GERMAN ECONOMIC ASSOCIATION OF BUSINESS ADMINISTRATION GEABA DISCUSSION PAPER SERIES IN ECONOMICS AND MANAGEMENT

Transcription

1 DISCUSSION PAPER SERIES IN ECONOMICS AND MANAGEMENT The Role of Patents in Acquiring High Technology Firms Tobias Pick, Deborah Knirsch & Rainer Niemann Discussion Paper No GERMAN ECONOMIC ASSOCIATION OF BUSINESS ADMINISTRATION GEABA

2 The Impact of the German Corporate Tax Reform 2002 on Stock Prices on the Anouncement Day of Share Repurchases Tobias Pick Deborah Knirsch Rainer Niemann University of Graz WHU Otto Beisheim School of Management University of Graz Center for Accounting Chair of Taxation and Institute of Accounting Research Accounting and Taxation Universitaetsstr. 15/G2 Burgplatz 2 Universitaetsstr. 15/G Graz Vallendar 8010 Graz Austria Germany Austria tobias.pick@uni-graz.at deborah.knirsch@whu.edu niemann@uni-graz.at Phone Phone Phone Fax Fax Fax Abstract In this study we analyse stock price reactions to share buyback announcements from a tax perspective in Germany. To determine the influence of taxes on stock prices on the announcement day of share buybacks two different tax regimes the full imputation system and the half income system are analysed. We also integrate the shareholder structure in our analysis, because it has an impact on the relative tax advantage of share buybacks over dividends. We find evidence that different tax regimes influence the share price reaction in the event period. Furthermore we find evidence for the substitution hypothesis and the dividend clientel effect: High dividend paying companies have smaller positive price reactions than non- or lower dividend paying companies.

3 1 Introduction In Germany, stock repurchases were generally prohibited until Since then, they have been an increasingly popular instrument for distributing corporate profits. Different aspects of share repurchases discussed in the literature are signaling effects, agency conflicts, aspects of behavioral finance and tax effects. 1 As capital gains are taxed preferentially over dividends in Germany and other countries tax law, taxation can be expected to have an impact on distribution policy and is widely discussed in the finance literature. 2 In this paper we focus on tax aspects only. In 2002, a major corporate tax reform took place in Germany. 3 The corporate full imputation system which was in effect since 1977 was replaced by a classical corporation tax with shareholder relief elements. Before, the corporate income tax on distributed profits was fully credited against the shareholder s personal income tax liability. In contrast, capital gains did not entitle to tax credits. Since the tax reform, the corporate income tax cannot be credited anymore independently of the corporate distribution policy. Only half of the dividends are subject to the individual income tax (halfincome system). Obviously, the relative advantage of share repurchases over dividends changed due to the tax reform. Capital gains from selling shares are taxable only if an individual shareholder owns a substantial interest in the corporation. Tightening capital gains taxation by lowering the critical threshold for substantial interest was another element of the tax reform. We analyse how this tax reform influences market reactions to stock repurchases. We measure the market reaction on the announcement day of share repurchases. The ad hoc announcement of a share repurchase is regulated by German law and is obligatory 1 See Dittmar (2000); Grullon / Michaely (2004); Jagannathan / Stephens (2003) and Jagannathan / Stephens / Weisbach (2000) for the US and Hackethal / Zdantchouk (2004) for Germany. 2 See f.i. Grullon / Michaely (2002) for the US; Amihud / Murgia (1997) for Germany. 3 See Schreiber (2000); Homburg (2000). 1

4 if the event has a likely impact on share prices. In an event study, we compare abnormal returns around the ad hoc announcement day before and after the tax reform, respectively. As capital gains taxation depends on the shareholder s stake in the corporation, we distinguish between shareholders with a stake of more or less than 1% or 10% (between 1999 and 2001), respectively. The investigation period covers the years 1998 to 2006 in order to include several years before and after the corporate tax reform. A handful of other studies have looked at share price reactions to the announcement of share buybacks in Germany. 4 In distinction to this study, there does not exist empirical work, which measures the effects of different tax systems on the share price reaction of share buyback announcements in Germany. Most closely related to this article are Grullon/ Michaely (2002). They find a different share price reaction to share buyback announcements before and after the Tax Reform Act in They conclude, that the decreasing tax advantage of capital gains over dividends after the tax reform is responsible for the stock price reaction in the US. 5 Methodically, this study can be differentiated from all other German studies because of the consideration of a tax variable in the regression analysis. This tax variable represents the tax advantage of capital gains over dividends, considering the shareholder structure. The analysed proxy variables for the German stock market are: dividends per share and the dividend yield. Beside the constructed tax variable, these proxy variables help to explain the relationship between share buybacks and dividends. 6 The remainder of the article is organised as follows: section 2 describes the legal framework, before our hypotheses are developed. Section 3 details the database and 4 Publications for the German stock markets are, for example: Schremper (2002), Gerke / Fischer / Langer (2003), Pertlwieser (2006), Seifert (2006). 5 See Grullon / Michaely (2002), P These proxy variables are used in: Pertlwieser (2006) and Seifert (2006). 2

5 the methodology. Section 4 presents the main results, including the event study and the regression analysis. Section 5 concludes. 2 Hypotheses 2.1 Impact of share repurchase announcements on stock prices Because of the tax advantage of capital gains over dividends there exists an incentive for the substitution of dividends through share buybacks since the adoption of the KonTraG in 1998 (substitution hypothesis). 7 Alternatively, share buybacks can be considered as complements to dividends (complement hypothesis). Companies can be reluctant to cut dividend payments (Lintner-Model 8 ) but want to implement share buybacks in order to transfer nonrecurring excess cash flows to the shareholders. A dividend increase would be interpreted as a signal for higher prospective dividend payments. Therefore a dividend increase is not practicable. In this case, share buybacks are used as a new opportunity to transfer cash to the shareholders. The basic fundamental principle to conduct this study is the semi-strong information efficiency of the stock market. 9 The semi-strong form efficiency implies that if the investor takes tax considerations into account the relative tax advantage of share buybacks over dividends must be priced into the stock market rates on the announcement day of the share buyback. In a first step the following alternative hypothesis proves whether the announcement of a share buyback provides additional information for the stock market: 7 The German act Gesetz zur Kontrolle und Transparenz im Unternehmensbereich (KonTraG) was enacted on May, 1st See Lintner (1956) 9 A semi-strong efficient stock market implies, that all publicly available information is already priced into the stock market rates. See Fama (1970), P. 383 pp. 3

6 H1: The announcement of the share buyback has an information content on the announcement day. If the alternative hypothesis H1 can be approved, then the announcement of share buybacks has an information content for the stock market. However, it is ambiguous if the tax reform has an influence on the stock market. To control for the tax reform 2002 (Change from the imputation system to the half-income system) a second alternative hyothesis is formulated: H2: The size of the abnormal returns differ between the half incomeand imputation tax regime, depending on the shareholder structure. To prove this alternative hypothesis it is necessary to create several portfolios, which represent the dominant shareholder groups, separated by tax law. Figure 1 shows the relevant shareholder structure from a tax perspective: Figure 1: Relevant shareholder structure from a tax perspective. The first portfolio consists of companies, whose shareholders (more than 50% of the rights to vote) are dominated by individuals with non-substantial interest. The second portfolio consists of companies, whose shareholders (more than 50% of the rights 4

7 to vote) are dominated by individuals with substantial interest. The third portfolio consists of companies, whose shareholders (more than 50% of the rights to vote) are dominated by corporations. A portfolio consisting of companies which are dominated by investment funds is not formed. Investment funds are taxed according to the transparency principle. Their taxation depends on the tax status of their shareholders. In the following, investment funds are subsumed under individuals with non-substantial interest. The portfolio creation takes the different relative tax advantage of capital gains over dividends depending on the shareholder structure into account. These different tax advantages should be visible in the stock price reaction on the announcement day of the share buyback. To show the effect of the tax reform 2002, every portfolio must be separated in two samples. The first sample corresponds to the observation period from May, 1, 1998 to December, 31, 2001 (imputation system), the second sample corresponds to the observation period from January, 1, 2002 to December, 31, 2006 (half income system). 2.2 Taxation of individual shareholders and corporate shareholders This section details the dividends and capital gains taxation prior and after the tax reform Especially, the taxation of different shareholder groups, separated by tax law, is taken into account. The different taxation of shareholders is solely considered on the first shareholder level. The taxation of downstream shareholders is ignored. 10 Prior to the tax reform, up to 2001, dividends received from domestic corporations 10 Implicitly, it is assumed, that the shareholders are optimising the cash flow between them and their company without recognizing in case of a corporation their own shareholders. This assumption is reasonable, because a corporation as a first level shareholder can optimise its cash flow to its owner after receiving dividends from its holdings. 5

8 entitled domestic individuals to a tax credit. As a consequence, distributed profits of corporations were effectively subject only to the individual income tax. 11 Due to crediting the corporate income tax against personal income taxes, the corporate income tax rate was irrelevant with regard to dividend taxation. The maximum personal income tax rate including solidarity surcharge 12 τ i decreased from 55.92% in 1998 to 44.31% in Long-term capital gains from stocks held for more than one year were tax-exempt for shareholders with non-substantial interest. The threshold was defined as a stake of less than 25% up to 1998, a stake of less than 10% from 1999 to 2001, and a stake of less than 1% as from The underlying profit was only subject to the corporate income tax. For retained profits, the corporate income tax rate including solidarity surcharge τ c decreased from 47.48% in 1998 to 42.20% in , and to 26.38% from In 2003, an increase to 27.96% was temporarily made to finance damages of a heavy flood. Figure 2 shows the time series of the threshold and the corporate income tax rate for retained earnings. Figure 2: History of threshold and corporate income tax in the observation period. 11 The German local business tax is neglected because it is levied on both retained and distributed earnings and has not been changed in the course of the 2002 tax reform. 12 In the following, we will only take top marginal tax rates including solidarity surcharge into account. The solidarity surcharge is a 5.5% tax on top of the personal income tax and the corporate income tax. Germany has introduced this tax in order to finance the German reunification. 6

9 The value of a retained profit was higher than its after-tax value because of the inherent tax credit. Thus, buyers of the shares were willing to pay for the tax credit. If we assume that the whole tax credit was comprised in the price calculation of the buyer, then the effective tax rate of the seller was zero, because the corporate tax payments were refunded by the buyer and no additional personal income taxes had to be paid. 13 In case of non-substantial interest, tax rates on dividends τ d and capital gains τ g under the full imputation system are computed as τ F I d = τ i (1) τ F I g = 0 (2) Tax rates on dividends τ d and capital gains τ g under the half-income system are computed as τ HI d = τ c + (1 τ c ) 0.5 τ i (3) τ HI g = τ c (4) In table 1, we show the effective tax rates on dividends and capital gains related to corporate profit. For example, in 1998 dividends were taxed at the personal income tax rate τ i of 55.92%. Capital gains from retained corporate profits were taxed at the corporate income tax rate of 47.48%, but the seller paid additional 47.48% for the share for the inherent tax credit. So the effective tax burden of the seller is 0%. The difference F I = τ F I d τ F I g of 55.92% indicates that capital gains are taxed preferentially over dividends. As a consequence, there is a strong tax-induced incentive in favor of share repurchases for 13 The maximum pretax value for 1 DM cash dividend is DM to a taxable German investor. McDonald (2001) finds that the empirically tested value is 1.26 DM, see McDonald (2001). 7

10 Year System Threshold Dividends tax rate Capital gains tax rate Difference τ d (%) τ g (%) (%) 1998 FI 25% FI 10% FI 10% FI 10% HI 1% HI 1% HI 1% HI 1% HI 1% Table 1: Effective tax rates on dividends and capital gains for shareholder with nonsubstantial interest. FI: Full imputation system, HI: Half-income system. high-income shareholders. Under the half income system in 2002 corporate profits were taxed at the corporate income tax rate of 26.38%. Additionally, distributed profits after corporate taxes were subject to the half-income system on the individual level. A total tax burden of τ HI g = 26.38% + ( %) % = 45.21% results. From the decreasing -values it can be seen that the tax incentive to repurchase stocks decreased compared to the full imputation system. This results in the following hypothesis: H2.1: Under the half-income system, the abnormal return for stock repurchase announcements is lower than under the imputation system for corporations which are dominated by individual shareholders with tax-exempt capital gains. Capital gains of shareholders with substantial interest were taxable under both tax systems. From 1998 to 2001, capital gains were fully taxed, but reduced tax rates were applied. In 1998, capital gains were taxed at half of the individual income tax rate. In the following years, preferential tax rates between 45.37% and 47.48% were applied 8

11 to capital gains of shareholders with substantial interest. Again we assume that the corporate income tax was refunded by the buyer because he received the tax credit. In case of substantial interest, the tax rates on dividends τ d and capital gains τ g under the full imputation system are computed as τ F I d = τ i (5) τ F I g = τ reduced i (6) The tax rates on dividends τ d and capital gains τ g under the half-income system are computed as τ HI d = τ HI g = τ c + (1 τ c ) 0.5 τ i (7) Year System Threshold Dividends tax rate Capital gains tax rate Difference τ d (%) τ g (%) (%) 1998 FI 25% FI 10% FI 10% FI 10% HI 1% HI 1% HI 1% HI 1% HI 1% Table 2: Effective tax rates on dividends and capital gains for individual shareholders with substantial interest. FI: Full imputation system, HI: Half-income system. Obviously, the full imputation system for major shareholders provided strong incentives to repurchase shares. Under the half-income system, major individual shareholders are indifferent between dividends and share repurchases The different incentives between the full imputation system and the half income system are even 9

12 Summarizing, the incentive for share repurchases are lower under the half-income system. H2.2: Under the half-income system, the abnormal return for stock repurchase announcements is lower than under the imputation system for corporations which are dominated by individual shareholders with taxable capital gains. Corporate shareholders are taxed differently from individual shareholders. Under the full imputation system, dividends were fully taxed on the corporate level. However, a full tax credit was granted to corporate shareholders. As a result, corporate income tax was levied exactly once, regardless of the corporate structure. In contrast, capital gains did not entitle to a tax credit. But again we assume that the buyer was willing to pay for the tax credit because it reduces his later dividend taxation. Thus the effective tax rate for capital gains was the corporate tax rate paid on the gain. The tax rates on dividends τ d and capital gains τ g of corporations under the full imputation system are computed as τ F I d = τ F I g = τ c (8) Under the half income system, the corporate tax rate was reduced and capital gains and dividend distribution were 95% tax exempt for corporations, but there was no tax credit granted. Thus, both alternatives are taxed identically. The tax rates on dividends τ d and capital gains τ g under the half-income system are higher when we take the decrease of the threshold for substantial interest into account. Capital gains that were formerly tax-exempt became taxable due to the reduction of the substantial interest threshold. The reduction of the critical threshold induces stronger effects compared to hypotheses H2.1 and H2.2, because more capital gains became taxable which reduces the incentive to repurchase shares. 10

13 computed as τ HI d = τ HI g = τ c + (1 τ c ) 0.05 τ c (9) Table 13 in the appendix shows the relative tax advantage of capital gains over dividends if the shareholders are corporations. If corporations are the dominant shareholder, then there is no tax advantage of capital gains (share buybacks) over dividends. 15 This result implies that the stock price reaction between the two samples should not differ significantly. H2.3: The abnormal return does not differ significantly between the two samples (imputation system versus half income system) if corporations are the dominant shareholders. 2.3 Alternative measures for tax effects Share buybacks are influenced by several factors. The focus of this study is to isolate the tax influence on the choice between share buybacks and dividends. For this reason a regression analysis is used to isolate tax effects from other relevant factors. Tax effects on the abnormal returns in the event window are measured, directly, through a tax variable and, indirectly, through proxy variables. As proxy variables for isolating the tax effect the adjusted dividend and the dividend yield are used. Hypothesis H3 approves the relationship between the abnormal return and the adjusted dividend payment in the event window. If there is a substitution effect, then the abnormal returns should decrease with an increasing adjusted dividend. Higher dividend 15 There could be different reasons to carry out share buybacks, such as changes of the capital structure or stock option plans. These reasons can lead to positive stock price reactions on the announcement day of share buybacks in spite of the tax irrelvance. 11

14 payments reduce the potential for prospective share buybacks. If share buybacks are used as complements, then the abnormal returns should increase with an increasing adjusted dividend. H3: An increasing dividend influences the share price reaction in the event window (substitution- versus complement hypothesis). A potential substitution effect should be higher under the imputation system compared to the half income system. The relative tax advantage of capital gains over dividends is lower under the half income system. This implies the following alternative hypothesis. H3.1: The substitution effect between dividend payments and share buybacks is higher under the imputation system. Apart from the substitution effect the dividend clientel effect could provide an explanatory content under a tax perspective. Corporations, which follow a high dividend yield strategy, could be dominated by shareholders with a low marginal income tax rate. If these corporations try to substitute dividends through share buybacks, their shareholders should profit less than shareholders with a high marginal income tax rate. H4: Corporations with a higher dividend yield show a lower share price reaction in the event window. 3 Sample and data description 3.1 Data base Electronically distributed ad hoc announcements serve as the data base to identify share buyback announcements The following data bases are searched for ad hoc announcements of companies which are listed on the stock exchange: Deutsche Gesellschaft für Ad-hoc Mitteilungen ( Hugin 12

15 The time frame investigated starts with the introduction of KonTraG, which facilitates the purchase of one s own stocks according to sec. 71 subsec. 1 nr. 8 AktG, on May, 1, 1998 and ends on December, 31, ad hoc announcements of 244 companies could be identified by analyzing the data bases. 17 Companies which were not listed on the CDAX 18 at the time of announcement were excluded from the analysis. The returns and abnormal returns were calculated by using data obtained from the Datastream data base by Thomson Financial. The share prices are adapted to the Return Index. The Return Index adjusts for dividends and capital changes and uses the (Xetra) closing prices of the respective domain stock exchange. For better comparability the CDAX Performance index, based on the closing prices of the domain stock exchange, is used to calculate the market rate of return. Datastream data base also provides dividend per share, the market value and the index membership to the New Market Index of all companies in the investigated sample. The shareholder structure is taken from the Hoppenstedt Aktienführer. During the investigated time period the trading hours at the Frankfurt stock exchange change. The trading hours of the floor trading 19 are 9 am to 5:30 pm from 1998 to June, 1, Ad hoc announcements which are published after 5:30 pm are assumedly processed on the following trading day. Starting on June, 2, 2000 the trading hours are prolonged to 8 pm. From this follows that share buyback announcements after 8 pm are allocated to the next trading day. ( Euro adhoc Portal ( 17 Companies which did not publish ad hoc announcements electronically were excluded from the analysis. 18 CDAX measures the development of the German stock market. Companies which are traded at the over-the-counter market are excluded. See Deutsche Börse (2007a), P Since stock repurchase announcements of all CDAX-companies are investigated only floor trading and not the electronic trading system Xetra, which only includes the prime standards, is considered as the relevant trading platform. See Deutsche Börse (2007b), P

16 3.2 Selection criteria It is necessary to adjust the dataset to investigate the effects on returns resulting from stock repurchase announcements. Table 3 shows the process of adjusting the data: Data Adjustments Numbers of Announcements starting point 418 preferred stocks 3 unknown event date -33 sec. 71 subsec. 1 nr. 1-7 Akt -3 contaminated ad hocs -50 overlapping ad hocs -16 over-the-counter trading -5 Total 314 Table 3: Data adjustment. Starting point of the analysis are 418 ad hoc announcements. 153 of these announcements are based upon our own investigation for the time period between April, 1, 2003 and December, 31, ad hoc announcements are based upon a data base used in Pertlwieser s dissertation (2006) for the time between May, 1, 1998 and March, 31, Our own investigation must be reduced by three announcements which lead to a stock repurchase according to sec. 71 subsec. 1 nr. 1-7 AktG. 20 Five more announcements are excluded because companies are traded over the counter. Ad hoc announcements for common share and preferred stock are considered twice in the investigation. This increases our own dataset by three ad hoc announcements. Since it was not possible to identify the exact date for 33 ad hoc announcements these were eliminated. A total of 50 ad hoc announcements contain further relevant information for the capital market besides the stock repurchase announcement. This means that market reactions may not result only from the share buyback announcements. Announcements containing the following additional information are eliminated: 20 An adjustment for ad hoc announcements according to sec. 71 subsec. 1 nr. 1-7 AktG is necessary because those share buybacks relate to non-payout policy conditions. 14

17 Suggestions to raise dividends Change of managing board or supervisory board Expectations about the business year Purchase of business units, acquisitions When analyzing the reactions of the capital market the consideration of these contaminated ad hoc announcements could lead to biased results, since the share price reaction does not depend only on the information about the stock repurchase. In a last step the announcements which overlap in the event window are deleted. As a result 14 announcements of our own investigation and two announcements from Pertlwieser s dataset are eliminated. Thus it is guaranteed that the calculated returns in the event window are not biased by additional occurrences. The final dataset consists of a total of 314 ad hoc announcements. 117 announcements are taken from our own investigation and 197 are based on the dataset from Perlwieser s dissertation (2006). 3.3 Methodology In order to analyse the derived hypotheses from Chapter 2, an event study is applied to quantify the capital market s reaction to stock repurchase announcements. 21 First we model the returns for the estimation window which is defined as the return that would be expected if the event did not take place. Second we calculate the deviation of the actual return from the expected return, and the so-called abnormal return is tested for significance. If the abnormal return differs significantly from the expected 21 See Schremper (2002), P

18 Figure 3: time line of an event study (according to: Campell et al. (1997), P. 157.) Note: The relevant trading days before and after the event are in parentheses. return at the time of the event (stock repurchase), then the announcement contains relevant information for the stock market. Then we investigate whether the abnormal return changes significantly as the full imputation system is replaced by the half-income system. The post-event window which serves to analyse long-term developments after the announcement is not considered. The market model 22 is estimated by calculating the alpha and beta factor for the estimation window 23 and serves to model the expected return of the event window: R it = α i + β i R mt + ɛ it, (10) E(ɛ it ) = 0, V ar(ɛ it ) = σ 2. (11) The deviation between the actual return (R it ) and the expected return ( ˆα i ˆβ i R mt ) results in the abnormal return during the event window and is calculated as follows: AR it = R it ˆα i ˆβ i R mt. (12) 22 See MacKinlay (1997), P The length of the estimation window is 125 trading days 16

19 R it is the logarithmic return 24 of security i on day t, whereas R mt is the logarithmic return of the DAX performance index on day t. α i und β i are the estimating parameters of the market model, which are calculated according to formula (10) for a 125-days window from -145 to -21 days by applying the least squares method and ɛ it is the error term of security i on day t. The market model assumes a linear timely uncorrelated relationship between the market return and the security return and a normal distribution of the abnormal returns with a zero conditional mean. 25 The calculated daily abnormal returns for each security i (AR it ) from a sample with N events are aggregated for every day of the event window to derive an average abnormal return (AAR t ): AAR t = 1 N N AR it mit t = 20,..., 20. (13) i=1 To draw conclusions concerning a longer period of time we aggregate the daily abnormal returns of every security i (AR it ) for the event windows [-20;-2] before [-1;1], during and after [2; 20] the announcement. This leads to cumulated abnormal returns CAR of security i for the respective periods from t 1 to t 2. For the average cumulated abnormal return (CAR i (τ 1 ; τ 2 )) we have 26 : CAR(τ 1 ; τ 2 ) = 1 N N CAR i (τ 1 ; τ 2 ) (14) i=1 We use the two-sided t-test and the Wilcoxon rank sum test for testing the hypotheses. 24 The logarithmic return opposed to the simple return follows a normal distribution which is assumed later for the significance tests (t-test) Another advantage of logarithmic returns is the possibility to determine the return of a longer period by simply adding up the daily returns. The logarithmic return is defined as: R i,t = ln( Pi,t P i,t 1 ), see Thomson (1988), P See Campbell / Lo / MacKinlay (1997), P An alternative aggregation is to aggregate the average abnormal returns AAR t. Both methods differ by the variance. See Campbell / Lo / MacKinlay (1997), P. 160 pp. 17

20 For the average abnormal return we have the following t-test statistic: t AAR(t) = 1 N 1 N 1 AAR t (15) n i=1 (AR it AAR t ) 2 The second multiplicand of the denominator is an unbiased estimator of the standard deviation of the population. Dividing by N leads to the estimated standard error of the mean of the population. footnotesee Sheskin (2000), P For the average cumulated abnormal return (CAR(τ 1 ; τ 2 )) we have this test statistic: t CAR(τ1 ;τ 2 ) = CAR(τ 1 ; τ 2 ) (16) n i=1 (CAR i(τ 1 ; τ 2 ) CAR(τ 1 ; τ 2 )) 2 1 N 1 N 1 Again, the second multiplicand of the denominator is an unbiased estimator of the standard deviation of the population. As before, the estimated standard error is divided by N. The t-test is based on the assumption of a normal distribution which could be contradicted. Thus, the Wilcoxon rank sum test 27 is used to check the test statistic. In the course of univariate statistics we derive abnormal returns and want to explain their influencing factors - especially tax aspects - by applying a regression analysis. As a first explanatory variable 28 for analyzing tax aspects in the regression we introduce the Tax-1 -variable which represents the tax advantages (Div V G t ) for the dominating group of shareholders (at least 50% share property) in a firm. 27 The Wilcoxon rank sum test is a non-parametric test which can also be used for small samples since it is not based on any particular distribution. 28 The explanatory variables regarding tax are calculated exogenously. Thus, they serve to approximately estimate the taxation of domestic shareholders. They are a form of artificial generated proxies. For a similar procedure on the American stock market, see Grullon / Michaely (2002), P

21 The Tax-1 -variable is calculated according to the following equation: T ax 1,t = (Div V G) unw,t if unw 50% (Div V G) wes,t if wes 50% (Div V G) Kap,t if Kap 50% (17) If there doesn t exist a dominant shareholder group, then the Tax-1 -variable is the weighted average of the tax advantages for every shareholder group. The calculation of the tax advantage for this company follows equation 18. unw stands for shareholders with non-substantial interest- in respect of taxes -, wes for shareholders with substantial interest and Kap for stock corporations as shareholders. The Tax-1 -variable is based on the idea that dominating shareholder groups can influence the management to favorably alter dividend policy. In a further regression an alternative tax variable, the TAX-2 -variable, is used. In this variable the tax advantage is weighted with the corresponding share property. Since foreign investors from different countries hold shares from German firms their advantageousness of stock repurchases is not considered (the sum of all shares from a firm may be below 100%). The following equation represents the advantageousness: T ax 2,t = (Div V G) unw,t x 1,t + (Div V G) wes,t x 2,t + (Div V G) Kap,t x 3,t (18) mit x 1,t + x 2,t + x 3,t 1 We also include dividend per share as another tax proxy in the regression. It serves as a further estimator of possible substitution effects between dividends and stock repurchases. 19

22 DP S i,t = div i,t number of shares i,t (19) The last tax relevant variable which we need for our regression is the dividend yield per share. It is used to show a possible dividend clientele effect. DivY i,t = dividend i,t stock price i,t 1 (20) As control variables we take the company size - approximated by the logarithmic market value -, a dummy variable for bull and bear markets, and a dummy variable for an index membership in the New Market Index, which was established in 1998 and closed on June, 5, The following equation shows the logarithmic market value: Size i,t = ln(marketvalue i,t ) (21) The CDAX is used for the categorization into rising and falling stock markets for which we need a dummy variable. A dummy variable of 1 stands for a rising stock market. (The figure in the appendix shows the categorization into rising and falling stock markets.) StockEx t = 1 t March, 07, t March, 13, otherwise (22) We use the control variable NewM to detect companies listed on the New Market Index. Thereby we can figure out the influence of the market segementation on the results. 29 Companies which are listed at the New Market have a dummy variable of 29 For univariate statistics Gerke et al.(2003) carried out different results (AAR t ) for different market 20

23 one. For the dummy variable we have: 1 N ewm arketindex NewM i = 0 otherwise (23) The regression equations used are estimated according to the method of ordinary least squares. We adjust standard errors if there is heteroskelasticity according to White (1980). 4 Results 4.1 Univariate statistics Analysis of share price reactions for the whole observation period Table 4 shows the results for the total time period considered: time period CAR(τ 1 ; τ 2 ) t value z value bzw. AAR t % % % bis % bis % bis % Notation: t- and z-values are significant with 1.65, 1.96 and 2.58 to the 10%-( ), 5%-( ) and 1%-( ) level of significance. Table 4: Average abnormal return (AAR t ) and cumulated average abnormal return CAR(τ 1 ; τ 2 ) for the whole observation period. segments at the announcement days, see Gerke / Fischer / Langer (2003). 21

24 The first column of the table shows the average abnormal return (AAR t ) on the day before [-1], on the event day [0] and on the day after the event [1]. The values [-20;- 2] and [2; 20] represent the time period before and after the event day. For this time period we calculate the cumulated average abnormal return (CAR(τ 1 ; τ 2 ) ). The second and third columns display the t-value and the z-value of the Wilcoxon signed-rank test. We find a significantly negative abnormal share price development of % for the time period [-20;2]. The management of enterprises tends to announce share buybacks especially after a significant decline of prices. A possible explanation for the announcement could be a predominant undervaluation according to the management (signaling). On the day prior the announcement [-1] there is no significant positive abnormal return. Thus insider trading before the stock repurchase announcement is unlikely. On the announcement day we find a significantly positive price jump of 4.595%. The announcement is judged positive by the stock market. This result conforms to all the studies considered for the German stock market in chapter 1. One day after the announcement day the average cumulated return further increases by insignificant 0.383%. This leads to the assumption that some players on the stock market display a delayed reaction. During the time after the announcement we find an insignificant increase of the average abnormal return by 0.981%. Delayed information processing of the announcement can not be proven statistically for the time period [2;20]. The alternative hypothesis H 1 can be confirmed. The stock repurchase announcement contains relevant information for the stock market. The considered sample of the total time period is divided into two samples with differing time periods. The first sample serves to investigate share price reactions to announcements at the time of the full imputation system. The second sample looks at share price reactions when the half income system is applied. Table 14 in the appendix shows the results for the two samples. Figure 4 displays the results graphically. 22

25 Figure 4: History of the average cumulated abnormal return CAR(τ 1 ; τ 2 ), separated by the tax regime. Figure 4 demonstrates that on the day of the announcement share price reactions are higher when the full imputation system is applied (4.977% vs %). The difference between the two samples is not significant. Looking at the average cumulated abnormal return on [-1;1] as opposed to the average abnormal return on the event day we observe a significant difference between the full imputation system and the half income system. The average cumulated abnormal return under the full imputation system is 6.068%. For the same time period [-1;1] it is 3.856% under the half income system. We can conclude that the average cumulated abnormal return [-1;1] differs significantly between the full imputation system and the half income system. This conclusion is based on the t-test as well as on the Wilcoxon signed-rank test which is not based on any particular distribution and does not overweight outliers. The alternative hypothesis which states that share price reactions are higher under the half income system can not be validated in the case of univariate statistics. The stock market displays the opposite result. The full imputation system leads to higher share price reactions. 23

26 4.1.2 The influence of the shareholder structure on abnormal returns The presented results for the total sample could be influenced substantially by the shareholder structure of firms. The different advantageousness of shareholder groups 30 regarding tax could bias the result of share price reactions. In this subsection we present the share price reaction in dependence of the shareholder structure for the univariate case. Table 5 presents the number of ad hoc announcements of companies with shareholder groups of substantial interest (criterion of dominance). Whole Period FI HI Domestic Non-Substantial Substantial Corporation Foreign Individuals/Fonds Corporation No Dominance Total Table 5: Total amount of the share buyback announcements, separated by the shareholder dominance. In table 5 we see that in 138 or 44.9% of all companies regarded we have shareholders with non-substantial interest. 31 The number of ad hoc announcements of these share- 30 We divided the domestic shareholder groups into individuals with substantial and non-substantial interest and stock corporations. Companies which are dominated by investment funds are counted as individuals with substantial interest according to the transparency principle. The differentiation between individuals with substantial and non-substantial interest is treated in chapter Shareholder structure has been taken from Hoppenstedt Aktienführer. Since the raw data is classified unequally and at times incomplete we have to make assumptions regarding the classification. The shareholder structure labeled as institutional investors belongs for us to the category of investment funds. We consider shares of the management, the supervisory board and family-owned shares as shareholders with nun-substantial interest since the exact number of people involved is unknown. Partnerships are regarded as shareholders with substantial interest. We do not consider the business tax in our analysis. If the shareholder structure is incomplete (shares are less than 100%) we increase the free float. If we cannot change the categorization we eliminate those ad hoc announcements. 24

27 holder groups divided into full imputation and half income system do not differ substantially. We only have 75 ad hoc announcements of companies which are dominated by shareholders with substantial interest. From these 75 announcements 49 (15.9%) belong to the time period of the full imputation system and 26 (8.47%) to the half income system period. At the same time we consider 39 (12.7%) announcements from companies which are dominated by corporations of which 14 (4.56%) were published during the full imputation period and the remaining 25 (8.14%) in the half income system period. A total of 9 (2.93%) foreign investors (among them 8 stock corporations) which dominate German companies publish ad hoc announcements during the time period considered. This small number is the reason why foreign investors are not further included in our study. From this we can see that corporations dominated by shareholders with non-substantial interest announce more share buybacks than corporations with dominant shareholder groups. For the investigation of the alternative hypotheses 2.1 to 2.3 we divide the total sample into three portfolios. Portfolio 1 contains the ad hoc announcenments of companies with at least 50% shareholders with non-substantial interest. Portfolio 2 comprises of companies which are dominated by shareholders of substantial interest with at least 50% of the votes. Companies in portfolio 3 are dominated by corporations with at least 50% of the votes. Figure 5 shows the development of the average cumulated abnormal returns CAR(τ 1 ; τ 2 ) for each shareholder group over time: Table 15 in the appendix displays the average abnormal returns (AAR t ) and the average cumulated abnormal returns CAR[ 1; 1] for all three portfolios structured according to substantial and non-substantial interest and stock corporations for the total time 25

28 Figure 5: History of the average cumulated abnormal return CAR(τ 1 ; τ 2 ) for the whole observation period, separated by the shareholder group. period considered. On the event day [0] ad hoc announcements of companies with shareholders of substantial interest show the most significant share price reactions of 5.836%. Regarding the average cumulated abnormal return [-1; 1] the share price reaction increases to significant 6.475%. In contrast, companies with shareholders of non-substantial interest display a significant abnormal return of 3.308% on the day of the announcement and an average cumulated abnormal return of 3.377% in [-1;1]. For companies with mainly stock corporations as shareholders the average abnormal return is with 4.443% on the event day [0] in the middle. The average cumulated abnormal return in [-1; 1] is 4.679% for this portfolio. From a tax perspective one would expect share price reactions to be the highest for companies with mainly shareholders with non-substantial interest since the shareholders of these companies profit most from a sale of their shares to the company. One would also expect share price reactions to be lower for companies with mainly share- 26

29 holders with substantial interest since these shareholders profit less in comparison to shareholders with non-substantial interest. This argumentation also applies for portfolio three. From a tax point of view stock corporations have no tax advantages from capital gains opposed to dividends. The share price reaction is expected to be the lowest if corporations dominate. A possible explanation of the results can be found in the signaling theory. Ad hoc announcements by companies with mostly shareholders of substantial interest display the highest abnormal negative share price development. Due to the corporate governance structure information policy could be worse and information asymmetry larger in the case of dominating shareholders with substantial interest. The reduction of increased information asymmetry could lead to higher price reactions on the announcement day. Another possible explanation for the results is liquidity and trading volume of the share. Shares with a high stake of non-substantial shareholders are more liquid than shares with a low stake. High liquidity could lead to lower price reactions due to better information processing as opposed to shares with lower liquidity. The alternative hypothesis H 2 cannot be validated. The portfolio with dominating shareholders of non-substantial interest has the lowest price reactions out of the three portfolios. Tax aspects could be irrelevant or overlapped by other effects. For testing the hypotheses H2.1 to H2.3 it is necessary to divide the portfolios into two time periods: the full imputation and the half income system. Table 6 shows the average abnormal return and the cumulated average return [-1; 1] for the full imputation and the half income system regarding companies dominated by shareholders with non-substantial interest. From table 6 we can see the significantly positive price reaction on the event day which is 3.483% under FI and 3.160% under HI. From the t- and z-values we know that the two samples do not differ significantly. Nevertheless we find a significant difference 27

30 FI (N=63) HI (N=75) Difference time period CAR t value z value CAR t value z value t value z value bzw. AAR t bzw. AAR t % % % % % % bis % % Notation:t- and z-values are significant with 1.65, 1.96 and 2.58 to the 10%-( ), 5%-( ) and 1%-( ) level of significance. Table 6: Average abnormal return AAR t and average cumulated abnormal return CAR( 1; 1) for individuals with non-substantial interest as shareholders, separated by the tax regime. between the two samples one day after the announcement day [+1]. The AAR 1 takes on significant 1.231% under FI and insignificant % under HI. The distributions of abnormal returns on the day after the event also differ significantly. If we calculate the cumulated average abnormal return for [-1;1] we get a significant value of 4.400% under FI and a significant result of 2.521% under HI. The distribution of the cumulated abnormal returns between the two samples is significant according to the t-value but not the z-value. The significant price reaction on the day after the publication [+1] of the ad hoc announcement under FI suggests a deferred reaction of private investors which assumedly were active on the New Market at the time of the full imputation system. The significant difference (t-value) of the distribution of the cumulated abnormal return is a result from the significant average abnormal return from the day after the announcement [+1]. Arguing with the average cumulated abnormal return [-1;1], we can validate the alternative hypothesis H 2.1. The stock price reaction under the full imputation system seems to be higher. From table 7 we can see the average abnormal return AAR t and the average cumulated abnormal return CAR( 1; 1) under the full imputation and the half income system for companies with dominating shareholders of substantial interest: On the day before the announcement [-1] we observe a significant positive price reaction of 1.211% for the FI period and an insignificant price reaction of % for 28

31 FI (N=49) HI (N=26) Difference time period CAR t value z value CAR t value z value t value z value bzw. AAR t bzw. AAR t % % % % % % bis % % Notation: t- and z-values are significant with 1.65, 1.96 and 2.58 to the 10%-( ), 5%-( ) and 1%-( ) level of significance. Table 7: Average abnormal return AAR t and average cumulated abnormal return CAR( 1; 1) for individuals with substantial interest as shareholders, separated by the tax regime. the HI period. The distributions of the abnormal returns do not differ significantly. Nevertheless, one might assume that during the FI period more insider trading took place. On the event day we find high price reactions of 5.830% under FI opposed to 5.847% under HI. The difference between the two samples is not significant. On the day after the announcement [+1] we have opposite reactions. Under FI the average abnormal return increases by insignificant 1.127% whereas under HI it decreases by significant 1.997%. Considering the average cumulated abnormal return [-1; 1] we find a significant difference which shows -according to the alternative hypothesis H 2.2 higher price reactions under FI. Tax aspects seem to play a role in this portfolio. Table 8 delivers the average abnormal return AAR t and the average cumulated abnormal return CAR( 1; 1) for companies dominated by corporations under the full imputation and the half income system. FI (N=14) HI (N=25) Difference time period CAR t value z value CAR t value z value t value z value bzw. AAR t bzw. AAR t % % % % % % bis % % Anmerkung:t- and z-values are significant with 1.65, 1.96 and 2.58 to the 10%-( ), 5%-( ) and 1%-( ) level of significance. Table 8: Average abnormal return AAR t and average cumulated abnormal return CAR( 1; 1) for corporations as shareholders, separated by the tax regime. Share buybacks have no advantages over dividends for corporations as shareholders. This is true for FI and HI. Therefore we do not expect differing price reactions between 29

32 HI and FI according to the alternative hypothesis H 2.3. Considering the results from table we can find no significant differences regarding the distribution of abnormal returns on the event day [0]. The average abnormal return AAR 0 under FI is 3.999% and under HI 4.692%. On the day prior the announcement we note a negative price reaction under FI opposed to a positive reaction of 0.327% under HI. Despite the big difference in returns there is no significant difference between the two samples in the distribution of abnormal returns. On the day after the announcement we observe positive reactions under both systems but no significant price reaction according to the z-value. The distribution of abnormal returns also does not differ significantly. The alternative hypothesis H 2.3 can be validated. Since dividend policy is irrelevant regarding the dominance of corporations there is no difference amounting to the price reaction. The significant price reaction on the event day can be attributed to non-tax aspects. Note that the small sample size reduces the significance of the results for portfolios with substantial interest and corporations drastically. Nevertheless, if we combined the two portfolios we could not validate the hypotheses which were developed on the basis of our calculations. This is why significance could only be increased and standard errors decreased by larger samples which can be obtained in the future. 4.2 Multivariate Statistics In this section a regression analysis is used to isolate the tax effects on the average cumulated abnormal return [-1; 1]. In a first step the conversion of the tax regime from the imputation system to the half income system is integrated with a dummy variable (system t ). Alternatively, in further regressions the dummy variable (system t ) is removed and replaced by the TAX-1 or TAX-2 variable. These TAX-variables are defined in section

33 As additional explanatory variables for the substitution effect the adjusted dividend per share (DP S i,t ) and the dividend yield (DivY i,t ) on the announcement day of the share buyback is used. As control variables the company size (Size i,t ) on the announcement day, a binary dummy variable, which controls for bull and bear markets (StockEx t ), and a second binary dummy variable controlling for New Market (NewM t ) companies are used. 32 The variable ɛ i is the residual. Formally, the following equation is estimated: CAR i [ 1; 1] = α + β 1 system t + β 2 DP S i + β 3 DivY i (24) +β 4 Size i + β 5 StockEx t + β 6 NewM t + ɛ i Table 9 shows the results of the estimated regression equation for the whole observation period (May, 1, 1998 to December, 31, 2006). Variable Predicted Sign Coefficient system DPS DivY Size StockEx -/ NewM Intercept N 268 R F (6.261) 5.44 Significance levels 10% 5% 1% Table 9: Regression results with the dummy variable (system t ). Annotation: Standard errors are adjusted according to White (1980). Multicollinearity is in line with the acceptable values. Table 16 in the appendix shows the Pearson-Correlation matrix for the independent variables. 32 The details of the definitions of these variables are explained in section

34 The goodness of the regression model is insufficient (R 2 =13.51%) for a clear explanation between the stock market reaction and the independent variables. Nevertheless the estimated regression has explanatory power for the tax influence on the cumulated abnormal return (CAR[-1; 1]). The coefficient of the tax system variable is significantly positive. The cumulated average abnormal return (CAR[-1; 1]) is significantly higher (2.17%) under the imputation system in comparison to the half-income system. 33 The alternative hypothesis H2, that the share price reaction should be higher under the imputation system in comparison to the half-income system, can be confirmed. The coefficient of the dividend yield (DIV Y i,t ) as well as the coefficient of the adjusted dividend per share (DP S i,t ) confirm the substitution hypothesis. If the adjusted dividends or dividend yields are decreasing, then the share price reactions are significantly higher. These results support the alternative hypotheses H3 and H4. The control variable Size is another influencing factor on the cumulated abnormal return [-1; 1]. An increasing company size leads to significantly lower share price reactions. The coefficient of the dummy variable for the bull and bear markets shows a significantly lower share price reaction (2.00%) in an increasing stock market compared to a decreasing stock market. Companies, listed at the market segment New Market and modelled by the variable NewM i,t show a 2.47% higher share price reaction in the whole observation period. Alternatively to the binary dummy variable (system t ), the conversion of the tax regime in 2002 is analysed by integrating the shareholder structure into the tax variable. The composition of the first analysed variable, Tax-1 -variable, is explained in section 3.3. The Tax-1 -variable uses the tax advantage (capital gains over dividend taxation) of the dominant shareholder group. If the shareholders are not dominated by one group, then the relative tax advantage is weighted with the fractions of the different shareholder groups on the shareholder structure. The following regression model shows the relationship between the cumulated abnormal return [-1; 1] and the independent 33 The same price reaction difference can be seen in table

35 variables. CAR i [ 1; 1] = α + β 1 T ax 1 + β 2 DP S i + β 3 DivY i (25) +β 4 Size i + β 5 StockEx t + β 6 NewM t + ɛ i Table 10 shows the results of the estimated regression equation. Variable Predicted Sign Coefficient t value T AX DPS DivY Size StockEx -/ NewM Intercept N 263 R F (6.256) Significance levels 10% 5% 1% Table 10: Results of the regresssion analysis considering the Tax-1 - variable. Annotation: Standard errors are adjusted according to White (1980). Multicollinearity is in line with the acceptable values. The R 2 of this estimated regression equation is decreasing with 1.35% in comparison to the estimated regression equation 24. The Tax-1 -variable provides a lower explanation than the (system t )-variable. The coefficient of the Tax-1 -variable is according to our forecasts with a value of positive and weakly significant at the 10% level (one-sided hypothesis). A higher tax advantage of capital gains leads to higher share price reactions on the announcement day of share buybacks. Because the Tax-1 - variable takes the tax advantage of every (dominant) shareholder group into account, the verification of the alternative hypotheses H 2.1 to H 2.3 is possible. The coefficient 33

36 of the Tax-1 -variable supports, considering the shareholder structure, the alternative hypotheses H 2.1 to H 2.3. The remaining independent variables of the regression analysis show the already predicted and explained signs. An alternative approach to integrate the relative tax advantage (capital gains over dividends) into the shareholder structure is to weight the tax advantage with the holdings of the shareholder groups. 34 For this reason the Tax-1 -variable is replaced by the Tax-2 -variable. Regression equation 26 shows the relationship. CAR i [ 1; 1] = α + β 1 T ax 2 + β 2 DP S i + β 3 DivY i (26) +β 4 Size i + β 5 StockEx t + β 6 NewM t + ɛ i Table 11 shows the results of the estimated regression equation. Variable Predicted Sign Coefficient t value tax DPS DivY Size StockEx -/ NewM Intercept N 263 R F (6.256) Significance levels 10% 5% 1% Table 11: Results of the regression analysis considering the Tax-2 -variable. Annotation: Standard errors are adjusted according to White (1980). Multicollinearity is in line with the acceptable values. The explanatory power (R 2 ) of the estimated regression equation (26) is 0.218% lower 34 See section

37 FI HI Obs. Coef. (DPS) Coef. (DivY) Obs. Coef.(DPS) Coef. (DivY) reg. (TAX-1 ) *** *** reg. (TAX-2 ) ** *** Anmerkung: t values are significant to the 10%-, 5%- or 1%- significance level. Standard errors of the regression coefficients are adjusted using White (1980). Table 12: Estimation of the coefficients of the adjusted dividend per share (DPS) and the dividend yield (DivY) separated according to the two tax regimes. than the estimated regression model (25). This implies, that the Tax-2 -variable explains the dispersion of the cumulated abnormal returns [-1;1] worse than the Tax-1 - variable. Nevertheless the coefficient of the Tax-2 -variable is positive ( ), but due to a rising standard error insignificant. This result supports, but doesn t confirm the alternative hypotheses H 2.1 to H 2.3. The coefficient of the adjusted dividend per share confirms like the two other regression models the substitution hypothesis (alternative hypothesis H 3 ). The coefficient of the dividend yield is significantly negative ( ). An increasing dividend yield leads to a smaller share price reaction. This result is also confirmed by the estimated regressions 9 and 10. This proves the dividend clientel effect. The alternative hypothesis H 4 can be confirmed. The remaining independent variables of the regression analysis show the already predicted and explained signs. To verify the alternative hypothesis H 3.1 it is necessary to divide the whole oberservation period into two subsamples. The first sub sample represents the time period under the imputation system, whereby the second subsample indicates the time period under the half income system. Table 12 displays the coefficients of the two proxy variables for tax effects (adjusted dividend per share and dividend yield) Analyzing the TAX-1 and TAX-2 -variable would be useless, because the tax factor in these two variables fluctuates minimally in the separated time periods. 35

38 Table 12 shows that the coefficients for the adjusted dividend per share (DPS) are significant under the imputation system and insignificant under the half income system. The level of the coefficients between the FI and HI is nearly indifferent. The dividend yield is higher and significant under the imputation system. Especially the significant coefficient of the adjusted dividend per share confirms a stronger substitution effect under the imputation system. The presented results confirm the alternative hypothesis H Conclusion Share buybacks can be an alternative form to dividends in the payout policy. From a tax perspective, share buybacks have a tax advantage over cash dividends. It can be shown, that the change of the tax system on January, 1, 2002 has a significant influence on the share price reaction on the announcement day of share buybacks. We see a significant higher share price reaction under the full imputation system in comparison to the half income system. This result is consistent with our predictions. Due to the tax reform 2002 the relative tax advantage of capital gains over dividends decreased. In a second step we created portfolios, consisting of shareholder groups, which are separated by individuals with non-substantial and substantial interest as well as corporations as shareholders. As a result, companies which are dominated by shareholders with substantial interest show the highest significant share price reaction and the portfolio, which is dominated by non-substantial shareholders show the lowest significant share price reaction. This result contradicts our prediction, because the tax advantage of share buybacks is the highest for non-substantial shareholders. Obviously, non-tax factors influence the price reactions stronger on the announcement day. If we divide the whole observation period into the full imputation system and the half 36

39 income system and if we retain the created portfolios, then we can show that the share price reaction for companies, which are dominated by individuals with substantial and non-substantial interest is larger under the full imputation system than under the half income system. This result is consistent with our prediction. The tax advantage for share buybacks over dividends is higher under the full imputation system. In a third step we use a regression analysis and try to separate tax factors from other factors influencing the share price reaction on the announcement day. It can be shown, that the adjusted cash dividend and the dividend yield influence the share price reaction significantly negative. Companies with high cash dividends and dividend yields show smaller share price reactions. These results support the substitution hypothesis and the dividend clientel effect. An open question is the share price performance of the company after the announcement day of a share buyback. With this analysis we could figure out if the share buyback is a signal for higher earnings of if the share buyback is used to reduce liquidity. Further, we do not know how many share buyback announcements are published without any actual share buyback. Game theory could give explanatory power in this field of research. 37

40 Appendix Figure 6: History of the CDAX. Year System Dividends tax rate Capital gains tax rate Difference τ d (%) τ g (%) (%) 1998 FI FI FI FI HI HI HI HI HI Table 13: Effective tax rates on dividends and capital gains for corporate shareholders FI: Full imputation system, HI: Half-income system. 38