EQUITY-BASED COMPENSATION AND INSIDER TRADING HONGYAN FANG. A dissertation submitted in partial fulfillment of the requirements for the degree of

Transcription

1 EQUITY-BASED COMPENSATION AND INSIDER TRADING By HONGYAN FANG A dissertation submitted in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY WASHINGTON STATE UNIVERSITY College of Business December 2010 Copyright by HONGYAN FANG, 2010 All Rights reserved

2 Copyright by HONGYAN FANG, 2010 All Rights Reserved

3 To the faculty of Washington State University: The members of the Committee appointed to examine the dissertation of HONGYAN FANG find it satisfactory and recommend that it be accepted. David A. Whidbee, Ph. D., Chair John R. Nofsinger, Ph. D. Donna L. Paul, Ph. D. ii

4 Equity-based Compensation and Insider Trading Abstract By Hongyan Fang, Ph.D. Washington State University December 2010 Chair: David A. Whidbee My dissertation consists of two essays related to executive equity-based compensation and insider trading. In the first essay, I examine the economic rationale for a well-debated compensation practice: option backdating. Using a sample of backdating firms and matching firms, I provide strong evidence in support of incentive and retention-based explanations for backdating. Backdating firms tend to be younger, faster growing, and operating in a more competitive labor market where competition for highly-skilled talent is fierce. Further, rather than experiencing poor performance, backdating firms tend to outperform matching firms in both prior- and post-backdating years. These results suggest that backdating reflects a firm s demand for valuable employees rather than a manifestation of agency problems. In the second essay, I examine the information content of prior investment announcement insider trading and investigate whether insider trading serves as a signal of the quality of the project or firm prospects. The results suggest that net managers (or other insiders) purchasing portends higher announcement period returns. The information content of managers trades varies and the signal effect is stronger for those trades made by managers with relative higher iii

5 compensation risk. Further, although managers signal firm value consistently through their trading, the market appears not to revise its valuation until large corporate event such as capital investment occurs. iv

6 Abstract... iii Dedication... vii CHAPTER ONE: GENERAL INTRODUCTION... 1 CHAPTER TWO:THE ECONOMIC RATIONALE FOR OPTION BACKDATING: INCENTIVE-BASED EXPLANATIONS Introduction Hypothesis Economic determinants of backdating: recruiting new talent or retaining existing employees Managerial power explanation Other possible explanation: Financial Reporting Costs/Liquidity Constraints Data and methodology The determinants of backdating Univariate analysis Multivariate analysis Firm performance surrounding backdating years: further evidence on incentive effect of backdating Heterogeneity in the recipients of backdated options CEO grant and Non CEO grant Operating performance for CEO grant and Non-CEO grant Summary and conclusion References Appendix: Definition of variables used in this study v

7 CHAPTER THREE: Insider trading and market valuation of captial expenditure announcements Introduction Hypothesis and sample selection Hypothesis Sample selection and descriptive statistics Insider trading and announcement period returns Insider trading surrounding capital expenditure announcements Insider trading and announcement period returns Compensation risk, managerial trading and announcement period returns The market response: investment opportunities or correction of previous mispricing? Further analysis on market valuation around the announcements: correction of previous mispricing Multivariate regression: is trading signal a competing explanatory variable for the announcement period abnormal returns? Robustness checks Endogeneity of EBC The lag between insider trading and investors receiving the signal Conclusion References APPENDIX A: Methodology for Measuring Abnormal Returns and Test Statistics APPENDIX B: Principal Component Analysis of Investment Opportunity Set vi

8 Dedication This dissertation is dedicated to my beloved family: my mother, my father and my sister, without whose emotional support, I might be where I am now. vii

9 CHAPTER ONE: GENERAL INTRODUCTION My dissertation consists of two essays related to executive compensation and insider trading. The first essay addresses the underlying rationale for a well-debated compensation practice: option backdating. My interests in this topic arise from two competing, yet not mutually exclusive arguments related to managers incentive compensation such as option and stock equity awards. On one hand, incentive compensation is granted to reduce agency costs and align the interests of managers and shareholders. On the other hand, some compensation practices, such as option repricing, rewarding CEOs for general market returns, and option reloading, etc. are considered to reflect inadequate governance structures and agency problems within the firm. In my first essay, I conjecture that these two explanations might also apply for the option backing practice. Using a sample of backdating firms and matching firms, I provide strong evidence in support of incentive and retention-based explanations for backdating. Backdating firms tend to be younger, faster growing, and operating in a more competitive labor market where competition for highly-skilled talent is fierce. Further, rather than experiencing poor performance, backdating firms tend to outperform matching firms in both prior- and postbackdating years. These results suggest that backdating reflects a firm s demand for valuable employees rather than a manifestation of agency problems. In my second essay, I examine the information content of insider trading prior to firms capital expenditure announcements. By doing so, I also intend to explore whether the market price adjusts towards manager s valuation around important corporate events when information asymmetry alleviates. The research question is motivated by my interests in the difference in market valuation and insider valuation induced by information asymmetry. Corporate organization structure provides managers with relative information advantage, thus managers 1

10 possess better knowledge of the firm s true value compared to outside investors. However, investors are able toinfer from managers trading activity and corporate decisions the prospects of the firm. Empirically, I find in my study that prior announcement insider trading, and managers in particular, provides a signal of the quality of the project or firm value: managers (or other insiders ) net purchasing portends higher announcement period abnormal returns. The information content of managers trades varies and the signal effect is stronger for those trades made by managers with relative higher compensation risk. Further, although managers signal firm value consistently through their trading, the market appears not to revise its valuation until large corporate event such as capital investment occurs. 2

11 CHAPTER TWO: THE ECONOMIC RATIONALE FOR OPTION BACKDATING: INCENTIVE-BASED EXPLANATIONS 1. Introduction Option backdating is the practice of granting stock options that are dated prior to the date the company actually granted the option. Generally, public firms compensation plans call for stock options to be granted at an exercise price equivalent to the stock s fair market value on the date of the option grant. Nevertheless, some companies have been accused of backdating stock option grants to coincide with the dates of low closing prices for the company's stock, thereby resulting in grants of in-the-money options to managers or other employees. Backdating, by itself, is not illegal, but it certainly raises criticism and it can violate shareholder approved compensation policies and, if not accounted for appropriately, violate tax laws and accounting rules. Much of the academic literature on backdating focuses on agency theory explanations for the practice and the shareholder wealth loss typically experienced upon the revelation of backdating (e.g., Collins, Gong, and Li (2009), Bizjak, Lemmon, and Whitby (2009), Narayanan, Schipani, and Seyhun (2007), Bernile and Jarrell (2009)). More recently, researchers have begun to examine why firms engaged in this practice at the potential risk of adverse public disclosure and human capital loss. Gao and Mahmudi (2009) assert that backdating is the consequence of efficient contracting and helps to reduce firm compensation costs and increase managerial incentives. Armstrong and Larcker (2009) propose behavior and economic theories that might explain backdating behavior. Dhaliwal, Erickson, and Heitzman (2009) provide a tax-based incentive to explain backdating. Both agency and incentive based theories are potentially valid in 3

12 understanding the factors driving backdating and the two alternatives might not be mutually exclusive. In this study, we empirically test both the agency related and the incentive based explanations for backdating. According to the agency based explanation, backdating is a manifestation of agency problems and, therefore, will be more likely to occur in firms with weak governance structures. According to the incentive based explanation, firms are willing to risk the potential adverse consequences associated with backdating options in order to attract or retain high quality employees. Using a sample of 344 pair-matched backdating and non-backdating firm-year observations from the period, we find that both agency and incentive based explanations contribute to the incidence of backdating, with incentive-based determinants exhibiting more explanatory power than governance characteristics. Firms with a larger proportion of insiders sitting on the board and with longer-serving CEOs have higher probabilities of backdating. We do not find any evidence, however, that firms with CEOs also serving as chairman, insiders sitting on the compensation committee, high managerial ownership, or less blockholder ownership are more likely to backdate options. In addition to governance features, we propose other firm characteristics as proxies for incentive-based determinants of backdating and find that backdating firms tend to be younger, experiencing high growth (low book-to-market ratio), and operating in a more competitive labor market. These incentive-based firm characteristics add much explanatory power to the incidence of backdating. We further examine the economic implications of backdating by examining the operating performance and shareholder value changes around its incidence. We do not observe any deterioration of performance in firms engaged in backdating. Rather, we find that market adjusted stock returns for backdating firms are significantly higher than those for a set of 4

13 matching firms in the years surrounding backdating. Further, we find that the board and ownership variables typically used as proxies for managerial power also act as proxies for other characteristics that capture a firm s demand for highly-skilled workers. Using methodology similar to Core, Holthausen, and Larcker (1999), we examine the association between subsequent firm performance and the predicted probability of backdating arising from board and ownership characteristics for the 344 pair-matched firm years and find no significant relationship between them. This finding is inconsistent with backdating representing a manifestation of agency problems and suggests that board and ownership characteristics may also be capturing some of the incentive-based motives for backdating. 1 Finally, we examine whether there are differences between firms that backdate CEO options versus those that backdate options granted to other employees. As asserted by some backdating firms, many backdated options are granted to technology workers when they initially become employed. If agency problems are driving the incidence of option backdating, we would expect to see differences between firms that backdate options granted to the CEO relative to firms that backdate options granted to other employees. Among the total 437 backdated grants in our firm-year sample, we find 43.71% of the grants include CEOs while 56.29% are non-ceo grants (most commonly other executive grants). However, there is no significant difference in the documented firm characteristics between CEO backdating and non-ceo backdating, suggesting that the determinants of option backdating are similar for options granted to CEOs and options granted to others. In fact, the dummy variable proxy for CEO backdating is 1 We restrain our interpretations here to this specific compensation practice of backdating and to our specific sample of 344 pair-matched firm years. Board and ownership structure may measures of agency problems in other cases. For example, contrary to our results, Core, Holthausen, and Larcker (1999) show that firms with weak governance structure reward CEOs with greater compensation and the predicted component of compensation based on firms board and ownership structure is negatively related to firms subsequent operating and stock return performance. 5

14 significantly negative when it is included in our primary logit model, indicating that, at least in our sample, CEOs are less likely to obtain backdated option than non-ceos. The remainder of this paper is organized as follows. In Section 2, we review recent empirical studies on backdating, propose possible economic theories underlying backdating, and develop our hypotheses. Section 3 describes the sample selection process and Section 4 provides univariate and multivariate analysis of firms decisions to backdate. In section 5 we examine firm performance surrounding backdating. Section 6 analyzes differences between CEO and non- CEO backdating. And a summary and conclusion is provided in Section Hypotheses Yermack (1997) finds that the 50-day cumulative abnormal return following CEO option grant dates is over 2 percent, suggesting that firms time option awards to coincide with favorable corporate news announcement. Lie (2005) documents negative market adjusted abnormal returns before the granting of unscheduled executive options and positive abnormal returns afterwards. He concludes that unless executives have a superior ability to forecast future short-term marketwide movements, the results indicate that at least some of the official grant dates must have been set retroactively. Heron and Lie (2007) find that the return pattern is much weaker after August 29, 2002 when the Securities and Exchange Commission requires that option grants must be reported within 2 business days following the grant date instead of 45 days after the fiscal year end. The return pattern is still present, however, even over a two-day window. Much of the existing studies associate backdating with firms governance characteristics. According to this agency-based view, managers in firms with weaker governance structures exert great influence over the boards in negotiating compensation packages. Collins, Gong, and Li (2009) find that firms with a higher percentage of inside and gray directors, a higher percentage 6

15 of outside directors appointed by incumbent directors, and the CEO serving as chairman of the board are more likely to engage in backdating. In addition, the presence of outside blockholders on the compensation committee, higher CEO stock ownership, and a higher proportion of stock options among CEOs total compensation also increases the probability of backdating. Bizjak, Lemmon, and Whitby (2009) find that backdating tends to spread to firms with interlocking board members. Consistent with option backdating representing a manifestation of agency problems, Narayanan, Schipani, and Seyhun (2007) find that the 21-day cumulative abnormal return surrounding public disclosure of backdating is about -8 percent for a sample of 48 firms, which is about $500 million for each firm. Bernile and Jarrell (2009) also document negative abnormal returns around firms public involvement in backdating. However, they do not find any direct cash flow consequences of backdating as evidenced by analysts earnings forecasts. In contrast to agency-based theories, Gao and Mahmudi (2009) show that backdating may be an optimal way to incent CEOs when taking into consideration tax and accounting concerns. Empirically, they find that backdating is negatively related to CEOs total compensation and cash compensation, and positively related to pay-performance sensitivity, suggesting that backdating may reduce compensation costs and increase managerial incentives. Armstrong and Larcker (2009) provide several behavioral and economic theories for backdating behavior. In their analysis, directors of firms engaging in backdating tend to have fewer degrees of separation between them than directors of firms not engaging in backdating. In addition, they suggest that managers personal characteristics might lead to a false consensus that backdating proliferates and is a legitimate compensation practice. They also acknowledge that firms might backdate options to attract talent. However, they do not provide comprehensive analysis of these potential behavioral and economic explanations. 7

16 2.1 Economic determinants of backdating: recruiting new talent or retaining existing employees Absent from much of the literature on backdating is an acknowledgment that many firms are in fierce competition for highly skilled employees. Much of the documented backdating occurred before 2002 when high-tech and internet-based firms were in their high-growth years and seeking talented workers in a competitive labor market. Setting option exercise prices at, say, monthly lows may have been a useful tool in recruitment and retention negotiation. During the 1990s, when financial markets were experiencing substantial returns, new hires at some firms complained that the firm s rapidly increasing stock price provided employees that were hired just days prior were receiving substantial rewards that the new employees were not receiving. According to anecdotal evidence, Silicon Valley companies would, in response to these complaints, set the actual grant date back by a week or two to give new employees the same opportunity for financial gains. 2 Some have suggested that the intense battle for talent among rapidly growing firms led to this practice being extended to existing employees and then up the corporate ladder. When Microsoft stopped pricing options at monthly lows, it faced skepticism whether this would reduce its ability to attract skilled talents. 3 Therefore, we suggest that option backdating may be a response to labor market conditions. Current literature and media reports tend to focus on the legal and ethical issues surrounding backdating and ignore the potential effectiveness of backdating as a tool to attract or retain talented employees. 4 Based on labor market considerations, if backdating is an effective tool for employee attraction or retention purposes, we would expect to observe that high growth firms, young firms, and firms with uncertain futures are more likely to engage in backdating. 2 Inquiry into Stock Option Pricing Casts a Wide Net, The New York Times, June 19, Dates from hell; Executive share options, Economist, July 22, A related possibility is that, in some industries, backdating may have become so common that firms were at a competitive disadvantage if they did not engage in the practice. 8

17 High growth firms and young firms are more likely to face recruitment and retention problems due to their earlier stage of development. In addition, young firms are less likely to have well developed succession plans, which renders turnover more costly for them (Chidambaran & Prabhala, 2003). Finally, stocks with higher return volatility have a greater potential for gains from option, thus providing more incentive to attract and retain employees. Therefore, departure from the normal option granting process does not necessarily indicate departure from optimal contracting and incentive dis-alignment. Rather, it might serve as a tool to hire and retain talented employees and align the interests of those employees with shareholders. 5 Although studies on backdating that document retention- or attraction-based motives are scarce, the literature on other compensation practices such as option repricing find that repricing serves to realign managerial incentives (Coles, Daniel, and Naveen (2005)). Carter and Lynch (2004) show that firms reprice stock options to retain key employees and repricing helps to reduce executive and other employee turnover. Chidambaran and Prabhala (2003) find that repricing firms tend to be younger, smaller and rapidly growth firms. Following these studies, we use B/M, book-to-market ratio, and R&D, firms research and development expenditures scaled by book value of assets, to proxy for firms growth prospects. 6 Both of these ratios are calculated using data from CRSP-Compustat merged data set (henceforward called CCM) in the fiscal year prior to the option grant year. Firms with lower B/M and higher R&D ratios are expected to have a higher probability of backdating. Following Chidambaran and Prabhala (2003), we define Firm age as the number of years since the first price for the firm is reported in the CRSP tapes. We compute volatility or uncertainty as the standard deviation of daily 5 For the relation between firm risk and certain compensation practice, see Bettis, Bizjak, and Lemmon (2001), Bizjak, Lemmon and Whitby (2009), Collins, Gong and Li (2009). For the relation between compensation level and firm risk, see Core Holthausen, and Larcker (1999), Cyert, Kang, Kumar, and Shah (1997). 6 Core, Holthausen, and Larcker (1999) proxy for firms investment opportunity set with the market-to-book ratio. Also see Coles, Daniel, and Naveen (2005). 9

18 common stock returns from CCM during the fiscal year prior to option grant year. Based on the previous argument, we expect a positive relation between backdating incidence and volatility. 2.2 Managerial power explanation Other than any incentive-based explanations, backdating might simply be a manifestation of agency problems facilitated by lax governance. Bebchuk, Fried, and Walker (2002) argue that some compensation practices, such as option repricing, use of at-the-money options, and pay for general market trends other than relative stock price performance are difficult to understand within the optimal contracting framework. Based on the managerial power argument, compensation schemes approved by directors deviate from optimal contracting theory because directors are captured or subject to influence by managers such that they are no longer at arm s length and able to make decisions that maximize shareholder value. Backdating might be another mechanism for the exploitation of managerial power. 7 According to this argument, backdating is more likely to occur in firms with weak governance. 8 Collins, Gong, and Li (2007) and Bizjak, Lemmon, and Whitby (2009) find evidence of an association between corporate governance and backdating. Based on their study, we use Lexis-Nexis to collect a set of board composition and ownership structure variables from firms proxy statements in the fiscal year prior to option grant years. We include board size (the total number of board members), inside director percentage (the percentage of inside directors relative 7 Narayanan, Schipani and Seyhun (2007) estimate that the potential gain for executives in backdating firms is $600,000 on average. 8 Core, Holthausen, and Larcker (1999) find that board and ownership structure accounts for a significant amount of variations of CEO compensation. Firms with a larger board and a board with a greater percentage of outside directors and gray directors have higher CEO compensation. Greater CEO ownership and the existence of blockholder tend to reduce CEO compensation. Other relating literatures on executive compensation and governance structure see Lambert, Larcker, and Weigelt (1993), Hallock (1997), etc. 10

19 as a percent of the total number of directors) 9, a dummy variable, CEO-Chair, to indicate whether the CEO also serves chairman of the board, and a dummy variable, Compensation conflict, indicating whether firm insiders sit on the compensation committee. In addition, we add CEO tenure as a measure of potential entrenchment. To account for differences in ownership structure, we include O&D ownership (the percentage equity ownership in the firm by officers and directors) and Blockholder ownership (the percentage of external blockholders ownership). We define external blockholders as those with over 5% stock ownership that are not related to the firm in any other way. 10 Smaller boards, having a higher proportion of insider directors, the CEO serving as chairman, and insiders sitting on compensation committees are characteristics of weak boards that are more likely to breed managerial power. We expect officer and director ownership and CEO tenure to be positively correlated with backdating since greater values might indicate deeper managerial entrenchment. We include percentage of blockholder ownership because previous literature suggests that external blockholder ownership has an effect on CEO compensation Other possible explanation: Financial Reporting Costs/Liquidity Constraints Backdating provides a covert method of granting at-the-money options while options with exercise prices below market prices are actually in-the-money. On one hand, some firms may backdate options to maximize executives option value for retention purpose or to attract new talent; on the other hand, granting in-the-money options incurs compensation expenses that reduce firms reported net income. The Accounting Principles Board (APB) Opinion 25 requires 9 Insiders are executive officers of the firm. We also include gray directors in our definition of Inside directors. A gray director is a director who is a family member of an executive officer, a former employee of the firm, an employee of a significant service provider, a supplier, or a customer. 10 The relationship includes: the blockholder is a family member of an officer or a director of the firm; the blockholder has business association with the firm; the blockholder is a former employee of the firm. 11 Collins, Gong and Li (2009) show that firms with an outsider blockholder sitting on the compensation committee are less likely to backdate options. Also see Core, Holthausen, and Larcker (1999) on relationship between blockholder and CEO compensation. 11

20 that the difference between grant date stock price and the exercise price (called the intrinsic value of options) be treated as an expense and deducted from income. The expense is zero if firms grant at-the-money options. 12 Thus, to minimize the stated cost of compensation and avoid reduction in reported net income, firms have the incentive to offer at-the-money options by falsifying option grant dates so that they correspond to dates with lower stock prices. This is more likely to occur when firms already face large financial reporting costs and liquidity constraints. The liquidity and financial reporting cost concerns in implementing certain compensation practices have been well documented (Matsunaga, Shevlin, and Shores (1992), Yermack (1995), Coles, Daniel, and Naveen (2005), etc.). Following Matsunaga, Shevlin, and Shores (1992) and Yermack (1995), we use Interest coverage, defined as income before extraordinary items plus interest expense divided by interest expense in the fiscal year prior to option grant year, to proxy for financing reporting costs and liquidity constraints. Lower interest coverage indicates higher reporting costs and higher liquidity constraints because firms with lower interest coverage have lower profitability relative to their debt obligations. We expect firms with lower interest coverage to be more likely to engage in backdating because expensing compensation costs will further reduce their interest coverage ratios Data and Methodology We use a sample of firms identified by the Wall Street Journal as being implicated in option backdating. The WSJ periodically updates the list of firms that have come under scrutiny for stock-option grants and the latest scorecard was updated in September Listed firms are generally implicated in at least one of the following ways: (1) The firm is under SEC formal or 12 Accounting rules of SFAS 123R, effective starting December 15, 2005, mandates expensing fair market value of the options at the time of the grant. However, because our sample period ends in 2004, we still adopt APB 25 as guidelines for our discussion. 13 However, under the guideline of APB 25, if backdating is discovered, firms must restate financial statements to reflect compensation expenses of the option intrinsic value. 12

21 informal investigation; (2) The firm is being investigated by the Justice Department; (3) Executives or directors step down from their position when backdating is disclosed; and (4) The firm announces the need to restate financial statements or take a charge because of incorrect option grants. As of September, 2007, 141 firms have been implicated in one or more of the four practices. Though these 141 firms are reported as being implicated in backdating practices, it is unlikely that all of the firms are actually engaged in backdating. Some firms are investigated by the SEC or the Justice Department with no punitive action being taken. In addition, the WSJ does not provide complete information about the specific grants that are backdated. To obtain information on the implicated grants, we use Thomson Financial s Insider Filing Database. Table 2 of the Thomson database provides data on insider derivative transactions as reported on SEC forms 3, 4, 5, and 144. We use the following filters to identify backdated grants for the total option awards in the insider database: First, we require the derivative type to be options (CALL or OPTNS), non-qualified options (NONQ), employee options (EMPO), or director options (DIRO or DIREO). Unlike most previous studies (Bizjak, Lemmon, and Whitby (2009)) (Collins, Gong, and Li (2009), Cai (2007)), we do not constrain our analysis to high level executives because the employee recruitment/retention argument applies to all employees. Next, we use transaction codes A (grant or award transaction pursuant to Rule 16b-3(c)), T (acquisition or disposition transaction under an employee benefit plan other than pursuant to Rule 16b-3) and J (Other acquisition or disposition). To maintain data quality, we only keep records with cleanse codes R (data verified through the cleansing process), H (cleansed with a verity high level of confidence) and C (corresponding record added). We then eliminate duplicate grants with the same document control number, transaction date, and expiration date, and those grants with no 13

22 CUSIP number reported. These procedures leave us with 398,678 option grants and 80,146 unique firm-grant-date observations that have stock price information from CRSP on the grant date. From the 80,146 firm-grant-date observations, we are able to find 2,143 firm-grant-date observations for 120 of the initial 141 listed firms implicated in backdating. We then identify backdating by observing stock performance surrounding the option grant dates. Detecting opportunistic timing of option grants generally falls into two categories: investigating pre- and post- option grant date stock return patterns (Lie (2005), Heron and Lie (2007), Narayanan and Seyhun (2008), Bizjak, Lemmon, and Whitby (2009)), or investigating the ranking of the stock price on the date of option grant relative to the distribution of stock prices during a specified window of time (Collins, Gong, and Li (2009), Bebchuk, Grinstein, and Peyer (2007)). 14 Based on these studies, we collect stock prices on the [-20, 20] days relative to the option grant date from CRSP for the 2,143 firm-grant-date observations. For each of the observations, we form the 41-day stock price into five groups and observe the quintile ranking of stock price on the option grant date. We identify an option as backdated if the stock price on the grant date falls into the lowest quintiles. We use the price-ranking methodology because Bebchuk, Grinstein, and Peyer (2007) show that the price ranking methodology is useful to detect backdating when stock price is relatively stable during the month. The 41-day window is chosen based on Bizjak, Lemmon, and Whitby (2009). 15 This procedure leaves us with Lie (2005), Heron and Lie (2007) calculate abnormal returns as the difference between raw stock returns of the granting firm and the predicted returns based on the Fama and French (1993) three-factor model. Bizjak, Lemmon and Whitby (2009) detect backdated grants if the prior and post-grant 41-day cumulative abnormal returns exceed a predicted cut-off level based on randomly selected grant dates. Collins, Gong and Li (2009) identify backdated grants if the stock price on the grant date falls into the lowest decile of the firm s stock price distribution over a 240- day window. Bebchuk, Grinstein and Peyer (2007) investigate the price ranking during the calendar month of the grant date. In their sample, around 30% of options were granted at the lowest three prices during the month. 15 We use the 41-day period rather than 240-day window used by Collins, Gong, and Li (2009) because a common practice of backdating firms is to backdate options to the monthly low. 14

23 backdated firm-grant-date observations which include 446 unique firm-year observations for 117 firms of our initial 141 firms. 16 We conduct our analysis at the firm-year level rather than at the individual employee or grant level. Ideally, we would analyze the incidence of backdating for individual employees, but that would require detailed information about personal characteristics of individual recipients such as their stock ownership, age, whether newly employed, etc. Unfortunately, this information is not available for any but the most senior executives. 17 Therefore, we resort to using available information and conduct the analysis at the firm-year level. Further, we focus on firm-year level data rather than firm level data because our analysis indicates that many of the 117 firms backdate options in more than one year. Selecting a single year for each firm might obscure some important time-series changes in firm characteristics, such as changes in firms governance structure, operating performance and recruiting circumstances, etc. These changes may consequently affect firms compensation decisions. For example, a firm s recruiting of new employees might concentrate in certain years when the firm is growing rapidly and then stop in subsequent years. In this case, either randomly selecting the firm in a certain year or selecting the first year when backdating occurs to form firm-level observations will result in an incomplete understanding of motives for backdating. Table 1 reports the year and industry distribution of the 446 backdating firm-year observations. The dating game becomes intensive in the late 1990s when technical firms are booming, with the highest frequency occuring in years 2001 and It declines to 5.38% in Collins, Gong and Li (2009) also find that backdating events decline in year We also delete observations when the firm has a SIC of or 49 in the grant year at this stage. 17 For example, backdated option recipients may include technology workers or non-high-ranking executives. Unfortunately, information concerning most of the technology workers and non-high-ranking executives is not available. Thus, we rely on firm-year level observations and use proxies for firm characteristics in a certain year to investigate the economic rationale for backdating. 15

24 Backdating is more likely to occur in industries with SIC code 36 (electronic and other electric equipment) and 73 (business service), with total yearly grants accounting for 22.42% and 23.99% of all observations. 16

25 Table 1 Summary statistics of 446 backdated firm-year observations This table reports the year and two-digit SIC industry distributions of the 446 backdating firm-year observations during the year A firm-year is identified as backdating if backdating occurs in the firm in the certain year. Panel A: Year distribution of firm-year observations Year Number of observations Frequency % % % % % % % % % % % % Total % Panel B: Industry distribution of firm-year observations Number of Frequency SIC code Industry Name observations 13 Oil & gas extraction % 15 General building contractors % 20 Food & kindred products % 28 Chemical & allied products % 32 Stone, clay, & glass products % 33 Primary metal industries % 34 Fabricated metal products % 35 Industrial machinery & equipment % 36 Electronic & other electric equipment % 38 Instruments & related products % 48 Communications % 50 Wholesale trade-durable goods % 51 Wholesale trade-non-durable goods % 52 Building materials & gardening supplies % 53 General merchandise stores % 56 Apparel & accessory stores % 57 Furniture & home furnishing stores % 58 Eating & drinking places % 59 Miscellaneous retail % 73 Business services % 78 Motion pictures mailing list % 80 Health services % 82 Educational services % 87 Engineering & management services % total % 17

26 We construct a matching group of 446 non-backdated option grants firm-years. The 446 matching firm-years are selected by the following procedures: first, we select from the rest of the 78,003 (=80,146-2,143) non-backdated firm-grant-date observations those with stock price on grant date falling into the highest 3 quintiles of the 41-day window price ranking. We double check these grants to make sure they are not granted by the 141 firms listed on WSJ; next, we match firm characteristics of non-backdated grants with firm characteristics of sample grants based on two-digit SIC code and firm size. For each of the 446 identified backdated firm-year observation, we identify book value of asset in the fiscal year prior to the backdating year, and find in the 78,003 non-backdated firm-grant-date observations those firms with book value of asset between 70% and 130% of backdating firms. We also require that matching non-backdated grants occur in the same fiscal year as our sample firm-year. If no match is found, we relax the last constraint by matching non-backdated grant that occurs most closely to the sample grant year. Following these procedures, we identify 446 control firm-year observations. For both sample and matching firm-years, we collect governance and ownership data from firms proxy statements in the year prior to the fiscal year when options are awarded. We obtain accounting and stock price data from CCM. We encounter a large number of missing values because: first, we are unable to find the firm s proxy statement in a certain year, hence we could not obtain the governance variables; second, CCM reports missing values for some accounting data. Missing values might cause asymmetric matching even though we start with 446 pair-wise matches. To handle this problem, we obtain sample firm-years with a complete set of variables that has nomissing values. We are able to identify 372 sample firm-years with complete information. For these 372 observations, we do a re-matching based on the previous criteria from the 446 matching firm-year observations that also have complete information. And this re-matching 18

27 procedure leaves with us 344 pairs that do not have any missing values for the variables in our later analysis. These observations cover the period and the main variables for our study are defined in the Appendix. 4. The determinants of backdating 4.1 Univariate analysis Table 2 reports univariate analysis of the differences between the 344 pairs-matched sample and matching firm-year observations. Consistent with backdating being the result of managerial power, backdating firms tend to have longer CEO tenure, more insider directors, and greater share ownership by officers and directors. Interestingly, board size is smaller in the backdating sample suggesting that when the board size is small, managers may have more influence on compensation contracts. The mean values for CEO tenure, board size, the proportion of insider directors, and share ownership by officers and directors for backdating firms are 9.451, 7.201, 0.349, and respectively. The values for non-backdating matching firms are 7.951, 8.177, 0.284, and Inconsistent with managerial power based explanations, however, backdating firms tend to have CEOs that also serve as the chairman of the board approximately as frequently as the non-backdating matching firms and there is no significant difference in outside blockholder ownership between backdating firms and matching firms. 19

28 Table 2 Descriptive statistics of backdating: firm-year observations This table reports mean and median various characteristics for the paired matches of 334 sample firmyear observations and 334 matching firm-year observations. The data covers the time period Sample firm-year is identified as a firm granting backdated option in a certain year. Matching firmyears are non-backdating firm-year observations that are chosen based on book value of assets (70%- 130%) and two-digit SIC code of sample firm-year observations in the fiscal year prior to backdating. t denote the fiscal year in which options are granted and t-1 is the prior fiscal year. All the variables are defined in Appendix. t statistics test the mean differences in the matched pairs of observations. zstatistics test the equality of distribution for matched pairs of observations using the Wilxocon signedranks test. The p-values for t-statistics and z-statistics are reported in parentheses. ***, **, * denote statistical significance at less than 1%, 5% and 10% levels, respectively. Variable Mean Median Sample Matching t-statistics Sample Matching z-statistics firm-year firm-year firm-year firm-year Firm size t (0.3941) (0.4534) O&D ownership t *** (0.0041) *** (<0.0001) Blockholder ownership t (0.4708) (0.7538) Board size t *** (<0.0001) *** (<0.0001) Insider director t *** (<0.0001) *** (<0.0001) CEO-Chair t (0.7022) (0.7021) CEO tenure t *** (0.0070) *** (<0.0001) Compensation conflict t (0.1087) (0.1087) Volatility t *** (<0.0001) *** (<0.0001) Firm age t *** (<0.0001) *** (<0.0001) (B/M) t *** (<0.0001) *** (<0.0001) R&D t *** (<0.0001) *** (<0.0001) Interest coverage t *** (<0.0001) (0.2746) N

29 Although some of the board and ownership structure characteristics are consistent with backdating representing a manifestation of agency problems, Table 2 also provides some evidence that incentive-based explanations may also play a role in explaining backdating. Backdating firms tend to be younger and have higher growth. The mean firm age is years for the backdating sample and years for the matching sample, consistent with backdating firms being in an earlier stage of development. The B/M ratio and R&D ratio is and for backdating sample and and for the matching sample. The evidence implies that backdating firms high valuation levels and potential growth may encourage potentially unethical behavior that is intended to either retain valuable employees or to attract new talent. This is also consistent with Chidamdaran and Prabhala s (2003) evidence that option repricing tends to occur in smaller, younger and rapidly growth firms. Carter and Lynch (2004) also show that firms reprice options to retain valuable employees. Overall, this evidence suggests that repricing is adopted to restore employee incentives by lowering the exercise price. Backdating may be a similar mechanism for retention and/or recruitment purposes. Consistent with Collins, Gong, and Li (2009) and Bizjak, Lemmon, and Whitby (2009), we find that backdating firms tend to have more volatile equity prices. Table 2 indicates that backdating firms have a mean equity volatility of while non-backdating firms have a mean of Higher volatility brings about greater potential gains by backdating. As some media reports suggest, backdating options to new employees may be firms responses to relatively new employees complaints about their inability to participate in the wealth gains that longer-tenure employees enjoy. Providing beneficial compensation through backdating may also be an effort to mitigate the risks associated with higher volatility. 21

30 We do not find any strong evidence that firms backdate when they have higher reporting costs, or liquidity constraints. The mean interest coverage for the backdating sample is higher than the matching sample while the median interest coverage shows no significant difference between the two groups. 4.2 Multivariate regression We use multivariate logit regressions to simultaneously assess the relative importance of firm characteristics in explaining backdating events. The dependent variable is 1 if the firm backdates options in a certain firm year, and 0 otherwise. To better differentiate the governance structure and incentive-based explanations, we first run the regression on the set of governance variables. We then run a regression on a set of variables that proxy for incentive-based factors. Last, we include both governance structure and incentive-based variables to provide evidence on the relative explanatory strength of each set of variables. We include Interest coverage in all the models. For each model, we report both maximum likelihood estimates and marginal effects. The marginal effect measures the instantaneous effect that a one standard deviation change in a continuous explanatory variable or a change from 0 to 1 in a dummy explanatory variable has on the predicted probability of backdating with all other explanatory variables held constant. Column 1 (maximum likelihood estimates) and column 2 (marginal effects) in Table 3 reports results from the first regression. Both maximum likelihood estimates and marginal effects of the logistic regression have the expected sign, as suggested by existing studies, but are largely insignificant except for the proportion of insiders on the board and board size. The likelihood of firms backdating is positively related to the proportion of inside directors and negatively associated with board size. Contrary to the results of our univariate analysis, officer and director ownership and CEO tenure have no effect on the likelihood of backdating. 22

31 The results from the first model provide only moderate evidence that weak governance structure is responsible for backdating behavior. Consistent with these results, Gao and Mahmudi (2009) find that firms implicated in backdating exhibit better governance structures than their non-backdating counterparts. The Interest coverage ratio is positively associated with the likelihood of backdating, indicating that firms with higher reporting costs are less likely to backdate options. At the same time, at least for our sample firms, backdating is not a reaction to liquidity constraints. A similar approach by Wu (2008) using cash constraints, as defined by cash holdings subtracting interest expenses, shows only weak evidence of the financial constraint hypothesis of backdating. 23

32 Table 3 Logistic regression estimates of the likelihood of backdating: firm-year observations This table reports the maximum-likelihood estimates and marginal effect from binomial logit model regressions of the determinants for option backdating for the 334 matched pairs of firm-year observations. Dependent variable is 1 if the firm-year is identified as backdated firm year and is 0 otherwise. The data covers the time period All the variables are defined in Appendix. Model 1 identifies the board and ownership structure characteristics associated with the incidence of backdating. Model 2 identifies incentive characteristics associated with the incidence of backdating. Model 3 combine all the characteristics in the regression. t denote the fiscal year in which options are granted and t-1 is the prior fiscal year. P-values of parameter estimates are reported in parenthesis. ***, **, * denote statistical significance at less than 1%, 5% and 10% levels, respectively. Variable Regression Model Model (1) Model (2) Model (3) ML estimator Marginal effect ML estimator Marginal effect ML estimator Marginal effect Intercept (0.5928) *** (0.0037) (0.1979) *** (0.0002) (0.5678) *** (0.0050) O&D ownership t (0.5116) (0.4963) (0.7114) (0.8553) Blockholder ownership t (0.6249) (0.6077) (0.7260) (0.8204) Board size t *** (<0.0001) *** (<0.0001) (0.1907) (0.1549) Insider director t *** (<0.0001) *** (<0.0001) *** (0.0022) *** (0.0010) CEO-Chair t (0.9226) (0.9624) (0.6926) (0.6537) CEO tenure t (0.1515) (0.1477) ** (0.0326) ** (0.0458) Compensation conflict t (0.8528) (0.8359) (0.8800) (0.9107) Volatility t ** (0.0307) ** (0.0310) * (0.0599) ** (0.0515) Firm age t *** (<0.0001) *** (<0.0001) *** (<0.0001) *** (<0.0001) (B/M) t *** (<0.0001) *** (<0.0001) *** (<0.0001) *** (<0.0001) R&D t (0.3434) (0.2923) (0.2872) (0.2435) Interest coverage t *** (0.0008) *** (0.0004) *** (0.0021) *** (0.0013) *** (0.0040) *** (0.0026) N Industry dummies Yes Yes Yes Yes Yes Yes Year dummies Yes Yes Yes Yes Yes Yes Chi-square Probability of chi-square <.0001 <.0001 <.0001 Pseudo R-square