THE IMPACT OF SHARES REPURCHASED ON THE LONG-TERM ABNORMAL STOCK RETURNS: EVIDENCE FROM U.S. FIRMS
J.A.F. VELTHOEN S2530988
Master’s Thesis Finance
University of Groningen Faculty of Economics and Business
Address: Huygensstraat 69 Tel: 06-30561480
e-mail: jurjenvelthoen@hotmail.com
Keywords: Share Repurchase, Undervaluation, Signaling Asset Pricing
Supervisor: dr. P.P.M. Smid
Word count: 12.240
Abstract:
The effects of the announcement of a share repurchase program on the returns of a company are the main focus of this long-term study. This study researches the effects on the three-year buy-and-hold abnormal returns among a sample of the 3000 largest firms in the United States, using ordinary least squares. This provides relevant outcomes for investors and for corporations that try to signal undervaluation. I find that the size of the repurchase program at the announcement does have a positive and significant impact on the long-term abnormal stock performance.
TABLE OF CONTENTS
1. Introduction 2. Literature Review
2.1. Payout Policy
2.2. Tender Offers vs. Open Market Programs
2.3. The puzzle of repurchases regarding payout and signaling 2.4. The exchange option
2.5. Actual repurchases following announcements 2.6. Frequency of open market repurchase programs 2.7. Secondary Equity Offerings (SEO) and Repurchases 3. Methodology
3.1. Sample selection
3.2. Measuring Abnormal Returns 3.3. Undervaluation
3.4. Econometric issues 4. Data
4.1. Price series data
4.2. Data on Abnormal Returns 4.3. Data on independent variables 4.4. Data on control variables
4.5. Data on the adjusted Undervaluation-index 5. Results
5.1. The size of the repurchase program at announcement 5.2. The frequency of repurchase programs
5.3. Actual Repurchases following the program announcement 5.4. Secondary Equity Offerings and repurchases
5.5. Additional Analyses – Undervaluation Index 5.6. Industry effects
5.7. Discussion of the control variables 6. Conclusion
1. INTRODUCTION
Firms have several options to use their excess cash from operations. They can employ capital for new investments, use cash for Research & Development (R&D), or distribute excess cash their shareholders. The latter can be done directly through dividends or indirectly through repurchasing shares back from its shareholders. Repurchasing shares has become increasingly popular among firms worldwide. Several appraised investors have publicly expressed their preference for companies that repurchase stock as they do not waste cash and signal undervaluation1. The all-time high levels of corporate stock repurchases, or buybacks, in 2018 lead to intense debates on this topic in the United States (US). Numerous US politicians recommended a “ban on buybacks” should be enforced. According to pessimists, an abundance of cash is being returned to shareholders in the form of buybacks, at the cost of investments in labor, capital expenditures (Capex), and Research and Development (R&D).
Thus, the pessimistic voices claim that capital flows out of companies instead of being reinvested into valuable growth opportunities. In response to that, many practitioners and scholars opposed these proclamations (Groman and Nathan, 2019).
The purpose of this paper is to expand our knowledge of factors driving the anomalous stock price behavior after the announcement of repurchasing programs. In other words, I will attempt to discover whether firms that announce an open market repurchase program, are signaling undervaluation. This prompts the question for investors “will these stocks earn abnormal returns in the long run?”. To realize this research objective, the main research question I will address is the following: “Do firms that announce an open market repurchase program signal undervaluation?”.
Given that earnings will be distributed among investors, it is of great interest to examine whether this is done through dividends or share repurchases. Followed by the reason why they distribute it by repurchasing firm shares in the open market. Many papers have established evidence that corporate executives try to signal mispricing of their firm’s stock by issuing a buyback program. Brav et al. (2005) interview 384 CFO’s and find that these financial executives view share repurchases as a more flexible way of a payout than dividends.
They use this flexibility to repurchase the stock when they perceive it to be undervalued.
1 http://www.cnbc.com/2018/08/31/warren-buffett-explains-the-enduring-power-of-stock-buybacks
Moreover, Brav et al. (p. 517) express that “nearly 90% of firms with low P/E ratios state that market undervaluation could lead to repurchases.”.
In the academic literature, a great deal of empirical evidence exists for long-term abnormal stock returns following a buyback announcement for the end of the 20th century [Lakonishok and Vermaelen, 1990; Ikenberry, Lakonishok and Vermaelen, 1995]. However, around the change of the century, scholars like Fama (1998) and Schwert (2003) have argued that the positive outcomes of earlier studies on repurchases could have been the consequence of chance or empirical issues like data mining. Moreover, they challenged that the gains would have disappeared over time as a result of arbitrageurs. However, later studies [Peyer and Vermaelen, 2009; Evgeniou et al. 2017; Yook, 2019] have shown the persistence of the buyback anomaly.
Most evidence of these papers has pointed in the direction of the repurchase announcement as a signal to the outside world that the company is undervalued.
In this thesis, I contribute to understanding the determinants of signaling undervaluation of firms through share repurchases. A number of factors that contribute to finding an answer will be addressed in sub-sections of this thesis. First, does the size of the open market repurchase program affect the long-term returns? Since the size should indicate the commitment that the firm is willing to make. Secondly, does the frequency of how often a firm is authorized to repurchase shares affect the signal and the long-term returns? This is likely to affect the credibility of the signal. The announcement is essentially an option to repurchase, and the firm is not obliged to actually buy back any stock. This prompts the third question.
Does the actual repurchasing of stock affect the long-term stock returns of the firm? Lastly, the repurchase announcement might be perceived by the market as a signal that a firm lacks any profitable investment opportunities to spend its cash on. When many firms in an industry seek funds through secondary equity offerings (SEO), it is often perceived that firms in that industry have profitable opportunities. Therefore, the fourth question relates to SEOs in the same industry as a repurchasing firm. Does SEO activity in the same industry affect the long-term returns of the firm?
In this thesis, I will investigate the relationship between share repurchases and undervaluation through an empirical approach. The firms within the sample are selected from the 3000 largest firms in the US based on market capitalization (Russel 3000-index). Data on the announcement and size of share repurchase programs are collected for the period 2003- 2013 from the Zephyr database. The data regarding stock prices and repurchases are gathered
from Compustat and CRSP. A long-term event study methodology as described by Barber, Lyon, and Tsai (1999) is constructed to analyze the sample firms’ returns.
Consequently, long-term buy-and-hold abnormal returns (BHAR) are calculated based on the CRSP equal-weighted index. The research window includes prices from 6 months before the announcement month to 36 months after the announcement month.
The outcomes of my research are largely in line with previous theoretical and empirical papers [Ikenberry, Lakonishok and Vermaelen, 2000; Yook, 2010; Evgeniou et al. 2017].
Furthermore, the results show that there is a positive and significant relationship between the announced size of the program and abnormal returns in the long run. This is sustained by further analyses on the level of repurchase announcements and the presence of other undervaluation indicators. There is no significant relationship observed for companies that announce infrequently or actually repurchase shares. The last sub question, regarding SEO activity is neither confirmed.
The remainder of this thesis is structured as follows. The next section addresses previous literature concerning repurchases and substantiates the hypotheses following from the sub- questions. The Methodology section describes the design of the empirical methodology and the sample selection process. The Data section lays out the structure and collection of my data.
The fifth chapter discusses the results of this research, these are further discussed and relate to other recent findings and theories of asset pricing and corporate finance. I conclude my research in the sixth chapter where I provide suggestions and limitations for future directions of research on the topic of repurchases.
2. LITERATURE REVIEW
2.1 Payout Policy
In this chapter, first the theoretical background on payout policy is reviewed. Followed by an overview of the theoretical and empirical literature on signaling in light of repurchases.
Finally, the hypotheses are introduced to fill the gap in the empirical and theoretical literature.
Let us revisit the seminal theory on payout policy as laid out by Miller and Modigliani (1958; 1961). They established the irrelevance of payout policy in frictionless markets. In essence, they reject value creation by capital structure (1958) or payout policy (1961). Firm value is solely driven by operating performance and investment decisions. Since then, various theories have tried to explain why companies distribute earnings among investors in distinct ways (e.g. repurchases and dividends). This fundamental theory has led to several directions of frictions that try to explain the real world and its deviations from the irrelevance theory of Miller and Modigliani. Share repurchases could not be adding any value to the firm in the view of the irrelevance propositions of Miller and Modigliani, thus there have to be frictions in place.
The most prominent frictions are 1) taxation, 2) agency problems, and 3) asymmetric information. This paper focuses mainly on the latter two frictions.
The taxation argument is believed to have been a driver in the past for firms to substitute their dividend payout with share repurchases. In the US, dividends are taxed at a higher rate than capital gains proceeding from repurchases. Since the US dividend-tax reductions of 2003, there has been a shift back from repurchases to dividends as evidenced by Brown et al. (2007).
However, Brav et al. (2005) find that most financial executives say that tax considerations matter but are not a dominant factor in choosing buybacks as a payout over increasing dividends. Therefore, this research limits to the undervaluation of firms in relation to share repurchases. Taxation is no topic of research in my thesis. It still bears relevance in the broader theme of share repurchases as a payout mechanism, Manconi, Peyer, and Vermaelen (2018) find that this is especially relevant in an international context due to a larger dividend tax disadvantage than in the US.
There are two main reasons for using repurchase programs instead of, or in addition to, dividends as a payout to shareholders. The first motivation for repurchases is that they are used to distribute cash flows among investors. Firms are reluctant to cut dividends (Lintner, 1956), since these firms are negatively affected by downward stock market reactions as Dennis et al (1994) observes. This makes it less attractive for firms to continuously increase dividends, as
it informally commits the firm to future dividends. In contrast, share repurchases are more flexible by nature, they usually arise when cash flows are less permanent (Jagannathan et al, 2000). Thus, it serves as an alternative way of payout policy in addition to dividends. Ha, Long, and Lee (2011) find that dividends and share repurchases are complementary in their signaling role for temporary and permanent earnings (p.517). The intuition of the technical effect is simple at first sight and works as follows. After a repurchase of common stock outstanding, the remaining holders of common stock end up with an increased fraction of the company, in contrast to before the actual repurchase, ceteris paribus. In the case that the price- to-earnings ratio (P/E-ratio) remains the same, the stock appreciates in value. Although this technical effect is easy to grasp, it would mean that markets are easily fooled and are not efficient processors of information, because the actual decrease in stocks outstanding doesn’t increase the value of the operations of the firm (Koller et al., 2015 p. 38). Thus, it is more likely that the total value of the company actually decreased. Hence, academics commonly argue that the increased equity stake is immediately offset by a decline in the price-to-earnings ratio.
Since total assets decrease as cash is paid out, leverage increases, ultimately amplifying the risk of the firm (Dobbs and Rehm, 2005).
The second reason for announcing a repurchase program is to signal good prospects for the company. This is done by the managers-insiders of the company who believe that the market price of the firm is undervalued in relation to its true market value (or fair value). This is the main topic of this paper.
The primary focus of the empirical research of this thesis is the companies that repurchase their shares because of undervaluation and not for other reasons. Other motives for repurchases are the alteration of capital structure, a tactic to counter stock dilution caused by (employee) stock options or to fend of hostile takeovers [Ikenberry and Vermaelen, 1996; Dittmar, 2000; Lei and Zhang; 2016].
Altogether, there has been a lot of empirical research conducted on the topic of buybacks and the mispricing of the companies that choose this form of payout structure or signaling device. The buyback anomaly, as it is titled by many practitioners and academics in the asset pricing literature, has survived empirical research in the past twenty years for several reasons (Evgeniou et al., 2017). First, the repurchase anomaly is most persistent in smaller firms. Loughran and Ritter (2000) argue that small firms are less covered by research analysts following their performance, thus increasing the likelihood of mispricing. Consequently, managers of small firms can take advantage of mispricing. Second, the repurchase anomaly has received real-world evidence from CFO’s (Brav et al., 2005) of whom 86% claim to use it
as an instrument to signal that the stock price is too low and does not reflect its true value.
According to Dittmar and Field (2015) managers are able to time the market in actually repurchasing shares. Finally, firms that repurchase shares do not appear to be mispriced based on general mispricing criteria as those presented by Stambaugh et al. (2015). Moreover, buyback stocks are generally experiencing negative returns prior to the announcement, in contradiction to the momentum anomaly (Jegadeesh and Titman, 1993). In addition, many undervalued stocks that announce repurchases miss earnings forecast, which is contradictory to the Mendenhall (2014) earnings drift anomaly predicting negative returns after missing earnings forecasts (Evgeniou et al., 2017). These contradictory anomalies might scare investors away when they suggest overvaluation.
2.2 Tender Offers vs. Open Market Programs
The two most common practices to repurchase shares are tender offers and open market programs. In tender offers, the company tenders a significant number of shares and agrees to pay a tender premium to buy shares from the shareholders in a short period, usually within a few months. While, in an open market program, the firm launches a program authorizing a total buyback value which can be bought in the open market for a longer period, on average one to two years (Oded, 2005, p.293).
In a theoretical model developed by Oded (2011) the trade-off between tender offers and open market programs is revealed. His underlying assumption for constructing the model is that the payout in the form of repurchasing shares is a way of reducing agency costs of free cash flow. In other words, the company does not have good investment opportunities at its disposal and the value of cash would gradually decrease in case it is held within the company.
Additionally, Jagannathan et al. (2000) find empirical evidence proving that this excess cash is best spent on repurchases instead of dividends, in case the underlying cash flows’ permanence is estimated to be not strong, and no long-term payout commitments want to be undertaken.
Consequently, it follows theoretically that tender offers are more efficient since they distribute the cash sooner, reducing the wasted cash by management (Oded, 2011). Moreover, the tender offer forces a wealth effect on current shareholders. In case the tender premium undervalues the true stock value, the tendering shareholders lose, whereas non-tendering shareholders gain.
Oded (2011) shows that this confrontation creates preparatory information gathering for various groups of investors and results in information asymmetry among shareholders. This information asymmetry engenders adverse selection and the shareholders need to be compensated for that by a tender premium if the firm wants the tender offer to succeed. On the
contrary, an open market program is gradual and therefore does not encourage information gathering nor does it require a tender premium, as shares are anonymously repurchased in the open market. Oded (2011, p. 3181) concludes that ‘there exists a trade-off between the two methods for the decrease in share value incurred in a tender offer and the waste of free cash flow in an open market program’.
One of the most interesting outcomes of Oded’s model is the contradiction towards existing repurchasing literature that tender offers’ long-term stock performance reverses after the tender premium and end up at the fair value, thus predicting negative long-term returns.
Moreover, firms pursuing tender offers will decrease their downside risk of overinvestment relative to their peers. Companies whose decreased risk outweighs the tender premiums paid will benefit long-term positive stock returns. One of the implications of the differences between open market programs and tender offers is that in tender offers a tender premium is paid, whereas in the open market program no premium is paid (Dittmar and Field, 2013). Since the differences in characteristics and consequences between open market programs and tender offers show different underlying motives of managers choosing between the two, the focus in this research is exclusively on open market repurchases.
2.3 The puzzle of repurchases regarding payout and signaling
Brav et al. (2005) find that 86.4% of the financial managers use repurchases to signal that the market price of their stock is undervalued in relation to its true value. Contrarily, the same interview results outlined that managers do treat repurchases as “the residual cash flow”
as implied by Modigliani and Miller. The first reason would signal better prospects for the firm, generally formalized by an increasing investment opportunity set. The latter would imply an absence of these positive NPV projects and serve the free cash flow hypothesis (Jensen, 1984) in that it would pay out excess cash in the hands of managers.
Furthermore, previous literature has found no significant increase in operating performance in the period following the repurchasing announcement and actual repurchases [Grullon and Michaely, 2004; Jagannathan and Stephens, 2003]. This complicates things since the signal of undervaluation would say something about the increasing operating performance driving the higher fair value of the company in the perception of the financial markets. In expectation, the actual positive abnormal performance of the shares following the repurchases would theoretically corroborate the undervaluation and would confirm a higher operating performance. Grullon and Michaely (2004) argue this to be in support of firms signaling
diminishing investment opportunities, succeeded by a decrease in risk, by cutting out new projects which bear significantly more risk than current assets in place. Since the decrease in risk leads to a lower cost of capital, the firm value increases.
2.4 The exchange option
Ikenberry and Vermaelen (1996) argue that the announcement of an open market repurchase program provides an exchange option for the firm where the market price can be exchanged for the true value of the shares, the repurchase option hypothesis. In open market programs, firms are authorized, though not obligated, to pursue the proposed buyback. It enhances the existing investment opportunity set by the authorization to buy back stock in the open market. This ultimately benefits the long-term shareholders who do not – unwittingly – sell their shares to the firm. The intuition of the exchange option is that, in case the current price differs significantly from the fair value, the creation of the option essentially bears the value of the rents. Ikenberry and Vermaelen (1996) predict the option value to be positively related to the volatility of the stock, the size of the proposed repurchase fraction, and the potential for mispricing in the future. Furthermore, Ikenberry and Vermaelen (1996) acknowledge the excessive use of open market repurchase programs by US firms – in the
’90s of the last century. This could be a sign for many of these firms to have other motives for buybacks than undervaluation. Nevertheless, these programs could still be credible if these programs are viewed as options. Managers have the ability to not exercise the option in case they see the mispricing gap disappear after the announcement. Subsequently, Ikenberry and Vermaelen (1996) theorize that firms can anticipate mispricing and request the board’s authorization early before there is undervaluation. This is called the “early adoption” strategy.
In contrast, waiting till there is a significant mispricing present would be a “wait-to-adopt”
strategy. If all firms would apply the wait-to-adopt strategy, assuming that markets are efficient, this would be a strong information signal at the announcement. The share price would rise extensively at the time of the announcement, and the mispricing would disappear, which would mean that the firm could no longer profit from the perceived undervaluation. Thus, killing the initial motivation for starting an open market repurchase program. In conclusion, Ikenberry and Vermaelen find that the average abnormal announcement-return increases with the fraction repurchased and volatility. The latter supports my first hypothesis that the size of the program at the announcement enhances the abnormal returns in the long run.
What one would expect from the above, is that in the absence of an immediate rise of the share price, the company could gradually buy back shares, thereby time the market by buying undervalued shares. Representing the early adoption approach. Oded (2005) enhances their model theoretically, by proving the announcement to be a non-dissipative separating tool for good (high true value) firms. In his model, he predicts that “firms that announce condition actual repurchases on a future realization of value” (p.291). Besides, he predicts that the abnormal returns will be low in the short run and high in the long run. Oded compares it to empirical evidence of Stephens and Weisbach (1998) who show that firms either repurchase close to all of the authorized shares or none at all. Furthermore, Stephens and Weisbach find that companies repurchase between 74% and 82% of the announced target level of repurchases.
They find that the market has some ability to forecast actual repurchases and payoffs to firms that announce but do not repurchase, however, this forecasting ability is minimal. In this research, I test if firms complete their targeted repurchase programs. Therefore, these firms materialize the exchange option and realize post-announcement returns following actual repurchases of shares.
Previous articles have theoretically argued the positive effect of the size of the fraction of shares repurchased [Ikenberry & Vermaelen, 1996; Zhang, 2005; Yook, 2010] in the long- run stock performance of the firm. The evidence from previous empirical articles is more than 10 years old or is empirically tested in other (less efficient) markets than the United States, therefore, these should be tested with new empirical evidence. Thus, this research adds to the literature of long-term post-announcement stock performance of share repurchase programs.
H1: The size of an open market repurchase program (at the announcement) has a positive impact on the long-term abnormal returns.
2.5 Actual repurchases and the completion of an announced program
The theoretical model of Oded (2005) predicts that post announcement long-run returns are positively correlated with the level of actual repurchases (p.292). Stephens and Weisbach (1998) and Yook (2010) have empirically researched the (stock) performance of firms who actually repurchase shares, in relation to those who don’t.
Ikenberry, Lakonishok, and Vermaelen (2000) show that managers trade strategically (p.
2375). They find empirical proof for an increased amount of actual stock repurchases when the stock drops in value, and decreasing actual stock repurchases after the stock rises in value. In
other words, managers buy the stock when it is more undervalued. This phenomenon is also observed by Dittmar and Field (2015) in a sample for the period 2004-2011. Gong et al. (2008) find that about 84.7% of their sample of announcing firms carry through with actually repurchasing shares in subsequent quarters following the announcement. Only these firms experience improvements in earnings in the subsequent quarters. Caton et al. (2016) report that about 87.6% of their sample firms report actual repurchases by the quarter following the repurchase announcement (p.165).
H2: Firms that actually repurchase following announcements realize positive long-term abnormal returns.
2.6 Frequency of open market repurchase programs
In the past decade, much has been documented on the catering hypothesis. Scholars like Jiang et al. (2013) and Kulchania (2013) have evidenced that managers cater to investor demand in the situation that investors place a higher valuation on repurchasing firms than those that payout through dividends.
In 2003, Jagannathan and Stephens disputed that companies that regularly buy back shares send credible signals to the market about the undervaluation of their stock. They found that the firms that repurchase stock infrequently, buy back larger fractions of their outstanding shares than the firms that repurchase more frequently. The infrequent repurchasing firms have relatively high asymmetric information, measured by institutional investor holdings, and experienced poor prior stock performance. Furthermore, Jagannathan and Stephens (2003 p.
71) find that infrequent open market programs are more often utilized by “smaller firms with higher variability of operating income, higher levels of capital expenditures, lower market-to- book ratios, and lower levels of institutional ownership”. Consequently, the writers argue that infrequent repurchases are associated with perceived undervaluation. Whereas “more frequent repurchases are more likely to be associated with the dividend substitution hypothesis or the exercise of stock options” (Jagannathan and Stephens, 2003 p. 74). Taken together, I assume the frequency of open market repurchase program announcements to have a negative impact on the abnormal returns in the long run. Particularly, I foresee infrequent repurchasing firms to experience positive abnormal returns in the post-announcement period.
H3: Firms that repurchase infrequently are signaling undervaluation and exhibit positive long- run abnormal returns.
2.7 Secondary Equity Offerings (SEO) and Repurchases
In addition to the empirical work of Grullon and Michaely (2004) on decreasing investment opportunities, Pham et al. (2019) claim to have found a proxy for measuring decreasing investment opportunities. They argue that the distinction between firms signaling better prospects and those that signal a reducing investment opportunity set can be made by looking at SEO’s activity in their industry in the past 6 months. Financing through SEOs is always used for capital expenditures or research and development expenses, no matter what the company says according to Walker and Yost (2008). Therefore, providing a good proxy for increasing investment opportunities according to Pham et al. (2019). Firms who would rather spend their cash on buying back shares than investing in the growth opportunities available to the industry sends a signal to the market that it lacks the growth opportunities.
This is in contrast with the free cash flow hypothesis developed by Jensen (1984) and the risk reduction hypothesis by Grullon and Michaely (2004) claiming that repurchases would increase the value by respectively reducing cash waste and decreasing the cost of capital. In these situations, the repurchase payout is the best alternative use of excess cash. Therefore, a regression of the abnormal returns on a dummy for recent SEOs in the industry will test for this hypothesis. It will provide insights if firms signal decreasing investment opportunities.
Loughran and Ritter (1995) address the high stock returns preceding SEOs, which are likely a sign that managers try to time the market when their stock is overvalued. This might further indicate that an industry with a lot of SEO activity is likely to be overvalued.
H4: Firms repurchasing shares in high SEO activity industries underperform regular repurchasing firms in the post-announcement period
In conclusion, this research builds on the traditional theories both from asset pricing and payout policy such as dividend irrelevance, signaling, information asymmetry, and agency costs. First established by Lintner (1956), Modigliani and Miller (1961), Vermaelen (1981), and Jensen (1986). Bonaimé and Ryngaert (2013) investigate the predictive relationship between insider selling, insider buying, and the repurchase decision of a firm. They find that insider selling during the same quarter as a share repurchase is only marginally consistent with undervaluation as a motive for repurchasing. Babenko et al. (2012) find that the market underreacts to insider trades of executives, which results in long-term buy and hold abnormal returns. Furthermore, they find that companies are more likely to complete the announced
programs if insider purchases are higher. However, I was unable to access data on insider trades, therefore, this subject is not empirically researched in this thesis.
3. METHODOLOGY
3.1 Sample Selection
The event firms in this research have been sampled in correspondence with the methods applied in prior research [Yook, 2010; Bhattacharya et al., 2015; Ditmarr & Field, 2015]. The US market is chosen for several reasons. First, it is a good proxy of an efficient2 (Fama, 1991) and liquid market with broad diversification among different industries. Second, other studies find different outcomes in an international setting due to other factors driving the results, for instance, higher dividend tax rates outside of the US (Manconi, Peyer and Vermaelen, 2018).
In finding results on the level of undervaluation and long-term returns, the US market is likely to be the best research environment to suppress confounding effects and reduce endogeneity.
Lastly, a very important factor is data availability, since CRSP and Compustat provide high quality data on US firms that are not as complete for other regions.
The sample data were extracted from Compustat, CRSP and Zephyr for the contemporaneous period of 2003-2013. This sampling period is chosen to exclude any impact of the 9/11 events on the US in the financial markets. The price series from CRSP were extended until the beginning of 2017 to measure the returns following announcements.
From the Zephyr Database 3275 share repurchase announcements have been gathered, of which 2534 announced a repurchase larger than 0.1% of outstanding shares at the time of announcement. All accelerated programs, tender offers and Dutch Auctions (238 announcements) were excluded from the dataset to counteract confounding effects of these differing types of repurchase events (Dittmar and Field, 2015). Firms from the industries financial services (SIC 6000-6999) and utilities (SIC 4900-4949) are removed from the sample.
Many firms in these industries are regulated and often have legal distribution requirements.
Dividend payments and if applicable, distributions through share repurchases, may be determined by law rather than chosen by management (Luttman 2008, p.40). These industries are consistently excluded by many other researchers [Ziang et al., 2013; Stephens and Weisbach, 1998]. Companies for which Compustat fiscal data is missing in the year prior to the repurchase announcement are removed. Firms announcing that the repurchase will be
2 This is based on the premise that the United States have lower transaction costs in terms of obtaining information and trading costs.
financed by borrowed funds are excluded, since their main focus is to alter the capital structure.
The firms that occur multiple times as a consequence of multiple announcements are cleaned, therefore, they don’t have overlapping buy-and-hold return periods in the final dataset.
Additionally, checks on firm characteristics in the form of control variables are added to the model to check for alterations in capital structure. Finally, announcement firms missing price data for the 6 months prior to the announcement or three years after the announcement are removed. This amounts to a sample size of 1619 repurchases by 602 unique firms. This sample has overlap in the calculated abnormal returns due to multiple announcements within the 36- month time horizon. After cleaning the sample from these confounding effects, there are 764 firm announcements left3. Table 1 provides more detailed information regarding the spread of the sample over the various sample years. Later in this thesis, sub-samples will be drawn for measures of undervaluation, size of announced program and industry. The resulting full sample is still a significant representation of the total population measured on B/M, size and industry in relation to the corresponding population (Russel 3000 index). In comparison to the total US market (CRSP universe), there is a significant overweight in larger firms. From all firms in the Compustat database, this sample is mostly representing the 4th and 5th quintiles based on Market Capitalization, see table 8 pane 2 in Appendix A. In sight of the B/M ratio, the firms in this sample are in the lower quintiles based on all Compustat firms (Table 8, pane 3). In terms of short-term momentum (six months) we have a near perfect representation of all CRSP firms (Table 8, pane 4).
The main research question of this paper “Do firms that announce an open market repurchase program signal undervaluation?” is researched by measuring the effects of the independent variables size of the repurchase fraction, actual repurchases, frequency of repurchase announcements and SEO-activity. The empirical model to estimate the impact of these variables on the dependent variable, the buy-and-hold abnormal returns, is specified as:
𝐴𝑅𝑖𝜏 = 𝛼 + 𝛽1𝑆𝑖𝑧𝑒 𝑖+ 𝛽2𝐴𝑐𝑡𝑢𝑎𝑙𝑖𝑘 + 𝛽3𝐹𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦𝑖+ 𝛽4𝐷_𝑆𝐸𝑂𝑖+
∑17𝑢=5𝛽𝑢𝐶𝑜𝑛𝑡𝑟𝑜𝑙 + ∑28𝑗=18𝛽𝑗𝑌𝑒𝑎𝑟𝑖+ 𝜖 ,
Eq. 1
3 Cleaning the dataset in this way preserves it from multiple counting firms’ abnormal returns. Throughout the rest of this thesis, the research sample will contain the 766 “clean” announcements. In other cases, specific references to the sample size and characteristics will be made.
Where 𝐴𝑅𝑖𝜏 is the abnormal buy-and-hold return for period 𝜏 on the security 𝑖, 𝑅𝑖𝜏 is the holding period return for period 𝜏 on the sample stock. The periods for 𝜏 are 6, 12, 24 and 36 months4. The long-term is viewed as 36 months and regressions are repeated for 24 months to check these long run results for a shorter time span.
According to the strong form of the Efficient Market Hypothesis: “prices fully reflect all available information about future values” (Malkiel and Fama, 1970, p383.). Thus, the price jump at the announcement should be measured. Therefore, I start measuring from the announcement month (t = 0). This will be further discussed later in the methodology section.
The 𝛽1𝑆𝑖𝑧𝑒 𝑖 is the announced repurchase program size, the 𝛽2𝐴𝑐𝑡𝑢𝑎𝑙𝑖𝑘 is the actual repurchases following the repurchase fraction with 𝑘 the for the following four or eight quarters upon the repurchase announcement. The empirical research of Yook (2010) and Stephens and Weisbach (1998) don’t include regressions, they merely research different panels of buy- and-hold returns and perform difference tests on these panels of returns. It is of importance to see the effect of actual repurchases during authorization programs in linear regressions.
Hereby, I decrease the chance of omitting relevant variables present in simple difference tests on panel returns. 𝛽3𝐹𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦𝑖 is the previous frequency of the same firm announcing before this repurchase announcement and 𝛽4𝐷_𝑆𝐸𝑂𝑖 is a dummy variable for the equity issuance activity of a firm’s industry. In addition, 𝛽𝑢𝐶𝑜𝑛𝑡𝑟𝑜𝑙 is a set of 13 control variables on several firm characteristics. In order to control for time-fixed effects, year dummies are included in the model by 𝛽𝑗𝑌𝑒𝑎𝑟𝑖. Where year is 𝑗 for the years 2003-2013. A series of cross-sectional regressions is performed on these variables. Since the abnormal return (independent variable) is static, there is no benefit in running panel regressions on this dataset.
3.2 Measuring Abnormal Returns
According to Chan, Ikenberry and Lee (2003) it is expected that in case of underreaction at the announcement followed by actual repurchases, the abnormal buy-and-hold returns are significantly high. Thus, the question if repurchases signal long-term abnormal returns brings up the important issue how these long-term abnormal returns should be calculated. In the previous methodology literature on abnormal performance, much attention has been devoted to varying calculation techniques of long run performance and their shortfalls
4 Throughout the rest of this paper, periods will be interchangeably indicated with “[starting month, ending month]”. For instance, the three-year months abnormal return will be indicated with BHAR [0,36].
(Barber, Lyon and Tsai, 1999). One of the most prominent techniques applied in previous research is the buy-and-hold abnormal return (BHAR). It is appraised since it is a precise estimator of investor experience (Chan, Ikenberry and Lee, 2003). Therefore, it has a practical application for investors which makes this methodology and its outcomes resourceful for a wide array of academics and practitioners.
Table 1: Summary Statistics for firms announcing open market repurchases by calendar year
Year Announced
Repurchases Unique Firms Industries
Number of firms announced size of
repurchase
Percentage Shares Announced
Frequency of Repurchase Announcements
Mean Median Mean
2003 97 88 33 92 7.5% 4.5% 1.1
2004 39 39 20 34 5.7% 4.8% 1.0
2005 109 98 33 105 7.2% 5.3% 1.1
2006 153 128 45 146 6.8% 4.9% 1.2
2007 262 227 51 254 7.0% 5.9% 1.2
2008 127 116 38 120 7.5% 5.9% 1.1
2009 47 43 19 45 5.8% 4.6% 1.1
2010 138 118 34 135 6.4% 5.2% 1.2
2011 222 179 45 220 6.1% 4.7% 1.2
2012 204 165 42 200 5.4% 4.0% 1.2
2013 221 161 41 218 5.3% 3.1% 1.4
Entire Sample 1619 602 59 1569 6.4% 4.8% 1.2
The sample selection starts from 2003 and ends in 2013. There are 1619 announced repurchase events included in the sample exercised by 602 unique firms. The industries column consists of the first two digits of the Standard Industry Classification code. In the sample period, 1569 of the total 1619 events have announced the size of the open market repurchase program.
Firms with announcements lower than 0.1% of shares outstanding are later removed, this results in 1419 announcements. The mean and median values of the size of the announced repurchase program are given by every calendar year. During the sample period, the firms in this sample announce repurchases 1.2 times per year on average. The dataset is further cleaned from multiple announcements within the three-year post-announcement period to prevent double measuring the effects on the returns. The resulting dataset includes 766 announcements.
Table 2: Summary Statistics for the frequency of firms announcing open market repurchases by calender year
Hypothesis
Included in Model
Variable Name Short Name Descriptive Statistics
Yes/No Observations Mean Median St. Dev Min Max 25th
Percentile
75th Percentile
H1 Yes Repurchase Announcement Size Repurchase Fraction 766 7.2% 5.7% 6.5% 0.1% 68.9% 3.1% 9.4%
H2 No First 4 Quarters Cumulative Actual
Repurchases - 766 4.9% 3.8% 4.5% 0.0% 35.6% 1.6% 6.6%
H2 No First 8 Quarters Cumulative Actual
Repurchases - 766 9.1% 7.5% 7.6% 0.0% 52.1% 3.4% 12.6%
H2 Yes Program Completion per Firm 4 Quarters
(Dollar Measure) Actual Repurchases Q4 766 111.2% 69.8% 157.1% 0.0% 1333.3% 27.6% 126.2%
H2 Yes Program Completion per Firm 8 Quarters
(Dollar Measure) Actual Repurchases Q8 766 225.1% 129.1% 157.1% 0.0% 3456.1% 62.2% 248.8%
H3 No Frequency of Repurchase announcement in
total sample (per uniqe firm) - 1619 4.45 4 3.6 1 19 2 6
H3 Yes Previous Frequency Frequency 766 0.85 0 1.35 0 8 0 2
H4 Yes SEO Industry Occurence (Dummy Variable) D_SEO 220 - - - 0 1 0 1
The actual repurchases are measured as firms that actually repurchased more than 0,1% of their stock in the first 4 (8) quarters following the quarter of the announcement. The cumulative actual
repurchases are based on the dollar measure as discussed in the data section (section 4.3). This measures the completion of the program, in this thesis, program completion and actual repurchases are used interchangeably. Please note the difference in the repurchase fraction mean (median) between the dataset before correcting for multiple announcements (in table 1) and after. This indicates that the announced fractions following earlier announcements are overall smaller. The frequency measure adopted in the model is back-ward looking. Therefore, it approaches the information incorporated by the market at the time. However, it only dates back to three years prior to the start of my sample period. The D_SEO measure indicates SEO activity within the same 3-digit SIC industry classification. In total, 220 out of 766 firms have experienced a secondary equity offering in the six months prior to their repurchase announcement.
One of the most common approaches to calculating BHAR is by using the Market Model to estimate the expected returns. This is a very popular approach for most short-term event studies. However, some limitations should be raised; it is expected to suffer from bad model problems in case applied for estimating long-horizon expected returns. For instance, the data used in the Market Model for calculating the expected returns may include other events such as dividend increases or merger activity, making it vulnerable for confounding effects.
Furthermore, using a value-weighted index like the S&P500 as a benchmark induces negative bias towards the abnormal return model, since the benchmark is periodically rebalanced whereas the compounded returns for the sample firm remain constant. This is an important reason to choose for an equal-weighted benchmark of the CRSP universe instead of a rebalanced (value-weighted) index. Alternatives for the Market Model would be a firm matching procedure as described by Barber, Lyon and Tsai (1999 p.173). However, this could create a cross-correlation bias in the returns since not all correlation can be removed by selection on firm-characteristics like industry, size and book-to-market ratio (Yook, 2010).
Therefore, the Market Model with an equal-weighted index seems the best predictor of expected returns, keeping in mind that it replicates the actual investments of an investor. The expected monthly return from an asset is calculated using the formula:
𝐸(𝑅𝑖) = 𝛼 + 𝛽 ∗ (𝑅𝑚) + 𝜀𝑖, Equation 2
Where 𝛼 is the intercept coefficient from time-series regressions of the simple return 𝑅𝑖𝜏 on the return of the market 𝑅𝑚. The (𝑅𝑚) is the observed simple return on the CRSP equal- weighted composite market index5. The 𝛽 is the synchronous movement of security 𝑖 with the return on the market index, called the market beta based on a five-year time series regression.
The 𝜀𝑖 is the error term which is a random variable with the expectation to be zero and to have finite variance. Furthermore, I assume the error term to be uncorrelated with the market return 𝑅𝑚 and the firm return 𝑅𝑖. In this paper, expected returns for the periods 6, 12, 24 and 36 months are calculated. The estimations for the betas for all expected return periods are based on the 5 year past returns, measured up till one month before the announcement. The abnormal returns are calculated as in equation 3:
5 In the appendix, results for the observed return on the S&P500 (value-weighted) composite market index are included. In enhancing the robustness of this empirical paper, regressions have been repeated with independent variables drawn from this alternative market index.
𝐴𝑅𝑖𝜏 = 𝑅𝑖𝜏− 𝐸(𝑅𝑖𝜏), Equation 3
However, this data has some limitations, expected returns of the event firms are based on the stock’s sensitivity to the market (𝛽) up till one month before the announcement. This is assumed to incorporate all past characteristics of the firm that will remain constant, which is not likely to happen in the real world. Furthermore, the 𝛽 estimation period of firms in the beginning of the sample are affected by systematic events like the tech bubble around the year 2000 and the 9/11 attacks. Furthermore, the market index only explains part of the variation of the returns, the idiosyncratic risk that is company specific, is not included in the equation for estimating the expected returns. This is a limitation of this study.
3.3 Undervaluation
The term undervaluation has been progressively adjusted throughout previous research.
The book-to-market ratio started mainly as one of the constituents driving factor models, throughout time it has gradually come to serve as an indicator of valuation levels of equities (Jagannathan and Stephens, 2003). Furthermore, the previous performance up to an announcement or event and its continuing performance in offsetting (ongoing) direction has been widely used as indicator of undervaluation (overvaluation) in ex-ante empirical research.
The anomalies literature is rich in providing ways of measuring undervaluation. One of the estimation techniques is the Mispricing Measure (Data) as comprised by Stambaugh et al.
(2015), consisting of 11 anomaly variables which comprehend an under- or over-pricing of securities. Recent empirical research (Evgeniou et al., 2017) on the anomaly of share repurchasing companies has shown insignificant approximations by applying this Mispricing Measure technique. Evgeniou et al. (2017) conclude that the Mispricing Measure, although efficient for regular stocks, might not be applicable to the special case of repurchased stocks.
Therefore, Evgeniou et al. use an enhanced version of the Undervaluation Index composed by Peyer and Vermaelen (2009). Peyer and Vermaelen (2009, p.1705) argue that firms’ stock prices that were penalized before the repurchase announcement, experience the largest positive abnormal post-announcement returns in the long run. Thus, they establish empirical evidence for the overreaction hypothesis that “managers use repurchases to respond to market overreaction of bad news”. This hypothesis states that prior returns are the best predictors of long-term returns. Therefore, the data will be partitioned into quintiles based on the stock performance from the sixth month prior to the announcement until one month before. In
expectation, the quintiles of lower prior returns will show higher buy-and-hold abnormal returns in the long-run.
The Undervaluation Index (U-Index) is comprised of a ranking (low to high) of companies on several characteristics based on the entire distribution of all CRSP firms.
Consequently, cutoffs are made on the quintiles of this population for the book-to-market value (5-1), size (5-1, proxied by Market Capitalization), previous stock price performance (1-5) and the stated motivation in the repurchase announcement (1,3 or 5 according to the relation with undervaluation). In this research, a restricted approach of the U-index is followed, companies are classified on Size (Market Capitalization), Book-to-Market ratio, and Previous Raw Returns. This is similar to the restricted approach of Evgeniou et al. (2017). According to Peyer and Vermaelen, post-announcement returns are positively related to the index and give reason to believe that managers are able to time the market when markets have overreacted to bad news. In an efficient market, as soon as the signal of mispricing would arrive to the market, in the form of the authorization of the repurchase program, the share price would rise to the equilibrium fair value resolving the mispricing and no actual repurchasing of shares would have to happen.
In expectation, by announcing a repurchase program, firms would attract attention from investors, who would research the firm on the merits of mispricing. This would result in an increase in the stock price if investors think that the firm was undervalued at time of the announcement. The announcement month return [0,0] captures the efficiency of the incorporation of new information in the market price, given that a firm is undervalued. In other words, it measures the closing of the mispricing gap. That is why the announcement month is included in the calculation for the abnormal returns ([0,24] & [0,36]). In theory, cases where the company doesn’t experience a closing of the mispricing gap around the announcement, companies can prove this to investors by actually repurchasing shares.
As an additional analysis for the robustness of my research, the independent variables from my full model will be tested on sub samples of undervalued stocks. Undervaluation will be measured by a ranking on size, book-to-market and past stock price performance for the six months before the announcement month according to the methodology of Peyer and Vermaelen (2009, p. 1706).
3.4 Econometric issues
In addition to the sample selection criteria tackled in the first part of the methodology section, further test on normality, multicollinearity and heteroscedasticity have been performed
to assess the appropriateness of OLS regressions and ensure BLUE. After every regression, a variance inflation vector (VIF) test is run to augment the potential multicollinearity between the variables. In the few cases where higher VIF values arise, it can be safely ignored. This is due to the fact that those VIF values concern control factors or variables that are products of other variables, these types of multicollinearity have no adverse consequences (Allison, 1998).
However, in cases where I suspect multicollinearity, variables are dropped, this is discussed in detail in the next chapter. Furthermore, problems arise in the skewness and kurtosis of the resulting panel of independent variables, the abnormal returns. In order to solve these issues, the independent variables are winsorized, this reduces the kurtosis. However, it remains a limitation, as the external validity of the tails of my sample is reduced. The potentially impactful issues arise in the non-normality of the residuals after running Shapiro-Wilk tests on the regression outcomes. Accordingly, the p-values from these tests become less trustworthy (Woolridge, 2015). The concerns for heteroscedasticity and non-normality are addressed by using White’s heteroscedasticity consistent standard errors in the regressions.
4 DATA
4.1 Price series data
For the entire period, from 2002-2016, the monthly holding period returns (CRSP item TMAVERET) have been used as a proxy for returns, since these account for stock splits and include dividends. Moreover, these resolve the issue of missing daily prices for certain stocks, since the holding period return is a calendar month measure. One of the limitations of this setup of the data is that the return around the days of the announcement is aggregated to the monthly holding period return, which is labeled buy-and-hold return [0,0] in the return data series.
4.2 Data on Abnormal Returns
The dependent variable in this thesis’ model is the abnormal return as observed over the various time periods after the announcement. The expected return estimates are based on the 5-year beta and the monthly holding period observed on the CRSP equal-weighed market index. The choice of the market index was largely based on comparison of the best fitted beta estimates of the sample on the different indices (CRSP equal- and value-weighted and S&P500). The CRSP equal-weighted index is chosen, since the market beta estimates of this index provided the highest fit (average R-square of 0.42) with the sample over the period 1999- 2017 in comparison to the S&P500 (average R-square of 0.21) and the CRSP value-weighted index (average R-square of 0.22). Consequently, the 5-year beta as estimated from the CRSP equal-weighted index provides more predictive power in the “normal returns” for the post- event period than does the S&P500 or the CRSP value-weighted indices. Furthermore, from Empirical Asset Pricing theory it is known that the equal-weighted index provides exposure to more risk factors, the most exposed being the small-firm effect (Plyakha et al., 2014). Since a considerable number of firms in the sample are not included in the S&P500 index, the equal- weighted CRSP benchmark serves as a better benchmark. Table 7 in Appendix A provides an overview of the descriptive statistics of the expected returns, the actual returns and the abnormal returns for the pre-announcement period and the post announcement period. The long-term returns, for the time period [0,36] have a skewness of 175.7 and a kurtosis of -7.5, suggesting that the distribution of these returns do not follow a normal distribution. As a solution the abnormal returns are winsorized. This seriously reduces the kurtosis (skewness) of the three-year BHARs, from 175.7 to 4.7 (-7.5 to 0.3), making it more in line with a normal distribution.
4.3 Data on the independent variables
The size of the repurchase announcement – Size Variable
From the Zephyr database, the variables announcement date and the size of announcement are extracted. The Zephyr database calculates the repurchase announcement based on the market Furthermore, different panels are constructed based on low repurchase fraction (0.1% - 2%, Panel B) medium repurchase fraction (2% - 7%, Panel C) and high repurchase fraction (>7%, Panel D), see table 10 in Appendix A. The high fraction sample is approximately the largest tercile of all the announcements in my sample. Whereas the low fraction is stratified based on the belief that the reason is not for signaling undervaluation (O’shaughnessy, 2015). For all the firm events in each sample, buy-and-hold abnormal returns are regressed for the three years after the announcement.
Actual repurchases – Actual Repurchases Variable
The data for analyzing the actual repurchases of the program is derived from the Compustat measure “Purchases of Common and Preferred Stock” (Compustat item 115). This has been widely used in academia [Dittmar, 2000; Grullon and Michaely, 2004; Jagannathan et al. 2000; Stephens and Weisbach, 1998; and Kulchania, 2012; Bonaimé and Ryngaert, 2013;
Caton et al., 2016]. Stephens and Weisbach (1998) point out some differences and side issues that this measure for the repurchase fractions brings along. First, it is measured in value terms instead of number of shares. No information is given for the price the firm repurchased its stock, thus it neglects the timing that firms may have had in buying the undervalued stock.
Second, it is an aggregate of all quarterly repurchases and retirements and can overstate the purchase of common stock, my variable of interest. Therefore, adjustments are made to take into account the increase per quarter and the corrections for the change in preferred stocks.
Through equation 4 the purchase of Common Stock is separated from the purchase of Preferred Stock. Then, the aggregate Purchases of Common Stock of the four and eight quarters is divided by the Market Value in the quarter of the announcement. Thereby, it is measured against the same fraction as in the announced repurchase authorization. This caters the ability to estimate the completion of the repurchase fraction trough a continuous scale. Dividing the actual repurchased fraction (equation 5), during four or eight quarters, by the size of the open market program at announcement gives the completion rate6 on a continuous scale (equation 6). In addition, the four quarter purchases are regressed on the BHAR for two years and the
6 Completion rate and actual repurchases are interchangeably in the rest of this thesis.
eight quarter purchases are regressed on the three-year BHAR. It doesn’t make sense to estimate the completion rate after the realization of the buy-and-hold abnormal returns (e.g.:
four quarters on two years is one quarter overestimated).
𝑃𝑢𝑟𝑐ℎ𝑎𝑠𝑒 𝑜𝑓 𝐶𝑜𝑚𝑚𝑜𝑛 𝑆𝑡𝑜𝑐𝑘𝑡 = ∑ 𝑃𝑢𝑟𝑐ℎ𝑎𝑠𝑒 𝑜𝑓 𝐶𝑜𝑚𝑚𝑜𝑛 𝑎𝑛𝑑 𝑃𝑟𝑒𝑓𝑒𝑟𝑟𝑒𝑑 𝑆𝑡𝑜𝑐𝑘𝑞,𝑖
4
𝑞=1
−
∆𝑃𝑟𝑒𝑓𝑒𝑟𝑟𝑒𝑑 𝑆𝑡𝑜𝑐𝑘𝑞,𝑖,
Equation 4
𝐴𝑐𝑡𝑢𝑎𝑙 𝑅𝑒𝑝𝑢𝑟𝑐ℎ𝑎𝑠𝑒 𝑓𝑟𝑎𝑐𝑡𝑖𝑜𝑛𝑖,𝑡= 𝑃𝑢𝑟𝑐ℎ𝑎𝑠𝑒 𝑜𝑓 𝐶𝑜𝑚𝑚𝑜𝑛 𝑆𝑡𝑜𝑐𝑘𝑡
𝑀𝑎𝑟𝑘𝑒𝑡 𝑉𝑎𝑙𝑢𝑒𝑞=0 , Equation 5
𝐶𝑜𝑚𝑝𝑙𝑒𝑡𝑖𝑜𝑛 𝑅𝑎𝑡𝑒𝑖,𝑡 =𝐴𝑐𝑡𝑢𝑎𝑙 𝑅𝑒𝑝𝑢𝑟𝑐ℎ𝑎𝑠𝑒 𝐹𝑟𝑎𝑐𝑡𝑖𝑜𝑛𝑖,𝑡
𝐴𝑛𝑛𝑜𝑢𝑛𝑐𝑒𝑑 𝐹𝑟𝑎𝑐𝑡𝑖𝑜𝑛 , Equation 6
Where 𝑞 is quarter, 𝑖 is the firm and 𝑡 resembles the year after the announcement, in this case I control for one and two years. An important limitation of my data is the overestimation of program completion due to multiple repurchase programs by the same firm. Therefore, program completion rates of more than 100% in eight quarters may be due to later initiated programs. However, many firms continue to repurchase under an existing mandate after announcing new programs, consequently, I have chosen to include eight quarters of actual repurchases after the program announcement in my model.
Alternatively, a measure based on the decrease of the shares outstanding (CRSP item
“SHROUT”) can be used. This simply estimates the monthly decrease in shares outstanding.
The method stems from Jiang et al. (2013) and is applied more than once. This measure develops a good estimate for the fractions of shares repurchased, although it is prone to measurement error. In detail, by the “SHROUT” measure, increases in the number of shares outstanding – caused by stock option plans or employee benefits – in the contemporaneous period, are neglected. Thus, the fraction of shares repurchased is underestimated. Therefore, I choose not to adopt this measure in my model7.
7 The data on this alternative variable based on a monthly decrease in shares outstanding has been gathered in accordance with previous research (Jiang et al. 2013). The outcomes from regression are mentioned in the results section of the corresponding independent variable, however, they will not be discussed in detail. The data is available upon request.
Frequency of announcements – Infrequent Repurchaser Variable
The terms used in this paper for the frequency of a repurchasing firm are “infrequent” or
“occasional” and “frequent”. The estimation of frequency follows a combination of the empirical framework of Yook (2010) and Jagannathan and Stephens (2003). However, I make an adjustment to increase the accuracy of the frequency measure by adopting a backward- looking measure of frequency. The backward-looking measure is the most realistic form as it represents the information that is present at the time of the announcement, this variable is included in the model. “Infrequent” under the backward-looking measure are repurchasers that have not repurchased in the last three years. In classifying the repurchasers I look back further than the start of my sampling period. The sampling period for this variable starts from 2000 onwards to be able to classify the frequency of the first year in the research sample.
SEO-activity – D_SEO dummy Variable
The SEO activity is represented by a dummy, if the repurchase announcement was preceded by the announcement of a seasoned equity offering in the same industry in 6 months prior to the repurchase announcement. The SEO announcement date is used instead of the actual offering date, since that is the moment that the public information is incorporated in the market.
The announcement data is gathered through the Zephyr database for the contemporaneous period of my sample data, starting six months before the first repurchase announcement. In the data selection, all deals were collected from Zephyr database. Specifically, all SEO deals involving bailouts (e.g. TARP) or any deal where the vendor was the US Government were excluded since these don’t proxy for increasing growth opportunities within the market but are an exit for the Treasury Department in bail-out programs. Moreover, Secondary offerings with the purpose of paying of debt were excluded from this sub sample. Pham et al. (2019) have a similar approach in their research.
