CEO Inside Debt and Insider Trading

Transcription

1 CEO Inside Debt and Insider Trading Eric R. Brisker The University of Akron Dominique Gehy Outlaw Hofstra University Aimee Hoffmann Smith Bentley University September 14, 2017 Abstract Managerial compensation theory suggests that both equity- and debt-type compensation should be included in the optimal compensation contract in order to align managers interests with both shareholders and debtholders of the firm. However, this also suggests that these two forms of compensation are directly in conflict with each other in that shareholders (debtholders) should react negatively to debt-type (equity-type) compensation. In this article, we consider how firm insiders react to CEO debt-type compensation by examining insider trading. We find that higher CEO debt-type compensation is associated with reduced selling of shares by firm insiders. This finding is robust to using two different measures of debt-type compensation as well as several different net and gross measures of insider selling, including definitions based on opportunistic trades and trades by directors and officers only. Further, we alleviate endogeneity concerns with the robust results of the instrumental variables tests we employ. Our results suggest that wellinformed insiders perceive debt-type compensation to provide a positive benefit to shareholders. JEL Classification: G14, G30, G32, G33, G34 Keywords: inside debt, insider trading, compensation, agency conflict 1

2 1. Introduction Corporations face two well-known agency conflicts: (1) the conflict between managers and shareholders and (2) the conflict between shareholders and debtholders. The former originates from a deficiency in the separation of ownership and control, while the latter is triggered by asset substitution (i.e., risk-shifting), whereby low-risk assets are exchanged for high-risk assets, resulting in the transfer of wealth from debtholders to shareholders. Jensen and Meckling (1976) first proposed a remedy for these agency conflicts through managerial compensation structure. To alleviate the conflict between managers and shareholders, firms could use equity-type compensation, such as stock and stock options, to align managers interests with those of shareholders. To alleviate the conflict between debtholders and shareholders, which now include managers, firms could use debt-type compensation, such as pensions and deferred compensation, to incentivize managers to take debtholders interests into account. 1 Theoretically, a manager whose compensation consists of both equity and debt equal to the firm s debt-to-equity ratio would consider the interests of both shareholders and debtholders equally. In this article, we empirically investigate how CEO inside debt holdings affect insider selling behavior. The incentive effects offered through equity compensation are widely expected to reduce the conflict between managers and shareholders, but the incentive effects offered through debt-type compensation appear to exacerbate the conflict between shareholders and debtholders. Given insiders unique information set related to the corporate policy decisions of the CEO, insiders offer us the ability to see how knowledgeable shareholders react to CEO inside debt holdings based on the perceived advantages and disadvantages they offer to shareholders, 1 Sundaram and Yermack (2007) suggest that certain forms of compensation widely used in practice, such as pension and deferred compensation, have debt-type payoffs that can be viewed as inside debt. In this paper, we use debttype compensation and inside debt interchangeably. 2

3 particularly with respect to the impact of inside debt on the conflict between shareholders and debtholders. Edmans and Liu (2011) provide foundational theory for deriving optimal compensation contracts for managers facing effort and investment (i.e., risk preference) choices that include both firm equity and debt. They first show that equity provides a solution to the manager-shareholder agency cost of equity in that it incentivizes more effort from the manager (i.e., it reduces shirking). They then show that debt provides a solution to the shareholder-debtholder agency cost of debt conflict in that it incentivizes managers to avoid risk-shifting and transfers of wealth away from debtholders toward shareholders. In particular, because the value of debt compensation hinges not only on the probability of bankruptcy, but also on the liquidation value if the firm fails, a manager holding a similar percentage of inside debt as the firm would have no incentive for asset substitution because he or she generally bears the same default risk that is faced by the firm s other unsecured creditors. The effect of debt compensation on the alignment of managerial interests with debtholders versus shareholders is in need of more in-depth explorations, especially given that debt and equity compensation appear to offer incentives to managers that are directly opposed to each other in many cases. For example, Wei and Yermack (2011) and Anantharaman, Fang, and Gong (2014) document a negative association between CEOs inside debt and the cost of debt financing, suggesting that debt-type compensation reduces agency costs of debt. These findings highlight the fact that debt compensation could actually benefit shareholders since it lowers the cost of debt financing and may allow otherwise financially constrained firms the ability to access debt markets. However, managerial inside debt holdings have also been found to lower managers risk appetite in setting corporate policies, such as R&D expenditures and financial leverage (Cassell et al., 2012), 3

4 as well as participating in mergers and acquisitions (Phan, 2014). This reduced appetite for risk may be in the best interests of debtholders but directly opposed to the interests of shareholders. The standard view in the managerial risk-preference and compensation literature is that undiversified firm managers are more risk-averse in setting firm policies than diversified shareholders would desire (Hirshleifer and Thakor, 1992; Klein and Coffee, 1996). The use of equity-type compensation in the optimal compensation contract is designed to encourage managers to not only exert more effort in managing the firm, but also offer risk-shifting incentives that lead to higher returns on equity, which benefits shareholders (Guay, 1999; Coles, Daniel, and Naveen, 2006; Low, 2009). However, given that managers receiving debt compensation face the same asymmetric payoff structure as outside debtholders (i.e., they receive fixed payoffs if the firm is successful but bear the loss proportionally if the firm fails), managers inside debt holdings incentivize them to act more in the interests of debtholders. This suggests that compensating managers with inside debt could neutralize some of the beneficial risk-shifting effects of equitytype compensation and lead managers to be excessively conservative (Bebchuk and Jackson, 2005). In summary, while debt-type compensation can be used to mitigate the shareholder-debtholder conflict, it could also exacerbate the divergence in attitude toward risk between managers and shareholders. As a result, managers may have the tendency to adopt excessively conservative business policies in the interest of debtholders, which actually compromises shareholders wealth. In determining the impact of CEO inside debt on insider selling, we define insiders as the officers, directors, and beneficial owners of more than 10 percent of a class of the firm s stock who are required to file Form 4 with the Securities and Exchange Commission for every transaction. If insiders believe that inside debt further misaligns the CEO s interests towards debtholders interests and away from shareholders interests, then it is likely that more insiders will sell their 4

5 personal shares as inside debt levels rise. Alternatively, if insiders believe that inside debt not only offers incentives to CEOs to better align their interests with debtholders, but also offers positive benefits to shareholders, then it is likely that fewer insiders will sell their personal shares as inside debt levels rise. We focus on trading by insiders for a few different reasons. First, insiders are an attractive subset of investors to examine because they have more information and devote more attention to the firm than do typical shareholders. For instance, Alldredge and Cicero (2015) find that insiders incorporate public information about their principal customers (i.e., those who comprise at least 10 percent of their sales or profits) in deciding when to buy and sell their company s shares. However, Cohen and Frazzini (2008) show that the typical investor does not. Trading on news that directly impacts linked customers and suppliers yields a profitable trading strategy because the typical investor does not pay attention to the linked firm. Not surprisingly, insiders pay closer attention. Thus, in the case of inside debt, it is reasonable to expect that insiders are more attentive to changes in the CEO s compensation structure, including those affecting inside debt holdings, as well as the implications of those changes for the CEO s strategy and incentives in risk-taking. Another reason for our focus on the subset of trades by insiders is that typical investors who are unaffiliated with the firm can benefit from a better understanding of insider trading activities. In fact, several studies document that in addition to insiders ability to earn abnormal profits on their company s stock, outsiders can also earn abnormal profits by mimicking insiders trades (Lorie and Neiderhoffer, 1968; Jaffe, 1974; Rozeff and Zaman, 1988). Piotroski and Roulstone (2005), along with several other studies, report predictability in insiders trades about the firm s future performance (Seyhun, 1999; Lakonishok and Lee, 2001; Beneish and Vargus, 2002; Jagolinzer, 2009). Even the size and frequency of the trades further disentangles insiders 5

6 trading motives. For example, Scott and Xu (2004) find that smaller sales, which are most likely motivated by liquidity and diversification, are positively associated with future firm performance. Cohen, Malloy, and Pomorski (2012) stratify the insiders trades based on two motivations: diversification and liquidity versus opportunistic trading. They report that opportunistic trades yield a profitable trading strategy. Gao and Ma (2015) find that the absence of insider transactions also contains important information about firm performance because insiders are least comfortable selling shares when they anticipate bad news about the firm. For example, insiders are unwilling to sell shares prior to shareholder lawsuits and large stock price drops in order to avoid exposure to securities litigation risk. Finally, empirical evidence suggests that insiders are more likely to trade in firms that have certain characteristics. For instance, Rozeff and Zaman (1988) show that the anomalous trading strategy of mimicking insiders trades is partially explained by the size and earnings-to-price factors. Their later 1998 article documents that insiders are more likely to sell growth stocks and stocks that experience a recent price run-up. In considering insider selling, it is reasonable to expect that the opinions of well-informed, attentive insiders at firms carrying inside debt will be reflected in their personal trades. If insiders believe that inside debt is detrimental to shareholders, then we would expect insiders at firms with high inside debt holdings to be net sellers of their firm s shares. This outcome might transpire if inside debt is believed to intensify the manager-shareholder agency conflict by creating a misalignment between the risk-taking propensity of the CEO and the level of risk-taking desired by shareholders. Furthermore, it may occur if shareholders are concerned about the possibility that inside debt will mitigate the shareholder-debtholder agency conflict by inhibiting asset substitution, which would reduce the probability of a welcomed shift toward riskier investments and a desired transfer of wealth from debtholders to shareholders. Alternatively, we would expect insiders to be 6

7 net buyers of shares in firms with high inside debt holdings if inside debt is thought to better shareholders interests, perhaps by improving the availability and affordability of debt financing for the firm. We employ two different measures of managerial inside debt holdings in order to examine the impact of inside debt on insider selling. Our first measure is the CEO s Debt-to-Equity Ratio, defined as the accumulated holding of pensions and deferred compensation divided by total equity holdings, which include stocks, stock options, and restricted shares. We also use a second measure of managerial inside debt holdings, the CEO s Inside Debt Ratio, which is defined as the accumulated holding of pensions and deferred compensation divided by the CEO s total compensation, which includes current salary and bonus, inside debt holdings (i.e., pensions and deferred compensation), and equity holdings (i.e., stocks, stock options, and restricted shares). Using a sample of U.S. firms during the period from 2006 through 2013, we find that the mean CEO s Debt-to-Equity Ratio ranges from a low of in 2006 to a high of in However, the median is considerably smaller, equaling less than 0.05 in all years except for The CEO s Inside Debt Ratio exhibits similar patterns over time; the mean ranges from a low of in 2006 to a high of in 2008, while the median ranges from a low of in 2011 to a high of in We also consider two different measures of insider selling, which we refer to as the Net Insider Selling Ratio and the Gross Insider Selling Ratio. The first of these measures was inspired by Lakonishok and Lee s (2001) net purchase ratio and is defined as the difference between the number of shares sold and purchased by insiders during the month, divided by the total number of shares sold or purchased by insiders during the month. Barber and Odean (2008) use a similar variable to measure buy-sell imbalance among retail investors. Our gross measure of insider selling 7

8 is computed following Khan, Kogan, and Serafeim (2012) by dividing the number of shares sold by insiders during the month by the total number of shares sold or purchased by insiders during the month. A similar variable appears in Rozeff and Zaman (1998) and Piotroski and Roulstone (2005), although that measure captures the proportion of insider trades that are purchases rather than sales. Over our sample period, we find that mean Net Insider Selling Ratio is and Gross Insider Selling Ratio is We independently examine the effects of our two measures of inside debt on our two measures of insider selling using a variety of empirical tests. For example, we conduct a univariate analysis in which we compare the net and gross measures of insider selling across the top and bottom quartiles based on each measure of inside debt. Furthermore, we explore a number of multivariate specifications in order to investigate the impact of inside debt on insider selling after controlling for insiders typical trading behaviors, firm financials, and an extensive array of characteristics and compensation measures pertaining to the CEO. We also include year and industry fixed effects as well as robust White (1980) standard errors clustered by firm. Our initial tests consistently indicate a negative relation between inside debt holdings and insider selling. We find that insiders tend to sell fewer shares, or perhaps even increase their equity holdings, when they observe a rise in inside debt, suggesting that firm insiders perceive inside debt positively. This finding supports the notion that insiders are receptive to the issuance of inside debt because they believe it benefits shareholders, possibly due to the fact that it reduces the cost of debt financing and provides financially constrained firms access to debt markets. The result also suggests that insiders perceived benefits of inside debt outweigh their concerns regarding its potential to exacerbate the agency conflict between shareholders and managers by encouraging managers to adopt excessively risk-averse strategies. Furthermore, insiders do not appear to be 8

9 particularly concerned about the possibility that inside debt will mitigate the agency conflict between shareholders and debtholders by inhibiting asset substitution, thus reducing the likelihood of a sought-after shift towards riskier investments and a transfer of wealth from debtholders to shareholders. In additional tests, we follow Cohen, Malloy, and Pomorski (2012) and differentiate opportunistic insider selling from routine insider selling in our empirical analyses. In these tests, we find that only opportunistic selling is negatively impacted by inside debt levels. This suggests that only opportunistic trades contain incremental information about insiders knowledge and expectations regarding inside debt. Routine trades, on the other hand, are not affected by inside debt and do not contain information about insiders beliefs about inside debt levels. This further highlights the fact that insiders react positively on average to inside debt levels when making nonroutine trades of their firm s shares. We also conduct additional tests that consider insider selling by the firm s directors and officers only rather than all individuals classified as insiders by the SEC. We exclude trades by beneficial owners because directors and officers should theoretically have greater insight regarding the effects CEO inside debt compensation due to their superior access to inside information. Our initial results show that inside debt has an even stronger negative impact on insider selling when we focus on trades by directors and officers only as opposed to trades by all insiders. Thus, directors and officers also view the use of CEO inside debt positively in regard to shareholders interests. When we consider opportunistic versus routine trades by directors and officers, we again find that CEO inside debt negatively impacts opportunistic trading only. Routine trading by directors and officers is not impacted by CEO inside debt levels. It is possible that endogeneity may arise as unobserved firm and/or CEO heterogeneity 9

10 determines both CEO inside debt and insider trading. We first address this possible endogeneity issue of CEO inside debt by controlling for a number of factors in our ordinary least squares (OLS) regressions. Specifically, we control for fixed industry and time effects, past insider trading activity, and an extensive array of firm and CEO characteristics in an effort to eliminate any omitted variable bias. Secondly, we estimate our specifications using a two-stage least squares (2SLS) instrumental variables model with the maximum state tax rate on individual income and the median industry inside debt ratio as instruments (Cassell et Al., 2012). Overall, our results are qualitatively similar even when employing these instruments. The incorporation of inside debt in the CEO compensation structure is an emergent phenomenon about which many questions remain. Using insiders trading activity as a lens through which to examine their perceptions of inside debt, we contribute to an underdeveloped yet growing literature on the effects of this recent practice. To the best of our knowledge, no prior study has directly examined how inside debt holdings impact shareholders conflicts with managers and debtholders or how their use is perceived by insiders, as reflected by their trading activity. Our research also offers guidance to practitioners regarding the optimal compensation structure for CEOs in light of potential agency conflicts. The remainder of the article is organized as follows. In Section 2, we review the related literatures on inside debt and insider selling. We then describe our data and methodology in Section 3. In the following section, we discuss the results of our empirical analyses. Finally, we summarize our findings and provide concluding remarks in Section 5. 10

11 2. Related Literature 2.1 Inside Debt In explaining the agency cost of debt, Jensen and Meckling (1976) argue that managers who are compensated with securities having a payoff structure similar to that of equity are incentivized to increase firm risk beyond the level preferred by debtholders and take actions that transfer wealth from debtholders to shareholders through asset substitution. They theorize that compensating managers with securities having a payoff structure similar to that of debt, often referred to as inside debt, could assist in reducing this agency cost of debt. Edmans and Liu (2011) provide foundational theory regarding the use of debt compensation in the optimal compensation contract to effectively address agency cost of debt issues. They argue that since creditors are concerned with both the probability of bankruptcy and liquidation values when bankruptcy occurs, optimal compensation contracts should be sensitive to such matters. When bankruptcy occurs, debt compensation offers managers a positive payoff that is proportional to the liquidation value of the firm. This incentivizes managers not only to avoid risk-shifting and transfers of wealth away from debtholders toward shareholders, but also to exert greater effort in protecting, or maximizing, liquidation values when bankruptcy becomes more likely. Debt compensation also benefits shareholders since it lowers the cost of debt financing and may allow otherwise financially constrained firms to access debt markets. Edmans and Liu (2011) then derive an optimal compensation contract that incorporates both equity and debt compensation. Equity compensation is used as a solution to the managershareholder agency cost of equity problem since it incentivizes managers to exert more effort (i.e., it reduces shirking). Debt compensation, on the other hand, is used as a solution to the shareholderdebtholder agency cost of debt problem since it incentivizes managers to avoid risk-shifting 11

12 behavior. Finding the optimal mix of equity and debt compensation to be used in the compensation contract is complicated due to the potentially conflicting incentives being offered through each of these components. Equity compensation incentivizes effort towards risk-shifting and the exploitation of growth opportunities, whereas debt compensation incentivizes effort towards riskshifting avoidance and the maximization of bankruptcy liquidation values. In most cases, the optimal compensation contract entails an equity bias in which managers equity stake percentage exceeds their debt stake percentage. This is because equity is more effective than debt in incentivizing managerial effort even though it also incentivizes risk-shifting. However, a debt bias may be optimal in cases where bankruptcy is likely or when effort is expected to be productive towards enhancing liquidation values. In more recent years, there has been a considerable increase in research focused on debt compensation due to the pervasive use of inside debt in large U.S. corporations (Bebchuk and Jackson, 2005) as well as the SEC disclosure rule that went into effect in 2006 and mandates detailed disclosure of executive holdings of pensions and deferred compensation. Sundaram and Yermack (2007) document a positive association between CEO inside debt holdings and the distance to default, indicating that inside debt moderates CEOs risk-shifting tendency, while Wei and Yermack (2011) report that inside debt holdings decrease (increase) the firm s cost of debt (equity) and decrease the firm s market risk levels (Wei and Yermack, 2011). Other empirical studies find that inside debt not only reduces accounting conservatism (Wang, Xie, and Xin, 2014) and decreases the riskiness of the firm s investment and financing policies (Cassell et al., 2012), but also lowers borrowing costs and discourages the use of debt covenants (Anantharaman et al., 2014). Furthermore, prior research shows that inside debt holdings are positively associated with firm cash holdings (Liu, Mauer, and Zhang, 2014) and M&A announcement abnormal bond 12

13 returns (Phan, 2014), and negatively associated with cash holding value (Liu et al., 2014) and M&A announcement abnormal stock returns (Phan, 2014). 2.2 Insider Selling The Securities and Exchange Commission requires all company insiders officers, directors, and beneficial owners of more than 10 percent of a class of the firm s stock to file Form 4 for any transaction involving the company s stock. In our study, the information content revealed in insiders trades is of particular importance because it allows us to examine whether insiders are net sellers or purchasers of their firms shares, which sheds light on insiders viewpoint on whether the CEO s risk-taking is more or less aligned with the interests of shareholders. Several studies document the abnormal returns that insiders earn on their trades and show how investors can mimic their trades to yield a profitable trading strategy (Lorie and Neiderhoffer, 1968; Jaffe, 1974; Rozeff and Zaman, 1988; Bettis, Vickeray, and Vickeray, 1997; Seyhun, 1999; Beneish and Vargus, 2002; Jagolinzer, 2009; Cohen et al., 2012). Further, insiders display tendencies to trade certain types of stocks more often than others. Piotroski and Roulstone (2005) document that insiders transactions provide incremental information about the firm s past and future financial performance. They find that insiders trades are positively related to the firm s book-to-market ratio and future earnings. Lakonishok and Lee (2001) find that firms with overall net buying by insiders experience positive long-run performance, while firms with overall net selling by insiders perform poorly. Their results are driven by small firms, growth firms, and glamour stocks that experienced a recent run-up in stock price. Most recently, Agrawal and Cooper (2015) present evidence that illegal insider trading is more prevalent than the literature suggests. Using a sample of 500 companies that faced accounting scandals, they find that managers sold significantly more stocks during restating periods. Their findings suggest that managers who 13

14 are more likely to sell their shares at overvalued prices are also more likely to engage in earnings manipulation. However, illegal trading on private information is not the focus of our study. It is reasonable to conclude that insiders trades provide incremental information about the company because insiders have access to more information and pay greater attention to the firm s performance than do most shareholders. Consistent with this notion, Alldredge and Cicero (2015) find that insiders profitably trade based on publicly available information. More specifically, insiders trade their personal shares based on public information about their company s principal customers (i.e., customers who comprise at least 10 percent of sales or profits). This attentiveness to publicly available information regarding the supply chain market constitutes a major distinction between insiders and typical shareholders. For example, Cohen and Frazzini (2008) document the customer-momentum strategy, which exploits return predictability in economically-linked customers and suppliers and is attributable to the limited attention displayed by typical investors. Additional studies further investigate the information content of insider trades by comparing the size and frequency of the trades to determine whether they are motivated by diversification and liquidity needs or profitable trading opportunities. Scott and Xu (2004) find that the size of insiders trades is important in determining return predictability. They find that smaller sales, which are most likely driven by attempts to improve liquidity and diversification, are positively associated with future firm performance. Small sales suggest that the insider is optimistic about the stock, yet needs to sell shares for liquidity and diversification reasons. On the contrary, large sales are associated with negative future performance. Gao and Ma (2015) find that the absence of insider trading also contains important information about firm performance. Insiders feel uncomfortable selling shares when they anticipate bad news about the firm because 14

15 they fear allegations of securities fraud. To avoid litigation risk, they are unwilling to sell shares prior to shareholder lawsuits and large stock price drops. To further investigate the information content of insiders trades, the literature also examines whether impositions to insider trading, such as corporate self-regulating policies and the presence of short sellers, impact the strategies and profitability of insiders. For example, insiders of firms that have restrictive trading policies, such as blackout periods where insiders are not allowed to trade, exhibit significantly lower trading profitability (Carr, Coles, and Lemmon, 2000). Khan and Lu (2015) find that short selling is higher prior to large insider sales, but not prior to small insider sales, especially for firms with low accounting quality. These patterns suggest that large insider sales contain information about the firm s future performance and reinforce short sellers negative perceptions of the firm. Using the Regulation SHO 2 pilot test as a proxy for short sale constraints, Massa, Qian, Xu, and Zhang (2015) confirm that the presence of short sellers accelerates insiders sales. Thus, insiders pay closer attention to the performance of their principal customers and the trades of sophisticated investors, and they trade accordingly. Overall, there is a comprehensive literature that captures the information content and return predictability in insiders transactions. However, to the best of our knowledge, no study has examined the effect of inside debt on insider trading. We expect insiders trades to shed light on whether inside debt is beneficial or detrimental for firms. 2 Regulation SHO was implemented by the Securities and Exchange Commission (SEC) on January 3, 2005 in order to restrict naked short selling, whereby the seller does not arrange to borrow the securities in time to deliver them to the buyer within the allowed three-day settlement period ( 15

16 3. Data and Methodology 3.1 Variables Main variables. The key variables in this study measure the CEO s inside debt holdings and insider selling. The literature provides a few measures of managerial inside debt holdings (Wei and Yermack, 2011; Cassell et al., 2012). Since our purpose is to examine the link between CEO inside debt holdings and insider selling, we focus on measures of inside debt that are specifically related to the CEO rather than the firm itself. Our first measure, the CEO Debt-to-Equity Ratio, is defined as the sum of the present value of accumulated pension and deferred compensation benefits divided by total equity holdings, which include stocks, stock options, and restricted shares. The value of stock is equal to the number of shares, including restricted shares, held by the CEO multiplied by the stock price at the firm s fiscal year end. Following the literature, particularly Cassell et al. (2012), we estimate the value of each individual tranche of options the CEO holds, including both exercisable and unexercisable options, using the Black-Scholes option model (Black and Scholes, 1973) adjusted for dividends based on Merton (1973). We then find the sum of these option tranche values to form the total value of option holdings for the CEO. 3 Our second measure of inside debt, the CEO Inside Debt Ratio, is defined as the sum of the present value of accumulated pension and deferred compensation benefits divided by the CEO s total compensation, which is the sum of current salary and bonus, inside debt holdings (i.e., pensions and deferred compensation), and equity holdings (i.e., stocks and options). Both of these inside debt measures are winsorized at the 1% and 99% levels. 3 For details, see Cassell et al. (2012), Appendix A. 16

17 Our first measure of insider selling is a net measure motivated by Lakonishok and Lee s (2001) net purchase ratio. We compute the Net Insider Selling Ratio as the difference between the number of shares sold and purchased by insiders during the month divided by the total number of shares sold or purchased by insiders during the month. Following Jeng, Metrick, and Zeckhauser (2003), we restrict the transactions to open-market transactions, excluding private transactions as well as those involving the exercise of options. For robustness, we also construct an alternative gross measure of insider selling. Following Khan et al. (2012), we define the Gross Insider Selling Ratio as the number of shares sold by insiders during the month divided by the total number of shares sold or purchased by insiders during the month. Since the executive compensation data is provided on an annual basis, we compute the mean level of insider selling over the 12 months following the fiscal year end associated with the compensation variables. For example, if a CEO s compensation is reported as of the firm s 2009 fiscal year end on August 31 st of 2009, then the insider selling measures reflect the mean level of monthly insider selling during the 12 months from September of 2009 through August of Control variables. In our multivariate analyses, we control for a wide array of characteristics and compensation measures pertaining to the CEO, firm financial variables, and other known determinants of insider selling. The first CEO characteristic for which we control is CEO Age. We control for this variable in light of Sundaram and Yermack s (2007) finding that inside debt increases with the age of the CEO. Serfling (2014) also documents a negative relation between CEO age and firm riskiness, indicating that older CEOs may avoid the use of debt. We separately control for CEO Tenure to acknowledge the argument that CEOs with longer tenures are more likely to be entrenched and therefore exhibit greater levels of risk aversion (Berger, Ofek, and 17

18 Yermack, 1997; Coles et al., 2006). We include a New CEO Flag to control for the possibility of elevated insider selling activity related to CEO turnover. Finally, we include the CEO Vega/Delta Ratio to control for potential effects of equity-type compensation on insider selling. The Vega measures the sensitivity of the CEO s equity-type holdings to a one-percentage change in the volatility of the firm s stock price. The Delta measures the sensitivity of the CEO s equity-type holdings to a one-percent change in the firm s stock price. Following Cassell et al. (2012), we use the CEO Vega/Delta Ratio as a parsimonious measure of equity incentives. We also control for a host of firm financials, all of which have appeared repeatedly in the insider selling literature. 4 In light of the fact that insiders prefer to trade small stocks, we control for firm size using the Market Capitalization, defined as the number of common shares outstanding multiplied by the closing stock price at the firm s fiscal year end (Seyhun, 1986; Rozeff and Zaman, 1988; Jeng et al., 2003; Core et al., 2006; Khan et al., 2012). We also control for the Market-to- Book ratio, computed as the firm s market capitalization divided by the book value of the firm s common equity, because several studies have linked insider trading to stock misvaluations and have demonstrated insiders propensity to purchase value stocks and sell growth stocks (Rozeff and Zaman, 1998; Lakonishok and Lee, 2001; Beneish and Vargus, 2002; Jeng et al., 2003; Piotroski and Roulstone, 2005; Jenter, 2005; Khan et al., 2012). Lagged 6-Month Returns, defined as the cumulative stock return using monthly returns over the last half of the fiscal year, is included to acknowledge the tendency of insiders to sell (buy) securities after periods of positive (negative) returns (Seyhun, 1986; Seyhun, 1999; Rozeff and Zaman, 1998; Lakonishok and Lee, 2001; Jeng et al., 2003; Khan et al., 2012). Following Massa et al. (2015), we also include Leverage, Stock 4 See Seyhun (1986), Rozeff and Zaman (1988, 1998), Lakonishok and Lee (2001), Beneish and Vargus (2002), Jeng et al. (2003), Jenter (2005), Piotroski and Roulstone (2005), Core, Guay, Richardson, and Verdi (2006), Khan et al. (2012), and Massa et al. (2015). 18

19 Return Volatility, and Share Turnover to control for the possibility that insider selling is affected by the firm s debt levels, total risk, or stock liquidity. Leverage is defined as the sum of the firm s current and long-tern non-convertible debt divided by total assets, while Stock Return Volatility is computed as the standard deviation of daily stock returns during the fiscal year. Additionally, Share Turnover equals the sum of the firm s monthly share trading volume during the fiscal year divided by the number of common shares outstanding at the firm s fiscal year end. Finally, we follow Core et al. (2006) and Khan et al. (2012) and include Lagged Insider Selling in order to control for insiders typical propensity to buy and sell shares of their firm s stock. This variable is designed to reduce the omitted variable bias, as it controls for unidentified omitted variables that affect insiders trading behavior. We rely on the appropriate insider selling variable from the month immediately prior to the 12-month period over which our mean measure of insider selling is computed. For example, for our hypothetical firm with a fiscal year end in August of 2009, our main insider selling variables would be computed based on the mean of the 12 monthly values from September of 2009 through August of 2010 and Lagged Insider Selling would reflect the value from August of Sample Construction We extract information on CEOs pension benefits and deferred compensation, as well as other CEO characteristics and compensation measures (i.e., salary, bonus, stock holdings, and option holdings), for S&P 1500 firms from the Standard & Poor s ExecuComp database. We begin our sample period in 2006 due to the fact that the Securities and Exchange Commission (SEC) s expanded executive compensation disclosure requirements, which mandate the provision of detailed information on executive pension and deferred compensation holdings, took effect that 19

20 year. Using the executive compensation data, we construct our two measures of inside debt as well as other variables that capture CEO characteristics and compensation structure. We then supplement the data set with annual accounting and financial data from Compustat and stock price data from CRSP. These data are used to construct the control variables for our multivariate analyses. We also use the Thomson Financial database to gather annual insider transactions data, from which we construct our two measures of insider selling. Finally, we merge the insider selling data with the executive compensation data, creating a one-year lag for the latter so that we can examine the effect of inside debt on insider selling in the subsequent year. The resulting sample consists of 12,586 firm-year observations over the 8-year period from 2006 through Figure 1 displays the distribution of inside debt by fiscal year. In Panel A, we observe an upward trend in the mean of CEO Inside Debt Holdings in recent years. Turning to Panels B and C, where we examine variation in the CEO Debt-to-Equity Ratio and the CEO Inside Debt Ratio over time, we note a marked increase in the mean proportion of inside debt within CEOs compensation structure in For example, the CEO Debt-to-Equity Ratio rose from 24.3% in 2007 to 44.3% in 2008 and the CEO Inside Debt Ratio rose from 11.7% to 15.5% over the same period. These findings are consistent with those of Anantharaman et al. (2014), who report an increase in inside debt at S&P 1500 firms from 25% to 43% of equity-based compensation from 2007 to This sudden increase in inside debt may be attributable to the financial crisis, as a growing number of firms experienced financial distress and therefore shifted CEO compensation toward debt-based measures in order to more heavily align the interests of the CEO with those of firm debtholders. 20

21 4. Results 4.1 Descriptive Statistics Table 1 presents descriptive statistics for our sample firms. All variables are lagged by one year relative to the fiscal year during which insider selling is measured. Additionally, with the exception of the two insider selling measures, all variables are winsorized at the 1% and 99% levels in order to mitigate the effects of outliers and data errors. We report mean values of insider selling equal to for the net measure and for the gross measure. For comparison, Khan et al. (2012) document a mean value of 0.40 for their gross measure of insider selling. The fact that our figure suggests relatively less selling by insiders may be attributable to differences between the compositions of the two samples. Since their analyses do not involve inside debt, meaning they are not limited to S&P 1500 firms from ExecuComp during the post-2006 period, Khan et al. examine a broader sample (nearly 314,000 firm-quarters) spanning both an earlier and longer time period ( ). Our measures of inside debt indicate that the average (median) CEO receives over $4.9 million ($606,000) in debt-type compensation, which constitutes 13.1% (3.6%) of his or her total compensation. The large standard deviation of for the CEO Inside Debt Ratio emphasizes the great amount of variation in inside debt holdings across firms belonging to the S&P Clearly, although inside debt accounts for a relatively small proportion of the total CEO compensation package at most firms, it represents a major source of compensation at some. Decomposing inside debt into its two constituents reveals that the average CEO is entitled to nearly $2.1 million in deferred compensation and $2.7 million in pensions. As expected, the mean (median) CEO Debt-to-Equity Ratio of 0.30 (0.04) demonstrates a strong equity bias, which may 21

22 reflect the fact that firms view equity compensation as a more effective means of inhibiting shirking and encouraging greater managerial effort (Edmans and Liu, 2011). The other CEO characteristics and compensation variables are also in line with our expectations. For example, the mean CEO Vega/Delta Ratio of is similar to the mean of reported by Cassell et al. (2012). Furthermore, the average CEO is nearly 56 years in age and has been with the firm for over 7 years. Not surprisingly, the firms in our sample are quite large; the average firm features a market capitalization of $7.3 billion, while the median firm exhibits a more modest figure of $1.8 billion. The average firm also commands a market-to-book ratio of 2.6, carries debt on the books valued at 22.0% of total assets, and boasts strong recent stock price performance with lagged 6-month returns of 6.3%. With a standard deviation of daily returns equal to 2.7%, the average firm s stock displays a decent level of volatility. However, the stock is quite liquid based on an average share turnover of In Table 2, we report pairwise Pearson correlation coefficients for the variables that will later be used in our multivariate analyses. We observe an exceptionally high correlation coefficient of 0.94 between the Net Insider Selling Ratio and the Gross Insider Selling Ratio, indicating that the two measures are closely related. This should come as no surprise considering the similarities between their calculations. We also report a high correlation of 0.87 between our two measures of inside debt, which reassures us that each one accurately captures the relative value of the CEO s debt-type compensation. However, we note that none of the control variables share a particularly high correlation, which reduces our concerns regarding potential multicollinearity due to correlated independent variables in our multivariate analyses. 22

23 Finally, all four of the pairwise correlations between our two measures of insider selling and our two measures of inside debt are negative, implying that insiders sell fewer shares at firms with high levels of CEO inside debt. This result suggests that insiders view inside debt favorably, possibly due to the fact that it reduces the cost of debt financing and provides financially constrained firms access to the debt markets. However, additional analysis is necessary to better understand the negative correlation between inside debt and insider selling. 4.2 Univariate Analyses In order to further explore the relationship between inside debt and insider selling, we conduct a univariate analysis in which we compare the net and gross measures of insider selling across the top and bottom quartiles based on our two measures of inside debt. We begin by partitioning our sample into quartiles based on the level of inside debt. We do this separately for the CEO Debt-to-Equity Ratio and the CEO Inside Debt Ratio. Next, we calculate the mean value of the Net Insider Selling Ratio and the Gross Insider Selling Ratio within each quartile. Finally, we compute the difference between mean insider selling in the top and bottom quartiles (i.e., Q4 minus Q1), and we test for the significance of this difference using two-sample t-tests. The results of this analysis appear in Table 3. The quartiles in Panel A are based on the CEO Debt-to-Equity Ratio, while those in Panel B are based on the CEO Inside Debt Ratio. In Panel A, we find that the mean Net Insider Selling Ratio is in the top quartile that consists of firms with the highest CEO Debt-to-Equity Ratio, and it is in the bottom quartile that consists of firms with the lowest CEO Debt-to-Equity Ratio. The difference of is statistically significant at the 1% level. Consistent with the negative correlation between insider selling and inside debt from Table 2, we conclude that there is significantly less selling by insiders at firms with high levels of inside debt. Once again, this evidence indicates that insiders are receptive to 23

24 issuances of inside debt; they sell fewer shares because they believe that a high level of debt-type compensation for the CEO will increase the value of their investment in the firm. We observe the same pattern for the mean Gross Insider Selling Ratio. The difference between the measure in the fourth quartile (0.232) and the first quartile (0.278) is , which is statistically significant at the 1% level. In Panel B, where we partition the sample into quartiles based on the CEO Inside Debt Ratio, we obtain qualitatively similar results. The difference between insider selling in the top and bottom quartiles is and for the net and gross measures, respectively. Both of these differences are significant at the 1% level. These findings further support the argument that insiders welcome the use of inside debt in the CEO compensation structure. Although the evidence in Table 3 corroborates the negative correlation between inside debt and insider selling that we reported in Table 2, it is interesting to note that insider selling does not decrease monotonically in the level of debt-type compensation. In fact, although the level of insider selling is consistently the lowest in the fourth inside debt quartile, we find that it is consistently the highest in the second quartile, followed by the third quartile and then the first quartile. In order to visually depict the relationship between inside debt and insider selling, we plot the means of the two insider selling measures as a function of inside debt in Figure 2. The sample is partitioned into quartiles based on the CEO Debt-to-Equity Ratio in Panel A and the CEO Inside Debt Ratio in Panel B. All four lines resemble an asymmetric inverted V where the height of the left endpoint exceeds that of the right endpoint and the peak appears in the second quartile. The steep rise in insider selling from the first quartile to the second quartile is followed by a gradual decline over the third and fourth quartiles. 24

25 The fact that the relationship between inside debt and insider selling is not monotonic emphasizes the intricate nature of the optimal compensation contract for CEOs and other firm executives. As Edmans and Liu (2011) point out, identifying the ideal combination of equity and debt in the managerial compensation contract is a complex task that is complicated by the potentially conflicting incentives that are offered by each component. While most agree that an equity bias is optimal for the majority of firms, our results indicate that the incorporation of debttype compensation is viewed favorably by insiders. However, we also find that insiders prefer low levels of inside debt over moderate levels, as evidenced by the lower values of mean insider selling in the first quartile relative to those in the second and third quartiles. This result suggests that low levels of inside debt more effectively incentivize managers to behave in the best interests of shareholders than do moderate levels, which might be expected considering that increases in managerial inside debt holdings should shift managers incentives towards those in favor of debtholders and away from those in favor of shareholders. In light of this pattern, it is surprising that trading activity by insiders suggests an overall preference for high levels of inside debt, as evidenced by the low means in the fourth quartile. Clearly, the topic of the optimal CEO inside debt level warrants further investigation. 4.3 Multivariate Analyses Though our results thus far support the notion that insiders view inside debt favorably due to its beneficial impact upon the firm, it is important to consider alternative explanations that may drive the significantly negative relationship between inside debt and insider selling. We therefore explore a variety of multivariate specifications in order to more thoroughly understand the impact of inside debt on insider selling. In Table 4, we report the results of four regressions. We include two different OLS models estimated using our two measures of insider selling as the dependent 25