Managerial Overconfidence and Share Repurchase

Transcription

1 Managerial Overconfidence and Share Repurchase Pei-Gi Shu Department of Business Administration, Fu Jen Catholic University Yin-Hua Yeh Department of Finance and International Buiness, Fu Jen Catholic University Yu-Hui Su Correspongind author Department of Accounting, Soochow University Tsui-Lin Chiang Graduate Institute of Business Administration, Fu Jen Catholic University

2 Managerial Overconfidence and Share Repurchase Abstract We use the keywords in press portrayals to gauge the level of managerial overconfidence. We argue the higher the level of managerial overconfidence, the more likely managers feel their firms value is undervalued, and the higher intensity of share repurchase programs being launched. Our empirical results using a data of 2,744 share-repurchase programs launched by 783 listed firms in Taiwan verify the positive relation between managieral overconfidence and the program intensity manifested in scale, execution, frequency, and price difference. The relation is sustainable after controlling alternative motives and endogeneity issue. However, the market lends little lenity to the share repurchase programs launched by overconfident managers. This is due to the fact that firms with a high level of managerial overconfidence tend to be firms that are not undervalued. Keywords: Managerial Overconfidence, Share Repurchase 1

3 2

4 1. Introduction Overconfidence is a prominent stylized fact from the social psychology literature indicating that individuals tend to overstate their acumen relative to the average (Larwood and Whittaker, 1977; Svenson, 1981; Alicke et al., 1985). This inclination is further reinforced by the self-serving bias that portrays individuals custom of attributing good outcomes to their actions and bad outcomes to luck (Miller and Ross, 1975). The notion of overconfidence in affecting managerial actions was first brought by Roll (1986) who proposes the hubris hypothesis arguing that managers are overconfident so as to overbid acquired targets. Heaton (2002) further elaborates the effect of overconfidence in affecting managerial actions. Empirically, the impact of managerial overconfidence has been examined in different corporate scenarios. For example, Malmediar and Tate (2005b) illustrate overconfidence heightens the sensitivity of corporate investment to cash flow and that is more pronounced to be found among equity-dependent firms 1. Ben-David et al. (2007) indicate that managerial overconfidence results in more investments, a more sensitive investment-cash relation, a lower cash dividend payout, and a higher long-term leverage. 1 Overconfident CEOs systematically overestimate the value of their investment projects and the current value of the firm. If they have sufficient internal funds for investment and are not disciplined by the capital market or corporate governance mechanisms, they overinvest relative to the first best. If they do not have sufficient internal funds, however, they are reluctant to issue new equity because they perceive the stock of their company to be undervalued by the market. As a result, they curb their investment. This would result in a tightened cash-investment relation. 3

5 In this study we investigate how managerial overconfidence is related to share repurchase program. Share repurchase has only been popularly used by the U.S. firms within the last two decades and was further fueled by an explosion in the use of open market repurchase programs. In the 1990s, this movement went global as countries like Canada and the U.K. also saw an increase in repurchase activity. Some countries such as Germany, Taiwan, Hong Kong, and Japan, that formerly prohibited stock repurchases adopted provisions allowing resident firms to repurchase equity in the open market for the first time. The magnitude of share repurchase programs was magnificent. The size of announced intentions to repurchase in the U.S. market in was roughly $750 billion worth of stock. Moreover, in 1998 its size exceeded cash dividend for the first time (Grullon and Michaely, 2002). The rationale of relating the two constructs is intuitive in the sense that overconfident managers tend to perceive their firms are undervalued and therefore to launch a share repurchase program aiming to fixing the price discrepancy. The higher the overconfidence level the larger the perceived price discrepancy to be fixed and a higher intensity of the repurchase program. The intensity of share repurchase program in this study is manifested in scale, execution rate, frequency, and price difference. We expect to find a positive relation between degree of managerial overconfidence and 4

6 intensity of share repurchase program. Our empirical results using 2,744 repurchase programs launched by 783 listed firms in Taiwan basically support this postulation. The positive relation between managerial overconfidence and the intensity of share repurchase program is sustainable after controlling other discerning variables 2 and the possible endogeneity issue. We further explore how market responds and to the share repurchase programs launched by overconfident managers. We explore whether firms with overconfident managers are undervalued so as to enhance the intensity of share repurchase programs. The result shows that these overconfident share-repurchase launchers are less likely to be undervalued. The result comes no surprise that the market responds negatively to managerial overconfidence. Investors in the first beginning appreciate large-scale share repurchase programs because the signal of undervaluation is reinforced. However, as the passage of time up to the maturity of a share repurchase program, outside investors realize that the intensified signal is affected by managerial overconfidence and therefore attach lower valuation. The potential contributions of this study are multifold. First, the relation between 2 The motives that firms engage in repurchasing shares are extenstively dicussed in literature. The first one is related to agency concern arguing that repurchasing share is to distribute excess cash flow (Jensen, 1986). This arugment is supported by the finding of a positive relation between repurchases and levels of cash flow (Stephens and Weisbach, 1998). The second one is related to signaling undervaluation (Vermaelen, 1981). The third one is related to leverage adjustment arguing that repurchasing stock could be used to increase firm s leverage ratio (Bagwell and Shoven, 1988; Opler and Titman, 1996). The fourth is related to fend off unwanted takeover attempts (Bagwell, 1991). 5

7 managerial overconfidence and share repurchase program has been sparsely explored in the literature. If any, the prior studies informally indicate the two are positively correlated without providing further empirical evidence. For example, Ben-David et al. (2007) using top financial executives estimation of the stock market prospect as a proxy of managerial overconfidence propose that overconfident firms are less likely to pay dividends and more likely to repurchase shares. Our finding supports this postulation. Taking a further step, we explore the relation between extent of managerial overconfidence and intensity of share repurchase program. The exploration provides readers a better understanding on how managerial inclination affects corporate policies. Second, we advance the press-based measurement of overconfidence proposed by Malmendiar and Tate (2005, 2008) and Brown and Sarma (2007). Specifically, our measure of managerial overconfidence tags along after the incumbent manager rather than the firms that launch share repurchase programs. Third, we find market responds negatively toward the programs launched by overconfident managers. As the passage of time outside investors are able to discern the overconfidence-induced intensified share repurchase programs and attach lower valuation. Finally, we find that repeated share-repurchase launchers and/or high-execution launchers are associated with lower market appraisal. The rest of this 6

8 paper is organized as follow. Section 2 reviews the related literature. Section 3 outlines the data, variables, and empirical models. Section 4 reports the empirical results. Section 5 discusses the robustness of the empirical results. Section 6 concludes. 2. Literature and Hypotheses In this section we firstly brief share repurchase program in Taiwan, followed by the literature review of share repurchase and overconfidence. Based on the literature, we develop hypotheses to explore the relations between managerial overconfidence and share repurchase program as well as investors reaction to open market share repurchases. 2.1 Share Repurchase Programs in Taiwan Comparing to other countries, Taiwan imposes stricter rules on repurchases disclosure and a short execution period of two months. The Article 28-2 of the Securities and Exchange Law was initially approved by the Legislative Yuan of Taiwan on June 30, 2000 and the Regulations Governing Share Repurchase by Exchange-Listed and OTC-Listed Companies applied to all listed firms on August 7, According to the rules, firms preparing to launch a repurchase program need to file their purposes that 7

9 are in accordance with at least one of the followings: (1) to prepare for employees bonus shares; (2) as reserves for the execution of derivatives such as warrants and convertibles; and (3) to protect shareholders rights with the repurchased shares to be written off against owners equity. The share repurchase announcement should be made to public in 2 days after board approval. In the meanwhile, the repurchasing firms should disclose the price range of the repurchase program. The ceiling price of the range is set at 150% of the higher of the thirty-day average price and the ten-day average price prior to the date of board approval. The floor price is set at the 70% of the close price at the date of board approval. In general, open market repurchases will start 2-3 days after public announcement, which is 4-5 days after board approval. The program should be finalized in 2 months and reported to the Financial Supervisory Commission the exact execution in 5 days after program expiration. Another board approval is needed when firms expect to extend their unexercised program. The ceiling of a repurchase program is confined to 10% of total issued shares and to the summation of retained earnings and capital surplus. The number of firms engaging in share repurchase programs was fast accumulated to 783 listed firms and 2,744 programs in December Quite a number of them were repeated players, being recorded of more than 15 times in our sampling period. 8

10 Share repurchase is pervasively recognized a tool for saving the undervalued shares. It is therefore welcome by the listed companies as well as the government endeavoring to uphold the market. A recent noted example is the share repurchase programs launch by Asustek and Pegatron; the two firms had just split and on June 24, Asustek launched a share repurchase program buying 10 million shares aiming to counterbalance the negative news such as CEO changeover, auditor changeover, and losses in exchange. Pegatron also claimed a share repurchase program of NT3.075 trillion buying back 75 million shares which is equivalent to 3.28% issued shares. The two programs haven t shown much of a pronounced impact in price rebound, however Motives of Share Repurchase Prior studies propose several hypotheses to explain why firms would engage in share repurchase programs. The first one is the excess capital hypothesis arguing that firms having excess cash can distribute it to shareholders through repurchasing shares (Easterbrook 1984; and Jensen 1986). Compared to dividend payout, repurchasing stock is associated with the flexibility in distribution and tax advantage and is therefore preferred to dividend payout 3. We expect firms with high levels of excess cash or cash flow to repurchase stock. The first two corresponding variables are free cash flow and 3 Brennan and Thakor (1990) and Lucas and McDonald (1996) present models in which repurchases are preferred to dividends for larger distributions. 9

11 firm s cash level. The second hypothesis is the undervaluation hypothesis indicating that outsiders undervalue the firm s share due to information asymmetry. According to this hypothesis, the market interprets share repurchase as an indication that the stock is undervalued (e.g. Asquith and Mullins, 1986; Comment and Jarrell, 1991; Dann et al., 1991; Hertzel and Jain, 1991; Lee et al., 1992). Ikenberry et al. (1995) imply that low market-to-book firms are more likely to be undervalued and therefore followed by a significant positive abnormal return up to 4 years subsequent to the repurchase announcement. The corresponding variable is book-to-market equity which is assumed to be positively correlated firm s launching share repurchase program. The third hypothesis is the optimal leverage hypothesis indicating that firms might use share repurchase to achieve the optimal capital structure when the current debt ratio is below the target one (Bagwell and Shoven 1989; and Opler and Titman 1996). In this paper we use debt deviation being calculated as the difference between current debt ratio and target debt ratio of Hovakimain et al. (2004) as the corresponding variable for this hypothesis. The fourth hypothesis is management incentive hypothesis indicating managers when holding more stock options are of particular interest in preserving price level 10

12 which could be attained via shares repurchases. Moreover, using treasury stocks provided for exercised options would not dilute the per-share value of the firm (Dunsby 1994; Jolls 1996; and Fenn and Liang 1997). A firm that compensates its executives with a large number of stock options may find it beneficial to repurchase stock. The corresponding variable to this hypothesis is executive stock options. The fifth hypothesis is the takeover deterrence hypothesis arguing that a repurchase can be used as a takeover defense because a repurchase can increase the lowest price for which the stock is available (Bagwell 1991). Thus, firms that are at a higher risk of becoming takeover targets are more likely to repurchase stock. The corresponding variable to this hypothesis is the takeover dummy. All the aforementioned variables are included in the models in examining the relation between managerial overconfidence and share repurchase programs Managerial Overconfidence Overconfidence relates to several streams of psychology literature. The first one is the better than average effect which illustrates the tendency of individuals to think themselves above average on positive characteristics (e.g. Kruger, 1999; Alicke et al., 1995; Alicke et al., 1985; Svenson, 1981). The tendency has been extensively examined and empirically supported (e.g. Svenson, 1981; Cooper et al., 1988). The better than 11

13 average effect is further reinforced by attribution bias which indicates that individuals have a self-serving inclination as to attribute good outcomes to their actions and bad outcomes to luck (Miller and Ross, 1975; Feather and Simon, 1971). Overconfidence is particularly applied to top managers for several reasons. First, skilled individuals who are in lack of an obvious comparison group tend to be overconfident (Kruger, 1999; Camerer and Lovallo, 1999). Second, overconfidence is more pronouncedly found when outcomes are abstractly defined (Moore and Kim, 2003; Nisbett and Ross, 1980). Third, individuals are the most optimistic about outcomes in which they believe are under their control (Langer, 1975) or to which they are highly committed (Weinstein, 1980). These attributes perfectly fit for top managers so that they persistently overestimate their own skills and are therefore too optimistic about the outcomes of their decisions. The overconfidence argument relating to managers is parallel to the hubris hypothesis proposed by Roll (1986). It also relates to the frameworks of Heaton (2002) and Landier and Thesmar (2004), who model managers who overestimate the probability of project success. In corporate finance, this form of overconfidence has been applied to different forums: contracting with managers (Gervais et al., 2003), succession tournaments (Goel and Thakor, 2000), dividend policy (Deshmukh et al., 12

14 2010), mergers (Malmendier and Tate, 2008), investment-cash sensitivity (Malmendier and Tate, 2005). The biggest challenge for empirical studies is to construct a plausible measure of overconfidence. Malmendiar and Tate (2005) use CEOs personal portfolio transactions to signal their beliefs about firm s future performance. This approach is dubbed as revealed beliefs approach which requires detailed information about CEOs personal portfolio transactions in their companies stock and options. Since CEOs human capital is invested in their company, they should diversify their portfolios even for modest risk aversion. CEOs against the rule of diversification such as holding more of company s stocks, reluctance to exercise deep-in-the-money options, or withholding the exercise of options to the final year of duration, are deemed as overconfident. Alternatively, they propose the press-based overconfidence measure which captures how outsiders perceive the CEO. In this study we adopt this press-based measure by tabulating the positive words against natural and negative words according to the definition of Malmendiar and Tate (2008). Our measures are time varying and tagging along after managers so that they are able to capture the notion of acquired managerial overconfidence. 13

15 2.4. Managerial Overconfidence and Share Repurchase The relation between managerial overconfidence and share repurchase is elaborated as follows. Overconfident managers tend to overestimate firm s future cash flow or underestimate the volatility associated with the cash flow. As a result, they tend to believe that the market misprices the value of the firm and the value of equity. Share repurchase programs are therefore launched for the purpose of fixing the price discrepancy gauged by overconfident managers. An extension of this argument is that overconfident managers who believe the undervaluation premise tend to overinvest in projects, such as value-detrimental mergers (Malmendiar and Tate, 2008), to prefer internal sources of financing (Malmendiar and Tate, 2005), and to be less likely to pay dividends (Deshmukh et al., 2010). In this study we take an up-close look on how the level of overconfidence affects the intensity of share repurchase programs. The level of managerial overconfidence is gauged by the relative counts of the positive words of overconfidence to the words of natural and passive words. In this sense, managerial overconfidence is a continuous variable being able to contrast the degree of overconfidence specific to a certain manager. The intensity of share repurchase program in this study is ramified into scale (percentage of shares repurchase and percentage of dollar amount repurchase), 14

16 execution, frequency, and price difference 4. The higher the level of managerial overconfidence the larger the anticipated price discrepancy to be fixed, the much intensified programs to be launched. We therefore assume that the both are positively correlated. Hypothesis 1: The level of managerial overconfidence is positively correlated with the intensity of share repurchase programs. A noteworthy issue for further investigations is how investors response to the share repurchases programs launched by overconfident managers. Deshmukh et al. (2010) find that the magnitude of the positive market reaction to a dividend-increase announcement is lower for firms managed by overconfident CEOs. This implies that investors are with reserve toward the signal that is jointly affected by the content of the signal and managerial overconfidence. The positive effect of share repurchase would at least be partially offset by managerial overconfidence. The extreme case is that 4 The price difference could be alternatively defined as the difference between the close price on the day of board approval and the mid-price of announced price range. Unfortunately, the price range is preset in the range between the cap of 150% of the reference price (higher of 10-day and 30-day average prior to the announcement day) and the floor of 70% of the close price of the announcement day. The price difference is therefore less meaningful to be further explored. In this study, we alternatively define the price difference as the difference between the mid-price of the price range and the 3-day average price after the execution of a share repurchase program. The price difference is used to signify program intensity on one hand, and to capture market response on the other. 15

17 managerial overconfidence dominates the positive effect of share repurchase program when investors find the firms launching share repurchase programs are firms that are less likely to be undervalued. That is, managers acquire overconfidence through their successful experiences. They learn to be overconfident. Overconfidence in turn makes them feel that the current price level is undervalued so as to launch share repurchase programs aiming to fix the price discrepancy. However, this does not concur with outside investors perception. The share repurchase program therefore is expected to end up with a negative price effect. Hypothesis 2: The market responds negatively toward the share repurchase programs launched by overconfident managers.. 3. Data and Variables Our data of share repurchases are collected from Taiwan Economic Journal (TEJ) in the sampling period from August 2000 through December Share repurchase program was firstly allowed for listed companies in August After excluding financial firms which are subject to different regulations, the final sample consists of 16

18 2,744 share-repurchase programs launched by 783 listed firms. Table 1 summarizes the sample distribution based on industry breakdown (panel A), frequency (panel B), and yearly breakdown (panel C). The results show that electronic appliance firms (379, 13.81%) comprise the most cases of the sample, followed by computer and peripheral firms (354, 12.9%) and semiconductor firms (322, 11.73%). Around 30% of the firms (233) only launch one share repurchase program in the sampling period. However, the statistics also shows that 146 firms (19%) are frequent players that launch more than 5 share-repurchase programs in this period. The very extreme case is 36 programs being announced in 8 years in the sampling period. The pervasive and repeated use of share repurchase could be something beyond signaling undervaluation because the market is less likely to undervalue shares at all time and across all firms. The industry breakdown in panel A illustrates that share repurchase program is more applicable to high-tech firms than firms in traditional manufacturing industry. For example, there are only 2 repurchase programs in the glass and ceramic industry. <<Insert Table 1 Here>> 17

19 Referring to Malmendier and Tate (2005) and Brown and Sarma (2007), we use the press portrayals to gauge managerial overconfidence. The source of data is collected from the News Clip file of Taiwan Economic Journal (TEJ) in the period from January 1996 and October This file having not been available until January 1996 is a complied version of all press reports specific to each company. The measure is associated with an advantage that it could capture the perception of outsiders so that any strategic move made by the manager would be reflected in the market price which mingles with the action itself and the manager s behavioral inclination. Specifically, the press-based managerial overconfidence (MOC) is defined as follows 5. MOC: a/(a+b) subject to managerial changeover (1) where (a) refers to the counts of the word confident and the word optimistic, and (b) is the counts of words not confident, not optimistic, reliable, cautious, practical, frugal and steady. We note that the measures are accumulated up to the yearend prior to share repurchase announcement so that the overconfidence measure for each manager are different through the passage of time. This might ameliorate the concern of possible autocorrelation for overconfidence measures. Moreover, the 5 We also alternatively define managerial overconfidence as a/b, a/(a+b), a/b subject to managerial changeover, a dummy that is assigned 1 when a>b and 0 otherwise, and a dummy that is subject to managerial changeover and is assigned 1 when a>(a+b) and 0 otherwise. The results from alternative definitions are qualitatively similar and are not reported for brevity, 18

20 measure is tagged along managerial changeover, meaning that the count of keywords would start over when there is a managerial changeover. The results in table 2 indicate that the average degree of overconfidence is The average number of share repurchase programs is The average percentage of repurchased shares is 2.03% with respect to total outstanding shares and 1.63% with respect to total equity value. On average, the announced programs were executed 70% of what they have been announced. The median execution rate is 85%. We define the price difference as difference between the mid-price of announced price range (MP) and the average price three days after the program (P +3 ) divided by the average price three days after the program. Price-Difference= ( (MP-P +3 )/P +3 ) (2) This measure is to gauge the price decline after share repurchase programs. If share repurchase programs were motived by managerial overconfidence, we would expect to find the number is positive and is positively correlated with the level of managerial overconfidence. The average price difference is 17.8%. 19

21 The statistics of the rest control variables are also reported in Table 2. The average book-to-market equity is The average free cash flow and cash level are 0.04 and 0.15; the two are surrogates of excess capital with which firms could distribute excess cash through share repurchase. The optimal leverage hypothesis indicates that firms use share repurchase programs to achieve the optimal capital structure when the current debt ratio is below the target one (Bagwell and shoven, 1988; Opler and Titman, 1996). Debt deviation is the difference between current debt level and optimal debt level which is estimated via the following regression: Debt* it = β 0 + β 1 MB i,t + β 2 SRT i(t-1) + β 3 ROA it + β 4 NOLC it + β 5 Size it + β 6 TanA it + β 7 SE it +β 8 RDE it +β 9 IL it + ε it (5 ) where Debt* is the optimal debt level, MB is the yearend market-to-book equity; SRT is the annual return, ROA is the net income divided by averaged total assets; NOLC is deferred taxes divided by total assets; Size is the nature logarithm of yearend total assets; TanA is the fixed assets minus depreciation and divided by total assets; SE is the selling and administrative expenses divided by revenue; RDE is R&D expenses 20

22 divided by total revenue; IL is the median debt-to-assets ratio for all firms in the same industry. The average debt deviation is -2.5%, implying that on average firms are below the target debt level and that urges firms to raise debt ratio through share repurchase programs, according to the optimal leverage hypothesis. The averaged stock option is 1%. The literature indicates that stock option could be one of the manager s concerns for preventing diluting share price. In our sampling period, 2% of the sampling firms are associated with rumors of mergers. When being potential targets, firms repurchasing shares could raise the margin price if the demand curve of shares is negatively sloped. The average payout being perceived as an alternative conduit of expending cash other than share repurchase is 45.8%. Using the traditional event study approach, we calculate the abnormal returns in different windows. Alternative windows (-1, 1), (-5, 5), (0, 20), (0, 40), and (0, 60) are selected to capture the information impact around the dates of repurchase program. The window (-5, 5) is to capture the impact from board approval to open market repurchase which usually takes place 4-5 days after approval. The window (0, 40) is to cope with the 2-month duration of a repurchase program. The window (0, 60) is selected to capture the overall market reaction after the execution of a share repurchase program. The corresponding averaged cumulative abnormal returns for the 5 windows are 1.16%, 21

23 -0.29%, 4.97%, 6.10% and 6.70%, respectively. The result illustrates a monotonic increase in abnormal returns. <<Insert Table 2 Here>> 4. Empirical Results In table 3 we investigate managerial overconfidence and valuation. In panel A we divide the sampling firms into quintiles based on size and market-to-book equity, and calculate the mean measures of press-based overconfidence. The purpose of so doing is to have a glimpse of how these overconfident firms or managers are characterized. The results show that in general the firms of median size and with high growth potential (high market-to-book equity) are associated with the highest overconfidence measure. Assuming that firms follow the life cycle paradigm, the statistics indicate that managers of firms in the stage of accelerated growth or of firms that have grown up to median scale exhibit the highest overconfidence. It could be case that managers in these firms have accumulated experiences of successes and therefore are nurtured to be overconfident. They have a strong sense that the firm s shares are undervalued and therefore are motivated to launch repurchase programs. However, as compared with 22

24 firms with low market-to-book equity, firms with high market-to-book equity tend to be glamour stocks that are less likely to be undervalued. In panel B we evaluate the firm s instrinsic value and investigate how managerial overconfidence is related to the valuation. We use two approaches to gauge the fair value of share repurchasing firms. The first one refers to Purnanandam and Swaminathan (2004) that the fair value for share repurchasing firms are calculated using price multiples, such as price-to-book and price-to-sales, of non-announced industry peers that are close in size, and then compare this fair value (V) to the average pre-announced price in the window (-10,-6) (P). [ (P/V ) Book = (P/B) Stock Repurchase Firm /(P/B) Matching Firm ] [ (P/V ) Sales = (P/S) Stock Repurchase Firm /(P/S) Matching Firm ] (4 ) The second approach refers to Rhodes-Kropf et al.(2005), the fair value is gauged by conducting the following cross-section regression V it =α 0 + α 1 b it + α 2 NI + it + α 3 I (<0) (NI) + it + α 4 LEV it + ε i, (5 ) 23

25 where b denotes book value of assets, NI denotes net income and LEV denotes leverage ratio. (P/V) CS denotes the ratio of the average price in the window (-10, -6) to the fair value estimated from the cross sectional regression. The value (P/V ) Book,(P/V ) Sales and (P/V) CS denotes overvaluation when they are higher than 1 and undervaluation when they are lower than 1. The results illustrated in panel B indicate that the average (P/V) book for high-moc firms (1.188) is significantly highr than that for low-moc firms (1.051). The average (P/V) sales for high-moc firms (1.279) is also higher than that for low-moc firms (1.252), albeit the difference is insignficant. Using the alternative metric referring to Rhodes-Kropf et al.(2005)we find that the average (P/V) CS is higher for high-moc firms than for their low-moc counterparts. The overall result indicates that firms with overconfident managers are less likely to be undervalued. In panel C we further investigate the relation between managerial overconfidence and the attributes of a share repurchase program. The attributes of interest include number of repurchases, percentage of shares repurchase, percentage of amount repurchase, percentage of execution, and price difference. The result from the univariate test that high-moc firms are associated with more share repurchase programs (4.315) 24

26 than that of low-moc firms (3.804). The percentage of share repurchase is higher for high-moc firms (2.066%) than for low-moc firms (1.898%). The price difference, being defined as the scaled difference between the mid-price of announced price range and the average price three days after the program, is higher for high-moc firms (0.214) than for low-moc firms (0.155). High-MOC firms on average are aslo associated with a higher percentage of dollar amount and a higher rate of execution than those of low-moc firms. However, the differences are insignficant. The overall result from table 3 indicates high-moc firms are less likely to be undervalued than low-moc firms. High-MOC firms are more active to launch more share repurchase programs, to claim a higher percentage of shares to be repurchased, and set a higher target price. We also conduct tests of differences in medians. The overall picture in intact. <<Insert Table 3 Here>> The empirical model referred to Dittmar (2000) is formulated as follow. 25

27 SR it =β 0 + β 1 MOC i,t + β 2 Size i,t + β 3 ROE i,t + β 4 BM i,t + β 5 Deviation i,(t-1) + β 6 Options i,(t-1) + β 7 Cash i,t + β 8 FCF i,t + β 9 INVO i,t + β 10 D(Takeover) i,(t-1) + β 11 Payout i,(t-1) + β 12 CGI i,t + ε i,t (6) where SR denotes the attributes of share repurchase program, MOC denotes managerial overconfidence, size denote the nature logarithm of firm s assets, BM denotes book-to-market equity, ROE is return on earrings, FCF denotes free cash flow being defined as earning before taxes and depreciation minus taxes, interest and cash dividends and divided by initial assets, Cash denotes cash divided by initial assets. Stock option is the expected employee stock options divided by the total outstanding shares in the prior yearend. INVO is the sum of market equity and total debt divided by total assets in the yearend. D (takeover) is a dummy that is assigned the value 1 when the underlying firm is associated with the rumor of being an acquirer or a target in the year or previous year and 0 otherwise. These variables are included in the models for controlling the alternative motives of launching share repurchase programs. Corporate governance index (CGI) being included in the model is to explore whether corporate governance plays a counterbalance role in deterring managerial overconfidence in launching share repurchase programs. Consolidating different dimensions into a composite index is to comprehensively capture the overall quality of a 26

28 firm s governance structure (Boehren and Oedegaard, 2003). The index is constructed by summing the following dummies: board control (1 for smaller than sample median and 0 otherwise), supervisory control (1 for smaller than sample median and 0 otherwise), voting-cash deviation (1 for smaller than sample median and 0 otherwise), pledge ratio (1 for smaller than sample median and 0 otherwise), related party sales ((1 for smaller than sample median and 0 otherwise), related party purchase (1 for smaller than sample median and 0 otherwise), related party loan (1 for smaller than sample median and 0 otherwise), related party guarantee (1 for smaller than sample median and 0 otherwise),cash flow rights (1 for higher than sample median and 0 otherwise), cash/voting ratio (1 for higher than sample median and 0 otherwise). Therefore, CGI is in the range between 0 and 10. Deviation refers to the difference between firm s debt ratio and optimal debt ratio estimated via the model of Hovakimain et al. (2004). In table 4 we conduct tobit regressions of the percentage of share repurchased. The results indicate that MOC is positively correlated with the percentage of shares repurchased. Moreover, firm size, the natural logarithm of firm s assets, is negatively correlated with the claimed portion of shares repurchase. This could be due to the possibility that large firms are more likely to be the focal point of investors and are 27

29 therefore less likely to be undervalued by the market. We also find that firms with high book-to-market equity tend to launch a higher large portion of shares repurchased, which is consistent with the undervaluation hypothesis. Again, we find that debt deviation is positively correlated with the percentage of share repurchase programs, which is inconsistent with the optimal leverage hypothesis while consistent with the overconfidence hyothesis. <<Insert Table 4>> In table 5 we conduct tobit regressions of the percentage of the dollar amount repurchased, an alternative measure of the scale of share repurchase program that simultaneously considers the number of shares and per share price. The results show that overconfidence is positively correlated with the proportion of dollar amount of share repurchase. The results from the rest control variables are qualitatively similar to the results from table 4. The only exception is that the takeover dummy is positively correlated with the dollar amount of share repurchase, indicating that firms that encounter the threat of takeover tend to launch large-scale share repurchase programs. 28

30 The results from table 4 and table 5 support our speculation that managerial overconfidence is one of major driving forces for managers to launch large-scale share repurchase program. Moreover, checking with the possible hypotheses that dictate firm s launching share repurchase program we find the undervaluation hypothesis receives supporting evidence in the sense that firms having been neglected by the market (high book-to-market equity and low investment potential) tend to launch large-scale. Moreover, the takeover deterrence hypothesis is also supported that firms being as merger targets tend to launch large-scale share repurchase program for deterring possible raiders. However, the excess capital hypothesis, the leverage adjustment hypothesis and management incentive hypothesis are not supported herein. These hypotheses might affect firms decision of whether to launch share repurchase program but not the decision of program scale. Since a large portion of the listed firms in our sampling have experiences of launching share repurchase programs, it would be less sensible to investigate the issue of whether firms would like to launch share repurchase programs. <<Insert Table 5 Here>> 29

31 In table 6 we investigate how managerial overconfidence is related to the percentage of execution of the repurchase program. The result is consistent with our intuition that overconfident managers tend to execute a high portion of what they have announced. MOC is significantly and positively correlated with the execution rate. Moreover, the investment opportunity is negatively correlated with the execution rate, which further renders supporting evidence to the undervaluation hypothesis because firms with insufficient investment opportunities tend to be out-of-favor stocks having shares been undervalued by the market. In order to uphold the price level to the anticipated level, overconfident managers tend to execute a high percentage of the claimed repurchased shares. The result also indicates that takeover dummy is negatively correlated with the percentage of execution. One possibility is that the repurchase signal is strong enough so that the anticipated price level set by overconfident managers would soon be attained. The spare cash from unexercised program could be saved for alternative promising investments. Since a low execution of the repurchase program could attain the anticipated price level, the firm could exempt from the threat of takeover. <<Insert Table 6 Here>> 30

32 In table 7 we investigate how managerial overconfidence effects on the number of share repurchase programs. The statistics in table 1 shows that more than 70% of announcement firms launching more than 1 share repurchase program. It is intuitively appealing to investigate the issue of how managerial overconfidence is related to the times of repurchase programs. The acquired theorists indicate that managers could learn to be overconfident through prior experiences. If this is the case, the successes from prior share repurchase programs will enhance the level of overconfidence and therefore further encourages the managers to launch another repurchase program. However, the result in table 7 fails to exhibit a significant positive correlation between number of repurchase program and the level of managerial overconfidence, albeit the regression coefficient is positive as expected. The results from the rest control variables such as size, book-to-market equity, and investment opportunity are consistent with prior findings that small-size and out-of-favor stocks need more runs of repurchases to reach the anticipated price level set by the overconfident managers. Moreover, firm s return on equity is negatively correlated with the number of programs, implying that good performing firms are less likely to be undervalued and therefore necessitate few runs of repurchase programs. 31

33 Free cash flow is negatively correlated with the number of repurchase programs, which is inconsistent with the prediction of excess capital hypothesis indicating firms having excess cash tend to distribute it to shareholders through repurchasing shares. Without jeopardizing the whole picture, one possibility to account for this phenomenon is that firms concentrate their use of free cash in the first few runs of repurchase programs. Moreover, we find that debt deviation is negatively correlated with the number of repurchases. This is consistent with the optimal leverage hypothesis implying the more deviation from the optimal capital structure the more runs of repurchase programs to fill in the gap. However, the result is inconsistent with the findings in table 5 where debt deviation is positively correlated with the percentage of dollar amount repurchased. A possible reconciliation is that firms aiming to adjust capital structure through share repurchase are through piecemeal repurchases rather than one-shot repurchase. This is also consistent with the general understanding that an abrupt adjustment in capital structure is associated with cost. Moreover, the takeover dummy is negatively associated with number of repurchase programs. Combining with the finding in table 5, we find that firms confronting with takeover threat might engage in one-shot large-amount repurchase to resist the threat. 32

34 << Insert Table 7 Here>> In table 8 we explore how managerial overconfidence is related to price difference, being defined the price rebound from the mid-price of announced price range to the average price after the program. The price difference is supposed to be positively correlated with the level of managerial overconfidence because the market would sooner or latter come to its sense that the exaggerated signal emitted by overconfident managers need to be rectified. The higher the level of managerial overconfidence, the larger the rectification is needed. The result basically supports our conjecture. Managerial overconfidence is positively correlated with the price difference. Moreover, debt deviation is also positively correlated with the price difference. We find that firms with a higher debt ratio are more zealous to engage in share repurchase programs. In other words, high leverage could be another manifest of overconfidence in the sense that managers are reluctant to issue undervalued equity and that results in a high debt ratio, or that they are confident about firm value as to issue more debt to finance the repurchase programs. High level of overconfidence is therefore associated with a significant price rebound. Finally, the takeover dummy is positively correlated with the price difference. This 33

35 could be due to the fact that outside investors interpret the share repurchase program is more for managerial entrenchment than for signaling undervaluation. <<Insert Table 8 Here>> In table 9 we further relate managerial overconfidence and the number of share repurchase programs to the announcement returns in different windows. The regression results show that managerial overconfidence is negatively correlated with CAR (0, 40) and CAR (0, 60). The relations between managerial overconfidence and short event windows, such as CAR (-1, 1), CAR (-5, 5) and CAR (0, 20), are insignificant. This implies that the market is relatively neutral to the share repurchase announcements made by overconfident managers. As the passage of time up to the deadline of the program execution, the market is able to discern managerial overconfidence and attaches lower valuation to the share repurchase programs launched by overconfident managers. Moreover, we find that the number of programs is negatively correlated with CAR (-1, 1) and CAR (-5, 5). This could be due to the case that managers intermittently issue promising news such as share repurchases to enjoy the positive price effect. 34

36 However, the marginal positive effect decreases as the number of signals accumulated. The results from the other control variables indicate that book-to-market equity is positively correlated with the announcement returns. This is consistent with the undervaluation argument that small-size and high book-to-market firms tend to be undervalued and therefore are associated with higher announcement returns. Moreover, we find that free cash flow is positively correlated with CAR (-5, 5), CAR (0, 40) and CAR (0, 60). Cash level is also positively correlated with CAR. The result is consistent with the excess capital hypothesis of Easterbrook (1984) and Jensen (1986), implying that repurchasing stock is flexible approach to distribute excess cash and to fix the agency problem embedded in the excessive cash. Moreover, we find investment opportunity is negatively correlated with the announcement returns. This could be explained by the undervaluation hypothesis that firms with high growth potentials are less likely to be undervalued, so their share repurchase programs are associated with lower valuation. From an alternative perspective, the market assumes that these high-growth firms should have alternative investment opportunities better than share repurchase programs and therefore attaches a lower appraisal on the share repurchase programs launched by high-growth firms. Finally, the takeover dummy is negatively associated with the returns, implying that the 35

37 share repurchase programs aiming to entrenching management are less welcomed by the market. <<Insert Table 9 Here>> We also include managerial overconfidence and the attributes of a share repurchase program, one at a time, in the regression analysis. The depedent variables are the announcment returns in various windows. All the other control variables are the same as in table 9. For brevity we only report the regression coefficient of managerial overconfidence and the coefficient of the attribute variables in table 10. The result shows that MOC is significantly negative in the regression of CAR (0, 40) and CAR (0, 60) when the attribute variables are in turn being included in the models. The result indicates that the percentage of shares claimed to be repurchased and the percentage of dollar amount are positively correlated with CAR (-1, 1) and CAR (-5, 5), while are negatively correlated with CAR (0, 20) and CAR (0, 40). Moreover, the percentage of execution is negatively correlated with CAR (0, 20), CAR (0, 40) and CAR (0, 60). This could be due to case that in the first beginning the firms that launch 36

38 large-scale repurchase programs, in terms of percentage of shares and percentage of dollar amount, draw favorable market appraisal in that outside investors interpret the embedded signal to be credible. However, as the passage of time the market in general realize the the large-scale repurchase is more for managerial overconfidence than for signal of credibility. As a matter fact, both the regression coefficient of MOC and the regression coefficient of scale variables are signficnatly negatively in the regression analysis of CAR (0, 40) and CAR (0, 60). We find that the percentage of execution is significantly negative in the regression analysis of CAR (0, 20), CAR (0, 40), and CAR (0, 60). We provide a possible explanation to this: the execution of a share repurchase program is costly and should be stopped when the anticipated price level is met. High percentage of execution, from the ex-post perspective, indicates the program is not that welcomed by the market and therefore associated with a lower abnormal return. In an unreported result we also explore the possible interactions between managerial overconfidence and the attributes of share repurchase programs, and include the interactions into the regression analysis of the announcement return. We find that the interaction between MOC and ln (times) is negative. The interaction between MOC and the percentage of shares repurchased is also negative. However, the interaction between between MOC and the percentage of execution is positive. Since all the interactions are 37

39 only marginally significant at 10% level, we would not overemphasize their effects on the announcement returns. <<Insert Table 10 Here>> 5. Robustness Check In the prior sections we deem managerial overconfidence is an exogenous variable. However, if managerial overconfidence is an acquired nature that is affected by other firm s characteristics, the result that managerial overconfidence drives the impulse to launch large-scale, numerous and high-execution programs might not be sustainable. For robustness, we conduct two-stage regression and use the first-stage regression to isolate the residual managerial overconfidence that is unrelated to other firm s characteristics. The first regression model is as follows. MOC i,t =β 0 + β 1 SIZE i,t + β 2 ROE i,t + β 3 BM i,t + β 4 FCF i,t + ε i. (7) The residual term is then included in the second stage regression as follows. 38