The Role of Corporate Governance and CEO Compensation in Backdating. of Executive Stock Options

Transcription

1 The Role of Corporate Governance and CEO Compensation in Backdating of Executive Stock Options Lalatendu Misra, University of Texas at San Antonio Yilun Shi, St. Edwards University * Corresponding author. Yilun Shi, School of Management and Business, St. Edwards University, 3001 S. Congress Ave., Austin, TX , Tel.: ; fax: yiluns@stedwards.edu

2 The Role of Corporate Governance and CEO Compensation in Backdating of Executive Stock Options Abstract We examine a sample of firms that are known to have engaged in option backdating to identify the determinants of backdating behavior. The sample firms are generally smaller, younger, exhibit higher sales growth and also potential for growth, have weaker operating performance and have greater variability in operating performance compared to non-backdating firms. We control for these variables and obtain a propensity score matched set of control firms which are not known to have backdated. Next we focus on the governance structure, compensation, and CEO characteristics and find that smaller audit committee size, higher proportion of independent directors, lower G-index, lower CEO pay concentration ratio, higher amounts of option compensation and lower cash compensation are associated with higher likelihood of backdating.

3 The Role of Corporate Governance and CEO Compensation in Backdating of Executive Stock Options 1 Introduction Executive stock options (ESO) currently constitute a major component in the total compensation of executives for most large corporations. The awarding of the ESO has seen substantial growth during the last two decades particularly for publicly traded companies. In keeping with the growth in ESO, suspected option abuses and opportunistic manipulations have also escalated rapidly. Recent ESO related corporate scandals include those associated with excessive size of awards, earnings and corporate performance manipulations, and backdating of option awards which constitutes the focus of the present study. Revelations of these activities have resulted in stock price declines and a generally negative impact on investor confidence. In his seminal study Lie (2005) presents evidence regarding the ubiquity of backdated option grants. As the scandal of actual backdating practices unfolded in the financial press in 2006, many companies restated their past financial statements and suffered substantial market value losses. Several high profile executives including Kobi Alexander of Comverse Technology were prosecuted or indicted by the regulatory agencies. The focus of our study is to examine which of the following factors; firm characteristics, CEO compensation packages, board structure, and corporate governance structure best explains the incidence of option backdating in firms which have allegedly engaged in such activities compared to firms that did not engage in backdating. Option abuse did not start with the press reports of backdating scandal in Firms had engaged in this practice for at least a decade. Additionally, earlier studies including Yermack (1997) and Aboody and Kasznik (2000) provide evidence that managers use their power to control either the timing of option grants or the timing of releasing important corporate news which can boost the value of their option grants. Backdating transactions, however, appear to escalate the egregiousness of the 2

4 opportunistic behavior to a new level as managers directly manipulate the granting price to exploit personal benefits. Typically, executive stock options are granted with the strike price set equal to the closing price of the stock on the grant date. Option holders gain as the stock price of the firm increases presumably on account of value enhancing corporate actions by the managers. Backdating occurs when a company sets the option strike price equal to the closing stock price on a date preceding the actual grant date and, in effect, grants the option retroactively at a lower strike price than the prevailing stock price on the actual grant date. The option is thus effectively in-the-money on the date of actual grant date. 2 Christopher Cox, the former Chairman of the SEC described the practice in testimony before a congressional committee on September 6, 2006, disguising an in-the-money option through backdating is to allow the person who gets the option grant to realize larger potential gains-without the company having to show it as compensation on the financial statement. The financial press has played an important role in disclosing the extent of the option backdating scandal. The Wall Street Journal (WSJ) started reporting a list of firms which have been alleged to engage in backdating in its Option Scoreboard in March These companies had either voluntarily initiated internal reviews, undergone investigations by SEC or by DOJ, or face civil law suits from investors with regard to their option granting practices. Bloomberg also published a list of backdating firms in By the end of 2007, more than two hundreds firms were named to have engaged in option backdating. Subsequent academic studies show that backdating has been a common phenomenon for a number of years (see, e.g., Lie, 2005, Lie and Heron, 2007, and Narayanan and Seyhun, 2007). Heron and Lie (2007), for example, provide an estimate that approximately 1/5 th of the unscheduled option grants were backdated during the period 1996 to While the incidence of backdating and the effect of backdating have been examined in the literature, there is limited evidence on the cross-sectional determinants of backdating. Our study aims to 2 Granting of in the money stock options is not illegal. The granting firm is required to record the in-the-money part as an expense. Non-reporting of this expense is potentially misleading to the investors and is likely to hold adverse tax consequences for the firm 3

5 close this gap. We examine how a variety of factors, including firm characteristics, corporate governance structure, CEO compensation structure, and board structure, contributes to the incidence of option backdating. In this regard, one stream of studies analyzes the influence of corporate governance system on backdating for a group of presumably backdating samples (see, e.g., Collins et al, 2007, and Bebchuk et al, 2006a, 2006b). Following Lie (2005), they construct their likely backdating samples by inferring from the relationship between option strike price and the stock price preceding the grant. Collins et al (2007) show that backdating firms exhibit weaker governance structure and that their CEOs are more powerful in determining the size of compensation. We employ a different study design which has lower measurement error for the sample. We focus on firms that are ex-post known to have engaged in backdating practices. Second, we control for a more complete set of firm characteristics and CEO features, such as firm age and CEO tenure, which may have influence on the likelihood of backdating. Last, following Erickson et al (2006), we compare backdating firms to two different control groups which are either termed unmatched or matched. The unmatched group includes all non-backdating firms appearing in the ExecuComp database (see Burns and Kedia, 2006 for similar application). The matched group is formed by applying propensity score matching which is more suitable in a multi-dimensional matching framework (see Chari, Chen and Dominguez, 2008 for an application). The use of two competing approaches allows for robust conclusions. We obtain a sample of 152 backdating firms from the WSJ or Bloomberg. This sample is used in the previous essay to estimate the losses on the announcement of the backdating event. The average 3- day market value loss for our sample firms is $67.25 million. We employ the same sample to study the determinants of the backdating decision. In order to do so, we also use all non-backdating firms in the ExecuComp database or a subset of these firms based on propensity score matching as control firms. We find that the backdating firms are significantly different from the control firms in size, age, operating performance, and available growth opportunities. Backdating firms are smaller, younger, less profitable, and exhibit greater volatility in their operating performance. These firms also experience 4

6 higher sales growth and display higher market-to-book ratios suggestive of potential for higher growth. We employ a propensity score matching approach and control for these firm features and create a matched group. We test whether CEO features, CEO compensations, corporate governance, and board structures vary systematically between the backdating companies and the control group. We find that, contrary to popular belief, corporate governances of backdating firms are not weaker but in fact are somewhat stronger. Both G-index and E-index are lower for backdating firms than their comparable peers. Backdating firms, however, have smaller boards and subcommittees possibly due to their smaller size. CEOs from backdating firms have less cash compensation but are granted more options. Even though they have a lower level of pay concentration, backdating CEOs rely more on option-based compensation. We conclude that backdating firms appear to be somewhat less mature firms with unstable operating performance, smaller board, and younger CEOs. They pay their CEOs more options. Relative to the control firms they do not appear to be worse in terms of governance structure or practice, albeit, they have smaller sized critical committees for compensation and audit. We review some related literature in the next section. Sample collection and the construction of the matched control are shown in Section 2.3. Empirical results are presented in Section 2.4 with Section 2.5 providing some concluding remarks for this essay. 2 Literature Review There are three strands of literature that are germane to our study. One group of research has examined a variety of managerial manipulations that can occur around option awards. The second set of studies focuses on the corporate governance structure and executive turnovers for firms which are reported to engage in financial fraud and/or accounting irregularities. The last strand of related literature includes studies which examine the ex-ante potential for option backdating based on the timing of option awards for a large set of firms. 5

7 Option backdating is just another, albeit more recent, manifestation of managerial manipulation of option awards. Opportunistic managerial behavior, including manipulation of earnings announcements or other corporate news around option exercise date or timing option exercises based on the reported earnings have been examined in the literature. Bartov and Mohanram (2004) report abnormal positive earnings prior to option exercising and negative stock price performance during the post-exercising period, evidence which is suggestive of the use of prior private information regarding weak earnings in the post-exercise period. Huddart and Lang (1996) document negative relationship between exercising and subsequent stock price performance in a smaller sample, suggesting the possibility that managers exercise their options ahead of poor news about corporate performance and before the maturity of the options. Yermack (1997) reports that managerial stock option awards tend to occur prior to the announcement of favorable corporate news. The awards, per se, do not drive up the stock price since stock price increases occur even though the news of the award is revealed, via proxy statements, at some future date. Aboody and Kasznik (2000) also study the relationship between stock return and option grants and report that companies manipulate option granting price by controlling the information flow. Managers report bad news before option grants and delay good news announcement after options are awarded (also see Chauvin and Shenoy, 2001). Our paper is also related to a group of studies that examines the relationship between compensation and the occurrence of financial fraud or accounting irregularity. The link between levels of compensation and the incidence of accounting restatement and financial fraud is somewhat murky and the evidence mixed. Erickson, Halon and Maydew (2006) report the absence of significant association between the sensitivity of top executives equity holdings and fraud after controlling for governance and other factors. Burns and Kedia (2006) show a positive association between accounting restatement and the sensitivity of CEO s stock option holdings. Efendi, Srivastava and Swanson (2007) examine whether 6

8 option based incentive is associated with the incidence of financial restatement. They show that as executives in-the-money option values increases, the likelihood of accounting misstatement also rises. The relationship between governance and financial fraud or accounting irregularity is examined in a group of studies. Using a sample of 159 companies with financial restatements and their industrysize matched control firms, Agrawal and Chadha (2005) report o relationship between corporate governance, measured by the independence of board and audit committee, and the odds of accounting fraud. Uzun, Szewczky and Varma (2004), however, show that fraudulent firms have fewer outside and independent directors compared to control firms. Lie s (2005) seminal work has sparked many subsequent studies on backdating. Extensive evidence regarding the potential for backdating is provided in a series of subsequent papers by Narayanan and Seyhun (2007), Lie and Heron (2007), Bebchuk et al (2006a, 2006b). Lie shows that stock price consistently under-performs before options are granted and uniformly outperforms afterwards, especially for unscheduled option grants. Unless corporate boards have superior ability to predict future movements in stock price it is unlikely that based on available information they can consistently grant options when the stock price hits its lowest level. The alternative explanation, and the one that appears plausible based on numerous subsequent studies, is that executives set the option granting date retroactively and pick the grant date when the stock price is the lowest during a period of time. Such unusual patterns of price movement around the option granting date have been observed less frequently since the enactment of the Sarbanes-Oxley Act (SOX) which required executives to report their option grants in a timely manner (Heron and Lie, 2007). As opportunities for manipulations get smaller, the abnormal gains to the executives also appear to have declined. Heron and Lie (2007) estimate that approximately 23% of unscheduled option grants were possibly backdated in the pre-august 2002 period, however, only 10% of the grants were backdated in the post-sox period. The potential for backdating is examined with an extensive sample of option awards by Narayanan and Seyhun (2007). They also provide evidence of forward-dating, by which executives set 7

9 option grant date in the future as stock price continues to go down. Bebchuk, Grinstein, and Peyer (2006a) identify factors relating to corporate governance and firm characteristics that appear to contribute to backdating. The incidence of lucky CEOs who get the options awarded on the day of the lowest stock price in a month, they argue, is too high to occur by chance alone. They report that the likelihood of backdating increases as the CEO becomes more influential in setting the terms of his option payment. Also, a CEO s longer tenure and a board s lack of independence increases the possibility of backdating. Backdating is more likely when the potential value capture by the executive is high. In a follow-up study, Bebchuk et al (2006b) discuss similar issues with regard to options granted to directors. More recently, Collins, Gong and Li (2007) find that weaker corporate governance structure is associated with backdating behavior. The percentage of insider director and CEO duality, i.e. CEO serves as the Chair of Board, are positively related to the incidence of backdating. Moreover, as executives rely more on stock-based compensation, the probability of backdating rises accordingly. Conversely, as managerial power is constrained, for instance, by the presence of large block-holders the incidence of backdating is also reduced. They also show that, when board members are interlocked with other backdating firms, the odds of granting backdated options increases, an issue that is also examined in Bizjak, Lemmon and Whitby (2008) with similar conclusions. The focus of the earlier studies on backdating is on the ex-ante likelihood of firms engaging in backdating behavior. Our approach is to examine the ex-post known backdaters and compare their characteristics with control firms. The use of known backdaters minimizes the measurement error compared to the studies that employ sample of firms who are likely to have engaged in backdating activities. We employ robust estimation methods by using propensity score matched control group so that each sample firms is paired with a unique matching firm which shares similarities in multidimensional firm characteristics. Most of the earlier studies focus on the role of corporate governance and CEO compensation in backdating. They do not include firm specific characteristics that are likely to influence the occurrence of 8

10 option backdating. Given certain firm specific features or managerial inclination to engaging in actions of uncertain legality, backdating could still occur in the presence of properly functioning and effective governance structures. That is, effective governance structure, though likely to mitigate such occurrence, does not provide an absolute guarantee against such incidents. We employ an additional set of independent variables capturing firm characteristics in addition to CEO compensations, corporate governance, and board structure. We believe that as in the case of other corporate scandals, backdating is the consequence of a complex interaction of a set of variables, rather than as a response to a simple structure of weaker corporate governance or excessive compensation package. Backdating may be related to firm age, operating performance, CEO pay, and board structure in addition to numerous other individual specific factors that may have a bearing on this issue, but ones that are not easily quantifiable or measurable by the researcher. We do find no evidence that backdaters are plagued by weaker governance structures. Backdating firms exhibit lower levels of G-index and E-index suggestive of better monitoring systems. Backdating CEOs are younger and rely more on option-based compensation, but based upon the pay concentration measurement, they do not appear to be more powerful than their peers. 3. Sample and Empirical Design 3.1 Sample We obtain the initial sample from WSJ Perfect Payday: Option Scoreboard column of January 10, This list includes 136 companies which were reported to have been involved in option backdating. During the later part of 2007, Bloomberg News reported a list of 208 firms involved in backdating. We use the Bloomberg cases with announcement dates prior to 2007 to augment our initial WSJ sample. We exclude firms which are incorporated outside of United States, firms with contaminated news announcements, firms that were acquired during 2005 or 2006, firms which were delisted before 9

11 2005, and firms which announced the backdating transactions in Our final sample consists of 152 firms that announced backdating transactions during the years 2005 and We collect the daily stock price from the CRSP database, obtain financial information from CompuStat, and compensation data from ExecuComp database. Governance data is hand collected from annual reports or proxy statements as applicable. We present summary descriptive statistics for our sample in Table I. Backdating firms suffered a significant value loss when the announcement of the backdating event is disclosed. The average 3-day market adjusted losses are -2.62% with a median loss of 1.83%. Our sample firms exhibit large variation in market capitalization size as shown in Panel A of Table I. With an average size of $51 millions, the largest one, Home Depot, has a market capitalization of $81 billion while the smallest one, Backyard Burger is only valued at $31 million. The median firm size is $15.45 billion. Our sample exhibits noticeable industry concentration with firms clustered in the computerrelated industry (SIC 2 digit code 36) and business services industry (SIC 2 digit code 73) as shown in Panel B. More than 30 firms belong to each of these sectors. Chemicals and allied products and machinery and computer equipment industry sectors have more than 10 cases each. In the following subsection, we describe the role of various types of variables such as measures of firm characteristics, compensation structures and CEO characteristics, and board structure variables. [Insert Table I] 3.2 Firm Characteristics: Sample and Unmatched Control We first compare the sample firms against control firms along size and other dimensions before pooling the firms for logistics regression. Following Burns and Kedia (2006), we begin our analysis by using all non-backdating firms in ExecuComp database as the control group. There are 1,206 control firms. 10

12 Size: Managers of larger firms earn more compensation and typically have higher levels of option holding and option grants compared to those of smaller firms (Murphy, 1999; Core and Guay, 1999). Thus, manipulation of option grant date may have greater financial consequences for the CEOs of the larger firms. CEOs of larger firms, however, have the potential for substantial loss of reputational capital which may act as an implicit deterrent to backdating. Conversely, CEOs of smaller firms may have greater opportunities to engage in backdating in view of their presumed ability to control smaller boards. Practice of dubious legality may be easier to commit if there are fewer prying eyes. Size, measured as the log of total assets, is expected to be negatively related to the incidence of backdating. We present the results of differences in firm characteristics across the sample and control groups in Table II. We find that backdating firms are substantially smaller relative to the control firms. The average size of backdating firms, in terms of total assets, is approximately $2.2 billion while the average of total assets of non-backdating firms is in excess of $10.6 billion. Other size measures such as sales exhibit similar patterns. Firm Age: Less mature firms are more likely to be cash constrained and rely more on equitybased compensation (Carter and Lynch, 2001). Larger amounts of equity based compensation may induce incentives for backdating. We measure a firm s age by the number of months that it has been reported as a publicly traded firm in the CRSP database. We expect backdating firms to be younger than the control group. As we show in Panel A of Table II, backdating firms are listed only for 128 months on average compared to 297 months for the control firms. Growth Opportunities: Firms with more growth opportunities may unintentionally provide their executives the incentives to manipulate stock prices and option grants. We measure growth opportunities by the average sale growth ratio over the period of and market-to-book ratio in 2002, and we expect backdating firms to have greater growth opportunities. Growth potential is used in the accounting restatement and fraud literature where one offered argument is that high growth firms tend to commit frauds in order to maintain or increase their growth ratios (see, e.g., Erickson et al, 2006; 11

13 Agrawal and Chadha, 2005). Effendi et al (2007) report that firms that undertake accounting restatements tend to exhibit higher market-to-book ratios prior to their first restatement year. Chidambaram and Prabhala (2003) compare the sales growth ratio of option re-pricing firms before and after the re-rpicing event and report that sales growth ratios tend to be higher in pre-period compared to the post-period. Backdating firms show significantly higher average sales growth ratios and book-to-market ratios compared to the control firms. Both these variables also indicate a higher growth potential for the sample firms. It is likely that backdating firms may have inflated market value which provides more incentive for option manipulation. 3 Operating Performance: Poor operating performance may result in backdating behavior as improving the stock price by way of superior corporate performance may be a difficult task for some firms. The compensation effect of the award can more easily be achieved by setting a lower strike price to boost the grant value. We measure operating performance by ROA and EBITDA following Carter and Lynch (2001). We use the average of annual ROA and EBITDA to smooth out the potential for year to year variations. We also test whether the volatility of operating performance, measured by σ ROA and σ EBITDA over the previous four years, is relevant to backdating. The operating volatility may be a proxy for market volatility and would be positively related to the option value of the grant (Chidambaran and Prabhala, 2003). We expect backdating firms to have higher operating volatility. As we show in Panel C of Table II, backdating firms exhibit significantly lower levels of ROA and EBITDA compared to the control firms. Also, as we present in Panel D, backdating firms exhibit significantly greater sales growth volatility and volatility in their operating performance, σ ROA and σ EBITDA compared to the control firms. The volatility in sales growth rate for the sample firms is 32.78% for the 3 We compare the P/E ratio of the sample firms against the entire control group of firms. Using the larger set of control firms (1,197), we find that the sample mean and median P/E are respectively and compared to the control firms mean and median P/E of and The tests of difference in means and medians are both significant at better than 5% level. This suggests that the sample exhibits higher P/E ratios compared to the control firms in addition to exhibiting higher sales growth and market-to-book ratios as shown in Table II. Further, the sales growth and the market-to-book ratios are highly significantly correlated by both parametric and non-parametric measures. 12

14 sample firms compared to 20.93% for the control firms. The standard deviation of ROA is 20.45% for backdating firms, while it is only 5.37% for non-backdating firms with similar results with regard to EBITDA volatility. Taken together, these indicate that the profiles of the backdating firms are substantially different from the control group of firms, especially along these dimensions. [Insert Table II] 3.3. Constructing the Matched Control Group Firm characteristics, governance mechanisms, board structure, and CEO compensation are possible factors that lead to backdating behavior. Given the considerable difference between backdaters and the unmatched control group with regard to observable firm characteristics, it could be problematic to compare the groups directly since the explanatory power of governance structures and the CEO compensation variables might also be influence by firm features. There are at least two approaches to take care of this issue. We could employ variables like size, growth potential, maturity and similar factors as control variables in the regression. We do this and report the result as our last table. An alternative approach to controlling for the observable firm characteristics is to employ a propensity matching technique to construct a matched control group that will allow us to directly measure the influences of governance, board structure, and CEO features and compensation after matching on size, growth and similar factors. We follow the propensity matching technique described in Chari, Chen, and Dominguez (2008) which is applied in a merger context. The matched control group approach yields a control group of the same size as the sample unlike the larger unmatched control group shown in Table II. We first estimate a probit regression by using all sample firms and all non-backdating control firms but only employing firm characteristics as independent variables including; size, market-to-book ratio, operating performance, operating volatility, and firm age. Our sample also exhibit clustering in specific industries as shown in Table I. Therefore, we add a dummy variable, splind, if a firm s two-digit 13

15 SIC code is 36 or if it is 73; otherwise, splind takes on a value of zero. The probit regression is specified as the following: PY ( = 1 x) = Φ ( β ' x) (1) where; the vector of firm characteristic variables include size, firm age, a dummy variable for splind, market-to-book ratio, average ROA, and σ ROA. The function Φ (.) refers to the standard normal distribution. We use the sample of 152 firms and the unmatched control of 1,206 firms in the probit regression. We obtain a propensity score, i.e. the predicted probability of backdating, from the probit regression. Next, we match sample firms with multiple-matched control firms and then rank each control firm according to its Mahalanobis metric, D M, which is computed as follows: D = X X S X X (2) T 1 M ([ S C] C [ S C]) where; X S and X C are the propensity scores of sample and control firms respectively, and S C is the covariance matrix of the matching variables. If one control firm is paired with more than one sample firms, the pair with the smallest Mahalanobis distance is picked. We drop this pair from the next iteration of the probit regression which includes the remaining sample and control firms. The matching process is repeated and pairs of firms are sequentially removed until each sample firm is paired with a unique control firm. To verify the accuracy of matching procedure, we compare all the variables we presented in Table II for the sample and the matched control firms in Table III. It may be noted that the size matched as total assets or sales are very similar across the two groups. The age of the control group is somewhat greater than that of the sample. The market-to-book ratios are similar across the two groups. Also, the operating performance is of similar magnitude as is the operating volatility. It appears that the matched control group is very close to sample firms in these observable firm characteristic aspects. In each Panel, matched firms present virtually no difference from the backdating sample, except the average growth sales ratio in Panel B. Since we use market-to-book ratio to proxy growth opportunities in the 14

16 propensity matching procedure the difference is realized sales growth does not come as a surprise. We revisit the regression with the unmatched control group as a robustness check in the last section. [Insert Table III] 3.4 Comparing Sample with Matched Control In the following univariate analysis, we compare the sample with the results of the matched group. Specifically, we focus on the stock price performance, corporate governance, board structure, and CEO compensation variables. Stock Return: We compute each firm s annual return in fiscal years We hypothesize that the backdating firms, on average, would have lower stock market returns compared to the control group. Consequently, backdating becomes an attractive alternative as lowering the strike price provide one easy way to stimulate the value of option grants. As shown in Table IV, the mean annual return of the sample firms and the matched control are not statistically significantly different in any year. Comparison of stock return volatility, however, reveals that sample firms generally exhibit higher levels of volatility than the control firms. There are two potential consequences of this. First, higher levels of stock return volatility likely makes the option grants more valuable. Chidambaran and Prabhala (2003) report that when young firms with high growth potential suffer sudden price drops, many such firms use the repricing mechanism to revitalize the incentive of option-based compensation. These features may also fit the profile our backdating firms. We predict that backdating firms may also be the ones which have inferior stock performance. Therefore managers may use backdating as an alternative to boost the value of their option grants. However, contrary to our expectation, backdating firms offer very similar levels of annual stock return as the matched control group. Backdating firms do show a higher return volatility, in four out of the six year, with particularly higher levels in years This finding is in line with the argument that backdating represents managerial rent extraction behavior since higher stock volatility is associated with higher option value. 15

17 [Insert Table IV] Governance Structures Effective boards monitor managerial behavior and restrict or punish rent-seeking behavior that is inimical to shareholder interests. The board is required to approve the size of the option grant and the grant effective date prior to the actual grant. Thus, backdating represents a failure of the board to perform its monitoring duty or, worse, suggests complicity in a legally dubious transaction. The agency cost of board failure is borne by the shareholders. CEOs may engage in fraudulent activities in the absence of appropriate controls, oversight, or monitoring mechanisms in the corporate environment. A priori, it is not clear which metric provides an appropriate measure of the corporate climate for misconduct or the tolerance of such misconduct. We examine the potential roles of the size of the board and relevant committees, the proportion of independent directors, and the frequency of their meetings. Choice of time period for the data: All variables with the exception of annual stock return are measured corresponding to fiscal year The choice of the time period to be employed in our analysis is based on a detailed examination of news. We find that many firms had histories of backdating going as far back as Using panel data from 1996 is not feasible since the lack of available data would severely restrict the sample. Furthermore, SOX became effective in August 2002 and as Heron and Lie (2007) report, the incidence of backdating reduced substantially in the post-sox period. We use data from year 2002 in the following analyses since it is the middle year of the largest amount of backdating activity. Average measures such as operating performance and growth are computed over the three years preceding Board Independence: Board independence is likely to reduce the likelihood of backdating since more independent boards are less likely to support retroactive grants. Bebchuk et al (2006a) find 4 For firms which eventually filed restatements, it appears that the period 2001 to 2003 is the period during which backdating activity was at its peak. This provides another justification for the use of 2002 cross-sectional data in the regressions. 16

18 that backdating firms have low levels of board independence. The percentage of independent directors in the board is somewhat higher for the sample group compared to the matched control group as shown in Panel A of Table V. Backdating firms have slightly higher percentage of independent directors. Board Meetings: It is reasonable to expect that active and effective boards meet more frequently. Vafeas (1999) indicates that the frequency of board meetings is related to corporate performance. Specifically, both firm value and operating performance are negatively related to the number of annual board meetings. The effect of board meeting frequency on backdating is unclear. If backdating firms are those experiencing weaker financial performance, their boards might meet more often. In that case, the propensity to backdate would be positively related to the frequency of board meetings. However, if the board meeting is a social occasion without substantial monitoring benefits, more board meetings could be associated with more backdating. The number of meetings hold by backdating firms is comparable to that of the matched control group as shown in Panel A of Table V. Board Size: Yermack (1996) provides evidence that firms with smaller boards exhibit better performance, greater market value, and more effective CEO pay performance relation. Based on his findings, we expect that board size would have a negative influence on backdating; i.e., firms with larger boards are more likely to engage in backdating. Board size, however, also acts as a proxy for firm size. To the extent that larger firms are subjected to greater scrutiny by external auditors, investors, and analysts the potential for backdating may be moderated. External monitoring provides a partial substitute for effective internal monitoring. Larger board size may thus mitigate the scope of backdating. From this perspective, smaller firms are more likely to be governed with smaller boards, and in the absence of strong external or internal monitoring, are more likely to exhibit backdating. Whether smaller or larger sized boards are more associated with backdating is an empirical issue. Since the control group is size-matched, their board size is also similar to that of the backdaters as shown in Panel B of Table V. Although the average size is slightly lower, there is no statistical difference between the numbers. 17

19 As we show in Panel B of Table V, we do find that backdating firms have both smaller audit and compensation committees, and they are statistically significantly different. Since we observe these committees to be smaller after matching the control firms on size and other dimensions, it provides one possible explanation that it may be easier for the CEO of the backdating firm to wield influence on smaller critical committees and managers to influence the board to provide favorable treatment to executive compensation and possibly enable option backdating. [Insert Table V] CEO Compensation Extant literature suggests that there is some relationship is expected between CEO compensation and firm s irregular activities such as CEO s stock option grants and earnings management (Bergstresser and Philippon, 2006) and CEO pay and restatement (Burns and Kedia, 2006). Efendi et al (2006) find that misreporting is related to the size of the CEO s holding of in-the-money stock options. The level of cash compensation arguably should be unrelated to the propensity to backdate. The size of compensation in the form of option grants is likely to be positively related to the propensity to backdate, which provides larger incentive to backdate option grants. We compare CEO s salary, bonus, total cash compensation and total compensation including option grants. We also include the value realized through option exercising and the value of option grants and conjecture that they are positively associated with the likelihood of backdating. CEO compensation results are summarized in Table VI. Backdating firms pay less cash and bonus. As shown in Panel A, however, the difference of means is not statistically significant whereas the median test indicates significantly lower salary and bonus for the backdating CEOs. The total cash compensation for the sample firm CEOs is significantly lower whether we examine the mean or the median. The total compensation across the two groups, however, is not statistically significant. 18

20 We present the value of the option-based compensation in Panel B of Table VI. Backdaters do grant more option and their CEOs gain more from option exercising. Nevertheless, neither difference is statistically significant. 5 [Insert Table VI] Governance and CEO Power Governance Index: Managerial opportunistic behavior could be as a result of neglect or weakness in the monitoring system. We analyze the influence of corporate governance, measured by Governance index (Gompers et al, 2003) and the Entrenchment Index (Bebchuk at al, 2009). We hypothesize that backdating firms have higher G-index and E-index, i.e. lower governance quality. We are able to collect the E-Index data from Professor Bebchuk s website and the G-Index data from Professor Gompers website. Data is available for a subset of the sample firms and a subset of the matched control firms. As we show in Panel A of Table VII, both the indices are smaller, and significantly so, for the sample firms. This is contrary to our expectations. The average G-index of backdating samples for 2002 is 7.84 while the average number for the control group is about 8.51 as shown in Panel A. The sample firms appear to have better governance structures. CEO Duality: The power of the CEO emanates from a number of sources. Some of the observable measures are summarized in Panel B of Table VI. Crutchley et al (2007) discuss the potential for accounting fraud when the CEO is also chairman of the board. They find that CEO duality does not have an impact on the potential for accounting fraud. CEO duality is an indicator variable taking a value of 1 if the CEO also serves as the chairman of the board and a value of 0 otherwise. We posit that a backdating firm s CEO would more likely have dual appointment as the chairman of the board. As 5 In unreported tables, we have compared the CEO compensation between backdating firms and the unmatched group. Backdating firms do pay their CEO more. Such difference is likely to be due to other firm features, such as size. 19

21 shown in Panel B of Table VII, the sample firms have a lower incidence of dual CEOs although the proportion is not significantly different from the control group. CEO Tenure and Age: CEO tenure and age are related to compensation levels and to optionbackdating in the literature. Bebchuk at al (2006a) provide evidence that longer CEO tenure in office is more likely to result in lucky grants as the CEO s influence on the setting of his compensation increases with the length of tenure. As shown in Table VII, the tenure of CEOs in both the groups is similar. The sample CEOs are younger than the control group CEOs by approximately 5½ years. The difference is significant. Thus, backdaters have younger CEO, who could be more aggressive and in favor of more risk-taking behavior. CEO s Power: The size of the CEO s total compensation relative to that granted to other executives can also be viewed as a measure of CEO s importance to the firm and as a measure of CEO s relative power. More powerful CEOs may be able to influence the board and other members of the executive team to backdate the option grants. Thus CEO power is likely to influence the propensity to backdate. We obtain the ratio of the CEO s compensation to the total compensation of all executives of the firm and refer to this as the Power Concentration (PC) measure. A similar measure is also employed by Bebchuk, Cremers and Payer (2007). A larger PC number indicates that greater power concentration rests with the CEO and a smaller number suggests a dispersion of power across the executive team. The possibility of backdating is greater with higher levels of power concentrations. The PC measure across the two groups is very similar. [Insert Table VII] 4. What Distinguishes the Backdating Firms? Based on the univariate analysis, we do not observe substantial differences in CEO features except that the CEOs of the sample firms are younger. CEOs exhibit similar tenure and comparable pay and pay concentration ratios. Taken together, this does not provide strong evidence to support the 20

22 conjecture that backdating CEOs have more power in controlling the firm and the board. This is opposite to the results found in Collins et al (2007) as we employ a different sample and matching procedure. Our sample firms do have higher proportion of independent board members but exhibit smaller compensation and audit committees. 4.1 Logistic Regression In order to further examine whether any of these variables are critical in determining backdating behavior we do so in a multivariate logistic regression framework. We report the regression results in Table VIII. The dependent variable is an indicator variable with 1 identifying the backdating firms and 0 identifying the control firms. We employ the logistic regression model where function: Λ (.) refers to the logit β ' x e PY ( = 1 x) = = Λ( β ' x) β ' x 1+ e (3) As discussed earlier, we have controlled for the firm characteristics in order to construct the matched control group with the propensity score matching process. We only examine the effect of sets of board structure, corporate governance, and CEO compensation in the logistic regression. Each set of variable is first examined in an isolated regression in Models I through Model III. We pool all the explanatory variables and estimate a grand regression in Model IV. We use board structure variables in Model I. Smaller board size or lower percentage of independent directors provides potentially lower obstruction costs for the CEO to manipulate option grants. However, consistent with the univariate results, board size is not significant and contrary top expectations larger proportion of independent directors lead to higher likelihood of backdating. Smaller compensation committee and audit committee size is associated with higher likelihood of backdating. Only the audit committee size is significant. 6 6 There is a strong correlation between the committee sizes. The There is strong positive correlation between the board size and the size of the audit and compensation committees. The correlation between the board size and the 21

23 In Model II, we examine the effect of governance on the occurrence of backdating. We expect that backdating firms have inferior corporate governance structure relative to their peers. CEOs of backdating firms are likely to have more power, i.e. higher PC and are more likely to be the Chairman of the board. But, similar to the findings presented in Table VI, backdating firms are in fact comparable in CEO duality and better in G-index in relation to the control group. In untabulate results, we substitute the G-index with E-index and find quantitatively similar conclusion. Governance variables seem to have very little influence in explaining backdating. Model III regress the incidence of backdating onto CEO compensation components controlling for stock volatility. We hypotheses backdating firms are more likely to be paid in option and less in cash related components and results are generally in line with our expectation. Cash compensation has a negative coefficient while the value of option grants is positively related to backdating. Both are statistically significant. We combine all variables and estimate a grand regression in Model V. The results of Model V are by and large consistent with the models estimated earlier. Backdating firms have smaller subcommittee and more independent directors, but no coefficient is significant any more. In terms of governance, G-index is lower for backdators, so is the pay concentration. Backdating CEOs are granted more options but show no difference in cash-based compensation. [Insert Table VIII] 4.2 Robustness Checks Table IX reports the logistic regression results of sample firms and the unmatched group. We begin with Model I, which includes only firm characteristics. As expected, backdating firms have higher volatility in stock return and operating performance. Also, backdators are younger. Then, we include board structure and governance variables in Model IIa and Model IIb. However, none variables seem to have explanatory power, except that CEOs of backdating firms have less chance to sever as the Chair of Board. audit committee is 0.4, between board size and the compensation committee size it is 0.335, and between audit and the compensation committee size the correlation is All of these correlations are highly significant. 22

24 Model III regress the occurrence of backdating onto CEO compensation and results are very close to the ones in Table VIII. Cash-based compensation has a negative effect while the option-based compensation is positively related to backdating. We combine all variables together in Model IV and results are no different except that the percentage of option grants shows a positive influence. [Insert Table IX] 5. Concluding Remarks Backdating strike many investors confidence and cause large value losses to many involved companies. In this study, we examine the factors that contribute to backdating. Many concurrent studies, using different samples, show that backdating firms are weaker in their governance mechanism. We employ two different control groups, the unmatched and the matched group. And, our results show that other factors are equally important. We find that backdating firms are younger and more volatile in their operating performance, compared to all non-backdators. These firm features are significant in explaining the occurrence of backdating in a pooled regression and more powerful than the governance and compensation variables. However, after we match the sample firms with a propensity close control group, we find backdatings are no worse than their peers. Their governance, measured by G-index, is in fact better. Neither backdating CEO is more power in determining their compensation. Option-based incentive seems the reason which drives backdating. In both regressions, backdating firms are granted more options and less cash. In sum, our study suggests that backdating is not necessarily related to presumably poorer governance. Such result could be an outcome of mismatching or measurement error. 23