ARTICLE IN PRESS. Journal of Accounting and Economics

Transcription

1 Journal of Accounting and Economics 47 (2009) Contents lists available at ScienceDirect Journal of Accounting and Economics journal homepage: Taxes and the backdating of stock option exercise dates $ Dan Dhaliwal a, Merle Erickson b, Shane Heitzman c, a Eller College of Management, University of Arizona, AZ, USA b Graduate School of Business, University of Chicago, IL, USA c Simon Graduate School of Business, University of Rochester, Carol Simon Hall 3-160A, Rochester, NY 14627, USA article info Article history: Received 3 January 2007 Received in revised form 20 August 2008 Accepted 24 September 2008 Available online 11 October 2008 JEL classification: D82 H24 J33 K22 abstract We investigate the backdating of stock option exercises. Before SOX, we find evidence that some exercises were backdated to days with low stock prices. Consistent with a taxbased incentive, these suspect exercises are more likely when the personal tax savings from backdating are higher. However, suspect CEO exercises generate average (median) estimated tax savings of $96,000 ($7,000). These savings appear modest relative to the costs insiders and firms face. We find that the likelihood of a suspect exercise increases in the likelihood of option grant backdating. This suggests that agency problems associated with backdating permeate option compensation in some firms. & 2008 Elsevier B.V. All rights reserved. Keywords: Stock option compensation Backdating Taxes Insider trading Regulation 1. Introduction Recently, the potential backdating of insider stock option grants has become the subject of intense scrutiny and analysis in the financial press as well as research by the academic community. The outcome of this attention has resulted in criminal investigations by the US Department of Justice, executive and director departures, and accounting restatements. While the focus to date has been on the manipulation of stock option grant dates, namely the backdating of option grants to days with low stock prices, little attention has been given to the manipulation of option exercise dates. 1 Executives can realize economic benefits through the manipulation of option exercise dates, and following recent research related to option grant backdating we investigate whether option exercises show similar evidence of backdating. $ We appreciate comments and suggestions from Michelle Hanlon, Bob Holthausen (editor), Gregg Jarrell, Bill Schwert, Terry Shevlin (referee), Cliff Smith, Bob Trezevant, Joanna Wu, Jerry Zimmerman and workshop participants at Carnegie Mellon University, Forensic Economics, the University of Rochester, and the 2007 Journal of Accounting and Economics Conference. Erickson appreciates financial support from the Graduate School of Business at the University of Chicago. Heitzman appreciates financial support from the Simon Graduate School of Business at the University of Rochester. Corresponding author. Tel.: ; fax: address: shane.heitzman@simon.rochester.edu (S. Heitzman). 1 For related research on option grant backdating, see Collins et al. (2005, 2006), Lie (2005), Narayanan and Seyhun (2008), Bebchuk et al. (2007a, b), Bernile and Jarrell (2008), and Heron and Lie (2006, 2007). For related research on option exercise backdating, see Cai (2007) and Cicero (2007) /$ - see front matter & 2008 Elsevier B.V. All rights reserved. doi: /j.jacceco

2 28 D. Dhaliwal et al. / Journal of Accounting and Economics 47 (2009) For those exercises in which the insider pays the exercise price in cash and plans to hold all of the acquired stock after the exercise (referred to in this paper as exercise-and-hold transactions), the executive can reduce his/her income tax liability by exercising the option when stock price is low (a strategy which could include both ex ante timing and ex post backdating). 2 For exercises followed by an immediate disposition of all or part of the acquired stock (referred to as exerciseand-sell transactions), the insider s wealth is maximized by exercising when stock price is high. Prior research typically assumes that option exercises are followed by a sale of shares (e.g. Carpenter and Remmers, 2001). This assumption leads to the prediction that option exercises occur at relative stock price highs. But whether insiders desire to exercise on dates with high or low prices depends on whether or not the insider intends to dispose of some of the acquired shares immediately, and empirical specifications can be improved by partitioning the data accordingly. Our data indicate that approximately one-third of CEO exercises are not followed by a disposition of shares within 30 days of the exercise, and we designate these as exercise-and-hold transactions. 3 We focus our investigation of opportunistic timing on exercise-and-hold transactions. Assuming that insiders who exercise and hold shares intend to hold the acquired stock long enough to qualify for favorable long-term capital gains treatment, their incentive to exercise when the stock price is low is driven solely by personal tax considerations; the strike price of the option is unaffected by the choice of exercise date. Further, exercise-and-hold transactions are usually accomplished in-house in a transaction between the company and the insider, whereas exercise-and-sell transactions are often done in coordination with a broker who may be unable or unwilling to participate in backdating the transaction. 4 Throughout the study, we partition the observations based on the exercise date relative to 29 August 2002 the effective date of the Sarbanes-Oxley Act (hereafter SOX). Prior to this date, insiders were required to report stock option exercises to the SEC by the 10th day following the calendar month in which the options were exercised. This reporting lag facilitated the potential to backdate an exercise to the most advantageous stock price during a given month. SOX shortened the deadline for reporting option exercises (as well as most other transactions, including option grants) to the second business day following the transaction date. Following prior studies on grant backdating (e.g. Heron and Lie, 2007), we exploit this regulatory regime shift to test whether stock option exercises were backdated. We first investigate stock returns surrounding stock option exercises by CEOs and non-ceo insiders. Before SOX, exercise-and-hold transactions are generally preceded by significantly negative returns leading up to and including the exercise date and are followed by significantly positive returns. While this is consistent with a tax-based incentive to time the exercise when the stock price is low, backdating is not the only possible explanation for the observed returns. That is, the returns patterns we observe could result from backdating as well as ex ante timing using private information. Moreover, positive returns after exercise could possibly reflect the market s favorable response to information contained in the insider s decision to exercise and hold the shares, and not reflect a tax-based incentive to time the exercise. In other words, signaling could induce patterns in the data that suggest opportunistic timing even where it does not exist. We perform several analyses. First, to provide evidence on whether signaling explains the observed stock return patterns before SOX, we analyze whether abnormal returns following the exercise date vary as a function of the time between exercise and disclosure of the exercise. If signaling explains price increases following an exercise-and-hold transaction, the increase should be stronger when insiders disclose their transaction to the public soon after exercise. However, we find no evidence that post-exercise returns are increasing in the timeliness of the option exercise disclosure. These analyses suggest that investors responses to news about exercises are not driving patterns in the data. Next, we compare stock returns surrounding exercise-and-hold transactions before and after SOX. Cumulative total returns over the 21 days ending on the exercise date average 1.58% before SOX and 1.52% after SOX for CEOs (p-value of difference o0.001). Over the 20 days following the exercise, returns average 3.81% before SOX and 2.24% after SOX (p-value of difference o0.001). Because SOX did not affect the tax-based incentive to opportunistically time such exercises ex ante, we interpret the returns-based evidence to be consistent with the existence of backdated stock option exercises prior to SOX. We also analyze the characteristics of stock option exercises most likely to be opportunistically timed, which we term suspect exercises. An insider minimizes the tax liability on an exercise-and-hold transaction by exercising when the stock price is as low as possible. Accordingly, we label an exercise suspect if it occurs on the day with the lowest closing stock price during the month. Before SOX, 13.55% of exercise-and-hold transactions by CEOs are classified as suspect, which is nearly three times the frequency expected if exercises occur randomly throughout the month (po0.001). After SOX, the frequency of suspect exercise-and-hold transactions falls to 7.20%. Non-CEO insiders experience a qualitatively similar reduction in the frequency of suspect exercise-and-hold transactions after SOX. The evidence that the frequency of suspect exercises drops significantly after SOX further supports the interpretation that some exercises were backdated before SOX, 2 The tax liability we model is the present value of taxes saved from a reduction in the exercise date stock price. Reducing the stock price on the exercise date lowers income taxed immediately at ordinary income rates, and increases future income to be taxed at lower capital gains tax rates as long as the shares are held at least 1 year. It does not change the strike price of the option. The present value of taxes saved from opportunistic timing depends on the relevant personal tax rates as well as how long the insider intends to hold the shares. 3 In sensitivity analyses, we employ alternative horizons to identify exercise-and-hold transactions. Results of these analyses (not tabulated) are qualitatively similar. 4 However, exercise-and-sell transactions which do not involve a sale of shares (such as a stock swap with the firm) could also be backdated. We find some evidence consistent with this conjecture later in the paper, but not to the magnitude observed in exercise-and-hold transactions.

3 D. Dhaliwal et al. / Journal of Accounting and Economics 47 (2009) and is not consistent with either ex ante timing or signaling as the main explanation for stock price patterns around option exercise dates. Taken together, the evidence and the results of various supplemental analyses discussed later in the manuscript imply that ex ante exercise timing and signaling are not the dominant explanations for the fortuitous timing observed in exercise-and-hold transactions. We estimate the present value of tax savings realized by insiders with suspect exercise-and-hold transactions. For 776 suspect exercises between January 1996 and August 2002, CEOs saved an average of $96 thousand per option exercise and over $74 million in total by exercising at the lowest closing stock price instead of the average price during the exercise month. The tax savings estimates are equivalent to about 5.18% of the taxable income on the exercise, and about 3.05% of the total value of the option position at exercise both metrics indicate relatively modest tax benefits from timing the option exercise date. 5 An average tax savings of $96 thousand ($7 thousand at the median) realized by CEOs in suspect option exercise transactions is surprisingly modest in comparison to the potential costs of engaging in backdating. 6 For example, the tax penalties for such illegal actions could easily exceed $100 thousand, and reputation costs to the insider could be substantial. 7 At the same time that suspect option exercises reduce the insider s taxable income and tax liability, the option granting corporation s income tax deductions are also reduced. That is, the issuing corporation receives fewer tax benefits from the option exercise than they would have otherwise received. 8 We estimate that foregone corporate tax benefits associated with the 776 suspect exercises by CEOs exceed $74 million. Thus, while suspect option exercises reduce the tax liability of the executive, they result in a cash cost to the corporation in the form of a higher tax liability. Further, the firm is liable for unpaid withholding taxes resulting from a backdated exercise. Consistent with the argument that personal tax incentives influence the probability of opportunistic exercise timing, we find that the likelihood of a suspect exercise-and-hold transaction prior to SOX is increasing in the individual s potential tax savings from exercising at the lowest stock price of the month. For the CEO, increasing the potential tax savings from the first to the third quartile increases the probability of a suspect exercise by 4.64%. We find some evidence that the tax status of the firm influences the likelihood of a suspect exercise before SOX. Additionally, we find that suspect exercises are more likely in smaller firms. To the extent firm size proxies for the strength of internal controls (Ashbaugh-Skaife et al., 2007; Doyle et al., 2007), this is consistent with the interpretation that an insider s ability to backdate an exercise is greater in firms with weaker internal control environments. 9 To link our results to the stream of literature on option grant backdating, we analyze the correlation between grant backdating and exercise timing. First, we focus on firms alleged to have backdated option grants to insiders as tracked by the Wall Street Journal s Options Scorecard. We find stronger evidence of exercise backdating among this group. Before SOX, over 21.5% of exercise-and-hold transactions by the CEOs of these firms occur on the day with the lowest stock price of the month compared to 13.5% of transactions for the full sample. Second, we use data on option grants to calculate insider and firm level estimates of the odds that option grants were backdated. For sample firms which were not named by the Wall Street Journal, we find additional evidence that the likelihood of a suspect exercise is increasing in the likelihood of grant backdating. In additional tests, we investigate the timing of exercise-and-sell transactions. We disaggregate these exercises into two groups: those in which shares were disposed of only to the company (stock swap), and those in which at least some of the stock was sold (stock sale). We expect that the latter group of exercise-and-sell transactions involve a broker and are not likely to be backdated. Thus, to the extent some of these exercises were backdated, it should be more evident among those exercises followed by stock swaps, and the results are consistent with that notion. The remainder of the study is organized as follows. In Section 2, we review the literature on option grant backdating and the role of private information in option exercises, discuss the institutional background on the mechanics of stock option exercises, and provide a simple model of the tax benefits resulting from backdating option exercise dates. In Section 3, we describe the sample and empirical methodology and present results. Section 4 contains the conclusion. 5 It is important to note that some of the $74 million estimate for CEOs is fairly ascribed to exercises that occurred legitimately on the day with the monthly low stock price. Non-CEO insiders also obtain similar relative tax savings at the exercise date, though of a smaller average magnitude than the CEO. 6 One possible reason for the small magnitude is that we underestimate the tax savings by backdating insiders since a large portion of the suspect exercises are likely to have occurred by chance. If the largest tax savings were obtained through backdating, then our estimate of tax savings from backdating based on all suspect exercises is biased downward. The average tax savings for the largest 50% of suspect exercises (by tax savings) is $189, See Internal Revenue Code Sections 7201, 7206 and For example, in United States v. Crum and United States v. Micke, defendants were accused and convicted of backdating documents to reduce tax liabilities and violating Code Section If a corporation is accused of violating these provisions of the tax law, the penalties can exceed $500 thousand. Moreover, violation of some of these Code Sections constitutes a felony. The economic ramifications for shareholders as a result of illegal actions on the part of a firm s insiders can be quite large. 8 Corporations receive a tax deduction at the date of exercise in an amount equal to the difference between the stock price on the date of exercise and the strike price of the option. If the stock price on the date of exercise is reduced, the corporate tax deduction is reduced correspondingly. 9 However, as discussed later in the manuscript, we find no systematic evidence that the likelihood of a suspect exercise is associated with board independence, CEO ownership, dual CEO-Chair title, and CEO tenure.

4 30 D. Dhaliwal et al. / Journal of Accounting and Economics 47 (2009) Background and literature review We begin this section with a short discussion of option grant backdating. We follow with a brief review of prior literature on the role of private information in stock option exercises and highlight the empirical implications of the assumption that all exercises are followed by an immediate sale of the acquired shares. We then discuss the mechanics of an option exercise. Finally, we discuss the tax consequences of a stock option exercise to the option holder and the firm and derive empirical predictions The timing of stock option grants Yermack (1997) and Aboody and Kasznik (2000) provide evidence that executives consistently receive stock option grants on days with relatively low prices. Yermack (1997) argues that CEOs time the option grant date ex ante to precede the release of positive information or follow the release of negative information. Aboody and Kasznik (2000) argue that this pattern is unlikely to hold for scheduled awards in which there is little ex ante discretion. In those cases, CEOs appear to time disclosures to minimize the stock price on the grant date. The implication of both studies is that option grant dates are likely to occur at relative stock price lows. An interesting feature of the time period studied by Yermack (1997) and Aboody and Kasznik (2000) was that option grants were not typically reported to the public until after the fiscal year-end on SEC Form 5. This created a substantial lag between the actual grant date and the date when insiders actually had to disclose it. Lie (2005) proposes that the systematic return pattern surrounding option awards documented in these earlier studies is so pronounced that it could be caused by executives exploiting the lenient reporting requirements to retroactively time these awards. In support of this proposition, he finds that CEO option grants appear to systematically time future market movements. Because insiders should not have private information about future market movements, evidence of such timing is consistent with the backdating hypothesis. 10 As a result of SOX, and effective 29 August 2002, the SEC required most option awards to be reported on Form 4 within 2 business days of the grant. This accelerated filing deadline eliminated many opportunities to retroactively time option grants, and presents a natural setting in which to test the backdating hypothesis. Collins et al. (2005) and Heron and Lie (2007) find that the systematic share price movements surrounding option grant dates are largely mitigated after SOX, and interpret the evidence to be consistent with the existence of grant backdating before SOX. 11 Over 130 firms have come under public scrutiny for alleged stock option grant backdating. To date, over four dozen executives and directors have been terminated or have resigned in connection with the issue, and many are facing civil and/or criminal charges. Heron and Lie (2006) attempt to quantify the prevalence of backdating during this time, and estimate that 23% of unscheduled option awards were retroactively dated. In a related study, Bebchuk et al. (2007a) estimate that 12% of firms provided backdated grants to their CEO and that the backdating increased the CEO s total compensation by more than 10%, on average. Bernile and Jarrell (2008) document significant market value declines for firms with alleged grant backdating practices and argue that the loss in value is largely explained by increases in perceived agency costs. Some have argued that stock option incentives are inefficient because executives have significant control over the composition, amounts, and timing of their pay (e.g. Bebchuk et al., 2002). To date, the evidence on backdating appears to support this position. That many firms have been accused of backdating option grants leads naturally to the question of whether other aspects of stock option plans have been manipulated to benefit insiders. 12 Recent coverage in the financial press (Dash, 2006) suggests that they have and Senate testimony by SEC authorities indicates that the SEC is actively investigating whether option exercise dates were also manipulated (US Senate Committee on Finance, 2006). Unlike option grants, backdating the exercise does not lower the strike price of the option. The only identifiable incentive to backdate an exercise to a date with a low stock price is to minimize the option holder s income tax liability. This tax incentive arises only if the insider plans to hold the acquired shares long enough to qualify for long-term capital gains treatment at least 12 months under current law Stock option exercises and private information Prior studies investigate the information content of insider stock option exercises for future stock returns. Consistent with the evidence in Ofek and Yermack (2000), these studies often assume that insiders sell the acquired shares immediately. If this is the case, insiders should time option exercises when stock prices are relatively high. Huddart and Lang (1996, 2003) and Carpenter and Remmers (2001) find that insider exercises are preceded by positive stock returns in the weeks prior to the exercise. If insiders have private information about future returns, then exercise-and-sell transactions should be followed by negative returns. Carpenter and Remmers (2001) find that option exercises by 10 In an earlier version of the paper, we performed analyses similar to those of Lie (2005). The results of those analyses support the conclusion that some option exercises were backdated before SOX. 11 Heron and Lie (2007) and Narayanan and Seyhun (2008) also find evidence that grants are backdated after SOX for transactions filed after the 2-day reporting deadline. 12 Some (e.g., Bar-Gill and Bebchuk, 2003) have alleged that the increased use of stock options in the mid and late 1990s created an incentive to commit accounting fraud. However, Erickson et al. (2006) do not find such an association.

5 D. Dhaliwal et al. / Journal of Accounting and Economics 47 (2009) % 8% 6% 4% 2% Cumulative total returns surrounding stock option exercises by CEOs - Before SOX 0% -20-2% Exercise-and-hold Exercise-and-sell Fig. 1. This figure plots the average cumulative total returns surrounding option exercises by CEOs before SOX. There are 14,839 total exercises, with 5,726 classified as exercise-and-hold transactions and 9,113 classified as exercise-and-sell transactions. executives preceded positive abnormal returns prior to May 1991, when these executives were required to hold the acquired shares for at least 6 months. After this date, however, they find no evidence that exercises are associated with future stock returns. Conversely, Huddart and Lang (2003) do find that among the seven firms in their sample, insider exercises are followed by negative stock returns for a broad sample of employees at various ranks. Because insiders do not always dispose of the acquired shares immediately, conflicting evidence on the association between insider exercises and future returns is not surprising when the research design does not control for the disposition of the acquired shares. 13 For example, insiders who plan to immediately dispose of all or part of the acquired shares clearly want to do so at the highest price. But what about insiders who plan to hold the shares? In the absence of taxes, the timing of these exercises should not depend on short-run price movements. But with taxes, the option holder faces an immediate ordinary income tax on the spread between the stock price on the exercise date and the exercise price of the option. Subsequent appreciation in the underlying stock will be taxed at generally favorable long-term capital gains rates when the shares are sold. Given the decision to exercise, and assuming the insider intends to hold the stock at least 1 year, the insider can minimize taxes by exercising when the firm s stock price is low. Fig. 1 plots the cumulative total returns surrounding both exercise-and-hold and exercise-and-sell transactions before SOX and indicates that the distinction between exercises that involve share dispositions and those that do not is important The mechanics of stock option exercises Unlike the decision to grant the option, the decision to exercise the option is ultimately up to the insider. 14 An exercise is initiated when the option holder gives notice to the grant administrator of the decision to exercise, the specific options to be exercised and the number of each. The exercise date is typically considered to be the latest date the notice is delivered and the option holder has paid the corporation, in existing shares or cash, the exercise price of the option plus withholding taxes. The option holder reports the spread between the market value of shares acquired on the exercise date and the exercise price of the options as ordinary income in the year of exercise. The option agreement specifies the allowable forms of financing the option exercise. The cash method is typically the default form whereby the option holder delivers the exercise notice along with the cash required to pay the strike price plus withholding taxes (in total, the exercise cost). Cash can be personal funds, and before SOX, could be obtained through a company loan. The acquired shares are then delivered to the option holder, and the option holder takes a tax basis in the shares equal to their fair market value at the exercise date. However, because the cash required at the exercise is likely to be significant, common methods of financing the exercise cost include exchanging existing holdings with the company (stock swap) as well as selling shares to an outside party (stock sale). All forms of financing must be approved in the option agreement See Aboody et al. (2008) for an exception. 14 This assumes that vesting schedules are met and no other insider trading restrictions apply, such as blackout dates (see Bettis et al., 2000). Blackout dates are designed to prevent violation of insider trading laws which apply to the purchase and sale of securities. Exercise-and-sell transactions are likely affected since they effectively involve a sale at market prices. However, exercise-and-hold transactions are unlikely to be affected since the transaction is between the insider and the issuer at a predetermined price (the strike price), and many firms explicitly exclude exercise-and-hold-type transactions from their insider trading policies. We are not aware of any study which documents the characteristics and determinants of exercise restrictions, however, to the extent blackout dates do apply in some transactions, our results may be affected. 15 In a stock swap, the insider pays the exercise price by delivering or attesting shares with a market value equal to the exercise cost. The value per share is usually determined on the exercise date. There is no additional tax due when appreciated shares are used to pay the exercise cost. To preserve the gain for tax purposes, a portion of the new shares acquired via the option exercise will take the basis and holding period of the swapped shares.

6 32 D. Dhaliwal et al. / Journal of Accounting and Economics 47 (2009) Tax consequences of stock option exercises To the option holder Personal tax savings drive the incentive to time an exercise-and-hold transaction on a day when stock price is low. But an important question is why the insider would exercise and hold the shares at all. McDonald (2003) and Scholes et al. (2005) discuss the optimality of exercising early to start the clock on capital gains treatment. In general, a strategy of borrowing funds to pay the exercise cost (the sum of the exercise price and taxes due) and holding the acquired shares is dominated by a strategy of using the borrowed funds to buy additional shares and hold the options. Each insider makes the decision to exercise and hold shares based on their information set, tax attributes, and risk aversion, subject to a set of constraints including upcoming option expirations and minimum ownership and holding period requirements. Without a full understanding of the choices and preferences of the insider, we are unable to address whether these exercises are optimal. Rather, we take the decision to exercise options and hold or dispose of the acquired shares as given, and instead investigate the determinants and consequences of the choice of exercise date, conditional on the exercise decision. To the extent that unobserved determinants of the exercise-and-hold decision also determine stock price patterns surrounding the exercise as well as the potential tax savings, our inferences may be affected. However, our empirical tests described later generally rule out such alternative explanations. We illustrate the tax effects and the potential magnitude of insider tax savings arising from exercise timing using a simple model. Let P GRT be the exercise price (strike price) of the options which is set at the grant date, P EX be the stock price at the exercise date, and P SALE be the price at which the insider expects to sell the acquired stock in the future. Further, let t ord represent the tax rate faced at the exercise date and t cg represent the expected tax rate faced at the future sale of the acquired stock. All options are assumed to be non-qualified stock options (NQSO). 16 The present value of the total tax incurred as a function of the insider s option exercise decision can be expressed as Tax ¼ N ðp EX P GRT Þt ord þ ðp SALE P EX Þt cg ð1 þ rþ n (1) N is the number of options exercised, r is the appropriate after-tax discount rate of the insider, and n is number of years the insider plans to hold the acquired shares. For exercise-and-hold transactions, the tax savings that can be achieved by exercising on a day with a low stock price can be shown by taking the derivative of Tax with respect to P EX,or qtax ¼ N t qp ord t cg EX ð1 þ rþ n 40 (1a) The term [t ord t cg /(1+r) n ] is the present value of converting $1 of ordinary income into capital gains to be taxed at a later date. Thus, Eq. (1a) captures the expected tax savings from reducing P EX by a dollar. Note that P GRT is not relevant for calculating the tax savings at exercise, and hence the incentive to time the exercise. As long as there is taxable income on the option exercise, the insider gets the same tax savings per dollar movement in P EX independent of the exercise price. To formalize the relation between tax savings and the exercise price implied by Eq. (1a), let P* be the price at which the insider would have exercised on an ex ante basis (i.e. no backdating), and P MLOW be the lowest price at which the insider could have exercised during the month. Thus, P* P MLOW is the potential reduction in taxable income per share achieved by exercising at P MLOW. Accordingly, we define the potential total tax savings for a given exercise-and-hold transaction as Potential tax savings ¼ N t ord t cg ð1 þ rþ n ðp P MLOW Þ (2) The potential tax savings is the product of three components: (1) the number of options exercised, (2) the present value of converting $1 of ordinary income into capital gains, and (3) the magnitude of the difference between the price at which the insider would have otherwise exercised and the lowest price of the month. Potential tax savings captures the direct taxbased incentive to time a given exercise on a day when stock price is low. Moreover, the only apparent meaningful benefit of backdating these exercises is to reduce the personal tax liability. 17 To empirically estimate the potential tax savings defined in (2), we assume that the exercise was equally likely to occur on each day of the month and use the average closing price during the month to proxy for P*. We assume that t ord is equal to the top marginal tax rate on ordinary income, t cg is equal to the current top marginal tax rate on long-term capital gains, 16 Stock options are non-qualified unless they meet the criteria for an incentive stock option (ISO) established in IRC Section 422. The stock acquired via an ISO must be held for at least 2 years after the ISO grant date and at least 12 months after the exercise date. An ISO cannot be granted in-the-money, and the stock value covered by the ISO cannot exceed $100,000 per employee per year. For a valid ISO, the employee is taxed at the long-term capital gains rate at the time the acquired stock is sold on the spread between the stock price on the sale date and the exercise price of the ISO. However, while the spread between market price and exercise price at ISO exercise is not included in the regular tax calculation, it is included for purposes of the alternative minimum tax (AMT) calculation under IRC Section 56(b)(3). Finally, the corporation gets no tax deduction at ISO exercise. See Scholes et al. (2005) for further detail and analyses of the differences between an ISO and an NQSO. 17 One plausible alternative reason to backdate the exercise is to salvage expired options. But such cases are likely quite rare. Furthermore, this alternative only suggests that the exercise would be backdated to a point before the expiration date. It does not, by itself, create an incentive to backdate the exercise to a day with a low price. No other plausible explanation for backdating these exercises was offered by conference participants.

7 D. Dhaliwal et al. / Journal of Accounting and Economics 47 (2009) the holding period is 5 years, and the insider s appropriate after-tax discount rate is 10%. With current rates of 35% for t ord and 15% for t cg, for example, the present value of tax savings per share approach $0.257 per dollar decrease in P EX [ /(1.1) 5 ]. Obviously, the total potential tax savings we estimate is sensitive to the assumed discount rate, the choice of holding period, as well as the choice of P* To the firm For non-qualified stock options, the spread between the stock price on the exercise date and the exercise price of the option is treated as deductible compensation expense to the option granting corporation. Therefore, for each dollar reduction in current stock price on the exercise date, the corporation loses a potential deduction by that amount. In other words, tax savings to the option holder are offset by a tax cost to the corporation. The total corporate tax benefit arising from an option exercise is Tax benefit ¼ NðP EX P GRT Þt c (3) where t c is the marginal corporate tax rate on an additional dollar of income or deductions. 18 The effect of an arbitrary reduction in the exercise date stock price on the corporate tax benefit is qðtax benefitþ ¼ Nt c 40 qp EX For firms facing the top marginal tax rate, the increase in federal tax liability per option is approximately $0.35 per dollar decrease in P EX, and backdating results in a wealth transfer from shareholders to insiders Summary and predictions We anticipate that stock price patterns surrounding option exercises are substantially different for insiders that divest some or all of the acquired shares relative to those who hold them. If insiders time the option exercise to maximize the after-tax value of the award, then exercise-and-hold transactions should occur on days with low stock prices, whereas exercise-and-sell transactions should occur on days with high stock prices. If backdating, rather than ex ante timing or signaling, explains the bulk of well-timed exercises, then any systematic evidence of opportunistic timing of exercise dates should be mitigated as a result of the 2-day reporting rule mandated by SOX. For exercise-and-hold transactions, we expect that the incentive to exercise an option on a date with a low stock price is increasing in the magnitude of the potential taxes saved by exercising on that date, where potential tax savings is a function of the size of the exercise, the relative tax rates on ordinary income and capital gains, and the magnitude of the potential discount achieved by exercising on the day with the lowest closing stock price of the month. We also make predictions with respect to firm characteristics. First, if insiders take shareholder wealth into account when deciding whether to backdate the option exercise to a date with a low stock price, the likelihood of a suspect exercise should be greater in firms with smaller potential tax benefits from the deduction (i.e. low tax rate firms). Second, the exercise decision is made by the option holder, but the actual exercise is coordinated through the option plan administrator. Therefore, the ability to backdate may be related to the strength of the internal control environment of the firm. Ashbaugh- Skaife et al. (2007) and Doyle et al. (2007) find that smaller firms, as measured by market capitalization, have weaker internal control environments. To the extent firm size captures the strength of internal controls in our setting, we expect that the likelihood of a suspect exercise is more likely in small firms. 3. Data, sample and methodology We obtain our sample of insider stock option exercises from the Thomson Financial Insider Filing database. We identify stock acquisitions arising from option exercises as reported on SEC Form 4 with transaction code M ( Exercise of in-themoney derivative security acquired pursuant to a Rule 16b-3 plan ) between 1996 and Prior to SOX, insider transactions for a given calendar month, including shares acquired via option exercises, were required to be reported on Form 4 by the 10th day following the end of the month in which the transaction occurred. We exclude transactions in which the insider exercised options on multiple days during the month to improve our ability to correctly classify exerciseand-hold transactions based on subsequent disposition of shares. 19 After merging with CRSP and Compustat, we obtain 138,917 insider-exercise observations. To deal with concerns of including firms with relatively illiquid and thinly traded 18 We do not estimate the GAAP earnings consequences of backdated option exercises. During the period in question, nearly all firms recognized stock option tax savings directly in shareholders equity (see Hanlon and Shevlin (2002) for a discussion of the financial accounting and tax treatment of employee stock options). Moreover, this estimate ignores cost sharing and transfer pricing effects. 19 In unreported tests, we ran the analysis for the CEO including all exercises, and used the same-day sale window to classify firms as exercise-andhold. We find that the frequency of suspect exercises as a percentage of all exercise-and-hold transactions before SOX is lower (12.56%), but so is the percentage of suspect exercises after SOX (6.93%). Because we do not have an accurate estimate of the frequency of legitimate exercises composing the suspect exercise group, we are unable to ascertain whether the incidence of actual backdating is overstated. Importantly, the regression results are robust to including all exercises.

8 34 D. Dhaliwal et al. / Journal of Accounting and Economics 47 (2009) Table 1 Summary statistics on stock option exercises by insiders N Number of unique insiders Number of shares acquired on exercise (K) Fraction selling shares within 30 days Market value of shares acquired ($M) Expected taxable income ($M) All exercises CEO 23,356 7, (20.00) (0.56) 1.65 (0.32) Non-CEO 107,715 40, (8.00) (0.22) 0.55 (0.13) By size of firm (CEO only) MVEo$500 M 7,876 3, (14.40) (0.22) 0.42 (0.13) $500 MoMVEo$2,500 M 7,625 2, (20.00) (0.55) 1.01 (0.32) MVEX$2,500 M 7,855 2, (35.80) (1.63) 3.50 (0.96) All exercises by CEO Before SOX 14,839 6, (19.50) (0.51) 1.74 (0.30) After SOX 8,517 3, (25.00) (0.62) 1.48 (0.35) Exercise-and-hold by CEO Before SOX 5,726 3, (12.00) (0.25) 1.17 (0.14) After SOX 2,291 1, (13.91) (0.28) 0.71 (0.15) Exercise-and-sell by CEO Before SOX 9,113 4, (24.00) (0.77) 2.09 (0.44) After SOX 6,226 2, (30.00) (0.86) 1.75 (0.47) This table reports sample means with medians in parentheses. The number of shares acquired on exercise is the sum of all exercises on a given exercise day, reported in thousands. The fraction selling acquired shares is the percentage of insider exercises followed by a disposition of any quantity of shares within 30 days of exercise. The market value of shares acquired is equal to the number of options exercised times the average price during the month. Expected taxable income is equal to market value of shares acquired less the product of the exercise price (when available) and the number of options exercised. equity in the analysis, we eliminate 5,668 observations in which the firm s market capitalization is less than $10 million or the average stock price during the exercise month is less than $3. Finally, we eliminate 2,178 observations in which the firm was engaged in a stock split during the month, because that event confounds our estimate of the potential tax savings from exercising at the lowest closing price of the month. 20 After these filters, we obtain a total of 23,356 exercises for CEOs, 18,720 for non-ceo executives, 58,160 for other officers, and 30,835 for non-officer directors. 21 We aggregate the latter three categories into a single non-ceo group hereafter. Based on this set of insider transactions, we then classify option exercises as: (1) those in which some or all of the acquired shares are immediately disposed of by the insider, and (2) those in which the insider holds all of the acquired shares. Exercise-related sales are likely to occur on the same day the option is exercised (or very soon thereafter). We classify an exercise followed by any disposition coded S, F, or D within 30 days of the exercise as an exercise-and-sell transaction. 22 Otherwise, we classify it as an exercise-and-hold transaction. In supplemental analysis (untabulated), we investigate the effect of using alternative horizons to classify exercises and the results are qualitatively the same as reported herein. Based on the entire sample period as reported in Table 1 (column 3), 66% of all exercises by CEOs are followed by a disposition of at least some stock within 30 days of exercise. This figure is 69% for non-ceos. Interestingly, 20 We eliminate observations where the firm had a stock split during the exercise month because we rely on the average of actual daily prices in order to calculate measures of potential tax savings in the analysis. 21 The insider classifications are made using the role codes reported by insiders and provided on the insider trading database. Following Bernile and Jarrell (2008) and Heron and Lie (2006), the CEO classification includes all insiders where the primary role code is CEO, President, or Chairman of the Board. Non-CEO executives include insiders where the primary and secondary role codes indicate the person is a CFO, COO, or both an officer and director. The other officer category includes all other officers who are not also directors. Non-officer directors include all other insiders where the primary role code is director. 22 Code S is described as Open market or private sale of a non-derivative or derivate security ; code F is described as Payment of option exercise price or tax liability by delivering or withholding securities incident to exercise of a derivative security issued in accordance with Rule 16b-3 ; code D is described as Disposition to the issuer of issuer equity securities pursuant to Rule 16b-3(e). These codes are defined by the SEC and reported by the insider.

9 D. Dhaliwal et al. / Journal of Accounting and Economics 47 (2009) % Cumulative total returns surrounding exercise-and-hold transactions by CEOs 3% 2% 1% 0% -20-1% % Before SOX After SOX Fig. 2. This figure plots the average cumulative total returns surrounding exercise-and-hold transactions before and after SOX. The CEO classification includes all insiders where the primary role code is CEO, President, or Chairman of the Board. Table 2 Stock returns surrounding insider exercise decisions Window CEO Non-CEO Before SOX After SOX Before After Before SOX After SOX Before After TR( 20,0) 1.58% 1.52% 3.10% 0.07% 1.94% 1.87% ( 5.79) (5.58) ( 6.69) (0.47) (17.34) ( 9.03) TR( 10,-1) ( 7.99) (1.35) ( 5.49) ( 4.48) (8.06) ( 7.36) TR(0) ( 3.64) (2.64) ( 3.74) ( 2.21) (3.96) ( 3.71) TR(1,10) (15.79) (7.41) (4.61) (20.84) (17.27) (3.56) TR(1,20) (15.70) (9.60) (3.80) (20.23) (21.17) (0.87) % Exercised at lowest price (19.57) (4.65) (8.01) (26.70) (8.26) (11.32) N 5,726 2,291 21,080 11,846 t-statistics in parenthesis. This table reports average cumulative total (TR) returns surrounding exercise-and-hold transactions by insiders before and after SOX. Exercise-and-hold transactions are those exercises which are not followed by share divestitures within 30 days. The difference between the means before and after SOX and the associated two-tailed t-statistic for the difference is also reported. Day 0 is the exercise date. % exercised at lowest price is the fraction of exercises in the respective group which fall on the day with the lowest closing price of the month. In the case of % exercised at lowest price, t-statistic tests whether frequency is significantly different from %. CEOs of small firms are less likely to exercise-and-sell (49% of the time) than are CEOs at medium and large firms (69% and 79%, respectively). The frequency of exercise-and-sell transactions increased from 61% before SOX to 73% after SOX Stock returns surrounding option exercises Unlike prior studies which focus on long window returns following exercise (e.g. Huddart and Lang, 2003), we are interested in the insider s ability to time the exercise within a concentrated window, similar to the work on stock option backdating. We focus on total stock returns since that is a more appropriate measure of the total gains from timing. In Fig. 2, we plot average cumulative total returns surrounding exercise-and-hold transactions before SOX for CEOs. In Table 2, we provide statistical analysis of the returns surrounding the exercise date. For CEO exercises before SOX, average total 23 One possible reason for this increase was the prohibition of company loans to insiders under Section 402 of SOX. Company loans allowed the insider to pay the exercise cost in cash and hold the acquired shares.

10 36 D. Dhaliwal et al. / Journal of Accounting and Economics 47 (2009) Table 3 Reporting lag and abnormal returns following exercise-and-hold transactions before SOX CEO Non-CEO Lago24 days LagX24 days Lago24 days LagX24 days AR(1, 5) 0.92% 1.08% 0.37% 0.55% (6.32) (6.77) (4.82) (6.50) AR(1, 10) (5.08) (4.99) (1.39) (1.97) AR(1, 20) (3.53) (2.48) ( 2.88) ( 1.38) Average lag in days Number of exercises 2,838 2,878 10,236 10,808 % Exercised at lowest price t-statistics in parenthesis. None of the differences between post-exercise abnormal returns for short and long reporting lags are statistically significant at the 10% level using a twotailed t-test. This table reports average cumulative abnormal returns (AR) following exercise-and-hold transactions by insiders as a function of the lag (in days) between the reported transaction date and the date the report was filed with the SEC. Day 0 is the exercise date. % exercised at lowest price is the percentage of observations where options were exercised on the day with the lowest closing price of the month. stock returns are 1.53% over the 10 days before the exercise, 0.23% on the date of exercise, and 2.56% over the 10 days after the exercise (po0.001 for all return windows). Qualitatively similar patterns are present for non-ceos. The stock return patterns surrounding exercise-and-hold transactions before SOX are consistent with insiders timing option exercises to minimize their tax liabilities. However, these findings can be attributed to alternative explanations which we now address. High post-exercise returns could be due to the market s reaction to the insider s decision to exercise and hold the shares. That is, the insider s action signals their private information to the market, and shareholders bid up the price of the stock in response. For the signaling explanation to hold, it must be the case that (a) investors respond positively to news of an exercise-and-hold transaction, and (b) investors learn about the transaction early enough for their response to be measured within our window. With respect to the first condition, Lakonishok and Lee (2001) find that insider stock purchases provide information about future returns. So finding a positive market response to exercise-and-hold transactions would be consistent with their results if shareholders view the acquisition of shares through an option exercise similarly to the acquisition of shares through an outright purchase. With respect to the second condition, reporting requirements before SOX resulted in significant lags between the exercise date and the actual filing date, which is when outside investors usually become aware of the transaction. Therefore, if signaling is the primary explanation for the observed price increases after exercise, it should be the case that insiders who disclose their exercise-and-hold transactions to the market soon after exercise should see a larger positive price response following the exercise relative to insiders who disclose their exercises in a less timely fashion. To test this explanation, we partition exercise-and-hold transactions before SOX into two groups based on the lag between the exercise date and the date the Form 4 is filed with the SEC. We provide the results of this analysis in Table 3. The first (second) column reports the post-exercise abnormal returns for exercises with reporting lags less than (at least) 24 days (the median lag of all CEO exercise-and-hold transactions before SOX). We focus on abnormal returns in this case because a signaling effect is more appropriately tested using firm-specific returns. 24 Over the first 10 days following the exercise, average cumulative abnormal returns are 1.01% for exercises with shorter lags between exercise and disclosure, and 1.03% for those with longer lags (p-value of difference 40.10). For all abnormal returns accumulated over various postexercise windows, the difference between exercises with short and long reporting lags is insignificant. In untabulated tests, we regress abnormal returns over various windows following CEO exercises on the number of days between the reported exercise date and the SEC filing date. In general, we find no reliable association between the timeliness of the exercise disclosure and abnormal returns after the exercise. Thus, while the signaling explanation is plausible, these tests suggest that it does not explain abnormal returns following exercise-and-hold transactions. High post-exercise-and-hold returns could also be driven by insiders exploiting private information. Similar to Heron and Lie (2007) and Collins et al. (2005), we compare stock returns before and after SOX to test this alternative explanation. Specifically, if exercise decisions are driven entirely by ex ante utilization of private information, the return patterns surrounding exercise should be similar before and after SOX. However, if some insiders exploited lenient reporting 24 For each insider-exercise, we estimate a three-factor model based on the risk factors identified in Fama and French (1993) using daily returns for the year ending 50 days prior to the exercise date. The daily abnormal returns surrounding each insider exercise equal the actual returns less the predicted returns based on the estimated coefficients. Results are similar if total returns used.