Social capital as a hedging tool: The relation between CEO risk-related incentives and corporate social responsibility

Transcription

1 Social capital as a hedging tool: The relation between CEO risk-related incentives and corporate social responsibility Kyumin Cho*, Hyeong Joon Kim**, Seung Hun Han *** School of Business and Technology Management, College of Business, Korea Advanced Institute of Science and Technology, Republic of Korea Address: N22, 291 Korea Advanced Institute of Science and Technology (KAIST), Daehak-ro, Yuseong-gu, Daejeon , Republic of Korea. Tell: Fax: * koc5251@kaist.ac.kr ** creatinghj@kaist.ac.kr *** Corresponding author at: #604, N22, 291 Korea Advanced Institute of Science and Technology (KAIST), Daehak-ro, Yuseong-gu, Daejeon , Republic of Korea. synosia@kaist.ac.kr 1

2 Social capital as a hedging tool: The relation between CEO risk-related incentives and corporate social responsibility Abstract This study examines the relationship between managerial compensation incentives and firm corporate social responsibility (CSR). We follow Hayes et al. (2012) to address potential endogeneity using the implementation of FAS 123R, representing an exogenous change that affects managerial incentives. We find that CEOs with higher CEO pay-performance sensitivity (delta) and inside debt holding (CEO relative leverage) have positive effect to firm's socially responsible activities. In addition, our results show that sensitivity of CEO wealth to stock volatility (vega and total sensitivity) have no impacts on CSR performance. Overall, our empirical evidence suggest that risk averse CEOs prefer less risky corporate policies and long-term investments in social activities, leading to better CSR performance. JEL Classification: G32, G02, M52 Keywords: Corporate social responsibility, Managerial compensation incentive 2

3 1. Introduction Corporate social responsibility (CSR) reflects the extent to which a firm engages in social initiatives in response to a wide range of stakeholder interests (Carroll 1979; Freeman 1983; McWilliams and Siegel 2001). According to Friedman (1970), business is to make as much money as possible while conforming to the basic rules of society, both those embodied in the law and those embodied in ethical custom". This means that CSR activity can affect the firm and the broader society. With regard to CSR and firm value, previous literature found that it is positive for shareholder value (Cheng, Ioannou, and Serafeim 2014; Edmans 2011; Flammer 2015). One particular way CSR might be positive for shareholders is as a hedging device (Godfrey 2005; Godfrey, Merrill, and Hansen 2009; Humphrey, Lee, and Shen 2012; Jo and Na 2012). Extant literature also shows that managerial compensation incentives affect corporate policies and long-term investments, including hedging (Anderson and Core 2017; Cassell et al. 2012; Coles, Daniel, and Naveen 2006; Wei and Yermack 2011). Coles, Daniel, and Naveen (2006) provide empirical evidence of a strong causal relation between the structure of managerial compensation and corresponding incentives and firm s riskier policy choices. The riskier policy choices lead to compensation structures with higher vega and lower delta. Cassell et al. (2012) use CEO relative leverage ratio as an alternative proxy of managerial compensation incentives and finds that CEOs with large inside debt holdings prefer investment and financial policies that are less risky. The present study shifts the attention more squarely to executive compensation structure as an explanation for CSR, showing that CSR initiatives may result from executive s personal investment decisions. Prior research on CSR and executive compensation has considered whether separate components of executive compensation are associated with CSR (Cai, Jo, and Pan 2011; Deckop, Merriman, and Gupta 2006; Mahoney and Thorn 2006). The more firms have a short-term focus on CEO compensation, as measured by the percentage value of bonus in the total pay package, the less 3

4 the firm s CSR (Deckop, Merriman, and Gupta 2006). This suggests that a short-term focus provides the CEO a disincentive to engage in CSR. These studies are limited to examining the amount of executive compensation itself rather than managerial incentives towards risk. The objective of this paper aims to fill this gap by exploring how managerial incentives might affect firm's involvement in CSR. To test our predictions, we follow prior studies (Anderson and Core 2017; Cassell et al. 2012; Coles, Daniel, and Naveen 2006) and managerial incentives such as vega, delta, CEO relative leverage ratio, and total sensitivity. Specifically, delta, vega, and CEO relative leverage ratio are measured based on Coles Naveen Firm s CSR performance is measured by net CSR, CSR index, CSR strength and concern, and adjcsr (Baron, Harjoto, and Jo 2011; Deng, Kang, and Low 2013; Jo and Harjoto 2012; Jo and Na 2012). Using a sample of U.S firm-year observations from KLD, Compustat, Execucomp and CRSP during the period of , our results show that firm s CSR involvement leads to compensation structures with higher delta and CEO relative leverage ratio. In addition, we find that vega and total sensitivity have no impact on the CSR performance. These results remain consistent even after we scale the managerial compensation incentives by CEO s total wealth. To mitigate potential endogeneity concerns, we conduct two-stage-least-square (2SLS) method and natural experiment setting from the implementation of FAS 123R following Hayes, Lemmon, and Qiu (2012). Aligned with the argument, our results remain consistent with the main findings even after controlling endogeneity concerns. Our findings complement previous studies, which find that the components of executive compensation affect firms involvement in CSR (Deckop, Merriman, and Gupta 2006; McGuire, Dow, and Argheyd 2003; Rekker, Benson, and Faff 2014). Since these studies do not address the important question as to whether managerial compensation incentives encourage or discourage managers to invest the level of CSR, our finding offer important policy implications regarding the importance and 4

5 impact of the structure of compensation on the firm s involvement in socially responsible activities. The findings also offer implications from an agency problem perspective. As CEOs with high risk taking incentives underestimate risk, they ignore the importance of the CSR investment, which may resolve conflicts between executives and stakeholders, and could mitigate the negative effects of damaging events on firm value. The remainder of the paper is organized as follows. Section 2 discusses the literature and hypothesis development; Section 3 discusses sample, variable construction and descriptive statistics; Section 4 presents the research design and empirical results; while Section 5 concludes the paper. 2. Related literature and hypothesis development The key relationships in this study that we develop into the testable hypotheses are those between CSR and hedging, managerial incentives and hedging, and CSR and managerial incentives. As risk averse CEOs pursue more hedging, we hypothesize that firms with risk-averse CEOs will engage in more CSR. The literature to develop this argument is considered below CSR and hedging From the 1980s, CSR grew in importance in firm strategy with development of stakeholder theory of the firm (Freeman 1983). Jones (1995) enhances the stakeholder theory by developing model that integrates economic theory and ethics, and find that firms conducting stakeholder s trust commit to ethical behaviors achieve a competitive advantage. Generally, the managers of firms adopting CSR appear to consider the impact of their decisions upon social good and broad stakeholder interest in the expectation that this will flow back as positive attribution or moral capital (Godfrey, Merrill, and Hansen 2009). In this sense, CSR is part of a firm s risk management strategy; it can therefore be regarded as a hedging tool. 5

6 Empirical evidences confirm that CSR performance can play a significant role in hedging a firm s risk (Godfrey 2005; Godfrey, Merrill, and Hansen 2009; Jo and Na 2012; Lins, Servaes, and Tamayo 2017). For example, the positive moral capital among communities and stakeholders from corporate philanthropy can provide shareholders with insurance like protection for a firm s relationship-based intangible assets. In addition, this protection contributes to shareholder wealth (Godfrey 2005). Also, during the financial crisis, high-csr firms experienced higher profitability, growth, and sales per employee relative to low CSR firms, and they raised more debt. In other words, investment in social capital can be thought of as an insurance policy that pays off when investors and the economy at large face a severe crisis of confidence. This evidence suggests that social capital, in addition to financial capital, can be an important determinants of firm performance, which coincided with hedging firm s risk (Lins, Servaes, and Tamayo 2017). Consistent with these arguments, CSR performance can help firms mitigate the risks of reputational loss and secure critical resource from stakeholder providing insurance-like protection when negative event occur (Gardberg and Fombrun 2006; Godfrey, Merrill, and Hansen 2009; Koh, Qian, and Wang 2014). In an event study of 179 negative legal and regulatory actions taken against firms between 1993 and 2003, Godfrey, Merrill, and Hansen (2009) show that firms that engaged in socially responsible activities were more likely to buffer against the threat of negative stakeholder reactions. Among a sample of U.S. public firms between 1991 and 2007, Koh, Qian, and Wang (2014) suggest that good CSR can generate positive firm value, and the value of doing CSR is greater to financially sound firms, firms that do not operate in socially contested industries and firms that are exposed to higher risks of litigation Managerial incentives and hedging 6

7 The link between CEO risk-aversion and hedging stems from the assertion that risk-averse CEOs take less risky projects, and invest more long-term investments to foster firm s sustainability than those of risk-taking CEOs. Since CEO risk preferences are not observable, the literature considers the CEO compensation structure and their portfolio holdings. The stream argues that managers with relatively low risk aversion tend to accept larger proportions of their compensation contingent on performance (i.e., stocks and options) than in assured pay (i.e., salary and bonus). From the perspective of agency theory, managers whose compensation is heavily related to firm performance have more incentives to take less risk than the level of risk diversified shareholders can afford (Jensen and Meckling 1976; Shavell 1979). However, the compensation contingent on performance, (i.e., options), create incentives for managers to take more risk (Smith and Stulz 1985; Smith and Watts 1992). Smith and Stulz (1985) suggest that managerial risk aversion can be affected by the design of compensation contracts. They argue that, given a manager s utility function is concave in expected wealth or firm value; managers could exhibit less risk-averse, risk-neutral or risk-taking behavior through different degrees of convexity in their compensation contracts. Previous literature addresses the relationship between risk and convexity in the compensation contracts using pay-for-performance sensitivity. Increasing pay-for-performance compensation induces managers to reduce overall firm risk in order to reduce their own risk exposure (Garen 1994). Aggarwal and Samwick (1999) test the relationship between the variation of stock return volatility and pay-for-performance sensitivity and show that the latter declines in the level of stock return variance. Coles, Daniel, and Naveen (2006) discuss that equity-based compensation is not exogenously determined and is likely to be correlated with CEOs risk aversion, firm s riskiness and many other features of the contracting environment. Therefore, researchers in this stream use the value or portion of stock options or the characteristics of the stock option compensation as measures of managerial risk aversion. Cohen, Hall, and Viceira (2000) find that there is a statistically significant association 7

8 between increases in executive option holdings and subsequent increases in firm risk. This evidence suggests that option grants lead to greater stock price volatility rather than the reverse. Rogers (2002) uses the observed characteristics of CEO portfolios of stock and option holdings to measure CEO risk preferences/aversion to study how CEO portfolio structure affects corporate derivatives usage. Abdel- Khalik (2007) uses the extent to which compensation choice is made up of stock-based awards as a measure of CEO risk-aversion. That study shows a negative relationship between CEO risk-aversion and the volatility of earnings and operating cash flows, supporting the argument that highly risk-averse CEOs act to reduce volatility. In the recent studies, Sundaram and Yermack (2007) use CEO inside debt holdings (pension benefits and deferred compensation) to proxy for CEO risk-seeking behavior. According to Sundaram and Yermack (2007) s finding, firm s default risk decreases with the level of inside debt, consistent with higher inside debt leading managers to take on lower risk. Also, mangers who hold large CEO relative leverage tend to reduce risks from increasing portfolio diversification and asset liquidity. Consistent with these arguments, Wei and Yermack (2011) examine the disclosure of CEO-firm relative debt-to-equity ratio, following the SEC regulation in 2006, led to bond prices rising significantly and stock prices falling. Cassell et al. (2012) document that CEO relative leverage is negatively associated with future stock returns, R&D expenditures, and financial leverage, which are widely used as proxies for risk-taking activities. Similarly, Anantharaman, Fang, and Gong (2013) find that firms whose CEOs have higher inside debt holdings have lower borrowing rates and fewer debt covenants. Furthermore, large portion of CEO relative leverage promotes high financial reporting quality (He 2015). Therefore, inside debt exposes the CEO to bankruptcy risk just as if the CEO held a piece of the firm s risky unsecured debt CSR and Managerial incentives 8

9 CSR has been linked to a range of corporate decisions leading to long-term investments, specifically executive s compensation structure (Cai, Jo, and Pan 2011; Mahoney and Thorne 2005; Rekker, Benson, and Faff 2014). Cai, Jo, and Pan (2011) indicate that the new CSR-compensation causation can sheds additional light on the issue of how socially responsible firms determine differently their executive compensation compare to that of socially irresponsible firms. As a result, they found that CSR engagement is negatively associated with executive s cash and total compensation. Also, using 90 Canadian firms, Mahoney and Thorne (2005) find that total CSR is positively related to higher levels of longer term compensation and CSR weakness is negatively related to higher longterm compensation. This in turn suggests that executive long-term compensation can be an effective tool in aligning executives welfare with that of the common good, which results in more socially responsible firms. However, very little research has considered the impact of compensation components on the level of CSR undertaken by the firm (Deckop, Merriman, and Gupta 2006; McGuire, Dow, and Argheyd 2003). Deckop, Merriman, and Gupta (2006) provide support for the argument that the way a CEO s pay is structured has an effect on CSR. The more firms have a short-term focus on CEO pay, as measured by the percentage value of bonus in the total pay package, the less the firm s CSR. This suggests that a short-term focus provides the CEO a disincentive to engage in CSR. In other words, investment in CSP is likely to have a limited or possibly negative effect on short-term financial performance, while representing an opportunity cost for scarce corporate resources. While this is important as it increases our understanding of the link between CSR and executive compensation, these studies do not address the important question as to whether risk preference incentives encourage or discourage managers to invest the level of CSR. This is the key contribution of the current study. Based on the literature discussed above, we are investigating the following: As CSR is a hedging tool, and risk averse CEOs hedge more, then, consistent with this, risk 9

10 averse CEOs will do more CSR. 3. Sample selection, variable measurement, empirical design 3.1 Sample selection We collect data from Standard and Poor s ExecuComp database for the sample period from 1994 to Since ExecuComp provides information on inside debt beginning only in 2006 (when the SEC began to require detailed disclosures) our total sensitivity measurement and CEO relative leverage ratio begin in Following Coles, Daniel, and Naveen (2006) and Hayes, Lemmon, and Qiu (2012), we remove financial firms (firms with Standard Industrial Classification (SIC) codes between 6000 and 6999) and utility firms (firms with SIC codes between 4900 and 4999). We merge the ExecuComp data with data from Compustat, CRSP and KLD database based on firm identifier (gvkey) and CEO identifier (co_per_rol) for each fiscal year. 3.2 CSR measures We retrieve data on CSR scores from the Kinder, Lydenberg, and Domoni (KLD) database, which is the most comprehensive and widely used data for CSR research for the sample period. The KLD has exclusionary screens relating to social ratings from involvement with alcohol, gambling, nuclear power, tobacco and the military. Since exclusionary screens only have negative social ratings, we only employ the inclusive screens relating to social ratings, which covers social rating criteria approximately 80 strength and concern ratings in seven major qualitative issue areas. In this research, we conduct three different CSR proxies: CSR, CSRIDX, and adjcsr CSR & CSRIDX 10

11 We include six qualitative issues such as Community Relations, Employee Relations, Environmental Issues, Product Quality, Human Right, and Workplace Diversity. We exclude the Corporate Governance scores in measuring a firm s CSR performance because corporate governance is viewed as a construct distinct from other CSR dimensions (Di Giuli and Kostovetsky 2014; El Ghoul et al. 2011; Kim, Park, and Wier 2012; Lins, Servaes, and Tamayo 2017). Following Cui, Jo, and Na (2016), we assign the value range from -1 to +1 and set zero, if not yet rated to construct net strength and net concern. Our CSR is constructed by the sum of net strength and net concern. Also, we follow prior studies to construct the CSRIDX (Baron, Harjoto, and Jo 2011; Hillman and Keim 2001; Jo and Harjoto 2012; Jo and Na 2012). CSRIDX is CSR scores scaled by sum of maximum strength and concerns in each year (Appendix Table 1 and 2) adjcsr Deng, Kang, and Low (2013) suggest that the summation of raw KLD CSR score from each major dimension scores approach has a serious drawback. Comparing scores across years and dimensions is not possible because the number of strength and concern indicators for most dimensions varies considerably each year (Manescu 2009). To overcome this issue, Deng, Kang, and Low (2013) construct adjusted CSR measure by dividing the strength and concern scores for each dimension by the respective number of strength and concern indicators to derive adjusted strength and concern scores for that dimension and then taking the difference between the adjusted total strength score and the adjusted total concern score. Thus, we expect that the adjusted CSR score give equal weight to the seven dimensions, not to the individual indictors, mitigating any bias caused by an indicator on the social performance of firms in relatively irrelevant industries. 3.3 CEO risk-related incentive measures 11

12 Detailed calculation to measure CEO risk-related incentives are in Appendix A. We use four proxies as CEO risk-related incentives: Delta, Vega, Total sensitivity, and CEO relative leverage CEO relative leverage We construct CEO relative leverage as the CEO s debt-to-equity ratio scaled by the firm s debt-to-equity ratio (Cassell et al. 2012; Edmans and Liu 2010). More details on calculating CEO inside debt, equity holdings, and firm s equity holdings are shown in Appendix A.1. In addition, we construct an indicator variable, CEO relative leverage dummy, set equal to one if CEO to firm debt/equity ratio exceeds one, and zero otherwise because Jensen and Meckling (1976) theorize that the incentive effects of CEO inside debt holdings are likely to be particularly acute when the CEO s debt-to-equity ratio exceeds that of the firm Delta & Vega CEO incentive-compensation features include delta (ratio of CEO total equity value change to a one percent change in share price) and vega (the sensitivity of the value of the CEO s accumulated equity-based compensation to a 1% change in the volatility of stock prices). The level of delta can represent for CEO s risk aversion (Coles, Daniel, and Naveen 2006; Murphy 2012; Ross 2004) while the level of vega can represent CEO s risk-taking incentive (Anderson and Core 2017; Coles, Daniel, and Naveen 2006; Grant, Markarian, and Parbonetti 2009; Guay 1999). More details on calculating delta and vega are shown in Appendix A.2 and A Total sensitivity Anderson and Core (2017) construct overall measure of a manager s risk taking incentives using the total sensitivity of the manager s debt, stock, and option holdings to firm volatility. They calculate each of debt, stock and option sensitivities, and add these sensitivities to construct total sensitivity. Since vega does not capture the sensitivities of stock and debt to volatility, the total 12

13 sensitivity can better reflects CEO incentives than that of vega. More details on calculating total sensitivity is shown in Appendix A Control variables The control variables that we use as determinants of CSR are based on existing literatures. For CEO control variables, we controlled for cash compensation and CEO tenure, which prior literature (Coles, Daniel, and Naveen 2006; Guay 1999) uses as proxies for the CEO s outside wealth and risk aversion. We also controlled CEO s age and gender since the impact of CEO s risk taking incentives on CSR may influenced by CEO s age and gender (Faccio, Marchica, and Mura 2016; Rekker, Benson, and Faff 2014). For controlling firm characteristics, following Anderson and Core (2017) and Hayes, Lemmon, and Qiu (2012), we used firm s size, MTB (market-to-book ratio), surplus cash, sales growth, volatility and modified version of the Altman Z-score (Altman 2000). Cai, Jo, and Pan (2011) and Jo and Harjoto (2011) show that CEO compensation structure is closely related to corporate governance. Therefore, along various specifications, we controlled corporate governance and adjusted corporate governance scores based on the KLD s corporate governance category. Finally, we controlled for industry fixed effect based on Fama-French twelve classification (Fama and French 1993) and year fixed effects. To mitigate the effect of outliers, we winsorize all firm control variables and CEO s cash compensation at first and 99st percentiles. [Insert Table 1] 3.5 Descriptive statistics 13

14 Table 1 presents the descriptive statistics of the key variables used in the main tests. As with the control variables described above, we also winsorize CEO risk-related incentive measures at the first and 99 th percentiles to make sure that our results are not influenced by outliers. The means for CEO relative leverage, delta, vega and total sensitivity are similar to means in the study of Anderson and Core (2017). The variable CSR has a mean of and the means for CSRIDX and adjcsr are and , respectively. The descriptive statistics for the other variables appear to be in reasonable ranges and are compare with those in the prior studies (Bouslah et al. 2017; Cassell et al. 2012). Table 2 reports the univariate analysis of CSR using T-test and Wilcoxon test. In panel A and B, CEOs with higher delta and delta to total wealth do more CSR performance than those with lower, and the difference between delta above median and that below median is statistically significant. Panel C shows that when CEO relative leverage dummy equals to one, CEOs increase CSR performance. [Insert Table 2] 4. Empirical Results 4.1 The results of OLS estimation We examine how the CEO s risk-incentive measures at time t are related to proxies for the level of CSR at time t+1 using regressions of the following form : CSR measure i,t+1 = β 0 + β 1 CEO risk incentive measure i,t + β k Controls i,t + Year dummies k + Industry dummies + ε it 14

15 [Insert Table 3] The table 3 reports our main results, which is a risk averse CEOs do more CSR. Panel A contains unscaled estimation results from regressing CSR, CSRIDX and adjcsr on lagged CEO relative leverage, CEO relative leverage dummy, delta, vega, and total sensitivity. Panel B contains scaled estimation results of Panel A. In panel A, CEO relative leverage and CEO relative leverage dummy have expected positive and significant coefficients in the models for all CSR measures. These results are consistent with the previous findings (Anantharaman, Fang, and Gong 2013; Anderson and Core 2017; Cassell et al. 2012). Our findings suggest that larger the CEO s inside leverage relative to firm leverage is, the stronger incentives to decrease firm risk. To reduce firm risk, CEO with higher CEO relative leverage encourage the firm s CSR policy as well. According to Guay (1999), the mix of vega and delta likely varies substantially across firms. Coles, Daniel, and Naveen (2006) include both vega and delta in the regressions to isolate the effect of each of these incentives on risk taking. Following above literatures, we consider both vega and delta in our regressions to mitigate this concern. However, in panel A, delta has unexpected negative and insignificant coefficients while vega has unexpected positive and significant coefficients in the models for all CSR measures. In addition, the total sensitivity is positively and significantly associated with the CSR measures. These unexpected results from delta, vega, and total sensitivity raise three possible concerns in our main regression. First concern is that we underestimate the effect of CEO s total wealth on CEO s risk taking incentives in line with the theory of Ross (2004). In addition, Guay (1999) finds that the risk-aversion effect is expected to be a function of a manager s total wealth, the degree of 15

16 diversification in a manager s portfolio of wealth, and a manger s utility function. In summary, CEO s total wealth is a critical determinant of a CEO s risk taking incentives because wealthier CEOs will respond less to the same amount dollar amount of incentives if wealthier CEOs are less-risk averse. To mitigate this concern, we compare incentives are relative to wealth, which is calculated by Dittmann and Maug (2007), the more a given risk increase will change the CEO s wealth and the greater the CEO s motivation to increase risk. Thus, the CEO s risk-sensitivity (delta & vega) scaled by total wealth, and CEO relative leverage with total wealth control at time t are related to proxies for the level of CSR at time t+1 using regressions of the following form (Anderson and Core 2017): CSR measure i,t+1 = β 0 + β 1 CEO relative leverage i,t + β 2 Total wealth i,t + β k Controls i,t + Year dummies + Industry dummies + ε it k CSR measure i,t+1 Delta i,t = β 0 + β 1 + β Total wealth 2 i,t + Year dummies + Industry dummies + ε it Vega i,t + β Total wealth 3 Total wealth i,t + β k Controls i,t i,t k CSR measure i,t+1 Delta i,t Total sensitivity i,t = β 0 + β 1 + β Total wealth 2 + β i,t Total wealth 3 Total wealth i,t i,t + β k Controls i,t + Year dummies + Industry dummies + ε it k The panel B of table 3 reports that the result is consistent with the panel A when we consider delta to total wealth and vega to total wealth in our regression. However, when we include delta to total wealth and total sensitivity to total wealth in our main regressions, coefficients of delta to total 16

17 wealth become statistically significant and positive. Also, the positive coefficients of total sensitivity to total wealth s significance level decrease compared to the result of panel A. [Insert Table 4] Second concern may come from decomposition of CSR activities. Prior literature decomposes CSR activities into socially responsible and socially irresponsible activities due to their different implications (Di Giuli and Kostovetsky 2014; Godfrey, Merrill, and Hansen 2009; Mackey, Mackey, and Barney 2007; Tang et al. 2015). Socially responsible activities are usually treated as activities that can weaken shareholder interests and divert valuable resources to society, while socially irresponsible activities are always considered as a cost-saving strategy to achieve short-term financial performance, which destroy stakeholder value. Additionally, firms that conduct more socially responsible activities do not necessarily participate in fewer socially irresponsible activities (Tang et al. 2015). Therefore, we examine the subcomponents of total CSR scores by examining the relationship between managerial incentives and CSR performance from both a socially responsible perspective and a socially irresponsible perspective. Consistent with previous arguments, Table 4 has two panel sets depending on unscaled variables and scaled variables. In case of CEO relative leverage and its dummy, both panel A and B show that these coefficients are only positive and significant in socially responsible activities, not in socially irresponsible activities. In columns (3) and (4) of Table 4, the estimated coefficients of delta and scaled delta in column (3) are significantly negative in socially responsible activities while the estimated coefficients of delta to total wealth in column (4) are positively and significantly associated with the socially responsible activities. In addition, the significance levels of total sensitivity decrease after we consider scaled variables, suggesting that scaled variables with the 17

18 CEO s total wealth can partially capture potential endogeneity such as measurement errors, or simultaneity problem. 4.2 Empirical design to address potential endogeneity issues Our last concern is that managerial incentives might be endogenous even after we control many firm-, manager-specific and CSR governance variables and fixed effects of year and industry. Our regression specification assumes that managerial incentive measures are exogenous. However, this might not be the case if some of the independent variables or unobservable variables that are not controlled for could affect both managerial incentives and CSR measures. Therefore, to mitigate endogeneity concern, we use two following approaches: Natural experiment and Two-stage-leastsquare method (2SLS) Natural experiment: the implementation of FAS 123R changes in accounting standards We conduct exogenous change following prior studies that using the period around the implementation of FAS 123R (Anderson and Core 2017; Chava and Purnanandam 2010; Hayes, Lemmon, and Qiu 2012), to address the concern about risk taking incentives being endogenous. The implementation of FAS 123R changes in accounting standards, which required firms to recognize compensation expense for employee stock options beginning in December Thus, firms responded to this accounting expense by granting fewer options. Our key assumption is that this change, which occurred in response to a change in the accounting rule, is exogenous to the firm s CSR activities. Therefore, we examine the effect of changes in managerial risk taking incentives around the years of rule change on the corresponding changes in CSR. For independent variables, we use changes in the mean value of the variables from 2002 to 2004 and from 2005 to 2008 (Anderson and Core 2017; Hayes, Lemmon, and Qiu 2012). For dependent variables (CSR measures), we use changes in the mean 18

19 value of the variables from 2003 to 2005 and from 2006 to Because this period includes data before inside debt became available in 2006, we can only focus on changes in the vega and delta. In other words, we cannot apply this approach to CEO relative leverage. [Insert Table 5] Table 5 presents the summary statistics for the pre-fas 123R period and post-fas 123R period. For the difference between post-fas 123R period and Pre-FAS 123R period, the results show that mean and median of the variables are significantly different. As Hayes, Lemmon, and Qiu (2012) mentioned, the mean or median of vega and delta are sharply decreased after the implementation of FAS 123R. Therefore, we believe that changing in accounting standards through FAS 123R is an exogenous shock and a suitable natural experiment setting to mitigate endogeneity issues that may arise between managerial incentives and CSR performance. CSR measure i,t+1 = β 0 + β 1 Delta i,t + β 2 Vega i,t + β k Controls i,t + Industry dummies + ε it k CSR measure i,t+1 Delta i,t Vega i,t = β 0 + β 1 ( ) + β Total wealth 2 ( ) + β i,t Total wealth 3 Total wealth i,t i,t + β k Controls i,t + Industry dummies + ε it k [Insert Table 6] 19

20 Table 6 shows the results of these regressions where dependent variables are CSR, CSRIDX and adjcsr. The columns (1) (3) of panel A and B include all controls except industry dummies, while the columns (4) (6) include all the control variables. From the panel A and B of table 6, we find that the change in delta has a positive and significant relation with the change in CSR measures while the change in vega has an insignificant but negative estimated coefficient. This result confirms that the positive coefficients of vega and vega to total wealth on the CSR performance in Table 3 might come from the endogeneity problems. Consequently, the delta and delta to total wealth explain the change in CSR measures around the introduction of option expensing under FAS 123R. This finding suggests that only delta, as a managerial incentive, has a significant and positive impact on the firm s CSR performance. For robustness check, we also test on STR and CON and report the result on Table 7. [Insert Table 7] SLS To address endogeneity for CEO relative leverage, state personal income tax rates as an instrumental variable (Anantharaman, Fang, and Gong 2013; Cassell et al. 2012). CEOs subject to higher state income tax rates have stronger incentives to defer income through pensions and ODC plans because the benefits of deferring income increase with their marginal tax rates. State personal income tax rates are, however, unlikely to directly affect firm s level of CSR, making them a plausible candidate for an IV. We therefore expect CEO relative leverage to be positively associated with the maximum state tax rate on individual income. Furthermore, we consider the industry median CEO to 20

21 firm s debt-to-equity ratio as additional instrument following Cassell et al. (2012). 1st Stage: CEO relative leverage i,t = β 0 + β 1 State tax rate i,t + β 2 CEO relative leverage_industry median i,t + β k Controls i,t + Year dummies + Industry dummies + η it k 2nd Stage: CSR measure i,t+1 = β 0 + β 1 CEO relative leveragei,t + β k Controls i,t k + Year dummies + Industry dummies + ε it Table 8 presents our results from 2SLS estimations of CEO relative leverage. Each of the panel A, B, and C involves different controls. In panel A, we only include CEO controls while the panel B includes CEO controls with few firm controls such as firm size and market to book ratio. In panel C, we include all controls that we used in OLS regression. These panels present results in three different CSR measures using year and industry fixed effects. Even after mitigating endogeneity concerns using two IVs (State tax rate and Industry median of CEO relative leverage), the impact of CSR on CEO relative leverage still remains positive and statistically significant at the 1 to 5% level. [Insert Table 8] 5. Conclusion This study examines the association between managerial compensation incentives for risk taking and CSR performance. Prior research shows that CSR has a hedging feature. CEOs who are risk 21

22 averse take relatively safe policies and long-term investments to diversify the firm s idiosyncratic risk, which leads risk averse CEOs to undertake more hedging. This predicts a positive relationship between CEO s risk aversion and the CSR performance. For measuring managerial risk taking incentive, we compare the results using delta and those obtained with vega, the CEO relative leverage ratio and total sensitivity used in the prior literature. In case of the firm s CSR performance, we conduct three different CSR proxies including CSR, CSRIDX, and adjcsr. Our results are relatively robust to alternative proxies for CSR and managerial compensation incentive and a rage of controls. Using a large sample of U.S firm-year observations during the period , we find that managerial compensation incentives influence the firm s CSR performance. The results reveal a positive association between CEO risk aversion, measured by the delta and he CEO relative leverage ratio, and the CSR performance. We also find that there is a no evidence of a significant relationship between the managerial risk preferences, measured by the vega and the total sensitivity, and the CSR performance. Our results hold up well even when we control for potential endogeneity using the natural experiment setting and IV approach. Our findings complement previous studies, which found that the components of executive compensation affect firms involvement in CSR (Deckop, Merriman, and Gupta 2006; McGuire, Dow, and Argheyd 2003; Rekker, Benson, and Faff 2014). Since these studies do not address the important question as to whether managerial compensation incentives encourage or discourage managers to invest the level of CSR, our finding offer important policy implications regarding the importance and impact of the structure of compensation on the firm s involvement in socially responsible activities. The findings also offer implications from an agency problem perspective. As CEOs with high risk taking incentives underestimate risk, they ignore the importance of the CSR investment, which may resolve conflicts between executives and stakeholders, and could mitigate the negative effects of damaging events on firm value. 22

23 Our contributions to the CSR-compensation literature in two major ways. First, we add to the body of research on the determinants of CSR by taking managerial compensation incentives, measured by vega, delta, CEO relative leverage ratio and the total sensitivity (Anderson and Core 2017; Cassell et al. 2012; Coles, Daniel, and Naveen 2006). Second, our result are particularly relevant for policy decisions, in light of the increased interest by stakeholders, such as investors and policy makers, in the relationship between corporate social responsibility and executive compensation. 23

24 Table 1. Descriptive statistics This table presents the descriptive statistics of sample. Detailed definition of variables are in the Appendix Table 2. The sample consists of U.S. non-financial firm-year observations over the period. Observation Mean Standard deviation Min p25 Median p75 Max STR CON CSR CSRIDX adjcsr CEO relative leverage CEO relative leverage dummy Delta Vega Total sensitivity Delta/Total wealth Vega/Total wealth Total sensitivity/total wealth Total wealth CSR_cgov Size MTB Salegrowth Zscore Cash Volatility Cash compensation Male dummy CEO tenure CEO age

25 Table 2. Univariate analysis of CSR This table presents the average of CSR, CSRIDX and adjcsr. Detailed definition of variables are in the Appendix Table 2. T-test and Wilcoxon test are used for t-statistics and z-statistics, respectively. t-statistics and z-statistics indicate that average of the CSR measures are significantly different. ***, ** and * denote significance at the 1, 5 and 10% level, respectively. Panel A: Delta Total Below Median Above Median T-statistics Z-statistics CSR *** *** CSRIDX *** *** adjcsr *** 8.124*** Observations Panel B: Delta / Total wealth Total Below Median Above Median T-statistics Z-statistics CSR *** *** CSRIDX *** *** adjcsr *** 8.730*** Observations Panel C: CEO relative leverage Total CEO relative leverage dummy = 0 CEO relative leverage dummy = 1 T-statistics Z-statistics CSR *** *** CSRIDX *** *** adjcsr *** 8.083*** Observations

26 Table 3. The effect of CEO risk-related incentives on CSR This table presents the result from OLS regressions where the dependent variables are CSR, CSRIDX and adjcsr. Detailed definition of variables are in the Appendix Table 2. Coefficients of control variables, year, and industry fixed effects are not reported for brevity. t-statistics based on robust standard errors, adjusted for heteroscedasticity and clustered by firm level, are reported in parentheses. ***, ** and * denote significance at the 1, 5 and 10% level, respectively. Panel A: Unscaled variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) CSR CSR CSR CSR CSRIDX CSRIDX CSRIDX CSRIDX adjcsr adjcsr adjcsr adjcsr CEO relative leverage *** *** *** (3.206) (3.201) (3.413) CEO relative leverage dummy *** *** *** (3.151) (3.158) (3.252) Delta (-1.143) (-1.152) (-1.008) (-1.140) (-0.875) (-1.460) Vega *** *** *** (5.813) (5.714) (5.153) Total sensitivity *** *** *** (4.122) (4.143) (3.863) CSR control Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Firm controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes CEO controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Year dummies Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Industry dummies Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 4,702 4,702 8,715 5,590 4,702 4,702 8,715 5,590 4,702 4,702 8,715 5,590 Adj. R-squared Panel B: Scaled variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) CSR CSR CSR CSR CSRIDX CSRIDX CSRIDX CSRIDX adjcsr adjcsr adjcsr adjcsr CEO relative leverage *** *** *** (3.208) (3.203) (3.333) 26

27 CEO relative leverage dummy *** *** *** (3.195) (3.201) (3.211) Delta / Total wealth * * * (0.050) (1.756) (0.106) (1.833) (0.675) (1.690) Vega / Total wealth *** *** *** (4.930) (4.868) (4.276) Total sensitivity / Total wealth ** ** ** (2.384) (2.390) (2.399) Total wealth (-0.849) (-0.858) (1.138) (-0.687) (-0.854) (-0.863) (1.248) (-0.673) (-1.158) (-1.167) (1.222) (-0.815) CSR control Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Firm controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes CEO controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Year dummies Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Industry dummies Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 4,632 4,632 8,589 5,486 4,632 4,632 8,589 5,486 4,632 4,632 8,589 5,486 Adj. R-squared

28 Table 4. The effect of CEO risk-related incentives on CSR strengths and concerns This table presents the result from OLS regressions where the dependent variables are STR and CON. Detailed definition of variables are in the Appendix Table 2. Coefficients of control variables, year, and industry fixed effects are not reported for brevity. t-statistics based on robust standard errors, adjusted for heteroscedasticity and clustered by firm level, are reported in parentheses. ***, ** and * denote significance at the 1, 5 and 10% level, respectively. Panel A: Unscaled variables (1) (2) (3) (4) (5) (6) (7) (8) STR STR STR STR CON CON CON CON CEO relative leverage *** (3.622) (0.138) CEO relative leverage dummy *** (4.204) (0.768) Delta * (-1.767) (-0.628) (-1.090) (1.109) Vega *** (6.115) (-0.397) Total sensitivity *** * (3.524) (-1.684) CSR control Yes Yes Yes Yes Yes Yes Yes Yes Firm controls Yes Yes Yes Yes Yes Yes Yes Yes CEO controls Yes Yes Yes Yes Yes Yes Yes Yes Year dummies Yes Yes Yes Yes Yes Yes Yes Yes Industry dummies Yes Yes Yes Yes Yes Yes Yes Yes Observations 4,702 4,702 8,715 5,590 4,702 4,702 9,125 5,590 Adj. R-squared Panel B: Scaled variables (1) (2) (3) (4) (5) (6) (7) (8) STR STR STR STR CON CON CON CON CEO relative leverage *** (3.817) (0.420) CEO relative leverage dummy *** (4.425) (0.942) 28

29 Delta / Total wealth * * *** (-1.795) (1.867) (-2.725) (-0.092) Vega / Total wealth *** ** (6.794) (2.188) Total sensitivity / Total wealth * (1.745) (-1.239) Total wealth (-0.190) (-0.159) (0.229) (-0.735) (1.223) (1.300) (-1.606) (-0.061) CSR control Yes Yes Yes Yes Yes Yes Yes Yes Firm controls Yes Yes Yes Yes Yes Yes Yes Yes CEO controls Yes Yes Yes Yes Yes Yes Yes Yes Year dummies Yes Yes Yes Yes Yes Yes Yes Yes Industry dummies Yes Yes Yes Yes Yes Yes Yes Yes Observations 4,632 4,632 8,589 5,486 4,632 4,632 8,999 5,486 Adj. R-squared

30 Table 5. Summary statistics for the pre- and post-fas 123R period This table presents the summary statistics for the pre-fas 123R period ( ) and post-fas 123R period ( ). For comparison of STR, CON, CSR, CSRIDX and adjcsr, we take pre-period ( ) and post-period ( ). Detailed definition of variables are in the Appendix Table 2. For Post- Pre-period, t-statistics and z-statistics indicate that mean or median of the variable in the post- and pre-periods are significantly different. For difference between the average of each variable for each firm pre- and post-fas 123R period, t-statistics and z-statistics indicate that each variable is significantly different from zero. T-test and Wilcoxon test are used for t-statistics and z-statistics, respectively. ***, ** and * denote significance at the 1, 5 and 10% level, respectively. Pre-FAS 123R period Post-FAS 123R period Difference between the average of each variable Post- Pre-period ( ) ( ) for each firm pre- and post-fas 123R period N Mean Median N Mean Median T-statistic Z-statistic N Mean T-statistic Z-statistic STR ** *** *** CON ** 2.802*** *** 8.736*** CSR ** *** ** CSRIDX ** ** ** 1.758* adjcsr *** *** Delta *** *** Vega *** *** *** *** Delta / Total wealth ** *** 4.372*** Vega / Total wealth *** *** *** *** Total wealth *** *** CSR_cgov ** *** *** *** adjcsr_cgov *** *** *** 4.016*** Size *** *** *** *** MTB *** *** *** *** Salegrowth *** 3.536*** *** 5.919*** Zscore ** 2.774*** *** 5.291*** Cash *** 2.380** *** 3.411*** Volatility *** 5.609*** Cash compensation *** *** *** *** Male dummy * * CEO tenure ** 8.643*** CEO age *** *** *** *** 30