The relationship between ownership and dividend
policy. Evidence from Germany.
Abstract
This study examines whether large shareholders have any influence on dividend policy in Germany. The potential influence is tested in four different setups and deals with the questions whether the size of the voting power of large shareholders influences the decision on whether or not to pay dividends at all, whether they influence the amount of dividends paid, and whether different shareholder types have different influences regarding the amount of dividends paid, and finally, whether shareholder types and their voting power influence the amount of dividends. The empirical results for 2005 to 2008 support Goergen et al. (2005) who finds no evidence for an influence on dividend policy of the largest shareholder in Germany for a 1984 to 1993 sample. With regard to the shareholder types, banks, financial institutions, companies, and families have a negative influence on dividends. When voting power is included, financial institutions and state authorities increase dividends, and family-owned Beteiligungsgesellschaften in some cases appear to have a slightly more positive influence on the amount of dividends paid than direct holdings of families and private blockholders.
JEL Classification: G32, G35
Key words: Dividend policy, Ownership structure, Germany, Beteiligungsgesellschaften
Matthias Smit S1544640
July 2009
Table of contents
1. Introduction 2. Literature Review
1. Ownership and payout policy
2. Dividend preferences of different shareholder types 1. Banks
2. State authorities
3. Families and private blockholders 3. Data
1. Sample and data sources
2. Transparency of ownership in Germany and limitations of the data 1. Regulation
2. The notification process 3. Variables
4. Descriptive statistics 4. Methodology
5. Empirical results 6. Conclusion
1. Suggestions for further research 7. Appendix
1. List of companies included in the sample 2. List of companies excluded from the sample 3. Detailed description of variables
4. Details of the 2001 change of the tax code
1. Introduction
This thesis examines how large shareholders influence the dividend policy of firms. In detail, it aims to clarify whether firms that have large shareholders pay on average more cash dividends than firms that have a dispersed ownership. To analyze this topic, data on German companies that are listed on the major indices DAX, MDAX, SDAX, TecDax, and GEX for the years 2005 till 2008 are used in this paper.
The focus of this research lies on the preferences of dividend policy of large shareholders. In this context, this study tries to answer four main questions. First, whether the relative size of the voting rights of the large shareholders influences the company’s decision whether or not to pay dividends. The second question is whether size of the voting rights of the large shareholders influences the relative amount of dividends paid to shareholders. Third, whether different types of the largest shareholders influence the amount of dividends paid, and, fourth, whether the size of the voting right and the type of the largest shareholder together influence the amount of dividends paid. In order to answer these questions it is necessary to further clarify the term ‘large shareholder’. In general, these are shareholders that own a significant stake in a firm. Those shareholders can be grouped into several categories, namely banks, financial institutions1, companies, state authorities, so-called Beteiligungsgesellschaften2, and families and private blockholders. Another feature that might be
important is whether these are German or foreign investors. Recent literature has shown that these different types of shareholders also have different preferences regarding dividend policies (these are explained in section 2.3). This paper will follow the methodology of Mancinelli and Ozkan (2006) and employ Probit and Tobit regression analysis to reveal the influence of ownership on cash dividends. This thesis focuses entirely on the “classical” way to distribute cash to shareholders, i.e. cash dividends, and leaves aside share repurchases. Von Eije and Megginson (2008) find that in a sample of 15 European countries the real dividends paid increased from 1989 to 2005. Thus cash dividends play an important role in the business world today despite advances in other cash distribution methods. This paper adds value to the existent literature in several ways. First, it reveals the influence of large shareholders on dividend policy on a sample of German firms using very recent data. Financial data from 2005 to 2008 is combined with ownership data from 2009 here. Prior to this period, the corporate tax system has changed completely3, compared e.g. to the earlier paper of Gugler and Yurtoglu (2003).
During their 1992-1998 sample period, dividends that were received by companies were taxed at much higher rates than after the tax change of 2001. Related to this change in the tax system, another contribution of this paper is the analysis of the dividend preference of Beteiligungsgesellschaften that are used amongst others by families and other private blockholders for indirect shareholdings. Due to
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
1 These are e.g. mutual funds, pension funds, investment companies, and insurance companies. 2 Those will be explained in detail in section 2.2.4.
the sample period, this study is able to partly capture also the effects of the current financial crisis. During the sample period, the German DAX started at 5,408.26 points into the year 2006, reached its highest closing price of 8,105.69 on July 16, 2007, and then fell to a closing price of 4,810.20 on December 30, 2008. Hence, half of the sample period used in this thesis covers the recent financial crisis. In addition, the results of this paper are well comparable to the studies of Short et al. (2002) for the UK and Mancinelli & Ozkan (2006) for Italy. Finally, since most of the existing literature covers the US, my thesis adds value in its field as it broadens the knowledge about dividend policies in Germany and, hence, continental Europe.
The remainder of this section will outline the background of the German economic system. The German economy is different from that of the United States or the United Kingdom in some areas. One important difference that also matters for dividend policies is the role of capital markets for financing. One has to distinguish between outsider (investors are at arm’s length) and insider systems (close relationship with investors), both of which are explained well in Leuz and Wüstemann (2003) and summarized here.
important feature of the German corporate governance system because it provides barriers to entry and therefore reduces the threat of competition for control from outside (Rajan and Zingales, 1998). This view is supported by Da Silva et al. (2004:132) who report that the control patterns in Germany have not changed much in the 1980s and 1990s. Legal changes in 1994 (Securities Act) and 1998 (Corporate Control and Transparency Act), however, indicate that the German financial system is transforming towards an arm’s-length system, according to Leuz and Wüstemann (2003). They reason that these changes were necessary because the reunified Germany had an immense refinancing demand and, thus, needed to adapt itself to international capital markets which have a demand for reliable public information.
These differences in the economic systems between the US and UK on the one hand and Germany on the other hand may have implications for the dividend policies, too. Lease et al. (2000) report that shareholders of German firms earn a lower dividend yield than those from the USA or UK. La Porta et al. (2000) find in a cross-sectional study on 4,103 listed firms from thirty-three countries that in civil law countries, like Germany, firms offer on average a lower degree of minority shareholder protection while they at the same time pay lower dividends than firms in common law countries. Da Silva et al. (2004), however, take a closer look at the data set of La Porta et al. (2000) and show that this pattern might not be as obvious as the authors indicate. One finding of their review is that, e.g., Germany as a civil law country has dividend payouts (measured by dividends over earnings) that are higher than the median payouts of the common law countries. Hence, Germany might be an exception from the group of civil law countries when it comes to the amount of dividends paid. This might also reflect upon the dividend preferences of large shareholders in Germany. This paper will show whether those dividend preferences are more similar to those in the UK (see Short et al., 2002), which is a common law country, or to those in Italy which is a civil law country (see Mancinelli and Ozkan, 2006).
In Germany, dividends are normally declared once a year (Da Silva et al., 2004:73). According to paragraph 58 AktG, management and the supervisory board make a proposition on how to use the annual profit, including a proposition on the dividend. The final decision on whether to accept this proposition or to decide for another way of allocating the profit is made by the voting shareholders in the annual general meeting. Da Silva et al. (2004) observe some facts about the actual dividend behavior of German firms. First, German firms often practice dividend smoothing, i.e. they do not so often change the dividend per share, irrespective of whether an increase or a decrease in the earnings per share or in the cash flow per share occurs. They report that after two consecutive years of increased profitability only 32 percent of German firms increase their dividends, whereas 75 percent of US-based firms do the same. Second, in 1991-92 approximately 25 percent of all German firms did not pay any dividends at all, whereas only 10-12 percent of UK firms abstain from paying dividends. Third, regarding the size of a firm there appears to be a U-shaped curve between firm size and dividend payout. The largest and the smallest quintiles exhibit below average payout ratios.
2. Literature Review
This literature review consists of two parts. The first one deals with the ownership structure of companies in conjunction with dividend policy. It outlines findings on how ownership structure changes after dividend events like initiations, omissions and changes of dividend size and whether large shareholders prefer different levels of dividend payout ratios. The second part concentrates on the preference of certain shareholder types concerning dividend policy. The three main types of shareholders discussed are banks, state authorities, and families and private blockholders.
2.1 Ownership and payout policy
Michaely et al. (1995) investigate market reactions to initiations and omissions of dividends using data of all companies on the New York Stock Exchange (NYSE) and the American Stock Exchange (AMEX) that initiated dividends between 1964 and 1988. They investigate whether the share of institutional ownership changes after a dividend omission4. They use data of the share of institutional ownership three years before the omissions and three years after the omission. With respect to ownership, they examine whether dividend omissions lead to a significant change in ownership but find that this is not the case. The average institutional ownership over all firms in their sample is 30.0 percent (18.1 percent standard deviation) before the omission and 30.9 percent (17.6 percent standard deviation) after the omission. Michaely et al. (1995) conclude that this adds support to the impression that dividend omissions do not produce dramatic shifts in ownership.
Crutchley et al. (1999) investigate the simultaneity of the financial variables leverage, dividends, insider ownership and institutional ownership and the link to agency cost. They use a three-stage least squares regression with four regression equations, i.e. one for each of the four variables. Their data consist of New York Stock Exchange and American Stock Exchange listed companies for the periods 1984 to 1987 and 1990 to 1993. In the first sample period, institutional ownership is found to be determined simultaneously with dividends5 while these two variables have positive impacts on each
other. Interestingly, though, the authors find a different result in the second sample period where dividends are negatively determined by institutional ownership while the latter is determined positively by the former. Crutchley et al. (1999) suggested that this change in the influence of institutional ownership on dividends may be caused by the fact that many institutional investors had become active monitors in the second sample period.
Dhaliwal et al. (1999) deal with the theory of tax clienteles for dividend policies and examine changes in the ownership of the equity of firms that initiate dividends and expect that corporate investors that
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
4 They also tried to analyze the change in institutional ownership before and after dividend initiations, but failed
to collect a large enough sample.
are tax-exempt/tax-deferred and for whom dividends are not tax disadvantaged will start buying shares of companies that initiated a dividend. Aggregate institutional ownership is used in that study as a proxy for ownership by tax-exempt and tax-deferred corporate investors. Dhaliwal et al. (1999) use a sample of 203 New York Stock Exchange (NYSE) / American Stock Exchange (ASE) / National Association of Securities Dealers Automated Quotation System (NASDAQ) listed companies that initiated or reinitiated6 a dividend payment between 1982 and 1995. Their finding is within their
expectation, namely that institutional ownership increases statistically and economically significantly after the initiation of the dividend payment.
Short et al. (2002) deal with the link between dividend policy and institutional ownership in the UK. In contrast to some papers mentioned above, they do not analyze initiations, changes, and omissions of dividends, but focus on the level of dividends. Their analysis is based on a sample of 211 firms that are listed on the London Stock Exchange and uses data from 1988 to 1992. The methodology of that paper uses four different dividend models, namely the Full Adjustment Model, the Partial Adjustment Model (Lintner, 1956), the Waud Model (1966) and the Earnings Trend Model (Fama and Babiak, 1968). For all these models they find strong evidence of a positive association between dividend payout policy and institutional ownership.
Gugler and Yurtoglu (2003) deal with dividends as informative signals that indicate the severity of conflict between large, controlling owners and small, outside shareholders. They analyze 736 dividend announcements of 266 companies in Germany between 1992 and 1998 and find that larger holdings of the largest owner reduce the dividend payout ratio while larger holdings of the second largest owner increase this ratio. Their study also reveals that substantial deviations from the one-share-one-vote rule also reduce the dividend payout. This is measured by the cash-flow-right-to-voting-right ratio of the largest ultimate shareholder and the result indicates that the smaller this ratio is the larger becomes the incentive of the large and controlling shareholder to seek compensation other than through pro-rata dividends. In line with this result, Gugler and Yurtoglu (2003) also find that the larger the cash flow right of the controlling shareholder is, the larger are the dividends that are granted to the rest of the shareholders.
Grinstein and Michaely (2005) examine the relationship between institutional holdings and payout policy in U.S. public firms. From corporate theory, they name three reasons why ownership structure and payout policy may be related. First, agency theories suggest that with better monitoring managers are likely to share more profit with investors. They refer to Jensen (1986) who states that with enhanced monitoring, firms are more likely to pay out their free cash flow. Grinstein and Michaely (2005) further state that building on the assumption that institutions are better monitors, agency theories imply that larger institutional shareholdings will increase payouts. The second reason that is
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
given by Grinstein and Michaely (2005) is based on Allen et al. (2000) who argue that firms want institutions to facilitate takeovers or to monitor in order to increase value. Due to common institutional charter, prudent-man rule restrictions and comparative tax advantages that some institutions have for dividends, institutional investors prefer higher dividends, and, hence, Grinstein and Michaely (2005) summarize that this second reason implies that higher dividends will lead to larger institutional holdings. Finally, their third reason deals with adverse selection problems that might lead uninformed investors to prefer dividend over repurchases (Barclay and Smith (1988), Brennan and Thakor (1990)). Grinstein and Michaely (2005) reason that since institutional investors are informed investors they might not face these problems. Furthermore, they might simply prefer the least costly payout policy, i.e. share repurchases. In their empirical analysis, Grinstein and Michaely (2005) obtain three major results. First, they find clear evidence that institutions prefer dividend-paying over non-dividend-paying stocks. Their second finding is that institutions do not show a preference for firms that pay higher dividends. Even more so, they find evidence that institutions prefer low-dividend stocks to high-dividend stocks. Consequently, they show that firms who increase their dividends do not attract more institutional investors. Third, Grinstein and Michaely (2005) find that institutions prefer firms that repurchase shares as opposed to firms that do not repurchase their shares. With share repurchases however, they find that firms that repurchase more as well as firms that repurchase regularly have higher institutional ownership. Finally, the authors report that they detect a trend for dividend paying and share repurchasing firms. Before the mid-1980s7 institutions preferred firms that paid more
dividends. However, after that time institutions show an aversion against high dividends and a preference for repurchasing firms.
Goergen et al. (2005) examine the flexibility and downward flexibility of dividends in Germany during the period from 1984 to 1993. In this context they also investigate whether ownership and control have an impact on dividend policy. They find that the presence of a large shareholder has no impact on the dividend decision.
Lee et al. (2006) investigate the changes in institutional ownership following changes in dividend policy for a Taiwanese sample8. They find strong evidence of a clientele effect for investors with
different tax brackets, e.g. private investors and institutional investors. In line with Michaely et al. (1995), they find that institutions as a whole show an insignificant response to dividend changes. When regarded separately, only foreign institutions show a positive and significant reaction to dividend increases. Lee et al. (2006) further find that institutions in Taiwan prefer higher dividends and higher payout ratios while they dislike firms that pay no dividends.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
7 This changed when the SEC rule 10b-18 was enacted in the mid-1980s which enabled firms to freely
repurchase their own stock (Grinstein and Michaely, 2005).
Graham et al. (2006) examine a sample of NYSE firms that make an initial dividend announcement between 1990 and 1998 and monitor the market reactions to these announcements. They divide their sample between firms that announce anticipated quarterly dividends on a regular basis and those firms that initiate unanticipated dividend payments. They find that compared to the unanticipated dividend initiating firms, the firms with anticipated announcements have a large number of institutional traders prior to the news release. Furthermore, Graham et al. (2006) find that after firms unexpectedly announced the initiation of dividend payments, institutional ownership rose from 32.2 percent of shares outstanding at the end of the quarter before the dividend initiation announcement to 50.6 percent at the end of the quarter after the announcement. In general, their results indicate that the trading volume increased for dividend-initiating firms after the announcement of dividend initiation was made public. They state that this is consistent with institutional investors increasing their ownership in those dividend-initiating firms. Contrary to a common assumption in the literature that institutional investors are the most informed investors, Graham et al. (2006) do not find evidence for this hypothesis. They rather find evidence of increased trading prior to unanticipated dividend announcements for stocks with low institutional ownership.
Similarly to the study of Short et al. (2002), Mancinelli and Ozkan (2006) conduct a study of the relationship between dividend policy and ownership structure for an Italian sample9. This study uses
2001 data on a sample that consists of 139 listed Italian firms. They use a methodology based on tobit regression and find that the voting right of the largest shareholder has a significantly negative impact on the dividend payout ratio. This contradicts the finding of Short et al. (2002) who find exactly the opposite for their UK sample. Mancinelli and Ozkan (2006) interpret their result as a support for their prediction that a higher level of concentration of ownership when measured in the voting rights of the largest shareholder leads to a higher probability of expropriation of minority shareholders. One possibility of how expropriation can occur in this context is called tunneling. In this approach, the large shareholders do not only have a majority stake in the firm, but also possess a large portion or complete ownership of the cash flow rights of other businesses that have relationships with the firm (Shleifer and Vishny, 1997). In this way, they are able to expropriate corporate wealth by setting unfair terms for within-group transfers of assets and control stakes or by sales of goods and services (Faccio et al., 2000). Hence, according to Mancinelli and Ozkan (2006), Italy seems to be a good example for expropriation of minority shareholders through tunneling. A second interpretation that they give is the extent to which top managers are related to the largest shareholder. They find support for the prediction that managers rather hold resources under their own control than distribute returns to
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
9 Mancinelli and Ozkan (2006) state that the ownership structure in Italy is highly concentrated and that this
shareholders and reason that this seems likely as in 70 percent10 of Italian firms the top managers come
from the largest shareholder.
Table 1: Summary of literature on the link between dividend policy and ownership
Paper Sample period Country Results
Michaely et al. (1995) 1964 – 1988 USA Dividend omissions do not change the ownership structure significantly.
Crutchley et al. (1999) 1984 – 1993 (in detail: 1984 -1987 and
1990 – 1993)
USA Institutional ownership and dividend policy are positively related (1984-1987). However, dividends are negatively related to institutional ownership (1990-1993) while institutional ownership is positively determined by dividends (1990-1993). Dhaliwal et al. (1999) 1982 – 1995 USA Dividend initiation leads to increased
institutional ownership.
Short et al. (2002) 1988 – 1992 UK Positive association between dividend payout policy and institutional ownership. Gugler and Yurtoglu
(2003) 1992 – 1998 Germany Larger holdings of largest shareholder reduce dividend payout, while larger holdings of second largest shareholder increase dividend payout. Substantial deviations from one-share-one-vote rule decrease dividend payout.
Grinstein and Michaely (2005)
1980 – 1996 USA Institutional investors prefer dividend-paying over non-dividend-paying stocks, and they also prefer low dividends to high dividends. Institutional investors prefer share repurchases to dividends.
Goergen et al. (2005) 1984 – 1993 Germany Find that the presence of a large shareholder does not have an influence on the dividend decision.
Lee et al. (2005) 1995 -1999 Taiwan Institutional investors prefer higher dividends and payout ratios while they show an insignificant response to dividend changes.
Graham et al. (2006) 1990 – 1998 USA Dividend paying firms have a high percentage and large number of institutional traders prior to an anticipated dividend announcement. Institutional investors increase their stake in firms after the initiation of dividend payments.
Mancinelli and Ozkan
(2006) 2001 Italy The voting right of the largest shareholder has a significantly negative impact on dividend payout.
2.2 Dividend preferences of different shareholder types
In the literature there is evidence that different types of large shareholders show different preferences for dividend policies for various reasons. Those shareholders are e.g. banks, financial institutions,
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
companies, state authorities, and families and private blockholders. One important variable in this context might also be whether the shareholder is of domestic or foreign origin. The following sections will summarize the empirical findings that can be found in the literature for the dividend preferences of banks, state authorities and families.
2.2.1 Banks
Banks play an important role in the German economy. Goergen et al. (2005) state that the German corporate governance system is sometimes called a ‘bankbased corporate governance system’ that is characterized by strong relationship banking. Da Silva et al. (2004:51) present three reasons as to why German banks play such an important role in corporate governance. First, through equity participation in industrial and commercial companies. Second, and due to the fact that most shares in Germany are bearer shares that many shareholders deposit with their bank, these banks have control over a considerable proportion of voting equity of large corporations at the annual general meeting, especially through proxy voting. Third, banks are well represented on the supervisory board of many corporations in Germany. Thus, banks are able to influence the outcome of the voting on the annual general meeting significantly. There is, however, a trend towards lower direct shareholdings by banks. Da Silva et al. (2004) show this for a sample of 221 German industrial and commercial quoted firms. They used the years 1984, 1989, and 1993 to analyze the first-tier and ultimate level ownership different investor types’ shareholdings exceeding 25 percent and 50 percent11. The result of this study
is that the percentage of banks within all shareholders that had a first-tier ownership in a company of at least 25 percent (50 percent) dropped from 12.1 percent (2.7 percent) in 1984 to 7.7 percent (2.4 percent) in 1993. For ultimate control, the figures dropped from 15.9 percent (5.5 percent) in 1984 to 10.1 percent (3.8 percent) in 1993. This development reflects the slight trend of the German economy away from the huge amount of cross-holdings among firms and towards a more open and foreign investor-friendly climate (Da Silva et al., 2004). Nonetheless, like France, Italy, and Japan, Germany is still regarded as a traditionally bank-dominated economy (Da Silva et al., 2004:51)12.
Regarding the dividend preference of banks as shareholders, Da Silva et al. (2004) further show that bank-controlled firms have significantly lower dividend payout ratios than firms controlled by dispersed ownership, or families. They show that bank control has a strongly negative impact on the dividend payout ratio. Amihud and Murgia (1997) point into the same direction when they conclude that the low dividend payout level in Germany might be due to the fact that banks are often in control of majority voting stakes in shareholders’ meetings. They reason that banks who are a major shareholder and lender of a company at the same time favor low dividend payouts in order to provide
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
greater security for their debt. In contrast to Da Silva et al. (2004), Gugler (2003) finds target payout ratios13 of bank-controlled firms to be larger than those of family-controlled companies. Goergen et al.
(2005) find in addition, that if banks are in control this increases the likelihood of dividend omissions in the wake of earnings losses, mitigates agency costs and reduce the need for dividends as a monitoring device. Furthermore, Gugler (2003) finds that in bank-controlled corporations lagged dividends determine this year’s dividend only marginally and that this might reflect the compensating incentives of holding debt and equity of those firms at the same time.
2.2.2 State authorities
Gugler (2003) reasons that in state-controlled firms there exists a double principal-agent problem, namely between the managers of the firm and the elected representatives of the government on the one hand, and between the elected representatives of the government and the citizens on the other hand. Gugler (2003:1301) assumes that “large numbers lead citizens to shirk on their monitoring role of
politicians, and thus the politicians themselves may not actively monitor the companies the state owns.” Therefore, higher dividends are needed as a substitution for monitoring in state-controlled
firms. Gugler (2003) finds in his sample of Austrian companies that state-controlled firms have the highest target payout ratio compared to firms controlled by banks, families and foreign owners. In addition, state-controlled companies show significant dividend smoothing. His explanation is that the elected politicians who are held accountable for all activities of government can be expected to have a strong interest in a steady flow of dividends. First, dividends may be able to convince citizens that the government successfully controls the company. Second, the steady flow of dividends will reduce the free cash flow in the hands of the managers and, hence, reduce the agency problems that can result from this free cash flow. Gugler (2003) computes the target payout ratios for state-controlled firms and finds them to be 42.90 percent (33.30 percent for firms doing R&D). In addition, he also looks at the dividend flexibility of firms. He finds that the ranking in flexibility to cut dividends is ‘family’ > ‘foreign firm’ > ‘bank’ > ‘state’ control. This, again, is consistent with the findings above where state controlled firms have a strong preference for stable dividends and families are not reliant on dividends as a means of reducing agency cost. In contrast to Gugler’s (2003) findings, Lee et al. (2006) find for their Taiwanese dataset that government entities prefer lower positive dividends and payout ratios. A reason for this is not mentioned by Lee et al. (2006). They only state that during the sample period some of the companies in the sample were in the process of privatization while the government was still owning shares and assume that it would be unlikely for a government to sell low dividend paying shares and buy high dividend paying shares just for the sake of tax-free dividends.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
13 These target payout ratios are measured as ! = "! / (1 - (1 - ")), where ! symbolizes the target payout ratio and
2.2.3 Families
Da Silva et al. (2004:129) also comment on the dividend payout ratios for firms that are family-controlled in Germany. They find those firms to have the largest dividend payout ratio (24.22 percent) compared to company-controlled (18.50 percent), bank-controlled (16.60 percent) and widely-held firms (19.03 percent). Interestingly, though, their analysis reveals that control by families does not seem to have an impact on dividend policy. Gugler (2003) finds contradictory evidence regarding the target payout ratios of family-controlled firms in Austria, i.e. that those firms choose significantly lower target payout ratios (25.00 percent) when compared to bank-controlled firms (34.40 percent) 14.
Gugler and Yurtoglu (2003) do not find a linear relationship between dividend payout ratios and family ownership for a sample of German firms. For family ownership, their analysis reveals a U-shaped relationship between the dividend payout ratio and the cash flow rights of the largest shareholder (with a minimum at 40.20 percent) and an inverted U-shaped relationship between the dividend payout ratio and the voting rights of the largest shareholder (maximum at 47.80 percent). These shapes are consistent with the patterns that were found for the whole sample in Gugler and Yurtoglu (2003).
Table 2: Summary of literature on different shareholder types
Paper Sample size
and period Shareholder type Country Dividend preference (target payout ratio in percent)
Gugler and Yurtoglu
(2003) companies. 266 1992-1998
Families Germany U-shaped relationship between dividend payout ratio and cash flow rights (minimum at 40.20) Gugler (2003) 214
non-financial firms. 1991-1999
State Austria 42.90
(33.30 for firms doing R&D)
Banks Austria 34.40
(19.10 for firms doing R&D)
Families Austria 25.00
(13.00 for firms doing R&D)
Foreign firms Austria 34.70
(13.90 for firms doing R&D) Da Silva et al. (2004) 416 German
industrial and commercial companies. 1993 Families Germany 24.22 Companies Germany 18.50 Banks Germany 16.60
Lee et al. (2006) 1,357 trading days. 1995-1999
State Taiwan Low (below sample median payout ratio)
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
14 The target payout ratios for firms that are active in R&D were found to be 13.00 percent for family-controlled
2.2.4 Other shareholder types
For the remaining shareholder types that are included in this research no empirical evidence was found in the literature. Thus this paragraph presents the author’s expectations regarding their effect on dividends.
First, financial institutions are expected to show a preference for lower dividends, as banks do. One reason therefore is that in four cases out of 35 in the sample of this study, the financial institution that is a largest shareholder is owned by a bank itself. There, the financial institution is the investment trust company of the bank. Hence, the dividend preference would be similar. A second reason might be that financial institutions are able to thoroughly assess a company they want to invest in. Investment professional have the knowledge to analyze the company and enough experience on how to exercise control effectively. This might be enough for them to monitor the management closely, so that high dividends as a substitute for monitoring is not needed in this case.
Second, companies that invest in other firms often do this for strategic reasons. This means, that the buying company has an interest in a cooperation with the firm it partly bought. In those cases, one can expect that the management team of the buying company has acquired plenty of information about the target company. In addition, close relationships between both management teams might also develop during the preparation of the acquisition. Hence, for the same reasons that are valid for banks and financial companies, companies are expected to have a negative influence on dividends when compared to widely-held ownership.
Third, the Beteiligungsgesellschaften are of special interest for this study. In Germany, those firms are
often called Beteiligungsgesellschaften, Vermögensverwaltungsgesellschaften,
Verwaltungsgesellschaften, or Holding. The reason for the special interest is a change in the corporate
tax policy in Germany in 2001 that determines the expectations of the dividend preference of these companies. Detailed information about the change of the tax regulation is outlined in the appendix, section 7.4. Before this policy change, dividends that were paid from German companies (Kapitalgesellschaften, such as AG and KGaA) to other companies (also Kapitalgesellschaften) that owned shares of the former were fully liable to taxation. The alteration of the paragraph 8b of the Corporate Tax Act (Körperschaftsteuergesetz) in the course of the tax reform 2001 brought a radical change. From then on until 2003, dividends that were exchanged between companies were completely tax-free. From 2004 on until today, 95 percent of those dividends are still tax-free and only 5 percent are liable to taxation. After these changes in tax policy, the possibilities to save taxes rose for companies. In addition, wealthy families or other private blockholders who set up a
Beteiligungsgesellschaft to which they transferred their shareholdings in other firms could also profit
that while direct holdings of families and private blockholders will still have a negative influence on dividends, their indirect holdings through Beteiligungsgesellschaften could have a positive impact on dividends.
3. Data
This section explains the data that are used in this study. It will start with the sample itself, the data sources and explain which companies are included and which are not. In addition, section 3.2 will provide an insight into the German law and institutions concerning the transparency of ownership and explain how the notification system actually works. Section 3.3 describes the variables used in this study. Finally, section 3.4 will present some descriptive statistics.
3.1 Sample
As mentioned in the introduction, the sample of this study comprises all German companies that were listed in the German stock indices DAX, MDAX, SDAX, TecDax, and GEX on February 19, 2009. From these 221 companies 30 belonged to the DAX, 50 to the MDAX, also 50 to the SDAX, 30 to the TecDAX and 83 to the GEX15. The stocks of those firms are all traded within the so-called Prime
Standard segment of the Deutsche Börse AG and have to fulfill the highest transparency requirements on the German stock market16. Due to the different structure of their financial reports, 6 banks, 3
insurance companies and one other financial institution were deleted from the sample. In addition, another 51 companies were excluded because they were no German companies or there were not enough data available. Finally, due to statistical reasons, four other firms were excluded as they were found to be extreme outliers within the sample17. The remaining sample of 156 companies can be
subdivided into seven different industries. 42 firms belong to the manufacturing, petroleum and technology sector, 20 other represent the electronics, software, and IT business. From the utilities, telecommunication and energy sector 13 companies are included in the sample and 17 companies are active in the chemical and agricultural sectors. 32 companies belong to the service and real estate industries, 17 are active in food or health care products, while the remaining 15 can be found in the consumer goods or other businesses.
3.2 Transparency of ownership in Germany and limitations of the data 3.2.1 Regulation
In this study, the most important variable in the analysis of dividend determination is the ownership variable, i.e. the relative amount of shares that the large shareholders of a company possess. To be able to acquire these data there are however some difficulties which will be highlighted in this section.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
15 22 out of those 83 were listed in the GEX and in one of the other indices simultaneously.
16 Among those transparency requirements are e.g. quarterly published company reports. Hence, the inclusion of
the earnings reporting frequency as a control variable does not make sense in this study.
It is said to be a German tradition to preserve the incumbents’ power by emanating as little information about the firm’s activities as possible (Becht and Boehmer, 1999). Nevertheless, in Germany the judicative basis for the publication of stockholdings that exceed certain thresholds of stock ownership in a firm is laid by the law named WpHG (Wertpapierhandelsgesetz). This law was passed in July 1994 and is the German interpretation of the guidelines 88/627/EWG and 89/592/EWG of the council of the European Community. The former deals with the information that has to be published due to acquisitions and dispositions of significant investments in exchange-listed companies. The latter deals with the coordination regarding the rules on insider trading. Despite the thirteen sections and 46 paragraphs of this law (details are discussed below), the efficacy of disclosure regulation is very low (Becht and Boehmer, 1999). Within the context of the “Zweites
Finanzmarktförderungsgesetz”, not only the WpHG was established, but also the Federal Securities
Supervisory Office, that is known as the “Bundesanstalt für Finanzdienstleistungsaufsicht” (BaFin; http://www.bafin.de) today18.
In practice, there are different regulations that require the publication of ownership data in Germany. As this paper only includes stock corporations, the so-called Aktiengesellschaften (AG) or
Kommanditgesellschaft auf Aktien (KGaA), only the applicable regulations will be dealt with here.
Companies that are AGs and KGaAs are not required to reveal the identity of owners who are individuals in its annual report. In contrast, any company must include in its annual report whether it owns more than 25 percent in another AG or KGaA itself (Becht and Boehmer, 1999). Furthermore, AGs and KGaAs are obliged to include their own ownership in other firms if they exceed 20 percent (§271 Handelsgesetzbuch (HGB)). Apart from the annual report, company statutes and contractual arrangements provide valuable ownership and control information because German companies frequently enter contracts where one company (controlled firm) fully surrenders claims on profits or decisions to another, controlling firm (Becht and Boehmer, 1999). The most frequent examples of those contracts are Profit and Loss Agreements (Gewinnabführungsvertrag) and Subordination of Management Agreement (Beherrschungsvertrag). The former implies a transfer of both profits and losses to the controlling company, the latter assigns the complete managerial control rights to the controlling company (Goergen et al., 2004; §291 AktG). If a company entered such a contract this gives a clear picture about the shareholder that is actually controlling the company. As this study deals with the largest exchange-listed companies in Germany, however, those kinds of contracts will probably be of less relevance because the firms in this sample will rather control other firms than be controlled by others through those kinds of agreements.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
18 In 1994, this entity was established as the Bundesaufsichtsamt für den Wertpapierhandel (BAWe). This
institution was merged together with the “Bundesaufsichtsamt für das Kreditwesen” and the
“Bundesaufsichts-amt für das Versicherungswesen” into the “Bundesanstalt für Finanzdienstleistungsaufsicht” on May 1st, 2002
when the Financial Services Supervisory Law “Finanzdienstleistungsaufsichtsgesetzes” (FinDAG) was passed. (Becht and Boehmer, 1999)
The sources from which primary ownership data can be obtained are company registers, the
Bundesanzeiger (http://www.bundesanzeiger.de), the BaFin and the Federal Antitrust Office
(Bundeskartellamt) (Becht and Boehmer, 1999). The company registers are managed by local courts and contain all filed documents from the businesses within the district of the court. Those documents comprise the statutes, charters, and annual reports. To gain access to their information, there are, however, four obstacles that have to be overcome (Becht and Boehmer, 1999): First, one must travel to the potentially distant court in whose district the business is located. Second, the courts mostly only have the most recent documents on site due to the storage space they have in their court buildings. It might thus become necessary to request older documents days or weeks in advance to have them on site when needed. Third, since the documents are sorted by company but not by type of document, it may require some time and manual search to find the required information. Finally, and most serious, most companies violate the law by not furnishing mandatory filings. Hansen (1996:56) assumes that the law is broken in this way by two-third of all German companies. Understaffed courts and fees of maximum EUR 5,000 are not able to stop or even sanction those business practices (Becht and Boehmer, 1999). To overcome those obstacles, the German government provides an online portal (http://www.unternehmensregister.de) where all publications from the electronic version of the
Bundesanzeiger, all filings and documents from the company registers and important company
information of securities-emitting companies can be obtained online. However, despite this appealing online platform, the final problem mentioned by Becht and Boehmer (1999), the incompleteness of the filings, persists.
The BaFin publishes all filings that are submitted to it under §21 to §29 WpHG on their website and makes it possible to search for all filings that were submitted per company. According to §21 WpHG anyone who reaches, transgresses or falls below 3%, 5%, 10%, 15%, 20%, 25%, 30%, 50%, or 75% of ownership of voting rights of a German company has to notify this company and the BaFin. From these filings, it is possible to obtain a good – yet not perfectly precise – impression of the current ownership structure of an AG or KGaA for all shareholders who possess at least 3% of the voting rights.
3.2.2 The notification process19
As mentioned above, if a shareholder exceeds, reaches or falls below the thresholds of stock ownership stated by the WpHG, a notification is required. This notification has to be sent to the company that issued the shares and the BaFin simultaneously within four trading days. The issuing company then has to publish this information without any delay but at least within three trading days. After this publication, the information has to be sent to the company register for storage. Hence, the maximum time allowed between the actual trade and its publication are seven trading days. Becht and Boehmer (1999) comment that before reaching the market, this information will have passed through many hands and, if it is price relevant, could easily be used for illegal insider trades. They further note that the German filing system appears antiquated, costly, and prone to errors compared to the filing systems in the United States (Electronic Data Gathering Analysis and Retrieval, EDGAR) and Canada (System for Electronic Document Analysis and Retrieval, SEDAR).
3.3 Data sources and Variables20
As mentioned before, this study follows the methodology of Mancinelli and Ozkan (2006) but it will also extend it in some cases. The main feature that distinguishes their methodology from many other papers is the fact that the ownership variable is collected for only one point in time. This study follows that approach, too, since it was not possible to obtain historic ownership data.
The dependent variable in this study is the amount of cash dividends paid. In Mancinelli and Ozkan (2006), the cash dividends are expressed in two ratios, namely cash dividends/earnings and cash
dividends/market capitalization. As a reason for those two ratios they state that although the cash dividends/earnings ratio is the most commonly used measure of dividend payouts, it might be subject
to earnings manipulation due to accounting practices. For that reason, the cash dividends/market
capitalization ratio is used to provide more robustness to the findings (Mancinelli and Ozkan, 2006).
This thesis, however, does not use the cash dividends/earnings ratio because there are some observations in the sample for which the earnings figure is negative. For that reason, cash
dividends/revenues is taken as the first dependent variable and, analogue to Mancinelli and Ozkan
(2006), cash dividends/market capitalization is used as a second dependent variable. For the construction of these variables, the financial data are averaged over three years. Data on cash dividends are obtained for the years 2005 till 2008, but only the years 2006 until 2008 are used here. According to the description of the cash dividend variable in Datastream, the amount that is given for e.g. 2008 shows the amount that the company paid out in 2008. Hence, this profit was earned in the year before and, accordingly, the cash dividend component of the two ratios is calculated as a mean of
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
19 This process is defined in the §§ 21 to 29a of the WpHG.
the years 2006 to 2008 while the revenue and market capitalization components were averaged from 2005 to 2007 in order to reflect the relevant time period in which the profits were earned that were paid out as dividends in the years 2006 to 2008.
Among the independent variables, the most important one is the ownership variable. Mancinelli and Ozkan (2006) use one variable for the voting rights of the largest shareholder, and another variable for the voting rights of all shareholders that have more than 2 percent of the voting rights, excluding the largest shareholder. In addition, they use a dummy variable that equals one if there is a second large shareholder and another dummy variable that equals one if the other large shareholders have more than 5 percent of the voting rights. In contrast, this thesis uses four variables for the size of shareholders.21
The first one contains the relative size of stock ownership of the largest shareholder if it exceeds 3 percent. The second contains the relative size of stock ownership of the second largest shareholder if it exceeds 3 percent. The third variable adds up the relative size of stock ownership of all other shareholders that hold at least 3 percent. Finally, the fourth ownership variable is a dummy that takes the value of one if there is at least one third largest shareholder who owns at least 3 percent of the stock of a company. As it is explained in the methodology section, combinations of these variables are used as well. In order to be able to detect the dividend preferences of different types of owners, dummy variables are used for banks, financial institutions, companies, state authorities,
Beteiligungsgesellschaften, and families and private blockholders. One final dummy variable is used
when the investor is of foreign origin. In the final analysis of this study (table 11), those dummies are multiplied with the size of the largest shareholder in order to include the shareholders’ size in the analysis of their influence on cash dividends.
The ownership data for this study were obtained from the Commerzbank’s Wer gehört zu wem? online service22. This database collects ownership information on about 12,000 German companies from the Bundesanzeiger and the German Federal Securities Supervisory Office (Bafin), both of which are
explained in the next section, as well as from the companies themselves23. For this reason, the
company data of this online service might be considered more complete and up to date than the official databases of the Bafin or the Bundesanzeiger. The database of the Wer gehört zu wem? online service was updated on May 11, 2009 before the data for this thesis were obtained from it. As it is not possible to download historic data from this source at reasonable cost, the ownership data could be collected only for one point in time, namely May 11, 2009. Since the financial data were downloaded for the years 2005 to 2008, there remains a gap in time between the ownership and the financial variables24.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
21 Size is always measured in percent of the shareholders’ equity. 22 http://wergehoertzuwem.das-buch-im-internet.de
23 In order to determine the ultimate shareholder of some Beteiligungsgesellschaften in the sample, the website
“Hoppenstedt Konzernstrukturen online” (http://www.hoppenstedt- konzernstrukturen.de/ nut_basis.htm) was used.
24 The company that experienced the most dramatic change in ownership that occurred during this gap is the
The effect of this difference in time might however be not too severe as ownership structures of German companies were found to be stable in the past, at least in the 1980s and 1990s (Da Silva et al., 2004:132). Despite there are roughly 12,000 companies listed in the database of the Commerzbank, it was necessary to obtain the ownership information for 30 companies from the Bafin because those firms were not listed in the Wer gehört zu wem? database.
In addition to the ownership, several other control variables are used that are based on lagged data, too. The year in which the company was founded is included in the Age variable. The data for this control variable are mostly taken from the company websites as the data on the year of incorporation that AMADEUS provides may be misleading in some cases25. Size is measured as the mean of the
market capitalizations from 2005 to 2007 and is based on Datastream data. Similar to von Eije and Megginson (2008), who use the standard deviation of the net income over five years divided by each year’s sales, this thesis uses the standard deviation of EBIT data from 2005 to 2007 divided by the mean of the revenues from 2005 to 2007 as a proxy for the Risk of the company. The leverage variable is defined as in Mancinelli and Ozkan (2006) as 2005 to 2007 average of total debt over total assets. The remaining financial data were obtained from Thomson Financial’s Datastream. Most variables that are used in this study can be obtained from this source. However, it has become necessary in multiple cases to add missing data to the dataset to fill the gaps that Datastream left open. In this case, data from the financial reports that can be found on the companies’ websites mainly filled the missing data points.
With regard to the control variables, the age of the company can be expected to have a positive impact on dividend payments. Von Eije and Megginson (2008:363) find this positive impact and reason that older companies may have lower growth opportunities, whereas their ability to accumulate funds might be better, compared to younger companies. The EBIT-based risk variable might be expected to have a negative impact on dividend policy since managers might prefer to pay low dividends during times of unstable profits. However, Goergen et al. (2005) find that the dividend flexibility in Germany is higher than, e.g. in the USA. If this effect is still present, then the influence of the risk variable might be negligible or insignificant. The size of the company will probably have a positive effect on dividends as mature and large companies have experienced a period of growth already and can now afford to pay out cash to their shareholders, whereas younger companies might need this cash for their own growth activities. Leverage is expected to have a negative influence on cash dividends
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
financial crisis in order to put it under state control and prevent this “system-relevant” bank from going bankrupt. Since, however, banks are excluded in this study, it does not have any effect here.
25 One example is the Daimler AG (former DaimlerChrysler AG). AMADEUS states the year of incorporation to
(Mancinelli and Ozkan, 2006). A possible explanation is given by Jensen (1986) who states that debt is a substitute for dividends because it is a more credible commitment of the management to pay out free cash flow and, hence, reduces the agency cost of free cash flow.
3.4 Descriptive statistics
In this sample of 156 companies, there are 122 firms that pay dividends in at least one year from 2006 to 2008, while there are 38 that do not pay any cash dividend in this period. Table 3 shows that state authorities seem to invest in companies with rather high dividend payout ratios, while banks prefer companies with low payout ratios. Another interesting fact is that the average market capitalization (Size) of the companies in which state authorities are the largest shareholder are much higher than the sample average, while family-controlled companies tend to be the smallest here.
Table 3: Averages of financial variables of the largest shareholders
Banks Financial institutions Companies State authorities Beteiligungs-gesellschaften Families and private blockholders Foreign shareholders Number of observations 3 35 32 3 29 49 40
Cash dividends paid / revenues
0.015 0.027 0.025 0.040 0.021 0.023 0.021
Cash dividends paid / market capitalization 0.010 0.019 0.016 0.038 0.023 0.014 0.018 Age 51 65.143 51.469 30.667 80 42.122 76.775 Size 1,142,040 7,723,079 1,567,767 31,0289,5 3 9,803,808 700,122 6,818,981 Risk 0.087 0.069 0.065 0.019 0.042 0.109 0.055 Leverage 0.396 0.246 0.179 0.304 0.199 0.204 0.218
With regard to the ownership situation, table 4 contains the ownership figures of the largest shareholders of all companies in the sample. 35 firms, or 22.4 percent of the sample are controlled by shareholders who own more than 50 percent of the company stocks and there are only two firms26 that
have no single shareholder who possesses at least 3 percent of the company stocks.
Table 4: Relative size of ownership of largest shareholder
Ownership in percent27 0 to 3 3 to 5 5 to 10 10 to 15 15 to 20 20 to 25 25 to 30 30 to 40 40 to 50 > 50 Number of companies 2 8 22 14 18 7 17 16 17 35
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
The data on direct ownership are summarized in table 5 which shows the number companies in which the different types of shareholders are the largest, second largest or belong to the group of other shareholders who own at least 3 percent of the company. Apart from the two companies mentioned above which do not have any shareholder that owns at least 3 percent, there are three other firms whose largest shareholders do not belong to either of the shareholder types in table 528. The final
column of table 5 indicates how many shareholders of the other types were found to be foreign-based. Contrary to the previous literature, banks are far from being the most important shareholder group and hold the largest block in only three out of the 156 companies. Much more often are companies, financial institutions, Beteiligungsgesellschaften, and families and individuals found to be the largest blockholders. The fact that 25.6 percent of the largest blockholders are foreign investors might be seen as support for the statement of Leuz and Wüstemann (2003) that Germany is transforming its economic system to attract foreign investors. Another striking fact is the high number of
Beteiligungsgesellschaften that hold large direct equity stakes in other firms. This can be seen as a
direct consequence of the 2001 reform of the Corporate Tax Act that created tax advantages for these kinds of companies. As one would expect, Beteiligungsgesellschaften are much less often found to be large ultimate shareholders of other firms. This suggests that these companies are mostly set up and used as a vehicle for direct shareholdings that can absorb dividend payments 95 percent tax free, independent of the tax treatment of the owner of the Beteiligungsgesellschaften. That becomes even more obvious when the ultimate shareholdings are taken into account. Among the 29
Beteiligungsgesellschaften that are the largest direct shareholders, 17 are found to have a family or an
individual as their ultimate shareholder. This figure is by far the highest when compared to those that have banks (1), companies (3), or state authorities (3) as ultimate shareholders. A similar pattern can be found for the second largest shareholders. Out of the 27 Beteiligungsgesellschaften there are 11 that are ultimately owned by families and individuals, whereas 4 are owned by state authorities. Banks, companies and financial investors are the ultimate shareholders of 2 Beteiligungsgesellschaften each.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
28 GFK SE has an association (GFK-Nürnberg e.V.) as its largest shareholder, Südzucker AG’s largest
Table 5: Number of companies that are owned by these types of shareholders (direct and ultimate ownership)
Banks Family and private blockholders Beteiligungs-gesellschaften State authorities Companies Financial institutions Foreign shareholders Direct ownership Largest shareholder 3 49 29 3 32 35 40 2nd largest shareholder 12 24 27 1 20 53 60
All others holding at least 3 percent
14 36 11 1 22 73 74
Ultimate ownership
Largest shareholder 8 74 6 11 23 27 40
2nd largest shareholder 21 47 8 7 16 40 61
All others holding at least 3 percent
28 45 4 3 20 65 73
Another pattern that can be observed in the data is that banks sometimes own parts of other companies not only directly, but also through their own mutual trust companies. Evidence therefore is present in 4 cases for the largest shareholders, and in 8 cases for the second largest shareholders.
Table 6: Relative size of average equity ownership of largest shareholders (direct and ultimate ownership)
Banks Family and private blockholders Beteiligungs-gesellschaften State authorities Companies Financial institutions Foreign shareholders Direct ownership Largest shareholder 18.4367 36.2446 35.8663 31.2600 33.3357 14.4463 18.8227 2nd largest shareholder 8.0797 9.4027 12.4106 20.2600 14.8455 8.5599 8.3998
All others holding at least
3 percent 21.7088 19.3645 18.7519 21.1600 21.6371 16.7231 16.8254
Ultimate ownership
Largest shareholder 16.1239 36.5623 36.9227 24.9961 31.1634 14.7877 19.3069
2nd largest shareholder 6.4355 11.8656 12.6319 14.5501 12.7886 8.2344 8.2687
All others holding at least 3 percent
4. Methodology
For this study it is not possible to use ordinary least squares analysis. Two types of regression techniques are used in this thesis. First, for the regression on the decision whether or not to pay dividends, the maximum likelihood technique for binary data is used. For the second type of regression, a method for censored data is used. The reason for that lies in the nature of the data of the dependent variable. Since dividends cannot be negative the dividend-based ratios that are used as dependent variables cannot be smaller than zero, too. Using ordinary least squares regression on such data would lead to biased and inconsistent parameter estimates (Brooks, 2008:533). Tobin (1958) developed a model to estimate regressions on censored dependent variables, which is known as Tobit analysis and which will be used in this study.
As it is outlined in the introduction, there are four main questions that shall be answered with this study. The first one is whether large shareholders have a preference for dividend-paying firms or for non-dividend-paying firms. To find an answer to this question, a dummy variable (DPNP; dividend payer / no dividend payer) is created that takes the value of 1 when the average value of cash dividends paid from 2006 to 2008 is positive, and zero otherwise. This variable then serves as dependent variable in the following regression equations.
(1) DPNP = !0 +!1 AGE + !2 RISK + !3 SIZE + !4 LEVERAGE + !5 LARGEST (2) DPNP = !0 +!1 AGE + !2 RISK + !3 SIZE + !4 LEVERAGE + !5 LARGEST
+!6 (2ND LARGEST + REST)
(3) DPNP = !0 +!1 AGE + !2 RISK + !3 SIZE + !4 LEVERAGE + !5 LARGEST +!6 2ND LARGEST + !7 REST(DUMMY)
(4) DPNP =
"
0 +"
1 AGE +"
2 RISK +"
3 SIZE +"
4 LEVERAGE +"
5 (LARGEST + 2ND LARGEST + REST)the company. This combination of ownership variables is used by Mancinelli & Ozkan (2006) as well. Finally, in the fourth case one variable containing the size of all shareholders that own at least 3 percent of the company is used. To account for the binary data of the dependent variable, an analysis based on maximum likelihood estimation is used in conjunction with an extreme value distribution of the residuals to determine the coefficients of this regression equation29.
The second main question outlined in the introduction deals with the amount of dividends that large shareholders prefer. Contrary to Short et al. (2002) and rather in line with Gugler and Yurtoglu (2003) and Mancinelli and Ozkan (2006), it is expected here that large shareholders will prefer lower dividends and, hence, have a negative influence on the cash dividends. The reason can be found in the economic system of Germany. Due to the insider system, monitoring the management can be achieved quite effectively even without high dividend payments because the important outside investors are reasonably well informed (Leuz and Wüstemann, 2003). In addition, also my expectation for the effect of the second largest shareholder is also in line with Gugler and Yurtoglu (2003) and Mancinelli and Ozkan (2006), who find it to be dividend increasing. In this way, as Mancinelli and Ozkan (2006) reason, it might protect the smaller shareholders from expropriation and tunneling. The empirical analysis of this question is based on Tobit analysis where the dependent variables (cash
dividends/revenues, and cash dividends/market capitalization) are censored at the lower end at zero.
The control variables used here are the same as in equations (1) to (4).
(5) RELDIVSIZE = !0 +!1 AGE + !2 RISK + !3 SIZE + !4 LEVERAGE + !5 LARGEST (6) RELDIVSIZE = !0 +!1 AGE + !2 RISK + !3 SIZE + !4 LEVERAGE + !5 LARGEST
+!6 (2ND LARGEST + REST)
(7) RELDIVSIZE = !0 +!1 AGE + !2 RISK + !3 SIZE + !4 LEVERAGE + !5 LARGEST +!6 2ND LARGEST + !7 REST(DUMMY)
(8) RELDIVSIZE =
"
0 +"
1 AGE +"
2 RISK +"
3 SIZE +"
4 LEVERAGE +"
5 (LARGEST + 2ND LARGEST + REST)RELDIVSIZE in this equations serves as a placeholder for the dependent variables of relative dividend
size, cash dividends/revenues and cash dividends/market capitalization. The ownership variables are used in exactly the same way as in equations (1) to (4).
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
29 Following the advice of Brooks (2008:539), robust covariances are computed according to “Huber/White” in
