Stock Repurchases in Canada: The Effect of History and Disclosure Comments welcome! James M. Moore PhD Candidate University of Waterloo October 10, 2005 jmooreca@sympatico.ca ABSTRACT Open market share repurchases (OMSR) are a common corporate event. Preliminary data collected in this study indicates that 10 to 20% of TSE listed firms announced a buyback program of their own stock during each of the years 1998 to 2004. Prior research on Canadian and U.S. share repurchases has found a positive reaction to the repurchase announcements, as demonstrated by the existence of cumulative abnormal returns surrounding the announcement and over the subsequent repurchase period. This evidence has often been interpreted as consistent with a signalling mechanism. However, firms often announce share repurchase programs, but then repurchase few or no shares raising the possibility that false signalling may be occurring. The Canadian repurchase environment differs from the U.S. environment in two important respects. First, there are mandatory disclosures that must be included in the repurchase announcement. Second, the number of shares actually repurchased must be reported on a monthly basis. These requirements may assist investors in assessing a current announcement. My proposed research focuses on whether these additional Canadian disclosures have information content. I propose to test information content broadly by examining the association of these additional disclosures to returns, trading volume and bid-ask spreads, both during the announcement period and the actual repurchase period. Evidence on the information content of these disclosures is useful to Canadian regulators in assessing the value of maintaining these disclosures and is useful to U.S. regulators who are currently considering similar requirements. Evidence on the effect of these disclosures also assists in assessing the two main theories used to explain stock repurchases, signalling theory and free cash flow theory.
1. Introduction Open market share repurchases (OMSR) are a common corporate event. In 1998, U.S. corporations announced plans to purchase $220 billion of their own shares 1. Grullon and Michaely (2002) document that total market wide share repurchases surpassed dividend payouts in the U.S. during 1999 and 2000. Preliminary data collected in this study indicates that 10 to 20% of TSE listed firms announced a buyback program of their own stock during each of the years 1998 to 2004. Most prior research on Canadian and U.S. firms has found a positive reaction to the repurchase announcements, as demonstrated by the existence of cumulative abnormal returns surrounding the announcement and over the subsequent repurchase period. This evidence has often been interpreted as consistent with a signalling mechanism. However, firms often announce share repurchase programs, but then repurchase few or no shares raising the possibility that false signalling may be occurring. The Canadian repurchase environment differs from the U.S. environment in two important respects. First, there are mandatory disclosures that must be included in the repurchase announcement. These disclosures include the type and number of shares to be repurchased, the reason for the repurchase and the number and cost of any repurchases within the last twelve months. Second, the number of shares actually repurchased must be reported on a monthly basis during the repurchase period. In determining the credibility of a current announcement, the requirement to disclose actual repurchases may be important since if a firm has followed through (not followed through) on past repurchase announcements, their current announcement may be viewed as credible (not credible). My proposed research focuses on whether these two specific disclosures required by TSE regulations but not by U.S. regulations have information content. 1 Ed McCarthy, 1999. Stock Buybacks: The Rules, Journal of Accountancy, May 1999, p.91. 2
I define information content similarly to Beaver (1968). Beaver (1968) considers earnings reports to have information content if the earnings release result in a price change or increased trading volume. Beaver considers both price and volume reactions since price reactions reflect aggregate changes in market beliefs while volume changes reflect individual changes in beliefs. An earnings release can still be considered to have information content if it results in an increase in trading volume even if there is no change in price. I apply a similar approach to repurchase announcements but also consider changes in bid-ask spreads. Bid-ask spreads reflect the premium charged by the market maker for dealing with informed traders. A reduction (increase) in bid-ask spread is considered to be a reduction (increase) in information asymmetry. I focus on whether the TSE requirement to disclose actual past repurchases and the reasons for the repurchase convey information incremental to the announcement itself. By testing whether these two components of the TSE requirement can be associated with changes in returns, volume and bid-ask spreads, I examine whether these two components have information content incremental to the announcement itself. Evidence on the markets reaction to the information contained in these disclosures is useful to U.S. regulators who are currently considering a requirement to have U.S. firms provide quarterly data on their repurchases activity 2. Evidence on the effect of these disclosures also assists in assessing the two main theories used to explain stock repurchases, signalling theory and free cash flow theory. 2. The Canadian Regulatory Environment An overall comparison of the requirements of Canadian and U.S. regulations for stock repurchases is provided in Appendix1. I focus on the effects of two unique Canadian requirements, the requirement to disclose actual repurchases and the requirement to disclose the 2 E. Lie. 2005. Operating Performance Following Open Market Share Repurchase Announcements. Journal of Accounting and Economics (Vol. 39), p.423. 3
reason for the repurchase. The specific announcement disclosure requirements for TSE firms are contained in Appendix 2. A sample of one firm s repurchase announcement is contained in Appendix 3. Most past research has focused on U.S. firms, but due to there not being equivalent repurchase program announcement requirement in the U.S., this research suffers from data problems. To overcome these data problems, Stephens and Weisbach (1998) use a variety of measures to approximate actual repurchases but their work is relatively recent and still involves approximation of the number of shares repurchased. Additionally, Canadian firms are required to disclose the reason(s) for their repurchase program. These reasons are summarized in Appendix 4 and include preventing dilution due to stock options, utilizing excess funds, indicating that the firm feels it shares are undervalued and to increase the proportionate interest of remaining shareholders. In addition to studying the effects of history, I also propose to investigate whether these disclosed reasons are important to the market. In this context, I will examine whether the disclosed reasons are associated with differences in returns, trading volume and bid-ask spreads during the announcement period and during the repurchase period. Use of Canadian data has advantages. As Ikenberry, Lakonishok and Vermaelen (2000) point out, Canada is the only other country that has a reasonable sample of repurchases 3. As a result, use of Canadian data allows researchers to confirm findings that have only been previously established using U.S. data. Unlike the approach used by Stephens and Weisbach (1998), no approximation of the number of shares repurchased is required, thus reducing the potential for measurement error. As noted in section 6, my sample is also very recent and includes a sub-period where repurchasing activity begins to decline. Hence, I will be able to confirm past empirical results in a substantially different setting. 3 Ikenberry,Lakonishok and Vermaelen. 2000. Stock Repurchases in Canada: Performance and Strategic Trading. Journal of Finance (Vol.55, No.5), p.2374. 4
3. Literature Review 3.1 Theory/Analytical Papers Several theories have been used in the literature to explain the motivation to repurchase stock and to explain the documented positive market reaction to stock repurchase announcements 4. Signalling theory holds that the repurchase announcement conveys new information to the market. As a result of receiving this new information, market participants react by re-evaluating their views of the firms future cash flows. If the news indicates that future cash flows are expected to be higher (lower) than previously thought, the price of the firm will rise (fall). Similar price behaviour can be explained by the free cash flow theory first proposed by Jensen (1986). Under the free cash flow theory, a positive market reaction to a stock repurchase announcement is attributed to the market interpreting the announcement as a commitment by management to reduce the agency costs associated with excess funds. A major difference between these two theories is that managers are typically assumed to altruistically maximize shareholder value under the signalling theory, while they are inclined to maximize their personal utility under the free cash flow theory. A more detailed discussion of these two theories is now presented, along with a discussion of other theories used to explain repurchases. a. Signalling Theory. Signalling theory emerged from the study of information economics under conditions in which buyers and sellers possess asymmetric information when facing a market interaction. Akerlof (1970) presents an important initial analysis on the used car market and demonstrates that information asymmetry can lead to market failure. Largely for his initial development of the theory in Spence (1973) and for his subsequent contributions to the development of signalling theory, Michael Spence was awarded the Nobel Prize in economics. 4 Nohel, E. and Tarhan, V. 2000. Share Repurchases and Firm Performance. Journal of Financial Economics 49, p.188. 5
Signalling theory has been applied to large number of settings such as job markets, consumer settings and share repurchases. Typically, in signalling scenarios, the seller knows the quality of the goods they are selling, but the buyer is unsure. The buyer would like to obtain information that allows them to separate the seller of high quality goods from the seller of low quality goods. If they are able to separate the seller of high quality goods from the seller of low quality goods, a separating equilibrium may occur. Oded (2005) notes In a separating equilibrium, the firm type is known to the market 5. To distinguish their goods as high quality, high quality sellers would like to be able to send a signal of high quality to potential buyers. However, low quality sellers would also like to send a signal of high quality in order to obtain the higher price that higher quality commands. In response to these incentives, the buyer seeks information (a signal) to resolve their quality concerns. If firms are rational, the buyer can look for signals that a high quality firm would find profitable to send, but a low quality firm would find unprofitable to send. In an environment where there is a cost to false signalling, a separating equilibrium will occur. The basic condition needed for a Spencian signalling solution is that the marginal cost (and or benefit) of the signal is decreasing (increasing) in true quality. In a separating equilibrium, market incentives lead high and low quality firms to send different signals. However, where there are no costs to false signalling, a pooling equilibrium will occur whereby all sellers utilize the same strategy and the buyer is hence unable to separate high from low quality. Prior studies focus on the costs that repurchasing itself incurs. If a firm is of low quality (its shares are overvalued), but tries to signal that it is high quality, in order to send a false signal to the market, the low quality firm will incur high costs because it will have to purchase overvalued shares in order to back up the signal. This is costly since the shares are priced at higher 5 Oded, J. 2005. Why do Firms Announce Open-Market Repurchase Programs? The Review of Financial Studies (Vol.18, No.1), p.288. 6
than they are worth. On the other hand, if the firm is of high quality and its shares are undervalued by the market, it need only purchase under-valued shares which are less costly to purchase than over-valued shares. Hence, the costs of the signal are decreasing in true quality, the key Spencian requirement for a separating equilibrium. One problem with this argument is that the market in the U.S. has a hard time determining if a firm is actually backing up its claim of undervaluation by repurchasing shares because there is no requirement to disclose actual repurchases to the market. I expand on the cost of signalling argument by considering another cost of false signalling, that is that the firm s credibility with respect to future repurchase announcements will suffer if the firm does not follow through on its current repurchase announcement. Credibility could be very important to obtaining a separating equilibrium since making non-credible announcements that will be subsequently discovered increases the cost of the signal since their will be a penalty imposed for false signalling (loss of reputation). If the buyer does not believe that the seller will incur costs from false signalling that exceed the benefits from false signalling, the signal will not be believed and a pooling equilibrium will result or there could be a complete market failure as in Akerlof (1970). My study considers the impact of credibility. I use the past follow through rate on repurchase announcements as a proxy for the credibility of the current announcement. Bhattacharya and Dittmar (2003) provide an analysis of the costs of signalling that differs from prior literature in its assumptions about the costs of signalling. They separate firms into two types: firms who announce repurchases and then make no repurchases and firms who announce repurchases and then do repurchase shares. They consider the first class of firms to be cheap talk firms and take the position that they incur no costs of signalling since they do not actually repurchase any shares. The no-costs of signalling view is in contrast to previous literature on signalling that assumes that there is a cost of signalling. The second group, who 7
actually repurchase shares, are considered to incur costs. Bhattacharya and Dittmar (2003) ignore the reputation effects of failing to follow through with their announcements, a consideration that I address in my research. They predict, and find empirically, that cheap talk repurchase announcements only work for firms that are very undervalued and ignored since only then will trading profits exceed the costs of scrutiny. Costs of scrutiny are the costs associated with uncovering information on the true value of the firm. Bhattacharya and Dittmar (2003) assume that there is other information available that will allow market participants to determine the true value of firms who engage in cheap talk. This assumption can be challenged in that it assumes that there is publicly information that is not already incorporated into the cheap talk firms price. Bhattacharya and Dittmar (2003) take the position that while the information may be publicly available, it will only be incorporated in price if the benefits of analysing the publicly available information exceed its price. A repurchase announcement signals that the benefits of this analysis may be worth the costs and may therefore prompt market participants to analyse already available information. Their empirical results support their predictions. I expand on their work by considering a cost of signalling that Bhattarcharya and Dittmar (2003) do not consider, that cost being the reputation cost of announcing a repurchase program and then not following through. b. Free Cash Flow Theory. Under free cash flow theory, firms with excess cash should distribute that cash to avoid the agency costs associated with the tendency of management to invest in self-serving investments or to utilize funds for perks, empire building, and excessive salaries. A positive market reaction to a stock repurchase announcement is attributed to a commitment by management to reduce these costs. Jensen (1986) makes the following prediction: Free cash flow theory predicts that, except for firms with profitable unfunded investment projects, prices will rise with unexpected payouts to shareholders or promises to do 8
so, and prices will fall with reductions in payments or new requests for funds or reductions in promises to make future payments 6. Like signalling theory, free cash flow theory predicts an increase in share prices when a repurchase announcement is made. It is important to note that voluntary repurchases to reduce agency costs are only one mechanism among many that can be used to control agency costs. Other mechanisms, such as the market for corporate control and the management market also serve to constrain agency costs. c. Other Repurchase Theory. Ikenberry and Vermaelen (1996) suggest that the documented positive reaction to stock repurchase announcements may be a result of management, through the repurchase program, being able to create an option to repurchase shares. Such an option is valuable when the share price falls below true value, or when repurchases can substitute for heavily taxed dividends. The option is more valuable the larger the potential for mis-pricing is. Oded (2005) provides further analysis of Ikenberry and Vermaelen (1996). Oded s model includes both the short-term costs of announcing and the long-term gains from informed trading. Oded notes that if the repurchase announcement gives the firm a valuable option, the market is short that option. As a result, no return should occur since the creation of this option represents a zero-sum game. This view ignores potential tax savings since if dividends are taxed higher than capital gains, there is still a wealth increasing effect for shareholders since after tax proceeds will be higher. Despite the repurchase option being a zero sum game overall, there is some wealth transfer between short-term and long-term shareholders. Since short-term shareholders will sell for less than the true value of the share during the repurchase period, they pay the cost of the option. Long-term shareholders enjoy the benefits of informed trading since they benefit from the firm repurchasing its shares below true value. Oded then constructs a signalling model 6 Jensen, M.C. 1986. Agency Costs of Free Cash Flow, Corporate Finance and Takeovers. American Economic Review 76, p.325. 9
where in equilibrium good firms will announce their repurchase programs, but bad firms will not. It is important to note, however, that Oded assumes that disclosure is voluntary. This is not the case for TSE firms. TSE firms are required to announce their share repurchase programs. 3.2 Empirical Considerable empirical literature exists on stock repurchases. Two streams of literature are readily identifiable, although many papers contribute to both streams. The first stream of literature considers the motivations to repurchase stock. The second stream focuses on the market reaction to repurchase announcements and actual repurchases. Many papers focus on the applicability of signalling theory and cash flow theory in explaining the market reaction. I review the empirical literature in chronological order of publication. Dann (1981) finds evidence of a significant increase in firm value within one day of a tender offer announcement. He attributes the increase in value to the announcement being a signal to the market that the firm is undervalued. Vermaelen (1981) was published almost concurrently with Dann (1981) and provides a similar finding on open market repurchases thus reinforcing the notion that repurchase announcements act as a signalling device. Bartov (1991) reports positive revisions to analyst forecasts and a lowering of systematic risk following an OMSR announcement, results consistent with the announcement having a signalling value. Ikenberry, Lakinoshok and Vermaelen (1995) examine long run returns performance following an OMSR announcement. They find an average four-year buy and hold abnormal return following an OMSR announcement of 12.1%. The abnormal return is 45.3% for high book-tomarket value stocks, but insignificant abnormal returns for low book-to-market glamour stocks. They interpret this as a market error to the initial announcement for value stocks. Bandyopadhyay and Kao (1996) provide the first analysis of the role of history in stock repurchase programs. They find that while all announcing firms enjoy abnormal returns during 10
the announcement period, firms that have made repurchases under past repurchase programs enjoy greater returns over the repurchase period than firms that did not make repurchases. I extend their work by using a continuous variable for repurchase history, by considering the effect of history on trading volumes and bid-ask spreads and by considering the effect of the disclosed reason(s) for the repurchase. Ho, Liu and Ramanan (1997) study a small sample of OMSR announcements occurring during 1978 to 1992. They find that the market reaction to the announcement is significantly associated with prior accounting performance, particularly past sales growth and profitability. They interpret this finding as evidence that the repurchase announcement acts as a signal that prompts the market to re-evaluate already known information from the financial statements. I extend their findings by considering whether the market will reevaluate already known information on the number of shares actually repurchased under prior repurchase programs. Ho, Liu and Ramanan (1997) also present evidence that this re-evaluation is more significant for firms with higher information asymmetry. They use firm size and analyst forecast dispersion as proxies for information asymmetry. Kirch, BarNiv and Zucca (1998) consider the effect of the number of shares actually repurchased on returns over the repurchase period. They find evidence that firms that actually repurchase stock have higher abnormal returns over the repurchase period than firms that do not repurchase stock. In other words, if the firm actually goes ahead with its planned repurchases, which is unknown to investors at the time of the announcement, returns will be higher. This is not surprising given the demand effects created by a firm entering into the market to purchase its own stock. They also find that non-fulfilling firms are less profitable before and after the OMSR program announcement. Barth and Kasnik (1999) find that firms with more intangible assets and idle cash are more likely to repurchase shares. They find that firms with more intangible assets enjoy higher announcement returns. Contrary to their expectations, they find that 11
information asymmetry is negatively related to repurchase likelihood. However, where information asymmetry is higher prior to the announcement, returns over the announcement window are higher. Li and McNally (1999) consider the characteristics of Canadian OMSR firms. They find evidence that repurchasing firms are smaller, have greater free cash flow and are more closely held than non-repurchasing firms. They interpret this evidence as being consistent with the cash flow hypothesis and inconsistent with the signalling hypothesis since their sample firms have characteristics of firms who have excess cash to distribute and mechanisms to control management, as opposed to characteristics of firms who are undervalued. Ikenberry, Lakonishok and Vermaelen (2000) study 1,060 Canadian OMSR announcements during the 1990s. They find evidence of a small announcement effect in the month of the repurchase announcement. By examining long-term windows coinciding with the repurchase period, they find evidence that, similar to the U.S., the market discounts the evidence in the repurchase announcements for value stocks. Using monthly repurchase data available in the TSE Monthly Review, they find that when prices fall, managers purchase more shares consistent with management backing up its undervaluation signal to the market with action. Their finding would be hard to replicate using U.S. data since U.S. firms are not required to disclose their repurchases. Guay and Harford (2000) find that firms use dividend increases to payout relatively permanent increases in cash flow and repurchases to payout transient increases in cash flow. Similarly, Jagannathan, Stephens and Weisbach (2000) also find that firms use dividends to payout permanent operating cash flows but use share repurchases to payout temporary, nonoperating cash flows. Brockman and Chung (2001) study the effect of repurchase announcements on the bid-ask spreads and market depth of Hong Kong listed firms. They find that during repurchase periods, bid-ask spreads widen and depths narrow. Further decomposition of the bid-ask spreads show 12
that adverse selection costs increase during repurchases, consistent with the presence of information asymmetry resulting from the presence of informed management trading. In contrast Ahn, Cao and Choe (2001) find that bid-ask spreads decline during tender offers, consistent with the competing market maker hypothesis, under which intensified competition for purchases reduces the bid-side spread. Grullon and Michaely (2002) find evidence that firms finance share repurchases with funds that would have otherwise have been used to increase dividends. Their evidence is consistent with firms increasingly using share repurchases instead of dividends as a means of returning funds to shareholders. They explain this trend using the substitution hypothesis that holds that firms will increasingly use repurchases instead of dividends because repurchases are more flexible than dividends. Jagannathan and Stephens (2003) study motives for multiple OMSR programs. They finds that frequent repurchasers are larger, have more stable operating income and higher dividend payout ratios. They also find that while the market views most repurchase announcements favourably, infrequent repurchasers announcements are the most positively viewed. Jagannathan and Stephens (2003) consider only the frequency of repurchasing, not the actual number of shares repurchased versus the targeted amount. This limitation is likely a result of U.S. firms not being required to report their repurchases on a monthly basis, thus limiting the data that researchers have. I extend their work by considering the actual number of shares repurchased, not just the frequency of announcements, and by controlling for the disclosed reasons for the repurchase. I also consider the effects of history and disclosure on trading volume and bid-ask spreads. Grullon and Michaely (2004) find that repurchasing firms operating performance, capital expenditures and research and development expenses decline compared to their peer firms. They interpret this as evidence in favour of Jensen s (1986) free cash flow theory. Lie (2005) finds that improvements in performance are limited to firms who actually repurchase shares during the same fiscal quarter. 13
Lie s (2005) results suggest that actual repurchases, not announcements, are more important in forecasting future performance. In his conclusion, Lie (2005) points out the benefits of a study that addresses the actual follow through rate of firms. My study partially addresses Lie s call for research into follow through rates. 4. The Effect of History and Disclosure on the Market Reaction to an OMSR Announcement: Theory, Hypotheses and Testing The objective of my research is to study the information content of Canadian stock repurchase disclosures. In section 4.1, I discuss why the signalling and free cash flow theories indicate that TSE stock repurchase disclosures should contain information. In section 4.2, I discuss empirical studies supporting the information content of TSE disclosures. I then present my returns hypotheses (section 4.3), my volume hypotheses (section 4.4) and my bid-ask spread hypotheses (section 4.5). I conclude with a discussion of why Canadian stock repurchase disclosures may not have information content (section 4.6). 4.1 Application of Signalling and Free Cash Flow Theories to TSE Requirements Signalling theory provides insights in the case of TSE open market share repurchase announcements. Management is representing the firm, which is selling its securities to investors. Management has an incentive to see the value of the firm maximized, so can be expected to make announcements (signals), such as share repurchase programs, which maximize the value of the firm. Investors realize these incentives and seek information to evaluate the credibility of the signals sent. In Canada, due to the requirement to report actual stock repurchases once a repurchase program has commenced and due to the requirement to disclose repurchases during the last twelve months as part of a repurchase announcement press release, investors have information on whether the current announcement is likely to be credible. Firms who have announced repurchase programs in the past but not repurchased any stock will stand out as less 14
credible, while firms that have a good history of fulfilling their repurchase announcements will be viewed as high quality. As noted earlier, the basic condition needed for a Spencian signalling solution is that the marginal cost (and or benefit) of the signal is decreasing (increasing) in true quality. Firms have an incentive to send credible signals since sending false signals will be visible, thus making it less likely that a subsequent repurchase announcement will be believed. Nevertheless it is still possible that some firms will attempt to fool the market by announcing repurchase programs that they have no intention of fulfilling. In this respect, the TSE requirement to disclose actual repurchases may fulfill an important role in constraining any incentives a firm may have to announce a repurchase program they have no intention of fulfilling. By imposing potential costs on non-fulfillers and by providing feedback on past repurchase programs TSE regulations may help markets to obtain a separating equilibrium. Specifically, a firm who intends on actually repurchasing it shares will incur a low cost to signal since it will not incur any reputation costs by making the announcement, since it will follow through with the announcement. On the other hand, firms who have no intention of repurchasing will undergo substantial reputation costs since the firm will have set itself up to be not believed with respect to future repurchase announcements. Evidence that Canadian firms historical repurchase disclosures affect the market reaction to the repurchase announcement would be consistent with those disclosures providing information to the market, and consistent with the predictions of signalling theory. On the other hand, signalling theory has some problems in being applied to the Canadian environment. Signalling theory works on the notion that management is deliberately trying to send information to the market. As Bernheim and Wantz (1995) point out on their tests of dividend signalling, it is difficult to separate deliberate attempts to release information from non-deliberate releases of information that still convey information. It is possible that management is not making a deliberate attempt to signal information to the market 15
when announcing a dividend increase: This evidence does establish that information is revealed through dividend announcements. However, it does not distinguish between the hypothesis that the dividend signals good news and the hypothesis that the dividend is the good news or conveys good news even though this is not the intent of management. Signalling is a very specific form of information revelation. 7 This analysis applies equally to stock repurchases. Management may announce the repurchase, not in attempt to signal undervaluation, but for other reasons such as to prevent dilution due to stock option plans, acquisitions, etc. (see Appendix 4). The market then interprets the announcement as good news, even though there was no real intention to signal good news. In this respect, the TSE requirement to disclose the reason for the repurchase may be helpful in separating out whether management is deliberately signalling undervaluation. In my third hypothesis, I predict that where the undervaluation is specifically cited as the reason for the repurchase, this reason will be incrementally informative in explaining cumulative abnormal returns. I make this prediction on the grounds that signalling theory is more likely to be applicable where management has specifically made it clear that it is signalling undervaluation. In interpreting free cash flow theory, it is important to consider that a repurchase announcement is a non-binding commitment to increase future payouts which will in turn reduce agency costs. Whether or not a firm has carried through on its past announcements may be indicative of whether it will do so in the future. In this respect, I expect the Canadian requirement to disclose past repurchases to have information content. For all announcing firms, I expect returns to be positively related to cash. I expect past repurchase follow through history to be positively related to announcement returns since a good repurchase history should increase 7 Bernheim, D. and Wantz, A. 1995. A Tax-Based Test of the Dividend Signalling Hypothesis. The American Economic Review (June, Vol.85, No.3): p.p.532-533. 16
the confidence of the market that the current promise of a payout of funds will result in an actual payout of funds. 4.2 Empirical Studies Supporting the Information Content of TSE Regulations My contention that prior repurchases may have information content is similar to Banker, Das and Datar (1993) who report that prior accounting information, such as dividend history and capital expenditures, is useful in explaining cross-sectional variations in the market response to current stock dividend announcements. Their results are consistent with investors considering the credibility of a signal before acting on it. If firms have made announcements of repurchase programs in the past but not actually repurchased stock, their current announcement may not be viewed as credible and hence investors will not react positively to it and may even punish the announcer for perceived false signalling. The idea that history may be relevant to the market is also consistent with the empirical findings of Bandyopadhyay and Kao (1996) who find evidence that past repurchase history is associated with the subsequent returns of announcing firms. It is also consistent with the findings of Ho, Liu and Ramanan (1997) who find that the market reevaluates known accounting information at the time of a repurchase announcement. I extend their work by considering whether the market considers known repurchase history. 4.3(a) Returns Hypotheses and Testing Announcement Window. To determine if repurchase history shows evidence of information content, I will begin by comparing the cumulative abnormal returns of firms based on their repurchase history. To help focus the discussion in this section, the main variables of interest to this research are discussed in the text. However, control variables are defined in Appendix 5 since they are ancillary to the main purpose of this research. Inclusion of these control variables allows me to confirm past findings and to control for variables already known to be associated with the markets reaction to a repurchase announcement. Consistent with Ikenberry, Lakonishok and Vermaelen (2000), 17
cumulative abnormal returns (CAR) will be calculated using two approaches for robustness. First abnormal returns will be calculated by subtracting the TSX index total return from the stock return. Second, a three factor Fama-French (1993) model will be used that controls for general market movements, value versus growth stocks and size. I will use two variables to measure repurchase history, a binary variable (HIS2) that measures whether the firm has had prior repurchase programs and a continuous variable (REP%) for firms with a history that measures the percentage of targeted shares actually repurchased (the follow through rate ) during the most recent repurchase program. Under the binary variable (HIS2), a firm will be considered to have a repurchase history (coded 1) if its current OMSR announcement is preceded by a previous announcement in the last two years, otherwise it will be considered to have no history (coded 0). This assumes that the market has a two year memory. As a validity check, I will expand this window to three years (substituting the variable HIS3 for HIS2) and compare the results, which I expect to be similar to the two year window. Jagannathan and Stephens (2003) find that as firms announce successive repurchase programs, abnormal returns decline consistent with the market being able to anticipate subsequent repurchase announcement. My first hypothesis does not depend on TSE disclosure requirements and relies on Jagannathan and Stephens (2003) result that returns decline with subsequent announcements: H1 Cumulative abnormal returns surrounding the announcement event are negatively related to having had repurchase programs in the past. I test this hypothesis using the following OLS model: (1) CAR = b 0 + b 1 HIS2 + b 2 CON + e 1 Where CON = a vector of control variables defined in Appendix 5. A negative b 1 coefficient supports H1. HIS3 will be substituted for HIS2 as a validity check. 18
For firms that have a good history of repurchasing shares, both signalling theory and cash flow theory indicate that the return reaction should be stronger where the announcement is credible. I expect the market to view the announcement as credible when past repurchase follow through history has been good. This leads to my second hypothesis H2 For firms with a past history of repurchasing shares, cumulative abnormal returns surrounding the announcement event are positively related to the most recent repurchase follow through rate. I test H2 using the following OLS model: (2) CAR = b 0 + b 1 REP% + b 2 CON + e 1 where CON = a vector of control variables listed and defined in Appendix 5. A positive b 1 coefficient supports the prediction and is evidence that the disclosure of prior repurchases has information content. I also seek evidence on whether the firm s stated reason for repurchase has information content. As seen in Table 2, 66.3% of my sample firms cite undervaluation as a reason for their repurchase program. The remaining firms cite other, non-undervaluation reasons for their repurchase programs such as preventing dilution due to stock options or acquisitions, provide vague reasons or make no disclosures at all about their reasons. For firms who cite undervaluation as a reason for their repurchases, signalling theory is particularly applicable since it is likely that, by stating that they are undertaking the program because they are undervalued, firms are signalling to the market that the price should rise. Where the firms do not cite undervaluation as a reason for their repurchases, firms are less likely to be signalling undervaluation and past follow through (REP%) should be less significant. As noted earlier, credibility is particularly important in a signalling environment. Also, signalling theory is dependant on the signal being deliberately sent. Given that it is not necessary to mention 19
undervaluation, mentioning undervaluation is a deliberate choice and appears more consistent with a signalling mechanism than a free cash flow mechanism. Since, it is likely that a statement of undervaluation is a deliberate signal and since credibility is particularly important in signalling, I expect there to be an interactive effect between repurchase follow through rate and providing undervaluation as a reason for the repurchase. My third hypothesis is as follows: H3 The follow through rate on the most recent repurchase program will be incrementally informative where undervaluation is cited as a reason for the repurchase. To test H3, I again use an OLS model: (3) CAR = b 0 + b 1 REP% + b 2 VALREAS + b 3 REP%*VALREAS + b 4 CON + e 1 where VALREAS=1 if the firm cites undervaluation as a reason for the repurchase, 0 otherwise and CON = a vector of control variables listed and defined in Appendix 5. Therefore, the appropriate test of H3 is b 3 > 0. A significant b 3 co-efficient will indicate that the providing of a statement that the repurchase is for valuation reasons has incremental information content beyond other known factors and repurchase follow through rate.. Finally, to investigate whether the reasons cited by firms are informative to the market, I will compare both the announcement period and the repurchase period cumulative abnormal returns of firms who cite each of the reasons cited in Appendix 4. My procedures for coding these reasons are discussed in section 6(d). T-tests will be used to determine if any differences are significant. The purpose of this analysis is to determine if there are any systematic differences in returns over the announcement period and over the repurchase period for each of these reasons. 4.3 (b) Returns Testing: Repurchase Period 20
Ikenberry, Lakonishok and Vermaelen (1995 and 2000) provide evidence that the market does not always appear to properly interpret the information contained in the announcement, particularly for value stocks. Ikenberry, Lakonishok and Vermaelen (1995 and 2000) find that by forming a portfolio of shares that is long in value firms (high book-to-market ratio firms) and short in glamour firms (low book-to-market ratio firms), abnormal returns could be earned over a period up to three years after the announcement date. They interpret this finding as being consistent with the market under-reacting to the news contained in repurchase announcements. In the context of this research, I will seek to determine if a portfolios formed by taking a long position in firms with a high REP% and a short position in firms with a low REP% are capable of earnings abnormal returns over the repurchase period. The purpose of this test is to provide limited evidence on whether the market adequately reacts to the available information about past repurchase follow through. If a trading strategy based on past repurchase follow through percentage is capable of earning abnormal returns, this finding will be consistent with the market not adequately reacting to past announcement follow through. If no abnormal returns are found, this finding will be consistent with the market adequately reacting to past announcement follow through. I will control for firms being value versus glamour stocks on the basis of book-tomarket, consistent with Ikenberry, Lakonishok and Vermaelen (1995 and 2000), and test to determine if the value versus glamour finding still holds in the presence of a history variable. 4.4 Trading Volume Testing. Following Beaver (1968), I also test the information content of TSE repurchase disclosures comparing trading volume during the pre-announcement period to trading volume during the announcement. As Beaver (1968) points out, volume changes reflect changes in the expectations of individual investors. Investors perception of risk may change as a result of the information contained in the repurchase announcement. Portfolio rebalancing may be required. Therefore, even in the absence of a consensus change in price there could still be a 21
substantial increase in volume if the announcement contains information. My first volume hypothesis focuses on the volume reaction to the announcement and is irrespective of the effects of history and disclosure: H4 Trading volume will rise in the period surrounding a repurchase announcement. To test this prediction, I will calculate the variable TVC% as the percentage change in mean daily trading volume during the announcement period over the previous six months mean daily trading volume. I will compare TVC% to zero using a standard t-test. A significantly positive TVC% supports H4. I also investigate the effects on trading volume of first time announcements and repeat announcements. Jagannathan and Stephens (2003) find that infrequent repurchasers have a greater return reaction consistent with infrequent announcements conveying more information than repeat announcements. For first time announcers it would be much more difficult to predict the announcement. As a result, the announcement itself will convey more information if it is a first time announcement and will convey less information as successive announcements are made. Therefore, my second volume hypothesis focuses on the volume reaction to the frequency of repurchasing. Again, the frequency of repurchase would be known to investors at the time of the announcement irrespective of the actual number of shares repurchased and the disclosed reason for the repurchase that is the focus of this study. H5 The change in volume declines with the frequency of announcement. H5 indicates that the change in volume (TVC%) will be higher for first-time announcers (HIS2=0) than for non first-time announcers (HIS2=1). H5 indicates that the change in volume (TVC%) will be lowest for firms who are announcing their third or subsequent repurchase program (FA=1) than for non-frequent announcers (FA=0) since the market will anticipate the announcement and hence the announcement itself will have little information value. I will use a 22
t-test to test these two predictions of H5. Evidence that first time announcers volume change exceeds that of non-first time announcers and that the volume change is lowest for third time announcers will be considered as support for H5. For firms with a repurchase history, I investigate whether firms that have followed through on past repurchase announcements enjoy a greater volume reaction to the announcement than firms that did not follow through on their past repurchase announcements. If a firm has a good follow through history when repurchasing shares, the market should view the current announcement as having greater credibility. Therefore my third volume hypothesis considers the incremental information content of the TSX requirement to disclose past repurchases and is as follows: H6 For firms who have a history of repurchasing shares, there is a positive relationship between the change in volume and the firms follow through history on its past repurchases. To test this hypothesis, I will test the following OLS regression on the sub-sample of firms with a history of repurchasing shares (HIS2=1 and HIS3=1 as a validity check): (4) TVC% = b 0 + b 1 REP% + CON + e 1 where CON = a vector of control variables found to be relevant in prior research (see Appendix 5). A significantly positive b 1 coefficient supports H6. Finally, consistent with my testing of returns I will investigate if there are any systematic differences in trading volume changes by reason for repurchase (Appendix 4). My procedures for coding repurchase reasons are discussed in section 6. T-tests will be used to determine if differences in volume change by reason for repurchase are significant. 4.5 Bid-Ask Spread Testing. To complete my analysis of the capital markets effects of repurchase disclosures, I will examine the effect of the announcement on bid-ask spreads, a 23
common proxy for information asymmetry. Prior analysis of the effect of repurchase announcements on bid-ask spreads has had mixed results. Ahn, Cao and Choe (2001) find a temporary reduction in bid-ask spreads during the repurchase period. Cook (2004) also finds that repurchasing activities lead to a reduction in the bid-ask spread. In contrast, Brockman and Chung (2001) find evidence that bid-ask spreads widen during repurchase periods, consistent with adverse selection costs increasing as a result of the presence of an informed trader, management. These prior studies did not consider a firm s past follow through history (HIS%), which I use as a proxy for the credibility of the current announcement. If the market views the repurchase announcement as credible (non-credible), information asymmetry should be reduced (increased) since the market maker and other market participants will now be privy to the information that management has that the shares are undervalued. I propose that the market will view the announcement as credible if past repurchases follow through (REP%) is high. Therefore, my first bid-ask spread hypothesis addresses the role of past follow through history on bid-ask spreads and is as follows: H7 For firms with a history of repurchases, the change in bid-ask spread is negatively associated with past repurchase follow through history. To test this prediction, I use the following OLS regression model on the sub-sample of firms who have a history of repurchasing (HIS2=1) (5) SPR = b 0 + b 1 REP% + b 2 CON + e 1 where SPR = mean bid-ask spread during the announcement period over the previous six months mean bid-ask spread and CON = a vector of control variables found to be relevant in prior research (see Appendix 5). A significantly negative b 1 coefficient supports H7. For first-time announcers, there is no history to use to assess the credibility of the current announcement. However, first time announcers have the option to provide a specific 24