Ownership concentration, investor type and dividend policy

(1)

dividend policy

Evidence from the European Union

Mathijs Koster1

Abstract

This study tests the relationship between ownership concentration, investor type and dividend pay-out. We use a sample consisting of over 1500 listed companies, all headquartered in countries that were member of the European Union before May 2004. We find that ownership concentration is negatively related to dividend pay-out. We reason that large shareholders expropriate other shareholders. Furthermore we find that institutional ownership is not significantly related to the dividend pay-out, where family ownership is positively related to the dividend pay-out. These results are somewhat surprising, as existing literature finds a strong positive relationship between institutional ownership and dividend pay-out and a negative relationship between family ownership and dividend pay-out. When combining the type of investor with the size of ownership of the largest investor, we find that institutional ownership is significantly positively related to the dividend pay-out. Family ownership combined with the size of ownership of the largest shareholder is also significantly positively related to the dividend pay-out.

Key words: Ownership concentration, Dividend policy, Investor type, Europe JEL Classification: G32, G35

1_{Student number: 1478052}

Master Corporate Financial Management

(2)

Table of content

1. Introduction

2. Theoretical background

2.1 Ownership concentration and dividend pay-out 2.2 Investor type and dividend pay-out

3. Data

3.1 Sample

3.2 Data sources and variables 3.3 Descriptive statistics 4. Methodology

4.3 Investor type, size of ownership and dividend pay-out 4.4 Robustness checks

5. Results

5.3 Investor type, size of ownership and dividend pay-out 5.4 Robustness checks

6. Conclusion

6.1 Suggestions for further research 7. References

8. Appendices

8.1 Detailed description of variables 8.2 Descriptive statistics

(3)

1. Introduction

In this paper we study the relationship between ownership concentration, investor type and dividend pay-out policy in Europe. Since the paper of Miller and Modigliani (1961) about dividend policy much literature has been dedicated to dividend (pay-out) policy. However the relationship between ownership concentration, investor type and dividend pay-out policy is a somewhat neglected one.

Most of the existing literature dealing with dividend policy focuses on United States based companies. But this is changing; Von Eije and Megginson (2008) researched dividend policy in the European Union using a dataset of 3400 companies spread over fifteen countries. Faccio et al. (2001) research the link between dividend policy and expropriation of minority shareholders. Mancinelli and Ozkan (2006) focus on the relation between ownership concentration and dividend policy in Italy. First results from these studies are that the relationship between ownership concentration and dividend pay-out is a negative one. However Renneboog and Szilagyi (2006) find this relationship to be non-negative in their study based on a sample consisting of Dutch companies.

The results on the relationship between the type of investor and the dividend pay-out are less clear. Different type of investors may have different dividend preferences. Following Short et al. (2002) and Kahn (2006) the presence of an institutional investor as the largest shareholder increases the dividend pay-out. Furthermore Mancinelli and Ozkan (2006) find that the presence of a family or individual as the largest shareholder decreases the dividend pay-out.

(4)

The goal of this paper is to research the relationship between ownership structure and dividend policy in Europe, in order to do this we will use Tobit regression techniques and we will test the following three hypotheses;

H1: The dividend pay-out is influenced by the concentration of ownership. H2: The dividend pay-out is influenced by the type of largest investor.

H3: The dividend pay-out is influenced by the type of largest investor in combination with the size of ownership of the largest investor.

In order to answer the hypotheses, we will use a sample of European listed companies, all headquartered in countries that were member of the European Union before May 2004. These countries are highly democratic and have well developed financial markets since a long time. Furthermore in line with existing literature, we delete all the financial and utilities companies from our sample. We base our results on the years 2007 and 2008, because data from these years is the most recent data available to us.

We add value to existing literature by focusing on the European market, this market is a relatively under-researched one, as most of the studies focuses on the United States. Secondly most of the existing research focuses on one or two countries; this research takes fourteen countries into account, which gives us the possibility to use a larger dataset, making it easier to generalize conclusions. Furthermore we use very recent data, which make it possible to provide valuable new insights and track recent changes in the field of ownership structure and dividend pay-out policy.

(5)

2. Theoretical background

In this section we present our theoretical background, it consists of 2 parts. Section 2.1 gives an overview of relevant theory relating to ownership concentration and dividend pay-out, like the agency theory and expropriation of minority shareholders. Furthermore we will discuss existing empirical findings on the relationship between ownership concentration and dividend pay-out. Section 2.2 deals with the relationship between the type of investor and dividend pay-out. Different types of investors may have different dividend preferences. We discuss the monitoring theory, tax implications, family ownership and we will discuss existing empirical findings on the relationship between the type of investor and dividend pay-out.

2.1 Ownership concentration and dividend pay-out

Agency theory

Sometimes economic literature assumes that managers are perfect agents who always act in the best interest of the investors. However in the real world this is highly unlikely to be the case, with the consequence that an agency conflict between the manager and the investor can arise, resulting in agency costs. Following Rozeff (1982) pay-out policy may act as a remedy to solve for these agency costs.

(6)

as consequence this lowers the risk for management and the creditors and transfers wealth form shareholder to bondholders.

Following Easterbrook (1984) both the monitoring and the risk-aversion problems are less serious when a company is constantly in the capital market. Large block holders can put pressure on the management to go to the capital market for financing of new projects. First of all this decreases the monitoring costs, as the creditors will take up the monitoring role and secondly it transfers wealth from the creditors to the shareholders as the debt-equity ratio increases and thus the risk of the company. The consequence of forcing the management to go to the capital market is that retained earnings can be used to be paid out in the form of dividends instead of being used to finance new projects internally.

(7)

Expropriation of minority shareholders

Faccio et al. (2001) study the expropriation of minority shareholders in Europe and Asia. They find that the presence of a strong block holder decreases the dividend pay-out and the presence of multiple large owners in Europe increases the dividend pay-out and decreases the dividend pay-out of Asian firms. This is an indication that in Europe, when there is no single large controlling shareholder the dividend pay-out increases. Faccio et al. (2001) reason that in Europe the presence of multiple large shareholders prevent the largest shareholder from expropriating the other shareholders. This contradicts with the findings of Asia were multiple large shareholders seem to collude in order to expropriate other shareholders, resulting in a lower dividend pay-out. Faccio et al. (2001) also state that this expropriation of minority shareholders is often done via tunneling techniques. Johnson et al. (2000) give the following definition of tunneling; ‘’the transfer of resources out of a company to its controlling shareholder’’. Techniques that are often used for tunneling are setting of unfair terms for intra-group sales and services and transfers of assets and control stakes, excessive executive compensation and specific loan guarantees. Weinstein and Yafeh (1998) for example find that Japanese companies affiliated to bank-controlled groups pay higher interest rates on their liabilities compared to unaffiliated companies. They interpret this as evidence that these banks expropriate minority shareholders. Some, but not all of these tunneling techniques that are used to expropriate minority shareholders are illegal.

(8)

They give two explanations for the negative relationship between ownership concentration and dividend pay-out. The first explanation of the expropriation of minority shareholders is that controlling shareholders use tunneling techniques in order to expropriate minority shareholders. The second explanation given by Mancinelli and Ozkan (2006) is that when there are large private benefits for managers, they tend to keep resources under their own control instead of distributing them to the shareholders. This is very likely to be the case in Italy, as 70% of the top managers comes from the largest controlling shareholder. This expropriation of minority shareholders can be seen in the light of the agency costs dividend explanation discussed by Easterbrook (1984) and Jensen’s (1986) free cash flow theory.

Gugler and Yurtoglu (2003) find in their German based study that larger holdings of the largest shareholder decreases the dividend pay-out, implying expropriation of minority shareholders. They also find that the presence of a second large shareholder increases the dividend pay-out; this is in line with the findings of Faccio et al. (2001) who find for Europe that the presence of multiple large shareholders increases the dividend pay-out. They reason that these shareholders monitor each other. The findings of Bena and Hanousek (2006) for the Czech Republic are perfectly in line with the findings of Gugler and Yurtoglu (2003), they find that larger holdings of the largest shareholder decreases the dividend pay-out and that the presence of a second large shareholder increases the dividend pay-out.

Summary

(9)

Smit (2009) do not find a significant relationship between ownership concentration and dividend pay-out.

(10)

Table 1: Summary of empirical findings ownership concentration-dividend pay-out

Author(s) Period Country Sample

size Main findings Rozeff (1982) 1974-1980 US 1000 companies

Negative link between

ownership concentration and dividend pay-out. Faccio et al. (2001) 1992-1996 Europe and Asia 5897 companies

For Europe positive link

between multiple large owners and dividend pay-out.

Gugler and Yurtoglu (2003) 1992-1998 Germany 266 companies

Larger holdings of largest shareholder decreases dividend pay-out, larger holdings of

second largest shareholder

increases dividend pay-out. Goergen et al. (2005) 1984-1993 Germany 221 companies

No link between ownership concentration and dividend pay-out. Renneboog and Trojanowski (2005) 1992-1998 UK 985 companies

Negative link between voting power of large block holders and dividend pay-out.

Kahn (2006)

1985-1997

UK 330

companies

Negative relation between

ownership concentration and dividend pay-out.

Mancinelli and Ozkan (2006)

2001 Italy 139

companies

Voting rights of the largest

shareholder are negatively

related to dividend pay-out. Bena and Hanousek (2006) 1996-2003 Czech Republic 238 companies

Presence of one large owner is negatively related to dividend pay-out. Presence of at least one strong minority shareholder increases dividend pay-out Renneboog and Szilagyi (2006) 1996-2004 The Netherlands 150 companies

Ownership concentration not negatively related to dividend pay-out.

Smit (2009)

2005-2008

Germany 156

companies

(11)

2.2 Investor type and dividend pay-out

Monitoring

Besides testing the relationship between ownership concentration and the dividend pay-out, Kahn (2006) also studies the relationship between the type of investor and the dividend pay-out, using UK based companies. He finds that institutional ownership is associated with a higher dividend pay-out and individual or family ownership is associated with a lower dividend pay-out. He reasons that institutional investors are poor

monitors compared to individuals and families. Following the agency theory2 this results

in higher agency costs and as a consequence these institutional investors demand a higher dividend pay-out. It is interesting to see that he finds a negative link between ownership concentration and dividend pay-out, but contrary to this, when he distinguishes between various types of investors, he finds that institutional ownership concentration is positively related to the dividend pay-out. Short et al. (2002) also state that institutional investors in the UK are poor monitors and they also find the relationship between institutional ownership concentration and dividend pay-out to be positive.

Goergen et al. (2005) state that the German economic system is a ‘’bank based’’ system, implying that banks have strong linkages with companies. Banks often hold places in the supervisory board and are active monitors in Germany, therefore he reasons that this should reduce the agency costs and as a consequence the dividend pay-out. However his results do not confirm this hypotheses, as he does not find a significant relationship between the type of investor and the dividend pay-out.

Smit (2009) finds in his German based study a negative relationship between institutional ownership and the dividend pay-out, this confirms the monitoring hypotheses of Goergen et al. (2005). When he combines the institutional investor type with the size of ownership he finds mixed results for different types of institutional investors (banks and financials institutions).

(12)

Taxation

Following Allen et al. (2000) taxation is one of the key elements explaining corporate pay-out policy. Various types of owners may be taxed differently and as a consequence they may have different dividend preferences. They conclude that this is one of the reasons why ownership structures influences the dividend pay-out. Weiyu and Jinlan (2008) tested the theory of Allen et al. (2000) using a sample of US based firms, they find support for the theory of Allen et al. (2000) and find that the presence of an institutional investor increases the dividend pay-out. Furthermore Allen and Michaely (2003) state that in the US dividends of individual investors are taxed at a higher rate compared to corporations and institutions. Some of these institutions do not have to pay taxes at all on their incoming dividends. They reason that this may be one of the causes why institutional investors prefer a higher dividend pay-out compared to individuals. Short et al. (2002) find the same for the UK; they reason that some shareholders, like pension funds, are tax-exempt. As a consequence these institutional investors prefer dividends above retained earnings and demand a higher dividend pay-out. Eckbo and Verma (1994) find similar tax advantages for institutional investors for their Canadian sample. Contrary to this Bena and Hanousek (2006) find for the Czech Republic that there are no differences in the tax rates on dividends between various types of investors.

(13)

as a consequence these pension funds may prefer stable high dividend paying firms in order to meet their own pay-out obligations.

Family ownership

Gugler (2003) finds the relationship between a family as the largest investor and the dividend pay-out to be negative. He reasons that often the director of a company comes from the controlling family, reducing the agency costs and as a consequence the dividend pay-out decreases. The findings of Smit (2009) confirm these results as he finds a negative relationship between family ownership and the dividend pay-out. This is also in line with the findings of Faccio et al. (2001) and Mancinelli and Ozkan (2006), who state that in Italy most of the time the top manager comes from the controlling family. They reason that this could reduce the agency cost and thus the dividend pay-out, but more important it makes the expropriation of minority shareholders much more likely. Via tunneling techniques resources are transferred to the controlling family or individuals and top managers from the controlling family enjoy private benefits, as a consequence the dividend pay-out decreases.

Summary

Following Kahn (2006) institutional investors in the UK are poor monitors, as a consequence they demand a higher dividend pay-out to be compensated for bearing the higher agency costs. Furthermore Eckbo and Verma (1994), Short et al. (2002) and Allen and Michealy (2003) find the relationship between institutional ownership and the dividend pay-out to be positive. They reason that this may be due to the lower rate the dividends of these investors are taxed at compared to individuals and families. Short et al. (2002) give another explanation for the positive link, they state that pension funds have pay-out obligations tot their pension holders and thus prefer stable high dividend paying firms.

(14)

expropriate other shareholders and enjoy private benefits, which reduces the dividend pay-out.

(15)

Table 2: Summary of empirical findings investor type-dividend pay-out

Author(s) Period Country Sample size Main findings

Eckbo and Verma (1994) 1976-1988 Canada 308 companies Institutional ownership is positively related to dividend pay-out. Faccio et al. (2001) 1992-1996 Europe and Asia 5897 companies

Family control decreases dividend pay-out.

Short et al. (2002)

1988-1992

UK 211

companies

Positive link between

institutional ownership and dividend pay-out.

Gugler (2003)

1991-1999

Austria 600

companies

Family control decreases dividend pay-out. Goergen et al. (2005) 1984-1993 Germany 221 companies

No significant results for relation between investor type and dividend pay-out. Bena and Hanousek (2006) 1996-2003 Czech Republic 1664 companies Institutional ownership positively influences dividend pay-out. Kahn (2006) 1985-1997 UK 330 companies Institutional ownership is positively related to

dividend pay-out. Family/ individual ownership is

negatively related to

dividend pay-out. Weiyu and Jinlan

(2008) 1980-2002 US - Institutional positively influences the ownership

amount of dividend paid out.

Smit (2009)

2005-2008

Germany 156

companies

Institutional and family

ownership negatively

influences the dividend pay-out.

(16)

3. Data

This section consists of a description of our sample, variables and an analysis of the data we use for our study. Section 3.1 deals with the sample and the sample selection process. Section 3.2 gives an overview of the dividend and ownership variables used in our study and provides an oversight of the control variables. Section 3.3 gives an overview of the descriptive statistics of the most important variables.

3.1 Sample

We examine the relationship between ownership concentration, shareholder type and dividend pay-out for the years 2007 and 2008. We use these two years, because it is the latest available ownership data we could retrieve. Our sample consists of listed companies headquartered in one of the 14 European countries that were member of the European Union before 2004. Countries that became member after 2004 are mostly eastern or central European countries. We exclude these countries, because the European Union members before 2004 are originally more capitalistic and democratic compared to the eastern and central European countries. Companies from countries that were member of the European Union before 2004 operate in a full market economy, were dividends may be of less relevance. Companies from eastern and central European countries are often (formerly) state-owned and this may have an impact on the dividend policy.

(17)

1511 companies divided over 14 countries3. Section 3.3 provides more information on the descriptive statistics of the sample.

3.2 Data sources and variables

The dependent variable that we will use in all our regressions is the dividend pay-out. We divide the total cash dividend a company paid out in a year by the total assets of that year. Existing literature sometimes divides cash dividend by market capitalization or earnings. The reason we choose for dividing cash dividend by total assets, is because we prefer using book values above the more fluctuating market values and the fact that earnings can be manipulated, are more volatile and sometimes are negative. Data on cash dividend and total assets is retrieved from DataStream for the years 2007 and 2008.

On the right hand side of the regressions the ownership variables are the most important. We use the total shareholdings4 of the largest shareholder as a measure for ownership concentration. Ideally we also would like to use total shareholdings of the other shareholders, but unfortunately these data is not available to us. Furthermore we use a variable which represents the ratio of direct ownership to total ownership, this makes it possible for us to test whether direct and indirect ownership differ in their influence on the dividend pay-out. We retrieved data of the shareholdings of the largest shareholder

for the years 2007 and 2008 from the Amadeus database5.

In order to test the relationship between the type of investor and the dividend pay-out, we distinguish between two different types of largest shareholders, respectively the institutional investor and the family/individual investor6. When the largest shareholder, for example, is an institutional investor, it takes on a value of 1 on the institutional investor type dummy and 0 otherwise. We constructed these dummies for the years 2007 and 2008 with data retrieved from the Amadeus database.

3_{All companies from Luxembourg are excluded from the sample after the merge of the two databases,}

correction for financials and utilities firms and deletion of companies with missing or incorrect data.

4_{Direct and indirect shareholdings of the largest shareholder.}

(18)

We also add several control variables to our regressions. We will explain each control variable in detail7_{and we will give following existing literature an estimate of the}

direction of the sign. The control variables we use are lagged by one year compared to the dependent variable. We do this for two reasons, first of all most companies determine their dividend pay-out on the (financial) performance of the previous year and secondly we do this to prevent the case of endogenity.

The first control variable we use is age. We measure age in years and extract the data from the Amadeus database. Following Von Eije and Megginson (2008) and Salas and Chahyadi (2006), we expect to see that older companies pay out more dividends. A reason for this can be that older companies have less growth opportunities and are better able to accumulate funds compared to younger firms (Von Eije and Megginson, 2008). The second control variable, size, is measured as the natural logarithm of market capitalization, which we extract from the DataStream database. We expect to see that larger companies pay out more dividends, this is in line with the findings of Denis and Osobov (2008) and Faccio et al. (2001). An explanation for this can be that larger firms have less growth opportunities and need less cash to finance growth compared to smaller firms and thus are able to distribute more dividends.

The third control variable we add to the regressions is leverage. We measure leverage by dividing total debt by total assets, we retrieve this data from DataStream. Following Jensen (1986) leverage may help to control agency costs, it can be seen as a substitute for dividends, so we expect to see a negative relationship between leverage and dividend pay-out. This is in line with the findings of Mancinelli and Ozkan (2006).

(19)

management to pay these cash balances out in the form of a dividend, they do this in order to reduce private benefits and misbehavior of the management from these large cash balances. Therefore, we expect to see a positive relationship between cash holdings and dividend pay-out.

Growth in assets is the sixth control variable. We measure growth as the growth in total assets from one year to another, were we retrieve data on total assets from DataStream. We expect to see a negative relationship between growth and dividend pay-out. We reason that companies that have good growth prospects pay fewer dividends, because they need the money to finance this growth and as a consequence have less cash available to return to the shareholder. This is in line with the findings of Faccio et al. (2001). We also add country variables to our regressions. We construct 14 dummies, each representing a country. When a company is for example headquartered in the Netherlands, it takes on a value of 1 on the Netherlands dummy and 0 on the other dummies. In order to save space the country dummies are not presented in the methodology and results section. We exclude the German country dummy, in order to prevent the case of multicollinarity.

(20)

3.3 Descriptive statistics

Our sample consists of 1511 companies divided over 14 countries; figure 1 in appendix 8.2 gives a geographical representation of our sample. Most of the companies come from the United Kingdom, followed by Germany and France.

Table 3: Descriptive statistics

Investor type

Institutional Family Other Total 2007 2008 2007 2008 2007 2008 2007 2008 Number of observations 571 534 337 368 603 609 1511 1511 Dividend 0.026 0.027 0.021 0.021 0.024 0.024 0.024 0.024 Total ownership 0.210 0.224 0.347 0.341 0.481 0.466 0.348 0.350 Relative direct ownership 0.486 0.710 0.841 0.841 0.875 0.876 0.719 0.810 Age 39.445 41.360 32.715 33.652 44.169 44.701 39.829 40.829 Size 8.354 8.403 7.816 7.880 8.349 8.438 8.232 8.290 Risk 0.188 0.136 0.052 0.047 0.131 0.210 0.135 0.144 Cash 0.114 0.105 0.106 0.097 0.119 0.110 0.114 0.105 Debt 0.133 0.131 0.161 0.164 0.139 0.139 0.142 0.142 Growth 0.086 0.140 0.107 0.123 0.075 0.113 0.087 0.132 Number of companies

from UK and Ireland 349 310 99 121 82 99 530 530 Number of companies

from continental Europe 222 224 238 247 521 510 981 981 We distinguish between two types of largest investors, the institutional investor and the family/individual investor, all companies with an other type of largest investor fall in the other category. Total represents the total sample. The dividend variable represent the average cash dividend pay-out divided by total assets of a company. Total ownership is the average percentage of shareholdings of the largest investor. Relative direct ownership is the average proportion of direct shareholdings of total shareholdings of the largest shareholder. Age is a one year lagged variable, which represents the age of the company in years. Size is a one year lagged variables which is measured by the natural logarithm of the market capitalization of each company. Risk is measured by dividing the standard deviation of net income for the period 2006-2008 by total revenues; an increase in this ratio means an increase in the volatility of net income, making the cash flows riskier. Cash is a one year lagged variable which is measured by dividing total cash by total assets of each company. Debt is a one year lagged variable which is measured by dividing total debt by total assets of each company. Growth is a one year lagged variable that is measured as the change in total assets from one year to another.

(21)

When looking at the total ownership variable we see that companies with an institutional investor as the largest shareholder have the smallest total shareholding of the largest shareholder, indicating that these companies have the lowest concentration in ownership. Also companies with a family/individual as the largest shareholder fall below the average value of total shareholdings of the largest shareholder of the whole sample. Just like the dividend pay-out, the shareholdings of the largest shareholder tend to be relatively stable over time.

The direct to total shareholdings ratio shows us that companies with an institutional investor are to a larger extent held via indirect shareholdings compared to the total sample. Contrary to this, companies with a family/individual as the largest shareholder are to a larger extent held via direct shareholdings compared to the total sample. Furthermore it is remarkable to see the large shift to direct shareholdings from 2007 to 2008 for the institutional investor type, as the ratio remains relatively stable for the family/individual and other investor type.

When looking at the descriptive statistics of the control variables we see that companies with a family/individual as the largest investor are in general younger and smaller companies compared to companies with an institutional investor as the largest shareholder. Furthermore these companies with a family/individual as investor have a less volatile net income, have less cash and are higher leveraged compared to the other investor types. The growth prospects of the companies are not stable over time, making it hard to draw conclusions about the growth prospects of the companies in the sample. Companies with an institutional investor as the largest shareholder are in general larger, have a riskier net income and are lower leveraged compared to the total sample.

(22)

4. Methodology

In this section we present the methodology we use in our paper. Due to the nature of our dependent variable it is not possible to use OLS (ordinary-least-square) regression techniques. As dividends cannot be negative, the dependent variables is always 0 or larger, which means that our dependent variable is censored. Following Brooks (2008), OLS regression techniques seriously bias the estimates of the coefficients in the case of censored data. So we use the Tobit regression model developed by Tobin, which is better suited for an analysis of censored data. As mentioned before in the data section, all the control variables are lagged one year compared to the dependent variable, we do this because otherwise these variables are endogenous variables, which could seriously bias our results. We test all the regressions for the years 2007 and 2008.

Section 4.1 deals with the methodology used to test the relationship between ownership concentration and dividend pay-out. Section 4.2 deals with the relationship between the type of investor and dividend pay-out. In section 4.3 we present the methodology relating to the combination of the type of investor and size of ownership of the largest shareholder and their relation with dividend pay-out. In section 4.4 we explain how we test the robustness of our results.

As stated in the introduction we will answer three main hypotheses in this study. The first one is the influence of ownership concentration on the dividend pay-out. We will use equation 1 to test this hypothesis.

1) Dividend=α0 + α1 age+ α2 size+ α3 risk+ α4 cash holdings+ α5 debt+ α6 growth+ α7 total ownership + α8 relative direct ownership

Here the dependent variable, Dividend, is the total cash dividend paid out by a company divided by the total assets of the company. Furthermore α0 is the intercept in the

regression and variables α1 till α6 are control variables that we add to the regression8. We

(23)

also add country dummies, which take on a value of 1 or 0 depending on the country were the company is headquartered. In order to save space these country dummies are not represented in this section and in the results section.

The most important ownership variable that we include in the regression is the total ownership variable. It represents the percentage of total ownership (direct and indirect) of the largest shareholder. We use this variable as a proxy for ownership concentration, because we do not have ownership data of other shareholders then the largest shareholder. Relative direct ownership represents the proportion of direct ownership to total ownership; it provides us with information about the influence of the proportion of direct and indirect ownership on the dividend pay-out.

With equation 2 we test the second hypothesis, which deals with the influence of the type of investor on the dividend pay-out. We use the following two investor types, respectively the institutional investor and the family/individual investor. Due to a lack of theoretical underpinning we exclude other investor types.

2) Dividend=α0 + α1 age+ α2 size+ α3 risk + α4 cash holdings + α5 debt+ α6 growth+ α7 institutional+ α8 family

The dependent variable and the variables α1 till α6 are the same variables as used in

(24)

4.3 Investor type, size of ownership and dividend pay-out

The third hypotheses deals with the influence of the type of investor in combination with the size of ownership of the largest shareholder on the dividend pay-out. In order to test this hypothesis we use equation 3, where we multiply the institutional investor and the family/individual variable with the total ownership variable.

3) Dividend=α0 + α1 age + α2 size + α3 risk + α4 cash holdings + α5 debt + α6 growth + α7 total ownership + α8 relative direct ownership + α9 institutional + α10 family + α11 institutional * total ownership + α12 family * total ownership

The dependent variable and variables α1 till α8 are the same variables as used in equation

1. Furthermore variables α9 and α10 come from equation 2. With variables α11 and α12 we

are not only able to test the dividend preference of a specific type of investor but we can also test the influence of the power the specific investor has on the dividend pay-out of the company.

4.4 Robustness checks

We conduct several tests in order to test the robustness of our results. First of all we test whether the relation between the dividend pay-out and total ownership is a true linear relationship. In order to do this we square the total ownership variable and test whether the function is a quadratic one or not.

(25)

5. Results

In this section we will present our results. In section 5.1 we discuss the relationship between ownership concentration and dividend pay-out. Section 5.2 deals with the relationship between the type of investor and dividend payout. In section 5.3 we combine the shareholder type with the size of ownership of the largest shareholder and discuss the impact on the dividend pay-out. And finally in section 5.4 we will discuss the robustness of our results.

The first hypotheses that we test is; the dividend pay-out is influenced by the concentration of ownership. From table 4 we see that for the years 2007 and 2008 total ownership of the largest shareholder significantly negatively influences the dividend pay-out. However the year 2008 is only marginally significant, falling just inside the 10% significance level. Furthermore we see that the coefficient decreases from 2007 to 2008, implying that the influence of ownership concentration on the dividend pay-out was stronger for 2007 compared to 2008.

Our findings are in line with the findings of Rozeff (1982), Mancinelli and Ozkan (2006), Renneboog and Trojanowski (2005), Kahn (2006) and Faccio et al. (2001), who also find a negative relationship between ownership concentration and dividend pay-out. This supports the view that other shareholders are expropriated by the large controlling shareholder. Faccio et al. (2001) state that this expropriation is done via tunneling techniques, were resources are transferred to the controlling shareholder. Furthermore Mancinelli and Ozkan (2006) find that 70% of the top managers in Italy come from the controlling family, making expropriation easier and these managers can also use resources of the company for their own private benefits.

(26)

Our results reject the agency theory of Easterbrook (1984) and Jensen’s (1986) free cash flow theory, both theories imply that a positive relationship is expected between ownership concentration and the dividend pay-out. They reason that investors want to reduce the free cash flow available to managers and that they want to be compensated for high monitoring and agency costs. Some authors argue that large shareholders are able to mitigate the agency problems and demand a lower dividend pay-out in order to overcome the problem of underinvestment by the management.

Our results contradict with the findings of Goergen et al. (2005), Renneboog and Szilagyi (2006) and Smit (2009) as they do not find a significant relationship between ownership concentration and the dividend pay-out.

Table 4: Shareholdings largest shareholder and dividend pay-out

Dependent variable: Cash dividend/Total assets

Year 2007 2008

Independent variable Coefficient Prob. Coefficient Prob. TOTAL OWNERSHIP -0.017 0.023 -0.012 0.099 RELATIVE DIRECT OWNERSHIP 0.005 0.148 -0.003 0.408

AGE 0.000 0.054 0.000 0.056 SIZE 0.021 0.000 0.020 0.000 RISK -0.036 0.000 -0.028 0.001 CASH 0.029 0.003 0.034 0.001 DEBT -0.056 0.000 -0.048 0.000 GROWTH 0.000 0.344 -0.001 0.828 C -0.155 0.000 -0.142 0.000 Number of firms 1508 1510

Left censored observations 472 467 In this regression the dependent variable is computed by dividing the cash dividend for 2007 and 2008 by total assets for 2007 and 2008. Prob. stands for the probability. Total ownership is the total ownership (direct and indirect) of the largest shareholder. Relative direct ownership is measured by dividing the direct percentage of ownership of the largest shareholder by the total ownership of the largest shareholder. We included country dummies for all countries, excluding Germany. Due to space restrictions they are not represented here. Age represents the age of a company in years. Size is a one year lagged variable, measured as the natural logarithm of the market capitalization. Risk is measured as the standard deviation of net income for the period 2006-2008 divided by sales. Cash is a lagged variable, computed by dividing cash by total assets. Debt is a lagged variable computed by dividing debt by total assets. Growth is a lagged variable that measures the growth in total assets from one year to another. C is the intercept in the regression.

(27)

When looking at the control variables, we see that all the control variables have the expected sign and with the exception of the growth variable, are significant. This implies that larger, older and cash-rich firms pay out more dividends to their shareholders. We see that firms with a riskier net income and firms who are higher leveraged pay less dividends.

The second hypotheses deals with the relationship between the type of investor and the dividend pay-out. In order to test this hypothesis we distinguish between two different investor types, respectively the institutional investor and the family/individual investor. From table 5 we see that the institutional investor type has a positive sign for 2007 and 2008, however for both years this variable is insignificant. Our results do not support the findings of Kahn (2006), Eckbo and Verma (1994), Short et al. (2002) and Allen and Michealy (2003), who find the relationship between institutional ownership and dividend pay-out to be positive. They give several explanations for this, first of all a monitoring explanation, which state that institutional investors are poor monitors and as a consequence ask for a higher compensation in the form of higher dividends. Second, they give a tax explanation, institutional investors pay less taxes on their dividends compared to other investors, which makes them more favorable to a high dividend pay-out. The third explanation deals with pay-out obligations of pension funds to their pension holders, as a consequence pension funds (institutional investors) may prefer high stable dividend paying firms in order to meet their pay-out obligations. Our results do not confirm the three explanations stated above, this is in line with the findings of Goergen et al. (2005) who also did not find a significant relationship between the type of investor and the dividend pay-out.

(28)

which makes it easier to expropriate other shareholders and to enjoy private benefits, which reduces the dividend pay-out. Our findings do not confirm their results, we do not have an explanation for the unexpected positive sign on the family coefficient. The results of Gugler (2003) and Mancinelli and Ozkan (2006) are based on companies from Austria and Italy, where shareholders are not very well protected. This implies that it is easier to expropriate shareholders in these countries and thus lowering the dividend pay-out. This may be an explanation why we do not find similar results. Our results are based on 14 countries in Europe, including a large set of companies from the UK where shareholders are better protected against expropriation. Furthermore the existing literature on the relationship between family ownership and the dividend pay-out is sparse for the European union.

We do not see changes is the signs and significance of the control variables compared to the results in table 4, with the exception of the growth variable, all control variables have the expected sign and are significant.

Table 5: Investor type and dividend pay-out

Year 2007 2008

Independent variable Coefficient Prob. Coefficient Prob. INSTITUTIONAL 0.004 0.276 0.005 0.157 FAMILY 0.007 0.070 0.010 0.010 AGE 0.000 0.080 0.000 0.058 SIZE 0.021 0.000 0.021 0.000 RISK -0.036 0.000 -0.027 0.002 CASH 0.029 0.003 0.033 0.001 DEBT -0.056 0.000 -0.048 0.000 GROWTH 0.000 0.369 -0.001 0.690 C -0.163 0.000 -0.164 0.000 Number of firms 1508 1510

(29)

5.3 Investor type, size of ownership and dividend pay-out

In order to test the third hypotheses, we test the relationship between a combination of the type of investor and the size of ownership of the largest shareholder and the dividend pay-out. From table 6 we see that the institutional variable multiplied by the size of ownership of the largest investor positively influences the dividend pay-out, the variable is significant for 2007 and 2008. We expected to find a positive relationship between the institutional investor type and the dividend pay-out in section 5.2, but we did not find a significant relationship there. Here we find that a combination with the size of ownership delivers the expected positive relationship between an institutional investor type and the dividend pay-out. This is in line with the findings of Kahn (2006), Eckbo and Verma (1994), Short et al. (2002) and Allen and Michealy (2003), who find the relationship between institutional ownership and dividend pay-out to be positive. We discussed three possible explanations for this in the previous section, respectively, a monitoring, a tax and a pay-out obligation explanation. Furthermore from section 5.1 we see that ownership concentration is negatively related to the dividend pay-out, but when combining it with an institutional type of investor we see that the sign switches to a positive one. We reason that the type of investor has the largest impact on the dividend pay-out compared to the relative size of ownership.

A combination of the family investor type and the size of ownership of the largest shareholder positively influence the dividend pay-out, however the variable is only significant for the year 2008. This supports our findings from section 5.2, where we find a positive relationship between the family investor type variable and the dividend pay-out. These findings contradict with existing empirical literature, who finds this relationship to be negative.

(30)

Table 6: Investor type, size of ownership and dividend pay-out

Year 2007 2008

Independent variable Coefficient Prob. Coefficient Prob. INSTITUTIONAL * TOTAL OWNERSHIP 0.038 0.027 0.030 0.063 FAMILY * TOTAL OWNERSHIP 0.027 0.150 0.033 0.054 INSTITUTIONAL -0.010 0.143 -0.007 0.266

FAMILY -0.005 0.523 -0.004 0.599

TOTAL OWNERSHIP -0.030 0.004 -0.025 0.015 RELATIVE DIRECT OWNERSHIP 0.004 0.278 -0.003 0.477

AGE 0.000 0.034 0.000 0.035 SIZE 0.022 0.000 0.021 0.000 RISK -0.036 0.000 -0.027 0.002 CASH 0.028 0.004 0.032 0.001 DEBT -0.058 0.000 -0.050 0.000 GROWTH 0.000 0.341 -0.002 0.643 C -0.155 0.000 -0.146 0.000 Number of firms 1508 1510

Left censored observations 472 467 In this regression the dependent variable is computed by dividing the cash dividend for 2007 and 2008 by total assets for 2007 and 2008. Prob. stands for the probability. We multiply the two investor type dummies with the total ownership variable in order to see how these two variables influence the dividend pay-out when they are combined. Institutional is a dummy that takes on a value of 1 when the largest shareholder is an institutional investor and 0 otherwise. Family is a dummy that takes on a value of 1 when the largest shareholder is a family/individual and 0 otherwise. Total ownership is the total ownership (direct and indirect) of the largest shareholder. Relative direct ownership is measured by dividing the direct percentage of ownership of the largest shareholder by the total ownership of the largest shareholder. We included country dummies for all countries, excluding Germany. Due to space restrictions they are not represented here. Age represents the age of a company in years. Size is a one year lagged variable, measured as the natural logarithm of the market capitalization. Risk is measured as the standard deviation of net income for the period 2006-2008 divided by sales. Cash is a lagged variable, computed by dividing cash by total assets. Debt is a lagged variable computed by dividing debt by total assets. Growth is a lagged variable that measures the growth in total assets from one year to another. C is the intercept in the regression.

(31)

companies that are headquartered in continental Europe and companies who are headquartered in the UK and Ireland. We make this specific distinction, because existing literature often finds differences in dividend pay-out between these two sets of companies (Faccio et al., 2001). We reason that companies who are headquartered in the UK and Ireland operate under a common law regime and companies from continental Europe under a civil law regime. Under a common law regime investors are better protected and a as consequence have more power to extract dividends from a company, which increases the dividend pay-out. We add a dummy variable representing companies from the UK and Ireland to all three regressions. The results from regressions 1 and 3, as presented in table 9 and 10 in appendix 8.4, show that the UK and Ireland dummies are not significant for the years 2007 and 2008. This implies that for these two regressions our results are robust across continental Europe and the UK and Ireland.

Table 7: Investor type and dividend pay-out UK and Ireland

Year 2007 2008

Independent variable Coefficient Prob. Coefficient Prob. INSTITUTIONAL 0.005 0.169 0.005 0.126 FAMILY 0.010 0.011 0.011 0.003 UK & IRELAND 0.006 0.095 0.006 0.085 AGE 0.000 0.068 0.000 0.058 SIZE 0.020 0.000 0.020 0.000 RISK -0.036 0.000 -0.029 0.001 CASH 0.028 0.004 0.033 0.001 DEBT -0.050 0.000 -0.044 0.000 GROWTH 0.000 0.414 -0.002 0.541 C -0.157 0.000 -0.161 0.000 Number of firms 1508 1510

(32)

From table 7 we see that for regression 2 the UK and Ireland dummy is significantly positive for 2007 and 2008. This implies that in 2007 and 2008 there where significant differences in the dividend pay-out between continental Europe and the UK and Ireland when testing for the influence of the type of investor on the dividend pay-out. The positive sign is expected, as we explained above these companies operate under a common law regime, where investors are better protected.

As a last test we look whether there are differences over time from 2007 to 2008. We do not find any changes in the signs of the variables from one year to another, furthermore the coefficients decrease or increase only marginally from 2007 to 2008. We also conducted a test by combining the data of 2007 and 2008, which resulted in 3022 observations instead of 1551 and we added a dummy representing the year 2008. We find that this dummy is insignificant, implying that there are no significant differences relating to the dividend pay-out between 2007 and 2008.

(33)

6. Conclusion

Using a dataset of over 1500 companies all from countries that were member of the European union before may 2004, we examine the relationship between ownership concentration, investor type and the dividend pay-out. We add value to the existing literature by basing our research on 14 European countries, while most of the other studies are based on one or two countries. As a consequence our dataset is larger than those used in most of the previous ownership structure and dividend related studies. When looking at the influence of ownership concentration on the dividend pay-out, we find the relationship between ownership concentration and the dividend pay-out to be negative. Our results support the view that large controlling shareholder expropriate other shareholders, which as a consequence reduces the dividend pay-out. The findings of Faccio et al. (2001) and Mancinelli and Ozkan (2006) support our results. Faccio et al. (2001) explain that shareholders are expropriated via tunneling techniques, where resources from the company are transferred to the controlling shareholder. Furthermore Mancinelli and Ozkan (2006) find that when top managers come from the controlling shareholder, the expropriation of other shareholders is easier and these top managers can use resources of the company for their own private benefits, which reduces the dividend pay-out.

In much of the theoretical literature the agency theory is used to build on the theoretical framework. However for the relationship between ownership concentration and dividend pay-out, we do not find support for the agency theory of Easterbrook (1984) and Jensen’s (1986) free cash flow theory. Both theories predict a positive relationship between ownership concentration and the dividend pay-out, where we find this relationship to be negative.

(34)

time top managers come from the controlling family, which makes it easier to expropriate other shareholders. We do not have an explanation for our positive sign, but we reason that differences between existing empirical findings and our results may come from the fact that existing literature focuses on companies from countries with low shareholder protection and in our sample there is a large set of companies from the UK and Ireland. In the UK and Ireland shareholders are better protected and as a consequence this makes it harder to expropriate these shareholders.

The insignificant relationship between institutional ownership and the dividend pay-out is somewhat surprising, because all existing literature finds this relationship to be significantly positive.

When we test the influence of the type of investor in combination with the size of ownership on the dividend pay-out, we find that a combination of the institutional investor type and the size of ownership is significantly positively related to the dividend pay-out. This finding is in line with the findings of Kahn (2006), Short et al. (2002) and Allen and Michealy (2003). We have several explanations for this positive relationship, first of all institutional investors are often poor monitors and they want to be compensated for this, we reason that dividends and monitoring are often seen as substitute devices. Second a tax explanation, institutional investors pay in most countries lower tax rates on their incoming dividends, which makes it for these institutional investors attractive compared to other investors to receive higher cash dividends. The last explanation, deals with the pay-out obligation of pension funds, they have to pay monthly pensions tot their pension holders and as a consequence they may prefer high stable dividend paying companies in order to meet their own pay-out obligations.

(35)

6.1 Suggestions for further research

(36)

7. References

Allen, F., Bernardo, A. and Welch, I. (2000), A Theory of Dividends Based on Tax Clienteles, Journal of Finance, 55(6), 2499-2536.

Allen, F. and Michaely, R. (2003), Payout Policy, Working Papers, Financial Institutions Center at The Wharton School, 1-148.

Bena, J. and Hanousek, J. (2006), Rent Extraction by Large Shareholders: Evidence Using Dividend Policy in the Czech Republic, Working paper, Charles University.

Brooks, C. (2008), Introductory Econometrics for Finance, Cambridge University Press, second edition.

Bureau van Dijk (2007), BvDEP Ownership Database, Bureau van Dijk Electronic Publishing.

Denis, D. and Osobov, I. (2008), Why do firms pay dividends? International evidence on the determinants of dividend policy, Journal of Financial Economics, 89(1), 62-82. Easterbrook, F. (1984), Two Agency-Cost Explanations of Dividends, The American Economic Review, 74(4), 650-659.

Eckbo, B. and Verma, S. (1994), Managerial shareownership, voting power, and cash dividend policy, Journal of Corporate Finance, 1, 33-62.

Faccio, M., Lang, L. and Young, L. (2001), Dividends and Expropriation, American Economic Review, 91(1), 54-78.

(37)

Gugler, K. (2003), Corporate governance, dividend payout policy, and the interrelation between dividends, R&D, and capital investment, Journal of Banking & Finance, 27, 1297-1321.

Gugler, K. and Yurtoglu, B. (2003), Corporate governance and dividend pay-out policy in Germany, European Economic Review, 47(4), 731.

Jensen, M. (1986), Agency Costs of Free Cash Flow, Corporate Finance, and Takeovers, American Economic Review, 76(2), 323.

Johnson, S., La Porta, R., Lopez-de-Silanes, F. and Shleifer, A. (2000), Tunneling, American Economic Review, 90(2), 22–27.

Khan, T. (2006), Company Dividends and Ownership Structure: Evidence from UK Panel Data, Economic Journal, 116(510), C172-C189.

Mancinelli, L. and Ozkan, A. (2006), Ownership structure and dividend policy: Evidence from Italian firms, European Journal of Finance, 12(3), 265-282.

Miller, M., and Modigliani, M. (1961), Dividend policy, growth, and the valuation of shares, Journal of Business, 34, 411–433.

Renneboog, L. and Szilagyi, P. (2006), How relevant is dividend policy under low shareholder protection?, Working paper, Tilburg University.

Renneboog, L. and Trojanowski, G. (2005), Control Structures and Payout Policy, TILEC discussion paper.

(38)

Salas, J. and Chahyadi, J. (2006), A decomposition of the dividend payout ratio trend in the 20th century, Working paper, University of Oklahoma.

Short, H., Zhang, H. and Keasey, K. (2002), The link between dividend policy and institutional ownership, Journal of Corporate Finance, 8(2), 105.

Smit, M. (2009), The relationship between ownership and dividend policy. Evidence from Germany, Master thesis, University of Groningen.

Von Eije, H. and Megginson, W. (2008), Dividends and share repurchases in the European Union, Journal of Financial Economics, 89(2), 347-374.

Weinstein, D. and Yafeh, Y. (1998), On the Costs of a Bank-Centered Financial System, Journal of Finance, 53(2), 635-672.

(39)

8. Appendices

8.1 Detailed description of variables

Table 8: Detailed description of variables

Variable Formula Components Source Dividend Cash dividend 2007/2008 divided

by total assets 2007/2008

Cash dividends for the years 2007 and 2008 and total assets for the years 2007 and 2008

DataStream

Age - Age in years Amadeus

Size Natural logarithm of market capitalization 2006/2007

Market capitalization for 2006 and 2007

DataStream Leverage Total debt 2006/2007 divided by

total assets 2006/2007

Total debt for the years 2006 and 2007 and total assets for the years 2006 and 2007

DataStream Risk Standard deviation of net income

for the period 2006-2008 divided by total sales for the years 2006 and 2007

Net income for the period 2006-2008 and total sales for the years 2006 and 2007

DataStream

Cash holdings Total cash 2006/2007 divided by total assets 2007/2008

Total cash for the years 2006 and 2007 and total assets for the years 2006 and 2007

DataStream Growth Increase in total assets from

2005/2006 to 2006/2007 divided by total assets 2005/2006

Total assets for the period 2005-2007

DataStream Country dummy9 Dummy variable which takes on a

value of 1 when a company is headquartered in the country as stated and 0 otherwise

Country where a company is headquartered

Amadeus

Total ownership Total ownership (direct and indirect) largest shareholder for 2007 and 2008

Total ownership largest shareholder for 2007 and 2008

Amadeus Relative direct

ownership

Direct ownership 2007/2008 divided by total ownership 2007/2008

Direct ownership largest shareholder and total ownership largest investor for 2007 and 2008

Amadeus

Institutional Dummy variable which takes on a value of 1 when the largest shareholder is a institutional investor for 2007 and 2008 and 0 otherwise

Investor type largest shareholder for 2007 and 2008

Amadeus

Family Dummy variable which takes on a value of 1 when the largest shareholder is a family for 2007 and 2008 and 0 otherwise

Investor type largest shareholder for 2007 and 2008

Amadeus

Institutional * Total ownership

Institutional dummy 2007/2008 multiplied by total ownership 2007/2008

Institutional dummy and total ownership for 2007 and 2008

Amadeus

(40)

Family * Total ownership

Family dummy 2007/2008 multiplied by total ownership 2007/2008

Family dummy and total ownership for 2007 and 2009

Amadeus UK and Ireland Dummy variable which takes on a

value of 1 when a company is headquartered in the UK or Ireland and 0 otherwise

Country where a company is headquartered

(41)

8.2 Descriptive statistics

(42)

8.3 Ownership data van Dijk

The Ownership Database intends to track control relationships rather than patrimonial relationships. This is why, when there are 2 categories of shares split into Voting/Non voting shares, the percentages that are recorded are those attached to the category Voting shares.

A link between two entities is indicated even when the percentage is very small (less than 1%) or unknown.

8.3.1 Direct Ownership (Bureau van Dijk, 2007)

A link indicating that entity A owns a certain percentage of Company B is referred to as a

direct ownership link. The example shows 60% Direct ownership of A in B.

8.3.2 Indirect Ownership (Bureau van Dijk, 2007)

It is possible that a source gives both a direct and an indirect percentage (for example 20% and 40%).

(43)

8.3.3 Total Ownership (Bureau van Dijk, 2007)

In some cases the information source indicates that entity A has a total stake in Company B without specifying the path through which the ownership is held.

This example shows a 60% Total ownership. Total ownership links are pictured with jagged lines.

This case can result from infinite situations involving three or more companies. In the examples below, direct links to and from company C are shown as dotted lines to indicate that the link is not recorded. The only known fact is that A indirectly owns 60% of B as shown above.

Even if these direct links between A and C and between C and B were known, BvD (Bureau van Dijk) would not compute the weighted average of company A's stake in B to calculate the Total ownership figure. This is because BvD cannot assert for sure that other indirect links between A and B do not exist which would impact the calculation.

(44)

Table 9: Ownership concentration and dividend pay-out UK and Ireland

Year 2007 2008

Independent variable Coefficient Prob. Coefficient Prob. TOTAL OWNERSHIP -0.019 0.007 -0.013 0.051 RELATIVE DIRECT OWNERSHIP 0.004 0.198 -0.002 0.564 UK & IRELAND 0.004 0.266 0.003 0.329 AGE 0.000 0.048 0.000 0.065 SIZE 0.019 0.000 0.019 0.000 RISK -0.037 0.000 -0.031 0.001 CASH 0.027 0.005 0.033 0.001 DEBT -0.050 0.000 -0.043 0.000 GROWTH 0.000 0.383 -0.002 0.648 C -0.141 0.000 -0.135 0.000 Number of firms 1508 1510

(45)

Table 10: Investor type, size of ownership and dividend pay-out UK and Ireland

Year 2007 2008

Independent variable Coefficient Prob. Coefficient Prob. INSTITUTIONAL * TOTAL OWNERSHIP 0.031 0.073 0.024 0.132 FAMILY * TOTAL OWNERSHIP 0.025 0.184 0.031 0.071 INSTITUTIONAL -0.007 0.279 -0.005 0.414

FAMILY -0.002 0.802 -0.002 0.783

TOTAL OWNERSHIP -0.030 0.004 -0.024 0.017 RELATIVE DIRECT OWNERSHIP 0.004 0.322 -0.002 0.664 UK & IRELAND 0.005 0.204 0.004 0.281 AGE 0.000 0.023 0.000 0.031 SIZE 0.020 0.000 0.020 0.000 RISK -0.036 0.000 -0.029 0.001 CASH 0.027 0.006 0.033 0.001 DEBT -0.053 0.000 -0.046 0.000 GROWTH 0.000 0.386 -0.002 0.511 C -0.146 0.000 -0.144 0.000 Number of firms 1508 1510