|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 TWO ESSAYS ON SHARE REPURCHASES By ALICE BONAIM A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2008
2 2008 Alice Bonaim
3 To my husband, who never doubted me
4 ACKNOWLEDGMENTS I thank all th e faculty members of the Departme nt of Finance at the University of Florida for their support and guidance. In particular, I am indebted to Mike Ryngaert for his patience and for challenging me to think fo r myself. I am grateful for the encouragement and helpful comments of David Brown and for the honesty and invaluable insights of Jason Karceski. I also thank Sarah Hamersma of the Department of Ec onomics for her contributions to my education, both in and outside of the classroom. Last but no t least, I am grateful for the support of the current and former finance PhD students, who helped to make this sometimes grueling process enjoyable.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ............................................................................................................... 4LIST OF TABLES ...........................................................................................................................7LIST OF FIGURES .........................................................................................................................8ABSTRACT ...................................................................................................................... ...............9 CHAPTER 1 REPURCHASES, REPUTATION, AND RETURNS ........................................................... 11Introduction .................................................................................................................. ...........11Hypothesis Development ........................................................................................................ 16Management Behavior afte r Prior Share Repurchase .....................................................16Other Credibility Factors .................................................................................................18Outside Signals ................................................................................................................19Long-Run Price Effects ...................................................................................................20Data and Sample .....................................................................................................................21Sample Selection .............................................................................................................21Proxies for Reputation a nd Signaling Credibility ........................................................... 22Returns Calculations ........................................................................................................26Control Variables .............................................................................................................27Results .....................................................................................................................................29Prior Repurchasing Behavior as a Predic tor of Future Repurchasing Behavior .............29Announcement Returns ...................................................................................................31Post-Announcement Returns ...........................................................................................33Conclusions .............................................................................................................................362 INSIDER TRADING AND SHARE RE PURCHASES: DO INSIDERS AND FIRMS TRADE IN THE SAME DIRECTION? ................................................................................. 46Introduction .................................................................................................................. ...........46Sample and Variable Construction .........................................................................................50Repurchase/Insider Trading Frequencies ...............................................................................53Determinants of Repurchasing/Insider Trading Frequencies ................................................. 54Univariate Tests ...............................................................................................................55Logit Regressions ............................................................................................................57Robustness of Results to Definition of Repurchasing Firm ............................................ 60Abnormal repurchases ..............................................................................................60Repurchase announcements ..................................................................................... 61Long Run Stock Returns .........................................................................................................62Methodology ................................................................................................................... .62Abnormal Returns Results ............................................................................................... 63
6 Post-Announcement Abnormal Returns .......................................................................... 64Interaction Regressions ................................................................................................... 66Conclusion .................................................................................................................... ..........67APPENDIX A VARIABLE DEFINITIONS ..................................................................................................90B LONG-RUN BUY-AND-HOLD R ETURNS CALCULATIONS ......................................... 94REFERENCES .................................................................................................................... ..........96BIOGRAPHICAL SKETCH .......................................................................................................100
7 LIST OF TABLES Table page 1-1 Summary Statistics. ....................................................................................................... .....401-2 Difference in means of percent comp leted by lagged percent completed group.. ............. 411-3 Repurchases regressions.. ................................................................................................. .421-4 Announcement returns. ..................................................................................................... .431-5 Long-run returns. ......................................................................................................... ......441-6 Univariate results for long-run returns. .............................................................................. 45A-1 Repurchasing variables .................................................................................................... ..90A-2 Returns variables ......................................................................................................... .......91A-3 Firm-specific characteristics ............................................................................................. .92A-4 Dividend announcement dummies a nd earnings announcement dummies .......................93
8 LIST OF FIGURES Figure page 1-1 Timeline of repurchase announcements. ............................................................................381-2 Percent completed by lagged percent completed. .............................................................. 392-1 Insider trading and re purchase frequencies. .......................................................................702-2 Repurchases and insider trading by si ze, market-to-book, and option exercise. ............... 722-3 Corporate governance. ..................................................................................................... ..762-4 Logit models. ............................................................................................................. ........772-5 Abnormally high repurch asing logit models. ..................................................................... 812-6 Repurchase announcement logit models. ........................................................................... 852-7 Abnormal returns. ..............................................................................................................862-8 Post-announcement abnormal returns. ............................................................................... 882-9 Interaction regressions. .................................................................................................. ....89
9 Abstract of Dissertation Presen ted to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy TWO ESSAYS ON SHARE REPURCHASES By Alice Bonaim December 2008 Chair: Mike Ryngaert Major: Business Administration Though open market repurchase announcements are generally viewed as positive signals and are associated with positive abnormal re turns, they are not binding commitments and investors are not guaranteed positive returns. My first study examined whether the market incorporates a firms reputation when evaluatin g the credibility of an announcement to buy back stock. Empirical results support th e theory that announcements made by firms that consistently meet or beat analysts expect ations are viewed as more cr edible. While short-run buyback announcement returns are unrelated to prior co mpletion rates, high prior repurchase plan completion rates are associated with greater abnormal long-run returns. For the subset of firms in the lowest quintile of returns after the prior announcement, I identified three-year cumulative abnormal buy-and-hold returns of 29.8% on average for firms whose prior completion rates were high. My second study examined the general di rection of insider transaction around share repurchase and the relationship between inside r trading and long-run returns after share repurchase. Net insider buying and net insider selling are observed more frequently in quarters during which firms are repurchasing non-trivia l amounts of stock than in non-repurchasing quarters, and the probability of simultaneous insider sales and company-level buybacks is greatest for firms with higher levels of option exercise, higher industryadjusted market-to-book
10 ratios, and weaker shareholder ri ghts. These results suggest that in certain cases managers may use share repurchases to advan ce their own agendas, as opposed to shareholder agendas. This study also investigates long run abnormal returns for repurchasing firms by direction of net insider trades. The findings are consistent with insider trades either validating or negating the undervaluation signal of the share repurchase: Re purchasing firms with concurrent net insider buying experience average buy and hold abnor mal returns of 4.95, 2.13, 1.72 percent during years one, two, and three, respec tively, after the repur chasing quarter compared to 0.41, -0.88, 1.85 percent for repurchasing firms with simultaneous net insider selling.
11 CHAPTER 1 REPURCHASES, REPUTATION, AND RETURNS Introduction Firms buy back their ow n stock for a multitude of reasons: to increase their leverage ratio (Bagwell and Shoven (1988)), to fund employee and management stock options (Fenn and Liang (2001), Kahle (2002)), to fend of f unwanted takeover attempts (Bagwell (1991), Billett and Xue (2007)), to mimic competitors (Massa, Rehman, and Vermaelen (2007)), to distribute excess cash flow (Jensen (1986)), and to signal undervaluation due to as ymmetric information between managers and shareholders (Stephens and We isbach (1998), Vermaelen (1981)). While these motivations are not mutually exclusive, some ar e arguably more important to management and to investors than others. Grullon and Michaely (2004) identify reducing agency problems associated with free cash flow and signaling under valuation as the two major theoretical reasons for buying back stock, and Brav, Graham, Harv ey, and Michaely (2005) provide empirical evidence in support of these motivations.1 Though announcements of open market share repurchase programs are not firm commitments to buy back stock, they are genera lly viewed favorably by the market (Stephens and Weisbach (1998), Jagannathan and Stephens (2003), Chan, Ik enberry, and Lee (2004), Chan, 1 In their survey of financial executives, the authors found that 86.4 percent of managers consider the market price of their stock (if the stock is a good investment, relative to its true value) as important or very important to the decision of whether or not to repurchase, and 84.5 percen t agree or strongly ag ree that repurchase decisions convey information about their company to investors. These percentages are higher than any other survey rationale cited by managers and support the notion that managers use repu rchases to signal undervaluation. The two next highest scores provide evidence in support of the free cash flow hypothesis: 80.3 percent of respondents consider the availability of good investment opportunities important or very important to their repurchasing decision, and 78.8 percent agree or strongly agree that re purchasing decisions are made after investment plans are determined. If good investments are nonexistent, then cash should be removed from the control of managers. The level of cash on hand also influenced repurchasing decisions: 61.9 percent believe that having extra cash or liquid assets, relative to their desired cash holdings, is important or very important to the companys repurc hase decision. This falls in 8th place behind the four mentioned above, merger and acquisition activity (#5, 72.3%), the stability of future earnings (#6, 65.5%), and a sustainable change in earnings (#7, 65.2%).
12 Ikenberry, Lee, and Wang (2006)). If cash di stribution and signaling undervaluation are the primary motives for announcing a repurchase, then announcement returns should be positively related to the amount of stock that the firm is likely to repurch ase and to the extent to which the firms stock is perceived to be undervalued. Cons istent with this conjecture, previous studies have found that share repurchase announcemen t returns are positive and increasing in the announced fraction of firm shares to be repurch ased (Comment and Jarrell (1991), Raad and Wu (1995)). Additionally, Stephen and Weisbach (1 998) and Chan, Ikenberry, and Lee (2004) find that announcement returns are generally higher for firms with low prior re turns, a common proxy for undervaluation. This paper asks a broader question: To what ex tent do capital market participants consider a firms past behavior when assessing the validity of a current announcement to buy back a specified number of shares? Since it is a nonbinding commitment the announcement of an open market share repurchase has been labeled by some as cheap talk (Bhattacharya and Dittmar (2004), Almazan, Banerji, and De Motta (2008)). A fi rms managers may engage in cheap talk in order to temporarily inflate the price of the st ock for their personal benefit. For example, if insiders are planning to sell stock in their personal accounts, they will want to obtain the highest possible price. Since the announcement of an open market share repurchase is generally associated with positive returns, management could easily announce a buy back then sell their shares at the new inflated price. Another motivat ion for cheap talk would be to avoid a takeover bid or proxy contest by temporarily increasing the stock price. The cheap talk problem may be alleviated if investors can ex ante identify which firms are likely to follow through on repurchases in a fashion consistent with th eir stock being underva lued. Hence, market
13 participants will want to assess managements beha vior in the past, especially with regards to past repurchase programs, when resp onding to a new repurchase announcement. A related literature has found that reputation is used to certify the accuracy of new information: investors are more responsive to forecasts provided by managers and analysts who have provided accurate predictions in the past. Hutton and Stocken (2007) find that when firms have a reputation for precise forecasting, the in formation content of voluntary disclosures is impounded in stock prices sooner. In a simila r study, Ng, Tuna, and Verdi (2008) identify forecasting credibility as a crosssectional determinant of the mark et reaction to news. They find that investors underreact less (i .e. the post-earning-announcemen t drift is lower) when the disclosing firm has issued credible forecasts in the past.2 Chen, Francis, and Jiang (2005) develop a model of Bayesian learning in which investors update their prior beliefs as the accuracy of analysts forecast is revealed. Their empirical results are consistent with stock market reactions to analyst forecasts increasi ng in the accuracy and length of the forecasting period. I develop a number of proxies for credibi lity relating to buyback announcements. In the case of buybacks, metrics of firm reputati on include how management behaved following previous share repurchase announcements: whethe r shares were actually repurchased in the amount suggested in the original repurchase announcement. The cr edibility and market revaluation of a repurchase a nnouncement might be undercut if management has not followed through on past repurchases, particularly if the stock being repurchased underperformed in the post-announcement period. If management was signa ling undervaluation, the lagging stock price performance should provide additional motivation to repurchase stock. Inside r trading activity in 2 Ng, Tuna, and Verdi (2008) use multiple proxies for credibility, including the precision (range) of the forecast, the accuracy of prior forecast, and whether or not the firm has pr ovided forecasts in the past.
14 the immediate aftermath of prior share repurchase announcements al so proxies for credibility. Net insider purchases (sales) in the period fo llowing past repurchase announcements suggest a strong (weak) conviction in th e undervaluation signal implied by the repurchase announcement. Stock sales may also imply an attempt to manage stock prices upward. Additional factors may relate to the cred ibility of managerial signaling via share repurchases. First, I examine whether the firms actual earnings have met or beat expectations over the past four quarters. The ability to meet or beat expectations influences a firms overall reputation. For instance, when Ge neral Electric missed earnings expectations in April of 2008, a spokesman for Goldman Sachs stated, We believe the miss and cut to guid ance raises credibility concerns for GE over the near-term.3 A management team known fo r consistently meeting or exceeding (missing) expectations may be viewed as less (more) inclined to engage in cheap talk. Second, I control for discretionary accruals, whic h may proxy for managerial intent to mislead investors. Chan, Ikenberry, Lee, and Wang (2006) find that repurch asing firms with low earnings quality experience poor operating performance and lower long-run returns than other highearnings quality repurchasing firms. Thus, higher levels of discretionary accruals may suggest an attempt by management to artificially increase earnings and share price. Finally, I examine whether the market react ion to buyback announcements complements or contrasts with other recent indicators of the attractiveness of the repur chasing companys stock. In particular, are announcement returns influenced by insider trading prio r to the announcement or by the average analyst opinion at the time of the announcement? Using a sample of share repurchases from 1982 to 1990, Raad and Wu (1995) identify a positive re lationship between net insider buying and subsequent repurchase annou ncement returns, and John and Lang (1991) 3 Moore, Heidi N., The GE Earnings Grenade, The Wall Street Journal, April 11, 2008.
15 document that returns surrounding dividend announ cements are significan tly lower for firms with insider selling leading up to the announcement. These studies suggest that recent optimistic signals may validate the information implied by a sh are repurchase, leading to a stronger market reaction. Nonetheless, it is po ssible that insider trading and analyst recommendations act as a substitute signal, that the mark et has already taken note of mana gements and analysts opinions of the stock. In this case, re purchase announcement returns may be lower for firms with recent insider buying or better analys ts recommendations because the buyback announcement has less of a surprise effect. I study how the market prices these items at the time of the announcement and if they can be used to predict post-announcement returns. For a sample of repurchase announcements ma de from 1990 to 2004, my results suggest that announcements made by firms that consisten tly meet or beat analysts expectations are viewed as more credible. Announcement returns for a firm that met or beat analysts forecasts for all of the past four quarters are at least 93 basis points higher on average than returns for a firm that did not meet or beat any forecas ts during the past four quarters. Also, I test the completeness of this initial market reaction. I find evidence that past repurchasing outcomes are predictiv e of future long-run abnormal returns, suggesting that the market does not fully inspect information related to management credibi lity at the time of the share repurchase announcement. Univariate and multiv ariate results imply that firms with higher completion rates regarding thei r prior repurchase plan have higher abnormal buy-and-hold returns. The results are most pronounced for the subset of firm s whose abnormal long-run returns after the prior announcement fall into the lowest quintile. In this case, firms with high levels of repurchases in their prior re purchase program have cumulative abnormal returns equal to 16.5 percent and 28.4 percent in year s one and two, respectively, sta tistically different from 3.0
16 percent and 7.1 percent for firms with low co mpletion rates after th eir prior repurchase announcement. Hypothesis Development Management Behavior after Prior Share Repurchase Consider the following simple example (Figur e 1-1), where a firm announces a repurchase program which it will either complete (R1 = 1) or fail to complete (R1 = 0). Assume that the market places positive value on the distribution of cash in the form of share repurchases in such a way that the expected initial market reaction to the announcement (r1ANN) is increasing in the probability that the firm will repurchase. dE(r1 ANN) / dP(R1 = 1) > 0 (1-1) Suppose that in the near future this sa me firm announces a second share repurchase program and that the previously observed behavior of the firm can be used to evaluate the probability that the firm will complete the second repurchase. P(R2 = 1|R1 = 1) > P(R2= 1|R1 = 0). (1-2) In other words, if the firm completed the prior buyback, then it is more likely to complete the subsequent buyback. Thus, since actual share repurc hases are valuable to market participants, the market reaction to a share repurchase announcement ( r2 ANN) will be stronger if the firm completed the prior repurchase. Taken together equations (1) and (2 ) imply the following: E(r2 ANN |R1 = 1) > E(r2 ANN |R1 = 0). (1-3) The above example can be extended to formulat e empirical tests concerning the relationship between firm behavior surrounding the prior share repurchase announcement and returns associated with the current announcement. First, I will verify the postulate that share repurchase completion rates are positively related to prior completion rates. Next, I will test the following hypothesis:
17 Hypothesis 1a: Returns surrounding share repurchase announcements are increasing with prior completion rates. The difference between the left-hand and right ha nd side of inequality (3) can be viewed as the discount factor for completion failure. This value, which I la bel the reputational effect, represents the difference in expected returns if th e firm repurchases versus if the firm fails to repurchase: Reputational effect = E(r2 ANN |R1 = 1) E(r2 ANN |R1 = 0). (1-4) Actual repurchases and long-run stock returns should be negativel y correlated given that the value added by repurchasing stock is decreasing in stock price. Thus, I now consider the two possible scenarios where the pos t-announcement stock returns fo llowing the first announcement are either high ( r1 P= H ) or low ( r1 P= L ). In the former case, low completion rates may be more acceptable to investors for two reasons. First, th e firm may have failed to complete its share repurchase plan because the primary motiv e for announcing a repurchase was to signal undervaluation and this objective was accomplishe d (e.g. the share price jumped without a repurchase). Stated differently, the actual buyback becomes unnece ssary if a firms stock price adjusts to reflect its fundamental value and perhaps undesirable if the stock moves enough to become overvalued.4 Second, investors are unlikely to puni sh management teams for shirking responsibilities if they simultaneous ly experience a significant increase in wealth. In the case of low returns, I argue that actual share repurchases become more salient. Shareholders who have recently experienced a decline in wealth proba bly question the credib ility of the current management team and hence monitor management more closely. Thus, th e penalty associated 4 Erin Callan, former Chief Financial Officer of Lehman Brothers Holdings Inc., recei ved criticism for repurchasing stock at a time when it was likely overvalued. A June 13, 2008 article in The Wall Street Journal reports that during her 6 month tenure as Lehmans CFO, the firm spent hundreds of millions of dollars to repurchase its own shares, at prices that turn out to be wildly high, which might have led investors to question her financial leadership.
18 with a past failure to repurchase may be greate r when the firm experienced low returns, leading to a larger reputational ef fect when returns are low than when returns are high. E(r2 ANN |R1 = 1 and r1 P = L) E(r2 ANN |R1 = 0 and r1 P = L) > (1-5) E(r2 ANN |R1 = 1 and r1 P = H) E(r2 ANN |R1 = 0 and r1 P = H) This inequality leads to the following empirically test able hypothesis: Hypothesis 1b: The relationship between announcement returns and prior completion rates is strongest when prior post-announcement stock returns are low. Other Credibility Factors Now suppose that following a share repurchas e announcement insiders choose to either buy ( IT1 = Buy ), hold ( IT1 = Hold ), or sell (IT1 = Sell ) company stock. Since one possibility is that managers announce a repurchase plan in or der to temporarily boost their companys stock price and unload a portion of their holdings at this inflated price, the market may consider how insiders behaved following the most recent prio r repurchase plan. In this case, announcement returns should be positiv ely related to the extent to which ma nagerial trades following the prior share repurchase concurred with company-level signals: E(r2 ANN |IT1 POST = Buy) > E(r2 ANN |IT1 POST = Hold) > E(r2 ANN |IT1 POST = Sell). (1-6) Hypothesis 2: Returns surrounding repurchase announcemen ts are increasing (decreasing) with prior ex post insider buying (selling). When reacting to a share repurchase announcement, the market may factor in indicators of firm credibility that are unrelated to share repurch ases. For example, if a firm consistently meets the expectations of analysts, then the market ma y view this company as more reputable and their announcements as more credible. On the other hand, if a firm has developed a reputation for inflating stock prices or earnings through the use of discretionary acc ruals, then the credibility of the current announcement may be dubious. If firms can be classified as either a good reputation firms ( Rep1 = Good ) or a bad reputation firms ( Rep1 = Bad ) based on factors unrelated to share
19 repurchases, then the market might react more positively to repurchase announcements made by the former group as opposed to the latter: E(r1 ANN |Rep1 = Good) > E(r1 ANN |Rep1 = Bad). (1-7) This example leads to two empi rically testable hypotheses: Hypothesis 3a: Returns surrounding repurchasing announcements are increasing in the portion analysts earnings forecasts that were met or exceeded in the past year. Hypothesis 3b: Returns surrounding repurchasing announcements are decreasing in the amount of discretionary accruals. Outside Signals Finally, investors ma y look to the actions of in siders as well as the opinions of informed outsiders to validate the underval uation signal inherent in a re purchase announcement. Consider a case where other signals about the value of the stock, such as insider trades and analyst recommendations, are either optimistic ( S1 = O ) or pessimistic ( S1 = P ). Optimistic (pessimistic) signals include recent insider buying (selling) and buy (sell) recommendations on the part of analysts. If these observable signals suggest that the companys stock may be trading below its fundamental value, then they may give creden ce to the undervalution signal implied by a share repurchase announcement and thus lead to higher announcement returns. E(r1 ANN |S1 = O) > E(r1 ANN |S1 = P). (1-8) However, it is also possible that the positiv e news implied by the announcement of a share repurchase is redundant in the case where th e market has already observed other recent optimistic signals. In this instance, returns ar ound share repurchase a nnouncements viewed as redundant might be lower. E(r1 ANN |S1 = O) < E(r1 ANN |S1 = P). (1-9)
20 The theoretical ambiguity of the relationshi p between announcement returns surrounding share repurchases and other prior underv aluation signals leads to two mutually exclusive hypotheses for the cases where repurchase announcements ar e either complementary or redundant signals: Hypothesis 4a: (Complementary signal hypothesi s). Returns surrounding repurchase announcements are increasing (decreasing) with pre-announcement insider buying (selling). Returns surrounding repurchase announ cements are higher (lower) for firms with optimistic (pessimistic) analyst ratings. Hypothesis 4b: (Redundant signal hypothesis) Returns surrounding repurchase announcements are decreasing (increasing) with pre-announcement insider buying (selling). Returns surrounding repurchase announ cements are lower (higher) for firms with optimistic (pessimistic) analyst ratings. Long-Run Price Effects The underlyi ng assumption behind the hypotheses outlined above is that investors price these reputation proxies at the time of the buyback announcement. Nonetheless, extant literature documents that investors often fail to incorporat e information into stock prices in a timely manner. For instance, the post-announcement drif t theory (Ball and Brown, 1968; Foster, Olsen and Shevlin, 1984; and Bernard and Thomas, 1989, 1990) suggests that the market initially underreacts to unexpected earnings a nnouncements and that stock prices thus continue to drift in the same direction long-term. More recently, Cohen and Frazzini (2008) find evidence that investors fail to incorporate information on eco nomically linked firms into stock prices. This investor inattention leads to the predictability of long-run returns. Applied to buybacks, this drift hypothesis implies that investors may not factor in past information concerning repurchases, returns, other credibility factor s, and outside signals at the time of the announcement. If not, the above hypotheses conc erning announcement returns can be extended to post-announcement long-run returns.
21 Data and Sample Sample Selection My initial sample of open ma rket repurcha se announcements reported in the Securities Data Corporation (SDC) database begins on Janu ary 1, 1986 and, in order to calculate three-year post-announcement returns, ends on December 31, 2004. Firms not reporting either the percent of shares outstanding sought in the repurchase program or th e intended dollar value of the repurchase program are omitted. Following prior studies, I delete announcements made in the fourth quarter of 1987, announcements made by financial firms and utilities, and announcements made by firms whose stock price was less than fi ve dollars two days prior to the announcement.5 In the rare case in which a firm announces two share repurchase plans in the same fiscal quarter, only the first repurchase announcement is retained.6 For each announcing company I gather daily and monthly stock price data from CRSP, quart erly accounting data from Compustat, analyst forecasts and actual earnings information from I/ B/E/S, and insider trading data from Thomson Financial. Finally, the analysis re quires at least four years of data prior to each announcement to calculate lagged variables.7 I establish a cutoff of four year s to account for possible changes in management. In addition, the market may have limited memory and, therefore, only consider 5 The excessive quantity of share repurchase announcements during this quarter suggests that they are likely to be fundamentally different from announcements made during other time periods (Ikenberry et al. (1995), Stephens and Weisbach (1998)). Billett and Xue (2006) also delete financial firms and utilities due to increased regulation in those industries. Low priced stocks are omitted because calculating their returns is problema tic due to bid-ask bounce. Billett and Xue (2007) also delete stocks trading at below five dollars. 6 A second repurchase announcement in the same qua rter may indicate that the SDC reported the same announcement twice because it appeared in two separate news sources. It is also possible that a firm announces two separate repurchase programs in the same fiscal quarter; however, since my estimate of actual repurchases is based on quarterly data, calculating the actual repurc hases between announcements is unfeasible. 7 Results are unchanged if a three or five year cutoff is used instead.
22 recent events. Because my analysis requires f our years of pre-announcement data, my final sample contains 5,519 announcements made between January 1, 1990 and December 31, 2004. Proxies for Reputation and Signaling Credibility My variables of interest measure firm reputation and returns around share repurchase announcem ents. Summary statistics for all variab les are presented in Table 1-1, and detailed variable definitions are given in the Appendix. My principle proxy for firm reputation in the co ntext of share repurchas es is the portion of the announced amount that the firm actually repu rchased during the prior share repurchase plan. I begin with percent sought the announced percentage of shares outstanding that the firm intends to buy back during the repurchase program. If this value is missi ng but the dollar value of the program is reported, then I calculate percent sought as the dollar value of the repurchase program divided by the market capitaliza tion of the firm one quarter prior to the announcement. The sample mean (median) of percent sought is 7.1 percent (5.2 percent) of shares outstanding. These repurchases are large investments: a firm in the 90th percentile of market capitalization that announces a plan of average size is intending to repurchase over one ha lf of a billion dollars worth of its shares. There is also wi de variation in this variable. The 10th percentile of percent sought is only 2.2 percent while the 90th percentile is 13.1 percent. Following Stephens and Weisbach (1 998) and Dittmar (2000), I calculate actual percent repurchased as spending on the repurchase of common and preferred stock reported in Compustat quarterly minus any decrease in redeemable preferred stock, scaled by the minimum price each quarter.8 Banyi, Dyl, and Kahle (2005) identify this measure as the most accurate 8 Scaling by the minimum price assumes that management possesses private information that allows firms to time the market when they repurchase shares. The results of th is paper are robust to relaxing this assumption and using average quarterly price instead.
23 proxy for actual shares repurchas es, especially for firms with high levels of employee stock option exercises. I sum this figure over two year s or until the quarter before following share repurchase announcement.9 The average firm repurchases approximately 4.4 percent of shares outstanding. This value is substa ntially less than the mean of percent sought, implying that the average firm buys back less than announced. I then calculate percent completed as actual percent repurchased divided by percent sought truncated at 100 percent.10 While the average firm buys back 66.5 percent of the announced amount, this variable is hi ghly skewed. The median value of percent completed is 80.5 percent, and the 75th percentile is 100 percent. These resu lts are similar to those of Stephens and Weisbach (1998), who find that firms on aver age acquire between 74 and 82 percent of the shares announced. Lagged percent completed the percent completed associated with the most recent prior deal, is available for 2,720 firm announcements. The mean (median) value of lagged percent completed is 70.1 percent (86.8 percent), suggesting that the unconditional completion rate is comparable to the completion rate conditional on announcing a second repurchase. The relationship between past completion rates and market reactions to current announcements is likely dependent upon past stock price performa nce. When returns are low, actual repurchases become more salient, and market reactions are more likely to be positively 9 By stopping calculations before the next repurchase announcement, this estimation may be biased downwards for cases where the next announcement happens before two years. Nonetheless, the mean completion rate conditional on an announcement within two y ears (71.8 percent) is greater than the unconditional mean. 10 While the SDC database is considered the most comprehensive source of share repurchase data, it is incomplete. I truncate percent completed at 100 percent to control for potentially missing subsequent share repurchase announcements. It is also possible that data is missing for the most recent prior share repurchase announcement. If there is no reported announcement during the four years pr ior to the current announcement, then the observation will not be included in analysis that depends upon past data. The case in which the most recent announcement is not reported in the SDC database but there was another anno uncement within four years will bias my results against finding a relationship between past repurchasing outcome and future repurchases and returns.
24 related to percent completed To measure stock performance, I calculate two-year abnormal returns, defined as the annualized si ze and book-to-market adjusted returns on the stock of the repurchasing firm from the month following the most recent prior announcement until 24 months following the most recent prior announcement. If the next share repurchase announcement occurs within this 24-month period, then return s calculations end the month before the current announcement. The annualized mean abnormal return equals 4.6 percent, but this variable varies considerably: the 10th percentile is -46.9 percent while the 90th percentile is 54.3 percent. A complete representation of firm reputat ion should include proxies for whether management intentionally misled investors. Lagged post-announcement insider trading measures aggregate trades following the prior repurch ase announcements. Similar to John and Lang (1991), insider trading variables are the dollar valu e of shares purchased minus the dollar value of shares sold, scaled by the sum of the value of shares bought and sold. In order to be included in this measure, the transaction must have taken place during the three months following the prior repurchase announcement and must have been reported to the Secu rities and Exchange Commission (SEC) before the current announcement. This variable is set equal to zero if no trading took place. The market may view a company as more reput able if other non-repur chase related factors reflect credible behavior. Therefore, I estimate percent of forecasts met the percentage of the last four quarterly mean earnings forecasts that the firm has either met or exceeded. If a firms management uses a repurchase announcement to in flate stock prices or ea rnings, then they may employ other methods of manipulation as well. Th erefore, I include disc retionary accruals to proxy for an alternative way in which managers could attempt to manipulate earnings and stock prices. Similar to the methodologies presented in Dechow, Sloan, and Sweeney (1995) and used
25 by Bergstresser and Philippon (2006) and Louis and Robinson (2005), accr uals are calculated using the Jones model with quarter and industry dummies, where industry is defined by the first two digits of a firms SIC code.11 Finally, I include a measure of inside r trading and analyst recommendations. Preannouncement insider trading calculated as the dollar value of shar es purchased minus the dollar value of shares sold, scal ed by the sum of the value of shares bought and sold, proxies for the aggregate trading behavior of insiders prior to the announcement and reveals whether insiders personally exhibit behavior consiste nt with an undervaluation signal. Analysts recommendations, which proxy for the outside opinion of the firm, range from 1 to 5, with 1 being a Strong Buy. Mean analyst recommendation, the mean recommendation of the repurchasing firms at the time of the announcement, is 2.1 on averag e, indicating th at the average repurchasing firm still holds a Buy recommen dation. The number of Buy and Strong Buy recommendations fell dramatically in 2002 and 2003 due to investor sentiment and SEC sanctions in 2003.12 To account for time variation in recommendations, I scale mean analyst recommendation by the yearly average consensus for all firms in I/B/E/S to create net mean analyst recommendation In addition to the level of outside opinion, I consider the extent to which a firm is able to guide market particip ations and consistently meet expectations. 11 Bergstresser and Philippon (2006) analyze yearly data and thus include year dummies to account for time variation. Louis and Robinson (2005) use quarterly data and include calendar quarter dummies to control for seasonality (e.g. all regressions using data from the same quarter are assigned the same dummy variable, regardless of the year). My approach accounts for both the time variation and the seasonality of discretionary accruals by assigning a dummy variable to each individual quarter. Both papers include industry dummies. The correlation among each possible pair of these variables is 0.99. 12 The Securities and Exchange Commission reached an agreement with the ten largest investment banks on Apri 28, 2003 which required then to separate their investment banking and research departments to avoid conflicts of interest (Barber, Lehavy, and Trueman (2007)).
26 Returns Calculations I hypothesize that prior post-announcement behavi or influences the credibility of future repurchase announcem ents, and that announcement returns measure th e markets assessment of whether the repurchase will be completed. Announcement returns are cumulative returns calculated from two days prior to the announcem ent until two days following the announcement minus cumulative equally-weighted market returns. Consistent w ith the prior literature, I find that abnormal returns surr ounding repurchase announcements ar e positive (1.2 percent) on average and significantly different from zero at the one percent level. This confirms prior research that the market views these announcements as good news. Additionally, I test the completeness of the initial market reaction by examining long-run returns. Chan, Ikenberry, and Lee (2004) document abnormal buy-and-hold returns of 23.6 percent during the four years following a repu rchase announcement, and Ikenberry, Lakonishok, and Vermaelen (1995) find positive and significant abnormal returns of 2.04 percent, 2.31 percent, and 4.59 percent during the first, second and third years, respectively, following an announcement. Abnormal returns are calculated for years one through three as the buy-and-hold return on the repurchasing firm minus the buy-and-hold return on a portfolio of firms matched on size and book-to-market. Cons istent with prior studies, abnormal returns are positive and significant following the announcement. I identify significant abnormal returns of 5.4 percent on average during the first year after the announ cement and an additional 3.7 percent during the second year. Unlike prior studies that provide ev idence of abnormal returns for three years, my results indicate that abnormal returns are in significant after the second post-announcement year. If announcement returns fail to incorporate the effect of reputation, then I expect higher long-run abnormal returns for firms that are more credible based on the empirical proxies presented in this paper.
27 Control Variables I construct control variables that mi ght infl uence share repurchase outcomes or market reactions to share repurchase announcements: rece nt stock returns, investment opportunities, information asymmetry, agency problems associat ed with excess cash on hand, capital structure, concurrent earnings and dividend announcemen ts, and frequency of share repurchases. Comment and Jarrell (1991) and Stephens a nd Weisbach (1998) identify a negative relationship between pre-announcem ent excess returns and announcemen t returns. Thus, I create the variable pre-announcement returns the buy-and-hold returns on the stock of the repurchasing firm from 40 to 6 trading days prior to the a nnouncement minus the buy-and-hold returns on the market during the same time period. Following Dittmar (2000), I control for investment opportunities, information asymmetry, agency problems, and capital structure. The ratio of market equity to book e quity is a standard proxy for i nvestment opportunities, where low market-to-book firms have more investment opportuni ties available. If mark et price setters fear that firms tend to overinvest instead of dist ributing cash, then a repurchase announcement made by a firm with few other opportunities may be viewed positively. Since market-to-book ratio varies widely by industry and time period, net market-to-book is the repurchasing firms marketto-book ratio minus the industry median ratio at the time of the announcement, where industry is defined by the first two digits of Compustats industry classification co de. Next, firm size is often used as a proxy for information asymmetry since more information is generally available for large firms than small firms. Size decile the firms market capitalization decile (with 1 being the smallest and 10 the largest) during the calendar quarter prior to the repurchase announcement, captures the surprise effect of the announcement since more information may be conveyed in announcements made by firms with higher information asymmetry. If firms repurchase stock to mitigate agency problems asso ciated with excess cash, then the amount of
28 cash on hand is likely to affect the level of repurchasing activit y. I control for this effect by including cash, which is scaled by market capitali zation. Finally, firms may use share repurchases to bring their capital structure closer to their target level. To account for this, I use net leverage the firms debt to assets ratio minus the industry average leverage ratio. All of the above variables are calculated at the end of the fiscal quart er prior to the announcement. Despite the fact that simulta neous earnings and dividend announcements may significantly drive announcement returns, few share repurchase studies control for these concurrent events. I control for earnings or dividend announcements made within two days of the share repurchase announcement. Positive surprise and negative surprise equal one if the announcing firm released quarterly earnings results above or below, re spectively, the consensu s mean of analysts forecasts. I also include a dummy variable, no surprise to account the effect of the presence of an earnings announcement, even if earnings were equal to their forecasted value. Increase, decrease, and no change equal one if the firm announced an in crease in dividends, a decrease in dividends, or a continuation of current dividends, respectively, w ithin two days of the share repurchase announcement. Special dividend equals one if the firm announced a special dividend within two days of the buyback announcement. A surprisingly large portion of repurchase announcements coincide with other announcements: 16.4 percent of open market share repurchases are contemporaneous with earnings announcements and 25.4 percent with dividend announcements. While repurchases are considered good news, they do not necessarily cluster with other good news events. Just as many ne gative earnings surprises are associated with repurchase announcements as posit ive earnings surprises, and a portion of the concurrent dividend announcements are for reductions in dividends.
29 Jagannathan and Stephens (2003) find th at announcements made by infrequent repurchasers are viewed more favorably by the mark et than those made by frequent repurchasers. To account for this effect, I create dummy va riables to measure frequency. Following their methodology, occasional repurchaser equals one if the current repurchase announcement is the second announcement in the past fi ve years, and zero otherwise. Frequent repurchaser equals one if the announcement is at least the third announcement in the past five years. Results I estimate three pr imary sets of results. First, I verify that there exists a positive relationship between past and future repurch asing outcomes through univariate statistics and Tobit models. Second, I estimate announcement retu rns to examine market reactions to share repurchase announcements and, more precisely, if the market considers past repurchasing outcomes and other related credib ility variables. Finally, I investigate the completeness of the initial market reaction by studying the relationshi p between firm reputatio n and long-run returns. Prior Repurchasing Behavior as a Pred ictor of Future Repurchasing Behavior Figure 1-2 shows a positive relationship betw een future share repurchase o utcomes and past share repurchase outcomes, and the magnitude and significance of the differences in the heights of the bars in this figure are presented in Table 1-2. For example, the average difference in percent completed for a firm that repurchased between zero and 25 percent of the announced amount following the prior share repurchase a nd a firm that repurchased between 75 and 100 percent is 16.071 (77.450 minus 60.479) percent and is significant at the 1 percent level. Among firms with a history of buybacks, repurchasing outcomes are significantly positively related to past performance. These results are economica lly important as well: a firm in the 0-25% lagged percent completed group buys back 15 percent less of their current announced amount on average than a firm that bought back at least 75 percent last time (i.e. in the 75-100% or 100%
30 lagged percent completed group). I also compare firms with a history of share repurchases to those that lack a recent prior repurchase. I find that firms with no recent prior announcements buy back significantly less than firms that repurch ased at least 25 percent during their last share repurchase program. On average, a firm that has not established a reputation related to repurchases buys back at least 15.5 percent less of their announ ced amount than a firm that repurchased 75 percent or more duri ng its last repurchase program. In Table 1-3 I estimate a Tobit model of completion rates on prior completion rates and other factors that influence buy backs. The tw o-tailed Tobit model accounts for the censored nature of percent completed which is naturally bounded by zer o and manually truncated at 100.13 The coefficient on lagged percent completed is positive and statis tically different from zero in all specifications. This fi nding supports the theory that past repurchasing outcomes are positively related to future behavior. Using estimates from Model (5), a one standard deviation increase in lagged percent completed is associated with a 10.1 pe rcentage point increase in percent completed Other firm-specific characteristics may influence actual repurchasing behavior. Percent sought should be negatively related to percent completed while cash may be positively related to actual repurchases. Firms with less informati on asymmetry and heavier monitoring, i.e. greater size decile may be more likely to follow through on their share repurchase (Bhattacharya and Dittmar (2004)). Furthermore, share repurchase outcomes may also be related to other investment opportunities, capital structure (Dittmar (2000)), and the frequency of repurchasing activity (Jagannathan and Stephe ns (2003)). Empirical results imply a negative relationship between the announced percent sough t and the subsequent completion rate. In other words, the 13 This controls for subsequent repurchase announ cements that are potentially missing from the sample.
31 larger the announced share repurcha se plan, the less likely the announc ing firm is to complete the plan. Completion rates are also negatively rela ted to net leverage and pre-announcement insider buying. Consistent with Chan, Ikenberry, and Lee (2004 ), Models (4) and (5) estimate regressions including lagged long-run returns variables and future long-run re turns variables, though future returns are clearly unavailable at the time of the announcement. Percent completed is negatively related to prior long-run returns, though this relationship is only ma rginally significant. Additionally, completion rates are negatively correlated with one-year abnormal returns, suggesting that firms repurchase less if their stoc k returns are higher during the year prior to or during the year following the announcement. The negative correlation between repurchases and returns is potentially due to companies no longe r considering their own stock undervalued. After controlling for long-run returns, regression results continue to provide justification for the assumption that share repurchase completion ra tes are positively related to prior completion rates.14 Announcement Returns I am interested in how the m arket res ponds to share repurchase announcements and, specifically, if investors incorporate past repurchasing outcomes in to their reactions to current announcements. Model (1) of Table 1-4 exam ines announcement returns around all share repurchase announcements in the sa mple and includes the four re putation proxies that are not conditional upon having a recent past histor y of repurchasing. The coefficient on percent of forecasts met is positive and significant, indicating that firms that have established a reputation 14 There may be concern about having percent sought as an independent variable since it is the divisor of the dependent variable, percent completed When Models (3) and (4) are estimated without percent sought, the coefficients on lagged percent completed remain almost identical (0.270 and 0.275, respectiv ely, and significant at the 1% level). All other coefficients also re main similar in magnitude and significance.
32 of communicating with market participants a nd meeting expectations are viewed as more credible at future announcements. Pre-announcement insider trading is negatively related to announcement returns, consistent with the theory that increased inside r buying has previously signaled that the firm was undervalued. A one standard deviation increase in this variable, which implies more insider buying relative to selling, is associated with a 58 basic point decrease in announcement returns. Since a higher analyst recommendation implies a more pessimistic one, the positive coefficient on net mean analyst recommendation suggests that firms that are considered in favor (out of favor) receive lowe r (higher) announcement retu rns. An increase of one standard deviation in net mean analyst recommendation indicating a decline in analyst optimism, suggests a 70 basis point increase in announcement returns. Hence, for both insider trading and analyst recommendati ons, results suggest that the information content of the repurchase announcement is redundant for firms al ready viewed as undervalued by insiders and analysts. The amount of discretionary accruals does not significantly a ffect returns in this model. Consistent with prior literature, the results suggest th at announcement returns are increasing in the size of the re purchase program and decreasing in pre-announcement returns. Also, market price setters respond more posi tively to announcements made by value firms, which have low market-to-book ratios and genera lly have less investment opportunities, and by smaller firms, whose higher levels of informa tion asymmetry may lead to more of a surprise effect at the time of the repur chase announcements. I include dummy variables to account for contemporaneous earnings and dividend announcemen ts but exclude coefficient estimates for brevity. As one might expect, announcement retu rns are significantly hi gher (lower) when the repurchase announcement coincides with a positive (negative) earnings surprise.
33 Models (2), (3) and (4) of Ta ble 1-4 are restricted to repe at announcements, defined as at least the second announcement made by a firm in the past four years. I interact lagged percent completed with a binary variable equal to one if the firms two-year abnormal returns fall into the lowest quintile. I calculate two-year abnormal returns as the annualized returns on the repurchasing firms stock minus the annualized return on a portfolio of control firms matched on size and book-to-market ratio from one mont h following the prior repurchase announcement until 24 months following the prior announcement. (See Appendices A and B for further discussion of control firms.) If the current shar e repurchase announcement occurs within this 24month period, then abnormal re turns calculations are stopped the month before the current announcement. The coefficient on the interaction te rm measures the difference in the extent to which investors incorporate prior repurchasing outc omes when returns are in the lowest quintile versus when they are not. It is po sitive, but fails to achieve statis tical significance, leading me to question to completeness of the initial market reaction. All othe r coefficients are similar in magnitude and significance to those in Model (1). Model (4) excludes the 586 observations concurrent with earnings or di vidend announcements to control for the confounding effects of these contemporaneous events. These results are robust to this alternative sample of share repurchase announcements. Post-Announcement Returns It is possible that the ma rket s reaction at the tim e of the announcement is incomplete and that prior repurchasing outcomes influence return s long-term. To test this hypothesis, I regress two-year abnormal buy-and-hold returns on lagged percent completed as well as other reputation and control variables that may affect stock price performance in Table 1-5. Abnormal buy-andhold returns are the annu al return on the repurchasing firms stock minus the annual return on a portfolio of control firms ove r the 24 months beginning the month following the repurchase
34 announcement. To construct control portfolios, I sele ct the twenty firms in the same size decile as the repurchasing firm at the time of the announ cement that are closest in market-to-book to the repurchasing firm. If a matched firm delists during that year, then I assume that the proceeds from that investment are invested in a market portfolio with returns eq ual to the returns on the value-weighted CRSP index. Control portfo lios are rebalanced on an annual basis. Model (1) of Table 1-5 includes lagged percent completed other variables related to reputation and variables potentially related to long-run returns. Consistent with the findings of Chan, Ikenberry, and Lee (2004) and Chan, Ikenbe rry, Lee and Wang (2006), the ratio of market value to book value is negatively related to longrun abnormal returns. Net leverage is positively related to long-run returns, but this coefficien t is only marginally significant. Pre-announcement insider trading is negativ ely related to two-year abnormal retu rns, indicating that insiders do not appear to time the market well. In Model (2) I condition upon low return s between the prior share repurchase announcement and the current announcement to id entify a positive and significant relationship between lagged percent completed and announcement returns. The coefficient on the interaction term measures the difference in the extent to which prior repurchasing outcomes can predict twoyear abnormal returns when retu rns are in the lowest quintile ve rsus when they are not. It is positive and statistically significant, implying that past repurchasing outcomes are more informative when prior returns are low. Conditio nal upon low returns, a one standard deviation increase in lagged percent completed is associated with a 13.1 per centage point increase in twoyear abnormal returns in Model (2). The results are consistent with the predictions that long-run returns and prior completions rates are positiv ely correlated and that the strength of the correlation is stronger when stock returns afte r prior repurchase announcements are low.
35 I verify the robustness of these results usi ng two alternative sample s. First, Model (3) excludes the two variables that require analyst coverage data, increasing the sample size by 435 observations. Next, Model (4) uses the subset of repurchase announcem ents that occurred independently of dividend and earnings announc ements. In both cases, I find that on average firms with low returns following the prior repurchase plan continue to be associated with low returns following the next repurch ase announcement. However, empi rical findings are consistent with the theory that firms w hose returns were low following th e previous announcement but that repurchased more stock will outperform firms w ith low returns that repurchased less shares, confirming prior results. Table 1-6 presents cumulative buy-and-hold abnormal returns for the first, second and third years following the repurch ase announcement using the entire sample of 5,419 repurchase announcements. Bootstrapping is used to infer the statistical significance of long-run returns. Specifically, I replace each firm in my repurch asing sample with a randomly selected nonrepurchasing firm in the same size and book-to -market quintiles at th e time of the buyback announcement. This process is repeated for a total of 1,000 random samples. The p-values reported in this papers reflect the percentage of the 1,000 random samples that have mean abnormal returns higher than the mean abnormal return for the repurchasing sample. A similar procedure is used to infer the statistical sign ificance of the difference in means in abnormal returns. (See Appendix B for more details on the empirical bootstrapping pr ocedure used in this paper.) Abnormal returns are pos itive and significant for the first three ye ars: 5.4 percent, 10.3 percent and 7.9 percent for years one, two, and three, respective ly. Abnormal long-run returns on firms that have announced a share repurchase in th e past four years are similar in magnitude and significance to those of the entire sample. I test whether prior repurchase activity is marginally
36 predictive of future stock returns. Firms that repurchased above the median lagged percent completed have higher CARs than those that repurcha sed less than the median amount, and these results are significant at the 10 per cent level for year s one and two. Next, I segment on returns following the prior share repurchase announcement to find that cumulative abnormal buy-and-hold returns are highe st and most significant for firms whose returns fell into the lowest quintile following th e prior share repurchase. Three-year abnormal returns for the group of firms in the lowest return s quintile after the last share repurchase are 23.8 percent on average, over three times as large as the abnormal returns for any other returns quintile. The firms with high levels of past repur chases help to drive these results: I identify three-year cumulative abnormal buy-and-hold return s of 29.8 percent for the subset of firms in the lowest returns quintile with high levels of repurchases. Of the 1,000 random samples used to construct p -values, none had long-run returns that surp assed this group. My fi ndings suggest that prior share repurchase activity is positively related to long-run abnormal returns, particularly in the case where returns were lo w after the prior repurchase. Conclusions My study examined short-run and long-run stock returns for a sam ple of 5,519 open market repurchase announcements made between January 1, 1990 and December 31, 2004. Since this type of share repur chase is a non-binding commitment market participants may look to external signals to determin e the likelihood that a firm will fo llow through with the repurchase plan and/or the probability that the announcing firm is undervalue d, one of the principle reasons for which a company might announce a buyback. A logical starting point for evaluating the credibility of an announcement is how the company behaved following the last repurchase announcement. Did the management team complete th e repurchase program, in particular in the case where returns are low?
37 About half (2,720) of the announcements in my sample were made by firms that had announced a open market share repurchase in the recent past. I use repurchase plan completion rates and long-run returns following the first announcement to explain market reactions to and long-run returns following the second announcement. After verifying that a link exists between prior and future repurchasing activity, I find th at a firms reputation for meeting or exceeding analysts forecasts is positively related to share repurchase announcement returns. Next, I examine the completeness of the ma rkets initial reaction. Univariate and multivariate results suggest that lagged share repurchases explain long-run cumulative abnormal buy-and-hold returns, particular ly in the case where returns were low between announcements. Firms whose prior completion rate fell above the median level experience greater abnormal returns that their low completion rate counterparts for two years following the current announcement. This discrepancy is most pronounced for the subset of firms whose lagged stock price performance fell into the lowest quintile, perhaps implying that the undervaluation signal sent through the first repurc hase announcement was ignored. My study adds to the extant l iterature on market responses, bo th short-term and long-term, to announcements of open market share repurch ases, and it also questions whether market participants incorporate a firms reputati on when formulating a reaction to corporate announcements. I identify persiste nce in repurchasing behavior, but find that the market only partially incorporates this relationship in the short-term, leading to long-run price effects.
38 Figure 1-1. Timeline of repurchase announcements. Th is figure presents a timeline of a simple example of repurchase announcements. rt ANN and rt P are announcement returns and post-announcement returns, respectively. Rt represents a binary variable equal to one if the firm completes its share re purchase plan and zero otherwise. Announcement 1 Announcement 2 r1 A N N r2 A N N R1, r1 P R2, r2 P
39 59.878 (N = 2529) 60.479 (N = 447) 71.706 (N = 280) 74.038 (N = 394) 77.450 (N = 423) 75.984 (N = 1119) 50.000 55.000 60.000 65.000 70.000 75.000 80.000 No prior repurchases 0-25% 25-50% 50-75%75-100% 100%Percent completedLagged percent completed Figure 1-2. Percent completed by lagged per cent completed. This figure depicts the re lationship between prior share repurchase outcomes and future repurchases. The horizontal axis repr esents the portion of the announced amount that the firm repurchased following the prior announcemen t (lagged percent completed), while the vertical axis represents the mean completion rate for the subsequent announcement (percen t completed).
40 Table 1-1. Summary Statistics. Th is table presents summary data from the sample of open market repurchase announcements made between January 1, 1990 and December 31, 2004. All variables ex cept for long-run returns variables, dummy variables, and insider trading variables are winsori zed at the 1st and 99th percentiles. Standard Category Variable NMeanDeviation10th50th90th Percent sought (%) 55197.0995.2182.2005.60613.095 Actual repurchases (%) 51934.4134.1040.2703.5289.622 Percent completed (%) 519266.54536.4035.33680.519100 Lagged percent completed (%) 272070.16835.1947.64186.831100 Occasional repurchaser dummy 55190.1960.397001 Frequent repurchaser dummy 55190.3600.480001 Announcement returns (%) 55181.2208.546-7.6881.08111.173 Pre-announcement returns (%) 5505-8.29815.123-27.777-7.5069.425 Abnormal returns year 1 (%) 54195.40465.027-53.553-1.32263.841 Abnormal returns year 2 (%) 54193.73460.071-50.760-1.87359.007 Abnormal returns year 3 (%) 54190.60458.061-52.714-4.25852.740 Two-year abnormal returns (after last repu rchase) (%)28084.59580.342-46.921-1.22754.311 Firm-specific characteristics Mean analyst recommendation 42792.0680.5651.3322.91 Scaled mean analyst recommendation 4279-0.0710.559-0.760-0.0870.700 Percent of analyst forecasts met (out of 4 quarters)49770.6820.3120.250.751 Net leverage 5481-0.0620.312-0.487-0.0380.317 Net market-to-book 54790.8892.394-1.0320.2353.450 Discretionary accruals 4574-0.0110.127-0.0530.0010.057 Market capitalization (millions of $) 5374443718731524277344 Size decile 55197.7142.0685810 Cash 53550.2122.9060.0070.0710.340 Pre-announcement insider trading 5519-0.0880.356-0.95300 Lagged post-announcement insider trading 3329-0.1840.485-1.00000 Increase 55190.0660.248000 Decrease 55190.0130.114000 No change 55190.1720.377001 Special dividend 55190.0030.057000 Positive surprise 52230.0690.254000 Negative surprise 52230.0800.271000 No surprise 52230.0150.121000 Dividend announcement dummies Earnings announcements dummies Repurchasing Percentiles Returns
41 Table 1-2. Difference in means of percent co mpleted by lagged percent completed group. This table presents the magnitude and signif icance of differences in mean percent completed by lagged percent completed. The va lues represent the differences in the heights of the bars depicted in Figure 12. The header row and column represent the range of values of lagged percent comp leted. The values within the table are calculated as the mean percent completed of the group represente d in that column minus the mean percent completed of the ro w. T-statistics are in parentheses below the estimates. *,**, and *** represent signifi cance at the 10%, 5% and 1% levels. 0-25%25-50%50-75%75-100% 100% No prior repurchases0.60111.82814.15917.571 16.106 (0.31) (5.40)***(7.77)***(10.84)***(13.28)*** 0-25% 11.22713.55816.970 15.505 (4.06)***(5.45)***(7.24)***(7.45)*** 25-50% 2.331 5.743 4.2782 (0.89) (2.29)**(1.98)** 50-75% 3.412 1.947 (1.55) (1.04) 75-100% -1.465 (0.82)
42 Table 1-3. Repurchases regressi ons. This table provides estimates of future share repurchases based on prior repurchasing outcomes. The equa tions estimated in this table are based on a two-tailed Tobit model where the dependent variable is percent completed, the portion of the announced amount that the firm subsequently repurchases, defined as actual percent repurchased divided by percent sought, truncated at 100 percent. Zstatistics are shown in italics below the estimated coefficients. *, **, and *** represent significance at th e 10%, 5% and 1% levels. ( 1 )( 2 )( 3 )( 4 )( 5 ) Lagged percent completed 0.2700.2660.2710.2830.286 7.71 *** 7.84 *** 7.88 *** 8.21 *** 8.29 *** Percent sought -2.494-2.430-2.425-2.364 -11.48 *** -10.82 *** -10.81 *** -10.48 *** Pre-announcement returns 0.0130.0330.040 0.15 0.370.45 Net market-to-book 0.7520.7500.620 1.47 1.421.16 Size decile 0.1060.1540.218 0.16 0.220.32 Cash 0.1400.1640.113 0.39 0.460.32 Net leverage -6.565-8.033-8.087 -1.56-1.89 -1.90 Pre-announcement insider trading -5.646-6.352-6.478 -1.84 -2.06 ** -2.11 ** Lagged two-year abnormal returns -0.028-0.028 -1.90 -1.91 Abnormal returns year 1 -0.064 -2.94 *** Abnormal returns year 2 0.010 0.46 Constant 72.89590.41786.85085.54184.796 26.63 *** 29.16 *** 13.66 *** 13.46 *** 13.33 *** N 26632663258725482535 Pseudo R-squared 0.00350.01130.01210.01280.013
43 Table 1-4. Announcement returns. This table presen ts coefficient estimates from ordinary least squares regressions of a nnouncement returns. The depe ndent variable in both regressions is announcement returns. Mode l (1) includes the full sample of share repurchase announcements. Models (2) and (3) include announcements made by firms that have announced a share repurchase program in the past 4 years. Model (4) is estimated with the sample of repeat announcements that were not made concurrently with earnings or dividend announcements. T-statistics are shown in italics below the estimated coefficients *,**, and *** represent significance at the 10%, 5% and 1% levels. Percent sought 0.0730.1270.1250.112 2.47 ** 3.56 *** 3.52 *** 2.52 ** Pre-announcement returns -0.047-0.041-0.039-0.043 -4.64 *** -3.24 *** -3.07 *** -2.78 *** Net market-to-book -0.278-0.158-0.119-0.134 -4.28 *** -2.19 ** -1.61-1.45 Size decile -0.0340.0170.0140.031 -0.37 0.14 0.11 0.22 Cash -0.061-0.090-0.090-0.090 -1.47-2.09 ** -2.11 ** -2.05 ** Net leverage 0.805-0.331-0.302-0.307 1.57-0.50-0.46-0.37 Occasional repurchaser dummy 0.061 0.15 Frequent repurchaser dummy 0.3820.2450.2570.472 1.08 0.59 0.61 0.90 Pre-announcement insider trading -1.339-1.194-1.161-0.783 -3.47 *** -2.86 *** -2.78 *** -1.45 Mean analyst recommendation 1.2431.3531.2781.618 4.28 *** 3.68 *** 3.48 *** 3.55 *** Discretionary accruals 1.1062.3412.2243.962 0.91 1.62 1.55 2.01 ** Percent of forecasts met 0.9271.2701.5041.443 1.73 1.98 ** 2.34 ** 1.80 Lagged post-announcement 0.1270.2980.616 insider trading 0.35 0.82 1.32 Lagged percent completed 0.0030.0010.002 0.57 0.09 0.22 Lagged percent completed 0.0160.021 Low returns dummy 1.24 1.38 Low returns dummy -0.306-0.403 -0.30-0.33 Intercept -0.156-1.377-1.474-1.602 -0.19-1.26-1.31-1.20 Concurrent earnings and dividends dummiesYESYESYES NO N 3416169816751089 Adjusted R-squared 0.02620.02690.02860.0346 (1) (2) (3) (4)
44 Table 1-5. Long-run returns. This table presen ts coefficient estimates from ordinary least squares regressions of two-year abno rmal returns. Models (1)-(3) include announcement made by firms that have announced a share repurchase program in the past 4 years. Model (4) is estimated with the sample of repeat announcements that were not made concurrently with earni ngs or dividend announcements. T-statistics are shown in parentheses below the estim ated coefficients. *, **, and *** represent significance at the 10%, 5% and 1% levels. (1) (2) (3) (4) Percent sought 0.7910.8630.8500.998 1.73 1.88 2.26 ** 1.60 Pre-announcement returns 0.1120.1450.1570.307 0.690.881.121.39 Net market-to-book -3.481-3.195-3.560-2.825 -3.71 *** -3.31 *** -4.20 *** -2.12 ** Size decile 0.4240.3650.8450.875 0.290.240.740.44 Cash -0.608-0.597-0.596-0.625 -1.11-1.09-1.16-1.02 Net leverage -13.796-11.428-10.129-14.329 -1.66 -1.36-1.46-1.22 Pre-announcement insider trading -12.464-12.618-12.407-20.691 -2.34 ** -2.35 ** -2.53 ** -2.73 *** Mean analyst recommendation 1.4370.584 0.011 0.310.12 0.00 Discretionary accruals -4.800-5.587-9.036-12.386 -0.26-0.30-0.61-0.45 Percent of forecasts met -11.783-11.183 -15.308 -1.45-1.36 -1.35 Lagged post-announcement -8.636-8.074-6.136-10.432 insider trading -1.86 -1.72 -1.53-1.59 Lagged percent completed 0.001-0.081-0.071-0.121 0.01-1.06-1.13-1.18 Lagged percent completed 0.3710.3370.603 Low returns dummy 2.27 ** 2.52 ** 2.82 *** Low returns dummy -22.242-19.816-35.643 -1.69 -1.89 -2.10 ** Constant 10.61615.2423.49916.922 0.771.060.320.90 N 1684166420991083 Adjusted R-squared 0.01420.01430.01580.0178
45 Table 1-6. Univariate results for long-run returns. This table presents returns duri ng the first, second a nd third year follow ing a share repurchase announcement. Repeat announcers refers to firms that have announced another share repurchase in the past four years. High (low) refers to firms whose lagged percent completed is above (below ) the 50th percentile. Q1 Q5 denotes the quintile of two-year a bnormal returns following the prior repurchase program. Raw returns are the returns on the repurchasing firm, control returns are th e returns on a size and book-to-market matc hed portfolio, and abnormal returns are the difference between raw returns and c ontrol returns. P-values correspond to the abnormal returns measure and are obtained through bootstrapping. See Appendix B for further explanations of abnormal returns and p-va lue calculations. Sample N RawControlAbnormalP-value RawControlAbnormalP-value RawControlAbnormalP-value All firms 541921.57316.1685.4040.000***41.07330.81710.2550.000***55.53047.6137.9170.000*** Repeat announcers269520.86215.0405.8220.000***39.46028.60910.8510.000***53.69945.5528.1470.000*** High repurchases134422.06814.3937.6750.000***41.03227.36513.6660.000***54.22144.5789.6430.001*** Low repurchases135119.66215.6833.9780.023**37.89629.8458.0510.010**53.17946.5216.6580.003*** High Low 26952.407-1.2903.6970.078*3.136-2.4805.6150.085*1.042-1.9432.9850.296 Q1 53521.72911.9479.7820.002***41.85424.10617.7480.002***61.35037.57623.7740.000*** Q2 53320.06115.4044.6570.067*40.81528.62012.1960.023**50.47544.4845.9920.053* Q3 53017.75214.2523.5000.082*35.01527.9907.0250.045**46.33946.0740.2650.217 Q4 53425.32917.8957.4340.008***41.74932.7449.0040.075*58.07550.4157.6600.031** Q5 52719.64614.7484.8980.065*38.52128.6789.8420.018**53.29148.4374.8540.062* Q1 & High 26728.24911.70116.5470.002***49.72121.32028.4010.000***66.16636.41129.7550.000*** Q1 & Low 26815.23412.1923.0410.21334.01626.8817.1350.11356.55138.73717.8150.110 High Low 53513.015-0.49113.5060.010**15.705-5.56221.2660.012**9.614-2.32611.9400.205 Q2 & High 28918.99414.0924.9020.083*37.21128.5398.6710.092*48.22043.0585.1610.105 Q2 & Low 24421.32416.9584.3660.16845.08428.71416.3700.037**53.14746.1726.9750.011** High Low 533-2.331-2.8660.5350.456-7.873-0.175-7.6980.770-4.927-3.113-1.8140.552 Q3 & High 25817.26012.2145.0450.085*34.06424.5029.5620.043**44.55543.0281.5260.216 Q3 & Low 27218.21816.1842.0340.27435.91831.2994.6190.20848.03148.962-0.9310.101 High Low 530-0.958-3.9703.0110.303-1.854-6.7964.9430.243-3.476-5.9342.4580.391 Q4 & High 26823.67217.3706.3020.061*42.14532.3569.7890.10556.27948.9757.3030.216 Q4 & Low 26626.99918.4248.5750.029**41.34933.1358.2140.15059.88451.8658.0200.344 High Low 534-3.327-1.055-2.2730.669 0.797-0.7791.5760.411-3.606-2.889-0.7160.516 Q5 & High 24723.06715.8357.2320.054*43.48529.40914.0760.027**56.89351.9784.9140.100 Q5 & Low 28016.62813.7892.8390.24834.14128.0336.1070.13450.11445.3124.8010.075* High Low 5276.4392.0464.3930.224 9.3441.3767.9680.208 6.7796.6660.1130.461 Year 1 Year 1-2 Year 1-3
46 CHAPTER 2 INSIDER TRADING AND SHARE REPURCH ASES: DO IN SIDERS AND FIRMS TRADE IN THE SAME DIRECTION? Introduction Share repurchases have long been viewed in the finance literature as a means to signal firm undervaluation (see e.g., Vermaelen (1981)). The signaling story entails managers initiating corporate share repurchases while retaining their own stock. If the firm is not actually undervalued when a repurchase is executed, ma nagers suffer a disproportionate loss on the shares they retain. Consistent with the signa ling story, share repurchases are greeted with positive abnormal stock returns on average both at the time of their announcement (Stephens and Weisbach (1998), Jagannathan and Stephens ( 2003), Chan, Ikenberry, and Lee (2004), Chan, Ikenberry, Lee and Wang (2006)) and in th e three years after the announcement (Chan, Ikenberry, and Lee (2004), and Ikenberr y, Lakonishok, and Vermaelen (1995)). Some have noted, however, that share repurcha ses might not convey positive information if management has an alternative agenda to give th e false impression of undervaluation so as to sell their stock at inflated values and/or preven t a hostile takeover (see e.g., Fried (2001)). Fried (2001) notes that in most repurchase programs there is no proh ibition on insider stock sales during the period it is in progre ss or shortly thereafter. A numb er of studies have examined insider trading around the announcement of majo r share repurchases. For instance, Lee, Mikkelson, and Partch (1992) find th at for a sample of self tende r offers, on average, managers sell fewer and buy more shares prior to a repurc hase. Insider trading returns to normal levels after the tender offer. Louis, Sun, and White (2 008) identify abnormally high net insider selling during the quarter of fixed-price and Dutch auc tion tender offer announcements. They also report a negative relationship between concurrent insider selling and both future operating performance and stock price performance.
47 While these studies are illuminating, most repur chases are not executed via tender offers, and for the most part, little attention has been paid to the frequency and nature of insider trading in conjunction with share repurchases. Additionally, we know little about the nature of firms where managers engage in insider selling while in the midst of a share repurchase program and to the subsequent share price performance of firms re purchasing shares condi tional on insider trading behavior. This study employs a simple empirical design. Each quarter we identify firms that have engaged in a non-trivial repurchase (at least 0.25% of the firms ma rket capitalization) and those that did not. We document firms where net insider trading (dollar buys less dollar sells), exceeds a threshold level of firm equity value in eith er the buy or sell direc tion. Insider trading is classified as net buying, neutral or net selling. We ask three fundamental questions. First, do insiders tend to trade in the same direction as th eir companies? One might expect that if stocks are deemed undervalued, then insiders will tend to be net buyers more frequently in quarters when corporate shares are being repurchased. We find that in quarters where companies are repurchasing stock, the frequency of net buying and net selling tends to be higher than during periods where there are no or minimal repurchases. An alternative way of framing the data is that, conditional on the pres ence of insider selling or insider buying, share repurchases are more frequently observed relative to the case where in sider trading is neutral. Curiously, repurchasing activity seems to pick up more in quarter with net selling ve rsus those with net buying. While increased repurchases make sense in qua rters with insider buying, the selling is more perplexing. We find that the results are primarily driven by cases where insiders are exercising large amounts of stock options and by larger firms. Hence, this result is consistent with the claim that firms repurchase stock to combat the dilutive impact of options (see Kahle (2002)). We also
48 find some evidence that firms with higher market to book ratios are more likely to repurchase in conjunction with insider selling. This result al so is with repurchases taking place at high valuations as insiders reduce their own stock exposure. Our second goal is to ascertain if the share re purchasing patterns are related to a firms corporate governance. One view of repurchases that coincide with insider selli ng is that this is an attempt to aid certain insiders with their stock sales by keeping prices hi gh. Alternatively, these might be firms that are overly conscious of the dilutive impact of options and fixated on short term earnings measures. Under these views, firms that repurchase when insiders are selling are not shareholder friendly. Using the G-index of corporate governance (Gompers, Ishii, and Metrick (2003)), we find that the unconditional probability of a firm repurchasing stock and insider selling in the same quarter is higher for firms with weaker shareh older rights. Conditional on the presence of insider selling, we also find mo re share repurchase activ ity. This suggests that more entrenched managers may use share repurchases to maximize their own wealth, rather than shareholder wealth. Third, we investigate stock market return patte rns for firms stocks based on their insider trading/repurchasing classificati on. Prior research has generally found that stocks perform well after open market share repurchase announcements (see e.g., Ikenberry, Lakonishok and Vermaelen (1995) and Chan, Ikenberry and Lee (2004)) or after self tender offer (see Lakonishok and Vermaelen (1990)). Seyhun (2000) summ arizes evidence that historically stock returns are also abnormally high (low) in the ye ar after insider buying (selling). We find that when insiders and corporations are both buying, st ock returns are abnormally high in the quarter of the actual repurchase, the subsequent quarter and in the three year s after the subsequent quarter. During the quarter of the share repurcha se with net insider buying, firms experience
49 statistically significant abnormal return of 2.59 percent and another 0.97 percent in the subsequent quarter. At the end of the quarter subsequent to the repur chase/net buying quarter, investors would have the information, in most cases, to implement a trading strategy of buying these stocks.1 The abnormal returns to this strategy are statistically signi ficant with abnormal returns of 4.95 in year 1, 2.13 percent in year 2 and 1.72 percent in year 3. The cumulative three year return is 9.03 percent. For firms that have repurchases and inside r selling we find much smaller abnormal returns of 0.34 percent during the repurchase quarter and 0.28 percent in the subsequent quarter. Investing in stocks of these firms one quarter after the combined net insider selling and repurchase quarter, the returns are 0.41 percent in year 1, -0.88 percent in year 2 and -1.85 percent in year 3. The cumu lative three year tradable retu rn is -2.33 percent. When we further segment our results for repurchasing firms based on the G-index and insider trading status, we find the weakest cumulative abnormal returns for the quarter of repurchase through nine quarters after repurchase for firms with ne t selling and a high G-I ndex (low shareholder protections). The highest abnormal returns among repurchase firms are for those with net buying and a low G-Index. We conduct two robustness checks on our resu lts. First, we alter the definition of a repurchasing firm in any given quarter to a fi rm that has repurchases that exceed median repurchase levels in the prior eight quarters by 0. 25% of the firms shares outstanding. We obtain similar results for insider trading frequency (not reported), logit models explaining which firms have net insider selling in conjunction with shar e repurchases, and the abnormal returns. We also provide a check on our results by examining a sample of open market share repurchase 1 In virtually all cases, except extremely late reporting, the insider trading data would be available. The repurchase data should be available in most cases within three months of the end of a fiscal quar ter. Current SEC rules require 10-Q and 10-K filings within three months of the end of a quarter, though some firms seek extensions.
50 announcements. Again, we examine whether in the 3 months after the repurchase announcements, there is abnormal insider buyin g and/or selling. We find less convincing evidence that the selling (condi tional on repurchasing) is comi ng from higher G-index firms2 and we find higher long run abnormal stock returns fo r repurchases in conj unction with insider buying versus repurchases in conj unction with insider selling. Overall, our evidence suggests that repurchases in conjunction with insider selling are not consistent with undervaluation as a motive for re purchasing and that the practice is associated with firms possessing lower levels of shareholder rights. Those considering long term investing strategies based on repurchases as a signal of undervaluation that is not fully priced in capital markets should take note that this result does not obtain when insiders are selling their stock at or about at the time that firms are repurchasing their stock. The paper proceeds as follows. Section II lays out the data employed in the study. Section III examines repurchase/insider trading frequencies. Section IV examines the determinants of repurchasing/insider trading freque ncies. Section V examines long run stock returns. Section VI offers our conclusions. Sample and Variable Construction In devising a samp le, we use data from Co mpustat Quarterly spanning the period from 1987 to 2007. We first eliminate firms that have only been in the database for two years. This eliminates firms that have just raised capital via a public offering and a llows for the calculation of lagged variables. We next el iminate firms with a negative b ook value at the beginning of the quarter since a firms market-to-book ratio is us ed in both univariate and multivariate analyses. Since part of our analysis is devoted to stock re turns, we delete firms whose stock price at the 2 Given the nature of the announcement event, the empirical design does not lend itself well to the probability of announcing a share repurchase conditional on insider trading.
51 beginning of the quarter less than one dollar beca use calculating returns may be problematic due to bid-ask bounce. This sample is used to ex amine how insider trading correlates with share repurchases for a broad set of firms over time. Our variables of interest include actual shar e repurchases and insi der transactions. We calculate share repurchases as the spending on the repurchase of comm on and preferred stock reported in Compustat quarterly minus any decrea se in redeemable preferred stock. Banyi, Dyl, and Kahle (2005) identify this measure as the mo st accurate proxy for actual shares repurchases, especially for firms with high levels of employ ee stock option exercises. We modify this proxy slightly by considering any increase in treas ury stock as a lower bound for share repurchases. Therefore, if our initial proxy is less than the ch ange in the value of treasury stock, we set share repurchases equal to the change in treasury stoc k. To report this measure as a percentage of shares outstanding, we divide our final proxy by th e firms market capitalization, defined as the number of shares outstanding at the beginning of the fiscal qu arter times the minimum monthly closing price during the current quarter. We obtain data on insider transa ctions reported in SEC form 4 from the Thomson Financial Insider Filing Data Files. For each firm fiscal quarter, we calculate the total dollar values of insider purchases, which are restricted to open market purchases (SEC code P) and sales, which include open market sales and sales to the company (SEC codes S and D). For the cases in which total insider sales exceed total insider purchases by at least 0.005 percent of the firms market capitalization at the end of the prior fiscal quarter, we label the behavior of these insiders as net selling. On th e other hand, if total insider purc hases surpass total insider sales by the same cutoff, then we say that these in siders are net buying. When the difference between insider purchases and sales is trivial or if no insider transactions occur during the
52 quarter, then these cases are classified as neutral.3 Finally, we are interested in the determinants of the relationship between share re purchases and insider trading, which potentially include firm size, market-to-book ratio, option exercise, and corpor ate governance. Firm size is measured as market capitalizat ion, and size deciles are calcula ted each calendar quarter to control for inflation and market fluctuation. S econd, since market-to-book ratios potentially vary with time and across industry, we calculate net market-to-book as the ratio of market capitalization to the book value of common equity of the firm minus the industry median marketto-book ratio for that calendar quarter. Industry is defined by the first two digits of a firms SIC code. These variables are calculate d using data from the fiscal quarter prior to the repurchase measurement quarter. Third, option exercise is the quarterly sum of the number of options exercised times the market pr ice on the day of exercise. Fourth, we use the corporate governance inde x developed by Gompers, Ishii, and Metrick (2003) available through the IRRC governance database. The database covers firms included in the S&P 1500 index as well as ot her publicly traded firms with high market capitalization and/or high institutional ownership leve ls. The governance index (G-index) is updated every two to three years and number of in cluded firms varies over time.4 This index measures the balance of power between managers and shareholders by co unting the number of provisions that reduce shareholders rights, implying that stronger shar eholder rights are associated with low scores. Firms in the bottom decile of governance index (G -index < 6) are classified as democracies and those in the top decile (G-inde x > 12) as dictatorships. Fo r our study these classifications are based on the most recent G-index score. To the extent one believes that firms with weak 3 We also used an alternative definition of net insider trading based on dollar values, with positive $100,000 being a net buyer and negative $100,000 being a net seller. The results were basically the same. 4 The most recent number of firms with a G-index is 1,896 total in 2006. Results for subsamples including the Gindex are smaller by construction.
53 governance may use repurchase programs to aid in siders in disposing of their shares at good prices or to manage earnings up in the face of option exercises, we would expect higher levels of the G-index to be associated with greater pa ired frequencies of in sider selling and share repurchases. In addition, when examining the probability th at a firm conducts a repurchase, we also include variables that have been shown to affect the unconditional odds of doing a share repurchase (see Dittmar (2000)). These include stoc k returns in the prior quarter, cash divided by total assets, cash flow scaled by total assets, and financial le verage defined as the industryadjusted debt to assets ratio. Repurchase/Insider Trading Frequencies Our first empirical question is whether or not in siders tend to trade in the same direction as their firm. Panel A of Figure 2-1 presents univariat e statistics that descri be the general direction of insider trades conditional on whether or not a firm is participating in a non-trivial share repurchase. In our sample firms are repurchasi ng non-trivial amounts of st ock in 14.33 percent of our 365,118 firm-quarter observations. We find th at insiders are net sellers (buyers) for 33.40 percent (18.64 percent) of our sample. This is roughly a 1.79 ratio of net selling firms to net buying firms for the entire sample. Panel B compares the frequency of observed insi der trading activity for firms that repurchase in a quarter versus those that do not, and we find si gnificant differences. Both insider purchasing and insider selling occur with grea ter frequency during the quarter in which the firm buys back shares. Conditional on repurchasing, 41.29 percent of firms have net insider selling, while only 32.08 percent of insiders are net sellers when no repurchasing is done in a quarter. Conditional on repurchasing, 19.98 percent of firms have ne t insider buying, while only 18.42 percent of insiders have net buying when the firm is not re purchasing shares. The rati o of sellers to buyers
54 is 2.07 conditional on repurchasing, which is not a ll that different from the unconditional sample. The obvious inference is that the amount of neutral insi der trading activity takes a big drop when the firm is repurchasing stock. It falls from 49.50 percent conditional on no repurchases to 38.73 percent conditional on repurchasing. In Panel C, conditional on insider trading behavior, we find that firms with net insider selling are the most likely to repurchase shares, follo wed by firms with net insider buying. Firms with net selling have repurchase activity 17.72 percent of the time, firms whose insiders are net buyers have repurchase activity 15.36 percent of the time, and firm s with neutral insider activity have repurchase activity onl y 11.57 percent of the time. While ex-ante we expected to find more fr equent insider buying in conjunction with repurchasing, the surprising result is the increas ed frequency of insider selling in conjunction with repurchasing. Developing a better understanding of these patterns is undertaken in the next section. Determinants of Repurchasing/Insider Trading Frequencies Next, we examine which factors influence in sider trading and repur chasing decisions. We segm ent our sample on four variables: size, market-to-book ratio, option ex ercise, and corporate governance. Assuming that smaller firms are m onitored less closely and that the level of information asymmetry is higher for this group of firms, their managers may be more willing than managers of large firms to engage in the ar guably suspicious behavior of selling stock in their personal accounts while th e company is buying back its stock. On the other hand, perhaps large corporations are more likely to initiate a share repurchase in the first place and to compensate their employees with stock opti ons. In this case, one would observe more simultaneous insider selling and corpor ate repurchases in larger firms.
55 Another factor that could influence insider and corporate transactio ns is the perceived attractiveness of the companys current stock pric e, specifically the market value of the firm relative to its book value. Assuming that the ma rket-to-book ratio is an appropriate proxy for undervaluation, companies and individuals are like ly to buy more and sell less stock (sell more and buy less stock) when market-t o-book ratios are low (high). Next, option exercise may affect the inte rsection of share repurchases and insider transactions. Companies may choose to repurchase shares in order to counter the dilutive effect of stock options being exercise d (Dittmar (2000), Kahle (2002)). In our investigation we do not examine whether the firm has a large number of options outstanding, bu t rather whether the options are actually being exercised. Actual ex ercises give companies extra cash to repurchase shares and a greater amount of diluted shares outstanding for purposes of calculating earnings per share (if the shares are not repurchased). Finally, we examine the possibility that managers use share repurchases as a means to maintain higher stock prices so that they can sell shares at a higher price, particularly in the case where managers are given more discretion and shareholders are less empowered. Therefore, we partition our repurchasing and announcing samples on the Gompers, Ishii, and Metrick (2003) corporate governance index, a pr oxy for the balance of manager versus shareholder power. Univariate Tests Figure 2-2 presents univariate st atistics describing the determin ants of insider trading and actual share repurchase frequencies. Panel A seg men ts the sample on size decile, where size is defined as the product of the sh are price and the number of sh ares outstanding and deciles are calculated quarterly to account for inflation. Results are consistent with size being positively related to the probability that a firm will simu ltaneously repurchase and experience net insider selling. Only 1.43 percent of firms in the lowest size decile engage in repurchasing and insider
56 selling while 15.93 percent of firms in the largest size decile de monstrate this behavior. These probabilities are statistically differe nt at the 1 percent level. To better understand th e dynamics of this interaction, we also examine conditional pr obabilities, i.e. the probability that a firm repurchases conditional on net insider selling/buy ing and the probability that insiders are net sellers/buyers conditional on the firm repurchas ing. We find that, condi tional on net insider selling or buying, large firms are more likely to repurchase. Additionally, for repurchasing firms, net insider selling increases with size while net insider buying decreases with size. Panel B of Figure 2-2 segments our actual repurchase sample on net market-to-book ratio. To the extent that a low level of this ratio is s uggestive of stock undervalua tion, we might expect to find more repurchases and net buying at lower levels of the ratio (Dittmar (2000)). We find that insiders of firms with higher net market-to-book ratios are 35.42 percentage points more likely to be net sellers while the company is repurchasing, but 11.41 percentage points less likely to be net buyers. Furthermore, conditional on repurchasing, firms in the top net market-to-book decile are almost twice as likely as those in the bottom decile to have net insider selling (6.57 percent versus 3.44 percent), but less likely to have net insider buying. As insiders appear to take advantage of potential mispricings in their personal accounts, there is also evidence that company level trades are related to net mark et-to-book. Conditional on net insider selling or buying, companies in the lowest net market-to-book decile are more likely to repurchase shares than those in the highest decile. In Panel C we examine how option exercises re late to insider and company transactions. A firm is identified as having a high level of option exercises if the total quarterly dollar value of options exercised (the number of shares exercised time the market price the day of exercise) is greater than 0.005 percent the market value of the firm at the end of the prior fiscal quarter.
57 Otherwise, the firms option exercise level is considered low. Our resu lts suggest that option exercise is highly correlated with repurchases and insider trades. Firms categorized as having high levels of option exercise are more likely to have simultaneous net selling and repurchasing and are less likely to have simultaneous net buyi ng and repurchasing. Conditional on net insider selling or buying, high option exercise firms are more likely to repurchase, and, conditional on repurchasing, high option exercise fi rms are more likely to be net sellers and less likely to be net buyers. Finally, in Figure 2-3, we calculate the relationship between corporate governance and the intersection of insider transacti ons and share repurchases. We fi nd that corporate governance is a determinant of simultaneous insi der net selling and repur chasing. Dictatorships (firms with a GIndex greater than or equal to 13) are 4.84 per cent more likely to buy back stock and have net insider sales than democracies (firms with a G-In dex less than or equal to 5), a result significant at the 1 percent level. In addition, conditional on net selling, dictatorships are 7.28 percent more likely to repurchase. Similarly, conditional on repurchasing, dictatorships are 8.14 percent more likely to have net insider selling. Logit Regressions In Figure 24 we conduct logit regressions which evaluate the proba bility of simultaneous insider selling and company-level repurchasing. We are most interested in this group of firms because insiders are behaving in a fashion incons istent with the undervaluation signal sent by the company. A potential explanation is that insiders are using sh are repurchases as a means to sustain high stock prices. Panel A includes the full sample and evaluates the probability that a firm has concurrent net insider selling and repur chasing. The dependent variable equals one if the firm simultaneously has (i) non-trivial repurch asing activity (at least 0.25% of its shares outstanding) and (ii) insiders w ho are net sellers (the difference of total insider sales and total
58 insider purchases is greater than 0.005% of last quarters market capitalization), otherwise zero. All models include fixed effects for the calenda r quarter since repurchasing and insider trading may depend on general time trends. Firm-level fixed effects are included for some models, but by construction these estimates must exclude firm s with no within-firm va riation, i.e., firms in the sample that never have simultaneous repurchases and insider selling. While the loss of these firms is a cost of firm fixed effects, the firm fixed effects model may capture idiosyncratic firm preferences and helps us to be tter understand those firms that sometimes repurchase stock and sometimes do not. If firm-level firm effects are not present, then standard errors are clustered at the firm level. We find results consistent with the notion th at the G-index of co rporate governance is positively related to the probability of concurrent insider sales and firm-level buybacks, implying that firms with stronger shareholder protection are less likely to ha ve this occur. Using estimates from Model (6), a one standard deviation increase from the mean value of the natural log of the G-index implies a 18.85 percent increase in the o dds ratio of simultaneous repurchasing and net insider selling (defined as the probability of th is event divided by one minus the probability of this event). Positive coefficients on size decile and option exercise provide evidence consistent with the prior results that large firms and firm s with high levels of option exercise are more likely to engage in simultaneous net insider se lling and share repurchases. Using estimates from Model (6), a one standard deviation increase from the mean value of size decile (natural log of options exercised) implies a 78.05 percent (100. 03 percent) increase in the odds ratio of concurrent repurchases and insider selling. As co ntrol variables, I include other factors that influence a firms decision to repurchase: cash to total assets, cashflow to total assets, leverage (total liabilities divided by to tal assets minus the i ndustry median leverage ratio for the same
59 calendar quarter), and la gged returns (the buy and hold stock re turns for the prior quarter minus the value-weighted returns on th e market during the same time period). Cash and cashflow, which are measured during the fiscal quarter pr ior to the repurchase measurement quarter, are positively related to the probability of repurchasi ng and insider selling while lagged returns are negatively related to this probability. The coeffici ent on the industry-adjusted leverage ratio is negative and significant for models that include firm-level fixed effects. An alternative approach is to explain conditional probabilities of repurchasing given insider selling and the conditional probabi lity of insider selling given re purchases. In a sense, we dont know if the idea to repurchase came first or the idea of inside rs to sell came first. Panel B presents logit regressions using only the subset of firms whose in siders are characterized as net sellers. The dependent variable equals one if the firm is conduc ting non-trivial share repurchases and zero otherwise. Our results are consistent with firms whose shareholders have less power being more likely to buy back stock at the tim e when insiders are se lling. Conditional on net insider sales, a one standard deviation increase from the mean value of the natural log of the Gindex implies a 17.89 percent increase in the odd s ratio of repurchasing using estimates from marginal effects from Model (6). Size and cas hflow are both positively related to the probability of repurchasing while lagged retu rns are negatively correlated with actual repurchases. Option exercise is positively related to the probability of repurchasing for models without firm fixed effects whereas, for models with firm fixed e ffects, the coefficient on cash is positive and significant while the coefficient on le verage is negativ e and significant. Panel C presents the probability of a firm having net insider sales, conditional on conducting a share repurchase. The sample only include s firms conducting non-trivial buybacks and the dependent variable is a binary vari able equal to one if the firms insiders are net sellers. In this
60 case, our proxy for governance is not significantly re lated to the probability of insiders selling. However, the likelihood of net in sider selling conditional on a firm repurchasing is positively related to firm size and option exercise. Robustness of Results to Defi nition of Repurchasing Firm Abnormal repurchases The possibility exists that ma ny firms engage in ongoing repurchase activity which would bring into question the validity of the test statistics due to correlated error terms. We redefine a repurchase quarter as a quarter in which the percentage of shares repurchased exceeds the median level of shares repurchased in the last eight quarters by 0.25% of shares outstanding. In this situation, we are argua bly looking at more opportunistic share repurchases. In Panel A of Figure 2-5, we continue to find results consistent with the notion that firms with stronger shareholder rights are associated with a lower probability of concurrent insider sales and firm-level buybacks. Coefficients on ma rket-to-book ratio, size decile, option exercise, cash, leverage, lagged returns, and cashflow are si milar to those in prior logit models. Panel B presents logit regressions using only the subset of firms with net inside r selling. The dependent variable equals one if the firm is repurchased at least 0.25 percent more than the median value of shares repurchased over the prior 8 quarters. The result that firm s with shareholders having more power are less likely to buy back stock at the time when insiders are selling continues to hold. Size, option exercise, cash, and cas hflow continue to be positively related to the probability of repurchasing while leverage and lagged returns are still negativ ely correlated with actual repurchases. Panel C presents the probability of a firm having net inside r sales, conditional on conducting an abnormally large share repurchase. Again, the relationship between corporate governance and the likelihood of having net insider selling is ambiguous. We continue to find a positive relationship between net insider sell ing and both size and option exercise.
61 Repurchase announcements To further verify the rob ustness of our results we examine insider tr ansactions surrounding announcements of a share repurchase programs. We might expect results to differ between actual share repurchases and repurchase announcements due to issues involving timing and level of commitment. The announcement of a repurchase progr am gives us an exact date at which the company reveals its plan to buy back stock, and, if a firm believes that its stock is undervalued, this sentiment may be strongest at the time of the announcement. Therefore, measuring insider transactions during the 90-day period during and immediately after the announcement should give us insight into how insiders behave at a time when the firm is publicly claiming that investing in their company is a wise decisi on. On the other hand, becau se the announcement of an open market share repurchase plan is not a firm commitment,5 actual repurchases could offer a stronger underval uation signal than announced repurchases. In any event, firms that may wish to falsely signal undervaluation to benefit insiders would exhib it selling activity in the period after the announcement. The sample used in Figure 2-6 consists of firms that announced an open market share repurchase program between January 1, 1987 and December 31, 2007 and whose announcement was reported in the Securities Data Corporation Mergers and Ac quisitions Database. Firms must have quarterly data available from Compustat an d monthly stock price da ta available from CRSP in order to remain in the sample. We examin e insider sales during the 90-day period beginning 2 days prior to each repurchase announcement. Hence, the dependent variable takes on a value of one when insiders of these firms are net sellers. Results are similar to those for the sample of actual repurchasers in that we identify an ambiguous relationshi p between proxies for corporate 5 Firms have no legal obligations to repurchase stock after the announcement of an open market repurchase plan, and Stephens and Weisbach (1998) find that firms acquire on average only 74 to 82 per cent of the announced amount.
62 governance and the probability of insider selling. The natural log of the G-index is not significantly related to the net insider selling, but size and option exercise are again positively related to insider selling. Long Run Stock Returns Calculating long run stock return s following repurchase activity is a natural empirical test for if and how much a companys stock was actually undervalued at the tim e that the actual buyback took place. Some argue that the undervaluation signal inherent in a share repurchase is invalid unless insiders retain their stoc k (Fried (2001)). At a minimum, insider purchases give some credence to an undervaluation si gnal while insider sales potentially detract from it. We further explore this idea by segmenting long run post-re purchase abnormal return s on the direction of insider trades. Methodology Quarterly and annual buy-and-hold abnorma l re turns are calculated using monthly returns from CRSP. Abnormal returns are the differen ce in buy-and-hold return s on the repurchasing firms stock and a size and book-to-market ma tched portfolio. Size and book-to-market are two important factors that explain long-run returns (Fama and French (1992), (1993)). The portfolio of control firms is constructed by identifying all firms in the same size decile as the repurchasing firm during the quarter of the repurchase. We then select the 20 firms that are closest in book-tomarket to the repurchasing firm, and equal weight s are assigned to each firm at the beginning of the investment period. If a firm delists during th e investment period, we assume that all proceeds from that investment are placed in a market portf olio with returns equal to the value-weighted CRSP index. The control portfolio is rebalanced at the beginning of each investment year. Quarter 0 is the fiscal quarter in which share repurchase and inside r trading proxies were
63 evaluated. Annual abnormal returns begin at quarters 2, 6, and 10 for years 1, 2, and 3, respectively. Bootstrapping is used to infer the statistical significance of long-run returns. Specifically, we replace each firm in our sample with a randomly selected firm not in that particular subset of firms, but in the same size and book-to-market quintiles at quarter 0. Using the procedure described in the paragraph above, we calculate the abnormal return s for years 1, 2, and 3 for each firm in our new random sample as well as the mean abnormal returns for the entire sample. This process is repeated for a total of 1000 random samples so that the simulated p-values reported in this paper reflect the percentage of the 1000 ra ndom samples that have mean abnormal returns higher than the mean abnormal return for the our sample. Abnormal Returns Results Figure 2-7 presents abnormal returns for fir m s that repurchased more than 0.25 percent of their shares outstanding or abnormally high amount s of stock during quarter 0. We find that abnormal returns for the repurchasing quarter, fo r the following quarter, and for the three years after the repurchasing quarter ar e correlated with the direction of insider trading. During the quarter of and the quarter immediately after the repurchasing activity we document abnormal buy-and-hold returns that are greate r for firms with net insider buyi ng than those with net insider selling. Furthermore, these results are stronge r in the long term: annual abnormal buy-and-hold returns for years 1, 2, and 3 are 0.41, -0.88, and 1.85 percent for firms with net insider selling compared to 4.95, 2.13, and 1.72 percent for firms w ith net insider buying. Long run returns for firms whose insiders transactions are considered neutral consiste ntly fall between those of net sellers and buyers, implying that abnormal return s are increasing in the extent to which insider trades are consistent with undervaluation.
64 These results hold for firms repurchasing abnormally high amounts of stock as well. In this case, annual abnormal buy-and-hold returns for firms with net sales are positive and significant during year 1, but negative duri ng years 2 and 3. On the other hand, firms with net insider purchases experienced positive and significant re turns of 5.36 and 2.52 percent during years 1 and 2. Firms whose insiders are classified as neutral traders obtain positive and significant abnormal returns for two years as well, though the ma gnitude of these returns is less than that of firms with net insider buying. Panel B of Figure 2-7 further segments the sample on the extremes of the corporate governance measure: dictatorships (G-index > 12) and democracies (G-index < 6). We find that democracies generally outperform dictatorships during the two years after the share repurchase. These results are consistent with the finding of Gompers, Ishii, and Metrick (2003) that firms with stronger shareholder power generally have better long run st ock performance than firm with weaker shareholder rights. For the group of firms that have net insider buying, we find that democracies have positive and marginally significant abnormal return s to the order of 5.22 percent in year one and an additional 6.55 percen t in year two. We also find that democracies experience lower stock returns during the quarter before the share repurchase quarter, implying that democracies may try to time the market more than dictatorships in or der to buy shares that appear to be undervalued. Post-Announcement Abnormal Returns We test the robustness of our long run return s results using a sam ple of open market repurchase announcements as opposed to actual share repurchases. Figure 2-8 presents abnormal returns for announcing firms, partitioned on insi der trading and corporate governance. Insider trading data is from quarter 0, the three calendar mont hs immediately following the month of the
65 announcement. Returns calculations for year 1 be gin at month 4 after the announcement, for year 2 at month 16, and for year 3 at month 28. We arrive at results similar to our previous findings, though in this case returns tend to be higher across the board. On average, firms announcing an open market share repurchase experience positive and significant abnormal returns during the year after the announcement, regardless of the direc tion of insider trading immediately after the announcement. However, the magnitude of these 12-month returns is lowest fo r the group of firms with net insider selling at 3.49 percent compared to 8.56 percent for neutral insiders and 4.05 percent for net buyers. Firms with net insider selling continue to have the lowest abnormal return s in year two at 2.84 percent, while firms with neutral insiders and net buying insiders have abnormal returns of 3.60 percent and 5.26 percent, respectively. In the third year after the repurchase announcement, firms with net insider buying are the only gr oup to have positive and significant abnormal returns while firms with net inside selling have negative abnor mal returns. These result s are consistent with our previous findings that net in sider selling appears to detract from the undervaluation signal inherent in a share repurchase; however, for an nouncing firms we do not necessarily find that insider buying strengthens an undervaluation signal. Finally, we look at how corporate governan ce measures affect long run returns for announcing firms. Similar to firms actually repurchasing stock, announcing firms that are democracies outperform announci ng dictatorships during the fi rst two years following the announcement. Again, the lowest returns are observ ed for dictatorships with negative cumulative returns from year 1 through year 3. Again, the signal seems least credible (after the fact) when insiders in dictatorships are the net sellers. Consistent with our prior findings, democracies with net insider buying immediately following the re purchase announcement have extremely high
66 abnormal returns during the first two years, 8.85 percent during year one and 39.22 percent during year two. The results for year one are economically important, but fail to achieve statistical significance. On the other hand, the resu lts for year two are highly significant. Only 3 of the 1,000 random matched samples had mean returns higher than the net buying/democracy group in year two. Announcing firms with neutral insider trading also perform extremely well, but only during the first year after the announcement. Abnormal buy and hold returns are 23.59 percent for this group and significant at the 0.5 percent level. Interaction Regressions A potential question tha t remains is whether the effect of repurchasing and insider trading on long run returns is additive or interactive. Fo r example, repurchasing firms tend to outperform non-repurchasing firms and firms with net inside r buying tend to outperfor m firms with neutral trading or net insider selling. Hence, is the e ffect of concurrent repur chasing and insider buying stronger than the sum of th e two individual effects? We test this question using re gression analysis on long run buy and hold abnormal returns, presented in Figure 2-9.6 We confirm that repurchasing and net insider buying positively affect long run returns, while net insider selling is negatively related to returns. Using estimates based on two-year abnormal returns from Model (2), we find that repurchasing firms have abnormal returns of 3.1 percent and firm s with net insider buying outperform size and book-to-market matched firms by 3.7 percent. On the other hand, a bnormal returns for firms with net selling are -1.6 percent on average for two years. However, we find no evidence that the eff ect of repurchasing and insider buying is interactive and only weak evidence in support of an interactive eff ect of repurchasing and insider 6 The t-statistics are likely overstated in these regressions due to data overlap. For example, a firm that is in our sample for two consecutive quarters would have two long-run returns observations that are highly correlated. Hence, one must be cautious in interpreting the results.
67 selling. Using estimates based on one-year returns, we identify marginally significant abnormal returns of an additional -1.8 pe rcent for firms that simultaneous ly repurchased and had insider selling. The combined effect of this interaction term and of in sider selling negates the positive effect of repurchasing. Though repurchasing and ne t selling interaction coefficients on models based on two and three-year returns fail to ach ieve statistical significance, they remain economically important. Conclusion Though share repurchases are generally viewed as a signal that a firms stock is trading below its fundam ental value, the validity of th e undervaluation signal is questionable when firm insiders are simultaneously sel ling significant amounts of stock. Th is study adds to the extant literature by examining the genera l direction of inside r trades duri ng actual share repurchasing, the determinants of insider trading around share re purchases, and the effect of insider trading on the strength of the undervaluation signal implied by a buyback. We find that both net insider buying and net insider selling increas e during the quarter of a share repurchase. Similarly, we find stock repurcha ses are more likely in quarters where insiders are selling and least likely when insider trading ac tivity is neutral. While insider buying during a share repurchase is consistent with expectations net insider selling during a share repurchase is more puzzling. We find that the likelihood of concurrent inside r sales and firm-level buybacks increases with size, market-to-book ratio, and option exercise. These results are not only consistent with firms repurchasing stock to counter the dilution associated with stock options, but they also suggest that repurchases are taking place as insiders reduce th eir personal exposure to the firms stock. A potential view of firms that enga ge in both repurchasing and inside r selling is that this is an attempt to sustain or artificially inflate stock prices while certain insiders sell stock. Furthermore,
68 these firms may be overly concerned about th e dilutive impact of options on standard performance measures such as earnings per share. Assuming that these firms can be characterized as less sharehol der friendly, the frequency of simultaneous repurchasing and insider net selling is likely related to proxies fo r corporate governance. Us ing the Gompers, Ishii, and Metrick (2003) governance index, we find re sults consistent with the probability of concurrent repurchases and net insider selling being higher for firms with weaker shareholder protection We also find that th e probability of a firm repurch asing conditional on net insider selling is higher for firms w ith lower shareholder rights. Finally, we investigate the relationship between insider trad ing around share repurchases and the strength of the undervaluation signal implied by the repurchase. We find that insider buying reinforces this signal while insider selling weaken s it. During the quarter of the repurchase, firms with net insider buying have abnormal returns of 2.59 percent, while firms with net selling have abnormal returns of only 0.34 percent. The trend c ontinues in the long run: during the three years beginning two quarters after the repurchasing quarter, firms with net insider buying have abnormal stock returns of 4.95, 2.13, and 1.72 per cent in years 1, 2, and 3, respectively, while firms with net selling have a bnormal returns of 0.41, -0.88, and 1.85 percent. Abnormal returns for firms whose insider trades ar e characterized as neutral fall between the returns of net buyers and net sellers, suggesting that the strength of the undervaluation signal is directly related to contemporaneous insider transactions. We verify the robustness of our results to alte rnative definitions of repurchasing firms. One might argue that a subset of firms will consistently repurchase stock as a means of distributing cash to shareholders. Therefore, we perf orm similar tests which segment firms on abnormal repurchases, defined as repurchasing at least 0.25 percent of share outst anding more than the
69 median value of share repurchases over the prio r 8 quarters. Additionally, we examine a sample of firms that announce a share repurchase. We find that the previously identified factors influencing the intersection of re purchases and insider trading con tinue to be contributing factors for these alternative definitions of repurchasing fi rms. Long run returns results are also similar. Taken together, our findings s uggest that insider selling durin g a share repurchase is not consistent with undervaluation as a motive for re purchasing and that the practice is associated with weaker shareholder rights. Long term inve sting strategies based on repurchases as a signal of undervaluation should incorporat e simultaneous insider transacti ons, which help to clarify the validity of the signal.
70 Insider trading 0.25%> 0.25%Row total 10034921604121953 27.48%5.92%33.40% 15483520264175099 42.41%5.55%47.96% 576111045568066 15.78%2.86%18.64% 31279552323365118 85.67%14.33%100.00% Share repurchases Net selling Neutral Column total Net buying A Figure 2-1: Insider trading and re purchase frequencies. This figur e presents summary statistics on the interaction of actual and announced sh are repurchases with insider transactions using quarterly share repurchase and insider trading data from 1987 until 2007. Actual share repurchases are calculated as the dollar value of purchases of common and preferred stock reported in Compustat quarterly minus any decrease in preferred stock. In the rare case that this figure is le ss than the quarterly increase in treasury stock, then share repurchases equal the in crease in treasury stock. Share repurchases are expressed as a percentage of shares outstanding. Net selling implies that the difference in total insider sa les and total insider purchase s is at least 0.005% of last quarters market capitalization. Net buying indicates that total insider purchases exceed total insider sales by at least 0.005% of the firms market capitalization at the end of the prior quarter. If the absolute value of the difference between insider purchases and insider sale s is less than 0.005% of the firms lagged market capitalization, then the firms insider trading is considered neutral.(A) Proportion of total sample Each box in this panel presents (1) the frequency of the intersection of the events represented in each row and each column and (2) the percentage of total firm-quarter observations that this inters ection represents. (B) Conditional probability of insider trading Panel B presents (1) the frequency of the intersection of the events represented in each row and each column and (2) the probability of observing the denoted insider trading behavior conditiona l on the repurchasing activity presented in that column. (C) Conditional probability of sh are repurchases This panel presents (1) the frequency of the intersection of the events represented in each row and each column and (2) the probabil ity of repurchasing the amount indicated in the column heading, conditional on the observed direction of insider trading.
71 Insider trading 0.25%> 0.25% 10034921604 32.08%41.29% 15483520264 49.50%38.73% 5761110455 18.42%19.98% 31279552323 100.00% 100.00% Share repurchases Net selling Neutral Net buying Column total B Insider trading 0.25%> 0.25%Row total 10034921604121953 82.28%17.72%100.00% 15483520264175099 88.43%11.57%100.00% 576111045568066 84.64%15.36%100.00% Net selling Neutral Net buying Share repurchases C Figure 2-1. Continued
72 Figure 2-2: Repurchases and in sider trading by size, market-t o-book, and option exercise. This figure reports statistics on the intersection of share re purchases with insider buying and selling, using quarterly data from 1987 until 2007. Net selling implies that the difference in total insider sa les and total insider purchase s is greater than 0.005% of last quarters market capita lization. Net buying indicates that total insider purchases exceed total insider sales by at least 0.005% of the firms market capitalization at the end of the prior quarter. Repurchasing firms buy back at least 0.25% of shares outstanding during the fiscal quarter. Values in bold represent the magnitude and significance (p-values) of tests of difference in means for the highest and lowest decile/group. Panel A: Size decile This pa nel segments firms on size, where size is defined as market capitalization and deci les are calculated each calendar quarter. Panel B: Net market-to-book. Panel B exam ines the effect of net market-to-book decile, where net market-to-book is the ra tio of market capitalization to the book value of common equity of the firm minus the industry median market-to-book ratio for that calendar quarter. Panel C: Option exercise. Panel C partitions the data on the level of option exercise. A firm is identifie d as having a high level of option exercise if the total dollar value of options exercise d (the sum of the number of shares times the stock price on the day of exercise) over the past 3 months is greater than 0.005% of the lagged market value of the firm.
73 Insider tradingFrequency % Net selling/ buying and repurchasing % Repurchasing conditional on net selling/buying % Net selling/ buying conditional on repurchasingSize decile 523 1.431 12.660 15.098 1 880 2.411 13.955 22.518 2 10442.859 13.180 24.623 3 12723.484 13.361 28.681 4 14463.962 12.616 32.789 5 17744.857 13.334 37.882 6 22406.134 15.063 44.103 7 27747.599 16.900 49.019 8 384110.520 21.663 53.362 9 581015.928 28.631 62.831 10 14.498 15.970 47.73310-1 (< 0.0001)(< 0.0001)(< 0.0001) 707 1.934 10.314 20.410 1 918 2.515 11.720 23.490 2 10762.946 13.475 25.377 3 11503.150 14.219 25.930 4 10932.994 14.601 24.785 5 10412.850 15.208 22.229 6 10342.832 16.061 20.358 7 10492.873 17.174 18.537 8 12633.459 21.627 17.547 9 11243.081 24.499 12.155 10 1.148 14.185 -8.25510-1 (< 0.0001)(< 0.0001)(< 0.0001) Net selling Net buying A Figure 2-2
74 Insider tradingFrequency % Net selling/ buying and repurchasing % Repurchasing conditional on net selling/buying % Net selling/ buying conditional on repurchasing Net M/B decile 12563.436 17.882 25.165 1 14493.968 17.731 28.412 2 18164.973 20.151 31.256 3 19205.259 18.628 33.672 4 20845.532 17.637 37.631 5 21276.018 17.119 41.837 6 25727.044 17.802 47.958 7 29137.980 18.193 51.971 8 30728.414 17.855 59.339 9 23956.566 15.420 60.587 10 3.130 -2.462 35.42210-1 (< 0.0001)(< 0.0001)(< 0.0001) 11633.181 15.636 23.302 1 11283.089 15.291 22.118 2 13683.746 17.514 23.546 3 13283.637 17.050 23.290 4 11823.138 14.831 21.343 5 11103.141 14.770 21.833 6 992 2.717 14.222 18.497 7 986 2.701 16.809 17.591 8 728 1.994 15.430 14.062 9 470 1.289 10.202 11.890 10 -1.893 -5.434 -11.41210-1 (< 0.0001)(< 0.0001)(< 0.0001) Net selling Net buying B Figure 2-2. Continued
75 Insider tradingFrequency % Net selling/ buying and repurchasing % Repurchasing conditional on net selling/buying % Net selling/ buying conditional on repurchasingOption exercise 74672.825 14.457 22.687 Low 1413714.020 20.109 72.834 High 11.195 5.652 50.147 High Low (<0.0001)(<0.0001)(<0.0001) 78912.986 14.357 23.975 Low 25642.543 19.568 13.210 High -0.443 5.211 -10.766High Low (<0.0001)(<0.0001)(<0.0001) Net selling Net buying C Figure 2-2. Continued
76 GovernanceFrequency % Net selling/buying and repurchasing % Repurchasing conditional on net selling/buying % Net selling/buying conditional on repurchasing Democracy (G 5)993 10.054 21.083 51.080 Other (5 < G < 13)9766 12.981 25.442 58.874 Dictatorship (G 13)1635 14.897 28.361 59.218 4.844 7.278 8.137 (<0.0001) (<0.0001) (<0.0001) Democracy (G 5)348 3.523 21.429 17.901 Other (5 < G < 13)2620 3.483 21.728 15.795 Dictatorship (G 13)469 4.273 24.881 16.987 0.750 3.452 -0.915 (0.0053) (0.0158) (0.4149) Net selling Net buying Dictatorship Democracy Dictatorship Democracy Figure 2-3: Corporate governance. This figure examines the inte raction of share repurchases and insider trading, partitioned on corporate governance factors. The measure of corporate gover nance is the G-index from Go mpers, Ishii, and Metrick (2003). The sample analyzed in this figure is restricted to firms that repurchase at least 0.25% of their shares outstanding and whose insiders are net sellers (the difference of total insider sa les and total insider purchas es is greater than 0.005% of last quarters market capitalization) or ne t buyers (the difference of to tal insider purchases and to tal insider sales is great er than 0.005% of last quarters market capit alization). Values in bold represent the magnitude and significance (p-values) of tests of difference in means fo r dictatorships and democracies.
77 Figure 2-4: Logit models. This figure presents es timates from logit regressions that examine the intersection of insider trans actions and firm-level share repurchases. Ln(G-index) is the natural log of the Gomper s, Ishii, and Metrick (2003) corporate governance index, for which larger values indicate more leni ent governance. Net M/B is the ratio of market capitalization to the book value of common equity of the firm minus the industry median market-to-book ratio for th at calendar quarter. Size is defined as market capitalization and deciles are calculated each calendar quarter. Ln(Options) is the natural log of the total dollar value of options exercised (the number of shares times the stock price on the day of exercise). Cash is cash and short-term investments, scaled by total assets. Cashflow is cashfl ow from operations, scaled by total assets. Leverage is total liabilities scaled by total assets minus the industry median leverage ratio for the same calendar quarter. Lagged returns are the buy a nd hold stock returns for the prior quarter minus the value-weight ed returns on the market during the same time period. Z-statistics are repo rted below the coefficient estimates in parentheses. When firm-level fixed effects are not present, standard errors are clustered at the firm level. *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively. (A) Simultaneous share repurchases and insi der selling. This panel includes the full sample of firms from Com pustat quarterly. The dependent variable is a binary variable equal to one if the firm simu ltaneously has (i) non-trivial repurchasing activity (at least 0.25% of its shares outstanding) and (ii) insiders who are net sellers (the difference of total insi der sales and total insider pur chases is greater than 0.005% of last quarters market capitalization), othe rwise zero. (B) Repur chasing conditional on insider selling. This panel includes the subset of firm s whose insiders are net sellers (the difference of total insider sales and total insider purchases is greater than 0.005% of last quarters market capitalization). The depend ent variable is a binary variable equal to one if the firm has nontrivial repurchasing ac tivity (at least 0.25% of its shares outstanding) and zero, otherw ise. (C) Insider selling conditional on repurchasing. This panel includes the subs et of firms conducti ng non-trivial share repurchasing (at least 0.25% of shares outstanding). The dependent variable is a binary variable equal to one if the firms insiders are net sellers (the difference of total insider sales and total insider purchase s is greater than 0.005% of last quarters market capitalization) and zero, otherwise.
78 (1)(2)(3)(4)(5)(6) Ln(G-index) 0.275**0.235**0.1870.610***0.500***0.590*** (2.574)(2.212)(1.604)(4.514)(3.602)(3.522) Net M/B 0.0000.0000.0000.0000.0000.000 (0.521)(0.443)(0.828)(1.104)(0.895)(0.737) Size decile 0.363***0.228***0.217***0.497***0.369***0.370*** (19.270)(12.060)(10.633)(22.180)(15.832)(13.579) Ln(options) 0.096***0.101*** 0.093***0.097*** (29.001)(28.627) (43.223)(37.855) Cash 0.717***0.425**0.297*1.206***1.108***1.059*** (4.383)(2.565)(1.689)(6.657)(5.944)(4.904) Leverage -0.0230.043-0.065-2.059***-1.977***-2.040*** (-0.142)(0.260)(-0.385)(-13.620)(-12.720)(-11.435) -0.138***-0.513***-0.529***-0.171**-0.511***-0.486*** (-2.725)(-9.159)(-8.354)(-2.568)(-7.270)(-6.114) Cashflow 3.164*** 1.179*** (11.643) (5.039) N 92,08992,02967,54665,16365,01446,591 Adjusted R2 0.0850.1340.1490.0900.1350.144 Time fixed effects YesYesYesYesYesYes Firm fixed effects NoNoNoYesYesYes Lagged returns A Figure 2-4.
79 (1)(2)(3)(4)(5)(6) Ln(G-index)0.319***0.314***0.282**0.480***0.482***0.562*** (2.853)(2.808)(2.278)(3.068)(3.082)(2.970) Net M/B-0.000-0.000-0.000-0.000-0.000-0.000 (-0.609)(-0.563)(-0.385)(-0.115)(-0.121)(-0.154) Size decile0.214***0.199***0.191***0.351***0.355***0.374*** (10.675)(9.699)(8.451)(12.936)(12.978)(11.711) Ln(options)0.013***0.014***-0.003-0.004 (3.918)(3.942)(-1.157)(-1.226) Cash0.294*0.2600.1711.302***1.299***1.285*** (1.658)(1.464)(0.904)(6.177)(6.161)(5.259) Leverage0.1780.1850.092-2.438***-2.446***-2.567*** (1.009)(1.049)(0.507)(-13.543)(-13.573)(-12.439) -0.976***-1.016***-1.003***-1.163***-1.153***-1.127*** (-15.070)(-15.678)(-13.870)(-14.296)(-14.116)(-12.211) Cashflow3.249***1.241*** (10.403) (4.559) N 46,90846,87434,15036,97736,91826,605 Adjusted R2 0.0630.0640.0740.1170.1170.127 Time fixed effects YesYesYesYesYesYes Firm fixed effects NoNoNoYesYesYes Lagged returns B Figure 2-4. Continued
80 (1)(2)(3)(4) Ln(G-index) 0.017-0.0820.533**0.197 (0.177)(-0.783)(2.496)(0.801) Net M/B 0.0010.0010.004*0.002 (0.853)(1.092)(1.956)(1.056) Size decile 0.305***0.129***0.363***0.120*** (17.598)(6.829)(10.694)(3.155) Ln(options) 0.159*** 0.174*** (40.138) (51.593) N 21,37421,35718,85018,827 Adjusted R2 0.0950.2590.0660.260 Time fixed effects YesYesYesYes Firm fixed effects NoNoYesYes C Figure 2-4. Continued
81 Figure 2-5: Abnormally high repurch asing logit models. This figure presents estimates from logit regressions that examine the intersection of insi der transactions and abnormally large share repurchases. The level of repurchasing is classified as abnormally high if the current value exceeds the median value of re purchases over the prior 8 quarters by at least 0.25% of share outstanding. Ln(G-index) is the natural log of the Gompers, Ishii, and Metrick (2003) corporate governance index, for which larger values indicate more lenient governance. Net M/B is the ratio of market capitalization to the book value of common equity of the firm minus the industry median market-to-book ratio for that calendar quarter. Size is defined as market capitalization and deciles are calculated each calendar quarter. Ln(Options ) is the natural log of the total dollar value of options exercised (the number of shares times the stock price on the day of exercise). Cash is cash and short-term invest ments, scaled by total assets. Cashflow is cashflow from operations, scaled by total assets Leverage is total liabilities scaled by total assets minus the industry median leve rage ratio for the same calendar quarter. Lagged returns are the buy and hold stock returns for the prior quarter minus the value-weighted returns on the market dur ing the same time period. Z-statistics are reported below coefficient estimates in parent heses. When the firm-level fixed effects are not present, standard errors are cluste red at the firm level. *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively.(A): Simultaneous abnormally high levels of repurchases and in sider selling. This panel includes the full sample of firms from Com pustat quarterly. The dependent variable is a binary variable equal to one if the firm simulta neously has (i) abnormally high repurchases (the median value of repurchases over the prior 8 quarters by at least 0.25%) and (ii) insiders who are net sellers (the differen ce of total insider sales and total insider purchases is greater than 0.005% of last quarters market capitalization), otherwise zero. (B): Abnormally high repurchasing c onditional on insider selling. This panel includes the subset of firms whose insiders are net sellers (the difference of total insider sales and total insi der purchases is greater than 0.005% of last quarters market capitalization). The dependent variable is a binary variable equal to one if the firm repurchases an abnormally high amount of stock. (C) Insider selling conditional on abnormally high repurchasing. This panel includes the subset of firms conducting abnormally high levels of share repurchases The dependent variable is a binary variable equal to one if the firms inside rs are net sellers (the difference of total insider sales and total insi der purchases is greater than 0.005% of last quarters market capitalization) and zero, otherwise.
82 (1)(2)(3)(4)(5)(6) Ln(G-index) 0.223***0.181**0.1060.578***0.522***0.541*** (2.585)(2.156)(1.123)(3.795)(3.362)(2.925) Net M/B 0.0000.0000.0000.0000.0000.000 (0.195)(0.013)(0.295)(0.839)(0.677)(0.993) Size decile 0.288***0.161***0.156***0.426***0.312***0.318*** (18.780)(10.273)(9.254)(17.313)(12.300)(10.771) Ln(options) 0.088***0.093*** 0.079***0.080*** (28.354)(26.797) (32.680)(28.504) Cash 0.816***0.558***0.383***1.641***1.566***1.446*** (6.061)(4.152)(2.658)(8.313)(7.783)(6.259) Leverage -0.344***-0.299**-0.402***-2.571***-2.491***-2.626*** (-2.608)(-2.304)(-2.989)(-15.082)(-14.359)(-13.253) -0.200***-0.550***-0.557***-0.282***-0.567***-0.537*** (-3.212)(-8.116)(-7.496)(-3.744)(-7.243)(-6.112) Cashflow 2.560*** 0.785*** (9.129) (3.087) N 92,08992,02967,54663,37763,22945,304 Adjusted R2 0.0660.1040.1170.0710.1030.109 Time fixed effects YesYesYesYesYesYes Firm fixed effects NoNoNoYesYesYes Lagged returns A Figure 2-5.
83 (1)(2)(3)(4)(5)(6) Ln(G-index) 0.245***0.241***0.178*0.512***0.516***0.490** (2.872)(2.832)(1.833)(3.085)(3.110)(2.454) Net M/B -0.000-0.000-0.0000.0000.0000.000 (-0.606)(-0.580)(-0.447)(0.139)(0.131)(0.541) Size decile 0.129***0.120***0.120***0.265***0.273***0.295*** (8.195)(7.313)(6.674)(9.425)(9.627)(8.983) Ln(options) 0.008***0.009*** -0.006**-0.009*** (2.762)(2.627) (-2.288)(-2.912) Cash 0.439***0.419***0.274*1.717***1.717***1.615*** (3.143)(2.984)(1.808)(7.871)(7.873)(6.451) Leverage -0.195-0.190-0.282**-2.887***-2.900***-3.083*** (-1.454)(-1.420)(-1.997)(-15.175)(-15.236)(-14.149) -0.979***-1.006***-0.988***-1.078***-1.062***-1.027*** (-12.825)(-13.138)(-11.843)(-12.480)(-12.247)(-10.545) Cashflow 2.456*** 0.686** (8.013) (2.475) N 46,90846,87434,15036,37436,31526,198 Adjusted R2 0.0480.0490.0570.0830.0830.092 Time fixed effects YesYesYesYesYesYes Firm fixed effects NoNoNoYesYesYes Lagged returns B Figure 2-5. Continued
84 (1)(2)(3)(4) Ln(G-index) 0.029-0.1010.452*0.051 (0.318)(-1.013)(1.774)(0.170) Net M/B 0.0010.0000.0030.002 (0.981)(0.962)(1.541)(0.810) Size decile 0.285***0.117***0.362***0.118*** (16.480)(6.080)(9.233)(2.649) Ln(options) 0.162*** 0.176*** (41.537) (41.973) N 14,86414,85312,33212,315 Adjusted R2 0.0960.2640.0810.287 Time fixed effects YesYesYesYes Firm fixed effects NoNoYesYes C Figure 2-5. Continued
85 (1)(2)(3)(4) Ln(G-index) 0.041-0.1260.1660.210 (0.262)(-0.731)(0.284)(0.296) Net M/B 0.0150.0130.0060.003 (1.127)(1.480)(0.600)(0.388) Size decile 0.248***0.115***0.134*0.003 (9.828)(4.180)(1.793)(0.037) Ln(options) 0.155*** 0.175*** (21.293) (13.600) N 2,9072,9051,5471,546 Adjusted R2 0.0870.2390.0900.317 Time fixed effects YesYesYesYes Firm fixed effects NoNoYesYes Figure 2-6: Repurchase announcement logit models. This figure presents estimates from logit regressions using the subset of firms th at announced a share repurchase program. The dependent variable is a binary variable equal to one if, during the 90-day period beginning 2 days prior to the announcement, the firms insiders are net sellers (the difference of total insider sales and total insider purchases is greater than 0.005% of last quarters market capitalization) and zer o, otherwise. Ln(G-index) is the natural log of the Gompers, Ishii, and Metrick (2003) corporate govern ance index, for which larger values indicate more lenient gove rnance. Net M/B is the ratio of market capitalization to the book va lue of common equity of the firm minus the industry median market-to-book ratio for that calenda r quarter. Size is defined as market capitalization and deciles are calculated each calendar quarter. Ln(Options) is the natural log of the total dollar value of opti ons exercised (the number of shares times the stock price on the day of exercise). Ca sh is cash and short-term investments, scaled by total assets. Cashflow is cashfl ow from operations, scaled by total assets. Leverage is total liabilities scaled by total assets minus the industry median leverage ratio for the same calendar quarter. Lagged returns are the buy a nd hold stock returns for the prior quarter minus the value-weight ed returns on the market during the same time period. Z-statistics are re ported below coefficient estimates in parentheses. When firm-level fixed effects are not present, standard errors are clustered at the firm level. *, **, and *** indicate signi ficance at the 10%, 5%, and 1% level, respectively.
86 FrequencyRepurchases Abnormal repurchases Insider trading Quarter -1Quarter 0 Quarter 1Year 1Year 2Year 321681> 0.25% Net selling-1.241%0.341%0.282%0.410%-0.879%-1.848%1.0000.0290.0690.1680.9440.99820404> 0.25% Neutral-2.086%1.034%0.971%3.028%0.582%0.953%1.0000.0000.0000.0000.0710.01710509> 0.25% Net buying-3.913%2.591%0.971%4.946%2.133%1.723%1.0000.0000.0030.0000.0090.03015226HighNet selling-1.534%0.602%0.564%1.203%-0.569%-2.343%1.0000.0040.0110.0230.7170.99815586HighNeutral-2.387%0.909%1.125%3.074%1.054%0.633%1.0000.0000.0010.0000.0420.1528188HighNet buying-4.229%2.557%1.083%5.364%2.519%0.488%1.0000.0000.0050.0000.0080.303 A Figure 2-7: Abnormal returns. Figure 2-7 pr esents quarterly and annual buy-and-hold abnormal returns. Panel A presents returns segmented on the level of repurchases and the direction of insi der trading. Panel B partitions the sample on repurchases, insider trading, and governance ch aracteristics. Repurchase and insider trading data are from quarter 0. Year 1 begins at quarter 2, year 2 at quarter 6, and year 3 at quarter 10. Repurchases are expressed as a percentage of shares outstanding. Repurchases are abnormally high if the current value is at least 0.25% greater than the median value of repurchases over the prior 8 quarters. Net selling implies that the difference of total insider sales and total insider purchases during the current quarter is greater than 0.005% of last quarters market capitalization. Net buying indicates that total insider purchases exceed total insider sales by at le ast 0.005% of the firms market capitaliz ation at the end of the prior quarter. If the absolute value of the difference betw een insider purchases and insider sales is less than 0.005% of the firms lagged market capitalization, then the firms insider trading is consider ed neutral. Dictatorship is a dummy variable equal to one is the corporate governance index is greater than 12 while demo cracy corresponds to a G-index of less than 6. (A): Returns by repurchasing activity and insider tr ading. (B) Returns by repurchasing activ ity, insider trading, and corporate governance
87 FrequencyRepurchases Insider tradingGovernance Quarter -1Quarter 0 Quarter 1Year 1Year 2Year 31635> 0.25%Net sellingDictatorship1.279%0.138%1.488%2.914%-1.489%-2.242%0.970.3760.0070.0230.7860.9993> 0.25%Net sellingDemocracy-2. 013%0.210%0.996%4.168%-1.670%-4.242%0.9820.3590.1210.020.7330.96657> 0.25%NeutralDictatorship2.997%2.097%1.039%2.122%0.215%-5.398%0.9960.0340.1590.1710.4030.95603> 0.25%NeutralDemocracy-6 .708%1.674%3.308%7.234%0.882%0.561%1.0000.0330.0010.010.3130.405469> 0.25%Net buyingDictatorship-5.020%1.690%0.327%2.592%3.419%-1.563%1.0000.0040.2520.0430.0290.795348> 0.25%Net buyingDemocracy-7.105%0.651%1.626%5.215%6.554%-0.471%1.0000.3660.2010.1440.0560.479 B Figure 2-7. Continued
88 Frequency Insider tradingGovernance Quarter -1Quarter 0 Quarter 1Year 1Year 2Year 3 2991Net selling-0.495%-0. 201%0.869%3.491%2.839%-1.336% 0.7790.6200.0820.0110.0410.801 3115Neutral0.535%0.568% 1.944%8.559%3.602%1.424% 0.1940.2250.0020.0000.0060.162 1688Net buying0.275%0.155%1.192%4.050%5.264%2.560% 0.3510.3890.0560.0440.0070.081 193Net sellingDictatorship-2.722% -0.069%0.086%-0.682%1.855%-6.218% 0.8820.5130.4720.5190.3250.867 141Net sellingDemocracy0.406% -0.980%3.778%10.422%3.396%-5.282% 0.4300.6650.0460.0470.2520.854 92NeutralDictatorship2.448% 1.193%2.492%7.395%-3.956%-1.939% 0.2140.3360.1690.1480.7130.556 94NeutralDemocracy-2.093% 1.900%12.621%23.588%-3.176%3.595% 0.6550.2430.0010.0050.6670.290 89Net buyingDictatorship3.818%-2.605%3.621%3.299%-0.222%6.740% 0.1310.7650.1250.3220.4580.191 62Net buyingDemocracy-3.161%0.260%0.260%8.851%39.220%-3.470% 0.7190.4760.4560.1990.0030.552 Figure 2-8: Post-announcement abnormal returns. Figure 8 presents quarterly and annual buy-andhold abnormal returns, partition ed the sample on insider trading and govern ance characteristics, for firms that have announced an open market share repurchase plan. Insider trading data is from quarter 0, the three calendar months immediately following the month of the announcement. Returns calculations for year 1 begin at m onth 4 following the announcement, for year 2 at month 16 and for year 3 at month 28. Net selli ng implies that the difference of total inside r sales and total insider purchases during the quarter is greater than 0.005% of last quarters market capitalization. Net buying indicates that total insider purchases exceed total insider sales by at least 0 .005% of the firms market capitalization at the end of the prior quarter. If the absolute value of the difference between insider purchases and insider sales is less than 0.005% of the firms lagged market capitalization, then the firms in sider trading is considered neut ral. Dictatorship is a dummy variable equal to one is the corporate governance index is greate r than 12 while democracy corresponds to a G-index of less than 6.
89 Year 1 Years 1-2Years 1-3 (1) (2) (3) Repurchasing 0.025*** 0.031** 0.035* (3.169) (2.338) (1.871) Net insider buying 0.027*** 0.037*** 0.032** (3.989) (3.085) (2.009) Net insider selling -0.008* -0.016* -0.018 (-1.773) (-1.661) (-1.223) Repurchasing*Net insider buying -0.007 -0.007 0.021 (-0.563) (-0.328) (0.758) Repurchasing*Net insider selling -0.018* -0.025 -0.036 (-1.941) (-1.576) (-1.547) Constant 0.005 0.001 -0.025** (1.361) (0.090) (-2.261) Number of observations 251,557 243,381 231,210 Adjusted R2 0.0003 0.0002 0.0002 Figure 2-9: Interaction regressi ons. Figure 9 presents OLS re gressions on long run buy-and-hold abnormal returns. The dependent variables are one-year abnormal returns, two-year cumulative abnormal returns, and threeyear cumulative abnormal returns. Repurchase and insider trading data are from quarter 0. Year 1 begins at quarter 2, year 2 at quarter 6, and year 3 at quarter 10. Repurchasing is a binary variable equal to one if the firm repurchases greater th an 0.25% of share outst anding during quarter 0. Net insider selling implies that the differe nce of total insider sales and total insider purchases during the current quarter is grea ter than 0.005% of last quarters market capitalization. Net insider buyi ng indicates that total insi der purchases exceed total insider sales by at least 0.005% of the firms market capita lization at the end of the prior quarter. T-statistics calculated using fi rm-level clustered standard errors are reported below coefficient estimates in parentheses. *, **, and *** indicate significance at the 10%, 5%, a nd 1% level, respectively.
90 APPENDIX A VARIABLE DEFINITIONS Table A-1. Repurchasing variables Percent sought The announced percentage of shar es outstanding reported by the Securities Data Corporation (SDC) that the firm intends to buy back during the repurchase program. If this value is missing but the dollar value of the program is reported, then I calculate percent sought as the dollar value of the repurchase program divided by the market capitalization of the firm one quarter prior to the announcement. Actual percent repurchased The spending on the repurchase of common and preferred stock minus decreases in redeemable preferred stock, scaled by th e minimum price each qu arter, summed over the two years following the announcement or until a subsequent announcement is made. Percent completed The portion of the announced amount that the firm subsequently repurchases, defined as actual percent repurchased divided by percent sought. The SDC database, while considered the most comprehensive source of share repurchase data, occasionally fails to report share repurchases. Hence, I trunc ate percent completed at 100 percent to control for potentially missing subsequent share repurchase announcements Lagged percent completed The percent completed asso ciated with the most r ecent prior announcement. Occasional repurchaser A dummy variable equal to one if the current repurchase announcement is the second announcement in the past five years. This definition is identical to the definition of occasional repurchases presented by Jagannathan and Stephens (2003). Frequent repurchaser A dummy variable equal to one if this announcement is at least the third announcement in the past five years. This definition is identical to the definition of frequent repurchases presented by Jagannathan and Stephens (2003).
91 Table A-2. Returns variables Announcement returns The cumulative stock returns from two days prior to the announcement until two days following the announcement minus cumulative equally weighted returns on the market for the same time period. Pre-announcement returns The cumulative stock returns from 40 days prior to the announcement until 6 days prior to the announcement minus cumulative equally weighted returns on the market for the same time period. Abnormal returns The annual return on the repurcha sing firms stock minus the annual return on a portfolio of control firms over 12 months. Year 1 begins the month following the repurchase announcement. To construct control portfolios, I select the twenty firms in the same size decile as the repurchasing firm at the time of the announcement that are closest in market-to-book to the repurchasing firm. If a matched firm delists during that year, then I assume that the pr oceeds from that investment are invested in a market portfolio with returns equal to the returns on the value-weighted CRSP index. Control portfolios are rebalanced on an annual basis. Two-year abnormal returns The annualized returns on the repurchasing firms stock minus the annualized return on a portfolio of control firms from one month following the prior repurchase announcement until 24 months following the prior announcement. If the current share repurchase announcement occurs within this 24-month period, then abnormal returns calculations are stopped the month before the current announcement. To construct control portfolios, I select the twenty firms in the same size decile as the repurchasing firm at the time of the past announcement that are closest in market-tobook to the repurchasing firm. If a matche d firm delists during that year, then I assume that the proceeds from that investment are invested in a market portfolio with returns equal to the returns on the value-weighted CRSP index. Control portfolio are rebalanced on an annual basis.
92 Table A-3. Firm-specific characteristics Pre-announcement insider trading The dollar value of shares purchased during the three months prior to the announcement minus the dollar value of shar es sold during the same time period, scaled by the sum of the value of shares bought and sold. If no trading took place during the time period, then I set this vari able equal to zero. To ensure that all information concerning insider transactions is publically available, I only use trades for which the report date with the Securities and Exchange Commission precedes the repurchase announcement date by at least two days. Lagged post-announcement insider trading The dollar value of shares purchased minus th e dollar value of shares sold during the 3 months following the prior announcement, scaled by the sum of the value of shares bought and sold. This variable equals zero if no trading took place. To ensure that all information concerning insider transactions is publically available, I only use trades in which the report date with the Secu rities and Exchange Commission precedes the repurchase announcement date by at least two days. Discretionary accruals I use the modified version of the Jones model from Louis and White (2007). Dechow et al. (1995) find that the modified-Jones model of di scretionary accruals provides the most powerful tests of earnings management among existing models. I calculate quarterly total accruals as the change in current assets (data item 40 in Compustat quarterly) minus the change is current liabilities (data item 49) minus the change in cash (data item 36) plus the change in debt in current liabilities (data item 45) minus depreciation (data item 5). If data item 5 is missing, then I calculate depreciation using data item 77 from the statement of cash flow. Using all firms with Compustat quarterly data, I estimate the following model for each calendar quarter and each two-digit SIC-code industry: Total Accrualsi,t / Assetsi,t-1 = 1( Assetsi,t-1) + 2( Salesi,t/Assetsi,t-1) + 3( PPEi,t/Assetsi,t-1) + i,t, with dummy variables for each fiscal quarter. The predicted value represents nondiscretionary accruals and the residual term represents discretionary accruals. Discretionary accruals correspond to the fis cal quarter prior to the share repurchase announcement unless the earnings announ cement follows the share repurchase announcement. In that case, discretionary accruals correspond to two fiscal quarters prior to the share repurchase announcement. Mean analyst recommendation The average of analysts recommendations re ported in I/B/E/S at the time of the repurchase announcement where 1 = Strong Buy, 2 = Buy, 3 = Hold, 4 = Underperform, and 5 = Sell. Net mean analyst recommendation The mean analyst recommendation scaled by the average mean analyst recommendation of all firms listed in I/B/E/S during the calendar year of the repurchase announcement. Percent of forecasts met the percentage of the last fo ur quarterly earnings forecasts reported prior to the repurchase announcement that the firm ha s either met or exceeded. Earning data comes from I/B/E/S.
93 Table A-3. Continued Cash Cash and short-term inve stments, scaled by market capitalization, at the end of the fiscal quarter prior to the announcement. Size decile The market capitalization decile in which the firm falls during the calendar quarter prior to the share repurchase announcement. Deciles are calculated quarterly using the entire database of Compustat quarterly. Decile 1 is the smallest. Net leverage The leverage ratio (total liabilities minus cash and short-term investments, scaled by total assets) minus the industry average leverage ratio at the end of the fiscal quarter prior to the announcement. Industry is defined using the first two digits of a firms SIC code. Net market-to-book Net market-to-book is the market-to-book (market value of equity divided by the book value of equity divided) minus the industry average market-to-book ratio at the end of the fiscal quarter prior to the announcement. Industry is defined using the first two digits of a firms SIC code. Table A-4. Dividend announcement dummie s and earnings announcement dummies Positive surprise A dummy variable equal to one if the announcing firm releases quarterly earnings results are above the mean of analysts fo recasts between the time period beginning two days prior to the repurchase announcement and ending two days after the announcement. Analyst forecast data is from I/B/E/S. Negative surprise A dummy variable equal to one if the announcing firm releases quarterly earnings results are below the mean of analysts forecasts between the time period beginning two days prior to the repurchase announcement and ending two days after the announcement. Analyst forecast data is from I/B/E/S. No surprise A dummy variable equal to one if the announcing firm releases quarterly earnings results are at the mean of analysts for ecasts between the time period beginning two days prior to the repurchase announcement and ending two days after the announcement. Analyst forecast data is from I/B/E/S. Increase A dummy variable equal to one if th e firm announced an increase in dividends between the time period beginning two days prior to the repurchase announcement and ending two days after the announcement. Decrease A dummy variable equal to one if the firm announced an decrease in dividends between the time period beginning two days prior to the repurchase announcement and ending two days after the announcement. No change A dummy variable equal to one if the firm announced a co ntinuation in current dividends between the time period beginning two days prior to the repurchase announcement and ending two da ys after the announcement. Special dividend A dummy variable equal to one it the firm announced a special dividend between the time period beginning two days prior to the repurchase announcement and ending two days after the announcement.
94 APPENDIX B LONG-RUN BUY-AND-HOLD R ETURN S CALCULATIONS All long-run returns are calcu lated using monthly returns from CRSP. The investment period begins the month following the buyback announcement; hence, one -year returns span from month 1 to 12, two-year re turns from month 13 to 24, and three-year returns from month 25 to 36. Abnormal returns are the difference in buy-and-hold returns on th e repurchasing firms stock and a size and book-to-market matche d portfolio. Size and book-to-market are two important factors that explain long-run returns (Fama and French (1992), (1993)). The portfolio of control firms is constructed by identifying all firms in the same size decile as the repurchasing firm during the month of the announcement. I then select the 20 firms that are closest in book-tomarket to the repurchasing firm, and equal weight s are assigned to each firm at the beginning of the investment period. If a firm delists during the investment period, th en I assume that all proceeds from that investment are placed in a ma rket portfolio with returns equal to the valueweighted CRSP index. The control portfolio is rebalanced at the beginning of each investment year. Bootstrapping is used to infer the statistical significance of long-run returns. Specifically, I replace each firm in my repurchasing sample with a randomly selected non-repurchasing firm in the same size and book-to-market quintiles at the time of the buyback announcement. Using the procedure described in the paragraph above, I ca lculate the abnormal returns for years 1 through 3 for each firm in my new random sample as well as the mean abnormal returns for the entire sample. This process is repeated fo r a total of 1,000 random samples. The p-values reported in this paper reflect the percentage of the 1,000 ra ndom samples that have mean abnormal returns higher than the mean abnormal re turn for the repurchasing sample.
95 A similar procedure is used to infer the statis tical significance of the difference in means in abnormal returns. For example, to calculate p -values associated with means returns of Group A and Group B, I pair the mean abnormal returns calculated using 1,000 random samples of firms that correspond to Group A in time periods, size quintiles, and book-to-market quintiles with the mean abnormal returns calculated using 1,000 random samples of firms that correspond to Group B. I then calculate the difference for each 1,000 pairs of averages. The p-value represents the percentage of the 1,000 differences in means that exceed the sample difference in means.
96 REFERENCES Alm azan, Andres, Sanjay Banerji, and Adolfo de Motta, 2005, Attracting attention: Cheap managerial talk and costly market monitoring, Working Paper. Bagwell, Laurie, 1991, Share repurchase and takeover deterrence, RAND Journal of Economics 22, 72-88. Bagwell, Laurie, and John Shoven, 1988, Share repurchase and acquisitions, Corporate Takeovers: Causes and Consequences Baker, Kent, Gary Powell, and Theodore Veit, 2003, Why companies use open-market repurchases: A managerial perspective, Quarterly Review of Economics and Finance 43, 483504. Banyi, Monica, Edward Dyl, and Kathleen Ka hle, 2005, Measuring shar e repurchases, Working Paper. Barber, Brad, Reuven Lehavy, and Brett Trueman, 2007, Comparing the stock recommendation performance of investment banks and independent research firms, Journal of Financial Economics 85, 490-517. Barber, Brad, and John Lyon, 1997, Detecting long-r un abnormal stock returns: The empirical power and specification of test statistics, Journal of Financial Economics 43, 341-372. Bergstresser, Daniel and Thomas Philippon, 2005, CEO incentives and earnings management, Journal of Financial Economics 80, 511-529. Bhattacharya, Utpal and Amy Dittmar, 2004, Costle ss versus costly signaling in capital markets: theory and evidence, Working paper. Billett, Matthew, and Hui Xue, 2007, The takeov er deterrent effect of open market share repurchases, Journal of Finance 62, 1827-1850. Brav, Alon, John Graham, Campbell Harvey, an d Roni Michaely, 2005, Payout policy in the 21st century, Journal of Financial Economics 77, 483-527. Chan, Konan, David Ikenberry, and Inmoo Lee, 2004, Economic sources of gain in stock repurchases, Journal of Financial and Quantitative Analysis 39, 461-479. Chan, Konan, David Ikenberry, Inmoo Lee, a nd Yanzhi Wang, 2006, Share repurchases as a tool to mislead investors: Evidence from earni ngs quality and stock performance, Working paper. Chen, Qi, Jennifer Francis, and Wei Jiang, 2005, Investor learning about analyst predictive ability, Journal of Accounting and Economics 39, 3-24.
97 Comment, Robert, and Gregg Jarrell, 1991, The relative signaling power of Dutch-auction and fixed-price self-tender offers a nd open-market share repurchases, Journal of Finance 46, 12431271. Cook, Douglas, Laurie Krigman, and Chris L each, 2004, On the timing and execution of open market repurchases, Review of Financial Studies 17, 463-498. Core, John, Wayne Guay, Scott Richardson, and R odrigo Verdi, 2006, Stock market anomalies: what can we learn from repurchases and insider trading?, Review of Accounting Studies 11,4970. Cudd, Mike, Harold Davis, and Marcelo Edua rdo, 2006, Mimicking behavior in repurchase decisions, Journal of Behavioral Finance 7, 222-229. DeAngelo, Harry, and Linda De Angelo, 2007, Capital structure, payout policy, and financial flexibility, Working paper. Dechow, Patricia M., Richard Sloan, and Am y Sweeney, 1995, Detecting earnings management, The Accounting Review 70, 193-225. Dittmar, Amy, 2000, Why do firms repurchase stock?, Journal of Business 73, 331-355. DMello, Ranjan, and Pervin Shroff, 2000, Equity Undervaluation and decisions related to repurchase tender offers: An empirical investigation, Journal of Finance 55, 2399-2424. Fenn, George, and Nellie Liang, 2001, Corporat e payout policy and managerial incentives, Journal of Financial Economics 60, 45-72. Fried, Jesse, 2001, Open market repurchases: signaling or managerial opportunism?, Theoretical Inquiries in Law 2, 865-894. Gompers, Paul, Joy Ishii, and Andrew Metrick, 2003, Corporate governance and equity prices, Quarterly Journal of Economics 118, 107-155. Gong, Guojin, Henock Louis, and Amy Sun, 2008, Ea rnings management and firm performance following open-market repurchases, Journal of Finance (forthcoming). Grullon, Gustavo, and Roni Michaely, 2002, Dividends, share repurchases, and the substitution hypothesis, Journal of Finance 57, 1649-1684. Grullon, Gustavo, and Roni Michaely, 2004, The information content of share repurchase programs, Journal of Finance 59, 651-680. Hirshleifer, David, and Danling Jiang, 2007, Commonality in misvaluation, equity financing, and the cross section of stock returns, Working paper. Huddart, Steven, and Henock Louis, 2006, Stock re turns, earnings management, and insider trading during the 1990s stock market bubble, Working paper.
98 Hutton, Amy, and Phillip Stocken, 2007, Effect of Reputation of on the credibility of management forecasts, Working Paper. Ikenberry, David, Josef Lakonishok, and Theo Vermaelen, 1995, Market underreactions to open market share repurchases, Journal of Financial Economics 39, 181-208. Ikenberry, David, Josef Lakonishok, and Theo Vermaelen, 2000, Stock repurchases in Canada: Performance and strategic trading, Journal of Finance 55, 2373-2397. Ikenberry, David, and Theo Vermaelen, 1996, The option to repurchase, Financial Management 25, 9-24. Jagannathan, Murali, and Clifford Stephens, 2003, Motives for multiple open-market repurchase programs, Financial Management 32, 71-91. Jagannathan, Murali, Clifford Stephens, and Mi chael Weisbach, 2000, Financial flexibility and the choice between dividends and stock repurchases, Journal of Financial Economics 57, 355384. Jensen, Michael, 1986, Agency costs of free cash flow, American Economic Review 76, 323-329. John, Kose, and Larry Lang, 1991, Insider tradin g and dividend announcements: Theory and evidence, Journal of Finance 46, 1361-1389. Kahle, Kathleen, 2002, When a buyback isnt a buyb ack: open market repurchases and employee options, Journal of Financial Economics 63, 235-261. Lakonishok, Josef, and Theo Vermaelen, 1990, Anomalous price behavior around repurchase tender offers, Journal of Finance 45, 455-477. Lee, Scott, Wayne Mikkelson, and Megan Partch, 1992, Managers trading around stock repurchases, Journal of Finance 47, 1947-1961. Li, Kai, and William McNally, 2003, The decision to repurchase, announcement returns, and insider holdings: a c onditional event study, Journal of Applied Finance 9, 55-70. Louis, Henock, and Dahlia Robinson, 2005, Do mana gers credibly use accruals to signal private information? Evidence from the pricing of discretionary accruals around stock splits, Journal of Accounting and Economics 39, 361-380. Louis, Henock, Amy Sun, and Hal White, 2008, Insi der trading around repur chase tender offers: Timing versus informed trading, Working paper Louis, Henock, and Hal White, 2007, Do managers in tentionally use repurchase tender offers to signal private information? Evidence from firm financial reporting behavior, Journal of Financial Economics 85, 205-233.
99 Massa, Massimo, Zahid Rehman, and Theo Vermaelen, 2007, Mimicking repurchases, Journal of Financial Economics 84, 624-666. Ng, Jeffrey, Irem Tuna, and Rodrigo Verdi, 2008, Management forecast credibility and underreaction to news, Working Paper. Nohel, Tom, and Vefa Tarhan, 1998, Share repurchases and firm performance: new evidence on the agency costs of free cash flow, Journal of Financial Economics 49, 187-222. Peyer, Urs, and Theo Vermaelen, 2005, The natu re and persistence of buyback anomalies, Working paper. Raad, Elias, and H.K. Wu, 1995, Insider tradi ng effects on stock returns around open-market stock repurchase announcements: an empirical study, Journal of Financial Research 18, 45-57. Seyhun, Nejat H., 2000, Investment Intelligen ce from Insider Trading, MIT Press. Stephens, Clifford, and Michael Weisbach, 1998, Actual share reacquisitions in open-market repurchase programs, Journal of Finance 53, 313-333. Vermaelen, Theo, 1981, Common stock re purchases and market signalling, Journal of Financial Economics 9, 139-183.
100 BIOGRAPHICAL SKETCH Alice received her BS in mathematics and BA in French from the College of Charleston before obtaining her Internationa l MBA (formerly masters in international business studies) and Certificate of Applied Statistics from the Univer sity of South Carolina. Her research interests include corporate finance, payout policy, and behavioral finance. Upon graduation, she will become an assistant professor of fi nance at the University of Kentucky.