Managerial Equity Holdings and Income Smoothing Incentives

Transcription

1 Managerial Equity Holdings and Income Smoothing Incentives Qing (Sydney) Shu Michael F. Price College of Business University of Oklahoma Wayne B. Thomas Michael F. Price College of Business University of Oklahoma October 2015 We would like to thank Charlene Henderson, Addy Noel, Jennifer Wu Tucker, workshop participants at Mississippi State University and University of Oklahoma, and participants at the 2015 FARS Midyear Meeting and 2015 AAA Southwest Region meeting for helpful comments and suggestions. This paper received the Distinguished Paper Award at the AAA Southwest Region meeting.

2 Managerial Equity Holdings and Income Smoothing Incentives Abstract Our study explores how managerial stock holdings and option holdings affect CEOs income smoothing incentives. Given the different roles of stock holdings and option holdings in solving agency problems, managers may smooth past earnings using discretionary accruals for the purpose of revealing information to help investors better predict future earnings, or alternatively, for the purpose of hiding volatility of past earnings. We find that the association between past smoothing and predictability of future earnings is increasing (decreasing) in CEO stock (option) holdings. The results are consistent with stock holdings aligning the interests of managers and shareholders, and managers using discretionary accruals to smooth past earnings to reveal information to investors about future performance. In contrast, options holdings have been linked with excessive risk-taking by managers, and managers use discretionary accruals to mask volatility of less predictable earnings. We also document similar effects of CFOs. Keywords: Stock holdings; option holdings; equity-based executive compensation; income smoothing; earnings predictability; earnings management JEL Classification: G10; G32; M40; M41

3 1. Introduction Managers use of discretionary accruals to smooth reported earnings is pervasive in practice (Graham, Harvey, and Rajgopal 2005). Such smoothing behavior is presumed by much of the academic literature to represent manipulative reporting. Managers are motivated to use discretionary accruals to dampen the volatility of underlying performance caused by their own opportunistic risk-taking behavior. By lowering volatility of reported performance, managers attempt to bias downward investors and creditors perceived risk of the firm. An alternative, more benign explanation for the existence of discretionary accruals-based income smoothing is managers desire to help investors better predict future performance. By reducing volatility of reported earnings caused by transitory items, managers better reveal to investors the firm s expected future performance. Thus, there are two competing motivations for the existence of discretionary accruals-based income smoothing. Our study explores these two motivations in the context of CEOs equity-based compensation (i.e., managerial stock holdings and option holdings). The literature has not provided a clear answer with respect to how equity-based compensation motivates CEOs to smooth their firms earnings to hide volatility of past performance or to communicate future performance (Carlson and Bathala 1997; Cheng and Warfield 2005; LaFond, Lang, and Skaife 2011). One the one hand, equity ownership is expected to be a natural solution to agency problem caused by separation of management and ownership (Jensen and Meckling 1976). Equity ownership is intended to incentivize the manager to align her actions with those in shareholders best interest. When managers actions are aligned, they have less to hide from shareholders and are more likely to use financial reporting to reveal their actions that benefit shareholders (Warfield, Wild, and Wild 1995; Klein 2002). On the other 1

4 hand, some forms of equity compensation may exacerbate the agency problem by motivating managers opportunistic risk-taking behavior. When such behavior is not in shareholders or other stakeholders best interest, managers will attempt to use financial reporting to hide their risky actions (Healy 1985; Aboody and Kasznik 2000; Nagar, Nanda, and Wysocki 2000). We differentiate between two types of CEO equity-based compensation stock holdings and option holdings. These two forms of compensation potentially provide the CEO with different incentives to use discretionary accruals to smooth income. Stock holdings are meant to better align shareholders and managers changes in wealth tied to firm performance and therefore incentivize managers to increase their efforts and make decisions in shareholders best interest. In this context, managers may want to smooth income to reveal private information related to their actions. In particular, managers can reduce earnings fluctuations arising from transitory events so that the firm s future economic performance is better signaled through the (smoother) trend in past profitability. Option holdings also tie managers wealth to stock performance, but they potentially affect managers income smoothing behavior in a different way. Stock options encourage risktaking behavior. The reason is that the value of stock options represents a convex payoff structure that increases with the volatility of stock price (Guay 1999). Such option-based compensation could therefore motivate managers to increase the firm s overall risk beyond a level that is optimal for shareholders, bondholders, employees, etc. Managers may use discretionary accruals to smooth income to hide such risk, which would otherwise be revealed through volatility of the firm s earnings stream. To the extent that managers income smoothing is motivated by hiding sub-optimal actions related to risk taking, past income smoothing will be associated with more volatile (i.e., less predictable) future performance. Thus, we offer different 2

5 expectations of the association between past income smoothing and predictability of future earnings based on whether equity compensation relates to stock holdings versus option holdings. To test our expectations, we adopt the approach in Tucker and Zarowin (2006). Tucker and Zarowin employ a prices-leading-earnings model developed by Collins et al. (1994) to test whether past income smoothing relates to the predictability of future earnings. They find that, in general, past income smoothing relates positively to more predictable earnings. The result is consistent with managers income smoothing representing a desire to better signal future performance to investors (information role) rather than hide volatility of performance (opportunistic role). The purpose of our study is to understand whether the positive association between past income smoothing and predictability of future earnings varies based on the type of equity compensation. As discussed above, stock holdings versus option holdings may have a different effect on managers motive for income smoothing. Our data of managerial equity holdings are collected from Execucomp. The final sample consists of 11,579 firm-year observations from 1994 to 2010 (with data required for future performance through 2013). We find that CEOs stock holdings enhance the positive association between past income smoothing and predictability of future earnings, whereas option holdings reduce the association. Thus, our findings support the following conjectures. As CEO stock ownership increases, her income smoothing behavior more likely represents a desire to convey private information about future profitability. As option-based compensation increases, however, the CEO s income smoothing behavior more likely represents an attempt to mask the firm s true underlying performance. In the case of options, CEOs are more likely to take actions that increase risk beyond the level optimal for shareholders, and they then use discretionary accruals to smooth earnings in an attempt to hide these sub-optimal activities. 3

6 We conduct several additional analyses and a variety of robustness tests to corroborate our primary tests. Our conclusions remain unchanged. In particular, we show that results are stronger in subsamples in which theory and prior empirical evidence suggest they should. For example, managers are expected to behave more opportunistically when the CEO also serves as the chairman of the board (i.e., CEO duality), and CEOs are expected to take on additional risks when options are out-of-the-money. We find greater evidence that option holdings relate to the opportunistic role of income smoothing in these settings. In contrast, manager-shareholder alignment is expected to improve with CEO tenure and with blockholdings, and we find greater evidence that stock holdings relate to the information role of income smoothing in these settings. Other tests show that restricted stock units (RSUs) have a similar effect on income smoothing as do stock holdings, as expected, and that conclusions are robust to measuring earnings predictability using an earnings persistence model (instead of a prices-leading-earnings model). Finally, we document evidence that CFOs (incremental to CEOs ) stock and option holdings have a similar effect on the association between past income smoothing and the predictability of future earnings. Our study offers several important contributions. First, in the debate over the impact of equity holdings on a firm s financial reporting quality, prior evidence has been mixed (Burns and Kedia 2006; Erickson, Hanlon and Maydew 2006; Efendi, Srivastava, and Swanson 2007; Larcker, Richardson, and Tuna 2007; Cheng and Farber 2008; Armstrong, Jagolinzer, and Larcker 2010; Armstrong et al. 2013). Our paper reconciles previous findings through the lens of CEOs income smoothing incentive. In particular, the uniqueness of our research design is to use the approach in Tucker and Zarowin (2006) to differentiate between the two competing explanations for the existence of income smoothing. In many different contexts, including equity 4

7 compensation, prior studies presume that a positive association with income smoothing is indicative of opportunistic reporting behavior (e.g., Grant, Markarian, and Parbonetti 2009). 1 However, as concluded by Tucker and Zarowin (2006) and confirmed in our study, a positive relation with discretionary accruals-based income smoothing could be indicative of either managers opportunistic behavior or managers informative behavior. We provide evidence that such smoothing is more likely to be informative as managerial stockholdings increase, and it is more likely to be opportunistic as managerial option holdings increase. Our study also adds to the interesting debate on the role of discretionary accruals in financial reporting in general (Watts and Zimmerman 1978; Christie and Zimmerman 1994; Dechow, Ge, and Schrand 2010) and in relation to executive compensation in particular (Christie and Zimmerman 1994; Bowen, Rajgopal, and Venkatachalam 2008). Graham, Harvey, and Rajgopal (2005) document that more than 96 percent of the executive respondents prefer smooth earnings streams, and the explanations offered by executives for this preference include achieving lower cost of equity and debt, preserving a higher credit rating, assuring customers and suppliers achieve better terms of trade, conveying higher growth prospects to investors, and helping analysts and investors predict future earnings. Our study validates the positive aspect of discretionary accruals to smooth earnings by linking improved earnings predictability with a higher level of managerial stock ownership. This result is consistent with the notion that shareholders benefit from earnings management when managers and shareholders interests are aligned (Linck, Netter, and Shu 2013; Louis and Robinson 2005). 2 This conclusion, however, 1 Prior studies offer a similar conclusion when documenting a positive relation between managers equity incentives and the absolute value of discretionary accruals (e.g., Cheng and Warfiled 2005; Bergstresser and Phillippon 2006; Cornett, Marcus, and Tehranian 2008; Jiang, Petroni, and Wang 2010; Armstrong et al. 2013). 2 Dechow and Skinner (2000) discuss that some earnings management is expected and should exist in capital markets, because judgments and estimates to implement accrual accounting are needed so that an earnings number can better inform shareholders of firm performance. Healy and Wahlen (1999) indicate that judgment and discretion 5

8 differs from some prior research which suggests that income smoothing associated with equity compensation is driven by an opportunistic reporting incentive (e.g., Cheng and Warfield 2005; Grant, Markarian, and Parbonetti 2009; Armstrong et al. 2013). 3 Our study also has timely and practical implications. The recent financial crisis raised concerns from regulators, standard setters, and politicians that equity-based pay schemes induce excessive risk-taking behavior (Bebchuk 2009; Bhattacharaya and Cohn 2010). 4 The SEC (2010, 10) states in its discussion of Proxy Disclosure Enhancements, Another concern expressed by commenters was that the linkage between risk-taking and executive compensation is not well understood, and that the disclosures provided under the proposed amendments would likely be boilerplate that could give investors a false sense of comfort regarding risk and risk-taking. 5 This sentiment has led recent studies to call for additional research on equity-based compensation (Bebchuk and Fried 2010; Bhagat and Romano 2009). Our study contributes to the current discussion of equity-based executive compensation by separately identify the effects of stock holdings and option options. 6 required in the financial reporting process allows managers to convey information to the capital market that best matches the firm s economic situation, possibly increasing the value of accounting information. 3 Cheng and Farber (2008) focus on managerial incentive to manage earnings upward and document a positive relation between earnings restatements and options granted to managers. Our study differs from theirs in that we analyze the effect of managers option holdings on their incentive to smooth income via discretionary accruals. Unlike intentional misrepresentation or fraud, estimates of discretionary accruals are allowed under GAAP and are necessary during the financial reporting process. We expect that earnings management through discretionary accruals can be either informative or opportunistic, depending on managers incentive to do so. 4 For instance, the Troubled Asset Relief Program (TARP) issued by the Treasure Department on June 10, 2009, limits cash pay to top executives and requires most compensation to be paid in the form of equity (Core and Guay, 2010). However, Treasury Secretary Timothy Geithner addressed in his report on the U.S. Treasury Budget in 2009 that what happened to compensation and the incentives in creative risk taking did contribute in some institutions to the vulnerability that we saw in this financial crisis In addition, U.S. CEOs typically hold several times more stock and options than their European peers (Conyon, Core, and Guay 2010). Our result relating option holdings to opportunistic income smoothing echoes the recent concern that, because U.S. executives already hold sufficient risk-taking incentive, requiring further option-based compensation induces managerial behavior to the detriment of shareholders (Core and Guay 2010). 6

9 In addition, the Securities and Exchange Commission (SEC) also recognizes that the CFO s compensation has important implications for shareholders because she, along with the CEO of the company, bears significant responsibility for the fair presentation of financial statements (SEC 2006). Since 2006, the SEC requires disclosure of CFO pay in addition to CEO pay. These additional disclosures have increased the attention of financial statement users and academics to the difference between CEO and CFO compensation-based incentives and the implications on financial reporting quality (Jiang, Petroni, and Wang 2010; Kim, Li, and Zhang 2011; Chava and Purnanandam 2010; Feng, Ge, Luo, and Shevlin 2011; Ge, Matsumoto, and Zhang 2011). Our study contributes to this growing stream of literature by showing an important and incremental role of CFOs equity holdings in explaining income smoothing behavior. In the next section, we discuss related research on managerial equity holdings and income smoothing and describe our hypothesis development. Section 3 introduces our research design including variable measures, model specification, and sample selection. Section 4 provides empirical results of our primary tests. Section 5 demonstrates additional analyses and robustness tests. In the final section we present concluding remarks. 2. Background and Hypothesis Development In this section, we first introduce the different roles of CEOs stock holdings versus option holdings in mitigating agency problems. We then address the competing explanations for managers income smoothing behavior (informative versus opportunistic). By combining these two streams of research, we develop our hypotheses. 7

10 2.1 Stock Holdings In a traditional agency-theory framework, the delegation of decision rights from shareholders to the CEO raises the concern that the CEO extracts private benefits and consumes perquisites from operating the firm in a way that is not in the interest of shareholders (Jensen and Meckling 1976; Williams 1987). To reduce such moral hazard problems stemming from executives owning too little of the firm, many firms adopt target-ownership plans that require their top executives to own a minimum level of company shares (Core and Larcker 2002). The rationale is that stock holdings require the CEO to share the gains and losses of the firm with outside shareholders, which motivates her to work harder and take actions that benefit shareholders (Smith and Watts 1992). In other words, stock ownership increases the CEO s incentive to undertake only projects with positive net present values to increase her firm s performance (Mehran 1995; Habib and Ljungqvist 2005). As evidence of this incentive, Lilienfeld-Toal and Ruenzi (2014) find that firms providing high stock holdings to their CEOs deliver higher stock market returns. Denis, Denis, and Sarin (1997) also document this incentive by showing a significant negative relation between value-destroying diversification and stock ownership of firm executives. Empirical evidence regarding the impact of stock holdings on earnings management, however, is ambiguous. Some studies document that stock ownership aligns the CEO s actions and shareholders interest and thus constrains the occurrence of financial reporting opportunism. For instance, LaFond and Roychowdhury (2008) show that, as managerial stock ownership declines, shareholders demand greater accounting conservatism to alleviate their concern about financial reporting manipulations in the firm. Warfield et al. (1995) document that managers 8

11 with increasing stock ownership choose less aggressive accounting choices through reducing discretionary accrual adjustments. On the other hand, once the CEO owns a significant fraction of her firm s stock shares, she can eventually, to the detriment of outside shareholders, accumulate sufficient influence to consume private benefits of control and protect herself from being dismissed for poor performance or being challenged by outside investors (Volpin 2002; Atanassov and Kim 2009). The harmful effects, for example, include excessive perquisite consumption, expropriation of minority shareholders wealth, and empire building (Kim and Lu 2011). In other words, with high stock ownership, the manager is more likely to indulge in self-serving activities inconsistent with shareholders interest. 7 Prior research, however, shows that such control advantage arising from managerial stock ownership only occurs after the CEO has reached a higher threshold level (Morck, Shleifer, and Vishny 1988). 2.2 Option Holdings Option holdings offer a different payoff structure compared to stock holdings. Option holders share the gains of the firm with shareholders but they do not share the losses. The convexity of this payoff structure makes the wealth of an option-holding CEO an increasing function of the firm s equity risk, or stock price volatility. 8 As a result, stock options potentially incentivize CEOs to engage in excessive risk taking beyond the optimal level desirable for 7 It may be worth noting that option-based compensation, as opposed to stock ownership, does not convey control benefits and thus may not motivate opportunistic financial reporting decisions through employment security (Core and Guay 2010). 8 A significant finance and accounting literature documents how stock-based and option-based compensation affect managerial behavior differently based on their different payoff structures (Jensen and Meckling 1976; Smith and Watts 1992; Haugen and Senbet 1981; Guay 1999; Tufano 1996; Cohen et al. 2000; Knopf et al. 2002; Chen et al. 2006; Brockmane et al. 2009; Dong et al. 2010; Burns and Kedia 2006; Kim et al. 2011). As opposed to the convexity inherent in option-based compensation, stock ownership has a linear payoff structure through creating a one-to-one relation between CEO wealth and stock price. In other words, stock holding exposes CEOs wealth to changes in stock price levels but not to changes in stock price volatility (Guay 1999). 9

12 shareholders (Lambert 1986). 9 Consistent with this conjecture, Gormley, Matsa, and Milbourn (2013) present theoretically that the desired stock price volatility for the CEO with option-based compensation is greater than it is with stock ownership. In particular, they address that relative to stock holdings, the risk-taking incentive stemming from option holdings reduces the sensitivity of managers expected utility to deviations from their desired level of risk. Furthermore, Gormley et al. (2013) provide empirical evidence that option compensation increases the firm s overall risk, such as higher leverage, more investment in R&D, and lower cash balance. Analogous to the lack of consensus on the role of stock ownership on earnings management, there are competing perspectives on the effect of option-based compensation on earnings management. On the one hand, a line of the literature suggests that options are granted to lessen management s desire to manipulate accounting numbers. Armstrong et al. (2010), for example, provide evidence of a negative relation between CEOs option-based incentives and the incidence of accounting irregularities. Their result is consistent with the notion that option compensation reduces financial reporting distortion. On the other hand, a substantial body of research indicates that firms granting high option-based executive compensation have greater earnings manipulation (Bergstresser and Philippon 2005; Harris and Bromiley 2007; Efendi, Srivastava, and Swanson 2007; Burns and Kedia 2006) and are more likely to be targets of fraud allegations (Denis, Hanouna, and Sarin 2006; Erickson, Hanlon, and Maydew 2006; Armstrong et al. 2013). In the next section, we explain how we use income smoothing to determine whether equity-based incentives to manage earnings help or hinder investors. 9 The positive relation between option-based compensation and managerial risk taking has been documented by a considerable literature. For instance, Coles, Daniel, and Naveen (2006) document that the payoff structure provided by options is positively associated with riskier policy choices, including more R&D expenditures, less PP&E investment, greater focus on fewer lines of business, and higher firm leverage. Rajgopal and Shevlin (2002) find that oil exploration risk is positively related to managerial option compensation. 10

13 2.3 Income Smoothing Income smoothing is depicted as managerial behavior to intentionally dampen the fluctuations of a firm s earnings realizations (Biedleman 1973). Through flexibility afforded in GAAP, managers can use judgments in estimating accruals to shift earnings from periods with high profits to those with low profits, thereby reducing the volatility in reported performance over time. The literature offers two opposing views regarding managers intent when smoothing income. One view is that managers smooth earnings to better communicate the firm s underlying economic performance to shareholders. Historical-cost GAAP-based earnings may contain transitory items that do not reflect managers long-term expectations of long-term economic performance. Managers can use discretionary accruals to dampen the fluctuations in earnings caused by these transitory items, giving investors a more accurate picture of future economic performance (Trueman and Titman 1988; Kirschenheiter and Melumad 2002). This type of income smoothing represents the information role of earnings management. When managers actions are more likely to be in shareholders best interest, there is less need for managers to hide performance, and they are more willing to signal information which reveals their actions that benefit shareholders (Sankar and Subramanyam 2001; Gu 2005). 10 A second view of income smoothing in the accounting literature is the opportunistic role of earnings management to hide information from shareholders (Leuz, Nanda, and Wysocki 2003; Jayaraman 2008). Opportunistic income smoothing also represents managers discretionary reporting to dampen fluctuations in earnings, but the motive for doing so is 10 Sankar and Subramanyam (2002) establish a setting where managers income smoothing is driven by their attempt to smooth consumption. They demonstrate that managers who have private information regarding future earnings have incentives to smooth income in ways that communicate their private information. As such, the market attaches greater weight to reported earnings than under a regime that does not allow managerial discretion to smooth income. 11

14 different. Rather than from the perspective of helping investors better predict future economic performance, opportunistic income smoothing is motivated to conceal volatility of past performance. When managers undertake excessively risky operations, those decisions will be revealed through volatile earnings performance. Managers can attempt to hide the true volatility of earnings arising from excessive risk taking by mimicking the earnings stream of firms without such opportunistic investment and financing policies (Dechow and Skinner 2000). This opportunistic perspective of income smoothing implies that discretionary accruals choices are made to deceive shareholders and enhance the manager s personal welfare (Healy 1985; Fudenberg and Tirole 1995; Leuz et al. 2003). Other studies conclude that income smoothing can be particularly misleading when such behavior makes it difficult for shareholders to predict future prospects (DeFond and Park 1997; Demski 1998) Hypotheses To summarize, the literature offers no consensus on the effect of managerial equity-based compensation on earnings management in general or income smoothing in particular (Carlson and Bathala 1997; Cheng and Warfield 2005; LaFond et al. 2011). We take into consideration the different incentives embedded in stock holdings and option holdings and distinguish between discretionary accruals choices motivated for informational reasons and those for opportunistic reasons. As discussed in more detail in the next section, we use the approach in Tucker and Zarowin (2006) to differentiate the informational role of income smoothing from its 11 The Office of Federal Housing Enterprise Oversight (OFHEO) confirmed an SEC investigation into the financial reporting practice of Fannie Mae in May The report states that Fannie Mae was accused of manipulating the timing of reporting nearly $11 billion dollars of net income and the manipulation was motivated by income smoothing for considerations related to executive compensation. The penalties include imposing a $400 million fine, freezing the company s loan portfolio, seeking restitution from former executives, and considering a possible criminal indictment (OFHEO 2006). 12

15 opportunistic role. They find that, in general, past income smoothing relates positively to the predictability of future earnings. The purpose of our study is to understand whether this positive association varies based on the type of equity compensation. The use of stock holdings to compensate executives is intended to align management s incentives with those of shareholders. Any action impairing firm value would lead to corresponding damage to the managers personal wealth, because they as shareholders bear a greater proportion of the costs caused by deviation from long-run firm value maximization. To signal their actions are aligned, managers are more willing to reveal information. In this case, their use of discretionary accruals to smooth income more likely represents an attempt to reveal information about future performance to investors. As a result, the association between past smoothing and predictability of future earnings is expected to increase in stock holdings. This discussion leads to our first hypothesis of the information role of income smoothing. H1: The association between past income smoothing and predictability of future earnings is increasing in stock holdings. While stock ownership is expected to provide an incentive for managers to improve the informativeness of earnings, there is some tension to the first hypothesis. High stock ownership potentially gives managers too much control over corporate resources (Cheng, Nagar, and Rajan 2004). Managers with high ownership may more easily exploit corporate resources for personal gains. To hide such activities, managers with high levels of stock ownership may engage in opportunistic earnings management, opposite expectations of the first hypothesis. We expect, however, that for most levels of stock holdings, they will serve to align managers and stockholders interest, and therefore income smoothing will reflect its informational role. Option holdings also represent equity-based compensation but they potentially have the opposite effect of stock holdings. Whereas stock holdings are intended to align the interests of 13

16 managers and shareholders, option holdings potentially misalign incentives for the following reason. Options offer a convex payoff to managers. The more volatile the expected payoff of the firm s underlying security, the more valuable the option is compared to the expected payoff of the underlying security. As a result, stock options potentially incentivize managers to make excessively risky decisions, which lead to higher performance volatility. Such excessive risk can reduce firm value and therefore may not be in stockholders best interest. Large fluctuations in firm performance can also directly affect the manager s tenure and job security (Carlson and Bathala 1997; Ronen and Sadan 1981), incentivizing managers to hide such activities. 12 To the extent that managers income smoothing is motivated by hiding sub-optimal actions related to excessive risk taking, past income smoothing will be associated with more volatile (i.e., less predictable) future performance. This leads to our second hypothesis of the opportunistic role of income smoothing. H2: The association between past income smoothing and predictability of future earnings is decreasing in option holdings. Option-based executive pay, nevertheless, is a form of equity-based compensation and therefore may ultimately serve to align managers actions and shareholders interest (Armstrong et al. 2009). To the extent that such alignment dominates the incentive to hide excessive risk, high levels of options may cause managers to use income smoothing to help investors better predict earnings. Therefore, while we state our second hypothesis from the perspective of the opportunistic role of income smoothing, there is some debate as to whether the information role will dominate. 12 Income smoothing based on discretionary accruals choices appears to be a less costly means of influencing perceived risk of the firm than altering operating, financing, and investing policies (Oswald and Zarowin 2007). 14

17 3. Research Design and Sample 3.1 Measures of Equity Holdings Consistent with the literature on agency problems between managers and shareholders, our measures of managerial equity holdings aim to capture the extent to which managers incentives are aligned with shareholders interests (Jensen and Meckling 1976; Cheng and Warfield 2005; LaFond and Roychowdhury 2008; Baker, Collins, and Reitenga 2003). Specifically, we calculate managerial stock holdings as the percentage of stock shares directly owned by the CEO (STK_HOLD), and calculate managerial option holdings as the percentage of stock options held by the CEO (OPT_HOLD). 13, Measure of Discretionary Accruals-Based Income Smoothing To measure discretionary accruals-based income smoothing (IS), we calculate the correlation between the change in discretionary accruals (DA) and the change in prediscretionary income (PDI) over the five previous years. PDI is measured as income before extraordinary items less DA. We multiply IS by minus one so that a higher value represents greater income smoothing. To control for industry and year effects, we convert IS into fractional rankings within each industry-year. 13 We also employ the dollar value of managers stock holdings and option holdings scaled by their total compensation, including their cash and bonuses, stock holdings, option holdings, and other long-term incentive plans. The tenor of our results remains unchanged. 14 In this study, we do not measure managerial equity holdings using delta and/or vega, which are defined as the sensitivity of managers equity wealth to the level and volatility of stock prices, respectively. This is because prior studies using delta and/or vega generally focus on managers equity incentives to engage in misreporting in the form of the absolute value of discretionary accruals, accounting restatements, or SEC Accounting and Auditing Enforcement Releases (AAERs). We instead examine managerial incentive arising from equity holdings from the perspective of investors ability to predict earnings (i.e., through the relation between firms future earnings response coefficient and income smoothing practice). 15

18 Our estimation of discretionary accruals is based on a modified version of the Dechow and Dichev (2002) model and accounts for the asymmetric timelier recognition of unrealized losses (Ball and Shivakumar 2006). Accrualst = α0 + α1ocft 1 + α2ocft + α3negocft + α4ocft*negocft + α5ocft+1 + α6δsalest + α7ppet + εt (1) We modify the Dechow and Dichev (2002) model by allowing the coefficient on operating cash flows (OCF) in the current year to vary between observations with positive and negative amounts (NegOCF). Accruals are calculated as the difference between net income and OCF. 15 Following McNichols (2002), we also include the growth in sales ( Sales) and property, plant, and equipment (PPE) for the current year. We estimate model (1) for each industry-year (defined by two-digit SIC code) with at least 10 observations. 16 The residuals from model (1) are discretionary accruals (DA) Earnings Predictability and Income Smoothing Incentive The information role of income smoothing improves the predictability of future earnings, whereas the opportunistic role distorts accounting information and creates earnings opacity. To distinguish the two components, Tucker and Zarowin (2006) adopt a prices-leading-earnings framework developed by Collins et al. (1994). Rt = β0 + β1xt 1 + β2xt + β3xt3 + β4rt3 + εt (2) 15 As total accruals also can be calculated as the change in working capitals ( WC) (Dechow and Dichev 2002), we compute WC as AR + Inventory AP TP + Other Assets (net), where AR is accounts receivable, AP is accounts payable, and TP is taxes payable. The results are qualitatively similar. 16 The results are qualitatively similar when we use industry-year combinations with at least 20 or 30 observations. 17 Tucker and Zarowin (2006) employ a model in Kothari et al. (2005), but Dechow et al. (2012) caution that the Kothari et al. model results in weak power and may lead to highly misspecified standard t-tests. Inferences are unchanged if we use the Kothari et al. (2005) model. 16

19 Rt is the ex-dividend stock return for fiscal year t. Xt 1 and Xt represent past and current EPS and are used to control for unexpected earnings in the current year. Xt3 equals the sum of EPS from year t+1 to t+3 and controls for expectations for future earnings in year t. All EPS measures are scaled by beginning price. Because using realized future earnings as a proxy for current expectations could introduce measurement error problems, realized stock returns for years t+1 through t+3 (Rt3) are used to control for future events not anticipated in year t. The coefficient on Xt3 is the future earnings response coefficient (FERC) and reflects the extent to which information about future earnings is impounded in current returns Rt. Because FERC captures revisions in shareholders expectations about future prospects, a higher FERC implies higher predictability of firms future earnings. To examine the impact of past income smoothing on predictability of future earnings, Tucker and Zarowin (2006) add IS to model (2) and estimate model (3). Rt = β0 + β1xt 1 + β2xt + β3xt3 + β4rt3 + β5ist + β6ist*xt 1 + β7ist*xt + β8ist*xt3 + β9ist*rt3 + εt (3) They find a significantly positive coefficient on the interaction ISt*Xt3 and conclude that past income smoothing is associated with more predictable earnings. 3.4 Primary Model In this study, we are interested in whether the association between Rt and the interaction of ISt and Xt3 changes with different levels of equity holdings. STK_HOLD and OPT_HOLD are added to model (3) to test H1 using model (4a) and to test H2 using model (4b) For ease in presentation and interpretation, we test H1 and H2 in separate models. Our inferences remain unchanged when both stock holdings and option holdings are included in the same model. 17

20 Rt = β0 + β1xt 1 + β2xt + β3xt3 + β4rt3 + β5ist + β6ist*xt 1 + β7ist*xt + β8ist*xt3 + β9ist*rt3 + β10 STK_HOLDt 1 + β11stk_holdt 1*Xt 1 + β12stk_holdt-1*xt + β13stk_holdt 1*Xt3 + β14stk_holdt 1*Rt3 (4a) + β15stk_holdt 1*ISt + β16stk_holdt-1*ist*xt 1 + β17stk_holdt 1*ISt*Xt + β18stk_holdt 1*ISt*Xt3 + β19stk_holdt 1*ISt*Rt3 + γncontrolsn + δncontrolsn*ist*xt3 + εt Rt = β0 + β1xt 1 + β2xt + β3xt3 + β4rt3 + β5ist + β6ist*xt 1 + β7ist*xt + β8ist*xt3 + β9ist*rt3 + β10opt_holdt 1 + β11opt_holdt 1*Xt 1 + β12opt_holdt-1*xt + β13opt_holdt 1*Xt3 + β14opt_holdt 1*Rt3 + β15opt_holdt 1*ISt + β16opt_holdt 1*ISt*Xt 1 (4b) + β17opt_holdt 1*ISt*Xt + β18opt_holdt 1*ISt*Xt3 + β19opt_holdt 1*ISt*Rt3 + γncontrolsn + δncontrolsn*ist*xt3 + εt We estimate models (4a) and (4b) using pooled, cross-sectional data, and the variables of interest are the three-way interaction terms STK_HOLDt 1*ISt*Xt3 and OPT_HOLDt 1*ISt*Xt3. 19 If the dominant effect of managerial stock holdings is to mitigate agency problems through creating mutual ownership between managers and shareholders, we expect a positive coefficient on STK_HOLDt 1*ISt*Xt3. A positive coefficient suggests that managers use of discretionary accruals to smooth past performance further enhances future earnings predictability as stock ownership increases. This result is consistent with the information role of income smoothing to enhance earnings predictability. Analogously, if the prevailing effect of option holdings is to exacerbate agency issues through encouraging excessive risk taking, we expect a negative coefficient on OPT_HOLDt 1*ISt*Xt3. A negative coefficient indicates that past smoothing is less positively related to predictability of future earnings as option holdings increase, consistent with 19 Given that our regression models include many interaction variables, we check the variance inflation factors (VIFs) of the interactions between our equity holdings variables, earnings variables and the income smoothing measure. The VIFs have an average value between 7 and 8. We also replicate the sample in Tucker and Zarowin (2006) and find that the variables of interest in their models have VIFs ranging between 4 and 6. 18

21 the opportunistic role of income smoothing to hide past earnings volatility induced by excessive risk taking. 20 To alleviate the potential concern that any statistical significance with respect to STK_HOLDt 1*ISt*Xt3 and/or OPT_HOLDt 1*ISt*Xt3 is due to omitted correlated factors, we follow Dou, Hope, and Thomas (2013) and include a set of control variables documented to be related with a firm s operating environment and affecting earnings smoothness. Specifically, we include control variables as main effects and interacted with ISt*Xt3. Firm size (SIZE) is measured as the natural logarithm of total assets at the beginning of year t. Growth opportunities (GROWTH) are proxied by the annual percentage change in sales from year t 1 to t. To gauge future earnings variability (EARNVAR), we use the standard deviation of EPS for years t+1 through t+3. The book-to-market ratio (BM) is calculated as the book value over the market value of common equity at the beginning of year t. To control for the extent of other implicit claims between a firm and its stakeholders, we control for investment intensity (INVEST) and fixed assets (PPE) at the beginning of year t. Prior evidence shows that debt contracting creates a demand for income smoothing, which helps firms to prevent adverse effects on debt rating (Trueman and Titman 1988; Minton and Schrand 1999). Accordingly, we control for the effect of leverage (LEV) using the ratio of long-term debt over the sum of long-term debt and the book value of common equity at the 20 We provide several alternatives to models (4a) and (4b). First, we split the sample into two groups based on the median value of STK_HOLD (OPT_HOLD) and estimate model (3) separately for high and low stock (option) holdings. Compared to the approach of interacting equity holdings with IS*Xt3, this alternative strategy allows all control variables to vary between high and low levels of equity holdings and, more importantly, alleviates concerns with multicollinearity. Second, we use indicator variables for IS, STK_HOLD, and OPT_HOLD based on sample median values. This approach is equivalent to running the model for the four subsamples, and then comparing the coefficient on X t3 across models. Third, we estimate the relation between managers equity compensation and income smoothing by taking the five-year average of equity holdings during the same time period (i.e., a five-year smoothing measure is examined in conjunction with a five-year compensation measure). Additionally, we examine a group of firms that experience no CEO turnover the five-year smoothing period. For each of these alternative specifications, our conclusions remain the same. We report several other sensitivity tests later. 19

22 beginning of year t. Incidence of negative earnings realization is represented by an indicator variable LOSS, which takes the value one if a firm has a negative current EPS and zero otherwise. 21 Finally, we measure analyst coverage (ANALYST) as the natural logarithm of one plus the average number of analyst coverage for each firm-year combination (Yu 2008). 3.5 Sample Selection The sample used for our primary test covers the period from 1994 to We obtain managerial equity holdings from Standard & Poor s (S&P) Execucomp and gather financial and accounting data from Compustat. Firm-year observations with missing data for past, current, and future three years earnings, operating cash flows, as well as those for current and future three years returns are deleted. To mitigate effects of extreme observations, all continuous variables are winsorized at their 1st and 99th percentiles. 23 We exclude firms in the financial and utilities industries (SIC codes between 4000 and 4999 and between 6000 and 6999) on Compustat before they are merged with firms on Execucomp. Our Compustat sample used to estimate discretionary accruals has 50,870 firm-year observations. After we merge the Compustat data with our Execucomp data, the final sample to conduct our primary test has a smaller size consisting of 11,579 firm-year observations with non-missing variable As a robustness test, we drop firm-year observations with negative current earnings, and our conclusions remain the same. 22 The sample period ends in 2010, as we require firms to have three future years of earnings and returns. 23 The results are robust to winsorizing the variables at the 2nd and 98th percentiles. 24 Execucomp contains information primarily for the S&P

23 4. Primary Results 4.1 Descriptive Statistics The descriptive statistics are reported in Table 1. Panel A demonstrates that the mean percentage of stock holdings by the CEO is and that the mean percentage of options held by the CEO is Both STK_HOLD and OPT_HOLD are considerably right-skewed. The relatively low values of managerial equity holdings are most likely due to the fact that levels of equity-based compensation in the S&P 1500 firms are restricted by various wealth constraints designated for large, publicly-listed firms. The earnings and returns variables are generally consistent to those reported in Tucker and Zarowin (2006). In particular, the average Rt of the final sample is , while untabulated statistics suggest that the mean Rt of the larger Compustat sample is The mean (median) value of Xt in the final sample is (0.0502), relative to the mean (median) of (0.042) in the Compustat sample. BM of the final sample has a mean (median) value of (0.3999), compared to the mean (median) of (0.447) in the Compustat sample. 25 The correlation matrix is shown in Panel B of Table 1. The below-diagonal portion presents the Spearman correlations, while the above-diagonal portion presents the Pearson correlations. As expected, STK_HOLD and OPT_HOLD are significantly and positively correlated. Consistent with prior research, Rt is negatively correlated with Xt 1 and Rt3 and is positively correlated with Xt and Xt LaFond and Roychowdhury (2008) document that Execucomp firms are more profitable and have higher marketto-book ratios than the population of firms on Compustat. 26 Given the valid concern that the control variable EARNVAR is potentially confounded with our income smoothing measure IS, it is worth noting that the Pearson (Spearman) correlation between the two variables is ( 0.191), which alleviates the concern of multicollinearity. 21

24 4.2 Benchmark Models of Earnings Predictability and Income Smoothing To compare with previous research, we present the results of the baseline model linking current returns with future earnings (model (2)) in Panel A of Table 2, using the Compustat sample and the final sample, respectively. As predicted, both the ERC and the FERC are significantly positive. This indicates that a significant amount of information about current and future earnings has been impounded in current stock returns, consistent with the prices-leadingearnings framework (Collins et al. 1994). Panel B of Table 2 reports the results of the benchmark model linking income smoothing with earnings predictability in model (3). Again, we use both the Compustat sample and the final sample. The significantly positive loading on the interaction ISt*Xt3 suggests that managers income smoothing behavior, on average, enhances earnings predictability. In addition, the estimated FERCs are highly comparable to the FERC in Tucker and Zarowin (2006, Table 3, Panel B), which gives us confidence that models (4a) and (4b) should provide reliable benchmarks to test our hypotheses. 4.3 Tests of Hypotheses Panel A of Table 3 provides the results of model (4a), and Panel B provides results of model (4b). The results for the variables used in the prices-leading-earnings model discussed in the previous section remain qualitatively unchanged after equity holdings variables are incorporated. Consistent with H1, the coefficient on the three-way interaction STK_HOLDt 1*ISt*Xt3 is positive and statistically significant. As stock holdings increase, the association between past income smoothing and predictability of future earnings increases. The result is consistent with the information role of income smoothing. Managers whose actions are 22

25 aligned with shareholders interest have an incentive to signal their actions and reveal information. The positive coefficient on STK_HOLDt 1*ISt*Xt3 is consistent with the use of discretionary accruals to smooth past performance to enhance investors ability to predict future performance. As for the result of option holdings, the coefficient on OPT_HOLDt 1*ISt*Xt3 is negative and statistically significant. This finding is consistent with H2. As option holdings increase, the association between past income smoothing and predictability of future earnings decreases. This result is consistent with the opportunistic role of income smoothing, which reflects the intent of managers to conceal past earnings volatility (rather than help investors predict future performance). Managers excessive risk-taking behavior is partially revealed through higher volatility of performance. To reduce investors and others perceived risk, managers use discretionary accruals to smooth past income. However, this type of opportunistic smoothing will be associated with more volatile (i.e., less predictable) future performance due to higher risktaking behavior. As a result, the relation between past income smoothing and predictability of future performance is less positive when managers opportunistically manipulate accruals. 4.4 Subsample Tests In this section, we test our hypotheses for subsamples based on manager and firm characteristics. To the extent our results reflect hypothesized constructs, we expect those results to be stronger for subsample in which theory and prior empirical evidence suggest they should. 23