Backdating Executive Stock Option Grants: Is It All Agency?

Transcription

1 Backdating Executive Stock Option Grants: Is It All Agency? Huasheng Gao Nanyang Business School Nanyang Technological University S3-B1A-06, 50 Nanyang Avenue, Singapore, Hamed Mahmudi Price College of Business University of Oklahoma 307 Brooks Street, Norman, OK This version: January 2016 Abstract: It is widely documented that managers tend to backdate their stock option grants so that a past date on which the stock price was particularly low is picked to be the grant date. Using a simple model of incentive contracting as a guide, we examine empirically whether some aspects of this practice may be an optimal response of firms to distortions in the institutional environment, in particular tax law and accounting rules. Some of our findings suggest, in fact, that firms may be attempting to efficiently lower the exercise price of the executive options in order to lower compensation cost for risk averse and poorly diversified executives, and in the presence of restrictive accounting and tax rules, backdating may be a mechanism to achieve this objective. Using data on corporate governance and executive compensation, we find the following evidence consistent with our theory of efficient contracting: (i) backdating is often associated with good corporate governance; (ii) backdating is often associated with better ex-post incentive structures; (iii) backdating is often associated with lower overall managerial compensation; (iv) after the Sarbanes-Oxley Act (SOX) of 2002 which greatly constraints firms from backdating, firms that were involved in more backdating in the pre-sox period have greater increase in CEO pay. These empirical results suggest that the managerial agency problems previously cited as an explanation for option backdating may be only part of the story underlying this complex and wide-spread practice. JEL Classification: G32; J33 Keywords: Backdating; Executive Compensation; Efficient Contracting We thank Alexander Dyck, Kai Li, Alan Kraus, Michael Lemmon (WFA discussant), Hernan Ortiz-Molina, Ralph Winter, Laurence Booth, Craig Doidge, Jan Mahrt-Smith, Adlai Fisher, Ambrus Kecskes, Feng Zhang, Zhongzhi Song, Yuan Gao, Konstantinos Zachariadis, Ivo Welch, and seminar participants at the University of British Columbia, University of Toronto, Shanghai Advanced Institute of Finance, the 2008 WFA meeting, the 2008 EFA meeting and the 2008 FMA meeting for insightful discussions. Gao acknowledges the financial support from Singapore Ministry of Education Academic Research Fund Tier 2 (Official Number: MOE2015-T ). All errors are ours.

2 The goal of backdating, it becomes clearer than ever, was to motivate employees at the lowest possible cost to shareholders. This was done by granting stock options that, at the date of issue, were "in the money". 1. Introduction Holman W. Jenkins, Jr., The Wall Street Journal, December 16, 2009, A25 There has been an active debate in the executive compensation literature between the efficient contracting view and the managerial power view. The former argues that the board of directors optimally designs compensation contracts to align managers interests with those of shareholders, while the latter argues that powerful managers can influence the terms of their own compensation and that the compensation practice is largely inefficient (see Weisbach (2007) and Larcker et al. (2011) for a review). The recent events regarding executive option backdating provide a setting to examine these viewpoints. The conventional wisdom is that backdating provides a powerful example of the weaknesses in corporate governance in the United States. For example, Arthur Levitt, former chairman of SEC, states Backdating is ripping off shareholders in an unconscionable way, (Forelle and Bandler (2006)). Bebchuk et al. (2010) also suggest that executive option backdating is the result of an agency problem whereby managers manipulate their compensation terms for their own benefits at the expense of shareholders. Moreover, there is evidence that option backdating is widespread, as Heron and Lie (2007, 2009), who brought backdating to public attention, find that this problem afflicts 29.2% of US corporations. In this paper, first, we examine the backdating evidence and, surprisingly, find that governance is at best a partial explanation. If backdating is a consequence of powerful managers manipulating their option grants at the expense of the shareholders, we should expect that weaker corporate governance predicts backdating. While it is true that worse governance predicts 1

3 backdating for some measures of governance, this is not what we find more generally. Specifically, for the larger sample better governance predicts more backdating, particularly when we use more general measures of governance and the full sample of backdaters. This relation is even stronger for riskier firms, for the period prior to the Sarbanes-Oxley Act (SOX), for longertenure CEOs, for larger option grants, and for the case when other top executives receive option grants along with the CEO. Since the governance explanation is at best a partial rationale for the patterns we observe in the data, we consider alternative explanations for backdating and conduct tests to examine their power in explaining the evidence. The dominant alternative to the managerial power view to explain compensation setting is efficient contracting. Here, we build on the work of Hall and Murphy and consider a context in which shareholders design option compensation to incentivize managers. Like Hall and Murphy (2000, 2002), the theory demonstrates that the optimal strike price for an option grant is usually lower than the grant-date stock price for an under-diversified and risk-averse CEO. The challenge with such a policy is the tax and accounting disadvantages. Under current tax rules, CEOs can benefit from a lower tax rate aligned with a capital gain and can defer the tax to the time of exercising the option grant. Generally, if all rules are complied with, the recipient of an option grant pays taxes on his entire option profit at the lower tax rate applicable to long-term capital gains. To qualify for the lower tax bracket, the option must be granted at or out of the money. Backdating sidesteps such issues, while of course raising legal and ethical quandaries, a topic we discuss at much greater length later. 1 This view helps to reconcile the broader evidence of a (weak) positive association between governance and backdating. Luckily, this explanation suggests further tests that help to 1 See Walker (2007) and Fleischer (2007) for a detailed discussion on the tax/accounting issues for executive option compensation. 2

4 differentiate between the efficient contracting and the managerial power views. A natural implication of the managerial power view is that managers would benefit financially from backdating, getting higher total compensation. The contracting view suggests the alternative: Backdating is associated with better alignment of shareholder-manager interest and reduces compensation level for executives. First, we examine the relation between backdating and managerial incentive portfolio. Our empirical results highlight a positive association between backdating and the pay-performance sensitivity in executive pay. This evidence is consistent with the view that option backdating could be part of efficient contracting which helps to strengthen managerial incentive. We then examine the relation between the level of the CEO pay and his backdating behavior. Supporting the efficient contracting view, we find that backdating is negatively related with the CEO s total compensation and his cash compensation. This evidence suggests that the board of directors simultaneously grants less cash payment to CEOs when backdating activities are ongoing. Of further importance is the finding that the total compensation cost is actually reduced in the presence of option backdating, which does not support the view that backdating makes shareholders overpay. To provide further evidence that backdating of stock option is conducted to lower the compensation cost, we examine the change in CEO pay around the implementation of SOX in Since SOX greatly constraints firms from backdating, we expect that firms increase their CEO pay after backdating is no longer viable. Consistent with our expectation, we find that firms involved in more backdating in the pre-sox period have greater increase in CEO pay compared to firms involved in less backdating in the pre-sox period. Although it is plausible to describe backdating as managers exerting influence over their own pay by retroactively timing their option grants, this explanation has certain drawbacks. It treats the compensation packages as being exogenously determined: Once the packages are set, 3

5 managers then backdate their option grants to increase the value of their option pay at the cost of shareholders. However, investors can adjust the whole compensation contract when anticipating option backdating. In other words, the managerial power explanation does not fully depict the equilibrium of a model in which option backdating results from managers manipulation of stock option grants. It is widely documented that firms stock returns are abnormally negative before executive option grants and abnormally positive afterward (see e.g., Yermack (1997), Aboody and Kasznik (2000), Chauvin and Shenoy (2001), and Lie (2005)). Backdating, that is, picking a past date on which the stock price traded particularly low to be the grant date, is believed to contribute to this stock price pattern (see, e.g., Heron and Lie (2007)). Our paper does not propose to deemphasize the self-interested behavior of some CEOs in manipulating the terms of their compensation contracts. However, efficient contracting motives, in which the managerial incentive is higher and total costs of compensation are lowered, can help us understand the documented empirical results, which are otherwise difficult to explain. We suggest that attributing backdating solely to agency problems is not necessarily the entire story and that backdating activity can also be justified as a form of efficient contracting from an economic perspective. It is worth noting that this paper is not intended to question the legality issues of backdating. But it explains backdating from the perspective of its efficiency to solve the executive compensation problem. Our definition of efficiency is from a purely economical perspective: An illegal behavior can be economically efficient if the ex-ante expected cost of getting caught is lower than the expected benefit. Our empirical findings imply that under existing tax/accounting rules at the time, an informed board could take advantage of the grey areas of the tax law and backdate the CEO s option grants while adjusting the other terms of his compensation. 4

6 Interestingly, practitioners and financial media have begun to change their opinions about backdating from being purely an agency problem. For example, in the Wall Street Journal (WSJ), Jenkins (2009, A25) states that most backdating cases amount to companies trying to behave rationally amid irrational accounting rules, rather than the media's standard trope of businessmen a-lyin' and a-stealin'. On the academic side, more evidence on the efficiency perspective of backdating is also found, including mitigating the investment timing problem (Dierker and Hemmer (2010)) and attracting valuable employee (Fang and Whidbee (2013)). 2 The remainder of this article proceeds as follows. In Section 2, we describe the data source and sample selection. Section 3 illustrates the empirical results on the relation between corporate governance and backdating activities. Section 4 presents the model and Section 5 tests the model s implications empirically. Finally, Section 6 concludes. 2. Variable Construction and Sample Selection We first seek to explore the power of governance to explain patterns in backdating in US firms. To do so we need to define what we (and others) consider backdating and to introduce the various measures of governance we explore Measures of Option Backdating Following Heron and Lie (2009), we infer the backdating activities from the stock price movement surrounding the option grant dates. In the absence of backdating or other types of grant date manipulation, the stock return distribution before grant dates should be largely similar to the distribution after the grant dates, as suggested by Heron and Lie (2009). For this reason, the difference between the stock returns for a certain number of days after grants and the returns of the same number of days before grants should be centered at zero. However, if the grants have 2 Overall, while our study is in many ways complementary to Dierker and Hemmer (2010) and Fang and Whidbee (2013), it captures a different source of efficiency as we provide an explanation based on CEO s risk aversion, under-diversification, and the existing tax law. 5

7 been backdated or otherwise manipulated, the difference will be positive. Following this logic, we estimate the extent of option backdating as follows: BBBBBBBB[t] = AA[1, t] AA[ t, 1] where AR[1,t] is the cumulative stock returns (AR) from day 1 to day t after the grant day (day 0), and AR[-t,-1] is AR from day -t to day -1 relative to the grant day. 3 An alternative approach to measure the degree of ex ante backdating is to focus on the most egregious examples of backdating. This is the approach taken by Bebchuk et al. (2010) and Collins et al. (2009) who define backdating as a dummy variable that equals one if the grant-date stock price is in the bottom decile of the firm s stock price distribution around option grants, and find some weak evidence that backdating is associated with poor corporate governance. Our preference is for the Heron and Lie (2009) measure as it is a continuous measure and captures the broad possible extent of backdating behavior. The continuous measurement allows us a more robust empirical investigation by including grants that have had strike prices lower than the grant-date stock price, but have not been issued with strike prices within the bottom decile of the stock price distributions around the grants. Furthermore, as we shall show in Section 4, in an efficient contracting setting with endogenous strike price, variables, such as executive risk preference and personal wealth, determine the optimal strike price and this strike price does not necessarily equal a firm s bottom-decile stock price. Therefore, using a continuous measure allows us to capture potential optimal strike prices in a wide range of time around the grant dates. 4 3 As a robustness check, we also delete observations with negative values of our backdating variables because those observations may be less indicative of backdating. Our results are insensitive to this alternative method. We also use cumulative abnormal returns using the market model based on CRSP value-weighted index returns instead of raw stock returns. The results are the same (untabulated). 4 Alternatively, we also follow Bebchuk et al. s (2010) method to define backdating as a dummy, which equals one if the grant-date price is within the lowest deciles of grant-month stock price, and zero otherwise. Our results are similar. 6

8 2.2. Measures of Corporate Governance To explore whether governance helps to explain patterns in backdating, we use several measures of corporate governance. One variable to capture weak firm governance is the degree of managerial entrenchment due to the number of anti-takeover provisions in a firm s charter and in the legal code of the state in which the firm is incorporated (the so-called G-index of Gompers et al. (2003), and the closely related E-index of Bebchuk et al. (2009)). Takeover protection provisions have a significant impact on firm decision-making since the market for corporate control is viewed as a strong external force for disciplining management. Second, we focus our attention on various attributes of the board of directors that are supposed to oversee compensation setting, specifically focusing on board size and board independence. Board efficacy plays a crucial role in making decisions related to executive compensation packages. Finally, we also look at ownership as an indicator of governance. Large shareholders with incentives to monitor management improve the firm s operations from within by taking steps to protect their own investments in the face of potential managerial agency conflicts. Our study focuses on these measures to provide insight into the effects of both internal and external governance on backdating activities. G-index. Our first measure is the Gompers et al. (2003) corporate governance index, G- index, which measures the number of anti-takeover provisions in a firm s charter and in the legal code of the state in which the firm is incorporated. The authors document that anti-takeover provisions, an indicator of poor corporate governance, decrease firm value. The Investor Responsibility Research Center (IRRC) assembles and reports the data for the index about every two years (1990, 1993, 1995, 1998, 2000, 2002, and 2004), and the index varies between 0 and 24. Note that current literature has extensively used this as a general measure of shareholder 7

9 rights (see, e.g., Cremers and Nair (2005) and Davila and Penalva (2006)). E-index. As a second measure, we replace the G-index with the entrenchment index developed in Bebchuk et al. (2009). This index is based on the same IRRC data but uses only six of the provisions in the firm s chapter. Bebchuk et al. (2009) show that out of the 24 provisions, these six have the greatest impact on firm value. Following Bebchuk et al. (2009), we denote the entrenchment index as E-index. The E-index ranges between 0 and 6; higher values indicate weaker shareholder rights or more entrenched management. BoardSize. We use the size of the board of directors to measure its effectiveness. As suggested by Jensen (1993), the cost of poorer communication and decision-making associated with larger groups will make a larger board less efficient. Yermack (1996) documents a clear inverse relation between firms market valuation and the size of their boards of directors. His result suggests that a smaller board is usually more efficient in monitoring CEOs. Board Independence. Starting with Weisbach (1998), many papers have found boards dominated by independent directors to be more likely to make decisions that are in the interests of shareholders. We measure the board independence as the fraction of independent directors on the board. Based on the IRRC, independent directors include retired executives of other firms, academics, private investors, and executives of unaffiliated firms. Institutional Ownership. As documented by existing literature, institutional shareholders maintain a strong monitoring role on managers. When institutional investors have larger amounts at stake in firms, they tend to have stronger incentives to devote resources to monitoring. In particular, Hartzell and Starks (2003) show that institutional ownership has a strong influence on corporate compensation policies. We measure institutional influence as the proportion of the 8

10 firm s shares owned by the top five institutional investors. 5 Of the above five governance measures, G-index and E-index measure external governance, while the other three measure internal governance Measures of CEO Incentive Based on existing empirical studies, we use two variables to proxy the CEO s incentive pay. Pay-Performance Sensitivity (PPS). Following Jensen and Murphy (1990), PPS is the dollar value of the CEO s wealth change for a $1,000 change in shareholders value. Although managers can receive pay-performance incentives from a variety of sources, the majority are due to ownership of stock and stock options (Jensen and Murphy (1990)). Similar to Aggarwal and Samwick (2003) and Core and Guay (1999), we compute this sensitivity as the dollar value change of stock and options held by a CEO to a $1,000 shareholder return. For common stock, PPS is simply the fraction of the firm that the executive owns. PPS for options is the fraction of the firm s stock on which the options are written multiplied by the options delta. We use the method developed by Core and Guay (2002) to estimate option deltas. Their method avoids the cost and difficulty of collecting option data from various proxy statements since it requires information from only the most recent proxy statements. More important, the authors show that their estimates are effectively unbiased and 99% correlated with the measures obtained if the parameters of a CEO s option portfolio were completely known. 6 Option-Grant Sensitivity (OGS). This sensitivity measures the dollar value change in a CEO s option grant per $1,000 change in shareholder value. Following Yermack (1995) and Hartzell and Starks (2003), we first calculate the delta of every option grant, and then multiply 5 As robustness check, we also measure institutional influence using the stock ownership by all the institutional investors in the firm or using the Herfindahl index of institutional ownership concentration. Our results are not sensitive to these alternative measures. 6 We also use Core and Guay s (1999) method to measure pay-performance sensitivity as the CEO s wealth change for 1% shareholder return; our results are qualitatively similar. 9

11 the delta by the number of options granted and divided by the number of shares outstanding at the beginning of the year. Finally, we multiply this number by 1000, which gives the sensitivity of the dollar value of option grants for per $1,000 change in shareholder wealth. Analyzing OGS independently is important because stock options have replaced base salaries as the single largest component of compensation (Murphy (1999)). Thus, if backdating is influential then we expect it to be quite prominent in this component of pay Control Variables We use various control variables in the regression analysis. We measure firm size as the natural logarithm of the firm sales, book value of equity as the sum of the common equity value and deferred tax, and market value of equity as common shares outstanding times fiscal year closing price. To control for firm growth opportunities, we compute market-to-book (M/B) as the ratio of market value of total assets over the book value of total assets, where the market value of total assets is obtained as the book value of total assets minus the book value of equity plus market value of equity. Return on assets (ROA) is measured as the ratio of operation income before depreciation over total assets. We compute Leverage as the ratio of long-term debt and current debt over total assets. To measure the firm risk, we use stock return standard deviation based on the firm s monthly returns over a five-year period. We also include the firm s annual stock return to control for the stock performance. We measure GrantSize as the number of options in each option grant deflated by the firm s total shares outstanding. The dummy OtherExecutive equals one if one of the firm s top five non-ceo executives also receives option grants on the CEO s option grant date plus/minus one day, and zero otherwise Data Sources We obtain our sample of stock option grants to CEOs from the Thomson Financial Insider 10

12 Filing database. This database captures insider transactions reported on SEC Forms 3, 4, 5, and 144. Like Heron and Lie (2007), we include only observations with a cleanse indicator of R ( data verified through the cleansing process ), H ( cleansed with a very high level of confidence ), or C ( a record added to non-derivative table or derivative table in order to correspond with a record on the opposing table ). Following Heron and Lie (2007), an option is regarded to be issued at the money if the exercise price is equal or close enough to the stock price on the grant date. We take the transaction date provided by Thomson Financial as the grant date if it is a trading day and the closest prior trading date in CRSP in other cases. A close enough price is defined as a price that is within 1% of the strike price. We exclude all grants not issued at the money, following Lie (2005) and Heron and Lie (2007). Consistent with prior research, more than 80% of the option grants are issued at the money. The existing literature generally separates scheduled grants from other ones, since it is unlikely that firms manipulate scheduled grants. Like Lie (2005), we define a grant as scheduled if it is issued on the same date plus/minus one day in the preceding year. Scheduled grants are eliminated from further analysis. We further require that our sample firms have available accounting data in Compustat, executive compensation data in ExecuComp, board data from the IRRC, and institutional ownership data from Thomson Financial. We obtain the G-index from Andrew Metrick s website and E-index from Lucian Bebchuk s website. 7 Our final sample consists of 8,486 unique option grants, 6,513 unique firm-year observations, and 1,971 unique firms from 1995 to Andrew Metrick s website is Lucian Bebchuk s website is 11

13 Except for the backdating variables, all the variables are measured at the end of each fiscal year. G-index and E-index are taken from the closest previous update. All of the dollar variables are measured in 2000-constant dollars. To ensure that data outliers do not drive our results, we winsorize all continuous variables at the 1 st and 99 th percentiles Summary Statistics Table 1 reports the distribution of CEO stock option grants across time. The number of option grants is increasing during our sample period. The accumulative stock return after the grant date is, on average, higher than that before the grant date, as indicated by our backdating variables. Consistent with Heron and Lie (2007), the degree of backdating becomes less obvious after 2002 due to SOX, because SOX requires that option grants must be reported within two business days. Table 1 also shows that backdating is economically important. For instance, the average difference in accumulative stock returns for the 50-day period before and after the grant date (Backdate50) is about 4.08 percentage points for our full sample. Figure 1 displays the average cumulative stock returns around the receipt of stock option grants by CEOs. Consistent with Lie (2005) and Heron and Lie (2007), the stock prices start to decline slightly more than a month before the award date. Nevertheless, there is a sharp turnaround of the price movement on the dates immediately after the grants; the prices increase. The price increase is more dramatic during the first few days after the awards; however, it continues to rise in the following 50 days after the awards. Our measure of backdating is the difference between post-grant and pre-grant stock returns. It is possible that this measure largely captures the firm s volatility. We therefore explicitly examine the correlation between our backdating variables and firm stock volatility and find their correlation coefficients pretty small in magnitude. For example, the correlation coefficients 12

14 between Backdate50 and ReturnStd and that between Backdate50 and ReturnStd 2 are only 0.12 and 0.11, respectively. This result suggests that our backdating variables are not indications of volatility. Table 2 shows the characteristics of our sample firms. The median value of a CEO s total annual compensation (ExecuComp Item TDC1) is $3,184,000 and the median cash pay is $1,173,000. The variable PPS has a mean of $21.6 per $1,000 shareholder return and a median of $6.5; this number is quite similar to that reported by Hall and Liebman (1998). 8 The average OGS in our sample is $0.63 per $1,000, with a median of $0.32, implying that option grants play an important role in aligning managers interests with those of their shareholders. The median firm has a G-index value of nine, E-index value of two, and nine directors sitting on the board with 70% of them being independent directors. Institutional investors are holding a sizeable amount of equity, with a median Top5holding of 25%. A median option grant includes 0.06% of the firm s total shares outstanding, and 55% of our sample grants coincide with the option grants to top five non-ceo executives. The median firm is quite large; its annual sales volume is $1,206 million. The sample firms are performing well, with a median M/B of 1.55, ROA of 13%, and yearly stock return of 10%. The firms are moderately levered with the median leverage ratio of 21%. 3. Does Governance Explain Option Backdating? Are firms with stronger governance associated with less backdating? The empirical answer, revealed in Figure 2 and explored more rigorously in regressions in Table 3, is no. Figure 2 provides a visual indication of the challenge for a governance explanation by providing splits of the data across the various governance measures described above. For example, Panel A of Figure 2 displays the average cumulative stock returns around the date of CEO option grants for 8 The pay-performance sensitivity reported in Hall and Liebman (1998) is $25 at the mean and $5.29 at the median. 13

15 sub-samples based on sample median for the G-index. Strikingly in the figure there is apparently far more backdating in the low G-index firms, where low G-index indicates better governance. The negative trend of stock price before the award date and the positive trend afterward are more evident for the sub-sample with a lower G-index. This difference between the two sub-samples is more significant during the 20 days before to 20 days after the awards. Panel B of Figure 2 shows the same relation between corporate governance and backdating when using E-index as the governance proxy. Given that the G-index and E-index focus on external governance mechanisms, we next turn to measures of internal corporate governance: board characteristics and ownership by institutional investors. Panel C of Figure 2 displays the average cumulative stock returns around the CEOs option grants for sub-samples based on BoardSize using the sample median BoardSize value as the cutoff. The negative trend of stock price before the award date and the positive trend afterward are more significant for the sub-sample with a smaller BoardSize. This difference is persistent between the two sub-samples during the entire sample from 50 days before to 50 days after the awards. Since a small board is usually considered more efficient than a big one, this figure also shows that better governance is associated with more backdating. Panel D of Figure 2 displays the stock price movement around the CEO s option grant dates for sub-samples made using the sample median Top5holding value as the cutoff. The negative trend of stock price before the award date is not noticeably different from the two sub-samples; however, the positive trend after the award date is slightly steeper for the sub-sample of high Top5holding. This figure suggests that firms under stronger shareholder control backdate slightly more. The only exception to this surprisingly negative relationship between governance measures and backdating is provided in Panel E of Figure 2 where we use the sample median of Board 14

16 independence to measure corporate governance. The V shape trend in stock returns is more pronounced in the sub-sample with low board independence, indicating that firms with fewer independent directors are backdating more. The result in Panel E is different from those in Panels A D. We address this difference in more detail when we investigate the relationship between backdating and governance in multivariate regressions. While interesting, these simple univariate relationships could be confounded for by other differences across the firms when they are broken into groups. To test for the power of governance to explain backdating we ran a series of regressions of governance on backdating, this time employing a wide range of control variables. Specifically, we estimate the following model: Backdate = a + a Governance + a FirmSize + a ReturnStd + a ROA it 0 1 it 1 2 it 1 3 it 1 4 it 1 + a M / B + a Leverage + a StockReturn + a PostSOX 5 it 1 6 it 1 7 it 1 8 it + a Tenure + a GrantSize + a OtherExecutive + IndustryDummy + e 9 it 10 it 11 it it where i indexes firms and t indexes year. To control for industry variation in the backdating activities, we include the Fama-French 48 industry dummies in each regression. We also include the dummy PostSOX, which takes the value of one if the option grant is given after September 1, 2002, and zero otherwise. Throughout the entire empirical test, p-values are computed based on robust standard errors clustered at the firm level. Finding a positive coefficient for a1 would be consistent with our prior result from Figure 2 that stronger corporate governance is associated with more backdating. For brevity, we only report the regression results based on Backdate50; using other four backdating variables, Backdate10 to Backdate40, provides the same results. In Column (1) of Table 3, we regress Backdate50 on G-index as well as the controls. The coefficient on G-index is and is statistically significant at the 1% level. A one-standarddeviation decrease in G-index is associated with an increase in Backdate50 of approximately

17 percentage points, given that the sample median of Backdate50 equals 1.47%. The result is clear: Better-governed firms conduct more backdating. We use E-index as an alternative measure for corporate governance in Column (2) of Table 3 and find a positive relation between corporate governance and option backdating. The regression result highlights that the coefficient of E-index is and is significant at the 1% level. The economic interpretation is that as E-index decreases by one standard deviation, Backdate50 will increase by 1.03 percentage points. Using Ln(BoardSize) as the proxy for the board effectiveness, Column (3) shows a significantly negative relation between the size of the board and backdating behavior. The coefficient of Ln(BoardSize) is and is significant at the 5% level. As a small value of BoardSize implies high board effectiveness, the column also suggests that option backdating is more prevalent when the board is more effective. To examine further the robustness of our results to the alternative governance proxies, we replace Ln(BoardSize) with Top5holding in Column (4). The coefficient on Top5holding is not significantly different from zero, which does not support the view that CEOs in firms with strong shareholder control are less likely to backdate. In Column (5), we use Board independence to measure corporate governance. Surprisingly, we find a negative coefficient on it, which indicates that backdating is less likely to occur when more independent directors are on the board. One possible reason why the result in Column (5) is different from those in Columns (1) (4) is that an outsider-controlled board does not necessarily imply better governance. Harris and Raviv (2008) predict that shareholders can sometimes be better off with an insider-controlled board when it is difficult for outsiders to acquire information about the firm s operation. Supporting Harris and Raviv s (2008) model, Duchin et al. (2010) find that when the cost of acquiring information is high, firm performance worsens after outside 16

18 directors are added to the board. To further investigate whether Harris and Raviv s theory can explain the negative relation between board independence and backdating, we use M/B ratio, R&D expense, Hightech indicator, and dispersion of analyst forecasts as four proxies for information asymmetry between insiders and outsiders, 9 and interact them with board independence in Table 4. We find that the coefficients on the four interaction terms are all negative and significant, indicating that the negative association between board independence and backdating is more pronounced when the cost of information acquisition is high (i.e. when having outside directors on the board is less beneficial to shareholders). Overall, the results from Table 3 and Table 4 show that strong corporate governance is usually associated with more backdating, which does not support the view that backdating is due to powerful managers manipulating stock option grants at the cost of shareholders. To better understand how corporate governance influences backdating, we investigate some other interaction terms in Table 5 by focusing on the G-index. We choose to focus on G-index because G-index is quite stable over time (Gompers et al. (2003)) and thus the negative relation between G-index and backdating is less likely to be driven by the possibility that firms adopt improved governance as a response to backdating. In Column (1), we interact corporate governance with stock return volatility; the interaction G-index ReturnStd has a significantly negative coefficient and the variable ReturnStd has a positive coefficient. This result indicates that firms with higher stock volatility are doing more backdating and, notably, strong governance strengthens this relation. Heron and Lie (2009) also find that firms with more volatile stock do 9 Existing literature suggests that firms with high M/B ratio, firms with high R&D expense, firms in the high-tech industry, and firms with great dispersion in analyst forecasts have more information asymmetry between insiders and outsiders (see for example, Barclay and Smith (1995), Opler et al. (1999), Harris and Raviv (2008), and Duchin et al. (2010)). 17

19 more backdating. They interpret this fact as evidence of managerial power explanation, since the CEO can gain more from backdating if the stock is volatile. But their interpretation is inconsistent with the result that the positive association between volatility and backdating is more evident for better-governed firms. In Column (2), we examine the interactions among SOX, backdating, and corporate governance. Consistent with existing literature, the coefficient of PostSOX is negative and significant, indicating that SOX has greatly curbed backdating. Furthermore, the coefficient of G-index PostSOX is significantly positive, which implies that the positive relation between governance and backdating becomes weaker after the SOX implementation. We interact G-index with CEO tenure in Column (3). The Tenure variable itself has a significantly positive coefficient, indicating that longer-tenure CEOs are backdating more. Bebchuk et al. (2010) interpret this result as that entrenched CEOs are more likely to manipulate their compensation terms via backdating. However, this interpretation is not consistent with the significantly negative coefficient of G-index Tenure, because this coefficient suggests that the relation that longer-tenure CEOs backdate more is more evident when the firm has stronger governance. In Columns (4) (5), we interact G-index with GrantSize and OtherExecutive, respectively. We find that CEOs are backdating more when they are receiving larger option grants and when other top executives are receiving option grants on a similar date. However, the above relations are stronger for better-governed firms. To formally examine the endogeneity concern that firms that struggle with backdating adopt improved governance as a response (Collins et al. (2009)), in Figure 3 we plot sample splits based on the median value of Backdate50, and track the firm s G-index from three years prior to 18

20 the option grant to three years afterward. We find that, despite a slight upwards trend, the G- index is largely stable in both sub-samples and the patterns of G-index are almost parallel across the two types of firms. Thus, there is no evidence that firms change their G-index in response to executive backdating. Overall, our empirical analysis does not support the view that poor corporate governance drives backdating option grants. Surprisingly, better-governed firms are at least as likely to engage in backdating and in some of our tests better governance predicts backdating activities. This is a puzzle for the managerial power explanation. If this is insufficient, what else might be going on? In the next section, we propose a theory to explain the rather puzzling results presented above. 4. What Explanations are Consistent with a Positive Relationship between Governance and Backdating? The data suggest a need for an explanation that shows a positive relationship between governance and backdating. What theory provides such a rationale? One candidate is the dominant alternative to managerial power explanations for compensation setting: efficient contracting. Does efficient contracting suggest such a relationship? 4.1. Optimality of In-the-Money Options In a standard principal-agent model, shareholders choose an optimal compensation package for the manager for maximizing stock price. The optimal pay structure, consisting of some base salary and stock-based income, balances the manager s incentive and risk. Building on the work of Hall and Murphy (2000, 2002) who previously explored the optimality of various option arrangements, we show that allowing managers to receive a low-strike-price option reduces the risk posited on risk-averse and under-diversified managers, and therefore, saves shareholders the expense of compensating them for bearing risk, while holding managerial incentive constant. 19

21 Like Lambert et al. (1991) and Hall and Murphy (2000, 2002), we distinguish the values of the option grants between executives and shareholders. The shareholders opportunity cost equals the gain that shareholders could have achieved by selling the options to outside investors, derived by Black-Scholes (BS) valuation model. However, the manager usually cannot trade or sell his options in the market. He is also less diversified than outside investors. Therefore, the manager values his option grants less than outside investors would. We measure the option value to executives by the amount of riskless cash compensation the executives would exchange for the option. We assume that the CEO has non-firm-related wealth w, holds s shares of the firm s stock, and is awarded n options to buy n shares of stock at the exercise price k in T years. 10 We also assume that w is invested at the risk-free rate of r f and the realized stock price at T is P T. The CEO s wealth at T is given by: W=w(1+ r f ) T + s P T +n max(0, P T k) If he receives cash V instead of the option grant and invests the cash in risk-free assets, then his wealth would be: W V = (V + w)(1+ r f ) T + s P T We solve Equation (1), listed below, for the certainty equivalent of cash V for the CEO to be indifferent between the two choices, using numerical methods. v U ( W ) f ( PT ) dpt = U ( W ) f ( PT ) dpt (1) 1 ρ W U (.) = is the CEO s utility function. We assume constant relative risk aversion ρ and 1 ρ, 10 Following Hall and Murphy (2000, 2002), we do not explicitly model restricted stocks in the compensation package. This is because in the model, restricted stock is a particular type of options with a strike price of zero. Moreover, if we allow the CEO to receive m shares of restricted stock together with n shares of options, the parameter s will be replaced with (s+m) in the simulation. Given that a CEO s existing ownership is usually much larger than his annual equity grant (Core, Guay and Thomas (2005) and Jensen and Murphy (1990)), s (s+m) and the model s implication will be largely the same. 20

22 that the stock price follows a geometric Brownian Motion with volatility σ and drift m = r f +β r r ) where β is the firm s systematic risk and r m is the return on the market portfolio. ( m f The simulations are derived assuming no dividends, σ = 0.30, β = 1, r f = 6 percent and r m r f = 6.5 percent, following Hall and Murphy (2000). We define incentives as the change in the certainty-equivalent option value for each $1 change in the stock price. 11 We imagine a two-stage process for deriving the optimal contract. The first stage minimizes compensation cost for an arbitrary incentive level, while the second stage solves for the optimal incentive level that maximizes firm value, given the results of the first stage. The second stage requires information on the production function linking executive actions to stock prices and the disutility function for those actions. While the second stage is beyond the scope of our paper, we focus on the first stage by calculating the strike price that minimizes the cost of the option grants for a given level of executive incentive. Without loss of generality we assume a one-to-one correspondence between the action level taken by the CEO and his incentive level. Extending Hall and Murphy (2000, 2002), we solve Equation (2) to find the exercise price that minimizes the company s cost of granting options, while holding CEO incentive constant. Min k nc (2) s.t. V = CONST P where V P is the CEO incentive, n is the number of options, and C is estimated cost of issuing options calculated by the BS model. 11 For example, suppose that the certainty-equivalent option value is V1 when stock price at time T is P1, and suppose that the certainty-equivalent value will be V2 when P1 increases to P1+1. In this case, the incentive in our model is simply computed as V2 minus V1, which measures the sensitivity of executives option value to the stock price. 21

23 In Figure 4, we assume that the CEO has initial wealth of $5,000,000 and 66% of his wealth is invested in the company stock, while varying incentive levels. The relative risk aversion is assumed to be two and the options are assumed to be held for ten years. The incentive is interpreted as the change in the certainty equivalent values of n options for each $1 change in the stock price. The grant-date stock price is normalized to be $100. Shareholders are aimed to provide managers with a certain incentive level. Without loss of generality, we set four different levels for incentive: $1,000, $2,000, $3,000, and $5, The four lines in Figure 4 correspond to the four incentive levels, respectively. As illustrated in Figure 4, the minimum BS cost is usually achieved when the strike price is set to be less than the grant-date price. For example, the minimum cost to the shareholders to achieve the executive incentive level of $3,000 is to issue option grants with the strike price of 75% of the grant-date price. The simulation results for a different range of model parameters are reported in Table 6. The relative risk aversion coefficient is assumed to be either two or three. The CEO incentive level changes from $1,000 to $5,000. The proportion of the CEO s wealth invested in the firm is assumed to be either 33%, 50%, or 66%. For a wide range of model parameter values, the optimal strike price is lower than the grant-date price. This phenomenon is more evident for the manager who is more risk-averse and more under-diversified. For the same level of managerial incentive, shareholders can save compensation costs by optimally setting the strike price. The percentage of cost savings, relative to setting the strike price to the grant-date price of $100, varies from 0.25 percentage points to 25 percentage points. Out-of-money options become optimal only when we assume a low level of risk aversion and a low level of CEO under- 12 These incentive levels are interpreted as if stock price at year T=10 increases by $1 dollar, the CEO s certainty equivalent value will increase by $1,000, $2,000, $3,000, and $5,000, respectively. 22

24 diversification. Poor diversification makes the CEO vulnerable to unfavorable outcomes. From the executives perspective, options issued in the money will be usually less risky, because they have a high probability of ultimately being in the money. This intuition is quite consistent with that advanced in Hall and Murphy (2002) Tax/Accounting Issues There exists a fundamental difference in the taxation of at-the-money versus in-the-money options. If the option s strike price is greater than or equals the grant-date stock price, then the executive who receives the option grant pays taxes on his option profit at the lower tax rate applicable to long-term capital gains. 13 The options recipient can also benefit from deferring the tax to the time of the option exercise. In short, unlike in-the-money options, at-the-money options qualify for this lower tax rate. By avoiding the tax disadvantages associated with in-the money options, the board can achieve the low cost of issuing options at the optimal strike price: Issuing at-the-money options which are backdated to a proper date. The proper date refers to the one when the daily stock price approximately equals the optimal strike price. Prior to 2005, the Generally Accepted Accounting Principles (GAAP) provided that the difference between the exercise price and the firm s grant-date stock price was only expense that had to be recognized by companies with respect to options issued on a fixed number of shares at a fixed exercise price. 14 Thus, the grant of an at-the-money option resulted in zero recognized expense for financial reporting purposes. On the other hand, an option that was granted in the 13 The recipients of option grants awarded at-the-money pay taxes on the profit from incentive stock options at favorable long-term capital gains rates. The incentive stock option grants sometimes make up as low as 10% of the outstanding options to top executives. However, given the size of executives entire option grants and the difference between the tax rates, the magnitude of the tax difference of the entire option profit is still quite sizable. 14 Prior to SFAS 123(R), most firms used the intrinsic-value method to expense for executive stock options because SFAS 123 allowed firms to choose either the fair-value based method or the intrinsic-value method to account for the options. Under the intrinsic-value method, firms could escape recording an expense associated with options if they granted a fixed number of options with a fixed exercise price set at or above the market price of the underlying stock on the grant date. 23