What Drives the Declining Wealth Effect of Subsequent Share Repurchase Announcements?

Recent academic studies document that open market share repurchase announcement period returns are much lower than those reported in early studies. This study finds that the lower announcement returns are attributed to repeat announcements that dominate the sample in the recent period. The announcement period return of the initial repurchase program launched by a repeat repurchasing firm on average is positive. However, the more this firm repeats its repurchase announcement, the lower its announcement period abnormal returns. Further tests reveal that firms with negative past repurchase announcement returns are more likely to repeat their repurchase programs, which result in lower announcement returns. Our results are consistent with overconfident managers drive the lower repurchase announcement returns.

What drives the declining wealth effect of subsequent share repurchase announcements?

Introduction
The purpose of this study is to examine why some firms keep repeating open-market share repurchase programs despite declining subsequent announcement period returns. Open market share repurchase programs have become the alternative to cash dividends as a payout method to distribute cash to stockholders (Grullon and Michaely, 2004). Even though such programs are not binding commitments, early studies document that stock markets react favourably to such announcements, with abnormal announcement period returns of more than 3% (see for example, Vermaelen, 1981;Ikenberry, Lakonishok and Vermaelen, 1995). 1 Recent studies, however, report that the magnitude The main motive for open market share repurchases is usually either to buy back undervalued stocks (Lakonishok and Vermaelen, 1990;Peyer and Vermaelen, 2005), or to distribute temporary free cash flows, in lieu of dividends, to shareholders (Stephens and Weisbach, 1998;Dittmar, 2000;Skinner, 2008). Other theories used to explain repurchases are: (1) to improve their leverage ratios (Bagwell and Shoven 1989); (2) to discourage takeover attempts (Bagwell, 1991); and (3) to counter the dilution effect of stock option plans (Fenn and Liang, 2001;Kahle, 2002). been increasing. Fu and Huang (2016) report that the frequency to announce subsequent share repurchase programs in the US since 2003 has increased three times compared to that in the period before 2003. 2 Another possible explanation on the lower open market repurchase announcement period returns could be due to lower past repurchase completion rates. Bonaime (2012), Mishra, Racine and Luke (2011) and Chang, Chen and Chen (2010) argue that when a firm launches an open market share repurchase program, but does not follow it through or repurchase less than the number of shares announced in the program, the market considers the firm as having a bad reputation. Consequently, the market will react less favourably when the firm announces a subsequent open market share repurchase program. Low past completion rates, however, cannot explain why some firms keep repeating open-market repurchase programs. If the motivation to repurchase is related to stock undervaluation, which is one of the most common motives to launch an open-market repurchase program, a positive market reaction to the repurchase announcement may be sufficient for the announcing firm to not fully follow through on its announced repurchase plan, and therefore We find that, during the sample period from 1996 to 2014, the number of repeat or subsequent announcements has increased over the years. The number of repeat announcements in a year, on average, is 68% of total open market repurchase announcements, and only in 1998 and 1999 was the number of repeat announcements (48% and 45% respectively) less than half of total announcements. We also find that, consistent with Jagannathan and Stephens (2003), firms that repeat their repurchase announcements, compared to those that do not repeat share repurchase announcements, are not only larger, more profitable, less undervalued, have significantly lower abnormal announcement-period returns (0.86% vs. 1.99% respectively), but also have more free cash flows and have higher growth opportunity. Therefore, it is an empirical question why these firms, having large cash flows and high growth opportunity, would invest in their own stocks.
Next, we document that, firms that repeat open market share repurchase programs experience an average positive announcement-period abnormal return of 1.33% from their initial announcements. However, when these firms repeat the repurchase program, the market reacts less favourably to the second announcement, that the announcement period return drops significantly to 0.9%. The decline in the announcement period return continues further 3 The finance literature documents that some managers are prone to self-attribution bias, which leads them to be overconfident. Ben-David, Graham, and Harvey (2007) find that among other corporate actions, are more likely to be associated with less efficient investments. Hayward and Hambrick (1997) find that CEO's hubris (or exaggerated self-confidence) is strongly positively associated with the size of premiums paid for acquisitions. Malmendier and Tate (2008) find evidence consistent with the view that overconfident CEOs overestimate their ability to generate returns. Hence, they overpay for target companies and undertake value-destroying mergers.
Another managerial trait -CEO narcissism has also been shown to be positively related to the number and size of acquisitions (Chatterjee and Hambrick, 2007). Billett and Qian (2008) and Karolyi et al. (2015) find evidence consistent with overconfident managers explain the declining returns of serial acquirers. Recent work by Aktas et al., (2016) show that both acquirer and target CEO narcissism affect the characteristics of the takeover process. No prior studies have studied stock repurchases using the lens of managerial hubris.
to only 0.27% when firms repeat more than 4 subsequent open market repurchase announcements.
We next examine if managerial behavioural bias can explain why these firms keep repeating their open market share repurchase programs, even though the market reacts less positively to such repeat announcements. We find that firms with negative past repurchase announcement returns, are likely to repeat their repurchase programs, and these repeat repurchase announcements are negatively associated with the subsequent announcement period returns. These results are consistent with overconfident managers repeating share repurchase announcements due to self-attribution bias that lower their shareholders' wealth.
Our study sheds more light in explaining the decline in the open-market share repurchase announcement returns over the years and contributes to the literature in several aspects. First, we employ a more recent sample period and report that the number of open market share repurchase announcements made by repeat repurchasers has significantly increased over the recent years, suggesting that there is a systematic change in the open market share repurchase behaviour during the recent period. Second, our study is the first to report that, not only the announcement returns of repeat announcements are lower than those of non-repeat announcements, but also that subsequent announcement returns are significantly decreasing in magnitude. Third, we propose a managerial motivation bias to explain repeated open-market share repurchase programs. Our results are consistent with the self-attribution bias hypothesis, that overconfident managers, supported with excess cash flows, are more likely to repeat their open market share repurchase program, even though their decisions may not be in the best interest of their shareholders.
The rest of the paper is organized as follows. In the second section, we discuss the hypotheses. In section 3, we describe our data collection processes and our sample. In section 4, we report our empirical results. In section 5, we offer our concluding comments.

Hypothesis development
Employing a US sample from 1986 to 1996, Jagannathan and Stephens (2003) find that the market reacts less positively to announcements made by repeat repurchasers than to those announced by non-repeat repurchasers. Our study covers a sample period from 1996 to 2014. To test if this result still holds in a more recent period, our first hypothesis is: H1: Announcement period returns of repeat repurchase announcements are lower than those of non-repeat announcements.
Managers who develop overconfidence bias can make suboptimal payout decisions because they have large excess cash flows. Free cash flows can be used to pay a return to capital providers through a direct cash payment (coupon interest or dividends) or in the repurchase of securities (bonds or shares). Malmendier and Tate (2005) find that overconfident managers are significantly more responsive to cash flows and tend to overinvest. 4 Therefore, managers of a firm with a large amount of free cash flows, may feel confident in their ability to meet not only the firm's obligations, but also, with the success of the initial repurchase program, feel over-confident in buying back their existing shares, even though the firm has a high growth opportunity to invest in the real sector instead. Jagannathan and Stephens (2003) find that repeat repurchasing firms have different firm characteristics from those of non-repeat repurchasing firms. They report that repeat repurchasing firms are 4 Lehn and Poulsen (1989) find that firms with undistributed free cash flows tend to pay a significant premium for stock repurchases related to going private transactions. Howe, He, and Kao (1992) investigate whether Jensen's (1986) free cash flow theory explains the market reaction to tender offer share repurchases and specially designated dividends where the cash distribution is not expected to be repeated. They find that free cash flows do not explain the announcement returns very well and conclude that their results are inconsistent with Jensen's free cash flow hypothesis but consistent with the information signaling hypothesis. They offer the entrenchment hypothesis as a possible explanation for their conflicting findings with those of Jensen's. larger, have more cash flows, are less undervalued, but have higher growth opportunities than those of non-repeat repurchasing firms. Thus, our second hypothesis is: H2: Repeat repurchasing firms are bigger, more profitable, less undervalued, have larger excess cash flows, but have higher growth opportunity than those of nonrepeat repurchasers.
Conceptually, managers should make payout decisions that maximize shareholders' wealth. When announcing the initial repurchase program, a manager of a repeat repurchasing firm may not have developed a hubris bias yet. The literature however, documents that when managers experience a success, they are prone to be overconfident. Therefore, when the initial open market share repurchase program is successful, a manager may develop over confidence based on this positive initial repurchase experience, and will repeat a similar program, to repeat the initial success. Overconfident managers are even likely to ignore negative feedback from the market (Taylor and Gollwitzer, 1995) and to overestimate market reaction of their decisions to repurchase their firms' stocks. This overconfidence bias may result in managers' overvaluation of their firms' stocks and, thereby, over-enthusiasm in launching a subsequent open-market repurchase program, even though it is not in the best interest of their shareholders. Consequently, the market would react less favourably to the firm's subsequent announcement. Therefore, we posit that overconfident managers are likely to repeat an open market repurchase program, even though announcement return from the previous announcement is negative (Roll, 1986;Billet and Qian, 2008). Furthermore, if repeated repurchase announcements arise from managerial overconfidence or hubris, then subsequent repurchase announcements should be associated with lower announcement period returns (Karolyi et al., 2015;Billet and Qian, 2008). Thus, we posit that overconfident managers making repeated stock repurchase announcements would result in lower abnormal stock returns, ceteris paribus.
Our third and fourth hypotheses therefore, are: H3: If repeat repurchasers are driven by managerial overconfidence, then firms with negative past announcement period returns are likely to repeat share repurchase announcements.
H4: These subsequent announcements will experience lower announcement period returns, and should be negatively related to announcement period returns.
An overconfident manager will also likely to repeat a subsequent share repurchase announcement in a shorter period. The market however, may or may not anticipate subsequent announcement made by repeat repurchasers. The degree of anticipation should be linearly related to the time between two subsequent announcements. The shorter the number of days between a previous and a current announcements, the lower is the expected current announcement period return. Thus, our last hypothesis is: H5: The time between two subsequent announcements is positively related to announcement-period return.

Data
We collect open-market share-repurchase announcement dates of non-financial and non-utility firms from Thomson Reuters SDC Platinum from 1996 to September 2014. We also collect price data of these firms from Thomson Reuters Datastream during the sample period. From these price data, we calculate the announcement abnormal return as marketadjusted abnormal returns with the market value weighted returns obtained from Kenneth French's website as the benchmark. We do not use the market model to estimate abnormal announcement returns as several firms in the sample repeat their announcements in less than a year. 5 After excluding those observations that do not have price data during the period required for our analysis, we obtain 3,122 announcement-year observations. Next, we match these observations with their firm accounting data from Thomson Reuters Datastream. To avoid the effects of outliers, we winsorise these variables at the 1% and 99% levels.
Insert Table 1 here

Empirical Results
Fu and Huang (2016)  split per-year announcement period returns based on whether they were the first or subsequent announcements made by the repurchasing firms, and report the results in Table 2.
The results show that, on average, announcement returns for subsequent announcements are significantly lower than those of the initial announcements. The results are also consistent when we split the sample period from before and after 2003. Thus, our results suggest that the decline in open market repurchase announcement-period returns cannot be explained by the increased efficiency of the US stock markets alone, but that the increased frequency of repeat repurchase announcements may also explain the lower repurchase announcement-period returns.
Insert Table 2 here Table 3 shows the differences of announcement returns sorted based on non-repeat and repeat announcement categories, and then based on the order of the announcements. In Panel A, the average event day abnormal return of non-repeat announcements is 1.99% and significantly higher than that of repeat announcements (0.86%). This result is consistent with Jagannathan and Stephens (2003)  Insert Table 3 here Table 4 shows the descriptive statistics of the sample sorted based on the frequency of announcements made by the repurchasing firms. Firm characteristics displayed in Table 4 are consistent with our second hypothesis. Repeat repurchasing firms are bigger, more profitable, have more cash flows and dividend payouts, less underpriced and have higher growth than those of non-repeat repurchasing firms. The less underpricing characteristics of repeat repurchasing firms suggest that the motivation of firms that frequently announce open-market share-repurchase program is not attributed to undervaluation, but seems to be more consistent with distributing excess cash flows. According to the free cash flows hypothesis, when there is no growth opportunities available, managers distribute excess cash to the firm shareholders to maximize their firm value. However, book-to-market ratio of repeat repurchasers is reported to be lower than that of non-repeat repurchasers. Furthermore, this ratio also decreases as the frequency of subsequent announcements increases, suggesting that repeat repurchasing firms have more growth opportunities than those of non-repeat repurchasing firms. This evidence suggests that managers of these firms choose to invest in the firm's stocks by announcing subsequent repurchase programs, even though there are growth opportunities available, which is inconsistent with the free cash flow hypothesis, but is more consistent with the self-attribution bias hypothesis. Self-attribution bias hypothesis predicts that over confident managers believe their firms' shares to be undervalued and therefore are likely to repurchase the shares. Thus, these managers are expected to repeat share repurchase announcements. Due to their illusory belief that they can repeat their past success of announcing such programs, these managers may also increase the size of the programs in subsequent announcements. The size of the programs announced by repeat repurchasers from the initial announcement to subsequent announcements increases from 6.78% of total outstanding shares to 7.29% in the second announcement and further up to 7.77% in the third announcement, with an average of 7.69% in more than 3 program announcements.
Insert Table 4 here To examine whether overconfident managers are more likely to repeat repurchase programs, we run a logit regression below and report the results in Table 5.  (1) where REPEAT ORDER is a dummy of one for each order of repurchase announcement (second, third and fifth announcement). We define HUBRIS as one if a previous repurchase announcement return is negative. RUNUP is the cumulative stock returns from 46 days to 6 days prior to an announcement. SIZE is the natural logarithm of equity market values. B/M is book to market ratio. LEVERAGE is leverage ratio. STDEV is the standard deviation of stock return from 100 days to 46 days prior to an announcement.
SIZEPROG is the change in the size of the repurchase program measured as percentage of shares authorized at initial announcement. ROA is return on assets. OPTION is a dummy of one if the purpose to conduct a share repurchase program is motivated by stock option plans.
All of the accounting variables are measured in the quarter before the announcement quarter. Table 5 shows that the coefficients on HUBRIS are positive and significantly related to subsequent repurchase programs, which is consistent with our third hypothesis, that overconfident managers are more likely to repeat share repurchase programs in the future, even though the past announcement period return is negative.

Insert Table 5 here
As firms' characteristics of repeat repurchasers are different from those of non-repeat repurchasers, and that these firms have the propensity to repeat open-market share repurchase programs, we conduct a logit analysis to examine the determinants or the likelihood of these firms to announce a share repurchase program. REPEAT REPURCHASE is a dummy of one for announcements made by repeat repurchasers. The results reported in Table 6 confirm the firm characteristics reported in Table 4. RUNUP is positively related to the likelihood to repeat repurchase suggesting that repeat repurchasers are not motivated by underperformance. Large firms with large cash flows and dividend yield are more likely to repeat share repurchase programs. Firms that repeat repurchase announcements are also likely to increase their program size.
Insert Table 6 here Next, we examine if subsequent repurchase announcements are negatively related to firm announcement period returns. In this analysis, we control for the probability to repeat a repurchase program measured as the fitted value from the logistic regression reported in Table 6. If the market is able to anticipate that a subsequent repurchase announcement will be made by a repeat repurchaser, then the coefficient on this variable should be significantly related to the announcement-period return, and the market should not react significantly to subsequent or repeat share repurchase announcements. In this analysis, we use four proxy variables for overconfidence. First, we use a dummy variable for firms that announce a share repurchase program more than once during the sample period (SECREPMORE). An overconfident manager is expected to repeat a repurchase announcement. However, firms that repeat their share repurchase announcements due to overconfidence are expected to experience lower announcement returns. Second, Taylor and Gollwitzer, (1995) argue that overconfident managers are likely to ignore negative feedbacks from the market. Therefore, we use a dummy variable for negative past announcement return (HUBRIS). If overconfidence drives the subsequent programs, then the relation between HUBRIS variable and the current announcement return should be negative. Third, we use the number of past repurchases (NPASTREP) as a proxy for overconfidence. A manager who has experience in launching more than one share repurchase programs can develop overconfidence to announce more repurchase programs, which results in negative announcement returns. Lastly, we use the number of days between two repurchase announcements (TBD). An overconfident manager will repeat a subsequent share repurchase announcement in a shorter period. The shorter the number of days between a previous announcement and a current announcement, the lower is the expected current announcement period return.
We report cross sectional regressions of share repurchasers' announcement abnormal returns on various deal specific and control variables in Table 7. Consistent with our fourth hypothesis, the results in Table 7 show that firms that announce an open market repurchase program twice or more during the sample period, and firms with negative past announcement-period returns experience lower announcement returns. Firms with higher number of past repurchases also experience lower subsequent announcement returns, but the relationship is not statistically significant. The coefficients on the probability to repeat repurchase are not statistically significant, suggesting that the market may fail to anticipate subsequent repurchase announcements. The last column in Table 7 shows that firms that repeat their announcements in a shorter period experience significantly lower returns, which is consistent with our fifth hypothesis.
Insert Table 7 here

Robustness tests
The beginning of our sample period (1996) may not represent the start of an initial open market repurchase program. Therefore, we followed Song and Walkling (2000), Cai et al. (2011), and Aktas et al. (2013) by imposing an initial time lag of 2 years (1996 and 1997) during which the repurchasing firms are not active. We include in the sample only those firms that have not undertaken any transaction during the initial dormant period (1996 and 1997).
The results are similar to those reported in our analysis.
We redefine our measure for repeat repurchasers as firms that announce subsequent open market share repurchase programs within five years. Though this alternative measure may suffer a sample selection bias due to this sample restriction, we find that the results are also similar to those reported in the tables.

Conclusion
We examine open market share repurchase announcements during the period of 1996 and September 2014. We find that the average open market share repurchase announcement return is much lower than that reported in early studies in the share repurchase literature. We document that, not only firms that repeat their repurchase announcements experience lower announcement-period returns than those of non-repeat announcements, but also the magnitude of announcement-period return is significantly decreasing in subsequent announcements. We also find that repeat announcements dominate the number of open market share repurchase announcements in the sample. On average, the number of repeat announcements per year is 68% of total announcements and since 2000 it has increased to an average of 75%.
We also observe the characteristics of firms that repeat their announcements are different from those of non-repeat announcements. Repeat announcement firms are larger, more profitable, have larger excess cash flows and are less undervalued, but have higher growth opportunities than those of non-repeat announcement firms. Firms with past negative repurchase announcement returns are more likely to repeat their repurchase programs even though the market reaction is less favourable to such announcements. Further test results are consistent with overconfident managers drive the declining wealth of repeat repurchasing firms' shareholders.  NON-REPEAT is firms that announce only once during the sample period. REPEAT is firms that announce to repurchase the first time and will repeat during the sample period. SIZEPROG is percent of shares authorized at initial authorization date.  Abnormal returns are measured as market-adjusted return. NON-REPEAT is firms that announce only once during the sample period. REPEAT is firms that announce to repurchase the first time and will repeat during the sample period. REPEAT 1 is firms that announce open market repurchase program for the first time and will announce subsequent programs during the sample period. REPEAT 2 is firms that announce the second time only. REPEAT 3 is firms that announce the third time only. *,**,*** denote statistical significance at 10%, 5% and 1% respectively. NON-REPEAT is firms that announce only once during the sample period. REPEAT 1 is firms that announce open market repurchase program for the first time and will announce subsequent programs during the sample period. REPEAT 2 is firms that announce the second time only. REPEAT2MORE is firms that announce twice and more. REPEAT 3 is firms that announce the third time only. REPEAT3MORE is firms that announce more than twice. CASHFLOW is measured as Cashflows/Total Assets. B/M is book to market ratio. DIVYIELD is dividend/market value of equity at time t-1. LEVERAGE is Total Debt/Total Assets. SIZE is the natural logarithm of market value of equity in the quarter prior to announcement quarter. EPS is Earnings Per Share. SALES is Change in Sales/Total Assets. ROA is Return on Assets. All accounting variables are measured in the quarter prior to the announcement quarter. RUNUP is cumulative market adjusted return measured from -46 to -10. STDEV is the standard deviation of market-adjusted return measured from day -100 to -46. Logit analysis of repeat repurchasers. The independent variable is one for a repeat repurchase made by repeat repurchasers. HUBRIS is one if past repurchase announcement return is negative. RUNUP is cumulative market adjusted return measured from -46 to -10. SIZE is the natural logarithm of market value of equity in the quarter prior to announcement quarter. B/M is book to market ratio. LEVERAGE is Total Debt/Total Assets. STDEV is the standard deviation of market-adjusted return measured from day -100 to -46. SIZEPROG is the change in program size. ROA is Return on Assets.
OPTION is a dummy variable of one if the motivation to conduct a share repurchase program is related to stock options. p-values are in parentheses. ***, **, * denote statistical significance at the 1%, 5% and 10% respectively. Logit analysis of the determinants of repurchasing made by repeat repurchasers. The independent variable is one for repurchases made by repeat repurchasers. RUNUP is cumulative market adjusted return measured from -46 to -10. CASHFLOW is measured as Cashflows/Total Assets. SIZE is the natural logarithm of market value of equity in the quarter prior to announcement quarter. B/M is book to market ratio. SIZEPROG is the change in program size. LEVERAGE is Total Debt/Total Assets. STDEV is the standard deviation of market-adjusted return measured from day -100 to -46. ROA is Return on Assets. OPTION is a dummy variable of one if the motivation to conduct a share repurchase program is related to stock options. pvalues are in parentheses. ***, **, * denote statistical significance at the 1%, 5% and 10% respectively. The independent variable is cumulative abnormal return (0,+1). SECREPMORE is a dummy variable for firms announcing twice and more open market share repurchase programs. HUBRIS is one if past repurchase announcement return is negative. NPASTREP is the number of open market share repurchase announcements made in previous period, calculated as natural log of one plus number of past repurchases. TBD is the number of days between two announcements. RUNUP is cumulative market adjusted return measured from -46 to -10. SIZE is the natural logarithm of market value of equity in the quarter prior to announcement quarter. B/M is book to market ratio. LEVERAGE is Total Debt/Total Assets. STDEV is the standard deviation of market-adjusted return measured from day -100 to -46. SIZEPROG is the change in program size. OPTION is a dummy variable of one if the motivation to conduct a share repurchase program is related to stock options. CASHFLOW is measured as Cashflows/Total Assets. Pr (Repeat repurchase) is the estimated probability of a repeat repurchase based on the logit results presented in Table 4. *,**,*** denote statistical significance at 10%, 5% and 1% respectively.