Effect of personal taxes on managers’ decisions to sell their stock

ARTICLE IN PRESS Journal of Accounting and Economics 46 (2008) 23– 46

Contents lists available at ScienceDirect

Journal of Accounting and Economics journal homepage: www.elsevier.com/locate/jae

Effect of personal taxes on managers’ decisions to sell their stock$ Li Jin a, S.P. Kothari b, a b

Harvard Business School, USA MIT Sloan School of Management, USA

a r t i c l e in fo

abstract

Article history: Received 19 May 2006 Received in revised form 8 May 2008 Accepted 28 May 2008 Available online 6 June 2008

We examine the effect of personal taxes on CEOs’ decisions to sell their equity, controlling for diversification, managerial overconfidence, and other determinants. While CEOs frequently sell large amounts of their unrestricted firm equity, the tax burden associated with the sale significantly deters them from selling equity even after controlling for other determinants like diversification. We also find that both taxable institutional investors and CEOs respond to taxes in their selling of equity, although CEOs appear to be less tax-sensitive. Our findings underscore the importance of taxes in corporate and managerial decisions and they have implications for executive compensation policies. & 2008 Elsevier B.V. All rights reserved.

JEL classification: H24 J33 M41 M52 Keywords: Executive compensation Taxation Overconfidence Behavioral finance Institutional investors

1. Introduction In the United States, CEOs’ ownership of company stock and options is a principal means of aligning their incentives with shareholders (Jensen and Murphy, 1990; Hall and Liebman, 1998; Core et al., 2003). CEOs’ holdings change through time with new grants, exercise of vested options, and sale of vested stock. We study the determinants of the decision of large, publicly traded US corporations’ CEOs to sell their vested equity. We find that CEOs frequently sell substantial amounts of stock in their own firms, the value of which often exceeds that of new equity grants. We pay particular attention to personal tax considerations and diversification motives for such sales. The compensation literature makes frequent reference to diversification as an important consideration in CEOs’ decisions to sell or retain firm equity (Aggarwal and Samwick, 1999; Jin, 2002; Garvey and Milbourn, 2003, among others). In contrast, tax as a determinant of CEOs’ selling behavior is less understood. Our results show that, controlling for other determinants of their selling, CEOs are less likely to sell their vested equity as the tax burden increases. We also find that CEOs react less to tax incentives than taxable institutional investors, although both do react to taxes. Somewhat surprisingly, the diversification benefit of selling does

$ We thank two anonymous referees, Raj Chetty, John Core, Austan Goolsbee, John Graham, Wayne Guay, Michelle Hanlon, James Poterba, Nagpurnanand Prabhala, Josh Rauh, Joel Slemrod, and especially Jerry Zimmerman (editor), and seminar participants at the American Accounting Association Annual Meetings, China International Conference in Finance, European Finance Association Annual Meetings in Moscow, Harvard Business School, MIT Sloan School of Management, NBER Summer Institute on Economics of Taxation, Samsung School of Business, S. Korea, Texas Tech University, University of Chicago, and University of North Carolina Tax Symposium for helpful comments.  Corresponding author. Tel.: +1 617 253 0994; fax: +1 617 253 0603. E-mail addresses: [email protected] (L. Jin), [email protected] (S.P. Kothari).

0165-4101/$ - see front matter & 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.jacceco.2008.05.002

ARTICLE IN PRESS 24

L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

not appear to overcome the disincentive to sell and incur an immediate tax liability. Other determinants affect CEOs’ selling decisions largely as predicted in the literature. Understanding CEOs’ equity-selling patterns is important for several reasons. First, how CEOs undo incentives from equity compensation informs the design of equity compensation contracts in more effectively aligning incentives. Second, it indirectly provides evidence on how corporations (managers) react to tax incentives. Existing literature suggests corporations rationally respond to tax incentives (see, for example, Auerbach, 2002; Gordon and Hines, 2002; Graham and John, 2003), and that firms trade off taxes and other considerations such as financial accounting costs and contracting costs (see, for example, Shackelford et al., 2001; Erickson et al., 2004). To better understand how corporations respond to incentives, it is helpful to know whether corporate decision makers respond to tax incentives and trade off taxes and other considerations in their personal portfolio decisions. Third, the existing literature on individual investors documents behavioral biases such as the ‘‘disposition effect’’ (see Odean, 1998), whereby investors sell winning investments too soon and thus forego tax deferral, but hold on to losing investments too long. CEOs as a group of individuals are well-advised and sophisticated. They also have large stakes affected by taxes, and therefore might care about tax optimization more than the average investor. Our study examines whether CEOs are tax savvy. In Section 2, we review the literature on the determinants of equity selling by CEOs. We describe our empirical methodology, and the data and measurement of all variables used in the empirical analysis, in Section 3. Section 4 provides descriptive evidence of CEOs’ selling behavior as well as evidence of the determinants of the selling of their equity. Section 5 concludes. 2. Related literature Previous research, discussed below, identifies and empirically explores several determinants of CEOs’ sale of equity. These include taxes, diversification, managerial behavioral biases, implicit and explicit contracts between shareholders and managers, CEOs’ liquidity needs, managerial opportunism (i.e., trading on insider information), and managerial ability to hedge their positions. 2.1. Taxes A substantial body of theoretical and empirical literature examines the effect of taxes on investors’ equity portfolio decisions, and also on security prices. This literature explores the equilibrium asset-pricing effects of taxes, and the effect of taxes on investors’ decisions to sell (or not sell) securities with capital gains or losses. The summary of this research below shows that this literature is primarily concerned with the effect only of taxes, not the effect of multiple factors affecting investors’ decision to sell equity. Assuming no short-sale constraints, Constantinides (1983, 1984) predicts that investors will buy and sell securities, i.e., rebalance portfolios, without triggering any capital gains tax until death. With short-sale constraints, however, it is not possible to defer taxes indefinitely (i.e., until death). In the context of CEOs, a short-sale constraint is likely binding because corporations typically disallow a CEO from selling short the firm’s equity. Absent short-selling, Klein (1999) expects security prices to adjust such that investors’ after-tax rates of risk-adjusted returns are equal. Viard (2000) extends Klien’s analysis by deriving the conditions for equilibrium asset prices and showing that prices of appreciated securities include a component that offsets the seller’s disincentive to sell the appreciated security and incur capital gains tax. This is consistent with the lock-in effect of capital gains taxes that discourages selling. Dammon et al. (2001a) study the effect of capital gains tax on investors’ optimal portfolio decisions in the presence of short-sale constraints. Not surprisingly, they show that an investor would increase the weight of an appreciated security in the portfolio because of taxes, especially as the investor ages, because capital gains tax can be effectively eliminated upon death. In related work, Shackelford and Verrecchia (2002) show that the differential between the long-term and short-term capital gains tax rates creates a trade-off between optimal risk-sharing and optimal tax-related trading strategy. They show that sellers are reluctant to sell appreciated assets sooner because they are subject to higher capital gains taxes. Thus, the overall theme of the literature is that taxes discourage selling of appreciated stocks in the presence of short-sale constraints. An equally voluminous empirical literature complements the theoretical work examining the effect of taxes on investors’ portfolio decisions and on asset prices. For example, many document the lock-in effect of capital gains tax.1 But taxes have received little empirical attention as a factor governing the timing of CEOs’ selling of equity in a multivariate setting that allows a variety of determinants of CEOs’ selling decisions in the empirical analysis. 2.2. Diversification Lambert et al. (1991), Hall and Murphy (2000, 2002), and Meulbroek (2001) explain that a CEO with an underdiversified portfolio that is over-weighted in the firm’s equity has an incentive to sell equity. Shareholders also bear some of 1 See Feldstein et al. (1980), Landsman and Shackelford (1995), Reese (1998), Goolsbee and Austan (2000a, b), Poterba and Weisbenner (2001), Klein (2001), Blouin et al. (2003), Ayers et al. (2003), Jin (2006), Ellis et al. (2006), and Dai et al. (2006).

ARTICLE IN PRESS L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

25

the consequences through its effect on manager’s risk-taking behavior. Consistent with incentive-compensation theory, empirically we observe that corporations tend to lower equity-based incentives in CEO compensation contracts when the cost of under-diversification is high (see, for example, Aggarwal and Samwick, 1999; Jin, 2002; Garvey and Milbourn, 2003). As a practical matter, CEOs might avoid direct selling by privately hedging equity exposure stemming from stock and option compensation. Bettis et al. (2001) report that CEOs often use derivative instruments such as zero-cost collars to hedge their exposure to firm specific risk. Hedging provides executives with the ability to reduce the firm-specific risk exposure associated with their company equity holdings without triggering a negative tax consequence. 2.3. Managerial behavioral bias Managerial behavioral biases might affect managers’ willingness to hold unrestricted firm equity. Malmendier and Tate (2005) hypothesize that managers might retain their vested equity due in part to their (overly) optimistic and confident assessment (hereafter, ‘‘overconfidence’’) of the firm’s prospects.2 Such overconfidence can translate into CEOs’ willingness to hold on to firm equity even in the face of the under-diversification cost.3 Another behavioral bias is the ‘‘disposition effect’’ (see Shefrin and Statman, 1985; Kahneman and Tversky, 1979) that retail investors exhibit (see Barber and Odean, 2000), whereby investors retain their losing investments too long and sell winners too soon. In the context of option exercise, Heath et al. (1999), and Core and Guay (2001) find that, controlling for economic factors, psychological factors influence managers’ and employees’ option exercise behavior in a fashion similar to the disposition effect. Overconfidence and the disposition effect yield different predictions. If managerial overconfidence dominates, CEOs might sell less firm equity after seeing the stock price go up because they attribute the superior firm performance to themselves and become more overconfident. If the disposition effect dominates, CEOs might sell more aggressively after seeing the price run-up. Which effect dominates is an empirical matter. 2.4. Insider trading A long-standing body of research examines corporate insiders’ purchase and sale of company stock to exploit insider information. Early studies such as Seyhun (1986) provide evidence that insider trades are profitable. Recent research (see Lakonishok and Lee, 2001; Jeng et al., 2003) suggests asymmetric informativeness of insider trades. Specifically, insiders’ purchases, but not sales, predict future price performance. The selling decision might be a consequence of portfolio rebalancing following option exercise and equity grants, that is, it might be motivated by liquidity needs rather than insider information. Stock purchases are more likely to reflect the exploitation of private information.4 2.5. Implicit or explicit employment contracts Finally, implicit or explicit clauses in employment contracts and labor market expectations might constrain managers from divesting equity ownership. Corporations usually grant stock and options to align managerial interests with shareholders. Core and Larcker (2002) find that many firms have explicit minimum share ownership requirements that could preclude CEOs from divesting. Core and Guay (1999) theorize that the optimal incentive from a portfolio of stock and options varies with hypothesized economic determinants such as firm size, growth opportunities, and proxies for monitoring costs (see Core and Guay, 1999; Ofek and Yermack, 2000; Li, 2004 for theory and evidence). The findings suggest that once executives reach some target level of incentives through equity ownership, they offset further efforts to increase incentives by selling equity. 3. Empirical methodology, data, and variable measurement 3.1. Empirical methodology Following the discussion in Section 2, we hypothesize that three major factors affect CEOs’ selling of vested firm equity: tax burden, under-diversification cost, and managerial behavioral biases (i.e., overconfidence and the disposition effect). Relating CEOs’ selling of equity to these three determinants is not straightforward, however, because of the confounding 2 The term ‘‘overconfidence’’ is pejorative. It suggests that management’s optimistic beliefs about firm value are irrational. Our analysis does not hinge on whether such beliefs are in fact rational or irrational because both predict the same managerial behavior with respect to ownership of vested equity. We therefore prefer to label the belief as ‘‘optimistic,’’ but for reasons of consistency with prior literature, we use ‘‘overconfidence’’. 3 It might also lead to value-destroying firm level investment decisions including overpayment for acquisitions, overinvestment using internal sources of funds, underinvestment using external sources of funds, and so forth (see Roll, 1986; DeMeza and Southey, 1996; Boehmer and Netter, 1997; Heaton, 2002; Malmendier and Tate, 2004, 2005; Bertrand and Schoar, 2003). 4 A substantial literature examines how employees, in general, and top executives, in particular, exercise options (e.g., Huddart and Lang, 1996). We make a distinction between option exercise and sale of equity, and do not expect a perfect correspondence between the reasons for CEOs’ option exercises and sales of equity. Carpenter (1998) models exercise patterns for executive options to show that a simple extension of the ordinary American option model, which introduces random, exogenous exercise and forfeiture, predicts actual exercise patterns quite well.

ARTICLE IN PRESS 26

L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

effect of firms’ past performance. Of the three determinants, tax burden is clearly an increasing function of past firm performance. But, past good performance is also likely a factor contributing to managerial overconfidence because ‘‘individuals expect their behavior to produce success, they are more likely to attribute good outcomes to their actionsy’’ (Malmendier and Tate, 2005, p. 2662; and also Miller and Ross, 1975; Gervais and Odean, 2001). Two additional consequences of superior performance are that, ceteris paribus, (i) a relatively larger percentage of a manager’s financial wealth is invested in the firm, leading to increased under-diversification, and (ii) fewer shares of the appreciated equity must be sold to raise a given amount of cash if the manager decides to sell equity to fulfill liquidity needs.5 Taken together, superior past performance could lead managers (i) to sell fewer shares of firm equity due to tax optimization, overconfidence, or a liquidity need of a given amount, or (ii) to sell more shares due to more severe underdiversification or the disposition effect. In the empirical analysis below, we attempt to disentangle the respective effects by including proxies for each (except the disposition effect) and adding a direct control for past performance. Assessing the tax effect after controlling for overconfidence, and vice versa, might still be complicated because managerial overconfidence and tax considerations both discourage outperforming managers from selling equity. In this respect, we also include additional tests where we compare CEOs’ sales of their equity with non-managerial tax-sensitive institutional investors’ sales. Outside investors’ selling decision is unlikely to be attributable to managerial overconfidence. Hence, their behavior conditional on the security’s past performance serves as a benchmark, and thus enables us to estimate the incremental impact of overconfidence on a manager’s equity ownership decision. In the analysis below, we estimate each factor’s impact on deferring the CEO from selling of vested equity by 1 year. This allows us to compare the magnitude of the effect of each determinant. We find that tax savings from delaying the sale of equity by 1 year, while smaller, are nonetheless of the same order of magnitude as the CEO’s incremental underdiversification cost of postponing by 1 year the sale of vested equity or exercise of vested options. More importantly, because both measures have a large cross-sectional variation, in some cases tax savings from deferring could dominate the under-diversification cost, and vice versa. Thus, a priori, taxes and under-diversification could both play a first order role in CEOs’ decisions to sell vested equity. One caveat is that unobservable CEO hedging on the side might impair our ability to empirically detect the importance of under-diversification in CEO equity selling: if CEOs facing severe under-diversification choose to hedge more rather than sell firm equity, the measured equity selling will be less responsive to the cost of underdiversification than would otherwise be observed. We include a number of other firm and managerial characteristics to control for their respective effects on managers’ proclivity to sell equity. This includes a variable to control for the effect of performance on share price, which might mechanically result in managers needing to sell fewer shares to meet liquidity needs. All of the variables and how they are measured are described in Section 3.3. Based on the preceding discussion, we model CEOs’ selling of vested equity as Equity shares ¼ f ðtax burden of selling; under-diversification cost; CEO overconfidence; other control variablesÞ:

(1)

3.2. Data sources Compensation data for CEOs are from Standard & Poor’s ExecuComp database. ExecuComp contains compensation information for the five most-highly compensated executives of firms in the S&P 500, S&P mid-cap 400, and S&P small-cap 600 indexes from 1992 to 2002. We use CRSP data to calculate the beta risk and return volatility as well as the risk-adjusted stock price performance. Institutional ownership data are obtained from the Spectrum 13F institutional investor holding database. We further supplement the institutional ownership data with characteristics of the institutions’ clients from the Investment Adviser Public Disclosure (IAPD) data obtained from the Securities & Exchange Commission. The IAPD database contains investment advisors’ self-reported clientele, broken down into 10 categories.6 Investment advisors are required to report the percentage of business represented by each clientele category. 3.3. Variable measurement 3.3.1. CEO holding CEO holding of stock and options is an input into the measurement of variables such as tax burden and manager’s exposure to firm risk. ExecuComp provides detailed information only on current year grants of stock and options and CEOs’ current stock holdings.7 It does not contain information that would enable a researcher to infer CEOs’ precise option 5

See Baker and Hall (2004) Schaefer (1998) and Huddart and Lang (2003). These are individuals (other than high net worth individuals); high net worth individuals; banking or thrift institutions; investment companies (including mutual funds); pension and profit sharing plans (other than plan participants); other pooled investment vehicles (mostly hedge funds); charitable organizations; corporations or other businesses not listed above; state or municipal government entities; and ‘‘other,’’ such as non-US government entities. 7 There is one caveat: ExecuComp reports stock holdings as reported in company proxy statements. The stock holdings are typically not for the fiscal year close, but rather at the time of the proxy statement, which is typically a few months after the fiscal close. Thus, there is a mismatch in the timing of 6

ARTICLE IN PRESS L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

27

positions (e.g., the grant date and vesting information for prior years’ grants). We estimate CEOs’ approximate option positions at various points in time using the Core and Guay (2002) measure. As a robustness check, we adopt a detailed measure developed in Jin and Meulbroek (2004) and Hall and Knox (2004), which, in theory, gives a more detailed estimate of the option positions (but with more assumptions). We find that the two measures are highly correlated (r40.95). We tabulate all the results in the paper using the Core and Guay (2002) measure, but the alternative measure produces qualitatively similar results. Appendix A provides details of the two methods. 3.3.2. Selling of equity To derive the net number of shares sold by a CEO in a year, we take last year’s stock holding, add the number of new shares acquired through option exercises and stock grants, and subtract the current stock holding.8 If the resulting number is positive, we infer that the CEO sold shares on the open market. Our procedure closely resembles that of Jenter (2005). We treat missing share ownership as a missing observation rather than zero ownership.9 A portion of the selling measured above might be attributed to selling in response to option exercises (see Ofek and Yermack, 2000). Although such selling still reduces a CEO’s incentive, the market might not perceive it as ‘‘information revealing’’ and a CEO consequently feels less constrained to sell. As a robustness check, we construct an alternative measure of selling that excludes the sales linked to option exercise and restricted stock grants and obtain similar results. We measure exposure to firm value as the sum of the exposures from a CEO’s stock and option positions. Following Yermack (1995) and Hartzell and Starks (2003), we measure the exposure from options using the option delta. The combined exposure from stock and options is measured by the ‘‘aggregate delta,’’ defined as the delta of options multiplied by the number of options granted, plus the number of shares. A CEO’s wealth rises by the aggregate delta for every dollar increase in the firm’s share price.10 3.3.3. Cost of under-diversification We use the Hall and Murphy (2002) approach to calculate a CEO’s incremental cost of under-diversification as a result of deferring the sale (exercise) of vested stock (options) by 1 year. For a given utility function, Hall and Murphy (2002) first calculate the ‘‘certainty equivalent’’ value of the firm equity to the CEO. This is the amount of cash, V, that would make the CEO indifferent between holding the firm equity and cash. Since the CEO is under-diversified, he derives lower utility from holding $X of the firm equity than from the same amount of cash. Thus, the CEO applies a ‘‘discount’’ for holding firm equity in his portfolio. We calculate the cost of under-diversification to be the percentage discount of the fair market value of firm equity owned by the CEO. For example, if the CEO holds $100 of firm equity and his utility is equivalent to holding $80 cash, the under-diversification discount is calculated to be (10080)/100 ¼ 20%. Following extant literature, for each year, we rank this percentage under-diversification cost across all CEOs and use the rank as the measure of under-diversification cost. Following Hall and Murphy, we adopt a constant relative risk aversion utility function and assume a relative risk aversion parameter of three. Hall and Murphy experiment with other parameters, such as a constant relative risk aversion parameter of two, and show that their measures of under-diversification cost are not sensitive to these assumptions. We also tried a constant relative risk aversion parameter of two and obtained qualitatively similar results. We make one modification to the Hall and Murphy measure. They assume a CEO’s firm specific wealth in any given year to be a fixed percentage of the CEO’s overall wealth. Since we analyze panel data spanning multiple years, we adjust over time the CEO’s firm-specific proportion of wealth (initially assumed to be 50%) according to firm performance, whereas the balance of the CEO wealth is assumed to grow at the rate of market. This modification explicitly captures the effect of firm performance on under-diversification, but we note that we obtain similar results without the modification. Extant literature documents that under-diversification costs should reduce the incentives provided by CEOs’ total holding of all equity—both vested and non-vested. Our tests focus on the selling (rather than holding) of vested equity (rather than total equity). Thus, the tests here do not constitute a direct test for or against the hypothesis that underdiversification would negatively affect the total level of incentives. In unreported tests, we do find that the total level of incentives in our data is significantly negatively affected by the under-diversification cost. (footnote continued) stock ownership data and option-ownership (based on fiscal-year) data. Such time discrepancies likely add noise but not systematic bias, and therefore they are likely to weaken our ability to detect significant effects. 8 To determine the exact number of shares granted in a year, we followed the suggestion given in WRDS to calculate it as the value of the grant dividend divided by the company’s stock price at the end of the calendar year. 9 A missing value of stock ownership in ExecuComp could be due to a misreport in the company proxy statement or an error in ExecuComp. We hand checked some of the missing stock ownership observations. For example, Jack Michaels, CEO of HNI Corporation, has a missing value in his shares owned for fiscal year 1999. Upon checking the Proxy Statement, the actual number of shares owned in this case is reported in the ‘‘INCUMBENT DIRECTORS’’ section as 208,012. Thus, this is an error in the company’s financial reports. 10 The aggregate delta measure enables us to calculate various ‘‘derivative measures.’’ For example, dividing the aggregate delta by the number of shares outstanding at the beginning of the year and multiplying it by 1,000 gives us the familiar dollar change in managerial wealth per $1,000 change in shareholder wealth, first used by Jensen and Murphy. Alternatively, if we want to measure how much the executive’s wealth will increase for a 1% change in share price, it is the aggregate delta divided by the share price. As another example, if we want to measure how many percentages the executive’s within-firm wealth changes for a 1% change in share price (the ‘‘elasticity’’ of executive wealth to firm value), it is $1/$P*delta/$V, where $P is the stock price and $V is the total dollar value of executive wealth in the firm (both stock and options value).

ARTICLE IN PRESS 28

L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

3.3.4. Managerial overconfidence We adopt the Malmendier and Tate (2004) measure of managerial overconfidence, which is based on indications of a CEO’s overconfidence as reported in financial press. Specifically, Malmendier and Tate (2004, p. 14) define their measure as: We also collect data on how the press portrays each of the CEOs during the sample period. We search for articles referring to the CEOs in The New York Times, Business Week, Financial Times, and The Economist using LexisNexis and for articles in the The Wall Street Journal using Factiva.com. For each CEO, we record four statistics: the total number of articles; the number of articles containing the words ‘‘confident’’ or ‘‘confidence;’’ the number of articles containing the words ‘‘optimistic’’ or ‘‘optimism;’’ and the number of articles containing the words ‘‘reliable,’’ ‘‘cautious,’’ ‘‘conservative,’’ ‘‘practical,’’ ‘‘frugal,’’ or ‘‘steady.’’ We hand-check each article to be sure that the terms are used to describe the CEO in question. In the process of scanning the search output, we separate out any articles specifically describing the CEO as ‘‘not confident’’ or ‘‘not optimistic’’. Malmendier and Tate transform the frequency measure of overconfidence into a dummy variable measure set ‘‘equal to 1 when the number of ‘confident’ and ‘optimistic’ mentions for a CEO in the LexisNexis and Wall Street Journal searches exceeds the number of ‘not confident,’ ‘not optimistic,’ and ‘reliable, cautious, practical, conservative, steady, frugal’ mentions.’’11 We exactly follow Malmendier and Tate’s procedure, but expand the media sources to include Fortune, Forbes, and Time.12 To test the robustness of the results, we employ an alternative measure of overconfidence inferred using the CEO’s ‘‘net buying of stocks,’’ as in Jenter (2005). The results are qualitatively similar. CEO overconfidence should not be confused with lower CEO risk aversion. Both could be operative, and would have a reinforcing effect on managers’ proclivity towards equity sales, but they are subtly different. Overconfidence is a behavioral attribute that is at odds with a (Bayesian) rational model. In contrast, a rational, less risk-averse CEO is expected to behave differently from an overconfident CEO. Specifically, a less risk-averse CEO would still divest firm equity when given the choice between firm equity and the market portfolio. The latter represents a better risk-return tradeoff compared to bearing the idiosyncratic risk of firm equity even for a less risk-averse individual.

3.3.5. Tax burden Tax burden is measured as the total tax liability, as a percentage of the total proceeds, assuming the CEO exercises all options and sells all stock today. In the simplest case in which the CEO holds only stock, the tax burden is the tax liability per dollar of proceeds from selling the stock. When the CEO holds both stock and options, we first adjust the options to derive the equivalent amount of stock before deriving the tax burden. Appendix B describes in detail the calculation of tax burden. The intuition underlying the procedure is summarized below. Our goal is to estimate the total dollar tax liability from selling/exercising the CEO’s entire unrestricted stock and option portfolio today. Tax liability is the product of (i) taxable income per share, (ii) the tax rate, and (iii) the number of shares (or share equivalents) sold. Taxable income per share is the difference between the current share price and the manager’s cost of acquisition, that is, the tax basis. The appendix describes how we estimate taxable income. For tax rates, we recognize that stocks and options are typically taxed at different rates. Stock sales from holdings of more than 1 year attract longterm capital gains tax, which typically is lower than the tax rate on ordinary income. Since most of the employee stock options are non-qualified options, they trigger ordinary income tax at the time of exercise. We assume that CEOs are at the maximum marginal federal tax rate for their ordinary income and capital gains tax. We use the applicable historical tax rate for each year during our sample period from the Tax Policy Center. Because we ignore state taxes, our tax burden is measured with error. This source of noise likely weakens the tax burden variable’s ability to explain CEOs’ selling decisions. The number of shares sold is the sum of the number of unrestricted shares held by the CEO and the number of ‘‘stock equivalent shares’’ held as a result of the CEO’s option holding. Each option gives d shares of exposure to firm risk, where d is the hedge ratio or the ‘‘delta’’ of options. We therefore transform the options position into the equivalent stock position by each option as d shares. Once we calculate the total dollar tax liability from selling firm equity, and the total ‘‘equivalent stock’’ exposure of the CEO, the tax burden is measured as Tax burden ¼ Total dollar tax liability=½ðdoptions  Noptions þ N stocks Þ  price, where doptions is the delta of the option, Noptions is the number of options held by the CEO, Nstocks is the number of stocks held by the CEO, and price is the end-of-year stock price. 11 In calculating the ‘‘confident’’ and ‘‘not confident’’ frequencies, references to the same story in different publications are counted multiple times, not just once. The assumption is that multiple references are indicative of the salience of the story, and thus are likely to positively contribute to managerial confidence. 12 Malmendier and Tate (2004, 2005) also use another measure that is based on a CEO’s persistent holding of options after vesting. As discussed in the paper, because a myriad of economic factors affect it, the practice is not merely a reflection of managerial overconfidence.

ARTICLE IN PRESS L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

29

3.4. Control variables We include a series of control variables. We summarize the intuition behind each. Corporate governance potentially affects the selling of vested equities for at least two reasons. (i) Governance substitutes for the need to provide incentives. A well-governed and well-monitored (for example, by institutions) firm might have less need to provide stock-based compensation incentive to its CEO. That is, governance substitutes for an equity-based incentive. (ii) Governance might also be complementary to equity incentives. If the better-governed firms also simultaneously require the CEOs to bear more consequences for their actions, the compensation contracts might include a preponderance of performance-based incentives. We control for the following CEO and corporate governance characteristics: joint CEO/chairman title, as the duality of these titles can lead to board decisions biased in the management’s favor (Jensen, 1993; Chen, 2004); institutional ownership, which is often linked to the strength and quality of corporate governance (Hartzell and Starks, 2003; Chidambaran and Prabhala, 2003). We include two additional CEO characteristics as control variables: CEO age and tenure, both to proxy for entrenchment, career concern, risk aversion, outside wealth, and to control for the mechanical effect of accumulated stock and options in a CEO’s portfolio on his selling behavior. Unfortunately, age and tenure information is missing for a large number of observations in the ExecuComp database. To avoid sacrificing the sample size, we present results using these variables only as robustness checks. CEOs’ sales of vested equity could be related to company grants in the current year: Ofek and Yermack (2000) report that CEOs offload most of the stock they acquire from option exercise immediately after the exercise. We therefore control for current year option grants by converting them into share equivalents using the option delta. Implicit and sometimes explicit contracts might dictate CEOs hold a certain amount of equity. If such contracts are related to industry characteristics, e.g., human capital specificity in different functions, then industry fixed effects would control for their confounding effect on the CEOs’ selling of equity. We also include additional firm-specific and CEO-specific control variables, e.g., the book-to-market ratio and firm size. Noe (1999) and Ke et al. (2003) suggest that even though CEOs’ equity selling is unrelated to immediate future earnings, it might be associated with the firm’s longer-term profitability. In particular, if the firm’s long-term prospects are promising, they might be disinclined to sell their equity. We proxy for this behavior using 1-year-ahead realized earnings deflated by share price. Ideally, it would have been desirable to also include 3–5-year-ahead realized earnings, but since our panel only spans 7 years, losing 3–5 years of data would have rendered the results meaningless. Finally, we control for firms’ past performance, measured by the past 3-year market model-adjusted return. (We also perform robustness tests using alternative measures of past performance, including 5-year performance and the CAPM and Fama-French three-factor model abnormal returns. The tenor of the results is unaffected.) Past performance (i) is positively, but imperfectly, related to the CEO’s tax burden13; (ii) might contribute to overconfidence; and (iii) raises the CEO’s underdiversification cost. In addition, it could affect the implicit and explicit labor market contracts because better performing CEOs might be less subject to the scrutiny of the board and might enjoy more freedom to dispose of shares in the firm. 4. Empirical results In Section 4.1 we present descriptive statistics for the data used in the empirical analysis. In Section 4.2 we discuss descriptive evidence of CEOs’ selling behavior, which reveals that selling of equity by the CEOs is quite prevalent. Sections 4.3–4.5 examine the economic determinants of CEOs’ selling of equity. 4.1. Descriptive statistics and cross-correlations Table 1 reports descriptive statistics for the main variables. The vested equity and many other variables contain outliers, as evidenced by skewness in the data. To ensure that our inferences are not unduly affected by the outliers, we truncate the following variables at the 1% and 99% levels: tax burden, new grant, institutional holding, past performance, assets, bookto-market, and future EPS. The tabulated results are based on outlier-truncated data, but we do not discern a qualitative change in the results due to truncation. We report statistics based on all the firm-year observations available individually for the variables. Therefore, there is considerable variation in the number of observations with far fewer data points for tenure and age variables. CEOs in our sample on average own 3.16% vested equity, which is the share-equivalent ownership of vested stock and options. CEO ownership is right-skewed, with 75th percentile ownership being only 2.46% whereas the maximum ownership is 35.15%. Average tax burden is 2% with a standard deviation of 12%. Although the average tax burden is quite small, the observed substantial variation in the burden across the sample of CEOs suggests that tests of the hypothesis that taxes influence CEOs’ selling decisions should have considerable power. The two overconfidence dummies show that 13% and 29% of the 13 Past performance is not perfectly correlated with the tax burden measure both because (i) our measure of past performance filters out the market return component, and (ii) the tax burden depends on the appreciation of the manager’s portfolio rather than of the firm and is thus affected by both (overall) performance and the timing and pattern of the acquisition of the manager’s portfolio.

ARTICLE IN PRESS 30

L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

Table 1 Summary statistics of main variables from 1996 to 2002 Variable name a

Holding of vested equity Tax burdenb Cost of under-diversification Overconfidence dummy 1c Overconfidence dummy 2d New grante Institutional holdingsf Past performanceg Future EPS/priceh Tenurei Agej

No. of obs.

Mean

Median

Std. dev.

Min

Quartile 1

Quartile 3

Max

13,536 11,123 13,536 12,055 13,536 13,536 13,536 13,536 10,649 7,005 6,795

3.16 0.02 0.50 0.13 0.29 0.17 0.52 0.21 0.01 16.64 57.59

0.79 0.03 0.49 0.00 0.00 0.08 0.55 0.06 0.05 14 57

6.20 0.12 0.29 0.33 0.46 0.28 0.20 1.06 0.28 12.22 7.82

0.00 2.87 0.00 0.00 0.00 0.00 0.06 0.91 2.20 0 34

0.28 0.00 0.24 0.00 0.00 0.01 0.32 0.39 0.02 6 52

2.46 0.07 0.74 0.00 1.00 0.20 0.68 0.43 0.07 26 62

35.15 0.19 0.99 1.00 1.00 1.75 0.94 5.93 0.19 60 90

a

Number of vested options times option delta, plus the number of vested stocks, divided by total shares outstanding, multiplied by 100. Tax liabilities per dollar selling, calculated as capital gains tax rate multiplied by taxable profit per share, divided by current stock price per share. c Overconfidence dummy 1 is the dummy for press mentioning of overconfidence, constructed following Malmendier and Tate (2004). d A dummy for CEO buying stocks on the open market in a year. e The new award of options multiplied by option delta, plus the new award of shares, dividend by shares outstanding, multiplied by 100. f The percentage holding by institutional investors, as reported in Thomson financials spectrum 13A data. g The 3-year market model adjusted cumulative abnormal performance for the stock, up to the end of the last fiscal year. h The next fiscal year’s earnings per share divided by price per share. i The number of years since the CEO joined the company. j The age of the CEO. b

CEOs are characterized as overconfident on the basis of press mentions and the CEOs’ purchase of firm equity in the open market, respectively. New grants of equity average 0.17% of the firm’s outstanding shares, but this variable also exhibits considerable right skewness. Institutional investors own over half of the equity of the typical firm in the sample. Average number of years that a CEO has been with his firm is almost 17 years. Table 2 reports Pearson product-moment correlations among the main set of variables. In general, the cross-correlations are fairly low even among variables that are expected to be positively correlated, for example, multiple proxies of the same variable. This suggests the presence of considerable measurement error. CEO equity ownership has a significant 0.06 correlation with tax burden, which is consistent with CEOs’ reluctance to sell appreciated equity to defer taxes. CEO ownership’s correlation with the overconfidence proxies exhibits mixed evidence. When we measure overconfidence on the basis of whether the CEO has purchased shares on the open market, it’s positively correlated with ownership, but the overconfidence proxy on the basis of press mentions is not significantly correlated. The lack of consistency might be due in part to the weak correlation between the two overconfidence proxies themselves. While we have followed previous literature in constructing the overconfidence measures, the lack of strong positive correlation suggests it is a daunting task to accurately capture managerial overconfidence, which suggests caution in interpreting the results about overconfidence and its attendant effects on managerial selling. Table 2 also shows that the cost of under-diversification is not significantly associated with ownership. CEOs’ ownership decision does not seem to be particularly influenced by the diversification consideration in a univariate correlation analysis. Both past performance and the dummy for CEO–Chairman dual title are significantly positively correlated with CEO ownership, which is not surprising. CEOs are rewarded for good past performance and some of this reward might be in the form of equity. As expected, tax burden has a positive correlation of 0.29 with past performance. 4.2. Descriptive evidence on CEOs’ selling behavior 4.2.1. Selling of stock Table 3 Panel A reports the results focusing only on the sale of unrestricted stock by the CEOs.14 On average, slightly more than 40% of the CEOs sell vested stock each year. The median (mean) sale is 15% (24%) of the existing stock position that includes the current year’s stock grant. Using the traditional measure of pay-for-performance sensitivity, the selling significantly affects CEOs’ incentives: the median (mean) sale would have affected the CEO’s wealth by $1.02 ($5.08) per $1,000 change in shareholder value. The sale represents a median decline in the CEO’s fractional ownership in the firm of 0.10% (0.51%), which is a sizable fraction of the CEO’s total stock ownership. Recall that the median (mean) CEO ownership from Table 1 is 0.79% (3.16%). 14 It is possible that when CEOs leave office they might sell a lot of stocks. The data we use in the analysis does not include CEOs who leave their jobs during the year, thus it does not capture the selling due to the CEOs job change.

Table 2 Pearson correlation coefficients between main variables of interest: data from 1996 to 2002 Vested equity holdinga

Tax burdenc

Dummy for dualCEO–Chairman Cost of underdiversificationd Overconfidence 1e Overconfidence 2f Past performanceg Future EPS/price

a

h

Institutional holding

DueCEO–Chairman

Cost of lost div.

Overconfidence 1

Overconfidence 2

Past performance

0.02 0.06 0.03 0.00 0.01 0.34

0.03 o.0001 0.01 0.19

0.04 0.00

0.02

0.03 0.06 o.0001 0.07 o.0001 0.09

0.09 o.0001 0.02 0.00 0.06

0.14 o.0001 0.07

0.06

o.0001 0.09

o.0001 0.02

o.0001 0.09

o.0001 0.02

o.0001 0.01 0.18 0.05 o.0001 0.05 o.0001 0.02 0.0699

0.06 0.02 0.01 0.02 0.00 0.02 0.01 0.09 o.0001

o.0001 0.05 o.0001 0.04 o.0001 0.29 o.0001 0.13 o.0001

0.02 0.02 0.04 0.05 o.0001 0.10 o.0001 0.08 o.0001

0.02 0.05 0.03 0.00 0.03 0.00 0.03 o.0001 0.07 o.0001

0.07 o.0001 0.39 o.0001 0.05 o.0001 0.05 o.0001

Share equivalent vested equity divided by shares outstanding, multiplied by 100. The share equivalent of the options is calculated as the number of options times the average delta of the options. The award of new equity divided by shares outstanding, multiplied by 100. The natural log of the ratio of the current price and the effective tax basis of the CEO’s equity. d The ranked measure of the Hall and Murphy (2002) measure of the percentage discount that managers place on the executive stock and options, assuming a relative risk aversion parameter of 3 and assuming that CEOs have 67% of their wealth invested in the firm’s equity. e The overconfidence measure 1 is based on the frequency of press mentions suggesting managerial overconfidence as a fraction of total press articles for the firm. This measure is constructed as in Malmendier and Tate (2004). f The overconfidence measure 2 is a dummy variable set equal to 1 if the CEO bought stock on the open market in the year of the observation. g Three-year market-model cumulative abnormal performance for the stock measured up to the end of the previous fiscal year. h The next fiscal year’s reported earnings per share divided by price per share. b c

ARTICLE IN PRESS

Institutional holding

Tax burden

L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

New grants of stock and options (total share equivalent)b

New grants (total share equivalent)

31

ARTICLE IN PRESS 32

L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

Table 3 Selling of vested stocks by chief executives: data from 1996 to 2002 Variable name

Panel A: Stock only calculation Selling (all inclusive) Selling/total holdinga Selling/shares outstanding  100b Selling (not including selling due to option exercises) Selling/total holdinga Selling/shares outstanding  100b Selling (not including selling due to option exercises or to offset current year restricted stock grants) Selling/total holdinga Selling/shares outstanding  100b Panel B: Stock and options combined calculation Reduction of combined positions in stocks and options Reduction/total holdinga Reduction/shares outstanding  100b Reduction of combined stock and option positions (not including reduction to offset current year option grants) Reduction/total holdinga Reduction/shares outstanding  100b Reduction of combined stock and option positions (not including reduction to offset current year option and stock grants) Reduction/total holdinga Reduction/shares outstanding  100b

As a percentage of whole CEO sample (%)

Mean

Std. dev.

1%

Quartile 1

Median

Quartile 3

99%

0.241 0.508

0.252 1.818

0.001 0.002

0.041 0.022

0.146 0.102

0.370 0.380

0.974 6.778

0.153 0.553

0.200 2.117

0.000 0.001

0.019 0.013

0.071 0.070

0.200 0.326

0.914 7.779

0.148 0.622

0.200 2.278

0.000 0.001

0.017 0.014

0.065 0.083

0.191 0.387

0.922 8.990

0.13 0.60

0.16 2.01

0.00 0.00

0.02 0.03

0.08 0.12

0.18 0.44

0.82 8.01

0.13 0.88

0.18 2.67

0.00 0.00

0.02 0.04

0.06 0.18

0.16 0.65

0.87 12.59

0.13 0.92

0.18 2.74

0.00 0.00

0.02 0.04

0.06 0.19

0.16 0.69

0.87 12.96

40.87

28.38

24.17

41.92

20.72

19.21

a Selling/total holdings is the selling as a percentage over total holdings, where total holdings is defined as last periods’ shares owned, plus this periods’ new shares acquired through either new shares granted or options exercised. b Selling/shares outstanding  100 is the ratio of shares sold and total shares outstanding, multiplied by 100.

Even though selling of shares lowers a CEO’s incentive, the net change in the CEO’s incentive from the previous year might still be positive because of new stock grants and option exercise during the year. The second measure in Table 3 excludes the stock sales that immediately follow option exercises. Over 28% of the CEOs sell their stock, which amounts to a median (mean) decline of 7% (15%) in their existing equity position. This represents a decline in the median (mean) pay-forperformance sensitivity by $0.70 ($5.53) of CEO wealth per $1,000 change in shareholder value. We also adjusted the CEOs’ equity sales for the amount of restricted stock grants. The tenor of descriptive statistics is unchanged. This is not surprising because restricted stock grants are relatively infrequent. Taken together, many CEOs sell their stock annually and the amount sold represents a non-trivial fraction of their ownership.

4.2.2. Analysis of stock selling with controls for options A weakness of the previous analysis is that it ignores options in measuring a CEO’s selling activity. Since options account for a large proportion of both new grants and total equity holdings of a CEO, excluding the effect of changes in option holdings could potentially lead to erroneous inferences. Below we examine the change in a CEO’s total exposure from stock and option ownership as a result of stock selling. We measure total exposure somewhat crudely by simply summing the number of shares and the number of options owned by a CEO. This overstates the exposure because the option delta is strictly less than one. However, measuring the exposure to firm risk from options as the number of options times the delta of the options is complicated because option delta changes with the stock price, so that even without a CEO actually selling or exercising options, the CEO’s exposure to firm risk fluctuates with stock price.15 At the end of each year, we add the restricted stock and option grants for the current year to a CEO’s total reported holding of stock and options at the end of the previous year. From this gross equity position we subtract the CEO’s reported stock and options position at the end of the current year. A positive difference implies the CEO sold stock in the current year. Using this definition of selling, Table 3, panel B reports that 42% of CEOs sell stock during a year. The median ratio of selling to the total combined holding of stock and options inclusive of the current year grants of stock and options is 8%, and 15 A ‘‘third approach,’’ which addresses the fact that option delta is less than one, but does not allow the exposure from options to fluctuate with stock prices, is to assume a ‘‘constant delta’’ for the options. Previous research such as that of Ofek and Yermack (2000), assume an average delta of about 0.6. This approach yields results that fall between the results reported in panels A and B of Table 3.

ARTICLE IN PRESS L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

33

the mean is 13%. Further adjustments that assume sale of stock in response to new grants of options and restricted stock slightly lower the magnitude of stock selling, but leave the tenor of the results unchanged. 4.3. Determinants of equity selling behavior 4.3.1. Determinants of the sale of vested stock We estimate the following regression based on Eq. (1) of the determinants of the sale of vested stock by CEOs: Selling of equity ¼ a0 þ a1 tax burden þ a2 negative tax burden þ a3 cost of under-diversification þ a4 overconfidence þ control variables þ error:

(2)

In the regression, we differentiate between positive and negative tax burden. Only a realized gain triggers an immediate tax liability. But even an unrealized loss can be a benefit if the CEO has other gains that can be offset by the loss. This suggests a non-linear effect of positive and negative tax burdens on the CEO’s decision to sell equity. We therefore include ‘‘Negative Tax burden’’ as another variable in the model. The total impact of tax, in the case of a negative burden, is the sum of a1 and a2. We include a series of control variables (see Table 4). Lagged ownership of vested equity is included because the level of ownership itself might affect the extent of selling of stock by the CEO in the current period. We also include industry-level fixed effects using two-digit SIC codes to address the concern that implicit labor market constraints might be industryspecific. We estimate Eq. (2) annually from 1996 to 2002 using only the observations where a CEO has sold stock. We report time-series averages of the estimated coefficients for each variable, and Fama and MacBeth (1973) standard errors corrected for first-order autocorrelation (see Pontiff, 1996). Table 4 reports the regression results for our main regressions. The first two columns show that the tax burden significantly discourages CEOs from selling their stock. The coefficient on the tax burden variable, 7.09 (t-statistic ¼ 4.07) is economically large as suggested by the standardized coefficient magnitude, which is reported under the ‘‘Economic significance’’ column in the table. Specifically, for a one standard deviation decrease in tax burden (0.12 from Table 1), the CEO’s stock selling reduces his exposure (i.e., pay-for-performance sensitivity) by 0.85% of shares outstanding (see the second column in Table 4). The reduction represents a nontrivial fraction of the mean (median) selling of a CEO at 0.51% (0.10%) reported in Table 3. In addition, the coefficient on negative tax burden is almost close to the coefficient on tax burden, so the overall impact of a negative tax burden on the selling of vested equity is close to zero. Compared to taxes, the coefficient on the overconfidence proxy is statistically and economically unimportant. Specifically, the coefficient on overconfidence is 0.11 (t-statistic ¼ 1.14). This implies the CEO’s exposure declines by 0.04% of shares outstanding when overconfidence decreases by one standard deviation. Results for the control variables show that the amount of current grant of stock and options and past stock return performance significantly increase the amount of stock a CEO sells in a year. Somewhat surprising, under-diversification cost is not significant in determining the selling of vested equity. We believe there are at least three explanations for this result. First, under-diversification cost is correlated with past performance and, as noted above, past performance significantly increases the likelihood of a CEO selling his stock. Second, CEOs might be less risk averse than previously believed in the literature. Thus, the discount they apply to their under-diversified positions might be lower than predicted in the existing literature. Consistent with this explanation, Mitton and Vorkink (2007) document investors’ preference for skewness. Since executives’ holding horizon for equity positions is typically long, and long-horizon security returns are significantly right-skewed, some CEOs might willingly hold under-diversified positions in their firms. Third, our estimate of the under-diversification discount might be noisy and biased downward because of significant correlated omitted variables such as hedging by the CEOs. Notwithstanding our attempt to follow the literature and be as comprehensive as possible in including potential determinants of equity selling, we acknowledge the possibility of omitted variables affecting our estimates. Noise in our proxy for the under-diversification cost, however, is unlikely to explain the significance of the tax variable. The error, in fact, is likely to bias against finding a significant tax coefficient. This inference follows from an errors-invariables analysis in a simplified setting of equity selling regressed on tax burden and under-diversification cost. In this case, the two independent variables are positively correlated and under-diversification positively affects the CEO’s propensity to sell equity. Therefore, measurement error in the under-diversification cost variable would positively (not negatively) bias the coefficient on tax burden (see Levi, 1973; Greene, 2000, pp. 377–378). We caution, however, that inferences about the direction of bias are muddled if multiple variables in the regression contain measurement error. We briefly discuss how our evidence squares with the findings of previous research. Our results on the relation between current equity grant and CEO selling are broadly consistent with those reported in Ofek and Yermack (2000), that highownership managers negate much of the increase in the incentive effect by selling previously owned shares. Our results show that past performance positively affects selling and negatively affects holding of vested equity, although the effect is not always significant. Our findings are also consistent with Huddart and Lang’s (2003) and Bartov and Mohanram’s (2004) findings that CEOs time the selling/exercise after stock run-up. In addition, our results suggest that future EPS, measured using 1 year forward-looking EPS normalized by the current share price, decreases the selling of vested equities. This result is broadly consistent with their interpretation, and with the results reported by Noe (1999) and Ke et al. (2003).

ARTICLE IN PRESS 34

L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

Table 4 Regression of selling of vested equity: stock alone and stock and optionsa Predicted Dependent variable: selling of vested stocks alonec sign Regression Economic Regression Economic coefficient significance coefficient significance

Dependent variable: selling of vested stocks and options Regression coefficient

Economic significance

Regression coefficient

Economic significance

Tax burden



0.92

+

Cost of underdiversification

+

7.82 (5.62) 8.43 3.36 0.32

0.94

Negative tax burden

7.64 (3.97) 7.61 (2.78) 0.22

Overconfidence



Independent variablesb

Lagged holding of vested equity Current grant of stock and options Past performance Size Book to market ratio Future EPS/price Institutional ownership Flag for CEO being chairman

7.09 (4.07) 8.16 (2.73) 0.29 (0.53) 0.11 (1.14) 0.13

0.98 0.08

0.04

7.39 (4.91) 8.00 3.40 0.37 (0.54) 0.14 (1.83) 0.11

0.89 0.96 0.11

0.05

(0.42) 0.10 (0.94) 0.13

0.92 0.00

0.00

(0.48) 0.15 (1.82) 0.11

(8.68) 0.03

(3.66) 0.08

(9.11) 0.00

(4.19) 0.04

(1.24) 0.13 (1.81) 0.04 (1.84) 0.02 (1.49) 1.77 (0.63) 0.43

(2.40) 0.15 (2.69) 0.06 (0.94) 0.07 (1.40) 2.02 (1.57) 0.51

(0.06) 0.11 (1.80) 0.01 (1.02) 0.02 (4.04) 1.97 (0.73) 0.41

(1.01) 0.14 (2.70) 0.02 (0.33) 0.07 (1.68) 1.70 (1.24) 0.36

(1.84) 0.12

(3.15) 0.13

(1.71) 0.10

(1.79) 0.15

(1.62)

(1.39)

2,218

(0.78) 0.00 (1.01) 0.01 (0.76) 1,159

2,218

(0.95) 0.00 (1.07) 0.01 (0.51) 1159

0.2484

0.2260

0.2518

0.2286

CEO tenure CEO age Number of observations Average adjusted Rsquared

0.85

1.02 0.09

0.05

a Fama and MacBeth (1973) cross-sectional regressions are estimated each year from 1996 to 2002. The standard errors of the time-series of the estimated coefficients are adjusted for potential serial correlation using Pontiff (1996) and assuming an AR(1) process on the residual terms. t-Statistics are reported under each coefficient in parenthesis. b Independent variables are as defined in Table 1, with the exception of negative tax burden, which is defined to be the same as tax burden if tax burden is negative, and zero otherwise. c The dependent variable, selling of vested stocks is the net reduction of vested stocks, after accounting for vesting of restricted stocks and exercising of stock options, divided by the total shares outstanding of the firm, and then multiplied by 100.

Like Huddart and Lang (1996) and Core and Guay (2001), we find that divesting of firm equity is positively related to stock returns during the preceding period. This might be reflective of a behavioral bias on the part of managers, but we refrain from such an interpretation because we have separately included at least one proxy for the manager’s behavioral bias, namely overconfidence. Results in the next two columns of Table 4 show that including controls for CEO tenure and age do not alter the findings qualitatively. The importance of tax burden in discouraging a CEO from selling his stock rises considerably. This is also true of the current grant of stock and options, past return performance, and of institutional ownership. The number of observations, unfortunately, is only about 25% as large as that without requiring CEO tenure and age data.

4.3.2. Determinants of the sale of both vested stock and options The preceding analysis focuses only on the changes in a CEO’s stock holdings. However, the CEO’s wealth and risk are affected by options as well as stock. Moreover, as a CEO exercises his options and receives shares as a result, the risk of his investment in the firm rises mechanically because option delta is strictly below one. It is possible that the CEO’s decision to sell stock is in part to undo the increase in risk from his exercise of options, or from an increase of the option delta as the stock price rises. Therefore, we examine the effect of taxes, overconfidence, and other determinants on a CEO’s decision to

ARTICLE IN PRESS L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

35

Table 5 Regression of selling of vested equity: stock alone from 1996 to 2002: above-median-age CEOsa Independent variablesb

Dependent variable: selling of vested stocksc Predicted sign

Regression coefficients

Economic significance

Regression coefficients

Economic significance

Tax burden



1.04

+

Cost of under-diversification

+

Overconfidence



8.29 (3.33) 7.08 1.94 0.34 (0.50) 0.27 (2.63) 0.08 (4.08) 0.07 (1.61) 0.09 (0.91) 0.03 (0.31) 0.10 (1.96) 2.12 (2.36) 0.15 (0.43) 0.26 (3.24) 0.00 (1.10) 0.03 (1.02) 835 0.2332

1.00

Negative tax burden

8.65 (3.55) 7.90 (2.41) 0.13 (0.24) 0.12 (0.84) 0.12 (9.26) 0.00 (0.12) 0.13 (2.30) 0.01 (0.50) 0.01 (0.53) 2.29 (0.54) 0.33 (1.38) 0.17 (4.43)

Lagged holding of vested equity Current grant of stock and options Past performance Size Book to market ratio Future EPS/price Institutional ownership Flag for CEO being Chairman CEO tenure CEO aged Number of observations Average adjusted R-squared

1,628 0.2612

0.95 0.04 0.04

0.85 0.10 0.09

a Fama and MacBeth (1973) cross-sectional regressions are estimated each year from 1996 to 2002. The standard errors of the time-series of the estimated coefficients are adjusted for potential serial correlation using Pontiff (1996) and assuming an AR(1) process on the residual terms. t-Statistics are reported under each coefficient in parenthesis. b Independent variables are as defined in Table 1, with the exception of negative tax burden, which is defined to be the same as tax burden if tax burden is negative, and zero otherwise. c The dependent variable, selling of vested stocks is the net reduction of vested stocks, after accounting for vesting of restricted stocks and exercising of stock options, divided by the total shares outstanding of the firm, and then multiplied by 100. d Above median age CEOs are the subsample of CEOs that are at least 57 years old in the year the analysis is performance on.

sell equity (i.e., stock and options). The next four columns of Table 4 report the results of estimating Eq. (2) except that the dependent variable is the annual change in total vested equity of a CEO. The results are similar to those using CEOs’ selling of vested stocks. For the remaining tables, we perform analysis on both the selling of stock alone and the selling/exercising of both stocks and options. The results are qualitatively similar. To save space, we only report the results for the stock selling. 4.3.3. Test on the sub-sample of CEOs who are less likely to be liquidity constrained The above analysis might still leave the possibility that liquidity considerations play a role in some CEOs’ decision to sell. To address the concern, we examine a sub-sample comprising older CEOs (above the age of 57, which is the median age of CEOs). Presumably, older CEOs are less subject to the liquidity constraint, given their accumulation of wealth over time. Table 5 shows that older CEOs are more tax sensitive, implying thereby their decisions are less influenced by liquidity and more by taxes. Also, we confirm the finding in Table 4 that only positive tax burden impacts selling. Therefore, below we focus only on observations with positive tax burden. 4.3.4. Time-series tests The tests so far have been based on annual cross-sectional regressions. We next check whether the pattern of a taxinduced (lack-of) selling is observed for a CEO over time. We accomplish this using the Fama-MacBeth methodology. Specifically, we (i) perform a time-series regression for each of the CEO in our sample and estimate the impact of tax and

ARTICLE IN PRESS 36

L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

Table 6 Time series regression of selling of vested stock on tax and under-diversification Independent variablesa

Dependent variable: selling of vested stocksb Predicted sign

Regression coefficients

Positive tax burden

–

Under-diversification

–

7.92 (2.89) 0.17 (0.36) 425

Number of observations

For this regression, we first run a time series regression for each executive over time, on all independent variables included in Table 4, except tax burden and under-diversification. We then take out the residual value of the dependent variable and use that residual as the dependent variable for the second round regression, where the independent variables are tax burden and under-diversification. The resulting coefficients of the second stage regression are treated in a fashion similar to Fama-MacBeth, to get the average and the standard deviation of the coefficients, and the t-statistics is calculated based on those. a Independent variables are as defined in Table 1, with the exception of negative tax burden, which is defined to be the same as tax burden if tax burden is negative, and zero otherwise. b The dependent variable, selling of vested stocks is the net reduction of vested stocks, after accounting for vesting of restricted stocks and exercising of stock options, divided by the total shares outstanding of the firm, and then multiplied by 100.

under-diversification, while ignoring all other factors. We then (ii) take all the CEOs, and aggregate the coefficient estimates for them, to calculate the mean and the standard deviation of these coefficients, and from those, we draw statistical inferences. We find that tax significantly decreases selling, while the impact of under-diversification is insignificant. The only problem with the above research design is that we must exclude some control variables included in the crosssectional regressions. We have a maximum of 7 years of data for each CEO, which means including all control variables leads to an over-identified regression. To overcome the problem arising from too few observations per CEO, we undertake an indirect method of incorporating all the control variables. We accomplish this in two stages. First, we estimate the crosssectional regression model reported in Table 4 column (1) to obtain a ‘‘residual value’’ of equity selling that is not explained by all of the control variables in the regression (i.e., other than the tax and under-diversification variables). The residual value is obtained for each CEO-year observation. Second, using these residual selling values for each CEO, we estimate a time-series regression with only tax and under-diversification as the independent variables. This approach accounts for the impact of all of the control variables, while examining the impact of tax and under-diversification in time series for individual CEOs. The results of implementing this time-series approach appear in Table 6. The evidence is consistent with taxes discouraging a CEO’s from selling equity over time. The impact of under-diversification continues to be insignificant.

4.3.5. Propensity of selling as a function of tax burden In addition to analyzing whether the quantity of shares sold might be affected by the tax burden, we study whether the probability of selling is decreasing in the capital gains tax burden. We use a Probit model to explain the incidence of selling. The dependent variable is one if the CEO sells stock in the current year, and zero otherwise. The explanatory variables are those used previously. The results reported in Table 7 show that a positive tax burden significantly reduces the propensity for equity selling. We also find that, controlling for tax burden, a higher cost of under-diversification increases the likelihood of selling, and the effect is statistically significant. Taken together, (i) both the probability of selling and the aggressiveness in selling are affected by the tax burden; (ii) there is some evidence that under-diversification affects, albeit to a lesser extent than tax burden, the probability of selling; and (iii) overconfidence does not affect the selling probability.

4.3.6. Does superior past performance reduce selling because of price appreciation? An important concern in concluding that tax burden reduces the CEO’s selling of stock is that the result could be merely due to the confounding effect of past performance, which correlates positively with tax burden.16 Specifically, superior past performance leads to stock-price appreciation. As a result, holding the CEO’s liquidity needs constant, the CEO can raise needed proceeds by selling fewer shares, which is precisely our result. To discriminate between the two explanations, we re-estimate the model in Table 4 except with the inclusion of the natural log of the stock price at the end of the year, Pt, and the natural log of the tax basis (i.e., the price at which the CEO acquired the stock), P0. If the negative coefficient on tax burden is spuriously capturing the effect of ending price on the CEO’s need for a given amount of proceeds, then we expect Pt, but not P0, to be significant. 16 We include past performance itself as a control variable, but it loads positively, not negatively. Therefore, the concern we address here is whether tax burden might be picking up the primary effect of price appreciation and loading negatively in the model because a CEO’s liquidity needs can be met by selling fewer shares of an appreciated security.

ARTICLE IN PRESS L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

37

Table 7 Probit regression of propensity to sell vested stocka Independent variablesb

Dependent variable: selling of vested stocksc Predicted sign

Regression coefficients

Regression coefficients

Positive tax burden



Cost of under-diversification

+

Overconfidence



3.70 (6.12) 4.25 (2.36) 0.02 (0.24) 0.02 (4.95) 0.04 (3.52) 0.28 (8.93) 0.21 (11.03) 0.08 (6.73) 0.11 (0.88) 0.28 (2.81) 0.05 (0.89)

5.19 (5.91) 4.27 (3.19) 0.14 (1.34) 0.02 (4.18) 0.05 (3.31) 0.38 (7.99) 0.18 (6.38) 0.10 (5.23) 0.30 (1.48) 0.56 (3.87) 0.05 (0.51) 0.02 (5.74) 0.00 (0.25) 2,918 0.2441

Lagged holding of vested equity Current grant of stock and options Past performance Size Book to market ratio Future EPS/price Institutional ownership Flag for CEO being Chairman CEO tenure CEO age Number of observations Average adjusted R-squared

4,867 0.2569

a Fama and MacBeth (1973) cross-sectional regressions are estimated each year from 1996 to 2002. The standard errors of the time-series of the estimated coefficients are adjusted for potential serial correlation using Pontiff (1996) and assuming an AR(1) process on the residual terms. t-Statistics are reported under each coefficient in parenthesis. b Independent variables are as defined in Table 1, with the exception of negative tax burden, which is defined to be the same as tax burden if tax burden is negative, and zero otherwise. c The dependent variable, selling of vested stocks is the net reduction of vested stocks, after accounting for vesting of restricted stocks and exercising of stock options, divided by the total shares outstanding of the firm, and then multiplied by 100.

Table 8 shows that both Pt and P0 are highly significant with predicted negative and positive signs, with approximately equal-sized coefficients. This is consistent with tax burden, as proxied in this regression using the combination of Pt and P0 variable, affecting the CEO’s selling decision. The results are inconsistent with the CEO’s liquidity need being fulfilled by the current stock price. In addition, past abnormal performance is separately included in the model, as in Table 4, to account for the effect of past performance via behavioral or other mechanisms on the CEOs’ selling decision. It has a significant positive coefficient as before. Other determinants have coefficients not too different from those in Table 4. Overall, the evidence is consistent with tax burden being a significant negative driver of CEO’s selling of equity. We repeat similar analysis by including Pt and P0 in the rest of the empirical analysis, and find no change in the tenor of the results.

4.4. Discriminating between taxes, under-diversification, and overconfidence Tables 4–8 show the importance of taxes in CEOs’ decision to sell equity. However, a remaining potential criticism is whether our proxies for under-diversification, overconfidence and taxes truly measure the distinct motivations affecting a CEO’s decision to sell equity. Specifically, one source of managerial overconfidence is likely to be managers’ biased attribution of past superior performance to their own skill, which could make managers overconfident. Since superior performance typically translates into a stock price run up and thus generates tax burden, our proxy for taxes might also be serving as a proxy for overconfidence. In addition, a higher company stock price compared to the market performance would also necessarily increase the percentage weight of the CEO’s portfolio on company stock, unless he takes offsetting actions to sell some firm equity. Thus it might also be related to the degree of under-diversification. Therefore, the results in Tables 4–8 might not unambiguously inform us about the importance of taxes independent of under-diversification and overconfidence. We use the behavior of tax-sensitive institutional investors to address the weakness. If tax, under-diversification and overconfidence are so intertwined, such that the tax variable merely proxies for underdiversification and overconfidence, then a comparison between managers and tax-sensitive institutions can uncover the

ARTICLE IN PRESS 38

L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

Table 8 Regression of selling of vested stock: separate coefficients on Pt and P0 termsa Independent variablesb

Dependent variable: selling of vested stocksc Predicted sign

Regression coefficients

Regression coefficients

log(Pt)



log(P0)

+

Cost of under-diversification

+

Overconfidence



0.38 (2.31) +0.29 (2.01) +0.39 (0.63) 0.14 (1.26) 0.12 (7.54) 0.03 (1.52) 0.09 (1.40) 0.00 (0.08) 0.02 (0.96) 1.86 (0.64) 0.39 (1.58) 0.12 (1.61)

0.31 (3.89) 0.26 (3.47) 0.16 (0.24) 0.16 (1.99) 0.11 (3.84) 0.08 (2.35) 0.11 (2.23) 0.03 (0.67) 0.06 (0.95) 1.56 (1.59) 0.58 (2.80) 0.15 (0.82) 0.00 (1.25) 0.01 (0.77) 1,159 0.2124

Lagged holding of vested equity Current grant of stock and options Past performance Size Book to market ratio Future EPS/price Institutional ownership Flag for CEO being Chairman CEO tenure CEO age Number of observations Average adjusted R-squared

2,218 0.2372

a Fama and MacBeth (1973) cross-sectional regressions are estimated each year from 1996 to 2002. The standard errors of the time-series of the estimated coefficients are adjusted for potential serial correlation using Pontiff (1996). t-Statistics are reported under each coefficient in parenthesis. b Independent variables are as defined in Table 1. c The dependent variable, selling of vested stocks is the net reduction of vested stocks, after accounting for vesting of restricted stocks and exercising of stock options, divided by the total shares outstanding of the firm, and then multiplied by 100.

incremental impact of under-diversification and overconfidence on managers’ equity-selling decisions. Assuming that the tax measure also proxies for under-diversification and overconfidence, we expect managerial selling to be more sensitive to such a tax measure than institutional selling, if we assume that institutions are not overconfident and also able to hold a well-diversified portfolio.17 Taxes, under-diversification and overconfidence should all discourage managerial equity selling, whereas only tax considerations should discourage institutions selling the equity. Thus, we predict a negative coefficient of larger magnitude on the tax variable for CEOs than for institutions in a matched sample. Table 9 reports results of comparing the selling behavior of CEOs with those of taxable institutions. Institutions with mainly taxable clientele are identified using the approach developed in Jin (2006). In particular, using the investor profiling information obtained from the Investment Adviser Public Disclosure (IAPD) data maintained by the SEC, we classify as taxsensitive those institutions whose clientele consists primarily (X50%) of tax-sensitive investors such as high net worth individuals. The analysis is conducted using a matched sample of CEOs selling equity and matching tax-sensitive institutions that have also engaged in selling the same stock during the year. Specifically, for CEO selling equity, we identify all tax-sensitive institutions selling the same stock during the year, and randomly choose one institution. If a matching institution cannot be identified, we drop the CEO observation from the regression. The estimated regression model is: Selling of stock ¼ a0 þ a1 CEO dummy þ a2 positive tax burden þ a3 CEO dummy  positive tax burden þ a4 CEO dummy  cost of under-diversification þ a5 CEO dummy  overconfidence þ control variables þ error:

(3)

17 Institutions might also exhibit irrational behavior such as overconfidence, but it might not be as salient as that of the CEOs. To the extent that institutional managers are also overconfident, our test shows whether CEO overconfidence affects their selling behavior above and beyond that of the institutions.

ARTICLE IN PRESS L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

39

Table 9 Managerial and institutional sellings combined regressions from 1996 to 2002a,b Independent variablesc

Manager dummy Positive tax burden Manager dummy  positive tax burden Manager dummy  cost of under-diversification

Manager dummy  overconfidence Past performance Size Book to market ratio Future EPS/price Institutional ownership Number of observations Adjusted R-squared

Dependent variables Selling of stock as a percentage of existing stock holdings

Selling of stock as a percentage of total equity (stock and options) holdings

Coefficient estimates

Coefficient estimates

0.03 (0.36) 1.59 (4.28) 0.14 (0.63) 0.18 (2.40) 0.01 (1.65) 0.02 (4.16) 0.01 (1.44) 0.00 (0.55) 0.14 (2.49) 0.00 (2.21) 3,760 0.1275

Economic significance

0.19 0.02

0.22 (6.93) 1.27 (4.16) 0.81 (3.52) 0.07 (1.87) 0.01 (1.09) 0.01 (3.47) 0.01 (2.20) 0.00 (3.10) 0.12 (2.14) 0.00 (1.24) 3,760 0.2867

Economic significance

0.15 0.10

a Fama and MacBeth (1973) cross-sectional regressions are estimated each year in 1996–2002. The standard errors of the time-series of the estimated coefficients are adjusted for potential serial correlation using Pontiff (1996) assuming an AR(1) process on the residual terms. t-Statistics are reported under each coefficient in parenthesis. b For each CEO selling we find a matching selling of the taxable institutional investors, and we only retain those CEOs for which we have a matching institutional selling. c Independent variables are as defined in Table 1.

where CEO dummy variable equals one for the CEO observations, and zero for the matching institutional investor observations. All other variables are as described earlier in the paper. In Eq. (3), the CEO dummy is interacted with the tax burden, cost of under-diversification, and overconfidence. These imply that the level of overconfidence and the cost of under-diversification for institutional investors are set to zero. In other words, we measure a CEO’s overconfidence above and beyond that of institutional investors, and we assume that such institutions are already welldiversified or could choose to be diversified. The coefficient on CEO dummy interacted with tax burden estimates whether CEOs’ propensity to optimize on taxes via their selling behavior is greater or less than that exhibited by institutional investors. Table 9 reports the regression results. For the institutional investor, selling refers only to stock selling in both sets of regressions. For the CEO and institutions, selling is expressed as a percentage of the respective CEO’s and institution’s stock holdings and not as a percentage of the shares outstanding for the firm. The reason is that the amount of ownership for institutions and CEOs is likely to be of different orders of magnitude, thus rendering a direct comparison of the number of shares sold non-comparable for the two groups. Results in Table 9 clearly demonstrate the importance of taxes on CEOs’ and institutions’ decision to sell stock. The coefficients on tax burden, 1.59 (t-statistic 4.28) and 1.27 (t-statistic 4.16), respectively, are highly significant in both sets of regressions. The tax burden variable interacted with the CEO dummy is positive, and significant in the second regression specification, suggesting that CEOs are not as sensitive to tax implications of selling as institutional investors with taxable clientele. The decline in CEOs’ sensitivity is more salient when their selling behavior is examined using their entire equity portfolio (i.e., stock and options), not just the stock portfolio. The coefficient estimate in the last column suggests CEOs are only one-third as sensitive as institutional investors to tax burden. In Table 9, the managerial overconfidence variable in the first column has a marginally significant coefficient, 0.01 (t-statistic ¼ 1.65), when stock sales are measured as a percentage of the CEO’s stock portfolio, and the coefficient in the third column is not significant when stock sales are a percentage of the CEO’s stock and option holdings. Recall that the overconfidence is assumed to be zero for the institutional investors included in the regressions. Overall, the results for the tax variable and the overconfidence variable together suggest; (i) taxes significantly influence CEOs’ and institutions’ decision to sell equity, and this effect of tax is not likely to be overconfidence in disguise (because otherwise the effect of

ARTICLE IN PRESS 40

L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

taxes on lack of selling should be stronger, not weaker, for managers), and (ii) CEOs exhibit a less tax-savvy behavior compared to institutional investors or that the CEOs have other means of reducing their tax exposure. While this would still be a behavioral bias, it is distinct from CEO overconfidence. It appears to be consistent with the disposition effect documented for individual investors because CEOs sell more (not less) than the presumably rational taxable investors following stock price run-ups. However, the usual caveats apply: (i) we might not have measured overconfidence accurately, and (ii) under-diversification and other considerations that are likely to be critical to CEOs, but not to institutional investors. These might be confounding the results and leading to a muted effect of overconfidence on the managers’ selling decision. 4.5. Additional tests and robustness checks The following robustness checks provide results that are qualitatively similar to those reported in the tables in the paper. 4.5.1. Explaining the level of vested total equity holdings Sale of stock by CEOs might not reduce their incentive because of new stock and option grants and option exercises. Therefore, although selling of stock is informative and of interest, CEOs might be optimizing on their level of equity holdings, and the sale of stock be only a byproduct of that optimization. Additionally, only about 40% of CEOs sell stock in any given year, which means the sample size more than doubles when we explain the level of their equity holdings. We perform additional tests explaining CEOs’ level of vested stock and option holdings. 4.5.2. Additional regression specifications All the tables present results using Fama-MacBeth regressions with a fixed industry effect to control for implicit industry-level labor-market contracts that might affect CEOs’ equity selling decisions as well as compensation practices across industries. We believe these results are conservative in controlling for various confounding effects and in gauging the statistical significance. However, we also estimated regressions using other approaches. For example, we use a regression clustering at the industry level, and include annual fixed effect. The panel data in our study are likely to exhibit both time-series and cross-sectional correlation. The regression with clustering at industry level with annual fixed effects accounts for a bias due to the correlations (see Petersen and Mitchell, 2005).18 4.5.3. Alternative measures of option positions through time We experimented with alternative methods of estimating CEOs’ options positions through time. The tabulated results are based on the Core and Guay (1999) method. We repeat the analysis with options positions estimated via a detailed method of accounting for each batch of options using the grant-date information, as developed in Jin and Meulbroek (2004). 4.5.4. Alternative measures of under-diversification We also tried the Meulbroek (2001) measure of the cost of under-diversification. Meulbroek calculates a deadweight loss associated with under-diversification in a non-utility-based way. Her approach assumes that CEOs assign the same risk-return tradeoff (or Sharpe ratio) as outsiders, except that CEOs (due to lack of diversification) bear some of the firmspecific risk, whereas diversified outsider investors bear only systematic risk. 4.5.5. Alternative measure of tax burden The tax burden measure is calculated as the tax based on selling the entire portfolio at any point in time. Thus, it is also a measure of the ‘‘average tax burden’’ faced by CEOs. Average tax basis is one common way to calculate the capital gains tax, but an equally popular alternative is to use the marginal tax that faces CEOs for selling the first 20% of their portfolios. We chose sale of the first 20% of the portfolio because the median amount sold is about 15% of the portfolio and the mean 24%. To implement the marginal tax rate, we further assume that when the CEOs sell, they first sell the most recently acquired shares. 4.5.6. Alternative measure of overconfidence: CEO’s stock purchase The measure of overconfidence used in the analysis exactly follows the Malmendier and Tate (2005) methodology. To check the robustness of the results to alternative measures of overconfidence, we construct another proxy for overconfidence. A CEO who purchases stock on the open market in addition to stock grants from the firm is likely more optimistic about the firm. Moreover, because the CEO’s stock purchase is incremental to the stock grants and option exercises during the year, such purchases cannot be motivated by tax considerations and are unlikely to be due to concerns about under-diversification. We attribute the buying behavior to CEO overconfidence. Empirically, this measure suffers from one problem: because CEOs who purchase their own firms’ shares are designated as overconfident, the measure is 18

We also verified the results in Table 9 to be robust to interacting the CEO dummy with all other control variables.

ARTICLE IN PRESS L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

41

mechanically positively correlated with managers’ level of equity holdings. The coefficient on the overconfidence dummy variable is expected to capture the marginal impact of overconfidence as well as the mechanical increase in CEOs’ equity holdings as a result of buying. Therefore, our primary goal in using this alternative measure of overconfidence is to examine whether taxes continue to play a role in managers’ decisions about level of equity holdings. More important, the estimated effect of tax burden is likely to be conservative because the mechanical negative relation between the new measure of overconfidence and selling of equity might inappropriately attribute some of the impact of taxes to overconfidence. Unreported results show, not surprisingly, that overconfidence is a stronger determinant of equity holdings than tax burden in this regression specification. However, notwithstanding a mechanical relation between overconfidence and equity holdings, tax burden continues to affect CEOs’ level of equity holdings.

4.5.7. Liquidity-motivated selling CEOs might sell stock and exercise options for liquidity reasons. CEOs’ liquidity constraints are not directly observable, but, presumably, younger CEOs and those with salary plus bonus significantly lower than other CEOs might face higher liquidity demands than other CEOs. We focus on CEOs who were compensated more than $1 million in the past year. Although the $1 million cut-off is obviously ad hoc, roughly half of the observations in our sample have total compensation in excess of $1 million.

5. Conclusion CEOs hold large amounts of firm equity, and such holdings change significantly through time with new grants, exercise of vested options, and sale of vested stocks. We analyze CEOs’ selling of equity. We find that CEOs frequently sell large amounts of their vested equity, which runs counter to the conventional wisdom that CEOs avoid selling firm equity for fear of an adverse stock market reaction. We also examine the determinants of CEOs’ stock selling behavior. In particular, we focus on assessing the relative importance of taxes, under-diversification, and managerial behavioral bias in a CEO’s decision to sell his equity in a given year. CEO ownership of vested firm equity beyond the levels justified by efficient portfolio considerations (i.e., diversification) is often attributed to managerial overconfidence (see, for example, Malmendier and Tate, 2005). However, an immediate tax liability associated with selling equity that has appreciated might also discourage managers from selling firm equity. In contrast, high under-diversification cost would motivate the CEO to sell equity. Our analysis suggests that taxes swamp overconfidence as a factor influencing CEOs’ decision to sell their vested equity, subject to the caveat that our measure of under-diversification is not very precise. Taxes also appear to be a stronger motivation than the under-diversification cost in the selling decision. The importance of taxes in a CEO’s decision to sell firm equity remains statistically and economically significant even after controlling for other determinants of the CEO’s stock selling behavior.19 We also compare a CEO’s selling decision with that of a tax-sensitive institutional investor, who also faces an immediate tax liability on realized capital gains upon selling a stock. We find that both the CEO and the tax-sensitive institution respond to tax incentives. However, the CEO’s response is weaker than that of the institutional investor. Taken together, our results suggest that in their selling decisions, CEOs exhibit a first-order influence of tax considerations, but CEOs might still exhibit some behavioral bias similar to a disposition effect, compared to fully rational financial investors. Notwithstanding the a priori lesser importance of taxes compared to under-diversification as explained in Section 3, our results suggest that taxes affect CEOs’ selling decision as much, if not more, than CEO under-diversification. The evidence suggests either (i) managers are excessively concerned about the immediate tax consequence of selling equity, and/or (ii) they have lower risk-aversion than indicated in the extant literature and thus do not value diversification nearly as much as existing studies might suggest, and/or (iii) labor contracts require CEOs to hold stock for incentive reasons and that these incentives are highly correlated with the tax burden (which seems implausible to us); or (iv) instead of selling equity to achieve better diversification, those CEOs facing the highest under-diversification costs privately hedge their exposure through capital market transactions, which achieves diversification without sacrificing the tax advantage of holding vested equity. We do not discriminate among these alternative explanations, but leave it for future research. Our findings have implications for corporate compensation policies as they relate to CEOs’ incentive level, and firms’ investment and financing policies. Specifically, (i) CEOs with tax overhang equity might have an under-diversified equity portfolio because these CEOs are less willing to sell vested equities. These firms thus might want to tailor the compensation package to avoid excessive equity incentive by giving more cash (bonus) instead of new grants of stock and options (see Core and Guay, 1999); (ii) unless a firm adjusts its compensation policy, the excessive incentive from tax overhang might distort the firm’s investment and financing policies; and (iii) CEOs are responsive to tax considerations and actively 19 There could be other, non-taxed actions that CEOs could take to reduce their total beneficial holdings of firm equities. For example, CEOs could gift their stock to charity or to their children. These actions reduce the CEO’s equity position and would be classified as ‘‘net selling’’ by our selling measure. To the extent that these activities are more than sporadic, they render the measured selling less responsive to the tax burden. Empirically, we document tax sensitivity of selling despite such mitigating factors, which suggests an even more significant true effect of taxes.

ARTICLE IN PRESS 42

L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

trade-off tax and other considerations in their personal portfolio decisions. We thus have reason to believe that they might exhibit similar behavior in their corporate decision making.

Appendix A. Procedure to estimate annual grants of options ExecuComp contains detailed information on options granted in the current year, but only limited information on the CEO’s option holdings as a result of grants from the past. In particular, ExecuComp contains information on the number of shares and realizable values (number of options times the excess of stock price over exercise price) of unexercisable and exercisable options held by the CEOs. The year-end option position of a CEO could result from multiple grants from the past and the current year. Detailed terms of those option grants (exercise price, remaining time-to-maturity) are not reported in ExecuComp. Hence, a researcher must estimate past annual grants to derive the total exposure from options.

A.1. Core and Guay approach Core and Guay (2002, pp. 615–619) provide a simple approximation method. Briefly, they first subtract the number of shares and dollar value of the current year option grant (if any) from the pool of unexercisable options to derive the number of shares and dollar value of unexercisable options from past grants. They treat the remaining exercisable and unexercisable options as two single grants, and estimate the exercise price and the remaining time-to-maturity of these two separate grants. The exercise price of each ‘‘grant’’ is derived from the reported realizable value. In particular, dividing the realizable, i.e., intrinsic values of unexercisable (excluding current year grants) and exercisable options by the number of unexercisable (excluding current year grants) and exercisable options of a CEO, respectively, yields estimates of how far each of these groups of options is ‘‘in the money.’’ Subtracting these average ‘‘intrinsic values’’ per option from the firm’s stock price generates estimates of the average exercise price of the unexercisable and exercisable options. Core and Guay (2002) estimate the remaining time-to-maturity of previously granted options by again discriminating between the unexercisable (excluding current year grants) and exercisable options. In particular, if a firm grants options in the current year, they set the time-to-maturity of previously granted unexercisable options equal to the time-to-maturity of the current year grant minus 1 year. If no grant is made in the current year, they set the time-to-maturity of unexercisble options to 9 years. They assume that exercisable options have a time-to-maturity 3 years less than that of the unexercisable options. Core and Guay (2002) note that their approximation, though simple, yields good fit in broad samples of actual and simulated CEO option portfolios, capturing more than 99% of the variation in option portfolio value and sensitivities. They acknowledge that the most severe bias occurs in samples of CEOs with a large proportion of out-of-the-money options, but argue that, even in that case, the explanatory power of their measure remains above 95%.

A.2. Jin and Meulbroek and Hall and Knox approach Jin and Meulbroek (2004) and Hall and Knox (2004) modify the Core-Guay approximation to mitigate the potential bias that might arise from a large proportion of out-of-the-money options. The modifications are based on the intuition that, while options granted in previous years are not fully described in the current proxy statements, they were fully disclosed in the proxy statement in the year they were granted. Thus, if a sufficiently long history of the proxy statements and knowledge of the firm’s option vesting rule (which likely does not change over time) are available, the exact composition of a CEO’s option holdings is feasible. This entails separating the option portfolio into tranches that correspond to the grants from each of the previous years. The exact terms of the options (exercise price and remaining time-to-maturity) can be identified. The method is in theory more robust to the existence of out-of-money options, but requires detailed assumptions about the firms’ option vesting rules.

Appendix B. Measure of tax burden Intuitively, tax burden is the tax liability per dollar of proceeds from the sale of stock, calculated as the capital gains tax rate multiplied by taxable profit per share, divided by current stock price. Whereas the intuition is straightforward in the case of stock sales, the corresponding measure for options is not. Since employee stock options are not traded, proceeds from the ‘‘sale’’ of stock options assumes options are exercised and sold immediately. Thus, in the case of stock options, (i) we estimate the proceeds from the sale of options to equivalent amount of proceeds from the sale of shares using the option delta, and (ii) we calculate the tax liability to account for the fact that option realizations are typically taxed at a rate different from the tax rate applicable to capital gains upon selling shares. We first summarize the tax code pertaining to the taxation of stocks and options, then describe how we measure tax burden.

ARTICLE IN PRESS L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

43

B.1. Taxation of restricted stock Under current US tax law, a recipient of restricted stock owes no tax at the time of the grant because it is subject to forfeiture and, hence, risky. The recipient owes ordinary income tax on the entire value of restricted stock grants at the time the stock vests. Vesting thus triggers a large tax liability, especially if the stock has appreciated significantly from the time of grant. Alternatively, the recipient can elect, within 30 days of the grant date under Internal Revenue Code section 83(b), to be taxed on the value of the restricted stock at the ordinary income tax rate, and subsequently pay the (lower) capital gains tax on any further appreciation. In most cases, however, invoking an 83(b) treatment is not optimal (see McDonald, 2003), and empirically we are not aware of CEOs of large public companies systematically invoking the treatment. We therefore assume that CEOs never invoke the 83(b) treatment. B.2. Tax burden of vested stocks Given our assumptions, a restricted stock grant enters the tax calculation only when it becomes vested. The executive owes ordinary income tax on the entire value of the restricted stock at the time of vesting. Any appreciation subsequent to the vesting date accrues capital gains tax. We collect from ExecuComp detailed information on CEOs’ holdings of unrestricted stock at the end of each fiscal year from 1992 to 2002. From these, we estimate the tax basis, with the simplifying assumption that all vested shares during a year were vested at the fiscal year end, with the fiscal-year-end stock price as the new tax basis. When we experiment with other assumptions about the vesting price to get a measure of the tax basis for future capital gains (e.g., trading volume weighted average price during the year), the results are qualitatively unchanged. If any vested stock is sold during a year, we assume that CEOs sell first the shares with the highest tax basis, which minimizes realized capital gains or maximizes realized capital losses. We use the previous 5 years’ holdings and grant information to estimate the first observation of tax burden measure for a CEO. Thus, given that our data starts in 1992, our first year of observation for the tax burden is 1996. The unrestricted stock held by a CEO in the first year of data available on ExecuComp is assumed to have been granted 5 years earlier. For example, if the CEO first appears in ExecuComp in 1993, we assume the reported unrestricted stock in 1993 was granted in 1988. This approximation is necessary due to the lack of more detailed information in ExecuComp. The assumption introduces noise in the tax burden measure, which is likely to reduce its explanatory power in our tests, and thus likely work against finding significant results for our tax burden measures. B.3. Tax burden for options The tax treatment on different types of executive stock options varies (see Hall and Liebman, 2000). The non-qualified stock options (NQSOs) are the most widely used. Executives with NQSOs are taxed at the personal income tax rate on option profits (the difference between that stock price and the exercise price times the number of options) when the options are exercised. If the executive continues to hold the shares after exercise, subsequent appreciation is taxed at the capital gains rate upon the sale of the stock. Other types of executive stock options include Incentive Stock Options (ISOs) and Stock Appreciation Rights (SARs). ISOs are untaxed at grant and exercise, and only taxed upon selling the underlying stock for capital gains. ISOs account for about 5% of all option grants, according to Hall and Liebman (2000). SARs are essentially the same as NQSOs, save that the exercise is cash-settled. These are rarely used for top executives, according to Hall and Liebman (2000). The ExecuComp database provides no information on the type of option granted. For simplicity, we assume all stock options are NQSOs. Therefore, at any point, the tax liability on a vested option is the ordinary income tax rate times the option’s intrinsic value, that is, the difference between current stock price and exercise price.20 B.4. Total tax liability and our tax burden measure The total tax liability from stocks and options is the sum of the liabilities from (i) selling vested stock, and (ii) exercising vested options. This is the total tax that would have to be paid were all of a CEO’s unrestricted equity to be sold. To calculate a ‘‘per share equivalent tax liability,’’ we divide the total dollar tax liability by the aggregate delta, that is, option delta times number of options, plus the number of shares. The final measure of tax burden is thus: Tax burden ¼ Total dollar tax liability=½ðdoptions  N options þ Nstocks ÞPrice,

20 CEOs often have to pay the Alternative Minimum Tax (AMT). Under AMT, executives have to pay taxes on the spread between the market price and the exercise price of incentive stock options (ISOs) granted by their employers. For example, if an executive exercised an incentive stock option to buy 100 shares of stock for $3 a share and the stock was trading at $10, the spread would be $7 a share, or $700 for the 100 shares. Under the regular tax rules, the executive wouldn’t pay current taxes on that amount since it is an incentive (qualified) stock option, but under the AMT it’s considered income and thus taxable at the current period. Because tax is already levied on the current profit, if the executive chooses to continue to hold the stock after the exercise, additional future profits will be taxed as capital gains. Thus, the AMT makes the ISOs behave more like NQSOs.

ARTICLE IN PRESS 44

L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

where ‘‘Total dollar tax liability’’ is the taxable gain times the tax rate applicable in the year of sale of equity, and the taxable gain is from stock as well as options for which the applicable tax rates might be different, as discussed above. This ratio represents the average percentage tax liability from selling equity in the company. Table B1 provides a simple example of the calculation of the tax burden measure. Table B2 provides evidence that the raw measure of tax burden is negatively related to the amount of total equity sold by the CEO.

Table B1 A simple example of tax burden calculations Year

Shares owned (000s)

Stock price at fiscal year end

New shares acquired during the year (000s)

New share*EOY stock price ($000s)

Total tax basis ($000s)

Average tax basis

Capital gains—tax rate at fiscal year end (%)

Total tax liability if selling everything today ($000s)

Tax burden (%)

1992 1993 1994 1995 1996 1997 1998 1999

0.467 0.951 1.492 18.103 42.708 70.230 85.546 97.927

33.125 37.125 32.875 40.500 41.125 51.625 47.063 32.125

0.467 0.484 0.541 16.611 24.605 27.522 15.316 12.381

12.26 17.97 17.79 672.75 1011.88 1420.82 720.82 397.74

12.26 30.23 48.01 720.76 1732.64 3153.46 3874.28 4272.02

26.25 31.78 32.18 39.81 40.57 44.90 45.29 43.62

28 28 28 28 28 20 20 20

0.90 1.42 0.29 3.48 6.64 94.43 30.35 225.22

5.81 4.03 0.59 0.47 0.38 2.60 0.75 7.16

Note: The data are for Dr. E. Linn Draper, Jr., Chairman, President & CEO, Americal Electric Power Company, Inc. Except for the first year, New Shares Acquired during the year is the differences between the current year stock holding and the previous year stock holdings. New shares times the end-of-year stock price is used as an input to calculate the total tax basis. The total tax basis is the sum of the incremental shares times the end-of-year stock price. Average tax basis is the total tax basis divided by the number of shares held. The unrestricted stock held by a CEO in his first year of data available on ExecuComp is assumed to have been acquired five years earlier. In this example, the first years holdings are assumed to be acquired in 1987, at the fiscal yearend stock price of $26.25 (quoted from COMPUSTAT). Perterbation around this assumption has negligible impact on the tax burden measures, since we are focusing on the observations after 1996. The long-term capital gains tax rate quoted is the rate at the end-of-fiscal-year. The company fiscal yearend is December. A change in the long-term capital gains tax from 28% to 20% occurs in May 1997. Thus the capital gains tax rate faced at the end of year 1997 is 20% instead of 28%. Tax burden ¼ Total tax liability/(total shares  current share price).

Table B2 Reduction of holding of vested equity: how it changes with tax burden measure 1996– 2002 Quintiles of tax burden

Average tax burden (%)

Mean reduction of total vested equity

Standard deviation of reduction of total vested equity

Number of observations

1 2 3 4 5

8.84 1.13 4.39 7.71 11.84

1.76 0.86 0.63 0.58 0.59

6.25 3.12 1.66 1.28 1.29

478 478 478 478 476

Each year, the CEOs are sorted into five quintiles based on the ranking of their tax burden measure. The data are then aggregated over all years. We then calculate the total share-equivalent reductions from vested stock and options for each CEO, and take the mean and standard deviations of the reduction for each quintile. t-Statistics for the hypothesis that mean reduction of vested equity are equal for quintiles 1 and 5 ¼ 87.88.

References Aggarwal, R., Samwick, A., 1999. The other side of the trade-off: the impact of risk on executive compensation. Journal of Political Economy 107, 65–105. Auerbach, A., 2002. Taxation and corporate financial policy. In: Auerbach, A., Feldstein, M. (Eds.), Handbook of Public Economics, vol. 4. Ayers, B., Lefanowicz, C., Robinson, J., 2003. Shareholder taxes in acquisition premiums: the effect of capital gains taxation. Journal of Finance 58, 2783–2801. Baker, G., Hall, B., 2004. CEO incentives and firm size. Journal of Labor Economics 22, 767–798. Barber, B., Odean, T., 2000. Trading is hazardous to your wealth: the common stock investment performance of individual investors. Journal of Finance 55, 773–806. Bartov, E., Mohanram, P., 2004. Private information, earnings manipulations, and executive stock option exercise. Accounting Review 79, 889–920. Bertrand, M., Schoar, A., 2003. Managing with style: the effect of managers on firm policies. Quarterly Journal of Economics 118, 1169–1208. Bettis, J., Bizjak, J., Lemmon, M., 2001. Managerial ownership, incentive contracting, and the use of zero-cost collars and equity swaps by corporate insiders. Journal of Financial and Quantitative Analysis 36, 345–370.

ARTICLE IN PRESS L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

45

Blouin, J., Raedy, J., Shackelford, D., 2003. Capital gains taxes and equity trading: empirical evidence. Journal of Accounting Research 41, 611–651. Boehmer, E., Netter, J., 1997. Management optimism and corporate acquisitions: evidence from insider trading. Managerial and Decision Economics 18, 693–708. Carpenter, J., 1998. The exercise and valuation of executive stock options. Journal of Financial Economics 48, 127–158. Chen, M., 2004. Executive option repricing, incentives, and retention. Journal of Finance 59, 969–1442. Chidambaran, N., Prabhala, N., 2003. Executive stock option repricing, internal governance mechanisms, and management turnover. Journal of Financial Economics 69, 153–189. Constantinides, G., 1983. Capital market equilibrium with personal taxes. Econometrica 51, 611–636. Constantinides, G., 1984. Optimal stock trading with personal taxes: implications for prices and the abnormal January returns. Journal of Financial Economics 13, 65–89. Core, J., Guay, W., 1999. The use of equity grants to manage optimal equity incentive levels. Journal of Accounting and Economics 28, 151–184. Core, J., Guay, W., 2001. Stock option plans for non-executive employees. Journal of Financial Economics 61, 253–287. Core, J., Guay, W., 2002. Estimating the value of employee stock option portfolios and their sensitivities to price and volatility. Journal of Accounting Research 40, 613–630. Core, J., Larcker, D., 2002. Performance consequences of mandatory increases in executive stock ownership. Journal of Financial Economics 64, 317–340. Core, J., Guay, W., Verrecchia, R., 2003. Price versus non-price performance measures in optimal CEO compensation contracts. Accounting Review 78, 957–981. Dai, Z., Maydew, E., Shackelford, D., Zhang, H., 2006. Capital gains taxes and asset prices: capitalization or lock-in? Working Paper 12342, NBER. Dammon, R., Spatt, C., Zhang, H., 2001a. Optimal consumption and investment with capital gains taxes. Review of Financial Studies 14, 583–616. DeMeza, D., Southey, C., 1996. The borrower’s curse: optimism, finance, and entrepreneurship. The Economic Journal 106, 375–386. Ellis, K., Li, O., Robinson, J., 2006. Capital gains taxes and IPO under-pricing. Working Paper, University of Notre Dame. Erickson, M., Hanlon, M., Maydew, E., 2004. How much will firms pay for earnings that do not exist? Evidence of taxes paid on allegedly fraudulent earnings. The Accounting Review 79, 387–408. Fama, E.F., MacBeth, J.D., 1973. Risk, return, and equilibrium: Empirical tests. The Journal of Political Economy 81 (3), 607–636. Feldstein, M., Slemrod, J., Yitzhaki, S., 1980. The effects of taxation on the selling of corporate stock and the realization of capital gains. Quarterly Journal of Economics 94, 777–791. Garvey, G., Milbourn, T., 2003. Incentive compensation when executives can hedge the market: evidence of relative performance evaluation in the cross section. Journal of Finance 58, 1557–1581. Gervais, S., Odean, T., 2001. Learning to be overconfident. Review of Financial Studies 14, 1–27. Goolsbee, Austan, 2000a. What happens when you tax the rich? Evidence from executive compensation. Journal of Political Economy 108, 352–378. Goolsbee, Austan, 2000b. Taxes, high-income executives, and the perils of revenue estimation in the new economy. American Economic Review 90, 271–275. Gordon, R., Hines, J., 2002. International taxation. In: Auerbach, A., Feldstein, M. (Eds.), Handbook of Public Economics, vol. 4. Graham, John, R., 2003. Taxes and corporate finance: a review. Review of Financial Studies 16, 1074–1128. Greene, W., 2000. Econometric Analysis, fourth ed. Prentice-Hall, Upper Saddle River, NJ. Hall, B., Knox, T., 2004. Underwater options and the dynamics of executive pay-to-performance sensitivities. Journal of Accounting Research 42, 365–412. Hall, B., Liebman, J., 1998. Are CEOs really paid like bureaucrats? The Quarterly Journal of Economics 113, 653–691. Hall, B., Liebman, J., 2000. The taxation of executive compensation. Working Paper No. 7596, NBER. Hall, B., Murphy, K., 2000. Optimal exercise prices for executive stock options. American Economic Review 90, 209–214. Hall, B., Murphy, K., 2002. Stock options for undiversified executives. Journal of Accounting and Economics 33, 3–42. Hartzell, J., Starks, L., 2003. Institutional investors and executive compensation. Journal of Finance 58, 2351–2374. Heath, C., Huddart, S., Lang, M., 1999. Psychological factors and stock option exercise. Quarterly Journal of Economics 114, 601–627. Heaton, J., 2002. Managerial optimism and corporate finance. Financial Management 31, 33–45. Huddart, S., Lang, M., 1996. Employee stock option exercises: an empirical analysis. Journal of Accounting and Economics 21, 5–43. Huddart, S., Lang, M., 2003. Information distribution within firms: evidence from stock option exercises. Journal of Accounting and Economics 34, 3–41. Jeng, L., Metrick, A., Zeckhauser, R., 2003. Estimating the returns to insider trading: a performance-evaluation perspective. Review of Economics and Statistics 85, 453–471. Jensen, M., 1993. The modern industrial revolution, exit, and the failure of internal control systems. Journal of Finance 48, 831–880. Jensen, M., Murphy, K., 1990. Performance pay and top-management incentives. Journal of Political Economy 98, 225–264. Jenter, D., 2005. Market timing and managerial portfolio decisions. Journal of Finance 60, 1903–1949. Jin, L., 2002. CEO compensation, diversification and incentives. Journal of Financial Economics 66, 29–63. Jin, L., 2006. Capital gains tax overhang and price pressure. Journal of Finance 61, 1399–1431. Jin, L., Meulbroek, L., 2004. Do underwater executive stock options still align incentives? The effect of stock price movements on managerial incentivealignment. Working Paper, Harvard Business School. Kahneman, D., Tversky, A., 1979. Prospect theory: an analysis of decision under risk. Econometrica 47, 263–291. Ke, B., Huddart, S., Petroni, K., 2003. What insiders know about future earnings and how they use it: evidence from insider trades. Journal of Accounting and Economics 35, 315–346. Klein, P., 1999. The capital gain lock-in effect and equilibrium returns. Journal of Public Economics 71, 355–378. Klein, P., 2001. The capital gain lock-in effect and long horizon return reversal. Journal of Financial Economics 59, 33–62. Lakonishok, J., Lee, I., 2001. Are insider trades informative? Review of Financial Studies 14, 79–111. Lambert, R., Larcker, D., Verrecchia, R., 1991. Portfolio considerations in valuing executive compensation. Journal of Accounting Research 29, 129–149. Landsman, W., Shackelford, D., 1995. The lock-in effect of capital gains taxes: evidence from the RJR Nabisco leveraged buyout. National Tax Journal 48, 245–259. Levi, M., 1973. Errors in variables bias in the presence of correctly measured variables. Econometrica 41, 985–986. Li, Y., 2004. Maintaining optimal CEO incentives Through equity grants and CEO portfolio rebalancing. Working Paper, CUNY Baruch College. Malmendier, U., Tate, G., 2004. Who makes acquisitions? CEO overconfidence and the market’s reaction. Working Paper 10813, NBER. Malmendier, U., Tate, G., 2005. CEO overconfidence and corporate investment. Journal of Finance 60, 2661–2700. McDonald, R., 2003. Is it optimal to accelerate the payment of income tax on share-based compensation? Working Paper, Northwestern University. Meulbroek, L., 2001. The efficiency of equity-linked compensation: understanding the full cost of awarding executive stock options. Financial Management 30, 5–30. Miller, D., Ross, M., 1975. Self-serving biases in the attribution of causality: fact or fiction. Psychology Bulletin 82, 213–225. Mitton, T., Vorkink, K., 2007. Equilibrium under-diversification and the preference for skewness. Review of Financial Studies 20, 1255–1288. Noe, C., 1999. Voluntary disclosures and insider transactions. Journal of Accounting and Economics 27, 305–326. Odean, T., 1998. Are investors reluctant to realize their losses? Journal of Finance 53, 1775–1798. Ofek, E., Yermack, D., 2000. Taking stock: equity-based compensation and the evolution of managerial ownership. Journal of Finance 55, 1367–1384. Petersen, Mitchell, 2005. Estimating standard errors in finance panel data sets: comparing approaches. Working Paper, Northwestern University. Pontiff, J., 1996. Costly arbitrage: evidence from closed-end funds. Quarterly Journal of Economics 111, 1135–1151. Poterba, J., Weisbenner, S., 2001. Capital gains tax rules, tax-loss trading, and turn-of-the-year returns. Journal of Finance 56, 353–368. Reese, W., 1998. Capital gains taxation and stock market activity: evidence from IPOs. Journal of Finance 53, 1799–1820.

ARTICLE IN PRESS 46

L. Jin, S.P. Kothari / Journal of Accounting and Economics 46 (2008) 23–46

Roll, R., 1986. The hubris hypothesis of corporate takeovers. Journal of Business 59, 197–216. Schaefer, S., 1998. The dependence of pay–performance sensitivity on the size of the firm. Review of Economics and Statistics 3, 436–443. Seyhun, N., 1986. Insiders’ profits, costs of trading, and market efficiency. Journal of Financial Economics 16, 189–212. Shackelford, D., Verrecchia, R., 2002. Intertemporal tax discontinuities. Journal of Accounting Research 40 (1), 205–222. Shackelford, Douglas A., Shevlin, B., Terry, C., 2001. Empirical tax research in accounting. Journal of Accounting and Economics 31 (1–3), 321–387. Shefrin, H., Statman, M., 1985. The disposition to sell winners too early and ride losers too long: theory and evidence. Journal of Finance 40, 777–790. Viard, A., 2000. Dynamic asset pricing effects and incidence of realization-based capital gains taxes. Journal of Monetary Economics 46, 465–488. Yermack, D., 1995. Do corporations award CEO stock options effectively? Journal of Financial Economics 39, 237–269.