Share Repurchases in Europe:

Master Thesis Finance Share Repurchases in Europe: A View over Managerial Hubris, EU Regulations and Country Legal Origins. Author: João Diogo Correia de Matos (ANR: 202419) Supervisor: Fabio Castiglionesi August 2014

Abstract This paper makes use of open share market repurchase announcements made in Europe to investigate explanatory theories of buybacks, assess the impact of new EU regulations, and to judge the influence of country legal origins in the market reactions. Consistent with the managerial hubris hypothesis this paper presents evidence of negative market reactions to buybacks for firms with past strong stock performance. The introduction of new EU-wide buyback regulations had a likely neutral impact in market reactions for the majority of countries. The prediction that certain legal origin systems create different levels of information asymmetries between managers and shareholders and thus originating different market reactions to buybacks, does not find strong support in data. In most of the sample countries no long run market under reaction to buybacks is detected. 1

Introduction This paper investigates the effect of open market share repurchase announcements made in Europe on the value of the repurchasing firm s common stock value. I address the topic from 3 different perspectives. Firstly, I test the most common explanations of repurchases cited in the literature. Those explanations are the signalling hypothesis, that assumes that firm s managers use share repurchases to signal their views to the market about either the firm s future prospects or about the firm s current market valuation, and also, the personal tax hypothesis, according to which pay-out methods impact firm s value due to the shareholder s personal tax exposures. While testing for these justifications I also reintroduce the managerial hubris or overconfidence hypothesis to further explain the observed market reactions to repurchases. Secondly, I investigate the impact on market reactions of new share repurchase regulations introduced by the European Commission Directive 2003/6/EC 1 (Market abuse directive Mad ). The Mad replaced the previous Council Directive 89/592/EEC 2 of 13 November 1989 adding further rules regarding market manipulation, specifically, restricting investors and firms from performing transactions or orders to trade which secure. the price of one or several financial instruments at an abnormal or artificial level 3. Since open market repurchases can be used to engage into price manipulation firms can be liable for manipulation charges and therefore firms could change their behaviour in response to the new regulations. Thirdly, I look at market reactions to share repurchases across countries of different legal origin. The main proposition under analysis is that if certain legal settings lead to more concentrated shareholder ownership than others, then, different levels of information asymmetry should arise between shareholders and the firm s management (La Porta et al, 1998). Henceforth, firms in English legal origin countries, characterized by relatively higher ownership dispersion should be subject to higher information asymmetries and therefore to display a stronger reaction to buyback announcements if these are motivated by managerial signalling (Lasfer, 2000). The event study methodology is the main tool employed in this paper to address the different topics. By using this method I focus on the firms abnormal stock performance before, 1 http://www.esma.europa.eu/mt/system/files/dir_03_6.pdf 2 http://www.esma.europa.eu/system/files/dir_89_592.pdf 3 European Commission Directive 2003/6/EC Article 1 2

surrounding and after repurchasing announcement dates. A brief illustration of the key findings follows in the next paragraphs. When analysing the firm s stock performance before repurchase announcement dates, no statistical significant abnormal returns are found. This suggests that firms are not timing their announcements in response to either worse or better stock performance than expected. However, the evolution of firms absolute raw stock returns in the previous six months to the announcement shows up to be a strong predictor of future abnormal returns. Specifically, firms with worse past six month stock performance tend to exhibit the strongest stock returns at announcement date and thereafter, while the reverse is also true,.i.e. announcing firms with strong and positive stock performance prior announcement exhibit near zero abnormal returns at announcement date and large negative abnormal returns after the announcement date. This can be interpreted in interesting ways, specifically that the market differentiates firms whose bad past performance and possible undervaluation are the main motivation to repurchase stock, thus the positive market reaction. On another hand, a buyback from a firm with strong past stock performance can be seen as an attempt by the management to keep stock prices artificially high (possibly due to overconfidence and/or management incentive structures), hence the negative market reaction. In the middle of the scale, firms with no extreme prior stock performance display abnormal returns close to and no statistically different form zero, hinting that the market treats such announcements as mere accounting transfers with no impact in valuations. The suggestion that firms with higher book-to-market ratios are more prone to be undervalued and that therefore, to display higher abnormal returns post announcement does not find support in the sample used. Due to the diversity of scenarios across countries it is hard to establish a causation link between the introduction of the Mad (Market abuse directive) and the evolution of market reactions to buybacks. A comparison of periods before and after year 2001 (Mad introduction) reveals increases in abnormal returns surrounding announcements in the UK and decreases in Norway and Finland, while in the reaming countries no change is detected. If the focus changes to long run market reactions then, only in Italy, Sweden and Austria, abnormal returns are statistically different form zero prior to year 2001. Thereafter only firms in Finland report long run (negative) abnormal performance. Firms in Sweden (Greece) report significant positive 3

(negative) abnormal returns depending on the investment horizon. With this landscape it can be reasonably concluded that the new regulations had a neutral impact in the majority of the countries under analysis. In what concerns market reactions to buybacks and country legal origin, prior to year 2001 abnormal returns surrounding event dates show no statistical difference between firms from different legal origin countries. After year 2001 abnormal returns of firms form English legal origin counties display a slight increase, firms from Scandinavian legal origin experience a sharp decrease in abnormal returns and firms in French and Germanic legal origin countries maintain the same profile of short term market reactions to buyback announcements. Therefore evidence of the hypothesis that buyback announcements carry a higher informational value in market based economies, where dispersed ownership is common and information asymmetries are expected to be high, as opposed to bank based economies where concentrated ownership is more expressive and information asymmetries are expected to be comparatively lower, can only be found in the period from year 2001 onwards. Long run abnormal returns are statistically undistinguishable across countries of different legal origins over the entire period under analysis. This research successfully replicates the findings of Dann (1981), Vermaelen (1981), Ikenberry, Lakonishok, and Vermaelen (1995) in regard to evidence of positive abnormal stock returns surrounding repurchase announcements. However it further adds that the market reaction is not homogenous on average and that the sign and magnitude of the abnormal returns surrounding announcements is highly dependent on firm prior stocks returns. Therefore the empirical finding that the market reacts positively to repurchase announcements can be just a matter of which type of firm (prior positive/negative past stock returns) dominates repurchase announcements in general. Ikenberry et al (1995) suggest that firms time their repurchase announcements based on the observance of negative and significant abnormal returns prior to repurchase announcements in the U.S. and Peyer and Vermaelen (2008) make use of prior firm returns together with bookto-market ratio to identify companies with higher future abnormal returns post announcement. Lasfer (2000) detects differences in abnormal returns between the UK (market-orientated economy with positive abnormal returns) and Continental European countries (bank-orientated economies with negative abnormal returns). 4

The rest of the thesis is organized as follows. Chapter I Literature review ; Chapter II - Research questions & Methodology; Chapter III - Data resources; Chapter IV - Results; Chapter V - Conclusion. Chapter I Literature review Share repurchases is one of the methods that allow companies to distribute cash to shareholders besides dividends. In comparison to the later, share repurchases can be regarded as being more flexible and tax favoured given that individual tax payers can time their share liquidations that trigger taxable income in a more efficient way. There has been vast research on the topic of share repurchases over the last 30 years. One major events in the topic was the passing of Rule 10b-18 by the SEC in the U.S. in 1982 which provided safe harbour provisions from market manipulation charges for U.S. companies listed on the stock exchange that choose to make share repurchases. Since this point in time the number of industrial companies performing share repurchases in the U.S. as increased very substantially as well the total amount of repurchased value (Skinner (2008), Grullon and Michaely (2004), Grullon and Ikenberry (2000)). Specifically, according to Skinner (2008), share repurchases have become the most important method of distributing cash to shareholders (in terms of total dollar value) surpassing the importance that of dividends payments in the U.S. In what concerns the impact of share repurchases on stock returns, there has been evidence reported of positive abnormal stock returns surrounding repurchase announcements (Dann (1981), Vermaelen (1981), Ikenberry, Lakonishok, and Vermaelen (1995), Guay and Harford (2000), Jagannathan, Stephens, and Weisbach (2000), Weston and Siu (2003), Maxwell and Stephens (2003), and Grullon and Michaely (2004)). Following this line of research, Lakonishok and Vermaelen (1990), Ikenberry, Lakonishok, and Vermaelen (1995) and Peyer and Vermaelen (2008) also present evidence of positive abnormal stock returns up to four years after the repurchase announcement, being this effect stronger for value stock companies and firms with worse past stock performance, consistent with the hypothesis of market under reaction. Also, Boudoukh, Michaely, Richardson, and Roberts (2007) show that including share repurchases in the firm s pay-out yields improves the performance of asset pricing models. Grullon and Michaely (2004) show that firm operational performance does not tend to improve following open market share repurchase announcements. 5

Focusing on Europe Eije and Megginson (2008) find that EU repurchase activity had a later start in comparison to the U.S. albeit it grew at a faster pace than in the US. They also describe that EU and American firm s dividend and repurchase policies are similar i.e. the fraction of firms paying dividends has decreased while the likelihood of firms making repurchases has steadily increased. Looking at the impact of share repurchases on stock returns in Europe Lasfer (2000) finds evidence of positive abnormal returns at share repurchase announcements and up to six months after the announcement over the period 1985 to 1998 in the UK, in contrast with insignificant or negative abnormal returns in other European countries in accordance with the view that institutional settings have an impact in the market valuation of share repurchases. Siems and Cesari (2012) find that the European Commission Directive 2003/6/EC (Market abuse directive Mad ) had a positive impact in the propensity for firms to repurchase stock. Chapter II - Research questions & Methodology A. The determinants of market reactions to buybacks and managerial hubris. As a first step of this research I discuss and test some of the most referred explanations for share repurchases and at the same time explore new empirical support for less cited ones. There are different explanations in the literature for share repurchases e.g. the personal tax hypothesis or the creditor expropriation hypothesis. In this paper I m going to address mainly the signalling hypothesis and the hubris hypothesis. Under the signalling hypothesis managers make use of repurchases to signal the market about the firm s better than expected future performance or to signal their disagreement about the firm s current market valuation. The prior version is rejected in Grullon and Michaely (2004) and the later in finds support in the studies of Lakonishok and Vermaelen (1990) and Ikenberry et al (1995). Additionally, Dann (1981) identifies firm size as a proxy for information asymmetries to justify higher returns at announcement date by smaller firms and, Vermaelen (1981) identifies the fraction of shares sought as a measure of the signalling strength. An alternative interpretation is signalling lack of internal investment opportunities (Ellis, 1965), however few incentives exist for managers to repurchase in these circumstances given the 6

firm value decreasing nature of the announcement and ready available investments in the capital market as alternatives (Dann, 1981). Further motives to repurchase are forward outside the scope of undervaluation. Ikenbetvy et al (1995) hypothesize the desire of managers to support artificially high stock prices or to avoid ownership dilution due to past option exercising as further reasons to repurchase. Also, managerial hubris can play a role for low book-to-market firms with relatively higher past stock performance as shares that are not under-priced get repurchased. The managerial hubris explanation is not yet tested empirically and no analysis is made to try to identify the firms that repurchase shares in those conditions and how the market reacts consequently. I take the additional step of clearly identifying those firms and describing the market reaction to those firms. B. The joint impact of the Market Abuse Directive 2003/6/EC (Mad) and Commission Regulation (EC) No 2273/2003 (The Regulation). As referred in the introduction, the Mad replaced the previous Council Directive 89/592/EEC of 13 November 1989 adding further rules regarding market manipulation, specifically, restricting investors and firms from performing transactions or orders to trade which secure. the price of one or several financial instruments at an abnormal or artificial level 3. Subsequent to the passing of the Mad, the Commission Regulation (EC) No 2273/2003 4 (The Regulation) introduced safe harbour exceptions for buyback programs if these comply with a set of conditions. Under the conditions, besides procedural requirements and various limits such as daily volumes, are the objectives of the buyback programmes which must fall into the following: to reduce the capital of an issuer; to meet obligations arising from debt financial instruments exchangeable into equity instruments; and to meet obligations arising from employee share option programmes or other allocations of shares to employees. The adoption of the Mad by the European countries entailed at the same time the adoption of The Regulation. For this reason I address from now the new regulations as Mad for simplicity of speech. 4 https://www.esma.europa.eu/system/files/madimplreg_2273_2003.pdf 7

This research will then, address the impact of the new regulations, which have both restrictive and safe harbour elements, in the market reactions to share repurchases. C. Country legal origins and market reactions Firms in English legal origin countries can be characterized to have relatively more dispersed ownership and to have higher information asymmetries between the firm s shareholders and the firm s management (La Porta et al, 1998). Therefore firms in English legal origin countries should display a stronger reaction to buyback announcements than comparable firms with more concentrated ownership since repurchases can be considered a form of managerial signalling (Lasfer, 2000). Another possible interpretation is that certain legal systems tend to offer better protection to shareholders than others, prompting firm managers to execute actions that are more aligned in general with the shareholders interests e.g. not to repurchase on the basis of ownership dilution. However, it is also reasonable to assume that those factors are already priced in the securities and that the relevant dimension to be tested for the informational value of repurchase announcements is that certain shareholders (high ownership) with better information access to the firm tend to bring the firm s stock price closer to its fundamental value. The sample under analysis includes countries from different legal origin systems which enables the testing of the hypothesis that legal origin systems also have an impact in the magnitude of the market reaction. Methodology To gauge the market reaction to repurchase announcements I use a standard event study methodology where abnormal returns are calculated by subtracting normal returns from actual returns, and tested for significance thereafter. I make a distinction between two set of abnormal returns based on two different benchmark models to estimate normal returns. The short run abnormal returns, calculated using a simple market model as benchmark, and the long run abnormal returns, calculated using a Fama French (1993) three factor model. Kothari and Warner (1997) show that the misspecification bias regarding the market model should not pose significant problems when the cumulating period of abnormal returns under analysis is less than 12 months. I also estimate the short run normal returns using CAPM as benchmark and verify 8

that the results are practically undistinguishable from the results rendered using a simple market model. Due to this outcome I present the findings using the market model for its simplicity of use. Nevertheless the conclusions for cumulating periods higher than 3 months will be based on the Fama French (1993) benchmark model. The estimation period for the market model benchmark is set to [-190;-40] days relative to the event date at t=0 and to [-27;-2] months relative to the event date at t=0 for the Fama French benchmark model. To guarantee the robustness of the long run event study I also make use of the Fama-French (1993) calendar-time buy-and-hold portfolio methodology as advocated by Fama (1998) and Mitchell and Stafford (2000). A more detailed analysis of both methodologies can be found in the appendix. Chapter III - Data resources The sample of repurchase announcements is extracted from the Securities Data Corporation s (SDC) Mergers and Acquisitions database for the period starting in 1990 and ending in 2013 for 16 European economies: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland-Rep, Italy, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, and United Kingdom. The selection of countries follows a very practical motive, specifically it is identical to the selection employed in the construction of Fama French (1993) portfolios (necessary for the long run analysis) made available on-line by Kenneth R. French at his personal web page from Tuck School of Business at Dartmouth College. The inclusion of Norway in the sample is sensible since it has undertaken to implement all European Directives regulating financial markets following its entry to the European Economic Area in 1992 (EEA Agreement). In the case of Switzerland, its observations are excluded when testing the effects of EU legislation. The sample used only considers announcements to repurchase ordinary shares in open market transactions and excludes new announcements that are less than two months apart from the same firm. The sample includes firms from financial and utility industries which are traditionally regulated. The reason to include regulated sectors is because they represent 20% of the observations gathered. The exclusion of these observations does not affect the results in any significant manner. The control variables percentage of shares sought and market cap are also extracted from the SDC database while firm returns, firm book-to-market ratio and market returns are obtained from Datastream and Worldscope from Thomson Reuters. 9

There are two significant comments about the gathered sample. First, most of the observations take place form year 1998 onwards. This has to do with changes in regulations in several countries that prompted firms to repurchase more often. Second, there are 129 observations in France for the year 1999 (40% of total observations from France) which are also related to changes in buyback regulations in 1998. To ensure that the conclusions regarding French firms are not biased I calculate all results with and without the French observations from year 1999. The conclusions presented in this paper are robust after this additional control. Table 1 Summary of count of sample repurchase announcements per year and per country. AU, BL, DK, FN, FR, DE, GE, IR, IT, NL, NW, PT, SP, SW, CH and UK stand for Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland-Rep, Italy, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland and United Kingdom respectively. Announcements by Nation year AU BL CH DE DK FN FR GE IR IT NL NW PT SP SW UK Total 1990 1 1 1 1 3 7 1991 1 1 2 4 1 3 12 1992 1 1 4 1 3 10 1993 1 1 1 2 5 10 1994 3 1 1 1 13 19 1995 2 1 1 1 1 1 1 14 22 1996 1 2 2 2 16 23 1997 5 1 1 2 1 2 1 26 39 1998 2 7 5 1 4 20 1 1 11 5 2 4 1 42 106 1999 1 4 9 27 6 12 129 2 12 2 5 1 11 221 2000 8 1 4 32 4 7 26 4 6 2 2 1 6 12 4 119 2001 5 1 4 16 3 11 9 4 1 1 3 1 59 2002 3 7 14 1 6 1 2 3 37 2003 2 2 5 6 1 1 19 3 5 3 2 2 6 57 2004 1 2 6 6 2 2 13 1 4 1 1 8 47 2005 2 4 7 10 1 1 5 1 1 3 2 1 6 44 2006 1 4 4 8 1 1 21 1 5 8 2 2 2 7 67 2007 5 5 13 27 3 1 17 2 1 6 15 2 2 1 5 10 115 2008 11 1 15 16 6 2 16 2 2 16 7 3 1 4 2 9 113 2009 2 10 2 5 1 3 1 6 6 6 3 45 2010 5 9 11 16 5 4 3 1 2 2 15 2 6 5 86 2011 5 3 11 19 26 38 14 3 1 1 3 44 2 45 9 224 2012 5 9 10 8 9 5 1 1 3 1 15 14 4 85 2013 1 5 14 3 29 1 2 1 2 2 1 1 9 71 55 33 126 233 87 89 340 37 15 95 66 101 4 38 99 220 1638 10

Table 2 Descriptive statistics for sample firms and repurchase announcements. Firm book-to-market ratio is measured at announcement date. Market-cap stands for firm market capitalization in millions at announcement date. % sought represents the intended fraction of shares to be repurchased expressed in percentage. p25, p50 and p75 stand for 25 th, 50 th and 75 th percentiles respectably. N refers to observations. Variable mean p25 p50 p75 N book-to-market 0.79 0.38 0.61 1.02 1597 market cap 8366 112 513 2789 1302 % sought 9.8 6.3 10.0 10.0 1311 Chapter IV - Results A. Full sample analysis In Table 3 I perform the most straightforward assessment of the market reactions to buyback announcements. That consists of measuring the cumulative average abnormal returns, in this case calculated using a market model as benchmark for normal returns, around the announcement date set at t=0 and thereafter. Table 3 Cumulative Average Abnormal Returns (CAAR) estimated with using a Market model for event windows [-2, +2],[+3, +60] days with event date at t=0. T-test stands for two-tailed test statistic robust for heteroskedasticity. The significance levels are indicated by,, and, and correspond to a significance level of 10%, 5% and 1% respectively. N refers to observations. event window CAAR t-test N [-2, +2] 1.2% 5.43*** 1638 [+3, +60] 0.1% 0.18 1633 The full sample Cumulative Average Abnormal Return (CAAR) at announcement [-2, +2] is positive (1.2%) and highly significant (t- test=5.43) while the post announcement drift up to +60 days after event [+3, +60] is statistically undistinguishable from zero. Hence, on average the market tends to react positively to buyback announcements and to display no significant post announcement drift. 11

Next, I test the hypothesis that firms time their repurchase announcements in response to past negative abnormal stock performance (Ikenberry et al, 1995). To do so I measure the firm s abnormal performance in the event window prior to the announcements. Table 4 Cumulative Average Abnormal Returns (CAAR) estimated with using a Market model for event windows [-40, -3],[-20, -3] days with event date at t=0. T-test stands for two-tailed test statistic robust for heteroskedasticity. The significance levels are indicated by,, and, and correspond to a significance level of 10%, 5% and 1% respectively. N refers to observations. event window [-40, -3] [-20, -3] CAAR 0,2% 0,0% t-test 0,8 0,0 N 1638 1638 The CAARs for the event windows [-40, -3] and [-20, -3] are 0.2% and 0.0% respectively and are both not statistically significant from zero. Thus it appears that the firms in the sample are not timing their repurchase announcements in response to out of ordinary bad stock performance. As a caveat, the fact that the CAARs prior the event are not different form zero is a suggestion of the good quality of the estimation period and of the normal returns. To conclude, I also perform a double control and check the distribution of positive and negative Cumulative Abnormal Returns (CARs) and of raw stock returns in the prior six months to the announcement. As table 5 demonstrates, the distribution of abnormal and raw prior returns is practically even. Table 5 Distribution of observations between negative and positive for the variables past six month, CAR [-40, -3] and CAR[-20, -3]. Past six month is the firm s past six month stock return measured in the event window [-190, -3]. CAR stands for firm Cumulative Abnormal Return estimated using a Market model for the event windows [-40, -3] and [-20, -3] with event date at t=0. positive/negative past six month CAR [-40, -3] CAR [-20, -3] < 0 48.9 50.3 48.9 > 0 51.1 49.7 51.1 Total 100% 100% 100% 12

B. Managerial hubris This section addresses the market reaction to buybacks from of firms that do not repurchase on the basis of firm undervaluation or signalling of future performance. Under the managerial signalling hypothesis, firm managers use buybacks to signal their disagreement about the firm s current market valuation or to signal their views about future firm performance. In turn the market reacts positively to this signal. However there can be alternative circumstances that induce firm managers to repurchase stock. Namely, the desire of keeping stock prices artificially high or avoiding ownership dilution (Ikenbetvy et al, 1995), and even plain managerial hubris in accessing the firm s value when repurchasing. In this paper I make use of firm s past stock performance to identify which firms are repurchasing stock on the grounds of undervaluation and which firms are repurchasing due to other motives. This division is intuitive and the expectation is that firms with worse past stock performance are the firms repurchasing due to perceived undervaluation by the management and that firms with strong past stock performance are repurchasing due to other reasons. It is of course possible that the management of a firm with strong past stock performance considers the firm to be undervalued, however in such occasions, so is higher the likelihood of overconfidence playing a role. After separating firms according to past six month stock performance I analyse the market reactions in terms of abnormal returns for firm category. Figure 1 presents the first set of results. 13

CAR Figure 1 Cumulative Abnormal Returns (CARs) for entire sample observations estimated with using a market model for the event window [-40, +60] with event date at t=0. On the xx axis the event periods are displayed. On the yy axis the size of the CARs is measured in percentage. The line Under -20% measures CARs form firms that have past six month stock performance prior to buyback announcement lower than -20%. Over +20% measures CARs form firms that have past six month stock performance prior to buyback announcement higher than +20%. Middle measures CARs form firms that have past six month stock performance prior to buyback announcement higher than -20% and lower than +20%. The groups Under -20% and Over +20% have approximately 25% of total sample observations each, while the group Middle has approximately 50% of total sample observations (see Table 6). 20% 15% Under -20% 10% 5% Middle 0% -40-36 -32-28 -24-20 -16-12 -8-4 0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60-5% -10% -15% event window Over +20% Middle Over +20% Under -20% 14

From Figure 1 it is clear that the market reaction to buybacks is not homogenous across firms with different levels of past stock performance. The group Under -20%, which comprises CARs from firms with past six month stock performance under -20%, has the highest abnormal returns at announcement date and also a positive return drift after the announcement. The group Over +20%, which comprises CARs from firms with past six month stock performance higher than +20%, has negative abnormal returns at announcement date and also a negative return drift the after announcement. The group Middle which comprises CARs from firms with past six month stock performance higher than -20% and lower than +20%, displays smaller positive abnormal returns at announcement date and no post announcement drift in returns (see Table 6). Grullon and Michaely (2004) show that firm operational performance does not tend to improve following open market share repurchase announcements, hence the market reaction to the announcements from firms in the Under -20% group fits better the hypothesis that managers are signalling that the firm is undervalued. This managerial signal translates into large and positive abnormal returns at announcement and into a post announcement drift in returns, hinting that the market also underreacts to the new information. In the Over +20% group, abnormal returns at announcement are slightly negative (albeit statistically significant). A possible interpretation for this market reaction is signalling lack of internal investment opportunities (Ellis, 1965), however few incentives exist for managers to repurchase in these circumstances given the firm value decreasing nature of the announcement and ready available investments in the capital market as alternatives (Dann, 1981). Therefore in association with the past strong stock performance in the Over +20% group, the overconfidence or hubris hypothesis is the explanation that better fits the observed market reactions i.e. the market may interpret the buyback as an attempt from the management to keep stock prices at high levels fowling past strong stock performance. Implicit in this interpretation is that the market recognizes the managements motives and as a result adjusts the security s price. Intelligibly, if the management believes that a buyback is necessary to keep the stock price high then it is because the firm s future results are likely to disappoint. Conversely, the market might also consider a buyback to be simply a bad use for cash, especially when the 15

underlying stock to be bought just increased over 20% in the last six months. Similarly for the Over +20% group, the market appears to be underreacting to this kind of new information. In regards to the Middle group, the positive abnormal returns at announcement are considerably lower (under 1%) and the there is no sign of post announcement drift in announcements. A reasonable explanation for this outcome is that the market sees these buybacks as mere accounting transfers with no impact in the firm s value per share. Table 6 Cumulative Average Abnormal Returns (CAAR) estimated with using a Market model for event windows [-2, +2],[+3, +60] days with event date at t=0. T-test stands for two-tailed test statistic robust for heteroskedasticity. Differences stands for difference of means with test statistics reported on the right side. Under -20% (Over +20%) is the group of observations with past six month returns below (over) -20% (+20%) and Middle is the group of observations with past six month returns between -20% and +20%. The significance levels are indicated by,, and, and correspond to a significance level of 10%, 5% and 1% respectively. Group Under -20%, Over +20% and Middle have for the event window [-2, +2] {[+3, +60]} 401, 467 and 770 {400, 464 and 769} observations respectively. CAAR t-test event window [-2, +2] [+3, +60] [-2, +2] [+3, +60] Under -20% 4.2% 7.9% 6.05*** 6.85*** Middle 0.9% -0.2% 4.36*** -0.36 Over +20% -0.9% -6.2% -3.08*** -7.77*** differences Under vs Middle 3.3% 8.0% -5.79*** -7.67*** Under vs Over 5.2% 14.1% -7.12*** -10.28*** Over vs Middle -1.8% -6.1% 5.17*** 7.04*** As an additional robustness check I test whether the effect of past six month stock returns in abnormal returns is still significant after controlling for variables known to impact the size of the abnormal returns. Summarizing past literature, firms tend to have higher (lower) abnormal returns at announcement if they are small (large) and if the size of the fraction of shares sought is higher (smaller). Moreover firms should show homogeneous reactions across different levels of book-to-market ratio at announcement. 16

Table 7 OLS regressions where the dependent variable is Cumulative Abnormal Return for [-2, +2] and [+3,+60] event windows, estimated using a Market model. The independent variables: past six month, is the firm past six month stock return measured in the event window [-190, -3] with event date at t=0; % sought represents the intended fraction of shares to be repurchased expressed in percentage; ln market cap, is the natural logarithm of firm market capitalization at announcement date; book-tomarket, is the firm book-to-market ratio measured at announcement date; country fixed effects assigns a dummy variable to each country and uses Austria a omitted dummy; industry fixed effects assigns a dummy variable to two digit sic codes (SDC s security industry code). Test statistics are robust for heteroskedasticity and are displayed in parenthesis under the coefficients. The significance levels are indicated by,, and, and correspond to a significance level of 10%, 5% and 1% respectively. N refers to observations. CAR / event window [-2, +2] [+3, +60] [-2, +2] [+3, +60] past six month -0.05-0.18-0.04-0.19 (-5.3) *** (-10.48) *** (-4.76) *** (-10.52) *** % sought 0.0004-0.0001 0.0004-0.0003 (1.79) * (-0.24) (1.76) * (-0.54) ln market cap -0.003 0.002 0.00 0.00 (-2.04) ** (0.89) (-1.59) (0.28) book-to-market 0.00-0.03 0.00-0.03 (-0.38) (-1.94) * (0.03) (-2.17) ** cons 0.03 0.02 0.01 0.00 (2.07) ** (0.7) (0.89) (0.01) country fixed effects NO NO YES YES industry fixed effects NO NO YES YES R-squared 0.06 0.18 0.13 0.24 N 1230 1226 1230 1226 The results from Table 7 indicate that the effect of past six month stock returns is significant at the 0.01% level across every model specification even after controlling for country and industry fixed effects. Moreover, given the size of the coefficients of the controlling variables, past six month has the biggest weight in determining the magnitude of the abnormal returns at and post buyback announcement. For a final remark, all controlling variables return the expected signs except book-tomarket ratio in the event window [+3, +60] with a negative and significant coefficient. For that 17

reason the hypothesis that high book-to-market firms tend to be relatively more undervalued and to experience higher abnormal returns post announcement is not confirmed in the sample in use. C. The joint impact of the Market Abuse Directive 2003/6/EC (Mad) and Commission Regulation (EC) No 2273/2003 (The Regulation) I. Short term market reactions To investigate the possible impacts of the Mad and Regulation on the market s reaction to intentions to repurchase I measure the differences in abnormal returns at announcement and post- announcement date, between buybacks performed before and after the new laws. The latest date of transposition of EU legislation to national law was June 2006 in Portugal and the earliest date was October 2004 in Germany. However to account for the fact that European law can be endogenous to the market I set the pre/post Mad reference date to May 2001 which is the date when the first proposal for a directive was submitted by the European Commission. Due to the fact that some countries have too few observations prior to year 2001 I decide to aggregate countries by legal origin (La Porta et al, 1998) and test the difference in market reactions pre/post Mad at the legal origin level. Table 8 Cumulative Average Abnormal Returns (CAAR) estimated with using a Market model, for event windows [-2, +2],[+3, +60] days with event date at t=0. English legal origin group comprises buyback events form IR and UK. French legal origin group comprises buyback events form BL, FR, GE, IT, NL,PT and SP. Germanic legal origin group comprises buyback events form AU and DE. Scandinavian legal origin group comprises buyback events form DK, FN, NW, and SW. Diff stands for difference of means in CAARs between Pre Mad and Post Mad periods. Test statistics are displayed in parenthesis under the CAARs and difference of means. The significance levels are indicated by,, and, and correspond to a significance level of 10%, 5% and 1% respectively. [-2, +2] [+3, +60] Legal Origin Pre-Mad prior year 2001 Post-Mad post year 2001 Diff of means Pre-Mad prior year 2001 Post-Mad post year 2001 Diff of means English 1.3% 3.5% 2.2% 2.3% 3.8% 1.5% (2.33) ** (4.21) *** (-2.24) ** (1.39) (1.79) * (-0.56) French 0.7% 1.7% 1.0% 1.6% 1.5% -0.1% (1.86) * (2.14) ** (-1.14) (1.43) (1.46) (0.07) Germanic 3.8% 2.3% -1.5% -3.7% -0.2% 3.5% (3.74) *** (4.58) *** (1.47) (-1.55) (0.84) (-1.49) Scandinavian 2.4% -1.2% -3.6% 3.8% -4.7% -8.5% (2.51) ** (-3.23) *** (4.01) *** (1.5) (-5.16) *** (3.72) *** 18

Table 8 shows an increase in CARs at announcement [-2, +2] for English legal origin countries (albeit not robust after controlling for industry fixed effects, (see Table 12)) and a statistical significant decrease for Scandinavian legal origin firms. All countries in this later group suffer decreases in CARs and the decreases are individually statistically significant for Norway and Finland and significant for Sweden and Denmark when taken together. Norway and Sweden drive the results of the Scandinavian legal origin group in decreases in CARs for the period [+3, +60]. The remaining countries of different legal origins do not suffer significant changes in CARs throughout different event windows. To interpret these evolutions under the light of the introduction of the Mad is difficult. For both UK and Ireland, Siems and Cesari (2012) consider the new regulations to be more restrictive and yet CARs [-2,+2] increase. The hypothesis that investors could fear firms to have a higher likelihood of receiving market manipulation charges is not reflected in the market reaction. When looking at Norway and Finland, due to the stark differences between individual tax rates for dividends and capital gains, the personal tax hypothesis to mind (both Norway and Finland had a 0% income tax on dividends in opposition to a 28% tax on capital gains). Since a change in the firm s pay-out method can have tax implications for the individual shareholder (Brigham, 1964), then as Dann (1981) points out, if a firm signals a modification in its investment decisions as well as in the method of cash distributions, the stock price should change in order to incorporate these combined modifications. Should the case be that the positive signal of an EU-wide safe harbour introduced by the Mad (Siems and Cesari, 2012) prompted repurchasing Norwegian and Finn firms to tilt their future pay-out decisions towards buybacks, then it is plausible to observe adverse stock price movements in consequence of repurchase announcements. The key assumption with this interpretation is that firms signal their pay-out policies changes at the same time as their repurchase announcements. Taking an overview of all countries it cannot be concluded that the introduction of the Mad produced necessarily significant changes in short term market reactions to buybacks. This assessment can carry a special importance to market regulators given that it is also applicable to economies that suffered a tightening in buyback regulations. 19

II. Long run market reactions In regards to the long run stock performance of repurchasing firms I firstly calculate Abnormal Returns (ARs) with normal returns estimated with a Fama French, (1993) three factor model. Again I aggregate countries by legal origin (La Porta et al, 1998) and test the difference in market reactions pre/post Mad at the legal origin level. Table 9 Cumulative Average Abnormal Returns (CAAR) estimated with using a Fama French three factor model, for event windows [+1, +6],[+1, +12] months with event date at t=0. English legal origin group comprises buyback events form IR and UK. French legal origin group comprises buyback events form BL, FR, GE, IT, NL,PT and SP. Germanic legal origin group comprises buyback events form AU and DE. Scandinavian legal origin group comprises buyback events form DK, FN, NW, and SW. Diff stands for difference of means in CAARs between Pre Mad and Post Mad periods. Test statistics are displayed in parenthesis under the CAARs or difference of means. The significance levels are indicated by,, and, and correspond to a significance level of 10%, 5% and 1% respectively. [+1, +6] [+1, +12] Legal Origin Pre-Mad prior year 2001 Post-Mad post year 2001 Diff of means Pre-Mad prior year 2001 Post-Mad post year 2001 Diff of means English 1.6% -0.6% -2.2% 3.6% -3.6% -7.2% (0.72) (-0.20) (-0.58) (0.99) (-0.71) (-1.14) French 5.5% 2.5% -2.9% 24.9% -0.6% -25.5% (2.29) ** (1.26) (-0.94) (6.52) *** (-0.21) (5.39) *** Germanic -2.5% 0.2% 2.7% 2.5% 0.3% -2.1% (-0.42) (0.08) (0.10) (0.27) (0.09) (-1.25) Scandinavian 10.1% -9.1% -19.2% 30.0% -5.5% -35.5% (2.80) *** (-5.42) *** (4.76) *** (5.02) *** (-2.24) ** (5.93) *** Under this methodology, firms form English and Germanic legal origin economies do not report any significant long run abnormal performance throughout the entire period. Of the French legal origin countries, firms from France, Italy and The Netherlands register positive and significant CARs before the Mad. After the Mad, all firms in this group have CARs undistinguishable from zero. Firms form Denmark, Norway and Sweden have positive and significant CARs before the Mad. In the post Mad period, Danish firms continue to have positive 20

and significant CARs, firms from Finland and Norway show negative and significant CARs, and abnormal performance disappears in Sweden. However I also take a second approach to analyse long run abnormal returns and create Fama-French (1993) buy-and-hold calendar-time portfolios as advocated by Fama (1998) and Mitchell and Stafford (2000). The portfolios are equally weighted for two reasons (Peyer and Vermaelen 2008); first, the objective of the analysis is to investigate whether some persistent behaviour exists for this small group of stocks and not so much to judge the efficiency of the market as a whole; second, larger firms are less likely to be miss-valued and thus could hide abnormal performance in the portfolio from smaller less represented firms. When using this methodology (Table 10), before the Mad only Italian and Swedish firms report positive long run abnormal returns after buyback announcements while Austrian firms have negative abnormal returns. After the introduction of the Mad Finn firms experience negative and significant abnormal returns and depending on the investment horizon, Swedish (Greek) firms experience positive (negative) abnormal returns. Thus, the evidence of a long run market under reaction to repurchase announcements is at most very thin and the introduction of new buyback regulations does not appear to have significantly changed the landscape of abnormal returns. The same conclusions can be made even when only firms from the top quartile of book-to-market ratio are considered (Table 13). These results can be relevant for investors because they practically rule out positive long run abnormal performance for repurchasing firms in Europe, and at the same time, call into question manager s ability to identify undervaluation. To demonstrate this point Figure 2 charts the return between 1 invested in the market portfolio and 1 invested in a portfolio made of stocks from firms that announced buybacks in the previous 12 months from years 1995 till 2014. 21

Table 10 Fama-French (1993) calendar-time buy-and-hold portfolio Abnormal Returns (AR) for holding periods 6, 12 and 24 months. The portfolios are constructed on a country basis and events are split between announcements before and after the Mad. The ARs represent monthly abnormal returns for portfolios constructed with firms that announced repurchasing programs in the last 6, 12, or 24 months. m. refers to months. t stands for two-tailed test statistic robust for heteroskedasticity. AU, BL, DK, FN, FR, DE, GE, IR, IT, NL, NW, PT, SP, SW, CH and UK stand for Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland- Rep, Italy, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland and United Kingdom respectively. The significance levels are indicated by,, and, and correspond to a significance level of 10%, 5% and 1% respectively. Before Mad (before year 2001) Country 6 m. t 12 m. t 24 m. t AU -4.0% -3.12 *** -2.7% -2.81 *** -2.1% -2.4 ** BL -0.3% -0.38-0.5% -0.85 0.0% -0.09 DK 1.5% 1.17 0.5% 0.61 0.6% 1.07 FN 1.5% 1.12 1.1% 1.23 0.9% 1.35 FR -0.3% -0.4 0.1% 0.12 0.2% 0.63 DE -0.7% -0.82-0.6% -0.97-0.5% -0.71 GE -0.4% -0.17-0.1% -0.07 1.5% 0.83 IR -2.6% -1.27-1.6% -1.23-0.6% -0.47 IT 2.4% 2.04 ** 1.8% 2.18 ** 1.3% 1.92 * NL -0.7% -0.54-0.1% -0.1 0.9% 1.03 NW 2.7% 1.08 1.9% 1.55 1.6% 1.65 PT 4.4% 1.82 0.8% 0.33-0.8% -0.67 SP -1.5% -1.33-0.6% -0.68 0.0% -0.02 SW 1.5% 1.47 1.8% 2.29 ** 1.2% 1.69 * UK -0.5% -0.83 0.0% -0.01 0.3% 1.09 After Mad (after year 2001) Country 6 m. t 12 m. t 24 m. t AU 0.1% 0.09 0.2% 0.35 0.2% 0.66 BL -0.3% -0.28-0.4% -0.71-0.3% -0.63 DK -0.5% -0.71 0.3% 0.56 0.5% 1.34 FN -2.4% -2.53 ** -1.3% -2.09 ** -1.1% -2.22 ** FR 0.2% 0.28 0.0% -0.07-0.1% -0.27 DE 0.5% 0.75 0.2% 0.51 0.1% 0.28 GE -0.6% -0.33-1.7% -1.61-1.3% -1.73 * IR -0.5% -0.26 0.0% 0.03 1.1% 0.93 IT -0.2% -0.29-0.2% -0.57-0.1% -0.27 NL 0.5% 0.51-0.3% -0.51-0.1% -0.15 NW -1.2% -1.58-0.6% -1.02-0.3% -0.78 PT -1.3% -0.68-0.8% -0.69-0.3% -0.37 SP 0.1% 0.26 0.3% 0.56-0.3% -0.71 SW 0.4% 0.62 0.7% 1.44 0.7% 2.01 ** UK 1.1% 1.64 0.4% 0.86 0.2% 0.52 22

Figure 2 Portfolio evolution from year 1995 till March year 2014 if 1 is invested in the market portfolio or in the 12 month buyback portfolio constructed with firms from all sample countries except Switzerland. 6 5 4 3 2 1 0 5 8112 5 8112 5 8112 5 8112 5 8112 5 8112 5 8112 5 8112 5 8112 5 8112 5 8112 5 8112 5 8112 5 8112 5 8112 5 8112 5 8112 5 8112 5 8112 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 20132014 Market portfolio 12 month buy-back portfolio 23

D. Legal Origin and Share Repurchases Market reactions to share repurchases are dependent on factors such as firm size, fraction of shares sought, book-to-market ratio and firm past stock performance. The next step of this research is to investigate whether legal origin also impacts the magnitude of the market response to buyback announcements. The analysis is made for pre and post Mad periods to control for the legal homogenization of buyback programs in 15 of the 16 sample countries. Table 11 To test the difference in Cumulative Abnormal Returns (CAR) between firms from different legal origin countries I run an OLS regression using observations of two legal origins at a time (e.g. English and French). The dependent variable is firm CARs [-2, +2] or firm CARs [+3, +60] estimated using a Market model. The independent variables are: firm past six month return measured in the event window [-190, -3]; % sought or, the intended fraction of shares to be repurchased expressed in percentage; the natural logarithm of firm market capitalization at announcement date; firm book-to-market ratio measured at announcement date; a dummy variable to distinguish firms between legal origins (e.g. English vs French). The tabulated dummy in the table represents the outcome of the dummy variable in the regression. T-test stands for test statistic and it is robust for heteroskedasticity. The significance levels are indicated by,, and, and correspond to a significance level of 10%, 5% and 1% respectively. For interpretation purposes, the first legal origin name is the one to which the dummy variable equals 1 in the OLS regression ( e.g. English is the first legal origin name in English vs French hence the dummy variable equals 1 if the firm is from a English legal origin country). Pre-Mad (prior year 2001) Post-Mad (post year 2001) [-2, +2] [+3, +60] [-2, +2] [+3, +60] dummy t-test dummy t-test dummy t-test dummy t-test English vs French 0.012 1.59 0.010 0.46 0.030 1.60 0.025 0.67 English vs Germanic -0.002-0.16 0.058 2.27 ** 0.024 1.84 * 0.065 1.80 * English vs Scandinavian -0.006-0.40-0.042-1.24 0.053 4.33 *** 0.074 2.15 ** French vs Germanic -0.016-1.52 0.058 2.66 *** -0.003-0.23 0.031 1.94 ** French vs Scandinavian -0.016-1.22-0.066-2.60 *** 0.022 2.11 ** 0.047 2.75 *** Germanic vs Scandinavian -0.006-0.38-0.098-3.05 *** 0.031 5.19 *** 0.011 0.89 Before the introduction of the Mad, CARs around announcement [-2, +2] cannot be considered statistically different across firms from different legal origin countries. After the introduction of the Mad, CARs [-2, +2] from English legal origin firms increase and CARs [-2, +2] from Scandinavian legal origin firms decrease. The increase in English legal origin CARs [-2, +2] is sufficient to make then statistically different from Germanic legal origin CARs, but not from French legal origin CARs. The decrease in Scandinavian CARs [-2, +2] makes them statistically lower than CARs from all other legal 24