The information content of dividend change announcements and share repurchase announcements in isolation and compared

1 The information content of dividend change announcements and share repurchase

announcements in isolation and compared

Albert Mol University of Groningen

S1462172

Abstract

This paper examines the stock price reaction to dividend change announcements and share repurchase announcements for firms of different sizes. It compares the stock price reactions following a dividend increase announcement and a share repurchase announcement corrected for the size of the cash flow. Small firms react more positive on a share repurchase than on a dividend increase, medium-sized firm investors do not seem to have a preference and large firms react more positive on a dividend increase than on a share repurchase. Furthermore, large firms do not react positive on a dividend increase announcement but a dividend decrease announcement leads to a devastating abnormal return of -9%.

JEL classification: C20, G14, G35.

2 Firms can distribute excess cash to their shareholders by repurchasing shares or by paying dividends. Both the payment of dividends and share repurchases are common practice for US firms. Without taxes and transaction costs, it does not matter how a firm distributes its excess cash (Miller and Modigliani, 1961). However, in the US dividends are on average taxed less favorable than capital gains and this makes the payment of dividends controversial. Only institutional investors are taxed more on realized gains than dividends and tax-exempt investors are taxed on neither, but there is no evidence that these groups are big enough to outweigh the effect of taxable investors (Black, 1976).

3 investors. Indeed, Schleifer (2000) presents evidence that investors in small firms are predominantly individual investors and investors in large firms are especially institutional investors. Small investors may be more sensitive to dividends as a marketing tool and institutional investors are taxed less on dividends.

4 as a dividend decrease? Which of the two announcements, dividend change announcements and share repurchases, has more effect on the stock price and are these effects different for firms of different sizes?

If one of the two announcements has more effect on stock prices this indicates which one is preferred by investors. Furthermore, if this relationship differs for firm size this indicates that different investors have different preferences. Firms can use this information in their pay-out decisions. If the reaction to dividend changes is asymmetric, then this may explain why firms slowly adjust dividends and have an aversion to dividend cuts. An asymmetric reaction to dividend changes can also be a sign for managers to better explain why dividends are changed. For example, a dividend cut can be positive when managers successfully convince investors that they have profitable opportunities.

5 1. Literature review

6 Haw and Kim (1991) examined the stock price reactions to dividend change announcements controlling for firm size. They conclude that stock prices reactions are positively related to the magnitude of the dividend change, but that the stock price reaction to these dividend changes is inversely related to firm size. It can be argued that the informational asymmetry is greater for small firms than for large firms, since investors have better access to information about large firms, and that the information content of dividend changes is bigger for small firms. This results in stock price reactions to dividend changes that are inversely related to firm size, as found by Haw and Kim (1991). There is also a theoretical explanation that predicts the opposite as Haw and Kim (1991). In the US, institutional investors are taxed less than individual investors on dividends. Since small investors are predominantly in small firms and institutional investors predominantly in large firms (Schleifer,2000), this could result in greater abnormal returns after a dividend increase for large firms than for small firms. Empirical evidence does not support this view. Alongside Haw and Kim (1991), Bajaj and Vijh (1990) also found that reactions to dividend changes are larger for low priced and small firm stocks.

7 repurchases seems to be superior over dividends. Brown (2007) examined the effect of share repurchase announcements for the Australian market. She found significant positive abnormal returns of 1.2% compared to 8% abnormal return in American studies. Masulis (1980), Dann (1981) and Vermaelen (1981) also report significant abnormal returns after share repurchase announcements. Ikenberry, Lakonishok and Vermaelen (1995) reported a significant 12% stock price increase over the subsequent four years after a share repurchase announcement. Ikenberry et al (1995) explained that if the market values of stocks are based on their fundamentals, it cannot be ‘’profitable’’ for a firm to repurchase shares immediately after an announcement. When managers are asked why they repurchase shares, the most cited reasons are ‘’undervaluation’’ and ‘’a good investment’’ (Ikenberry et al, 1995). The term ‘’profitable’’ can then be interpreted as a wealth transfer between investors selling the stock and long-term shareholders. However, firms are not obligated to repurchase the stocks after an announcement. Indeed, some firms do not repurchase stocks after an announcement. This means that open market announcements of a share repurchase provides firms with an option to buy back their outstanding shares, but do not commit firms to do so when all mispricing is eliminated. If all mispricing is eliminated after a share repurchase, it is not profitable anymore for a firm to actually repurchase the stocks (Isagawa,2002).

8 otherwise be wasted by management (Jensen, 1986). Guffey and Schneider (2004) found that large firms are more likely to engage in share repurchase programs and that one of the main reason is the elimination of excess free cash flow, consistent with Jensen (1986). Guffey and Schneider (2004) also argue that large firms may repurchase shares to satisfy needs of employee stock options. Large firms are more likely to have employee stock options than small firms and are more likely to repurchase shares. With the share repurchase, the firms then locks in the cost of shares for future shares expected to be issued to satisfy employee stock option exercises (Rogers, 2006).

10 undervalued stocks use dividend increases to signal this. This could explain the results from Choi and Chen (1997) that a share repurchase announcement has a much greater effect on stock prices than a regular dividend increase. However, Choi and Chen (1997) do not answer the question whether this effect is the same for firms of different sizes. As argued earlier, institutions are taxed less on their dividends in the US and since these institutions invest predominantly in large firms, the effect of a share repurchase announcement vis-a-vis a dividend increase on stock prices could differ for firm size.

It can be concluded that both methods for redistributing money, share repurchases and dividends, have pros and cons. Furthermore, few studies have examined whether the announcement effects are the same between dividend increases and share repurchases corrected for cash flow. Choi and Chen (1997) indeed compare whether the announcement effects are the same, but does not make a difference between small, medium and large sized firms. In this paper we try to present a full picture of the information content of dividend changes and share repurchases.

11 leads to under reaction in the short run. This under reaction to news like dividend changes and share repurchases is also found by other authors as mentioned above.

12 2. Data

13 Table I: Descriptive statistics

15 3. Methodology

The information content of share repurchase announcements and dividend change announcements is tested before, on, and after the event date. Three samples are needed for upside dividend change announcements; one for large, medium and small firms. Three samples are needed for downside dividend change announcements; one for large, medium and small firms. Three samples are also needed for share repurchases; one for large, medium and small firms. For each sample, 50 events are used. This means that in total 9 samples are used to answer the research questions. Market capitalization is used as a proxy for firm size. The firms are then equally divided in the small, medium and large sized firms group based on their market capitalization. The firms are divided in three groups based on firm size because, for reason mentioned earlier in the literature review, relationships can differ for firms of different sizes. Chen and Choi (1997) do not differentiate for firm size and make one general statement about the difference in information conveyed between share repurchases and dividend increases for firms of all sizes. We think that this is not appropriate because firms of different sizes have different investors with different preferences. By dividing firms in three groups based on firm size we try to capture this.

The estimation period and event windows

16 estimation window. A 40 days event window is used consisting of 20 days before the event date and 20 days after the event date, consistent with Mackinlay (1997). In this event window, the following periods are of particular interest: T= [-5,-1] , T=0 and T= [1,5]. This is done because the objective of the paper is to check the information content on and after the announcement day but also if information is leaked before the announcement date. These periods are chosen based on the cumulative abnormal return graph of the event window. Although the events can have a small impact in the period before day -5 and after day 5, the most effect seems to be in the period 5 days before and after the event. The estimation period will be 120 trading days starting from 21 days before the event date. The estimation period starts 21 days before the event to exclude the possibility that the event has effect on the mean return.

Model used to estimate abnormal returns

The model used in this paper is the mean adjusted model. The reason for using this model is that it is the easiest model and yields almost the same results as more complex models (Brown and Warner, 1985). The mean adjusted model adjusts excess return with the following formula:

] [ i IT IT R E R AR _{= − (1)} IT

AR _{is the abnormal return for security I on day T.} R_{ITis the return of security I on day}

17 Analyzing abnormal returns

The average abnormal return on every day in the event window of all securities in the sample is computed with the following formula:

∑

Where the first term is the average abnormal return, N is the sample size, i is the individual security and the last term is the abnormal return for that period.

A Jarque Bera test is used to test whether the distribution of abnormal returns is normal or non-normal for each of the 9 samples under the null hypothesis that the abnormal returns are normally distributed.

) 4 ) 3 ( ( 6 2 2 − + = N S K JB ₍₃₎

18 robustness check with non-normal distributions and the Corrado test with normal distributions.

Parametric t-test

The H0 for the event date T=0 is tested by dividing the average mean adjusted return by the standard deviation of the mean adjusted returns. This is shown in the following formula. 2 / 1 ) var( 1 T T AR AR = ϕ (4)

Where the variance of the mean abnormal return is computed as follows:

∑

For the multi-day periods T= [-5,-1] and T= [1,5], the average abnormal returns of all securities were cumulated using the following formula:

19 where the first term is the cumulated abnormal return in the given event window and the term on the right side is the sum of the daily average abnormal returns computed earlier.

The variance of the average CAR is computed with the following formula

∑

H0 is tested by dividing average CAR by the standard deviation of the average CAR.

2 / 1 )) 2 , 1 ( var( ) 2 , 1 ( 2 T T CAR T T CAR = ϕ (8)

The left term is the test statistic, which is used to make inferences about the sample.

Non-parametric Corrado test

20 Each firm’s abnormal return is ranked over the combined estimation and event window (Serra, 2002) where: ) ( il il rank AR K _{= (9)} is il is il AR K K AR > ⇒ _> (10)

The test compares the ranks in the event period with the average rank under the null hypothesis of no abnormal returns (Ki =0.5+Ti/2).

∑

Where ϕ3the test statistic, N is the sample size, K is the rank of return, S is the standard deviation of the ranks and 2L _{is the highest rank of the sample.}

With multi-day periods the following formula is used:

L K S K R L I I 1 ) ( 1

∑

21 It is also tested if the relationship between dividend change announcements and stock returns and the relationship between share repurchase announcements and stock returns differs for firm size. This is done using a regression to correct for the magnitude of the dividend change and the share repurchase. For example, the average dividend change for large firms may be larger than for small firms. This would make it improper to compare the two T-statistic from the event methodology. The following regressions are used: SIZE DIV CAR 2 1 0 β β β + + = (13) SIZE REP CAR 2 1 0 β β β + + = (14)

This regression is performed 9 times; for dividend increases for the periods T= [-5,-1] , T=0 and T= [1,5], for dividend decreases for the periods T= [-5,-1], T=0 and T= [1,5] and for share repurchases for the periods T= [-5,-1], T=0 and T= [1,5]. CAR is the cumulative abnormal return for the periods T= [-5,-1] , T=0 and T= [1,5]. DIV is the dividend change in percentage. REP is the amount of shares repurchased in percentage. SIZE is the log of the firms market capitalization.

22 adjusted form of the formula used in Chen and Choi (1997), is used to test which of the two announcements has more effect on stock prices.

VOL SIZE DIV FES DUM CAR 5 4 3 2 1 0 β β β β β β + + + + + = (15)

This regression is performed nine times; for small, medium and large firms for the periods T= [-5,-1] , T=0 and T= [1,5]. CAR is the cumulative abnormal return for respectively T= [-5,-1] , T=0 and T= [1,5). Unlike Chen and Choi (1997) we divide the firms in three samples based on firm size because firms of different sizes have different investors with different preferences. Chen and Choi (1997) make a general statement that share repurchases convey more information than dividend increases for firms of all sizes. For reasons mentioned in the literature review, this does not have to be the case for firms of all sizes. SIZE is the market value of equity two days before the announcement. VOL is the stock return volatility, defined as the standard deviation of daily returns for the period -140 to -21 days. DUM is one for a share repurchase and zero otherwise. FES is the fraction equity sought and is the percentage of shares repurchased. DIV is the incremental dividends as a fraction of market value of equity and is computed as follows.

23 Where INCD is incremental quarterly dividends, SHARE is the number of shares outstanding and MV is the market value two days before the announcement. DIS is the discount rate defined as 0.049 + 0.055 β, following Devaney (2008). Devaney (2008) calculates the average 10-year US treasury return for the period 1870-2002 as an approximation for the risk free rate. He calculates the average S&P 500 return for the period 1870-2002 as an approximation for the market return. Although is it possible to write a thesis about the actual risk free rate and market return, we think that this is a reasonable approximation.

If a share repurchase announcement has more effect on the stock price than dividend change announcements, the coefficient on DUM in equation 15 should be significantly positive.

4. Results

24 4.1 Results and analysis share repurchases

Figure 1 shows the CAR for small, medium and large firms surrounding a share repurchase announcement. The figure shows a clear difference in the stock price reaction after a share repurchase for firms of different sizes.

Figure 1: Car graph of share repurchases (small; medium and large firms)

25 Table II: Test statistics share repurchases for the three event windows and robustness check

Appropriate test CAR Test Statistic Robustness check

Small firms T= (-5,-1) Corrado -1,20% -1,26 -1,17 T= 0 Corrado 0,53% 1,74* 0,50 T= (+1,+5) Medium Firms Corrado 3,26% 2,10** 2,21** T= (-5,-1) Corrado -1,63% 0,33 -1,32 T= 0 Student T Test 1,25% 2,20** 2,57*** T= (+1,+5) Large firms Corrado 1,14% 0,58 0,77 T= (-5,-1) Corrado -0,56% -0,36 -1,00 T= 0 Corrado 0,15% 1,20 0,48 T= (+1,+5) Student T test -0,94% -1,71* -0,23

***significant at a 1% level, **significant at a 5% level, *significant at a 10% level. The corrado test is used as an robustness check when the appropriate test is the Student T test and the Student T test is used as an robustness check when the appropriate test is the Corrado test.

26 on dividends. We test this more formally in section 4.4. Another explanation is that investors prefer firms to retain earnings and use them for future investments.

The previous analyses could be flawed, because a event study methodology does not correct for the magnitude of the share repurchase. The average percentage of shares repurchased is 10,64% for the small firm sample, 6,92% for the medium-sized firm sample and 5,27% for the large firm sample. Table 3 shows the results of the regression, which corrects for the magnitude of the shares repurchased.

Table III: Regression tests of the relationship between CAR and firm size and percentage shares repurchased.

T=[-5,-1] T=0 T= [+1,+5] C -0,04 (0,72) -0,03 (0,70) 0,35 (0,03) ** Logsize 0,00 (0,81) 0,00 (0,67) -0,03 (0,02) ** Sharerepurchase 0,04 (0,63) 0,05 (0,44) 0,12 (0,24) Adjusted R^2 F statistic 0,05 0,10 0,04 0,28 0,08 7,17 p-value 0,90 0,75 0,001 Ramsey Reset test 0,18 0,14 0,18

***significant at a 1% level, **significant at a 5% level, *significant at a 10% level For all dependent variables, the first number is the coefficient and the second the p- value. The Chi-Square probability of the ramsey reset test is shown in the last row.

27 Furthermore, stock prices of small firms could be more undervalued since there is more information asymmetry for small firms. The negative reaction on share repurchase announcements for large firms indicate that investors in large firms do not think that a share repurchase conveys positive information or they would like the money to be distributed through a dividend increase. The non-significance of the coefficient on the share repurchase variable indicates that it does not really matter how much of the shares are repurchases. Investors base their opinion on the announcement of the share repurchase and not on the actual amount of shares repurchased.

4.2 Results and analysis dividend increase

Figure 2 shows the CAR for small, medium and large firms surrounding a dividend increase announcement.

28 Both small and medium firms seem to react positive on a dividend increase announcement while large firms do not react at all. Table IV shows that the only significant stock price reaction is for small firms at day zero. At first glance, when looking at figure 2, medium sized firms also react positively on the announcement of a dividend increase. This is largely due to some large positive outliers in the medium-sized firm sample. The corrado test takes this into account and therefore is not significant different from zero.

Table IV: Test statistics dividend increase for the three event windows and robustness check

Appropriate test CAR Test Statistic Robustness check

Small firms T= (-5,-1) Corrado -1,45% -0,58 -1,23 T= 0 Corrado 0,62% 2,18** 1,10 T= (+1,+5) Medium Firms Corrado 2,19% 1,27 2,20** T= (-5,-1) Corrado 0,34% 0,82 -1,00 T= 0 Student T test -0,47% -0,72 -1,74* T= (+1,+5) Large firms Corrado 1,03% 0,52 2,00* T= (-5,-1) Corrado -0,72% -1,28 -0,71 T= 0 Corrado -0,09% -0,23 -0,21 T= (+1,+5) Corrado 0,67% 0,10 0,74

***significant at a 1% level, **significant at a 5% level, *significant at a 10% level. The corrado test is used as an robustness check when the appropriate test is the Student T test and the Student T test is used as an robustness check when the appropriate test is the Corrado test.

29 and are attracted to dividends. If dividends signal that firms have enough cash to finance their investment projects then this should also be the case for medium-sized and large firms. As mentioned earlier, the t-test does not correct for the magnitude of the dividend increase. This means that, it would be improper to compare firms of different sizes without correcting for the magnitude of the dividend increase. The average dividend increase is 38,59% for the small firm sample, 29,58% for the medium-sized firm sample and 18,88% for the large firm sample. This indicates that the percentage dividend increase is inversely related to firm size. Table VI shows the results of the regression, which corrects for the magnitude of the dividend increase.

Table V: Regression tests of the relationship between CAR and firm size and percentage dividend increase.

T=[-5,-1] T=0 T= [+1,+5] C -0,11 (0,15) 0,05 (0,24) 0,10 (0,17) Logsize 0,01 (0,12) -0,01 (0,20) -0,01 (0,28) Dividend increase -0,01 (0,48) 0,00 (0,49) -0,02 (0,28) Adjusted R^2 F statistic 0,02 2,25 0,01 0,96 0,01 1,32 p-value 0,11 0,38 0,26 Ramsey Reset test 0,43 0,43 0,91

***significant at a 1% level, **significant at a 5% level, *significant at a 10% level For all dependent variables, the first number is the coefficient and the second the p- value. The Chi-Square probability of the ramsey reset test is shown in the last row.

30 reactions for firms of different sizes. While the results of the t-tests indicate that only small firms react significant and positive on the announcement of a dividend increase there is no evidence that there is a significant difference in the stock price reaction of firms of different sizes. This does not confirm our hypothesis that a dividend increase announcement will result in a positive abnormal return that is most present for small and large firms.

4.3 Results and analysis dividend decrease

Figure 3 shows the CAR for small, medium and large firms surrounding a dividend decrease announcement.

Figure 3: Car graph of dividend decreases (small; medium and large firms)

31 is present differs for firms of different sizes. A dividend decrease seems to have the most impact on large firms. The cumulative abnormal return for large forms is -9% on day 2 after the announcement. Almost the entire part of this abnormal return is cumulated before the announcement for large firms. Table VI below indicates that a dividend decrease results in a significant negative abnormal return for small firms in the period T=[+1,+5], for medium-sized firms in the period T=[-5,-1] and for large firms in the period T=[-5,-1] and on day zero. The effect for medium sized firms is only significant at a 10% level and therefore not strong. The stock price effect of a dividend decrease for large firms is significant at a 1% level for the period T=[-5,-1] and at a 5% level for day zero.

Table VI: Test statistics dividend decrease for the three event windows and robustness check

Appropriate test CAR Test Statistic Robustness check

Small firms T= (-5,-1) Student T Test 0,16% 0,11 -0,49 T= 0 Corrado -1,55% -1,28 -1,66 T= (+1,+5) Medium Firms Corrado -0,89% -2,16** -0,44 T= (-5,-1) Student T Test 2,46% 1,75* 1,38 T= 0 Corrado -1,42% -0,28 -1,39 T= (+1,+5) Large firms Corrado -0,11% 0,12 -0,05 T= (-5,-1) Corrado -5,30% -2,68*** -2,31** T= 0 Student T Test -3,07% -2,35** -2,24** T= (+1,+5) Corrado 1,00% -0,06 0,48

33 Table VII: Regression tests of the relationship between CAR and firm size and percentage dividend decrease.

T=[-5,-1] T=0 T= [+1,+5] C 0,22 (0,02)** 0,07 (0,43) 0,13 (0,27) Logsize -0,02 (0,03)** -0,01 (0,35) 0,01 (0,38) Dividend decrease 0,10 (0,13) 0,03 (0,32) -0,04 (0,47) Adjusted R^2 F statistic 0,05 4,27 0,02 1,51 0,01 0,76 p-value 0,02 0,22 0,49 Ramsey Reset test 0,00 0,11 0,91

***significant at a 1% level, **significant at a 5% level, *significant at a 10% level For all dependent variables, the first number is the coefficient and the second the p- value. The Chi-Square probability of the ramsey reset test is shown in the last row.

34 Table VIII: Regression tests of the relationship between CAR and firm size and percentage shares dividend

decrease for the period T=[-5,-1] with a non-linear term .

T=[-5,-1] C -1,49 (0,02)** Logsize 0,37 (0,05)** Logsize^2 -0,03 (0,05)** Dividend decrease 0,09 (0,19) Adjusted R^2 F statistic 0,07 4,36 p-value 0,01 Ramsey Reset test 0,18

***significant at a 1% level, **significant at a 5% level, *significant at a 10% level For all dependent variables, the first number is the coefficient and the second the p- value. The Chi-Square probability of the ramsey reset test is shown in the last row.

35 the dividend cut for large firms. Another argument that could explain this is that of Wansley, Shilling and Lee (1991). They found that the market anticipates on the dividend cut when they are made late. Unfortunately, we do not have measures in this paper that makes a distinguish between an announcement being late or early.

4.4 Testing for differences in the information content of share repurchases and dividend increase

36 Table IX: Regression tests of the differential information content of share repurchase announcements and

dividend increase announcements.

***significant at a 1% level, **significant at a 5% level, *significant at a 10% level For all dependent variables, the first number is the coefficient and the second the p- value.

As shown in table IX, the coefficient of DUM is significantly positive for small firms for the period T= [+1,+5] and significantly negative for large firms for the period T= [+1,+5]. The medium-sized firms’ samples do not show a significant coefficient on the DUM variable. The positive sign on the DUM variable for small firms indicates that the stock price reaction to a share repurchase announcement is more positive than for a dividend increase announcement. For large firms, the inverse is true. A dividend increase announcement conveys more positive information than a share repurchase announcement for large firms. For medium firms, there is no significant difference in the stock price reaction to a dividend increase and share repurchase announcement. These

Small Medium Large

37 results are consistent with the fact that large investors, who are predominantly in large firms, pay less tax or no tax at all on dividends. Analog to this, small investors are predominantly in small firms and small investors are taxed less on share repurchases. The non-significant results for the medium-sized samples can be explained by the fact that these are not dominated by one group of investors. Therefore, there is no preference for the cash distribution of medium-sized firms based on tax considerations.

5. Conclusion

Firms redistribute money with share repurchases and dividends. This paper investigates whether the stock price reaction to a dividend change announcement and a share repurchase announcement differ for firms of different sizes and how this can be explained.

38 returns for small firms in the period after the announcement and for large firms in the period before the announcement.

40 References

Aharony, J., and Swary I., 1980, Quarterly dividend and earnings announcements and stockholders’ returns: an empirical analysis, Journal of Finance 35, 1-12.

Allen, Franklin and Roni Michaely, 2003, Payout policy, working paper, Financial Institutions Center at the Wharton School.

Alangar S., Bathala, Chenchuramaiah T., and Rao Ramesh P., 1999, The effect of institutional interest on the information content of dividend change announcements, Journal of Financial Research 22, 429-448.

Asquith, P., and Mullins Jr. David W., 1983, The impact of initiating dividend payments on shareholders’ wealth, Journal of Business 6, 77-96.

Bajaj, M., and Vijh Anand M., 1990, Dividend clienteles and the information content of dividend changes, Journal of financial economics 26, 193-219.

Bhattacharya, S., 1979, An exploration of nondissipative dividend- signaling structures, Journal of Financial and Quantitative Analysis 14, 667-668.

Black, F., 1976, The dividend puzzle, Journal of Portfolio Management 23, 10-14.

Brown, C., 2007, The announcement effects of off-market share repurchases in Australia, Australian Journal Of Management 32, 369-385.

Brown, Stephen J., and Warner Jerold B., 1985, Using daily stock returns, the case of event studies, Journal of Financial Economics 14, 3-31.

Choi, D., and Chen S.S., 1997, The differential information conveyed by share repurchase tender offers and dividend increases, The Journal of Financial Research 20, 529-543. Dann, Larry Y., 1981, Common stock repurchases, Journal of Financial Economics 9, 113-138.

Devaney, M., 2008, Will future equity risk premium decline?, Journal of Financial Planning 21, 46-53.

Ghosh, C., and Woolridge J., 1988, An analysis of shareholder reaction to dividend cuts and omissions, The Journal of Financial Research 11, 281-294

41 Haw, I.M., and Kim W.S., 1991, Firm size and dividend announcement effect, Journal of Accounting, Auditing and Finance 6, 325-344.

Jagannathan, M., Stephens, Clifford P., and Weisbach Michael S., 1999, Financial flexibility and the choice between dividends and stock repurchases, Journal of Financial Economics 57, 355-384.

Jagannathan, M., Stephens, C., and Weisbach M., 2000, Financial flexibility and the choice between dividends and stock repurchases, Journal ofFinancial Economics 57, 355-384.

Jensen, M., 1986, The Takeover controversy, Midland Corporate Finance Journal 4, 6-32. John, K., and Williams J., 1985, Dividends, dilution and taxes: a signaling equilibrium, Journal of Finance 40, 1053-1070.

Ikenberry, D., Lakonishok, J., and Vermaelen T., 1995, Market under reaction to open market share repurchases, Journal of Financial Economics 39, 181-208.

Isagawa, N., 2002, Open-market repurchase announcements and stock price behavior in inefficient markets, Financial Management 31, 1-17.

Lakonishok, J., and Vermaelen T., 1990, Anomalous price bahavior around repurchase tender offers, Journal of Finance 45, 455-477.

Lintner, J., 1956, Distribution of incomes of corporations among dividends, retained earnings and taxes, The American Economic Review 46, 97-113.

Mackinlay, Craig A., 1997, Event studies in economics and finance, Journal of Economic Literature 35, 13-39

Masulis, Ronald W., 1980, Stock repurchase by tender offer: an analysis of the causes of common stock price changes, Journal of Finance 35, 305-319.

Miller, Merton H., and Modigliani F., 1961, The cost of capital, corporation finance and the theory of investment, The American Economic Review 48, 261-297.

Miller, Merton H., and Rock K., 1985, Dividend policy under asymmetric information, Journal of Finance 40, 1031-1051.

42 Rogers, Daniel A., 2006, Repurchases, employee stock option grants and hedging, School of Business Administration.

Shefrin, Hersh M., and Statman M., 1984, Explaining investor preference for cash dividends, Journal of financial economics 13, 253-282.

Serra, Ana P., 2002, Event study tests - a brief survey-, working paper 117, University of Porto.

Shleifer, A., 2000, Inefficient markets; an introduction to behavioral finance (Oxford University Press, New York).

Shefrin, Hersh M., and Statman M., 1984, Explaining investor preference for cash dividends, Journal of Financial Economics 13, 253-282.

Short, H., Zhang, H., and Keasey K., 2002, The link between dividend policy and institutional ownership, Journal of Corporate Finance 8, 105-122.

Vermaelen, T., 1981, Common stock repurchases and market signaling, Journal of Financial Economics 9, 139-183.

Wansley, James W., Shilling, James D., and Lee Y.J., 1991, Dividend change

announcement effects and earnings volatility and timing, Journal of Financial Research 14, 37-49.

43 Appendix A:

A Jarque Bera test is used to test whether the distribution of abnormal returns is normal or non-normal for each of the 15 samples. The result of the Jarque Bera test for normality are shown in table 1 and indicates whether it is appropriate to use a parametric or a non-parametric test for testing the null hypothesis. The other test is used as a robustness check.

Table II: Jarque bera test statistic and appropriate test per sample

Event window T= (-5,-1) T= 0 T= (+1,+5)

Dividend change positive small firms 572,12 (Corrado test) 26,76 (Corrado test) 35,08 (Corrado test)

Dividend change positive medium firms 87,10 (Corrado test)

6,27 (T-test)

52,48 (Corrado test) Dividend change positive large firms 138,61

(Corrado test) 9,70 (Corrado test) 38,78 (Corrado test)

Dividend change negative small firms 3,93 (T-test) 17,68 (Corrado test) 35,36 (Corrado test)

44 Dividend change negative large firms 26,37

(Corrado test)

5,53 (T-test)

8,99 (Corrado test) Share repurchase small firms 465,14

(Corrado test) 15,99 (Corrado test) 6247,45 (Corrado test)

Share repurchase medium firms 362,34 (Corrado test)

5,96 (T-test)

1230,67 (Corrado test) Share repurchase large firms 14,04

(Corrado test) 28,95 (Corrado test) 3,46 (T-test)