cap Social index. In 2003, all firms of the Russell 2000 and the Broad Market Social index are added into the MSCI ESG data. Table 1 summarizes the number of strengths (S) and concerns (C) considered by MSCI ESG from 1991 to We use the notation GOVˮ, COMˮ, DIVˮ, EMPˮ, ENVˮ, HUMˮ, and PROˮ to represent the dimension of corporate governance, community, diversity, employee relations, environment, human rights, and product 7

8 characteristics, respectively. As Table 1 presents, the number of strengths and concerns considered by MSCI ESG varies over the sample period. Taking the dimension of community as an example, in 2012 there are 2 (1) strengths (concerns) considered by MSCI ESG, compared to 7 (4) strengths (concerns) in The number of strengths and concerns also differs among dimensions. In 2012, while there are 9 (9) strengths (concerns) considered in environment, only 2 (4) strengths (concerns) are examined in human rights. To examine the relation between changes in corporate goodness and financial performance, we focus on the seven dimensions that include a list of strengths and concerns. Firms involved in the six dimensions of controversial businesses are fundamentally different from the other firms in terms of corporate social responsibility and thus are excluded from our study. Because MSCI ESG may have a firm on its list but not examine its performance in all corporate social dimensions, we follow Statman and Glushkov (2009) by excluding firms from our analysis if they do not receive any strength or concern indicator in the dimension we examine. To control for the variation in the number of strengths and concerns considered by MSCI ESG overtime, for each CSR dimension, we standardize each firm s strength (concern) scores by dividing its strength (concern) scores by the total number of strengths (concerns) considered by MSCI ESG in the same year. The standardized strength and concern scores range between zero and one. We then measure a firm s social responsible performance in each dimension by subtracting the standardized concern score from its standardized strength score. Table 1 presents the average performance of corporate social responsibility of sampled firms in each dimension. The performance of corporate social responsibility ranges between -1 and 1 by construction, with -1 (1) meaning the firm receives scores of one in all concerns 8

9 (strengths) and zero in all strengths (concerns). We find that the average social performance changes over time and in most cases not too much deviates from zero. For example, the average performance of employee relations in 1991 and 2012 are equal to 0.02 and 0.16, respectively, while in 2010 the average performance is Except employee relations, we don t find a significant change in the average social performance during the market crises in 2001 and The average performance of employee relations drops from 0.04 to in year 2001 after the technology bubble burst. Table 1 also suggests that there is improvement in the performance of community, environment, human rights, and product characteristics during the past three years. Additionally, Table 1 shows a great variation in corporate social performance among the dimensions considered. For instance, while overall firms have positive performance in community, the average performance in environment has been negative during most the sampled years. Due to the variation in corporate social performance across industries, we adjust for heterogeneity by constructing the industry-adjusted MSCI ESG performance. For each year and each dimension, we compute the average performance for each industry and subtract it from the performance of companies with the same SIC code. 4 We obtain industry-adjusted MSCI ESG performance for sampled firms from 1991 to Then, for each year from 1992 to 2012, we measure changes in corporate social performance by subtracting the industry-adjusted MSCI ESG score in that year from the industry-adjusted score in the previous year and rank the differences in performance from high to low. 4 We follow Statman and Glushkov (2009) by using the SIC-based industry classifications of Moskowitz and Grinblatt (1999). 9

10 We then construct the up (down) portfolio for each dimension and each year by identifying firms that are ranked within the top (bottom) 20 percent of all firms in the performance differences. We construct the long-short portfolio by going long in the up portfolio and short in the down portfolio. Further, we construct the up (down) portfolio of overall social responsible performance by identifying firms that are in at least two up (down) portfolios but not in any of the down (up) portfolios. Since corporate governance reflects the conflicts of interest between managers and stockholders, we also consider overall social performance but excluding the dimension of corporate governance. The up (down) portfolio of overall social performance without corporate governance consist of firms in at least two up (down) portfolios but not in any of the down (up) portfolios while the up and down portfolios of corporate governance are excluded from the construction. ALL and ALL_exG are used to denote the overall social performance with and without corporate governance, respectively. Using annual data from COMPUSTAT, we estimate the mean of total assets, market value, net income, revenue, debt ratio, and return on equity (ROE) of firms in the up and down portfolios for each year during That is, we examine the financial data of the same year when firms are identified being in the up and down portfolios. Table 2 presents the average, standard deviation, and median of the aforementioned mean of financial variables over the sample period from 1992 to We also conduct a paired t-test to see if significant differences exist in these financial variables between firms in the up and down portfolios. The result shows that for most dimensions there is no significant difference in the total assets, market value, sales revenue, financial leverage, and profitability between firms in the up and down portfolios. 5 The data of market value before 1997 is not available from COMPUSTAT, so the statistics of market value are based on the data during the period of

11 However, the up portfolios of diversity and product characteristics show significantly higher ROE than the corresponding down portfolios. The average ROE of the up and down portfolios of diversity (product characteristics) are equal to 0.13 (0.16) and 0.04 (0.09), respectively. The total asset of the up portfolio in diversity is also significantly higher than that of the down portfolio. Besides, we find the up portfolio in human rights has significantly higher net income than the down portfolio, with an average value of $3,349 and $2,073 million, respectively. II. Methodology To examine the relation between corporate goodness and financial performance, for each month in the following year after the construction of up and down portfolios, we use monthly returns obtained from CRSP to compute equally weighted portfolio returns. We are interested in testing if the risk-adjusted return of the up, down, or long-short portfolio is significantly different from zero. If investment in CSR has a positive impact on financial performance, we expect to see the up portfolio shows a positive risk-adjusted return. On the contrary, if engaging in CSR reflects an agency problem between managers and stockholders, a decline in corporate goodness should produce superior financial performance. A superior performance of long-short portfolio suggests that firms with improvement in CSR outperform firms with deterioration in CSR. Unlike previous studies that examine the impact of corporate goodness by comparing the financial performance of firms with different levels of social performance (see, e.g., Statman and Glushkov, 2009), our study compares the performance of firms with changes in corporate social performance. Our new approach helps to shed light on current mixed evidence about the relation between corporate social responsibility and financial performance. 11

12 To estimate the abnormal portfolio return, we run regressions of monthly portfolio returns based on the four-factor model of Carhart (1997). The model is specified by the equation below: R t R f t m f ( Rt Rt ) 2 SMBt 3 HMLt 4 MOM, 1 t t where R represents portfolio return in month t, t f Rt stands for risk-free rate, and m Rt represents the return of the market portfolio. SMB, HML, and MOM denote the size, book-to-market, and momentum factors, respectively. Nofsinger and Varma (2014) show that during periods of market crises, socially responsible mutual funds perform much better than conventional funds and indicate that investors of socially responsible mutual funds are willing to accept lower returns during noncrises in exchange for higher returns during market crises. To examine if the relation between corporate goodness and financial performance changes during non-crisis and crisis periods, we expand previous models by allowing a separate relation between excess portfolio returns and risk factors during non-crisis and crisis periods. The model is specified by the equation below: R t R f t NC NC m f NC NC NC 1 ( Rt Rt ) 2 SMBt 3 HMLt 4 MOM t I t C C m f C C C C ( R R ) SMB HML MOM I, 1 t t 2 t 3 t 4 t t t NC where I ( I C t NC t ) is a dummy variable equal to 1 if month t is within the crisis (non-crisis) periods, and 0 otherwise. Following Varma and Nofsinger (2014), we define the periods of 12

13 March October 2002 and October March 2009 as market crisis periods. 6 The risk free rates, excess market returns, and SMB, HML, and MOM factors are obtained from Kenneth R. French's website. III. Empirical Evidence A) Regression Results For each dimension, we run regressions on the monthly returns of the up and down portfolios. To see if there are differences in the performance between the up and down portfolios, we form a long-short portfolio by going long in the up portfolio and short in the down portfolio. Table 3 presents the regression results of the Carhart four-factor model. For the overall social responsible performance (ALL), the up and down portfolios show a monthly abnormal return of 0.41 and 0.36 percent, both of which are significantly different from zero. The abnormal return of the long-short portfolio is positive, but not significant. As to the overall social performance without corporate governance (ALL_exG), only the up portfolio has significantly positive abnormal return. The evidence is somewhat supporting the argument that improvement in CSR is positively related to superior financial performance. Since ALL and ALL_exG contain several corporate social responsibility dimensions, it is possible that some social responsibility dimensions are positively related to firm performance whereas some other dimensions are negatively related to company values. In the following, we take a closer look at the relation between financial performance and corporate goodness in each dimension by comparing the performance of up, down, and long-short portfolios of each social performance dimension. 6 Varma and Nofsinger (2014) identified the two stock market crisis periods based on S&P 500 index. The first crisis occurred after the technology bubble burst and the second crisis period is related to the global financial crisis. 13

14 Our results suggest that the relation between corporate goodness and financial performance varies with social performance dimensions. Both the up and down portfolios in employee relations have a significantly positive abnormal returns, equal to 0.28% and 0.29% per month, respectively. The long short portfolio has a negative alpha but not significantly different from zero. The evidence that the up has significantly positive abnormal returns is somewhat consistent with the result in Statman and Glushkov (2009), Edmans (2011), and Derwall, Koedijk, and Ter Horst (2011) that investors are rewarded by investing in companies with high employee satisfaction. However, we find that deterioration in employee relations does not necessarily decrease company value. As to environment, the abnormal returns of up and longshort portfolios are positive and statistically significant, while the abnormal return of the down portfolio is not significantly different from zero. The up portfolio has an abnormal return of 0.29% per month and the long-short portfolio shows an abnormal return of 0.32%. The positive relationship between environmental investment and financial performance is consistent with the conclusions in Derwall, Guenster, Bauer, and Koedijk (2005), Kempf and Osthoff (2007), Statman and Glushkov (2009), and Jiao (2010). The portfolios of firms with improvement in community and human rights also have superior performance. The up portfolios in community and human rights have an abnormal return of 0.21% and 0.49%, respectively. The estimated alpha of the long-short portfolio in human rights is 0.98% with statistical significance, while the abnormal return of the long-short portfolio in community is not significantly different from zero. The down portfolio in corporate governance has a positive abnormal return with statistical significance and significantly higher than that of up portfolio. Our result is somewhat different from the evidence of positive impact of corporate governance on financial performance indicated in Bebchuk, Cohen, and Ferrell 14

15 (2009). The discrepancy may be attributed to the different factors considered in our study to measure corporate governance performance. As to product characteristics, both the up and down portfolios have significantly positive abnormal returns with values equal to 0.31% and 0.38%, respectively. The long-short portfolio has a negative abnormal return of -0.07% but not statistically different from zero. It seems to suggest that compared to firms with improvement in product characteristics, firms with deterioration have comparable risk-adjusted returns. Finally, we do not find significant relationship between diversity and financial returns. B) Non-Crisis and Crisis Periods Varma and Nofsinger (2014) compare the performance of socially responsible and conventional mutual funds and find that socially responsible funds outperform conventional funds during the crisis periods but not the non-crisis periods. Their result suggests the relation between corporate goodness and financial returns may change with the market condition. In the following, we relax the constraint by allowing separate effects during the non-crisis and crisis periods. The regression results based on the Carhart four-factor model are presented in Table 4. For the non-crisis and crisis periods, the abnormal returns of the up and down portfolios of ALL and ALL_exG are positive with statistical significance. Interestingly, the abnormal returns of the long-short portfolio of ALL and ALL_exG are positive during the non-crisis period and negative during the crisis period, although not statistically significant. It seems to suggest that improvement is CSR is better perceived by the market during the non-crisis period. We then study each dimension of corporate social responsibility and compare the portfolio returns of up 15

16 and down portfolios. The result in Table 4 suggests that the relationship between corporate goodness and financial performance differs and sometimes flips between the non-crisis and crisis periods. For employee relations, the abnormal returns of the up and down portfolios are 0.34% and 0.33% with statistical significance during the non-crisis periods. During the crisis period, the estimated alphas of the up and down portfolio are significantly positive with values equal to 0.94% and 1.16%, respectively. The abnormal return of the long-short portfolio is positive during the non-crisis period, but turns to be negative during the crisis period. Thus, although firms with improvement in employee relations have superior performance during both the noncrisis and crisis periods, declining in the performance of employee relations can be value enhancing from the stockholders point of view. As to environment, the up and long-short portfolios have significantly positive abnormal returns during the non-crisis periods, with values equal to 0.53% and 0.43%, respectively. During the crisis period, on the contrary, the down portfolio has superior performance. It suggests that improvement in environment is positively related to financial returns only during the non-crisis periods. Both the up and down portfolios in community have significantly positive abnormal returns during the non-crisis periods, and the abnormal return of the long-short portfolio is positive but not significantly different from zero. During crisis period, however, only the abnormal return of the down portfolio is significantly positive and the long-short portfolio has a negative abnormal return of -0.42%. It suggests that investment in community is positively perceived by the market during the non-crisis period, whereas during the crisis period, the portfolio of firms declining in community investment shows better performance. With respect to human rights, during both the non-crisis and crisis periods, only the up portfolio has significantly positive abnormal returns, with values equal to 0.53% and 1.24%, respectively. It suggests that 16

17 improvement in human rights is positively perceived by the market, regardless of the market s condition. For diversity, the risk-adjusted return of the down portfolio is 1.35% during the crisis periods, which is significantly different from zero and significantly higher than that of the up portfolio by 1.23%. On the contrary, during the non-crisis period, the up portfolio outperforms the market on a risk-adjusted basis, but its abnormal return is not significantly different from that of the down portfolio. We find evidence suggesting that the relation between corporate governance and financial performance flips during the crisis and non-crisis periods. During the non-crisis periods, the long-short portfolio underperforms the market on a risk-adjusted basis. During crisis periods, only the up portfolio has positive abnormal returns with statistical significance. Thus, although the portfolio of firms with deterioration in corporate governance has superior performance during the non-crisis periods, firms with improvement in corporate governance outperform the market during the crisis periods. Both the up and down portfolios of product characteristics show significantly positive abnormal returns during the non-crisis and crisis periods. The estimated alphas of the up and down portfolio are 0.31% and 0.64% (1.49% and 0.58%) during the non-crisis (crisis) period, respectively. The abnormal return of the long-short portfolio is significantly positive during the crisis period, but not significantly different from zero during the non-crisis period. The evidence suggests that firms with improvement in product characteristics outperform firms with deterioration during the crisis period. In sum, the market responds differently to firms' engagement in various corporate social activities and its reaction may change between the non-crisis and crisis periods. Both the up and down portfolios of employee relations and product characteristics have superior performance 17

18 during both the non-crisis and crisis periods, but during the crisis period, the up portfolio of product characteristics outperforms the down portfolio. Improvement in human right shows significantly positive abnormal returns, regardless of the market condition. We find firms with improvement in environment, community, and diversity outperform the market during the noncrisis period, whereas firms with deterioration in these dimensions show superior performance during the crisis periods. On the contrary, during the non-crisis periods, the portfolios of firms declining in corporate governance have superior returns, whereas during the crisis periods, firms with improvement perform better. IV. Robustness Checks This section reports results of additional tests that evaluate the robustness of our findings to alternative sample periods, to a different selection criterion, and to the potential problem caused by changes in number of strengths and concerns used in the MSCI ESG database over time. A) Subsample Periods To check if the findings hold under alternative sample periods, we consider the subsample periods ( and ) and conduct a regression analysis on the monthly portfolio returns for each of the subsample periods. 7 We select these two subsample periods because during years MSCI ESG expanded the number of firms covered and thus making the sample less comparable over those years. Tables 5 and 6 present the regression 7 The portfolios are constructed based on changes in social performance in and , respectively. We compare the returns of the up and down portfolios in the subsequent years, i.e., and

19 results of the Carhart four-factor model with the separation of crisis and non-crisis periods for the subsample period and , respectively. The abnormal returns of the up and down portfolios in ALL and ALL_exG are mostly positive and statistically significant during the crisis and non-crisis periods. The results for each social responsible dimension are also qualitatively similar to previous finding when the entire sample period is used. The risk-adjusted returns of the up and down portfolio in employee relations are positive and mostly significant during the non-crisis and crisis periods. As to environment, the up portfolio has significantly positive abnormal returns only during the noncrisis periods in both subsample periods. Consistent with previous finding, portfolio of firms with improvement in human rights shows positive abnormal returns and higher than that of firms with deterioration in human rights regardless of the general market condition. Improvement in community, on the other hand, is related to positive market reaction only during non-crisis periods. During crisis periods, the portfolio of firms with declining in community seems to have better performance. The portfolio of firms with deterioration in diversity shows positive abnormal returns during the crisis periods, and during the subsample period of outperforms the portfolio of firms with improvement in diversity. With respect to corporate governance and product characteristics, the evidence obtained from the subsample periods is qualitatively similar to what we show when the entire sample period is used. The portfolio of firms with improvement in either corporate governance or product characteristics shows higher abnormal returns during the crisis periods than that of firms with deterioration in the respective dimension. However, the pattern is reversed during the non-crisis periods. Overall, the results of subsample periods are similar to what we find when the entire sample period is considered. 19

20 B) Selection Criterion We construct the up (down) portfolio by selecting firms whose rankings of performance differences are within the top (bottom) 20 percent among all firms. To see if our finding is robust to a different selection criterion, we use 30 percent as an alternative cut-off point to construct the up and down portfolios and conduct the regression analysis again. The results are presented in Table 7. Similar to previous finding, both the up and down portfolios of employee relations outperform the market on a risk-adjusted basis. The abnormal returns of the up portfolios in environment, community, and human rights are positive, but only the result of human rights is statistically significant. Consistent with previous findings, the down portfolio in product characteristics has positive abnormal returns with statistical significance and significantly higher than that of the up portfolio. As to corporate governance, the down portfolio has significantly positive abnormal return of 0.24% and higher than that of the down portfolio,. Table 8 reports the regression results of four-factor model with the separation of noncrisis and crisis periods. The abnormal returns of the up and down portfolios of employee relations are positive during the crisis and non-crisis periods, and there is no significant difference between the performance of the up and down portfolios. The up portfolios of environment, community, and diversity have superior performance during the non-crisis periods, whereas the down portfolios in these dimensions perform better during the crisis periods. Consistent with previous finding, the up portfolio of human rights has superior performance regardless of the market condition. As to corporate governance and product characteristics, the finding that the relation between corporate goodness and financial returns is reversed during noncrisis and crisis periods still holds. The portfolios of firms declining in corporate governance or 20

21 product characteristics have superior returns during the non-crisis periods, whereas during the crisis periods, portfolios of firms with improvement in either dimension outperform the market. In sum, our findings are robust when a different selection criterion is used to construct the portfolios. C) Adjust for Factor Changes Although the change in the number of strengths/concerns is modest over time, to assess if the results are robust to changes in the number of strengths/concerns, we construct portfolios as before but exclude monthly returns in those years if their portfolios are constructed in years with a change in numbers of strengths/concerns. The regression analysis based on the four-factor model is presented in Table 9. Consistent with previous results, the up and down portfolios of employee relations have superior performance. Similarly, the portfolio of firms with improvement in community outperforms the market on a risk-adjusted basis. The portfolios of firms with improvement in environment and human right have positive abnormal returns and significantly higher than the abnormal returns of firms with deterioration in these dimensions. For corporate governance and product characteristics, both the up and down portfolios have positive abnormal returns with statistical significance. Consistent with previous finding, the down portfolios seem to perform better than the up portfolio, but the evidence is not significant. Overall, we find similar evidence and the conclusion still holds after the consideration of changes in the number of strengths and concerns over years in the MSCI ESG data. 21

22 V. Conclusions This study examines the relation between corporate goodness and financial performance by comparing the portfolio return of firms with improvement and of firms with deterioration in various social performances. We find the relation between corporate goodness and financial returns varies with corporate social responsible dimensions. Overall, we find evidence somewhat supporting the argument that improvement in CSR is positively related to superior financial performance. The portfolios of firms with improvement in environment, community, and human rights have positive abnormal returns, whereas the up and down portfolios of employee relations and product characteristics outperform the market in the four-factor model of Carhart. We then examine the relationship between corporate goodness and financial returns with the separation of non-crisis and crisis periods. The up and down portfolios of employee relations show significantly positive abnormal returns during the non-crisis and crisis periods, but there is no significant difference between the returns of these two portfolios. The portfolio of firms with improvement in human rights outperforms the market regardless of the market condition, whereas the portfolios of firms with improvement in environment and community show superior performance only during the non-crisis periods. We also find deterioration in diversity is positively related to portfolio return during the crisis periods. For corporate governance and product characteristics, the relation between corporate goodness and financial performance during the non-crisis and crisis periods is reversed. During the non-crisis periods, the portfolios of firms declining in either corporate governance or product characteristics have better performance than that of the up portfolios, whereas firms with improvement in either dimension outperform firms with deterioration during the crisis periods. The findings are robust to 22

23 alternative sample periods, to the cut-off point used to construct portfolios, and to the changes in the number of strengths and concerns used in MSCI ESG database over time. 23

24 Table 1 Summary of MSCI ESG Data This table presents the numbers of strengths (S) and concerns (C) evaluated by MSCI ESG from 1991 to 2012 and the average standardized social performance of firms in the sample. We use the notation GOVˮ, COMˮ, DIVˮ, EMPˮ, ENVˮ, HUMˮ, and PROˮ to represent the dimension of corporate governance, community, diversity, employee relations, environment, human rights, and product characteristics, respectively. Year CGOV COM DIV EMP ENV HUM PRO S/C Average S/C Average S/C Average S/C Average S/C Average S/C Average S/C Average / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / /

25 Table 2 Summary Statistics of Firms in the Up and Down Portfolios This table shows summary statistics of sample firms in the up and down portfolios. For each year during , the mean of total assets, market value, net income, revenue, debt ratio, and return on equity (ROE) in the up/down portfolios is estimated. This table presents the average, standard deviation (S.D.), and median of the aforementioned mean financial variables over the sample period from 1992 to Paired t-test is conducted to examine if significant differences exist in the financial variables between firms in the up and down portfolios. ***, **, and * indicate statistical significance at the 0.01, 0.05, and 0.10 level, respectively. Market Value ($ million) Net Income ($ million) Total Assets ($million) Revenue ($ million) Debt Ratio ROE ALL Down Average 36,242 20, , S.D. 40,377 15, , Median 17,621 15, , Up 36,662 19, , ,932 11, , ,107 16, , ALL_exG Down 38,837 24, , ,191 14, , ,674 18, , Up 34,314 20, , ,306 8, , ,107 18, , EMP Down 27,503 16, , ,191 7, , ,242 13, , Up 26,943 17, , ,612 8, , ,270 18, , ENV Down 29,844 21, , ,529 8, , ,231 21, , Up 23,730 20, , ,374 10, , ,972 19, , COM Down 61,559 31,577 1,446 18, ,868 12,550 1,038 8, ,501 29,300 1,226 16, Up 57,732 32,937 1,391 18, ,446 12,108 1,108 9, ,704 31,114 1,022 17, DIV Down 25,201 17, , * 16,542 11, , ,890 14, , Up 36,999 19, , ,009 11, , ,803 17, ,

26 Table 2 (Continued) Market Value ($ million) Net Income ($ million) Total Assets ($million) Revenue ($ million) Debt Ratio ROE GOV Down Average 41,989 20, , S.D. 46,497 14, , Median 23,550 16, , Up 38,873 22,219 1,020 15, ,503 13, , ,918 23, , HUM Down 130,173 43,645 2,073* 30, ,532 24,922 1,817 25, ,536 38,472 1,534 29, Up 110,002 60,031 3,349 40, ,560 26,299 3,145 30, ,253 55,030 1,870 30, PRO Down 52,129 28,441 1,194 17, * 45,123 9, , ,976 24,782 1,137 18, Up 48,089 22,283 1,116 16, ,160 7, , ,305 22, ,

27 Table 3 Performance of Equally Weighted Returns of Up, Down, and Long-Short Portfolios Based on Carhart Four-Factor Model This table presents results of regressions on the monthly returns of the up, down, and long-short (L-S) portfolios from 1993 to 2013 based on the Carhart four-factor model. For each dimension, the up (down) portfolio is constructed in the previous year with firms whose rankings in performance differences are within the top (bottom) 20 percent among all firms. The long-short portfolio is formed by the strategy that goes long in the up portfolio and short in the down portfolio. Rm and Rf respectively denote the market return and risk-free rate. SMB, HML, and MOM represent the size, book-to-market, and momentum factors. ***, **, and * indicate statistical significance at the 0.01, 0.05, and 0.10 level, respectively. Constant (Rm-Rf) SMB HML MOM R-squared ALL Up *** *** *** *** *** 0.85 Down ** *** *** *** *** 0.84 L-S ** ** 0.06 ALL_exG Up *** *** *** *** *** 0.83 Down *** *** *** *** 0.82 L-S ** ** 0.05 EMP Up ** *** *** *** *** 0.89 Down ** *** *** *** *** 0.90 L-S ** * 0.03 ENV Up * *** *** *** *** 0.79 Down *** * *** *** 0.78 L-S * * 0.02 COM Up * *** ** *** ** 0.87 Down *** *** *** 0.86 L-S ** * *** 0.10 DIV Up *** *** *** *** 0.88 Down *** *** *** *** 0.87 L-S *** *** GOV Up *** *** *** *** 0.83 Down ** *** *** *** *** 0.81 L-S * ** 0.03 HUM Up * *** *** *** 0.62 Down *** *** * 0.55 L-S *** PRO Up ** *** ** *** *** 0.83 Down *** *** *** *** *** 0.84 L-S