Internal Corporate Governance and Insider Trading *

Transcription

1 Internal Corporate Governance and Insider Trading * Lili Dai, Renhui Fu, Jun-Koo Kang, and Inmoo Lee October 2013 Abstract This paper examines how internal governance systems limit insiders ability to profit from their information advantage. We find that compared with insiders of poorer-governed firms, those of better-governed firms earn significantly smaller abnormal profits from their sales transactions, but not from their purchase transactions. This result is more pronounced when firms have greater monitoring needs. Moreover, better-governed firms are more likely to place ex-ante preventive measures to restrict insider trading and take ex-post disciplinary actions against CEOs who earn large abnormal profits from their sales transactions. We also find that good governance restricts insiders use of private information but not that of public information. * Lili Dai is from Erasmus School of Economics, Erasmus University, Rotterdam, Netherlands (ldai@ese.eur.nl), Renhui Fu is from Krannert School of Management, Purdue University, West Lafayette, Indiana (fu106@purdue.com), Jun-Koo Kang is from Nanyang Business School, Nanyang Technological University, Singapore, (jkkang@ntu.edu.sg), and Inmoo Lee is from KAIST College of Business, KAIST, Seoul, Korea (inmool@kaist.ac.kr). We thank Tim Loughran, David Veenman, and seminar participants at the University of Amsterdam, Erasmus University, KAIST, National Taiwan University, Seoul National University, and Queen s University Belfast for their helpful comments. All errors are ours.

2 Internal Corporate Governance and Insider Trading Abstract This paper examines how internal governance systems limit insiders ability to profit from their information advantage. We find that compared with insiders of poorer-governed firms, those of better-governed firms earn significantly smaller abnormal profits from their sales transactions, but not from their purchase transactions. This result is more pronounced when firms have greater monitoring needs. Moreover, better-governed firms are more likely to place ex-ante preventive measures to restrict insider trading and take ex-post disciplinary actions against CEOs who earn large abnormal profits from their sales transactions. We also find that good governance restricts insiders use of private information but not that of public information. Keywords: Internal corporate governance, Insider purchases, Insider sales, Profitability of insider trading, Legal risk. JEL Classification: G34, J33, K31, M52.

3 1. Introduction There has been much debate over the usefulness of insider trading. On the one hand, some researchers argue that insider trading contributes to the market efficiency and helps firms compensate managers for their successful entrepreneurship (e.g., Manne 1966; Roulstone 2003; Piotroski and Roulstone 2005). 1 On the other hand, others argue that informed insider trading discourages outsiders from actively collecting value-relevant information since it limits potential gains from such information. They further suggest that insider trading should be banned since it allows insiders to exploit their information advantage over other market participants and thus enables them to extract private benefits (e.g., Seyhun 1986; Fishman and Hagerty 1992; Bettis, Coles, and Lemmon 2000; Jagolinzer, Larcker, and Taylor 2011). Consistent with this latter view, policymakers have put various restrictions on insider trading, such as Rule 10b-5 of the Securities Exchange Act of 1934, the Insider Trading and Securities Fraud Enforcement Act (ITSFEA), and the Stock Enforcement Remedies and Penny Stock Reform Act (SERPSRA). 2 In this paper we extend the latter view on insider trading by examining how internal governance systems limit insiders ability to profit from their information advantage. 3 If managers engage in insider trading for their own benefits at the expense of shareholder interests and if internal governance systems are effective in aligning the interests of shareholders with those of top managers, we expect these systems to discourage managers from engaging in informed insider transactions, especially informed sales transactions (i.e., selling shares before stock price declines) because they are likely to increase litigation risk more than informed 1 Roulstone (2003) show that insider trading restriction significantly affects executive compensation and firms restricting insider trading tend to provide higher compensation for their executives, possibly to make up for the lost opportunity to earn abnormal returns from insider trading. 2 The ITSFEA and the SERPSRA make both top management and the firm responsible for employees illegal trading. 3 Even under the former view, increased legal risk due to informed insider transactions can still be a serious concern for shareholders and therefore, internal governance systems are expected to play an important role in restricting such insider transactions. 1

4 purchase transactions. Moreover, allowing insider sales cannot be viewed as an optimal way to compensate individuals for managerial success since they are more likely to reflect managerial failure rather than entrepreneurial success. These arguments suggest that internal corporate governance significantly affects the profitability of insider trading, particularly that of insider sales. By examining the role of internal governance in insider trading, our study attempts to provide the answers to the following questions: i) does good internal governance restrict the information content of insider trading?, ii) does the restrictive effect of internal governance on the informativeness of insider trading differ between sales and purchase transactions?, and iii) is this restrictive effect simply due to well-governed firms ex-ante preventive policies that discourage informed insider trading or is it due to their use of ex-post disciplinary actions (e.g., demotion or termination of employment) against insiders engaged in informed insider trading? 4 We focus on internal governance systems rather than external governance systems since informed insider trading tends to occur mainly due to a lack of internal controls and therefore, it is likely to be more effectively controlled by internal systems than by external systems. 5 Previous studies show that firms with good internal governance, such as those with outsidedominated boards (Hermalin and Weisbach 1998), those with high CEO pay-performance sensitivity (Li and Srinivasan 2011), and those with independent large shareholders (Shleifer and Vishny 1986), are effective in monitoring top management, suggesting that to protect shareholder wealth, these firms are more likely to take actions to discourage informed insider trading. In a related study, using a governance index constructed by Gompers, Ishii, and Metrick 4 Other examples of disciplinary actions include oral or written warning, suspension, removal of job duties and responsibilities, and reduction in compensation. 5 Acharya, Myers, and Rajan (2008) emphasize the importance of internal corporate governance in mitigating shareholder-manager conflicts and show that it reduces firms agency problems even without any external governance mechanisms. 2

5 (2003) (hereafter, GIM index), Ravina and Sapienza (2010) show that market-adjusted abnormal returns earned by executives and independent directors are closely related to firms GIM index. Although their study provides evidence on how the quality of a firm s governance affects the profitability of insider trading, to the extent that the GIM index is constructed using the firm s antitakeover provisions and thus, is not able to explicitly capture the effectiveness of internal governance mechanisms, it does not provide the direct answer on whether internal governance plays a role in preventing insiders from exploiting private information. Moreover, their main research question is not to examine the relation between corporate governance and profits earned by insiders, but to compare the trading performance of independent directors with that of other executives. Using a large sample of insider trading on firms stocks listed on the NYSE, AMEX, or NASDAQ from 1998 to 2011, we find that better internal governance significantly reduces the six-month profitability of insider (officers and directors) sales but not that of insider purchases. 6 This asymmetric effect of internal governance on the profitability of insider trading is consistent with the compensation argument suggested by Roulstone (2003). 7 We also find that the above result holds after controlling for insider trading policies, general counsel pre-approval requirements, and other determinants of the profitability of insider trading (Bettis, Coles, and Lemmon 2000; Jagolinzer, Larcker, and Taylor 2011) and is robust to a 6 We focus on six-month profitability because of the short swing rule (Section 16(b) of the Securities Exchange Act of 1934) that prohibits insiders from earning profits from a round-trip transaction within a six-month interval, which is likely to force insiders not to reverse their position for at least six months. 7 Roulstone (2003) shows that insider trading restrictions lead to higher executive compensation, suggesting that firms recognize insider trading as a source of executive compensation and increase compensation when insiders are restricted in trading their shares. Since allowing profitable insider purchases and prohibiting opportunistic insider sales are the ways to reward and discipline managers for their success and failure, an optimal compensation structure will focus on restricting only insider sales, suggesting that good internal governance can asymmetrically affect the profitability of insider trading. Consistent with our result, Lee, Lemmon, Li, and Sequeira (2013) also find that while insiders of firms with voluntary restrictions on insider trading take advantage of positive information in their trading, they become more cautious when exploiting negative information. 3

6 variety of internal corporate governance measures and regression model specifications, alternative measures of insider trading profitability, and the consideration of the endogeneity concern. We obtain similar results when we measure the profitability of insider trading using a short-term event window (i.e., the three-day window around insider trading filings) instead of using a long-term event window of six months. Moreover, we find that the impact of a firm s internal governance on the informativeness of insider trading is more pronounced for firms with greater ex-ante litigation risk and for opportunistic transactions that are more likely to attract attention from the Securities and Exchange Commission (SEC) (Cohen, Malloy, and Pomorski 2012). These results are consistent with the view that good internal governance is particularly effective in restricting informative insider trading that can potentially increase legal risk. We also find that firms with better internal governance are more likely to adopt policies to discourage informed insider trading and take expost disciplinary actions (i.e., forced CEO turnover) against CEOs engaged in informed insider transactions, especially for sales transactions. Finally, we find that better internal governance restricts insiders use of private information but not their use of public information. In particular, insiders of better-governed firms earn significantly lower abnormal returns from their sales transactions made immediately before the announcements of bad news such as earnings restatements and stock price crashes. Thus, good governance plays an important role in reducing the profitability of insider transactions that are likely to be motivated by negative information and thus, result in high legal risk. Overall, these results suggest that firms with good governance effectively discourage officers and directors from exploiting negative private information, but not from exploiting positive private information. Good governance reduces the incentives of insiders to engage in profitable 4

7 insider trading by influencing firms to adopt ex-ante preventive measures on informed insider trading, such as voluntary insider trading policies, and to take ex-post disciplinary actions against the CEOs who earn large abnormal profits from their sales transactions. Our paper contributes to the literature in at least two important ways. First, in terms of corporate governance, it sheds new light on the role of internal governance in preventing insiders from exploiting private information. Prior literature shows that corporate governance affects a firm s idiosyncratic risk, stock liquidity, and valuation (Ferreira and Laux 2007; Aggarwal, Erel, Stulz, and Williamson 2009; Chung, Elder, and Kim 2010). Our paper further shows that better governance benefits shareholders by discouraging insiders from taking advantage of private information and that well-designed internal governance systems are effective in aligning the interests of shareholders with those of top managers. Second, our paper contributes to the literature on insider trading by documenting that internal corporate governance matters for the informativeness of insider sales, but not for the informativeness of insider purchases. This asymmetric effect of internal governance on the informativeness of insider trading is particularly pronounced for transactions made by insiders of firms with high litigation risk and for opportunistic transactions. These findings add to the literature on the link between insider trading and executive compensation (Roulstone 2003). We show that insiders of a well-governed firm are not necessarily compensated from their sales transactions since their incentives to engage in these transactions are likely to be motivated by extracting private benefits at the expense of shareholder wealth. The remainder of the paper is organized as follows. In Section 2, we discuss the construction of our key variables, namely, the measure of the quality of a firm s internal governance and the profitability of insider trading, and describe the data and sample characteristics. Section 3 5

8 outlines our empirical methodology. Section 4 presents the empirical results. In Section 5, we examine how firms with better internal governance discourage insiders from taking advantage of negative private information. We present the summary and concluding remarks in Section Variable Construction and Sample Description 2.1. Measure of the quality of internal governance To measure the quality of a firm s internal governance, we focus on three aspects of governance mechanisms: board independence, compensation structure, and institutional ownership. Previous studies emphasize the importance of board independence in internal governance (e.g., Weisbach 1988; Rosenstein and Wyatt 1990; Byrd and Hickman 1992). We use the percentage of outside directors on the board and the percentage of outside directors in the compensation committee as the measures of board independence. Weisbach (1988), Rosenstein and Wyatt (1990), and Byrd and Hickman (1992) show that outside directors protect the interests of shareholders when their interests are diverged from those of managers, suggesting that the percentage of outside directors on the board can serve as an important measure of board independence. Roulstone (2003) shows that executives of firms with restrictions on insider trading tend to receive a premium in their total compensation, suggesting that insider trading policies are closely related to executive compensation. Given the importance of the compensation committee in determining executive compensation structures, as in Laksmana (2008), we use the percentage of outside directors in the compensation committee as another measure of board independence. Compensation structure can also play an instrumental role in internal governance. Previous studies show that agency problems between managers and shareholders can be reduced by 6

9 optimally designing compensation structures (Jensen and Murphy 1990) and that firms with better governance tend to have higher pay-performance sensitivity (Li and Srinivasan 2011). When executives and directors have compensation schemes that are tied to firm performance, their interests are more likely to be aligned with those of shareholders, suggesting that compensation structure plays an important role in mitigating shareholder-manager conflicts. We measure the effectiveness of compensation structure as an internal governance mechanism using two variables (Hoechle et al. 2012): CEO pay-performance sensitivity and an indicator for firms in which non-executive directors receive shares or options. Previous studies also show that large shareholders perform an important monitoring function (Shleifer and Vishny 1986). For example, Hartzell and Starks (2003) find that institutional investors have strong incentives to closely monitor managers when they have a large stake in the firms. We use two measures to capture the governance role of large shareholders: the percentage of ownership held by institutional investors and the percentage of shares held by the top five independent, long-term, dedicated/quasi-indexer institutional investors as defined in Chen, Harford, and Li (2007). To measure the overall quality of a firm s internal governance, we aggregate the above six measures of three main internal governance attributes into one composite measure using an approach similar to the one used in Larcker, Richardson, and Tuna (2007) and Armstrong, Ittner, and Larcker (2012). Specifically, we first standardize each variable to have zero mean and unit variance, and then sum this standardized value of all individual governance variables to obtain our main composite internal governance score, ICG. We collect the information on directors, executive compensation, and institutional ownership from the RiskMetrics, S&P s ExecComp, and Thomson Reuters institutional 7

10 holdings (13F) databases, respectively. The RiskMetrics provides information on director ownership and compensation committee membership starting from 1998 while the ExecComp provides information on executive compensation for about 2,800 firms that are or were members of the Standard and Poor s (S&P) 500 index, the S&P MidCap 400 index, or the S&P SmallCap 600 index. This data availability allows us to measure the quality of internal governance for those S&P 1500 firms covered by the ExecComp from To check the robustness of our results, we experiment with two alternative measures of the quality of a firm s internal governance. First, we use a more comprehensive list of governance variables including board size, the percentages of old directors and busy directors, an indicator for at least one independent director who is a blockholder, the fraction of directors whose tenure predates the CEO, an indicator for CEO/Chairman duality, and the ownership held by insiders as in Armstrong, Ittner, and Larcker (2012), Hazarika, Karpoff, and Nahata (2012), and Hoechle, Schmid, Walter, and Yermack (2012). We first standardize these variables and then add their standardized values to ICG to obtain a more comprehensive governance measure, ICG1. Appendix 1 summarizes the detailed description on the construction of each governance variable used in the analysis. Second, we use the governance score based on the Corporate Governance Quotient (CGQ) that is constructed by Institutional Shareholder Services (ISS). This measure has been used by several papers including Aggarwal, Erel, Stulz, and Williamson (2009), Bernile and Jarrell (2009), Chung, Elder, and Kim (2010), and Chung and Zhang (2011), and mostly captures the quality of a firm s internal governance. ISS covers a large set of U.S. firms included in the S&P 500 index, the S&P SmallCap 600 index, or the Russell 3000 index starting from The ISS 8 ISS also covers the firms that are required to file various documents and forms with the SEC through the Electronic Data Gathering, Analysis, and Retrieval system (EDGAR). For a detailed description of CGQ for U.S. 8

11 governance score is the number of minimally acceptable governance attributes met by a firm out of 64 governance attributes along eight major dimensions, with higher ISS scores indicating better governance Measure of the profitability of insider trading We measure the profitability of insider trading by estimating abnormal returns over the 180 calendar days following the transaction date. 10 We use two approaches to calculate abnormal returns. First, similar to Ravina and Sapienza (2010), we use the market-adjusted abnormal return defined as the difference between a firm s buy-and-hold return over the 180 calendar days following the transaction date and the corresponding buy-and-hold return for the market (BHAR6m), where we use the CRSP value-weighted index as a proxy for the market portfolio. Second, following Jagolinzer, Larcker, and Taylor (2011), we define the abnormal return as an intercept from the Carhart (1997) four-factor model estimated over the 180 calendar days subsequent to the transaction date (Alpha6m). Specifically, we estimate the following regression model and use the intercept, α, as our measure of daily abnormal returns from insider trading: (R i - R f ) = α +β 1 (R mkt - R f ) + β 2 SMB + β 3 HML + β 4 UMD + ε, (1) firms covered by ISS, see Aggarwal and Williamson (2006) and Aggarwal, Erel, Stulz, and Williamson (2009). ISS stopped updating CGQ in March 17, Detailed information on CGQ is available at the ISS web site ( 9 Appendix 2 provides these 64 minimally acceptable governance standards set by ISS and the proportion of firms that meet these standards from 2001 to If a firm meets the minimally acceptable standard in one governance attribute, then we assign a value of one to this attribute and zero otherwise. The ISS score equals to the sum of these assigned values. For example, an ISS score of 34 means that out of 64 standards, 34 are met. Since the number of governance attributes covered by ISS has increased over time and some governance attributes are not available during our full sample period, we replace missing values of governance attributes in a particular year by the values available in prior or later years. Our results remain similar when we use the ISS score based on governance attributes available during the sample period from 2001 to In untabulated tests, we use the log of total dollar profits earned from insider trading as an alternative measure of the informativeness of insider trading and find similar results. Since total dollar profits are confounded by size and individual wealth effects, we focus on the measures based on abnormal returns as in previous studies (Ravina and Sapienza 2010; Jagolinzer, Larcker, and Taylor 2011). The results using dollar profits are available upon request. 9

12 where R i is the daily stock return of firm i, R f is the daily risk-free interest rate, R mkt is the CRSP value-weighted index return, and SMB, HML, and UMD are the size, book-to-market, and momentum factors, respectively (Fama and French 1993; Carhart 1997) Sample and descriptive statistics Our initial sample includes insider transactions of firms listed on the NYSE, AMEX, or NASDAQ covered in Thomson Financial Insiders Data Feed (IDF) over the period 1998 to Thomson Financial IDF contains trade information on directors, officers, and large stockholders with holdings greater than 10 percent of a firm s stock, all subject to disclosure requirements as defined in Section 16 of the Securities Exchange Act of Since our main hypothesis regarding the role of governance in preventing insiders from exploiting private information is relevant only for officers and directors, we exclude transactions made by large shareholders from the sample and use only transactions made by officers and directors in most of our analyses. We focus only on valid purchase and sales transactions of common shares. 12 To be included in our sample, we also require that firms be covered by RiskMetrics, ExecuComp, and Thomson Reuters institutional holdings (13F) data, and that their stock return and financial data be available in CRSP and Compustat, respectively. Following previous studies, we further limit the sample by requiring that share codes in CRSP be 10 or 11 and we exclude the following transactions from the sample: (1) transactions with less than 100 shares or those with trading prices less than $2; (2) transactions with traded prices outside the range between the 11 To check the robustness of the results, in untabulated tests we also use the market-adjusted buy-and-hold abnormal returns over the 1, 3, and 12 months following the transaction date, and the alphas from the Carhart (1997) four-factor model estimated over the same time periods. We find that our results are robust to using these alternative measures of the profitability of insider trading. 12 A valid transaction is one without a cleanse code of A or S in Thomson Financial Insiders Data Feed database. 10

13 daily low and high prices reported in CRSP; (3) transactions with the number of shares exceeding the total number of shares outstanding in CRSP; (4) transactions with the number of shares traded exceeding the total daily trading volume in CRSP; and (5) firms in the financial or utilities industries (firms with SIC codes between 6000 and 6999 or between 4900 and 4999). These restrictions result in a final sample of 11,310 firm-year observations and 463,527 insider transactions. 13 We obtain data on analyst forecasts from I/B/E/S and data on litigation from Stanford Securities Class Action Clearinghouse. Because the data requirements differ across tests, the sample size for each test varies depending on its data requirements. Table 1 reports descriptive statistics for the sample of 463,527 insider transactions. Detailed description on each variable is available in Appendix 3. Panel A reports the summary statistics of variables used in our regression analyses while Panel B reports the statistics of variables used in the construction of internal corporate governance measures. The numbers reported in the Mean column can be interpreted as the insider transactions-weighted average numbers across sample firms. The mean BHAR6m, Alpha6m, and cumulative market-adjusted excess return over the 180 calendar days prior to the insider trading date (CAR_6m) are -0.31%, basis points, and %, respectively, and the mean (median) ICG, ICG1 and ISS scores are 0.00 (0.48), 0.00 (0.32), and (38), respectively. The mean and median market values of equity (Size) are $18.51 billion and $3.26 billion in 1998 purchasing power, respectively, and the mean and median market-to-book equity ratios (MB) are 4.25 and 2.99, respectively. 13 There were a total of 1,429,547 transactions made by insiders of NYSE, AMEX and NASDAQ firms that are available in Thomson Financial s IDF database during our sample period. Due to the stock returns and financial data requirements from CRSP and Compustat, 40,659 and 45,589 observations were excluded, respectively. In addition, 212,710 observations were excluded due to the unavailability of analyst information or due to the requirement of a minimum of 3 analysts following the firm from I/B/E/S. Finally, 667,062 transactions were excluded from the sample due to the fact that the data was unavailable in RiskMetrics, ExecComp, or Thomson Reuters insider holdings (13F) data. Our results are similar when we exclude transactions made on a date when both insider purchases and sales are made in the same company. 11

14 The mean ratio of the number of shares purchased and sold by insiders on the transaction date to the total number of shares outstanding (TradeSize) and the mean ratio of the number of shares traded by all insiders of the same firm during the ten days prior to the transaction date to the total number of shares outstanding (RecentTrades) are 0.27% and 0.58%, respectively. We also find that 25.5% and 7.6% of our sample transactions are made by insiders of firms reporting non-zero R&D expenditures (RND) and by insiders of firms with negative net income before extraordinary items during the most recent fiscal year (Loss), respectively. The mean (median) standard deviation of financial analysts earnings per share (EPS) forecasts over the average forecasted EPS (Dispersion) is 10.0% (3.4%). Finally, the mean fraction of transactions made by insiders of firms with insider trading policies (general counsel pre-approval requirements), ITP (GC), is 48.1% (8.9%). 14 In Panel B, we report the descriptive statistics of governance variables used to construct our internal governance measures. Reported numbers are based on raw values before the standardization. The mean board independence (proportion of outside directors on the board) is 73.3% and the mean compensation committee independence (proportion of outside directors in the compensation committee) is 95.2%. The mean CEO pay-performance sensitivity indicates that on average, options and shares account for 75% of CEO total annual compensation. In our sample, 57.4% of non-executive directors receive stocks or options as compensation. On average, 77.6% of shares are owned by institutional investors and of which, 13.5% are owned by the top five independent, long-term, dedicated/quasi-indexer institutional investors. The mean board size and the mean percentages of old directors who are older than 72 and busy directors who hold three or more outside directorships in other firms are 9.4, 6.5% and 15.2%, respectively. About 14 Following Jagolinzer, Larcker, and Taylor (2011), we collect the data on insider trading policies and general counsel pre-approval requirements by manually searching firms websites. 12

15 4.4% of our sample transactions have at least one outside director who is a block shareholder with more than 10% of share ownership. We also find that on average, 11.7% of directors are appointed as directors before the CEO took the position. Finally, we find that the mean (median) insider ownership held by all board members and top 5 officers is 11% (4%). Table 2 reports pairwise Pearson and Spearman correlation coefficients of key variables used in our analyses. Since both correlation measures are similar in most cases, we only discuss the results of Pearson correlation coefficients that are reported above the diagonal. The correlations between our two abnormal returns measures, BHAR6m and Alpha6m, and between two internal governance measures, ICG and ICG1, are 0.77 and 0.69, respectively, while the correlation between ICG (ICG1) and log (ISS) is relatively low at 0.40 (0.08). More importantly, we find that abnormal returns are generally negatively correlated with internal governance measures, consistent with our hypothesis that insiders of firms with better internal governance are discouraged to exploit private information. We also find that our governance measures are significantly positively correlated with the indicator variables for firms with insider trading policy (ITP) and general counsel pre-approval requirements (GC). 3. Empirical Methodology To examine whether internal governance plays a role in limiting the profitability of insider trading, we estimate the following regression: BHAR6m (Alpha6m) = α + β 1 Gov + β 2 log(size) + β 3 MB + β 4 CAR_6m + β 5 TradeSize + β 6 RecentTrades + β 7 RND + β 8 Loss + β 9 Dispersion + β 10 ITP + β 11 GC + β 12 Fixed Effects + ε, (2) where 13

16 BHAR6m (Alpha6m) = Market-adjusted buy-and-hold returns over the 180 calendar days following the transaction date (intercept from the Carhart (1997) four-factor model estimated over the 180 calendar days subsequent to the transaction date), 15 Gov = Various governance measures including board independence, compensation structure, institutional ownership, ICG, ICG1, and log(iss) measures, Size = Inflation-adjusted market capitalization at the end of the most recent fiscal quarter (based on 1998 dollars), MB = Ratio of the market value of equity to the book value of equity at the end of the most recent fiscal quarter, CAR_6m = Cumulative market-adjusted abnormal returns over the 180 calendar days prior to the insider trading date, RND = Indicator that takes the value of one if the firm reports non-zero research and development expenditures in the most recent fiscal year and zero otherwise, TradeSize = Number of shares purchased and sold by insiders on the transaction date divided by the total number of shares outstanding, RecentTrades = Number of shares purchased and sold by all insiders of the same firm during the ten days prior to the transaction date, scaled by the total number of shares outstanding, Loss = Indicator that takes the value of one if the firm reports negative net income before extraordinary items for the most recent fiscal year and zero otherwise, 15 To make higher abnormal returns imply more profits earned by insiders regardless of types of transactions, we multiply -1 for sales transactions. 14

17 Dispersion = Standard deviation of earnings per share (EPS) forecasts for the current fiscal year divided by the average of EPS forecasts made during the month of transaction date, ITP = Indicator that takes the value of one if the firm has insider trading policies and zero otherwise, and GC = Indicator that takes the value of one if the firm has a general counsel pre-approval requirement and zero otherwise. Following Lakonishok and Lee (2001), we include Size and MB to control for size and bookto-market effects (Fama and French 1997). In addition, following Brochet (2010), we control for CAR_6m, RND, TradeSize, RecentTrades, and Loss. We include CAR_6m to control for insiders contrarian behavior. RND and Dispersion are included since insider sales and purchases are likely to be more informative in firms with higher R&D intensity or those with greater analysts forecast dispersion, for which information asymmetry problems are perceived to be greater than firms with lower R&D intensity or those with smaller forecast dispersion (Aboody and Lev 2000; Healy and Palepu 2001). We further include TradeSize to control for the possible link between the importance of private information and trade size, and RecentTrades to control for either preemptions of a trade s information content or reinforcements of prior signals. We also include Loss to control for the potential reversal of poor accounting performance. Finally, we include ITP and GC in the regression to control for previously documented effects of these insider trading policies on the profitability of insider trading (Bettis, Coles, and Lemmon 2000; Jagolinzer, Larcker, and Taylor 2011). Our key variable of interest is Gov. If the internal governance system reduces the profitability of insider trading, abnormal profits earned by insiders from their transactions are likely to be less 15

18 for better-governed firms than for poorer-governed firms and thus we expect the coefficient estimate on Gov to be negative. We mitigate the potential bias caused by omitted unobservable industry characteristics by including industry fixed effects in the regressions (using Fama-French (1997) 48 industry indicators). We also include year fixed effects to control for potential time trend effects. To incorporate the guidance suggested by Petersen (2009) regarding the use of panel data sets, we also use clustered standard errors at the individual firm level to calculate t- statistics. As robustness tests, we calculate t-statistics using clustered standard errors at the individual insider level and find qualitatively similar results. 4. Empirical Results 4.1. Internal governance and profitability of insider trading To examine whether the profitability of insider trading is related to the effectiveness of a firm s governance, we estimate equation (2) using ordinary least squares (OLS) regressions in which the dependent variable is BHAR6m. To examine the differential effects of internal governance on insider purchases and sales, we run the regressions separately for these two types of insider trades. We also run the regressions separately for six alternative governance measures: board independence, compensation structure, institution ownership, ICG, ICG1, and log (ISS). The results are reported in Table 3. We find that the coefficient estimates on all six governance measures are insignificant for purchase transactions but significantly negative at least at the 5% level for sales transactions, suggesting that better internal governance significantly reduces the profitability of insider sales but not that of insider purchases. In other words, the results imply that better internal governance discourages insiders from exploiting negative private information, but not from exploiting positive private information. We also find 16

19 that the effect of governance on the profitability of insider sales is economically large and significant. For example, when ICG is used as the governance measure, its coefficient estimate is for sales transactions, suggesting that one standard deviation increase in ICG (3.061) is associated with 1.79% lower abnormal returns earned by insiders from their sales transactions. Among control variables, we find that the coefficient estimates on log (Size) are significantly negative only for purchase transactions. On the other hand, the coefficient estimates on MB and Loss are significantly negative for purchase transactions while they are significantly positive for sales transactions. The coefficients estimates on ITP and GC are not significant for purchase transactions whereas they are significantly negative for sales transactions. These results suggest that internal governance matters for the informativeness of insider sales above and beyond the effects that insider trading policies and general counsel pre-approval requirements have on the profitability of insider trading Sensitivity analyses To check the robustness of the above results, we perform several additional tests. We first examine whether our results are robust to alternative model specifications. The results are reported in Panel A of Table 4. To save the space, we report only the coefficient estimates on ICG. Using ICG1 or log (ISS) does not change the results. First, instead of using the marketadjusted excess return (BHAR6m), we use the alpha estimate from Carhart s (1997) four-factor model (Alpha6m) as the dependent variable. Our results do not change. Second, we replace Dispersion with alternative transparency measures, the number of analysts following (log (Following)) and earnings quality. Earnings quality is measured as the absolute 16 Abnormal performance of insider sales is likely to be underestimated if firms are excluded from the sample due to significant price drops following insider sales. This sample selection bias problem is not a major concern for our study since our sample includes both active and inactive S&P 1500 companies as explained earlier. 17

20 value of discretionary accruals estimated using the modified Jones model (Dechow, Sloan, and Sweeney 1995). Again, the results remain similar. Third, we examine whether our results hold for the subsample of insider trades that took place after the 2002 enactment of the Sarbanes-Oxley Act (SOX). Since the requirements for corporate governance practices as well as insider trading regulations were significantly strengthened after the adoption of SOX, it is possible that both variation in governance quality and profitable insider trading opportunities have decreased during the post-sox period. 17 To address this issue, we reestimate regressions using only observations in the post-sox (2003 to 2011) period and find that for insider sales, the coefficient estimate on ICG remains negative and significant, suggesting that firm-level differences in governance quality affect the profitability of insider sales even after the change in regulatory environments. For insider purchases, the coefficient estimate on ICG is not significant. 18 Fourth, we examine whether the cross-sectional dependence that arises when different insiders of a firm simultaneously purchase (sell) stocks several times on the same trading date and these trades are counted as separate observations in the regression leads to biased estimation of t-statistics. To do so, we aggregate the trades made on the same date by all insiders of the same firm and then rerun regressions. 19 We find that the results are qualitatively similar to those reported in Table Section 403 of the Sarbanes-Oxley Act specifies an amendment to Section 16 of the Securities Exchange Act of 1934, effective as of August 29, Insiders are required to report their transactions to the SEC within two business days (rather than within ten days following the trading month, which used to be the case prior to the amendment). 18 The results for the pre-sox period are similar to those for the post-sox period. 19 If the number of shares purchased (sold) during the day is greater than the number of shares sold (purchased) during the same day in a given firm, we define the aggregate transaction as an insider purchase (sale) for this firm. 18

21 Fifth, to address potential endogeneity bias caused by omitted unobservable firm and insider characteristics, we include firm and insider fixed effects, respectively, in the regressions. We find that our results do not change. Sixth, we examine whether the changes in ICG, rather than the levels of ICG, matter by replacing the level of ICG with the change in ICG. Consistent with the previous results, we find that the coefficient estimate on ICG is negative and significant at the 5% level for sales transactions while it is positive but insignificant for purchase transactions. Thus, the profitability of inside sales is affected by both the levels of and changes in ICG. 20 In Panel B, we test whether the role of governance in discouraging insiders from exploiting private information is different between firms that adopt insider trading policies (ITP and GC) and those that do not by including the interaction terms between ICG and ITP (GC) indicators. We find that for sales transactions, the coefficient estimates on ICG and its interactions with ITP and GC are negative and significant at least at the 5% level, indicating that internal governance discourages insiders from exploiting negative private information and its effects become stronger when accompanied by insider trading policies and general counsel pre-approval requirements. Finally, we examine whether internal governance affects the profitability of insider trading made by blockholders. Given that internal governance cannot control the transactions made by outside large shareholders, we expect that ICG does not affect the profitability of insider trading made by them. To address this issue, we add the transactions made by large shareholders to the sample and reestimate the regressions by including the following five variables: ICG*Officer, ICG*Director, ICG*Blockholder, Officer, and Director, where Officer, Director, and 20 In untabulated univariate analyses, we compare abnormal returns earned by insiders of firms with an above median ICG score (change in ICG score) and those earned by insiders of firms with a below median ICG score (change in ICG score). Consistent with the regression results, we find significant differences in abnormal returns between the two groups only for sales transactions. The average differences in ICG scores and the changes in ICG scores between the two groups are about 3, which is close to the standard deviation of ICG (3.061). 19

22 Blockholder are indicators for transactions made by officers, directors, and blockholders, respectively. We use the same control variables as those used in the Table 3 regressions. The results are reported in Panel C of Table 4. As expected, we find that ICG matters only for sales transactions made by officers and directors but not for those made by blockholders. Overall, these results suggest that our key results in the previous section are robust for an alternative measure of the profitability of insider trading and a variety of model specifications Internal governance versus GIM index As discussed earlier, Ravina and Sapienza (2010) use the GIM index as a measure of the effectiveness of a firm s governance and find that insiders earn more profits from both purchase and sales transactions when their firms have a higher GIM index (i.e., poorer corporate governance). However, given that the GIM index mainly captures the extent to which the firm adopts antitakeover provisions and thus is less likely to be effective in controlling the incentives of officers and directors to exploit private information than internal governance measures, it is still an open question as to whether internal governance matters for the profitability of insider trading. To address this issue, we first replicate the results of Ravina and Sapienza (2010) using the same regression model (i.e., GIM index as a measure of the quality of a firm s corporate governance and no risk control), the same sample period ( ), and the same sample transactions (transactions made by officers, directors, and blockholders) as theirs. 21 We use BHAR6m as the dependent variable. Although Ravina and Sapienza (2010) report the regression results only for purchase transactions, we estimate the regressions for both purchase and sales 21 Although Ravina and Sapienza (2010) do not explicitly mention, we conjecture that their sample period starts from 1990, not from 1986, since the data needed to construct the GIM index in their Table 6 are available from Therefore, our analysis that replicates their results uses the sample that starts from

23 transactions. The results are reported in columns (1) and (5) of Table 5. The number of observations for purchase transactions is close to theirs (36,333 compared with 37,263) and the results are qualitatively similar: insider purchases are more profitable to insiders of poorergoverned firms than those of better-governed firms when we use the GIM index to capture the effectiveness of corporate governance. 22 We find similar results for insider sales. Next, we reestimate regressions in columns (1) and (5) using the subsample period, , during which the data on both the GIM index and ICG are available. The results reported in columns (2) and (6) are almost identical to those in columns (1) and (5), respectively. In columns (3) and (7), we replace the variables including the GIM index in columns (2) and (6) with those including ICG, i.e., ICG and its interactions with the indicators for independent directors and blockholders. We find that ICG matters only for sales transactions. In columns (4) and (8), we include all variables used in the previous regressions: GIM index, the interaction terms between GIM index and the indicators for independent directors and blockholders, ICG, and the interaction terms between ICG and the indicators for independent directors and blockholders. We find that the coefficient estimate on ICG in column (8) is significantly negative at the 1% level, suggesting that the profitability of insider sales is significantly smaller for firms with better internal governance than for firms with poorer internal governance. On the other hand, the coefficient estimate on ICG in column (4) is insignificant, suggesting that the profitability of insider purchases is statistically indistinguishable between firms with better internal governance and those with poorer internal governance. These results suggest that the role of internal governance in limiting insiders ability to exploit private 22 Following Gompers, Ishii, and Metrick (2003), we assign firms into ten groups according to their GIM index levels. GIM6 through GIM13 are indicators for firms with a GIM index of 6 through 13, respectively. GIM14 is an indicator for firms with a GIM index greater than or equal to 14. We omit the indicator for firms with a GIM index less than or equal to 5, so that the intercept captures abnormal returns for firms with a GIM index less than or equal to 5. 21

24 information is different from that of GIM index. To the extent that informed insider trading generally reflects the lack of an internal system that controls managerial incentives, which can be better managed by internal governance mechanisms rather than by antitakeover protection as measured by GIM index, our results provide more relevant evidence that internal governance plays an important role in limiting insiders ability to profit from their information advantage Alternative approaches to measure the informativeness of insider trading To check whether our results are robust to using an alternative approach to measure the profitability of insider trading, we use a time-series portfolio approach to measure abnormal returns. Specifically, in each month between July 1997 and December 2011, we count the number of shares purchased (sold) by insiders of each sample firm during the month, and then assign a sample firm into a net purchase (net sales) group if the number of shares purchased (sold) by insiders of the firm is greater than the number of shares sold (purchased). In each month starting from January 1998, we then form net purchase (sale) portfolios using all firms classified in the net purchase (sales) group at least once over the past six-month period. We also separately construct two portfolios according to the sample median of firms ICG: a portfolio of firms with high ICG and a portfolio of firms with low ICG. We calculate both value- and equally-weighted returns on each portfolio during the month of portfolio formation. Next, using net purchase (sales) portfolios, we form a zero-cost investment (i.e., hedge portfolio) strategy that is long in the portfolio of firms with high ICG and short in the portfolio of firms with low ICG and compute the average return on this hedge portfolio. Finally, using these monthly hedge portfolio returns (168 observations), we run time-series regressions of the Carhart (1997) four-factor model as described in equation (1). 23 The correlation coefficient between ICG and GIM index is small (0.055). 22