Corporate Social Responsibility Exposure and Performance of Mutual Funds

Transcription

1 Corporate Social Responsibility Exposure and Performance of Mutual Funds Xi Dong Shu Feng Sitikantha Parida Zhihong Wang * Abstract We study the performance consequences of exposure to corporate social responsibility (CSR) through stock holdings for mutual funds. Using a large sample of U.S. domestic mutual funds, we find that funds overweighting low-csr stocks outperform funds underweighting them by between 1.7% and 2.6% annually. This outperformance, however, reverses during the financial crisis. We also find similar performance patterns among stocks. An equal-weighted high-minus-low CSR stock return spread can explain the CSR-based fund performance spread, whereas a value-weighted spread cannot. These results are consistent with the interpretation that low-csr funds overweight low-csr small-cap stocks that offer high returns to investors who are averse to low-csr investments due to either social norms against low-csr assets and/or risk of underperformance in such assets when overall trust in corporations suffers a negative shock (such as during a financial crisis). JEL classification: G01; G11; G14; G23; M14 Keywords: Corporate Social Responsibility, Mutual Fund Performance, Stock Return * Xi Dong is Assistant Professor of Finance at Baruch College, City University of New York, New York, USA, xi.dong@baruch.cuny.edu. Shu Feng (corresponding author) is Assistant Professor of Finance at Clark University, MA, USA, sfeng@clarku.edu. Sitikantha Parida is Assistant Professor of Finance at Clark University, MA, USA, sparida@clarku.edu. Zhihong Wang is Assistant Professor of Accounting at Clark University, MA, USA, ZhihWang@clarku.edu. We thank Richard Spurgin and seminar participants at the Clark University for valuable comments and discussions. We thank Ying Duan and Yawen Jiao for sharing the list of the funds that explicitly include their SRI objectives in their brochures.

2 Corporate Social Responsibility Exposure and Performance of Mutual Funds ABSTRACT We study the performance consequences of exposure to corporate social responsibility (CSR) through stock holdings for mutual funds. Using a large sample of U.S. domestic mutual funds, we find that funds overweighting low-csr stocks outperform funds underweighting them by between 1.7% and 2.6% annually. This outperformance, however, reverses during the financial crisis. We also find similar performance patterns among stocks. An equal-weighted high-minuslow CSR stock return spread can explain the CSR-based fund performance spread, whereas a value-weighted spread cannot. These results are consistent with the interpretation that low-csr funds overweight low-csr small-cap stocks that offer high returns to investors who are averse to low-csr investments due to either social norms against low-csr assets and/or risk of underperformance in such assets when overall trust in corporations suffers a negative shock (such as during a financial crisis). JEL classification: G01; G11; G14; G23; M14 Keywords: Corporate Social Responsibility, Mutual Fund Performance, Stock Return

3 1. Introduction Socially responsible investment (hereafter, SRI) has become an increasingly important concept in recent years. US SIF Foundation estimates that, as of 2015, more than 22% of money under professional management in the United States, about $8.10 trillion or more, is invested according to various SRI strategies. 1 SRI investors achieve their SRI goals either at a firm level, by engaging in the management of the firms to make firms' operations more aligned with their personal goals of corporate social responsibility (CSR), or at a portfolio level by including (or excluding) CSR securities based on environmental, social, and corporate governance (ESG) criteria. Mutual funds are one of the most important and fast-growing portfolio-level investment vehicles for SRI conscious investors. Existing literature in SRI has mainly studied the performance of dedicated SRI mutual funds, i.e., funds that explicitly indicate their SRI objectives in their prospectus. In general, this strand of literature does not provide consistent evidence on the difference in the performance between socially responsible mutual funds and conventional mutual funds. 2 One of the major limitations of these studies is the size of the sample as the number of dedicated SRI funds is rather small (e.g., only 49 funds in Geczy, Stambaugh and Levin 2005 and 59 funds in Bollen 2007). Given the increasing popularity of SRI concepts in recent years, both mutual fund investors and managers might have consciously or subconsciously taken into consideration social responsibility as one of the factors in their investment decision-making process. Therefore, fund portfolio choices may reflect SRI concepts to various degrees due to either clientele preferences 1 See SIF Foundation s biennial report on 2 E.g., Bauer, Koedijk, and Otten (2005), Goldreyer and Diltz (1999), Hamilton, et al.(1993), Statman, M. (2000), and Renneboog, Ter Horst, and Zhang (2008). 1

4 or managers beliefs. The purpose of this study is to utilize a more inclusive sample, relative to the small sample of SRI funds in previous studies, to investigate the performance effects of the differential CSR exposure across the entire cross-section of U.S. domestic equity mutual funds. We draw our sample from 2003, which was the year the MSCI ESG STATS database increased its CSR measure coverage to include Russell 3000 stocks. This comprehensive coverage makes it possible for us to estimate fund-level CSR measures with high precision for the majority of U.S. domestic equity mutual funds, as most funds can invest in any of these 3000 stocks. We develop a new fund level SRI measure for every mutual fund in our sample to overcome the limitation of relying on only a few dedicated SRI funds in prior studies. This measure is estimated at the beginning of each quarter by value-weighting CSR scores of individual fund holdings obtained from the MSCI ESG database. This approach allows us to broadly capture the levels of funds exposures to the SRI concepts and it may also be effective in scenarios where funds do not invest according to their SRI pledges in their prospectuses. We generate a monthly ranking of mutual funds based on this new measure and classify the top group as high-csr funds (funds that invest more in CSR stocks) and the bottom group as low-csr funds (funds that invest less in CSR stocks). Over , we find that high-csr funds underperform low-csr funds by between 1.7% and 2.6% a year after controlling for various risk factors. These results are both economically large and statistically significant (at the 1% level). We then explore the channels behind these fund return patterns. The CSR literature has provided several theories that suggest low-csr assets should earn high returns. These theories can be summarized into two categories. One category is risk-based. A firm s CSR activities may generate more sustainable financial growth due to their social capital and stakeholder trust (Putnam (1993)). Such trust may pay off with a 'flight to quality' effect during 2

5 periods when there is a negative shock to the overall social trust in the corporations such as during the recent crisis period (see Lins, Servaes, and Tamayo, (2016)), 3 as investors consider high-csr stocks as safer/ better quality stocks. Hong and Kacperczyk (2009) argue that higher CSR stocks may be less risky as they carry less litigation risk. The other category is based on social norm. Hong and Kacperczyk (2009) argue that the social norm against investing in the low-csr stocks drives down these stocks price and increase their expected return. Based on these theories, we argue that the significant abnormal returns (alphas) generated by low-csr funds is because low-csr funds invest disproportionally more in low-csr stocks that deliver abnormally high returns. These abnormal returns either reflect the possibility that conventional factor models fail to fully account for the risk embedded in investing in low-csr stocks, or that there are social norms against investing in low CSR stocks. This interpretation suggests several testable predictions. First, the aforementioned risk-based explanation suggests that low-csr assets outperform high-csr assets on average in the long term, but a 'flight to quality' effect should be observed in certain periods. Second, both risk and social norm explanations suggest low-csr stocks should outperform high-csr stocks. Third, if CSR-based fund performance spread is entirely due to investing in low-csr stocks, then a high-minus-low CSR stock return spread should explain away the alpha of the CSR-based fund return spread. We provide evidence supporting all three predictions. First, following Lins, Servaes, and Tamayo (2016), we use the recent financial crisis as a proxy for periods when social trust experiences a negative shock. We partition our sample period into non-crisis period and crisis period to examine the differential effects of the recent global financial crisis on the performance 3 As also noted by Lins et al. (2016), the notion that the crisis led to a decline in public trust in corporations has also been corroborated by surveys, such as the Edelman Trust Barometer 2009, which shows that 62% of respondents, from a twenty-country survey, had lower trust in corporations in the aftermath of the financial crisis. 3

6 of high-csr and low-csr funds. The results suggest that the underperformance of high-csr funds is primarily driven by the non-crisis period. During the crisis period, high-csr funds reverse their underperformance and significantly outperform low-csr funds by between 2.1% and 4% a year. Next, we rank all the common stocks based on their CSR measures and divide them into tercile groups. The top and the bottom groups are classified as high-csr and low-csr stocks, respectively. We find that there is a significant return spread by about 3.6% a year between equalweighted low- and high-csr stocks over the sample period (Carhart 4-factor alpha). This performance spread is primarily driven by low-csr stocks as they generate a significant positive alpha, whereas high-csr stocks generate insignificant alpha. Similar to the high-csr funds, the high-csr stocks reverse their underperformance during the crisis period and significantly outperform the low-csr stocks. However, we find that although the performance spread between value-weighted low-csr and high-csr stocks is between 2.1% and 2.3% a year, the spread is statistically insignificant. The results imply that it is mainly the small market-cap, low-csr stocks, which receive higher weights in the equal-weighted portfolios, that generate abnormal returns unexplainable by conventional risk factors. Last, we build two CSR-based stock factors. They are the return spreads of the equalweighted and value-weighted low-minus-high CSR stock portfolios. We find that the performance difference between high- and low-csr funds is completely explained by a 5-factor model with equal-weighed CSR stock portfolio return spread as the 5 th factor added to the Carhart four-factor model. In contrast, a 5-factor model with value-weighted CSR stock portfolio return spread as the 4

7 5 th factor does not explain the fund performance difference. Our results are robust to controlling for various fund characteristics and alternative CSR measures. Taken together, these results suggest that low-csr funds outperform high-csr funds because their holdings of low-csr stocks offer high unconditional returns to attract investors, who otherwise would avoid such stocks either due to trust / litigation related risks or social norm concerns. Our evidence is also consistent with the 'flight to quality' effect during the crisis period, as low CSR funds (and stocks) outperform only during non-crisis period. These results support that the risk explanation at least plays a significant role in explaining the CSR-based fund performance difference. Furthermore, the results also suggest that high-csr funds underperform low-csr funds because low-csr funds invest disproportionally more in low-csr stocks with smaller market caps. This is because a value-weighted CSR factor has lower weight on small-cap stocks and hence, it is unable to explain the returns of low-csr funds which overweight low-csr small-cap stocks. Small low-csr stocks play a major role in explaining our fund performance findings, possibly because the risks related to trust and litigations may particularly affect the returns of small-cap low-csr firms, as they have less diversified business to withstand related shocks. Our findings contribute to the literature in several aspects. First, to the best of our knowledge, this is the first academic study that provides evidence on the impact of funds investments in CSR stocks on their performance in a comprehensive cross-section of U.S. domestic equity mutual funds. Earlier works focus on a small sample of dedicated socially responsible mutual funds relative to the traditional mutual funds. They do not find a significant CSR-based fund performance effect, potentially due to their sample limitations (e.g., Bauer et al., 2005; 5

8 Renneboog et al., 2008) or the fact that many funds may not explicitly mention their CSR pledges in their prospectuses. In contrast, we employ a novel methodology to measure the degree of CSR exposure for mutual fund investments. This allows us to study a more comprehensive sample of mutual funds. Another related study, Geczy et al. (2005), compares the optimal fund portfolio from dedicated CSR funds (i.e., a constrained optimization) with the optimal fund portfolio from all other mutual funds. They, however, do not study the cross-sectional difference in performance. Second, predicting mutual fund performance is an important task, and a tremendous amount of resource has been devoted to identifying fund managers with good performance. We add to this literature (e.g. Kacperczyk, Sialm, and Zheng, 2005; Kacperczyk, Sialm, and Zheng, 2008; Cremers and Petajisto, 2009; Amihud and Goyenko, 2013; Dong and Massa 2017; Dong, Feng, and Sadka, 2017) by exploring the effects of the CSR dimensions on fund performance. Third, our findings are also related to previous studies attempting to understand the channels that affect returns of SRI investments. This literature identifies risk (Hong and Kacperczyk (2009), Lins, Servaes, and Tamayo (2016) etc.) and social norm (Hong and Kacperczyk (2009)) as two main channels that can explain a negative relation between CSR and returns. While our findings are consistent with both channels, our results contrasting the financial crisis with non-crisis periods helps support the risk explanation as an important channel. Lastly, we employ a more comprehensive measure to evaluate the relationship between CSR and stock performance. The evidence on this relationship in the literature has been rather mixed so far. One of the approaches in the literature was to build stock portfolios based on various social performance indicators and study if these portfolios produced any superior return. For example, Brammer, Brooks, and Pavelin (2006) use a sample from UK stock market, and Hong and Kacperczyk (2009) focus on an industry based concept of social responsibility, e.g., stocks in 6

9 the sin industries. Both works have found a negative relationship between social performance indicators and stock portfolio performance. Other works, such as Edmans (2011) and Cai and He (2014), found a long-run positive relationship between CSR and stock returns. Each of these papers focuses on one specific aspect (e.g. employee satisfaction) of CSR. 5 In contrast, we focus on a broader concept of CSR that is independent of industries and is not driven by a sub-category of the CSR concept. 6 Our main CSR measure is a comprehensive measure based on five social responsibility criteria: employee relations, community relations, environmental protection, diversity, and products. Our findings add to the literature by providing a fuller picture of the CSR and stock return relationship. Lins et al. (2016) use a similar measure and study the relationship between CSR and stock returns by focusing only on the financial crisis period. 7 They find a significant positive relationship during the crisis. In contrast, our results show a significant negative relationship between CSR and stock returns over a much longer sample period including both the crisis and the non-crisis period. We also differ from their paper in that our main focus is to document a performance effect in mutual funds, whereas they only examine stocks. The rest of the paper is structured as follows: Section 2 describes the data and the summary statistics, Section 3 discusses the empirical analysis and Section 4 concludes. 2. Data and Summary Statistics 5 Kempf and Osthoff (2007) find that portfolios consisting of stocks with higher corporate social responsibility (CSR) ratings have significantly higher alphas among the stocks. However, they use a total of around 650 stocks included in the S&P 500 and the Domini Social 400 (DS 400) over the period and therefore, their findings are not representative of the average stock universe. 6 E.g., even in SIN industries, companies could still obtain high or low CSR scores. 7 As a result, they do not update a firm s CSR measure every year. 7

10 Our sample starts from the MSCI ESG STATS (formerly the KLD database which is originally developed by Kinder, Lydenberg, Domini Research and Analytics.) 8 The database has been widely used in a large number of studies examining the effects of CSR on firm performance (e.g. Hong and Kostovetsky 2012; Krüger 2015; Borisov, Goldma, and Gupta 2016). MSCI ESG STATS provides information on 63 indicators to capture strengths and concerns attributes in seven categories that include community, employee relations, environment, diversity, human rights, products, and governance. It uses a binary reporting of ESG ratings. If a company meets the criteria established for one of the 63 indicators, this indicator is marked with a 1. If a company does not meet the criteria for one indicator, it is marked with a 0. We then sum the strength and concern attributes across each category. Following Kim et al. (2012), we calculate the strengths (concerns) for each stock by adding up strengths (concerns) for each of the five dimensions (community, environmental, employee relations, diversity, and product). Our stock level CSR measure, KLDNET 9, is then determined by subtracting total concerns from total strengths for each stock. Our sample spans from 2003 to We retrieve the performance and holding data of mutual funds from the Center for Research in Security Prices survivorship bias free mutual fund (CRSP MF) database and Thompson Financial CDA/Spectrum holdings database. The CRSP MF database provides information on fund returns, total net assets (TNA), and other fund characteristics. The Thompson Financial CDA/Spectrum holdings database provides stock holdings of mutual funds. We follow Kacperczyk, Sialm and Zheng (2005) to merge the two databases. We focus on open-end US 8 The database was developed by KLD Research and Analytics, and then was acquired by RiskMetrics Group in 2009, which was subsequently acquired by MSCI in We follow previous studies to continue name the stock level CSR measure as KLDNET, which reflects the original name of this database. 10 We choose to start our sample from 2003 because that is the first year the ESG STATS database increases its sample coverage to Russell 3000 companies. The universe of companies covered by ESG STATS varies by time: S&P 500 companies from 1991 to 2000, Russell 1000 companies from 2001 to

11 domestic equity mutual funds and eliminate balanced, bond, money market, international, and sector funds, as well as funds not invested primarily in equity securities. The reported objectives do not always indicate whether a fund portfolio is balanced or not, and hence, we exclude funds that, on an average, hold less than 80% or more than 105% in stocks. We also exclude funds that hold fewer than 10 stocks and which in the previous month managed less than $5 million. If a fund has multiple share classes, we eliminate duplicate funds and compute the fund-level variables by aggregating across different share classes. For asset under management, we sum the TNAs of the different share classes. To construct a CSR measure at the mutual fund level, we calculate the value-weighted portfolio average of KLDNET for the fund. There are 4838 unique funds in our sample. Table 1 reports the mean, the median, the standard deviation, the 25th and the 75th percentile of the Age, Fund CSR Measure, TNA (Total Net Assets), Expense ratio and Annual turnover of all the funds in the sample. Variable definitions are presented in the Appendix. {Insert Table 1 about here} 3. Empirical Analysis 3.1 CSR-Fund Performance Following the method described in section 2, we estimate the fund level CSR measure and rank the funds based on this measure into tercile portfolios (i.e., low-csr, mid-csr, and high-csr) to analyze the effects of CSR on fund performance. In the Robustness Checks section, we also sort funds (stocks) into quintile portfolios. We estimate Model (1) to examine how the mutual funds that invest more in high-csr stocks (i.e., high-csr funds) perform compared to mutual funds that invest more in low-csr stocks (i.e., low-csr funds): 9

12 Returni,t = β0 + β1 High-CSR i,t 1 + β2 Mid-CSRi,t 1 + β3 Xi,t-1 +Time Dummies+ εi,t (1) where Returni, t is the monthly raw return or risk-adjusted return. We use CAPM, Fama-French 3- factor model, and Carhart 4-factor model to estimate risk-adjusted returns. High-CSR and Mid- CSR are indicator variables that take on the value of one if a fund belongs to the top tercile or middle tercile respectively, according to its fund-level CSR measure and zero otherwise. Since funds are sorted into terciles, a dummy for low-csr funds is omitted due to multicollinearity. The coefficients β1 and β2 effectively capture the return difference between highand low-csr funds and between mid- and low-csr funds, respectively. X i,t-1 is a vector of control variables, which include the fund and fund family characteristics employed in Table 2 and are defined in the Appendix. The independent variables have been lagged by one month. All standard errors are double clustered by fund and year and year dummy variables are included. Table 2 reports the results for the whole sample ( ). We observe that the high CSR funds under-perform the low CSR funds by between 14 and 22 bps a month on a risk-adjusted basis. The results are both economically and statistically significant (at 1% level). {Insert table 2 about here} 3.2 CSR-Fund performance and Financial Crisis Next, we investigate how high-csr funds and low-csr funds perform during the recent financial crisis. Following Lins et al. (2013, 2016), we define the crisis period as from Aug 2008 to March 2009 and divide our sample into two subsamples: the crisis period (Aug 2008 March 2009) and 10

13 the non-crisis period (Jan 2003 July 2008 and Apr 2009 Dec 2012). 11 We estimate Model (1) separately for both the non-crisis period and the crisis period. The first four columns of Table 3 report the results for the non-crisis period. We observe that, as in the whole sample, high-csr funds underperform low-csr funds by between 17 and 24 bps a month on a risk-adjusted basis. The last 4 columns in Table 3 reports the results for the crisis period. We find a reversal in the relative performance during the crisis, that high-csr funds now outperform low-csr funds by between 17 and 34 bps a month on a risk-adjusted basis. These results are both economically and statistically significant (at 1% level). Overall, we find that high-csr funds under-perform low- CSR funds during normal periods, whereas they reverse this underperformance and significantly outperform the low-csr funds during the financial crisis periods. 12 {Insert Table 3 about here} 3.3 CSR-Stock Performance In this section, we investigate how high-csr stocks perform compared to low-csr stocks in our sample. Similar to the tests at the fund level, we rank the stocks based on their annual CSR measures and form tercile equally weighted and tercile value-weighted portfolios. We hold these portfolios for the next 12 months, until new CSR measures for the stocks are announced. We use the CAPM, the 3-Factor (Fama-French) and the 4-Factor (Carhart) models to evaluate their performance. Panel A of Table 4 reports the results of the equal-weighted CSR stock portfolios. Between 2003 and 2012, an equal-weighted portfolio of low-csr stocks outperforms an equal-weighted portfolio of high-csr stocks by about 30 bps a month. Panel B of Table 4 reports the results for 11 Lehman Brothers filed for bankruptcy in September 2008 and S&P 500 hit its lowest point of the crisis in March Glushkov and Statman (2016) find somewhat opposite results to that of ours. 11

14 the value-weighted CSR stock portfolios. Results show that a value-weighted portfolio of low- CSR stocks outperformed a value-weighted portfolio of high-csr stocks by about 18.5 bps a month, however, these differences are not statistically significant. We also found that stock results for the crisis period are consistent with Lins et al. (2016), which report that high-csr stocks significantly outperform the low-csr stocks by between 4-7% over the whole crisis period (Aug 2008-March 2009). Taken together, the results of funds in Table 3 and stocks in Table 4 are consistent with a risk-based argument (e.g., Putnam 1993). That is, high-csr is less risky than low-csr investments (either investing in funds or in stocks) in the sense that there is a flight to quality effect during financial crisis (i.e., high CSR investments earn higher returns than low CSR investments); in contrast, during non-crisis periods, low-csr significantly outperform high-csr investments. {Insert Table 4 about here} In addition, Panel A of Table 4 also suggests that the return spread between the high-csr and low-csr equally weighted stock portfolios is primarily generated by low-csr stocks. The risk-adjusted returns from the low-csr stock portfolios are significant, whereas those from high- CSR stock portfolios are not. These results from the equal-weighted portfolios, together with the finding that there is no significant return difference between the value-weighted low- and high- CSR stock portfolios, indicate that high-csr funds under-perform low-csr funds during normal times (i.e., non-crisis period) because low-csr funds overweight low-csr, small-cap stocks. We further test this conjecture in the next section. 3.4 CSR-Fund Performance with CSR Factors 12

15 In this section, we further examine if the return difference between low- and high-csr funds is primarily driven by low-csr funds overweighting on small-cap low-csr stocks. We construct an equal-weighted and a value-weighted CSR factor, calculate the monthly 5-factor CSR adjusted fund returns (the other 4 factors are the factors used in Carhart (1997), i.e., MKT, SMB, and HML from the three-factor model of Fama and French (1993) and a momentum factor) and estimate the following model: Returni, t = β0 + β1 High-CSR i,t 1 + β2 Mid-CSRi,t 1 + β3 Xi,t-1 +Time Dummies+ εi,t (2) Where, Returni, t is the equal-weighted CSR factor adjusted return or the value-weighted CSR factor adjusted return. For the equal- (value-) weighted CSR factor adjusted return, we use the 4 factors from the Carhart (1997) model and construct the fifth factor as returns from a portfolio long on an equal-(value-) weighted portfolio of low-csr stocks and short on an equal- (value-) weighted portfolio of high-csr stocks. The independent variables High-CSR, the Mid-CSR and X i,t-1 are as defined earlier. The independent variables have been lagged by one month. All standard errors are double clustered by fund and year and year dummy variables are included. Table 5 reports the results. We observe that the coefficient on the High-CSR dummy variable for the equal-weighted CSR adjusted return regression is not significant. This implies that the equal-weighted CSR factor completely explains the return difference between high-csr and low-csr funds. However, the coefficient on High-CSR dummy variable for the value-weighted CSR adjusted return regression is still significant, suggesting that high-csr funds underperform low-csr funds by about 12 bps a month, after controlling for the value-weighted CSR factor. These results are consistent with our conjecture that the outperformance of low-csr funds over high-csr funds is due to the fact that low-csr funds overweight on the small-cap low-csr stocks. Our findings are also consistent with previous studies (e.g., Putnam, 1993; Hong and 13

16 Kacperczyk, 2009) arguing that higher CSR investments should earn lower returns unconditionally. {Insert Table 5 about here} 3.5 Robustness Checks In this section we perform some additional analyses to examine the robustness of our results Alternative CSR Measure First, we adopt an alternative CSR measure and re-estimate Model (1). The alternative CSR measure includes two additional criteria, human rights, and governance, provided by the MSCI ESG STATS database. These two additional criteria are excluded by most previous studies in the literature and also from our main analysis. The results, as reported in Table 6, suggest that high- CSR funds underperform low-csr funds by between 12 and 18 bps a month. These results are both economically and statistically significant (at the 5% level) and are very similar to our main findings, suggesting that our analysis is robust to different specifications of the fund CSR measure. {Insert Table 6 about here} CSR-Stock Portfolio Performance Controlling for Sin Factor Hong and Kacperczyk (2009) find that Sin stocks (publicly traded companies involved in producing alcohol, tobacco, and gaming) have higher expected returns than other comparable stocks. Their finding is consistent with the hypothesis that they are neglected by norm-constrained investors and they face greater litigation risk heightened by social norms. In this section, we test if the risk-adjusted return spread between low- and high-csr stocks can be explained by a Sin factor, a factor constructed as returns from a portfolio long on an equal-weighted portfolio of Sin stocks and short on an equal-weighted portfolio of comparable stocks. 14

17 Table 7 reports the results. As shown in the table, the results are similar to our main findings. An equal-weighted portfolio of low-csr stocks outperform an equal-weighted portfolio of high- CSR stocks by about 29 bps a month (significant at 5% level); whereas a value-weighted portfolio of low-csr stocks outperform a value-weighted portfolio of high-csr stocks by about 17 bps a month, which is not statistically significant. These results suggest our main findings are not simply driven by the stocks in the Sin industries. Rather, they suggest that using more comprehensive CSR measures provide economically important incremental information for predicting returns related to the CSR concept. {Insert Table 7 about here} Quintile Fund and Stock Portfolios In addition to the fund tercile portfolios in the main analysis, we sort funds into quintile portfolios in this section. Tables 8 reports the results. The lowest CSR fund category, csr_rank_0, is the reference category that takes the value of one, if a fund belongs to the bottom 20% of the funds according to its CSR measure or zero otherwise. csr_rank_4 is a dummy variable that takes the value of one, if a fund belongs to the top 20% of the funds according to its CSR measure or zero otherwise. The other CSR rank dummy variables are defined accordingly. We find similar results as those reported in the previous sections: the coefficient of the top CSR dummy variables is negative and significant (at the 1% level) during non-crisis periods and positive and significant (at the 1% level) during crisis-periods. {Insert Table 8 about here} Similarly, as a robustness test, we sort stocks into CSR quintile portfolios. Table 9 reports results for the quintile equal-weighted stock portfolios in Panel A and quintile value-weighted stock portfolio in Panel B. Low-CSR (high-csr) is the portfolio of the bottom (top) 20% stocks 15

18 according to their CSR measure. As expected, the risk adjusted alpha of the low minus high CSR decile portfolio return spreads have gone up compared to the tercile portfolios results found earlier. {Insert Table 9 about here} Dedicated SRI Funds We also examine whether our results hold, if we exclude the dedicated SRI funds (funds that explicitly include their SRI objectives in their prospectuses) that have been studied by prior studies. We exclude those funds from our sample and run the same analysis to examine whether our results are driven by those SRI funds. The results of the sample (not tabulated in the paper) without the dedicated SRI funds, are very similar to the results of the whole sample. Hence, we conclude that our results are driven by the CSR stock holdings, not by the dedicated SRI funds. Overall, we find that our results are robust to a number of different model specifications. 4. Conclusion In this paper, we study the relationship between mutual funds exposure to socially responsible investments and fund performance in a broad cross-section of U.S. domestic equity mutual funds. Using data from 2003 to 2013, we find that mutual funds that invest more in stocks rated high in corporate social responsibility measures (high-csr funds) underperform those that invest more in stocks rated low in corporate social responsibility measures (low-csr funds) by between 1.7% and 2.6% a year. This underperformance reverses during the financial crisis period. Correspondingly, we find similar return patterns for high- and low-csr stocks. An equal-weighted high-minus-low CSR stock return spread is able to explain the CSR-based fund performance spread, whereas a value-weighted spread cannot. The results are consistent with the interpretation that low-csr funds overweight low-csr small-cap stocks that offer higher returns as many investors tend to avoid such stocks due to social norms against investing in low-csr stocks or risk 16

19 of underperformance when overall trust in corporations and markets suffers a negative shock (such as during a financial crisis). 17

20 References Amihud, Y., Goyenko R., Mutual fund's R2 as predictor of performance, Review Financial Studies, 26(3), Bauer, R., Koedijk, K., Otten, R., International evidence on ethical mutual fund performance and investment style. Journal of Banking and Financ, 29, Bollen, N.P., Mutual fund attributes and investor behavior. Journal Financce and Quantitative Analysis, 42(03), Borisov, A., Glodman, E., Gupta, N., The corporate value of (Corrupt) lobbying. Review Financial Studies, 29, Brammer, S., Brooks, C., Pavelin, S., Corporate social performance and stock returns: UK evidence from disaggregate measures. Financial Management. 35(3), Cremers, M., Petajisto A., 2009, How active is your fund manager? A new measure that predicts performance, Review Financial Studies, 22, Cai, L., He, C., Corporate Environmental Responsibility and Equity Prices. Journal of Business Ethics, 125, Carhart, M.M., On persistence in mutual fund performance. Journal of Finance, 52(1), Dong, X., Massa, M., Excess Autocorrelation and Mutual Fund Performance. AFA 2014 Philadelphia Meetings Paper. Dong X, Feng, S.,Sadka, R., Liquidity risk and mutual fund performance. AEA 2012 Chicago Meetings Paper. Management Science, Forthcoming. Edmans, A., Does the stock market fully value intangibles? Employee satisfaction and equity prices. Journal of Financial Economics, 101(3),

21 Geczy, C., Stambaugh, R.F., Levin, D., Investing in Socially Responsible Mutual Funds. Working Paper. Statman, M., Glushkov, D., Classifying and Measuring the Performance of Socially Responsible Mutual Funds. Journal of Portfolio Management, 42 (2), Goldreyer, F. Elizabeth, Diltz, J. David, The performance of socially responsible mutual funds: incorporating sociopolitical information in portfolio selection, Managerial Finance, 25 (1), Hong, H., Kacperczyk, M., The price of sin: the effect of social norms on markets. Journal of Financial Economics 93, Hong, H., Kostovetsky, L., Red and blue investing: Values and finance. Journal of Financial Economics 103(1), Kacperczyk, M., Sialm, C. and Zheng, L., On the industry concentration of actively managed equity mutual funds, Journal of Finance, 60, Kacperczyk, M., Sialm C., and Zheng L., 2008, Unobserved actions of equity mutual funds, Review Financial Studies, 21 (6), Kempf, A., P. Osthoff, The effect of socially responsible investing on portfolio performance. European Financial Management, 13(5), Krüger, P., Corporate goodness and shareholder wealth. Journal of Financial Economics, 115(2), Kim, Y., Park, M. S., Wier, B., Is earnings quality associated with corporate social responsibility? Accounting Review, 87(3),

22 Lins, K.V., Servaes, H., Tamayo, A., Social Capital, Trust, and Firm Performance: The Value of Corporate Social Responsibility during the Financial Crisis. Journal of Finance, Forthcoming. Lins, K.V., Volpin, P. Wagner, H.F., Does family control matter? International evidence from the financial crisis. Review Financial Studies, 26(10), Putnam, Robert D., Making Democracy Work: Civic Traditions in Modern Italy, Princeton University Press, Princeton, NJ. Renneboog, L., Ter Horst, J., Zhang, C., The price of ethics and stakeholder governance: the performance of socially responsible mutual funds. Journal of Corporate Finance, 14, Hamilton, S., Jo, H., & Statman, M., Doing Well while Doing Good? The Investment Performance of Socially Responsible Mutual Funds. Financial Analysts Journal, 49(6), Statman, M., Socially Responsible Mutual Funds. Financial Analysts Journal, 56(3),

23 Appendix Variable Definitions Variables MRET Fund CSR Measure Total Net Assets Fund Age Expense Ratio Turnover Ratio Family Net Assets Load Dummy High-CSR Mid-CSR Low-CSR High-CSR Alternative Alternatigve Mid-CSR Alternative Low-CSR Alternative Definitions Monthly fund raw returns Monthly fund level CSR measure estimated from aggregated stock level KLD scores Total net assets of a fund expressed in millions Age of a fund Annual expense ratio of a fund Annual turnover ratio of a fund Total net assets of a fund family expressed in millions Equals to 1 if a fund has a load; 0 otherwise Equals to 1 if a fund (a stock) belongs to the High tercile according to its fund-level (stock level) CSR measure; 0 otherwise Equals to 1 if a fund (a stock) belongs to the middle tercile according to its fund-level (stock level) CSR measure; 0 otherwise Equals to 1 if a fund (a stock) belongs to the bottom tercile according to its fund-level (stock level) CSR measure; 0 otherwise Equals to 1 if a fund belongs to the High tercile according to its alternative fund-level CSR measure; 0 otherwise Equals to 1 if a fund belongs to the middle tercile according to its alternative fund-level CSR measure; 0 otherwise Equals to 1 if a fund belongs to the bottom tercile according to its alternative fund-level CSR measure; 0 otherwise 21

24 Table 1: Summary Statistics This table presents the summary statistics for the sample of domestic equity mutual funds over the period of 2003 to Please refer to Appendix A for variable definitions. Variables Mean Median Std. Dev. 25% 75% Fund Age Fund CSR Measure Expense Ratio Total Net Assets Family Net Assets Turnover Ratio Total Number of Funds 4838

25 Table 2: CSR-Mutual Fund Performance for the Whole Sample from 2003 to 2012 This table reports the results of regressions for the whole sample with risk adjusted fund returns (alphas) as dependent variables. For the CAPM alpha, we use the one-factor Capital Asset Pricing Model; for the 3-Factor alpha, we use the three-factor model of Fama and French (1993), which includes MKT, SMB, and HML; for the 4-Factor alpha, we use the four-factor model of Carhart (1997), which includes MKT, SMB, and HML from the three-factor model of Fama and French (1993) and a momentum factor. The returns are expressed in percent per month. The independent variables have been lagged by a month. Year dummy variables are included and the standard errors are clustered by fund and year. Adjusted R-squared are reported at the bottom. Please refer to the Appendix for variable definitions. The sample includes the CRSP-mutual fund universe for the period from January 2003 to December CAPM Alpha 3-Factor Alpha 4-Factor Alpha Est. (%) t value Est. (%) t value Est. (%) t value Intercept High-CSR Mid-CSR Total Net Assets Load Dummy Expense Ratio Turnover Ratio Fund Age Family Net Assets Year Dummies Yes Yes Yes Adjusted R

26 Table 3: CSR-Mutual Fund Performance: Crisis versus Non-Crisis Periods This table reports the results of regressions with risk adjusted fund returns (alphas) as dependent variables for the crisis and non-crisis periods. We define crisis period as Aug 2008-March 2009 and non-crisis period as Jan July 2008 and Apr Dec For the CAPM alpha, we use the one-factor Capital Asset Pricing Model; for the 3-Factor alpha, we use the three-factor model of Fama and French (1993), which includes MKT, SMB, and HML; for the 4-Factor alpha, we use the four-factor model of Carhart (1997), which includes MKT, SMB, and HML from the three-factor model of Fama and French (1993) and a momentum factor. The returns are expressed in percent per month. The independent variables have been lagged by a month. Year dummy variables are included and the standard errors are clustered by fund and year. The significance levels are denoted by *, **, *** and indicate whether the results are statistically different from zero at the 10-, 5- and 1-percent significance level. Adjusted R-squared are reported at the bottom. Please refer to the Appendix for variable definitions. The sample includes the CRSP-mutual fund universe for the period from January 2003 to December Non-Crisis Period Crisis Period CAPM Alpha 3-Factor Alpha 4-Factor Alpha CAPM Alpha 3-Factor Alpha 4-Factor Alpha Intercept *** *** *** High-CSR *** *** *** *** *** *** Mid-CSR *** *** *** *** *** Total Net Assets Load Dummy ** ** ** ** ** Expense Ratio *** ** ** *** *** Turnover Ratio ** *** * * Fund Age * * Family Net Assets *** *** *** Year Dummies Yes Yes Yes Yes Yes Yes Adjusted R

27 Table 4: CSR-Stock Portfolio Performance for the Whole Stock Sample from 2003 to 2012 This table reports the risk adjusted fund returns (alphas) of the low and high CSR-tercile portfolios, as well as the low minus high CSR portfolio. Stocks are sorted into tercile portfolios on an annual basis based on annual CSR measures. Panel A reports the results for the equal-weighted CSRstock portfolio. Panel B reports the results for the value-weighted CSR-stock portfolio. For the CAPM alpha, we use the one-factor Capital Asset Pricing Model; for the 3-Factor alpha, we use the three-factor model of Fama and French (1993), which includes MKT, SMB, and HML; for the 4-Factor alpha, we use the four-factor model of Carhart (1997), which includes MKT, SMB, and HML from the three-factor model of Fama and French (1993) and a momentum factor. The returns are expressed in percent per month. Low-CSR portfolio contains the bottom tercile stocks according to their CSR measure, High-CSR portfolio contains the top tercile stocks according to their CSR measure and Low-High CSR portfolio is long on Low-CSR and short on High-CSR stocks. The sample includes all common stocks for the period from January 2003 to December Panel A: Equal-weighted CSR-Stock Portfolio Performance CAPM Alpha 3-Factor Alpha 4-Factor Alpha Est. (%) t value Est. (%) t value Est. (%) t value Low-CSR High-CSR Low-High CSR Panel B: Value-weighted CSR-Stock Portfolio Performance CAPM Alpha 3-Factor Alpha 4-Factor Alpha Est. (%) t value Est. (%) t value Est. (%) t value Low-CSR High-CSR Low-High CSR

28 Table 5: CSR-Mutual Fund Performance (with CSR Factors) This table reports the results of regressions for the whole sample with 5-Factor alphas as dependent variables. In the first regression, the 5 th factor is returns from an equally weighted portfolio long on Low CSR stocks and short on High CSR stocks (the other 4 factors are the excess market return, SMB, HML and MOM). In the second regression the 5 th factor is returns from a value weighted portfolio long on Low CSR stocks and short on High CSR stocks. The abnormal returns are expressed in percent per month. The independent variables have been lagged by a month. Year dummy variables are included and standard errors are clustered by fund and year. Adjusted R- squared are reported at the bottom. Please refer to the Appendix for variable definitions. The sample includes the CRSP-mutual fund universe for the period from January 2003 to December Equal-weighted Value-weighted Est. (%) t value Est. (%) t value Intercept High-CSR Mid-CSR Total Net Assets Load Dummy Expense Ratio Turnover Ratio Fund Age Family Net Assets Year Dummies Yes Yes Adjusted R

29 Table 6: CSR-Mutual Fund Performance with Alternative CSR measure This table reports the results of regressions for the whole sample with risk adjusted fund returns (alphas) as dependent variables and an alternative CSR measure as the independent variable in addition to the other control variables. For the CAPM alpha, we use the one-factor Capital Asset Pricing Model; for the 3-Factor alpha, we use the three-factor model of Fama and French (1993), which includes MKT, SMB, and HML; for the 4-Factor alpha, we use the four-factor model of Carhart (1997), which includes MKT, SMB, and HML from the three-factor model of Fama and French (1993) and a momentum factor. The returns are expressed in percent per month. The alternative CSR measure include two additional CSR criteria: human rights and governance, as provided by the MSCI ESG STATS database. The independent variables have been lagged by a month. Year dummy variables are included and the standard errors are clustered by fund and year. Adjusted R-squared are reported at the bottom. Please refer to the Appendix for other variable definitions. The sample includes the CRSP-mutual fund universe for the period from January 2003 to December CAPM Alpha 3-Factor Alpha 4-Factor Alpha Est. (%) t value Est. (%) t value Est. (%) t value Intercept High-CSR Alternative Mid-CSR Alternative Total Net Assets Load Dummy Expense Ratio Turnover Ratio Fund Age Family Net Assets Year Dummies Yes Yes Yes Adjusted R

30 Table 7: CSR-Stock Portfolio Performance Controlling for Sin Factor Stocks are sorted into tercile portfolios on an annual basis based on their annual CSR measures. This table reports the risk adjusted fund returns (alphas) of the low and high CSR-tercile portfolios, as well as the low minus high CSR portfolio. The second and third columns report the results for equal-weighted and value-weighted CSR-stock portfolios respectively. For the alpha, we use a five-factor model with four factors (MKT, SMB, HML and momentum factor) from Carhart (1997) and a fifth factor constructed as returns from a portfolio long on an equal-weighted portfolio of sin stocks and short on an equal-weighted portfolio of comparable stocks. The returns are expressed in percent per month. Low-CSR portfolio contains the bottom tercile stocks according to their CSR measure, High-CSR portfolio contains the top tercile stocks according to their CSR measure and Low-High CSR portfolio is long on Low-CSR and short on High-CSR stocks. The sample includes all common stocks for the period from January 2003 to December Factor EW Alpha 5-Factor VW Alpha Est. (%) t value Est. (%) t value Low-CSR High-CSR Low-High CSR