Corporate Social Responsibility and Insider Trading: The Role of Reputational Costs. Abstract

Transcription

1 Corporate Social Responsibility and Insider Trading: The Role of Reputational Costs Abstract We investigate the constraining effects of reputational concerns on insider trading by examining whether a firm s corporate social responsibility (CSR) orientation affects executives insider trading activities. To the extent that CSR activities increase a firm s reputational capital and therefore the reputational costs of informed trading, executives of CSR conscious firms are likely more constrained from insider trading. Consistent with this, we find that executives of CSR conscious firms make significantly lower profits from insider purchases than executives of non-csr conscious firms. We also find that executives of CSR conscious firms are less likely to trade on future corporate news. Collectively, our results provide evidence suggesting that reputational costs constrain insider trading. Keywords: corporate social responsibility, insider trading, reputation JEL Classification: M14, D82, G10

2 1. Introduction Corporate executives have incentives and disincentives to trade on their insider information. While they can make significant profits from informed trades, they also face various constraints. The literature has so far focused on the constraints imposed by regulations and firm-level policies (e.g., Seyhun, 1992; Bettis et al., 2000; Jagolinzer et al., 2011). The constraining effects of reputational concerns, on the other hand, although often conjectured (e.g., Manne, 1985; Piotroski and Roulstone, 2008), lack empirical support. We fill this gap by investigating whether a firm s reputational capital created by its corporate social responsibility (CSR) activities constrains executives insider trading behavior. There is ample evidence that insider trading brings significant financial gains to corporate executives (e.g., Finnerty, 1976; Ravina and Sapienza, 2010; Jagolinzer et al., 2011). These profitable trading activities, however, are commonly perceived as a result of insiders unfairly exploiting their private information at the expense of uninformed shareholders (e.g., Schotland, 1967). This appearance of impropriety raises significant negative sentiment against insider trading among regulators, media, and the general public. Consequently, firms and executives often experience substantial negative publicity when information-based trading is detected and revealed (Bettis et al., 2000). Such negative publicity and the resulting reputational losses likely have a deterring effect on informed trading (Manne, 1985; Piotroski and Roulstone, 2008). Ceteris paribus, we expect that executives of firms with more reputational capital, thereby having more to lose from negative publicity, to profit less from insider trading. On the other hand, CSR activities, which have become an issue of growing interest for regulators, investors, and other stakeholders in recent years, are widely 1

3 viewed as a firm s strategic investment in reputation building and maintenance (Fombrun and Shanley, 1990; McWilliams et al., 2006; Linthicum et al., 2010). A recent survey of Fortune 1000 executives reveals that reputation is the primary motivation behind CSR initiatives (Adam Friedman Associates, 2012). The literature also suggests that CSR activities enhance a firm s reputation by raising brand value, customer awareness, and investor appreciation (Brown and Dacin, 1997; Servaes and Tamayo, 2012; Rosen et al., 1991). To the extent that CSR conscious firms have more reputational capital at stake and therefore higher reputational costs associated with insider trading than non-csr conscious firms, we expect that their executives are more constrained from profiting from informed trades. We measure a firm s CSR orientation by examining the social ratings data issued by MSCI ESG STATS (MSCI hereafter; previously known as KLD) from 1991 to Similar to the extant literature (e.g., Hong and Kostovetsky, 2012; Kim et al. 2012), we assess a firm s CSR performance with a summary CSR score that reflects various aspects of social responsibility, including environment, community, employee relations, diversity, customers, and involvement in controversial industries. Following the prior literature (e.g., Piotroski and Roulstone, 2008; Huddart and Ke, 2007; Jagolinzer et al., 2011), we proxy for executives insider trading profits using the average daily abnormal returns from the Fama-French four-factor model estimated over the 180 days following each trade. We examine the association between a firm s CSR orientation and its executives subsequent insider trading profits. Consistent with the prior literature (Aboody and Lev, 2000; Huddar and Ke, 2007; Jagolinzer et al., 2011), we find that executives make profits from purchase transactions 2

4 but not from sales. We further find that, for CSR conscious firms, i.e., firms with more CSR strengths than concerns in the prior year, executives trading profits from purchases are significantly lower than for non-csr conscious firms. The difference in trading profits persists after we control for firms book-to-market ratio, size, a variety of proxies of information asymmetry, and insider trading restrictions. Exploring various dimensions of the aggregate CSR score, we find that environmental consciousness is a dominant factor that constrains executives insider trading profits. This is consistent with environmental initiatives receiving the most attention from regulators and the general public (PricewaterhhouseCoopers, 2010; Adam Friedman Associates, 2012) and thus likely contributing more to a firm s reputational capital compared to the other dimensions of CSR. To reinforce our inference that CSR constrains executives insider trading profits, we conduct several inter-temporal and cross-sectional tests. First, we examine whether a firm s adoption of new CSR initiatives is associated a subsequent decline of executives insider trading profits. Our analysis indicates that, after a firm turns from non-csr conscious to CSR conscious, its executives insider trading profits decrease significantly. Second, we examine trading profits of newly appointed executives from outside the company. These new executives do not create the firm s CSR orientation but are likely constrained by it. Consistent with our primary findings, these new executives of CSR conscious firms also make lower insider trading profits than their counterparts of non-csr conscious firms, supporting our main inference that insider trading is constrained by CSR. 3

5 Third, we investigate whether the constraining effects of CSR on insider trading profits are more pronounced absent effective alternative constraining mechanisms. Prior research suggests that strong corporate governance limits insider trading profits (Ravina and Sapienza, 2010; Jagolinzer et al., 2011). We thus expect the constraining effects of CSR to be more evident in firms with weak governance. Consistent with this, our analysis shows that executives of CSR conscious firms make lower insider trading profits than executives of non-csr conscious firms only in the subsample of weakly governed firms. We perform a number of robustness tests. First, a firm s CSR orientation could be endogenously determined by the characteristics of the firm. To address this issue, we use the propensity score matching technique and match each CSR conscious firm-year with a non-csr conscious firm-year with the closest propensity score. Our matching model accounts for all explanatory variables in the trading profit regression and other potential determinants of CSR, such as free cash flows and institutional ownership. We continue to find that executives of CSR conscious firms make significantly lower profits from insider trading than executives of matched firms. Second, we include additional controls for potentially omitted correlated variables. We control for media coverage, measured by the average daily number of news articles covering each firm in the year of the trade, as mass media can be a source of information that reduces insider trading profits (Frankel and Li, 2004) and a source of pressure that promotes CSR practices (Zyglidopoulos et al., 2012). We also control for top executives total compensation since prior research suggests that insider trading can be a form of compensation (e.g., Manne, 1966, 2011). Our inferences are robust to including the additional controls. 4

6 Third, we expect the constraining effects of CSR to be less prominent on lower level managers than on executives, as lower level managers informed trades are likely less damaging to a firm s reputation given their lower visibility. We find that, while CSR constrains executives trading profits, it does not constrain the trading profits of lower level managers. Finally, we examine whether a firm s CSR orientation affects the likelihood of executives engaging in insider trading prior to future corporate news. Consistent with insider trades being driven by superior information, we find that, for non-csr conscious firms, executives are more likely to buy prior to good news and/or sell prior to bad news. In contrast, there is no such evidence for executives of CSR conscious firms, suggesting that these executives are constrained by the CSR orientation of their firms and therefore are less likely to trade on insider information. Our paper contributes to the literature in at least two ways. First, we add to the insider trading literature by examining the role of reputational concerns on insider trading. While a number of studies have investigated how regulations and firm-specific policies constrain insider trading, there lacks empirical support for the often conjectured constraining effects of reputational costs. Using CSR orientation as a proxy for reputational costs, we provide evidence on the constraining role of reputational concerns. Second, we contribute to the CSR literature by suggesting a new channel through which CSR initiatives might affect managerial behavior. We find that a firm s CSR orientation could act as an economic constraint on potentially opportunistic managerial decisions, such as informed insider trading, via its connection with the firm s reputational capital. 5

7 Some studies suggest that executives of CSR conscious firms are ethical (e.g., Kim et al., 2012). If so, they will likely restrain from trading on private information. However, we believe our results are better explained by the economic constraint perspective of CSR than ethics. Executives of CSR conscious firms, just like executives of non-csr conscious firms, make economically and statistically significant profits from insider trading, inconsistent with these executives being ethical. The remainder of this paper is organized as follows: Section 2 discusses related studies, hypothesis, and research design. Section 3 describes the sample and main analyses. Section 4 reports additional tests and Section 5 concludes. 2. Related Studies, Hypothesis Development and Research Design 2.1. Related studies and hypothesis development Corporate insiders have access to a continuum of material non-public information about their own firms. Prior studies suggest that insiders trade on (e.g., Frankel and Li, 2004; Piotroski and Roulstone, 2005) and profit from (e.g., Aboody and Lev, 2000; Skaife et al., 2012) their superior knowledge about the firms prospects. 1 Rozeff and Zaman (1988) point out that insider trading profits persist even after taking into account transaction costs. While insiders can extract profits from informed trading, they also face a number of constraints. Regulators impose penalties on insider trades based material non-public 1 Under the semi-strong form of market efficiency, only trades based on private information can earn profits. Therefore, these persistent trading profits suggest that insiders trade on their private information. 6

8 information. 2 However, Seyhun (1992) shows that enactment of tighter regulations does not seem to reduce insider trading profits. Piotroski and Roulstone (2005), on the other hand, find evidence that legal costs do constrain insider trading in the sense that insiders attempt to reduce the risk of investigation by avoiding trades immediately prior to major news events. Firms also voluntarily adopt policies to restrict insider trades. Bettis et al. (2000) document that the specification of blackout periods is effective in restricting insider trades. Jagolinzer et al. (2011) further show that a more effective approach is to require approval from the general counsel before insider transactions. A relatively unexplored constraint of insider trading is a firm s reputational concerns. Profitable insider trading is associated with a general perception that insiders exploit uninformed shareholders for personal financial gains. 3 Schotland (1967) claims that, since such gains are unfair, unacceptable wrong is inflicted by insiders informed trading. This perception of unfairness raises significant negative sentiment against insider trading. The media widely embrace the view that profit from insider trading is reward without risk and wealth generated with injury done to others. 4 Consequently, when insider trading is detected and revealed, firms often experience substantial negative publicity that results in reputational losses (Bettis et al., 2000). For example, in the case of Diamond v. Oreamuno, the court noted that informed trading can cast a cloud on the 2 The Securities Exchange Act of 1934 requires insiders to forfeit (to their company) the profit realized from round-trip transactions (e.g., a buy followed by a sale) made within six months of each other. The Insider Trading Sanctions Act of 1984 increases the maximum civil penalty for illegal insider trading to three times the amount of profit gained or losses avoided, and imposed a maximum criminal penalty of $100,000 and a maximum jail sentence of 5 years. The Securities Fraud Enforcement Act of 1988 further increases the maximum criminal penalties to $1,000,000 and the maximum jail sentence to 10 years. 3 Our paper is silent on whether insider trading enhances or harms economic efficiency. The literature is inconclusive on this topic (e.g., Manne 1996; Ausubel, 1990; Fishman and Hagerty, 1992). 4 See, for example, the article by Pulitzer-Prize-winner George Will, Keep Your Eye on Guiliani, on Newsweek, March 2,

9 corporation s name, injure stockholder relations and undermine public regard for the corporation s securities. CSR, on the other hand, is widely considered a strategic investment in reputation building and maintenance (Fombrun and Shanley, 1990; McWilliams et al., 2006; Linthicum et al., 2010). McWilliams and Siegel (2001) present a theory of the firm perspective of CSR, stating that firms accrue reputation capital through their commitments to the social good. Adam Friedman Associates (2012) survey CSR executives at Fortune 1000 firms and identify reputation as the primary motivation behind CSR initiatives. Consistent with CSR being effective at enhancing reputation, the literature shows that CSR conscious firms experience increased brand value, more customer awareness, and greater investor appreciation (Brown and Dacin, 1997; Servaes and Tamayo, 2012; Rosen et al., 1991). With more reputation capital at stake, CSR conscious firms face higher potential reputational losses from informed trading. Such expected costs, ex ante, will constrain executives from insider trading. The constraining effects could come through multiple channels. For example, CSR conscious firms likely entertain a corporate culture that values reputation. This culture would discourage individual behavior that can damage firm reputation, such as informed insider trading. 5 Furthermore, managerial wealth is tied to the value of the firm directly through the ownership of stock or stock options and indirectly through salaries and bonuses based on various financial and nonfinancial performance measures. Thus, reputational losses of the firm can translate into personal 5 Organizational research suggests that corporate culture is a social control mechanism that influences individual behavior by encouraging conforming activities and discourages nonconforming activities (O Reilly and Chatman, 1996; Sunstein 1996; Sørensen, 2002). 8

10 costs of insider trading and thereby provide disincentives for informed trading. 6 In addition, CSR conscious firms might adopt explicit corporate policies to constrain insider trading and protect their reputation. 7 In summary, we expect that executives of CSR conscious firms are more constrained from insider trading due to higher reputational costs and therefore would make lower trading profits than executives of non-csr conscious firms. We thus form the following hypothesis: H1: Executives of CSR conscious firms make lower insider trading profits than executives of non-csr conscious firms Research Design Measure of CSR orientation We measure firms CSR orientations with the CSR ratings issued by MSCI, which uses a combination of surveys, financial statements, and articles in the popular press and academic journals, as well as government reports, to assess firms CSR performance with approximately 100 indicators. It examines strengths and concerns in each of following seven categories: environment, community, human rights, employee relations, diversity, customers, and governance. Additional concerns are noted for 6 Informed insider trading may also harm executives personal reputation. If a firm s CSR activities increase the personal reputation of executives who make CSR decisions, the personal reputation costs of their insider trading would increase, thereby constraining their trading. While this may be possible and is consistent with reputation constraining insider trading, we are not aware of any literature support regarding the connection between a firm s CSR orientation and its executives personal reputation. 7 We do not find our measure of CSR consciousness and our proxy of the existence of firm-level black-out period restrictions on insider trading to be positively correlated. While formal insider trading policies might be a channel through which CSR conscious firms regulate insider trading, it is possible that the implicit constraining effects of corporate culture and compensation schemes reduce the benefits and therefore the likelihood of adopting costly explicit policies. 9

11 activities in the following controversial industries: alcohol, gambling, tobacco, firearms, military, and nuclear power. MSCI data have been used extensively in various disciplines of business academic research to measure the construct of CSR performance (e.g., Turban and Greening, 1997; Berman et al., 1999; Johnson and Greening, 1999; Hong and Kostovetsky, 2012; Kim et al., 2012; Servaes and Tamayo, 2012). Chatterji et al. (2009) commend MSCI s social ratings as one of the most influential and the most widely accepted CSR measures used by academics. Szwajkowski and Figlewicz (1999) evaluate the validity and reliability of the MSCI database, and conclude that their ratings have substantial and discernible validity with especially strong internal discriminant validity. Waddock (2003) further assert that the MSCI data are the de facto research standard for measuring CSR in scholarly research. Similar to prior studies (e.g., Kim et al., 2012), we construct a CSR score, measured as the total number of strengths minus the total number of concerns in all MSCI s rating categories excluding the human rights dimension and the corporate governance dimension. 8 We identify a CSR conscious firm as a firm with a positive CSR score in the prior year Measure of insider trading profits 8 Our results are robust to including the human rights dimension and the corporate governance dimension, excluding the concern indicator variables for involvement in controversial industries, and excluding observations with a CSR score of zero. 9 Our inferences are unchanged if we also classify firms with a CSR score of zero as CSR conscious firms, or if we use the continuous CSR score to measure CSR consciousness. 10

12 Following the extant literature (e.g., Frankel and Li, 2004; Huddart and Ke, 2007; Jagolinzer et al., 2011), we measure executives insider trading profits as the average daily abnormal return over the six months following each trade. 10 The trading profits are estimated using the following transaction-specific regression of daily returns on four common factors: R i R f = α + β 1 (R mkt R f ) + β 2 SMB + β 3 HML+ β 4 UMD + ε R i is the daily return of firm i s equity, R f is the daily risk-free interest rate, and R mkt is the CRSP value-weighted market return. SMB, HML, and UMD are the size, book-to-market, and momentum factors. Trading profits are equal to α (-α) for purchases (sales). We examine differences in executive s insider trading profits between CSR conscious firms and non-csr conscious firms. If CSR constrains insider trading, we would find that executives of CSR conscious firms make lower trading profits than executives of non- CSR conscious firms. 3. Main analyses 3.1. Sample and descriptive statistics We gather the CSR ratings data from MSCI, a database that covers about 650 largest U.S. companies from 1991 to 2000, about 1,100 largest U.S. companies from 2001 to 2002, and about 3,000 largest U.S. companies since We obtain insider trading information from the Thomson Financial Insider Filing Data (TFN), which contains insider trading activity reported on SEC Forms 3, 4, and 5. We focus on trades 10 Defining trading profits alternatively as the dollar amount of profits, i.e., abnormal returns multiplied by the value of the trades, yields qualitatively similar results. 11 Our results are robust to using only the sample period 2003 and beyond. 11

13 of common shares by top executives including CEO, CFO, CIO, COO, CTO, and Presidents. Following the literature (e.g., Frankel and Li 2004), we include only openmarket transactions and require a minimum transaction price of $2 a share and 100 shares per transaction. We delete transactions with more shares than the CRSP daily volume and prices outside the bid-ask price range. After merging with the Compustat database to obtain control variables, our final sample includes 56,966 insider trading transactions corporate executives made from 1992 to Table 1 reports descriptive statistics. The mean CSR score is and the median is 0. About 28% of our sample have a positive CSR score and are therefore classified as CSR conscious firms. The summary statistics for control variables used in the main analysis are also reported. Variable definitions are summarized in the appendix. Not surprisingly, our sample firms are large, due to the composition of the MSCI database Univariate and regression analyses Panel A of Table 2 reports the univariate analysis of executives insider trading profits in CSR conscious firms vs. non-csr conscious firms. We present the trading profits following each trade across various windows, i.e. 180 days, 120 days, and 90 days. Since our results are similar using different windows, we focus on the 180-day window in our discussions. Following prior studies (Frankel and Li, 2004; Brochet, 2010; Jagolinzer et al., 2011), we report insider purchases and sales separately. The mean daily trading profits from executive purchases is % for CSR conscious firms. In contrast, for non-csr conscious firms, the mean daily trading profits from executive 12

14 purchases is %, over 40 percent more than for CSR conscious firms. The difference is significant, both statistically and economically. The comparison of the median profits from executive purchases between CSR conscious firms and non-csr conscious firms exhibits the same pattern. These results suggest that executives of CSR conscious firms make lower insider trading profits than executives of non-csr conscious firms. Consistent with the prior literature (Aboody and Lev, 2000; Huddart and Ke, 2007; Jagolinzer et al., 2011), we find that insiders do not profit from sale transactions. The average trading profits from their sales is negative for both CSR conscious firms and non-csr conscious firms ( % and %, respectively). 12 This is consistent with insider sales being driven by multiple factors, some unrelated to information. For example, executives may sell shares due to the need for diversification or portfolio rebalancing following option grants (Ofek and Yermack, 2000). In addition, insider sales are more likely to trigger litigation if followed by a plunge in stock price, a potential deterrent for insiders to profit from selling on private information (Cheng and Lo, 2006; Chen et al., 2012). As insider trading profits are more likely to create an appearance of impropriety, in the rest of the paper, we focus on purchases only, similar to Ravina and Sapienza (2010). Panel B of Table 2 reports the insider trading profit regression results based on the following model: Trading Profit t = α + β 1 CSR t-1 + β 2 BTM t-1 + β 3 LogMV t-1 + β 4 R&D t-1 + β 5 Loss t-1 + β 6 Volatility t-1 + β 7 LogAnalyst t-1 + β 8 Turnover t-1 + β 9 Restrict t-1 + ε t (1) 12 The difference is %, statistically significant at the 10% level, which is consistent with our hypothesis. However, we acknowledge that this difference is likely not economically significant. 13

15 The variable Trading Profit t is the estimated average daily trading profits over 180 days following each trade. The CSR t-1 indicator variable is equal to one for CSR conscious firms, i.e., firms with a positive aggregate CSR score in the year prior to the trade. Following Lakonishok and Lee (2001), we control for firm risk with two proxies, the book-to-market ratio (BTM t-1 ) and firm size (LogMV t-1 ). We also control for information asymmetry following prior research. Aboody and Lev (2000) document higher insider trading profits in firms with R&D activities. We define R&D t-1 as a dummy variable equal to one if a firm has positive R&D expenses. Following Huddart and Ke (2007) and Brochet (2010), we also include a loss dummy (Loss t-1 ). As in Ravina and Sapienza (2010), Volatility t-1 is computed as the variance of daily stock returns over the interval (-380, -20) before each trade. We also follow Frankel and Li (2004) to include analyst following (LogAnalyst t-1 ). We add share turnover (Turnover t-1 ) to control for differences in the intensity of investor interest between CSR conscious and non-csr conscious firms. Bettis et al. (2000) find that insider trading restrictions with blackout periods reduce insider trading profits. We follow Roulstone (2003) and Brochet (2010) and control for Restrict t-1, a dummy variable set to one if 75% or more of trades in a fiscal year occur in a 30-day window following an earnings announcement. 13 Following Ravina and Sapienza (2010) and Cao et al. (2011), we cluster standard errors by executive Bettis et al. (2000) collect data on blackout periods from a survey. Roulstone (2003) and Brochet (2010) construct a proxy for blackout periods based on publicly available data on insiders trading patterns and we follow their approach to control for blackout periods. 14 Our results are robust to clustering standard errors by executive-year or by firm, and controlling for industry and year fixed effects. 14

16 Panel B reports the results of model (1) for purchases. Consistent with the univariate analysis, our results are similar across different trading windows. Therefore, we focus on the 180-day window in our discussions here and in subsequent analyses following prior studies (Seyhun, 1998; Huddart and Ke, 2007; Brochet, 2010; Jagolinzer et al., 2011). We find a negative and significant coefficient on CSR t-1 ( , p=0.0074), suggesting that executives of CSR conscious firms make significantly lower profits in insider purchases than executives of non-csr conscious firms, consistent with our hypothesis. In an untabulated analysis, we also estimate model (1) for sale transactions across different trading windows. The coefficient on CSR t-1 is insignificant at conventional levels. Thus, there is no evidence that trading profits (losses) from insider sales differ between CSR conscious and non-csr conscious firms Dimensions of CSR Our CSR score is a composite index summarizing various dimensions of social responsibility. Next, we explore which aspect drives the relation between a firm s CSR orientation and executives insider trading profits. We follow the prior literature (Turban and Greening, 1997; Kim et al., 2012) and examine each of the following five dimensions of CSR: environment, community, employee relations, diversity, and product. 15 Environmental issues tend to attract more attention from regulators and the general public. Adam Friedman Associates (2012) suggests that environmental issues are a top focus for CSR executives, dominating all the other dimensions of CSR. A similar view is 15 We do not examine the controversial industries as a separate dimension because they are noted for concerns only. 15

17 expressed in another survey by PricewaterhhouseCoopers (2010), where over 90% of surveyed companies had a dedicated discussion about the environment in their CSR reports. Since investing in environmental projects is more likely to effectively accrue reputational capital for the firm, environmental consciousness is more likely to constrain executives insider trading profits than the other dimensions. Similar to the definition of CSR t-1, we construct an indicator variable for each of the following five dimensions, CSR_Environment t-1, CSR_Community t-1, CSR_Employee Relations t-1, CSR_Diversity t-1, and CSR_Product t-1. These indicator variables are equal to one if the CSR score in the corresponding dimension is positive, i.e., the total number of strengths is greater than the total number of concerns in that dimension. We examine the relation between these CSR dimensions and executives trading profits using the following regression model: Trading Profit t = α + β 1 CSR_Environment t-1 + β 2 CSR_Community t-1 + β 3 CSR_Employee Relations t-1 + β 4 CSR_Diversity t-1 + β 5 CSR_Product t-1 + β 6 BTM t-1 + β 7 LogMV t-1 + β 8 R&D t-1 + β 9 Loss t-1 + β 10 Volatility t-1 + β 11 LogAnalyst t-1 + β 12 Turnover t-1 + β 13 Restrict t-1 + ε t (2) We present the regression results in Table 3. The coefficient on CSR_Environment t-1 is negative and significant ( , p=0.0315) whereas coefficients on all the other CSR dimensions are insignificant at the 10% level. Our results suggest that the environment dimension of CSR is a dominant constraining force limiting executives insider trading profits Effect of new CSR initiatives To reinforce our inference that CSR constrains insider trading, we complement the main analyses by examining whether a firm s adoption of new CSR policies affects 16

18 executives subsequent insider trading profits. Since we do not have data on CSR initiations, we identify likely adopters of new CSR policies based on the MSCI data, i.e., we select a sample of firms rising from non-csr conscious (CSR = 0) to CSR conscious (CSR = 1). For this subsample, we estimate the following model using executive trades before and after their rise to CSR consciousness. Trading Profit t = α + β 1 Post + β 2 BTM t-1 + β 3 LogMV t-1 + β 4 R&D t-1 + β 5 Loss t-1 + β 6 Volatility t-1 + β 7 LogAnalyst t-1 + β 8 Turnover t-1 + β 9 Restrict t-1 + ε t (3) The indicator variable Post is equal to one for executive trades during the first year after the firm turns CSR conscious, and zero for their trades during the most recent year before that. 16 We expect the coefficient on Post to be negative if executives insider trading profits decrease after the adoption of new CSR policies. Table 4 summarizes the estimation results of model (3). The coefficient on Post is negative and significant ( , p=0.0067), suggesting that executives insider trading profits decrease as a firm turns CSR conscious. This finding reinforces our primary inferences, suggesting that insider trading is more constrained with enhanced CSR consciousness New executives from outside the firm An alternative explanation for our finding of a negative relation between CSR and executives insider trading profits is that executives who are less likely to exploit their 16 Executives do not trade in all years, so we examine firm-years with trades to identify rises in CSR consciousness. If we identify the rises by comparing adjacent firm-years only (i.e., we require these adjacent years to have executive trades), our sample size is reduced by half but our results are robust, i.e., the coefficient on Post is still negative and significant. 17 Using an alternative specification of a firm-year level regression that regresses changes in yearly average executives insider trading profits on an indicator variable for rising from non-csr conscious to CSR conscious and changes in all control variables in model (1) yields the same inference. 17

19 private information for trading profits are also more likely to take on more CSR. To explore this alternative explanation, we conduct an additional analysis focusing exclusively on newly appointed executives from the outside. These new executives do not create the firm s CSR orientation. Thus, under the alternative explanation, we do not expect these new, outside executives trading profits to be systematically correlated with the existing CSR orientation of their new employer. However, since they are constrained by CSR and the reputational concerns at the new company, our hypothesis should extend to these new executives and predict a negative association between their trading profits and the CSR consciousness of the new employer. We identify newly appointed executives from the outside using two alternative approaches. The first approach relies on the TFN insider filing data. We examine all transactions, including option exercising, of executives and lower-level officers of all firms to track the employment history of each person. If an executive was affiliated with another organization, as an executive or as a lower-level officer, and the gap between his/her first trade as an executive at the current company and last trade at the other organization is within five years, we identify the executive as recently hired from the outside. 18 The second approach relies on the Execucomp data. Execucomp provides information on the year in which an officer joins a company for about 30 percent of executive/company combinations. We then manually merge this subsample with the insider trading data by executive name. Under either approach, we estimate model (1) with the purchase transactions made by these new, outside executives only. 18 Requiring the gap between trades to be within three years, two years, or one year does not change our inferences. 18

20 The estimation results are reported in Table 5. Column (1) shows the results for new executives identified using the TFN data. The coefficient on CSR t-1 is negative and significant ( , p=0.0176), suggesting that new executives of CSR conscious firms make lower profits from insider trading than their counterparts of non-csr conscious firms. Column (2) reports the results for new executives identified using Execucomp data. Consistent with the first column, the coefficient on CSR t-1 is also significantly negative ( , p=0.0532). These results provide further support for our main inference that CSR acts as a constraint of insider trading Corporate governance and the constraining effects of CSR The constraining effect of reputational concerns on insider trading is likely more pronounced in the absence of effective alternative constraining mechanisms. Ravina and Sapienza (2010) suggest that, when corporate governance is weak, executives have more opportunities to expropriate shareholders and to profit from informed trading. In this case, we expect the reputational capital captured by CSR orientation to exhibit a stronger constraining effect, i.e., the difference in insider trading profits between CSR conscious and non-csr conscious firms is more evident under weak corporate governance. We partition our sample into strong- vs. weak-governance firms based on the median of the governance index proposed by Gompers et al. (2003). 20 Among the 2, A potential selection issue in the appointment of new executives is the tendency of CSR conscious firms to hire executives from other CSR conscious firms. Our inferences are unchanged if we drop new executives who used to work for CSR conscious firms. 20 Since G-index data is not available after 2006, we apply the values of 2006 to future years (i.e., ). Our results are qualitatively the same if we drop years after 2006 from our sample. We follow Ravina and Sapienza (2010) to use Gompers et al. s (2003) governance index but not use MSCI s corporate governance data as our proxy for corporate governance because MSCI measures the strengths and weaknesses of governance based on the quality of a firm s CSR/sustainability reporting, a firm s attitude 19

21 observations in the weak-governance sample, 1,020 observations are from CSR conscious firms; 847 of the 2,465 observations in the strong-governance sample are from CSR conscious firms. We estimate model (1) for firms with strong and weak corporate governance separately and report the regression results in Table 6. We find a negative and significant coefficient on CSR t-1 for firms with weak corporate governance ( , p=0.0031), suggesting that executives of CSR conscious firms make lower insider trading profits than executives of non-csr conscious firms when governance is weak. However, the coefficient on CSR t-1 is insignificant (0.0130, p=0.3095) for firms with strong corporate governance. These results are consistent with our expectation that the constraining effects of a firm s CSR orientation on executives insider trading profits is more pronounced under weak corporate governance. Gompers et al. s (2003) governance index captures overall governance based on corporate charter provisions that may not be specifically designed for monitoring insider trading. Thus, in addition to this general measure, we also explore the impact of insidertrading-specific governance mechanisms (i.e., insider trading policies) on the constraining effects of CSR on insider trading profits. We search firms websites as of November 2012 to collect data on whether they disclose a stand-alone insider trading policy or caution against insider trading in the code of conduct. We focus on a subsample of insider trades occurring in 2011, the most recent year in our sample, assuming that firms maintain consistent insider trading policies and disclosures between 2011 and towards public policies on environment, communities, employees, or consumers, the severity of controversies related to a firm s executive compensation and governance practices, and the severity of controversies related to a firm s business ethics practices (MSCI ESG STATS User Guide & ESG Ratings Definition, June 2011). These dimensions are very different from the traditional measures of corporate governance that researchers use in academic research. 20

22 Such disclosures highlight their concerns over insider trading and therefore we consider it an indicator of strong insider-trading-specific governance. About 61% (304 out of 498) of our sample make these disclosures. We find a significant difference in executives trading profits between CSR conscious and non-csr conscious firms only in the subsample where firms do not emphasize insider trading restrictions on their websites, i.e., firms likely with weak insider-trading-specific governance. Since these inferences are the same as those from Table 6 on overall governance, the results are untabulated. 4. Additional analyses 4.1. Endogeneity of CSR and propensity score matching A firm s CSR orientation can be endogenously determined by the characteristics of the firm. To the extent that certain firm characteristics drive a firm s CSR orientation and are omitted from our prior analyses, the relation between CSR and executives insider trading profits could be spurious. We follow the prior literature (e.g., Armstrong et al., 2010; Jagolinzer et al., 2011) and adopt the propensity score matching technique to mitigate this concern. To generate the propensity score, we construct a first-stage logistic regression model for CSR t-1 at the firm-year level as follows: Prob(CSR t-1 =1) = logit(α + β 1 BTM t-1 + β 2 LogMV t-1 + β 3 R&D t-1 + β 4 Loss t-1 + β 5 Volatility t-1 + β 6 LogAnalyst t-1 + β 7 Turnover t-1 + β 8 Restrict t-1 + β 9 Free Cash Flow t-1 + β 10 Institutional Ownership t-1 + β 11 Firm Age t-1 + β 12 Market Share t-1 + ε t-1 ) (4) 21 We cannot follow Bettis et al. (2000) or Jagolinzer et al. (2011) to construct their insider-trading-specific governance variables due to data constraints. Bettis et al. s (2000) analysis is based on a survey of companies on insider trading policies. Jagolinzer et al. (2011) conduct a comprehensive web search for detailed insider trading policies and find only 437 policies out of the universe of public firms. In our random sample, only 14 firms disclose detailed insider trading policies. 21

23 We include all the control variables in model (1). In addition, we control for Free Cash Flow t-1, which measures the difference between cash flow from operations and cash flow used in investing activities. We expect that firms with more free cash flow have more resources to invest in CSR. Since institutional investors care about CSR (Johnson and Greens, 1999; Kim et al., 2012), we also control for Institutional Ownership t-1, measured as the percentage of institutional holdings. Dhaliwal et al. (2012) suggest that older firms are more likely to invest in CSR, so we control for Firm Age t-1. We also follow Dhaliwal et al. (2012) and use Market Share t-1, measured as a firm s fraction of sales in its industry, to capture the firm s visibility and public pressure for CSR performance. Panel A of Table 7 tabulates the first stage logistic regression results. The regression is estimated with 4,273 firm-year observations in our full sample, 1,240 of which are classified as CSR conscious firms. Consistent with our expectations, we find that larger firms, and firms with more R&D investments, higher investor interests, and higher institutional ownership are more likely to be CSR conscious firms. Next, we calculate the propensity scores using predicted probabilities from the logistic regression and match each CSR conscious firm-year to a control firm-year with the closest propensity score, provided that the propensity score of the closest match is within a distance of 0.1. This procedure results in 1,109 matched pairs (i.e., about 90% of the CSR-conscious firm-year observations are matched). The mean (median) propensity score of the CSR conscious sample is (0.3662), while that of the matched control sample is (0.3540). As in Armstrong et al. (2010), we perform a parametric t-test of the difference in means between the CSR conscious sample and the matched sample and a non-parametric Kolmogorov-Smirnov test of the difference between the two 22

24 distributions. Neither test shows the difference to be statistically different from zero at conventional levels. We then compare executives insider trading profits in CSR conscious firms vs. the propensity score matched control firms. 22 The mean trading profits for executives of CSR conscious firms is significantly lower than the mean trading profits for executives of the control firms (0.0390% versus %, p<1%). 23 Comparison of the medians shows the same pattern. These results are similar to our main results reported in Table Impact of media coverage Zyglidopoulos et al. (2012) argue that media attention could potentially pressure firms to be more socially responsible. Meanwhile, Frankel and Li (2004) predict that news coverage, as a source of information, helps decrease insider trading profits. Thus, the negative association between executives trading profits and CSR t-1 may be driven by CSR conscious firms being covered more extensively by the media. To mitigate this concern, we collect from Factiva the number of news articles covering each firm during the fiscal year of the trades. We include all sources available on Factiva, both voluntary disclosures and press coverage. A simple correlation analysis indicates a positive and significant correlation between CSR t-1 and the average daily number of news articles, Media t, (0.1474, p<0.0001). The correlation between trading profits and Media t is negative but insignificant ( , p=0.6898). 22 Following Jagolinzer et al. (2011), we perform the first-stage propensity score matching procedure at the firm-year level, but conduct this insider trading test at the transaction level. 23 The propensity score matching procedure creates a pseudo random sample such that univariate differences in means between the treatment and control groups should be sufficient to estimate the treatment effects (Dehejia and Wahba, 2002; Dehejia, 2005). Nonetheless, our inference remains unchanged if we rerun model (1), a multivariate analysis, for the pooled sample of CSR conscious firms and the propensity score matched control firms. 23

25 We estimate model (1) again, including Media t as a control variable. The results are reported in the first column of Table 8. The coefficient on Media t is insignificant (0.0047, p=0.2435). 24 Importantly, it does not explain away the negative association between executives trading profits and CSR t-1. The coefficient on CSR t-1 remains negative and significant ( , p=0.0319) Insider trading as a form of compensation Manne (1996, 2011) argues that insider trading can serve as a compensation device. Consistent with this argument, Roulstone (2003) finds firms that restrict insider trading pay a premium in total compensation relative to firms not restricting insider trading. This argument suggests that executive compensation can be an omitted correlated variable if CSR conscious firms are more likely than others to pay higher executive compensation. To address this issue, we include the level of executive total compensation in model (1). We obtain executive compensation data from Execucomp. Since matching TFN and Execucomp at the person level is a manual operation, we take the natural logarithm of median total compensation of the top five most highly paid executives as a firm-year level measure of compensation. 25 The estimation results are reported in the second column of Table 8. We find a positive coefficient on Compensation t (0.0152, p=0.0355), contrary to the expectation of 24 Frankel and Li (2004) find a positive correlation between media coverage and trading profits, although they predict a negative association. 25 Alternatively we use the logarithm of the average total compensation and the dollar amount of average or median compensation as robustness checks. While the coefficient on the compensation variables differs across different specifications, the coefficient on our variable of interest, CSR t-1, is consistently negative and significant at better than 5% level. 24

26 Manne (1996, 2011). Importantly, the coefficient on CSR t-1 remains negative and significant ( , p=0.0576), consistent with our prior results CSR and lower-level managers Given their lower visibility than executives, informed trading by lower-level managers is likely less damaging to a firm s reputation. Therefore, the constraining effects of reputational concerns on insider trading are likely less pronounced for lowerlevel managers. To compare with our previous results for executives, we re-estimate model (1) for trades made by lower-level managers. Table 9 reports the estimation results. The coefficient on CSR t-1 is not significantly different from zero (0.0056, p=0.3382), whereas the estimation results of model (1) for trades made by executives (Table 2 Panel B) reveal a negative and significant coefficient on CSR t-1 ( , p=0.0074). This contrast is consistent with our expectation that insider trading less damaging to a firm s reputation is less constrained by CSR CSR and the likelihood of news-based insider trading In addition to trading profits, we also examine the likelihood of executives trading on private information as an alternative measure of insider trading activity. We expect that CSR also lowers the likelihood of executives engaging in informed purchases. We adopt the following model in Piotroski and Roulstone (2008), which focuses on trading 25

27 ahead of changes in earnings and returns, and estimate it separately for CSR conscious firms and non-csr conscious firms: 26 Prob(Purchase t =1) = logit(α + β 1 Earn t+1 + β 2 MARet t+1 + β 3 BTM t + β 4 LogMV t + β 5 MARet t + β 6 R&D t + ε t ) (5) We define an indicator variable Purchase t, equal to one if any executive purchased shares in an open-market transaction during year t and total shares purchased by executives are greater than or equal to total shares sold by executives. Earn t+1 is one of the two proxies for future news, measured as the change in earnings from year t to year t+1. MARet t+1 is a second proxy for future news, measured as the twelve-month buy-and-hold marketadjusted return over year t+1. We control for insiders contrarian trading tendency with the twelve-month buy-and-hold market-adjusted return over year t (MARet t ). We also control for the book-to-market ratio (BTM t ), firm size (LogMV t ), and R&D activities (R&D t ). 27 We report the results of regression model (5) separately for CSR conscious firms and non-csr conscious firms in Table 10. For non-csr conscious firms, we find a positive coefficient on both proxies for future news, i.e., change in earnings (0.6606, p=0.0399) and future returns (0.1163, p=0.0901), suggesting that executives are more likely to buy prior to good news. In contrast, neither coefficient is statistically significant 26 Examining the trading patterns before future news also mitigate a potential concern of measuring insider trading profits with post-trade abnormal returns (Ke et al, 2003; Piotroski and Roulstone, 2005) since one may argue that the post-trade price movements might have been caused by the trades themselves and thus post-trade abnormal returns may not indicate insiders profiting from superior information (e.g., Ravina and Sapienza, 2010). 27 We also include variables capturing whether executives received grants of restricted stock or options during year t and whether executives exercised options during year t as additional control variables. Untabulated results show that adding these controls reduces our sample size by more than 50%, but our inferences remain unchanged. 26