Corporate governance and dividend policy: A worldwide investigation

Corporate governance and dividend policy:

A worldwide investigation

Master thesis

by

Wouter Struyvenberg

DD MSC International Financial Management

Supervisor (IFM): J.H. Von Eije

Second Supervisor (IFM): dr. H. Gonenc

University of Groningen

Uppsala University

Date: 10-01-2014

This thesis tests two agency models of dividends, the outcome and the substitute model. I analyze a cross section of 3,825 companies from 40 countries from 2008 to 2012 with different levels of minority shareholder rights and I find support for the outcome model. Univariate analysis indicates a difference in dividend payouts in countries with different degree of shareholder protection. Regression analysis indicates that sales growth (investment opportunities) of companies in common law countries has a negative effect on dividend payouts. The tax advantage of dividends encourages dividend payouts. Finally, robustness checks, in the presence of share repurchase variable, validate the obtained results.

Key words: corporate governance, agency costs, dividends, payout policy, share repurchase JEL Classification: G3, F4, F3

1

2

II. Literature Review ... 6

2.1 Corporate governance, dividend policy and agency problems ... 6

2.2 Institutional and legal environment ... 7

2.3.1 Outcome model ... 9

2.3.2 Substitute model ... 10

2.4 Investment opportunities ... 12

2.4.1 Lower dividend payouts ... 12

2.4.2 Higher dividend payouts ... 12

2.5 Dividend tax advantage ... 13

2.6 Share repurchases ... 13 2.7 Hypotheses ... 14 III. Methodology ... 16 3.1 Data ... 16 3.2 Econometric model ... 18 IV. Results ... 21 4.1 Univariate analysis ... 21 4.2 Regressions analysis ... 24 4.2.1 Regressions ... 24 4.2.2 Common/civil regressions ... 27

4.2.3 First robustness check ... 29

4.2.4 Second robustness check ... 31

V. Discussion/Conclusion ... 34

VII. Appendices ... 37

3

I. Introduction

1.1 Intro

Although corporate financial payout policy has been a common topic of academic research and one of the most debated topic in the corporate circles for the last five decades, there are still a lot of gray areas in companies’ payout policies.

Dividend policies have been an important subject in the finance literature since the seminal study of Miller and Modigliani (1961). They argue that paying dividends is irrelevant for rational investors in perfect markets. The only way for shareholders to gain from any type of dividend policy is through market imperfections or investor behavior that increases the relevance of the dividend policy. Their dividend irrelevance theory has resulted in numerous explanations trying to solve the so-called ‘dividend puzzle’ of why firms do pay dividends.

Several theories have tried to explain corporate dividend behavior such as the signaling or the agency theories. According to the signaling theory future earnings and income taxes are believed to be key factors that the board considers when determining payouts (Bartram et al., 2010). Yet, another promising approach is based on corporate governance and more specifically, the role of dividends in companies with a different quality of shareholder protection.

4

In the beginning of the 2000’s, studies based on agency theory arguments have explored the relation between a firm’s corporate governance quality and its payout policy. La Porta, Lopez-de-Silanes, Shleifer, and Vishny (2000), hereafter referred to as LLSV, argue that the intensity of agency problems to which minority shareholders are exposed differs greatly across countries. One of the causes for this is the differences in legal protection of shareholders across countries (La Porta et al., 1997; 1998). They propose two competing hypotheses for a causal relation between shareholders rights and dividend policy, an outcome hypothesis and a substitution hypothesis. On one hand, the outcome model predicts that strong legal protection will result in better protection of minority shareholders that empower them to extract higher dividend payments in firms whereas on the other hand the substitute hypothesis predicts that weaker protection of minority shareholders will result in higher dividend payments to establish a reputation. As an example of the outcome model one may refer to 2003 when Microsoft was forced to pay dividends by their shareholders. This illustrates the ability of common law shareholders to force the disgorgement of cash by corporate managers. More recently, Comcast, a firm that had not paid dividends since 1999, announced in 2008 that it would start paying a dividend in response to shareholder desire for the return of more cash. A further implication of the outcome model is that companies with high investment opportunities in countries with strong shareholder protection have lower dividend payouts than in those countries where the level of shareholder protection is low (Bartram et al., 2010). This shows that shareholders of companies in high protection countries are willing to delay a return on their investment to a later period in time in investment prospects are good. On the other hand, the substitute model suggests that companies with growth prospects may want to establish a reputation, because they might have a higher need of external financing in the future, and thus, they may choose higher dividend payout ratios (LLSV, 2000).

Accordingly, the main research question is: ‘is there a difference in sensitivity of dividend payouts to

shareholder protection and sales growth/investment opportunities between companies in high and low shareholder protection countries?’

To find an empirical answer on the main research question this paper attempts to answer the following sub questions:

1. Does the quality of shareholder protection influence the degree of dividend payouts?

2. Is the effect of investment opportunitites on dividend payouts different in countries with a different quality of shareholder protection?

5

countries around the world for the period 2008 to 2012. To examine the impact of shareholder protection, two dummies are used that measure the quality of shareholder protection, namely the civil law (versus the common law) dummy and the low shareholder protection (versus the high shareholder protection) dummy.

In this paper, I find empirical evidence in favor of the outcome model. My univariate analysis shows that companies in common law countries (i.e. high shareholder protection) have higher dividend payout ratios than those civil law countries (i.e. low shareholder protection). Furthermore, the regression analysis (e.g. for dividend-to-earnings and for dividend-to-total-assets) shows that companies with good investment opportunities (i.e. high growth companies) in common law countries have significantly lower payouts compared to companies with poor opportunities (i.e. low growth companies). These results are found to be robust when checked in the presence of an additional control variable (share repurchases). Finally, in civil law countries with at the same time high investor protection, high growth companies have lower dividend payouts than low growth companies.

1.2 Scientific contribution

Recently, studies on the corporate dividend payout policy often focus on single countries, with varation in firm-level corporate governance (e.g., Mitton, 2004; Brockman and Unlu, 2009; Jiraporn et al., 2011, Floyd, Li, and Skinner, 2013; Hauser, 2013), whereas the focus of my investigation will be global, which will improve the generality of my findings on corporate payout decisions.

Despite the global importance of cash dividends and share repurchases, very little has been published about these subjects on a continent-wide basis (Von Eije and Megginson, 2008). Denis and Osobov (2008) do investigate dividend policies of companies in three European countries, but focus on the propensity to pay dividends and neglect share repurchases. LLSV (2000) perform a global study of 33 countries around the world but also omit share repurchases. Pinkowitz, Stulz and Williamson (2006) also have a global scope in their investigation to value cash holdings and dividend payments in 35 countries with different corporate governance regimes but again neglect share repurchases. My thesis will contribute to these global studies by focusing also on share repurchases and by investigating influences over time. I will include a time frame of five years, from 2008 until 2012, to capture changes among these years and to include potential consequences of the financial crisis on the dividend policies of companies.

1.3 Structure

6

review, which consists of six subsections. The third section goes into the methodology and will explain the research method. The actual research will be presented in the fourth section. Finally, the discussion/conclusion section will present and discuss the most important results, answer the research questions, mention limitations of this study and suggest areas for future research.

II. Literature Review

This section consists of six subsections. The first subsection is on corporate governance, dividend policy and agency problems. The second subsection is about the institutional and legal environment around the world. The next two subsections explain the two agency models of dividend policy and the impact of investment opportunities on dividend payouts. The fifth subsection elaborates on dividend tax advantage and the last subsection discusses share repurchases.

2.1 Corporate governance, dividend policy and agency problems

Mustapha and Ahmad (2011) define corporate governance as: ‘a term often used to explain the

processes and structures used to direct and manage the business activities of a company in order to enhance its shareholders’ wealth.’ According to Shleifer and Vishny (1997) the goal of corporate

governance is to make sure that the suppliers of capital or funds receive a return on their investment. A return on investment to suppliers of capital or funds can be achieved through diverse ways, like through dividends or capital gains.

7

shareholders have more incentives to monitor the management and no real dominant manager-shareholder conflict is apparent here. Shareholders are able to add value, since they do not have free-rider incentives (Gugler and Yurtoglu, 2003). In this thesis, I examine corporate governance and dividend policies across countries, thereby diverting my main focus towards country-level mechanisms.

Von Eije and Megginson (2008) mention multiple major issues addressed in dividend policy studies, like taxation, agency costs, transaction costs, asymmetric information and behavioral biases. Adjaoud and Ben-Amar (2010) argue that dividend payments can alleviate agency costs. They claim that the agency theory of dividends is relatively successful because it relates corporate governance to dividend policy. The agency theory indicates that shareholders prefer dividends to capital gains because of the possibility of expropriation by insiders (Easterbrook, 1984). Various forms of agency costs exist, like the cost of monitoring managers and risk aversion of managers. The former is costly for shareholders, if they don’t apply collective action to deal with these costs. Managers’ risk aversion suggests that risk-averse managers will select safe projects with lower expected returns compared to diversified shareholders that prefer riskier projects with higher expected returns. Therefore, the focus of my thesis will be on agency costs.

It is of critical importance to understand the difference between insiders and outsiders in widely held and controlled firms. In widely held firms, that have highly dispersed ownership structures, corporate insiders are the firm’s managers and corporate outsiders are the firm’s shareholders. In controlled firms, corporate insiders are the controlling shareholders (and the managers under their control) and the corporate outsiders are the minority shareholders. A conflict between these groups signifies that insiders can use profits for personal use such as investing in unprofitable projects, which provide them with private benefits. Therefore, outside shareholders will prefer dividends to retained earnings. In the existing literature, theories differ in the ways outside shareholders can force firms to distribute dividends. According to LLSV (2000) the key point to acknowledge here is that the failure of dividend payments results in diversion of resources and waste, with the potential of harming the interests of outside shareholders (minority shareholders). The main agency problem addressed in my thesis is therefore the conflict between insiders (e.g. managers or controlling shareholders) and minority shareholders.

2.2 Institutional and legal environment

The common law is mostly situated within the area of the Anglo-Saxon countries with a high degree of investor protection (LLSV, 1997). It is shaped primarily by judges with the purpose to resolve specific conflicts. Legal rules of civil law countries are derived from the Roman law and are understood as ‘rules

8

compared to common law. Three examples of branches of the Roman law are the German, French and Scandinavian ones (Farinha and López-de-Foronda, 2009).

Several studies indicate that the institutional and legal environment of a firm is able to influence the dividend payouts of firms. According to LLSV (2001) one possible solution to agency problems is the law. They argue that corporate and other laws foresee outside investors from a certain protection against expropriation by insiders. LLSV (2001) examined investor protection in 49 countries around the world. They point out that the degree of legal protection of investors varies greatly across countries, but investors in most countries tend to have limited rights. Particularly, common law countries tend to protect investors better than civil law countries. Especially, the French civil law has a weak protection for their investors and enforcement of laws. The German civil law and the Scandinavian countries have an intermediate protection. This is consistent with the findings of Faccio, Lang and Young (2001). They argue that in common law countries (e.g. United States), that are well regulated, have transparent financial markets and dispersed share ownership. Consequently, the agency problem is mostly between managers and shareholders. In East Asia, the predominant ownership structure is control by a family. More recent, the study of Gompers, Ishii and Metrick (2003) argues that dividend payouts are influenced by the intensity of agency costs and that these costs are related to the strength of shareholder rights. Their findings indicate a relationship between the distribution of dividends and the strength of shareholder rights.

The following analyses across countries provide empirical evidence on the interaction between dividend policy and corporate governance structures. The study of Ferris, Jayaraman and Sabherwal (2009) indicates that firms in common law countries acknowledge investors preferences for dividends more than those in civil law countries do. They argue that managers in civil law countries are influenced by controlling shareholders and have little interests in the minority shareholders.

9

They argue that in common law countries (e.g. United States) that are well regulated, have transparent financial markets and dispersed share ownership, the agency problem is between managers and shareholders. In East Asia, the predominant ownership structure is control by a family, where often one of the family members is the top manager.

More recently, the study of Gompers, Ishii and Metrick (2003) argues that dividend payouts are influenced by the intensity of agency costs and these costs are related to the strength of shareholder rights. Their findings indicate a relationship between the distribution of dividends and the strength of shareholder rights. Adjaoud and Ben-Amar (2010) have investigated this expropriation in Canada and their findings indicate that dividend payouts may serve as a device to protect investors against expropriation by management and large shareholders. More often, shareholders use their power to force managers to distribute more dividends when their rights are well protected. Otherwise managers will spend these excess cash flows on private benefits.

In this thesis, the quality of investor protection is viewed as resulting in lower agency costs and two indicators of shareholder protection are used. The first is based on a country’s legal regime, common or civil law, and the second is based on the revised country’s index of antidirector rights. I have chosen to use the revised antidirector rights index of Djankov et al. (2008)2, as several authors have criticized the original antidirector rights index of LLSV (1998) for several conceptual ambiguities and plainly errors in coding (Pagano and Volpin 2005; Spamann 2005; and Djankov et al., 2008).

2.3 Two agency models of dividends

In the literature a distinction between two agency models of dividends is made, namely the outcome model and the substitution model

2.3.1 Outcome model

The first model is based on the free cash flow hypothesis of Jensen (1986). It implies that dividend can be seen as an outcome of an effective system of legal protection of shareholders (LLSV, 2000). This imply that shareholders with stronger rights use these legal powers to extract more dividends from the firm to prevent insiders from using the earnings for private benefits. Minority shareholders can use various methods to achieve this (i.e. voting rights, monitoring). The diversion of assets gets legally riskier and costlier for insiders when investors are well protected, therefore the distribution of dividends will be

2 _{The revised antidirector rights index of Djankov (2008) not only clarifies, but also modifies some of the index}

10

more attractive for them. Thus, in this model the expected dividend policy is the outcome of the governance regime (Jiraporn, 2011). Managers of firms with strong legal protection (governance) are more likely to pay higher dividends than firms with weak legal protection (governance). I want to emphasize here that the outcome model of LLSV (2000) and Jiraporn (2011) does not depend on minority shareholders having specific rights to dividends per se, but rather on that they have more generic rights with respect to voting for directors and protection against wealth expropriation. In fact, the outcome model asserts, other things equal, the stronger the rights of minority shareholders, the more dividends they will extract from the company (LLSV, 2000).

A number of researchers have used the outcome model as the basis of their study. The paper of Faccio et al. (2001) builds on the study of LLSV (2000), and provides a comprehensive analysis of the ownership and control structure of East Asian corporations in comparison with West European corporations. They relate the dividend payout ratios to the dissimilarities between the ownership rights and control rights of the controlling shareholders. Their results show that dividend payouts are influenced to the extent that a firm’s minority shareholders are vulnerable to expropriation by controlling shareholders. The study of Bartram et al. (2010) is similar to the study of Faccio et al. (2001) because they investigate the roles of firm and country level agency conflicts. Their findings confirm the generality of the outcome model of corporate dividend payouts. The main difference with LLSV (2000) is the degree of legal protection not only differs across countries, but also agency costs differ across firm within one country. Furthermore, Mitton (2004) includes a sample of 365 firms from 19 countries and finds that firms with higher corporate governance ratings have higher dividend payout ratios. His results suggest that a shift from the worst to the best corporate governance means on average a higher dividend payout ratio. Likewise, Brown and Caylor (2004) use a governance score based on the Institutional Shareholders Services governance standards. Their findings indicate a positive link between the governance score and the dividend payouts in the United States. Finally, Farinha (2003) reports similar results for the United Kingdom using the governance guidelines of the Cadbury report.

2.3.2 Substitute model

11

Jiraporn (2011) argues that dividend payments are more important for companies with weak governance as they are more susceptible to managerial entrenchment and rationally anticipate the bigger scope of the free cash flow problem. Furthermore, dividend payments may result in the costs of not being able to pursue profitable projects and the costs of the need to come to the external market for raising funds (Officer, 2006). Companies need to establish a certain reputation to raise external funds on good terms. Officer (2006) suggests that it is more important for companies in countries with weak shareholder protection to establish a reputation of taking care of shareholders. While the need to establish such a reputation for companies in countries with strong shareholder protection is lower.. Accordingly, the substitute model argues that companies in countries with low shareholder protection will have higher dividend payments compared to countries with strong shareholder protection, other things equal (Jiraporn, 2011).

The following studies suggest that dividend can be seen as a substitute for legal protection. Jiraporn and Ning (2006) examine the relationship between dividends and shareholder rights. Their results show that companies distribute higher amounts of dividends when shareholder rights are more suppressed. A possible explanation for the difference in result with LLSV (2000) is that the latter investigates dividend policies across different legal systems, whereas Jiraporn and Ning (2006) examine dividend policies within the US legal system. Adjaoud and Ben-Amar (2010) suggest that strong corporate governance practices can also protect from the expropriation of investors, which reduces the need for dividend payments. In other words, corporate governance mechanisms could also be a substitute for dividend payouts. Following this idea, companies would determine the level of dividend payout ratios based on the level of corporate governance indicating the following relationship: companies in countries with weak corporate governance induce higher dividend payouts than those with strong corporate governance. John and Knyazeva (2006) confirm this finding. They report that the probability of dividend payouts and total payouts (i.e. share repurchase and dividend payouts) is significantly higher when measures of internal and external governance suggest weak governance.

12

2.4 Investment opportunities

2.4.1 Lower dividend payouts

The second implication of the outcome model provides an explanation for companies with investment opportunities and growth prospects in countries with strong shareholder protection that have lower dividend payouts. In some situations, shareholders with strong rights may accept lower dividend payout ratios. For example, when shareholders believe that there are profitable projects available for the excess cash (Bartram et al., 2010). Assuming that in countries with string shareholder protection, growth opportunities and dividend payout ratios are negatively related. This would suggest that shareholders would be more willing to accept a lower dividend payout ratio, when they know that their rights are well protected and believe that they will receive a share of the return from the profitable investments in the future. On the contrary, shareholders in low protection countries may try to extract dividend payments instantly without considering the investment opportunities of the firm, as shareholders are more cautious about the use of the cash and possibility of being expropriated.

The results of LLSV (2000) indicate that firms with high investment opportunities and growth prospects in high shareholder protection countries have significantly lower dividend payouts than companies with low growth firms. The paper by Mitton (2004) confirms the findings of LLSV (2000). By investigates the coefficient on the interaction term along with the main effects on corporate governance and growth, he found evidence that suggests that among firms with stronger corporate governance, the there is a negative relationship between dividend and growth opportunities. Finally, the study of Bartram et al. (2010) also finds evidence in favor of the second implication of the outcome model. They suggest that high growth firms have significantly lower dividend payouts in a situation if agency costs are low at both firm and country levels.

2.4.2 Higher dividend payouts

The substitute model also contains a second implication regarding investment opportunities. According to this implication, companies with high investment opportunities may want to establish a reputation in order to have the possibility to receive external financing in the future (LLSV, 2000). This implies that companies with high growth prospects may distribute higher dividend payouts than companies with low growth opportunities in countries with a weak legal protection of shareholders.

13

rights (e.g. preference shares, priority shares and certificates of the Dutch-style poison pills). Their results suggest that Dutch companies make few dividend payouts and these payouts do not respond to changes in earnings and have a weak relationship with investment opportunities. Therefore, these findings do not support the substitution model. LLSV (2000) also did not find support in favor of the second implication of the substitute model. Their findings indicate that investors in countries with poor legal protection appear to take any dividend payout ratio they can get without looking at the growth opportunities of companies.

2.5 Dividend tax advantage

The evidence on the effects of taxation on dividend payments is controversial and is divided into the ‘traditional’ view and the ‘new’ view (Chetty and Saez, 2005). According to the traditional view, the corporate cost of capital will decrease through lowering dividends and may lead to higher levels of investments (Poterba and Summers, 1984). A lower dividend tax rate will generate a higher the dividend tax advantage and will decrease the tax burden on taxable investors (shareholders), which in turn will purchase new equity issues because they expect dividend payouts in future. In other words, a higher dividend tax advantage (i.e. lower dividend tax rate) encourages dividend payouts. The results of Poterba (2004) further indicate that dividends in the United States react to alterations in the relative tax burden on dividends and capital gains, which is consistent with the ‘traditional view.’

On the contrary, the ‘new’ view on dividend taxation assumptions that marginal investments are fully financed by retained earnings instead of new share issues (Chetty and Saez, 2005). This assumption encompasses that investment decisions of companies are not affected by dividend tax and profits, therefore dividend payments should not alter either. Therefore, the dividend tax cut is irrelevant for corporate decisions and will only result in private gains for shareholders by lowering their tax burden (Poterba, 2004). In this thesis, a measure of the tax advantage of dividends is included to estimate the effect of taxes on various dividend policies around the world.

2.6 Share repurchases

14

in an attempt to increase earnings per share or time the equity market. Several studies in the existing literature demonstrate positive effect of repurchase announcements compared to increases or initiation of dividend payouts (Dann, 1981; Vermaelen, 1981; Ikenberry, Lakonishok and Vermaelen, 1995; Guay and Harford, 2000; Jagannathan, Stephens, and Weisbach, 2000; Weston and Siu, 2003; Maxwell and Stephens, 2003; Grullon and Michaely, 2004).

Skinner (2008) even claims that shareholders prefer share repurchases over dividend payouts in the United States. Hisresults show a higher total annual value of share repurchases compared to dividend payouts. The results of John and Knyazeva (2006) indicate a mixed effect of corporate governance on share repurchases. Strong external corporate governance motivates managers to increase share repurchases and to use this as the only form of payout, while strong internal corporate governance decreases the necessity for repurchases.

Von Eije and Megginson (2008) examine cash dividends and share repurchases in the European Union. They found that firms headquartered in a common law country have a higher chance of paying cash dividends. However, they are not more likely to repurchase shares than firms headquartered in a civil law country. As argued by the outcome model, shareholders in high protection countries are more capable of extracting dividends from firms by using their legal powers. This suggest that they can dictate the payout form they prefer, such as devoted dividend payments. Bartram et al. (2010) argue that dividend payouts bring more obligations than share repurchase, because shareholders demand managers to keep a stable dividend policy rather than a stable repurchase policy.3 Officer (2006) confirms this by arguing that dividend payouts are implicit pre-commitments for paying the same or larger amounts of dividends in the future. Brown, How and Verhoeven (2008) claim that this especially applies for high protection countries.

Therefore, dividend payouts are expected to be the payout form in high protection countries. On the contrary, in low protection countries no relationship is expected because shareholders prefer any form of payout without considering the growth opportunities. Hence, share repurchases could have a negative impact on dividend payouts. Therefore, in my thesis, share repurchases will be used as a control variable in the robustness tests to validate my results.

2.7 Hypotheses

Based on the research objective of this thesis and previous research in the field the following hypotheses emerged:

3

15

Hypothesis 1:

H10: There is no difference in dividend payouts between companies in countries with high shareholder

protection and low shareholder protection

H1a-1: There are higher dividend payouts of companies in countries with high shareholder protection

compared to weak shareholder protection countries.

H1a-2: There are lower dividend payouts of companies in countries with high shareholder protection

compared to companies in weak shareholder protection countries.

The null hypothesis (H10) indicates that dividend payouts are similar between companies with a

varying degree of country based shareholder protection, while the first alternative hypothesis (H1a-1)

implies that companies with high shareholder protection (e.g. common law) have higher dividend payouts compared to those companies with weak shareholder protection (e.g. civil law). The first alternative hypothesis confirms the view of the outcome model. The second alternative hypothesis (H1a2) implies that

companies in countries with weak shareholder protection have higher dividend payouts compared to companies in countries with high shareholder protection. The second alternative hypotheis (H1a2) is in

line with the substitution model of dividends. Univariate analysis will indicate the dividend payout ratios specific for each country, the type of legal system and shareholder protection. The difference in dividend payouts will be tested for significance, difference in means (T-test) and difference in medians (Mann-Whitney-Wilcoxon Test adjusted for ties). The results of univariate tests will lead to accepting or rejecting the null hypothesis (H10).

Hypothesis 2:

H20: There is no difference in sensitivity of dividend payouts to sales growth between companies in high

and low shareholder protection countries

H2a-1: There is a higher sensitivity of dividend payouts to sales growth for companies in high shareholder

protection countries compared to companies in low shareholder protection countries

H2a-2: There is a lower sensitivity of dividend payouts to sales growth for companies in high shareholder

protection countries compared to companies in high shareholder protection countries

The null hypothesis (H20) indicates that the dividend payouts are similar for all countries regardless

of growth opportunities and level of shareholder protection. The alternative hypotheses, H2a-1 and H2a-2,

indicate a potential difference between companies in high and low shareholder protection countries due to varying investment opportunities/growth prospects. On the one hand, the first alternative hypothesis (H2

a-1) indicates a negative effect of sales growth on dividend payouts in high shareholder protection countries

16

indicate that growth/investment opportunities lower dividend payouts, and will confirm the outcome model view. On the other hand, the second alternative hypothesis (H2a-2) indicates a lower effect of sales

growth on the dividend payouts in the high protection countries, which will contradict the ourcome model view. Furthermore, a positive coefficient on the interaction variable between the sales growth and civil law dummy will indicate that for the weak shareholder protection countries, the sales growth positively influences dividend payouts. This fact will confirm the substitution model view, in which firms with better growth prospects might have higher dividend payout ratios. However, a negative coefficient on the interaction variable will indicate that sales growth has a negative impact on dividend payouts of companies in weak shareholder protection countries and will oppose the view of the subsitution model. Based on the regression analysis the null hypothesis (H20) will be accepted or rejected.

III. Methodology

3.1 Data

My sample was obtained from the Datastream database using company identifiers found in ORBIS after initial filtering, see Table I. My analysis uses data from 2008 until 2012. I select only very large and large active companies, with recent detailed financials. After, I have filtered on publicly listed firms in 56 countries that left me with 38,578 firms. From the original database, I eliminate financial firms4 with a SIC code between 6,000 – 6,999; public authorities, state, government firms5; firms without consolidated balance sheet from 2008 through 2012; firms that have no data available; firms with zero dividend payouts; firms with a negative cash flow in 2008-2012; firms with missing dividend, sales, total assets or net cash flow data in 2008-2012; firms whose total cash dividends exceed net sales. This leaves me with a basic sample of 3,825 firms.

For my analysis I have eliminated 15 countries out of initial 55, see Table I Panel B. First, have I eliminated socialist countries, such as China. Then, countries with mandatory dividend payments are eliminated6, like Brazil, Chile, Colombia, Greece and Venezuela (Martins and Novaes, 2012). Finally,

4 _{Financial firms are eliminated primarily due to their difference in capital structure and difference in accounting}

standards to which they comply.

5

Public authorities, state, government firms are eliminated due to the same reason as the financial firms (capital structure and reporting standards).

6 _{Proponents of mandatory dividend rules argue that they reduce the amount of cash available for investments,}

17

countries that do not meet the data requirements (e.g. data that contain errors, data that states ‘‘not available’’ and data that contains missing sales) are eliminated.

Table I

Construction of the Sample

Panel A: Firms in the sample

ORBIS Sample (listed companies) 38,578

Financial firms (primary and/or secondary SIC between 6,000 and 6,999) -9,724

Public authorities, state, government -3,053

Unconsolidated balance sheets in 2006-2012 -1,565

Clean data from errors -4,506

Firms with zero dividend payouts -4,655

Missing sales (2007 -2012) -1,768

Negative cash flows in 2008-2012 -5,286

Negative net income before extraordinary items before common and preferred dividends -4,159

If total cash dividends are higher than net sales -37

Basic sample 3,825

Panel B: Countries in the sample

Countries in ORBIS 55

Socialist countries (China) -1

Mandatory dividend countries (Brazil, Chile, Colombia, Greece, Venezuela) -5 Countries that do not meet data requirements (Afghanistan, Albania, Azerbaijan, -9 Czech Republic, Israel, Lebanon, Russia)

Countries in the sample 40

In my paper, I use two proxies for minority shareholders protection. First proxy is a dummy equal to one if the commercial code or the country’s firm law is civil law and zero for common law. I do not distinguish between French, German and Scandinavian law because of the limited data available on these countries. The second proxy is the low protection dummy that is equal to one if the antidirector rights index is equal or lower than three. For the analysis I use the revised antidirector rights index7 of Djankov et al. (2008), which is based on the antidirector rights index of LLSV (1998). Appendix C presents the initial and revised antidirector rights index and the construction of the dummy variable.

To accommodate differences in accounting standards across countries and changes over time, I have computed several measures of the dividend payout ratio as used in different studies (LLSV, 2000; Bartram et al., 2010). For this ratio the numerator is the same, namely total cash dividend paid to common and preferred shareholders. The denominators are cash flow from operations, earnings, sales and total

7

18

assets. According to Amidu and Abor (2006) the liquidity or cash flow position is an important determinant of the dividend payout ratio. Their results indicate that firms with good and stable cash flows are more capable of paying dividends than firms with unstable cash flows. The dividend-to-earnings ratio is the most widely used calculator of dividend payouts. However, these two ratios have several problems. First, earnings (and to some extent cash flows) can be manipulated by managers. Secondly, diversion of resources can take place before cash flows or earnings are reported, which means that they might overestimate the share of actual earnings that the firms pay as dividends (LLSV, 2000). Given this fact, I included two extra variables, specifically the dividend-to-sales ratio and the dividend-to-total-assets ratio, since these ratios are harder to manipulate and they are more comparable between countries and companies.

The past growth in sales is used to capture investment opportunities of the firm. The disadvantage of this method is that it relies on the past as a predictor of the future. However, the major advantage of sales growth is that it is relatively independent of accounting practices. I calculate for each firm the annual nominal sales growth rate using the years from 2007 to 2012.

In my analyses, I control for the tax advantage of dividends, measured as the ratio of the value of 1 US dollar distributed as dividend income divided to the value of 1 US dollar received in the form of capital gains, when kept as retained earnings. The dividend tax advantage is calculated for each country on the per year basis.8 The detailed calculation of this ratio is described in Appendix B. In my robustness analyses I also control for share repurchases.

3.2 Econometric model

In order to analyze the impact of shareholder protection and investment opportunities on the firm’s dividend payouts, several regressions are set up. For my analysis I use a panel data structure with cross-section fixed effects. 9 As the fixed effect model focuses of the within-subject variations and ignores the between-subject variation, there is a tradeoff between bias and sampling variability. Fixed effects model tend to reduce bias10 at the expense of the sampling variability. Given many possible reasons of the bias in the data and the fact that fixed effects models are generally much less restrictive than random effects

8 _{Data on taxes is obtained from sources like Ernst & Young’s Worldwide Corporate and Personal tax guide,}

KPMG’s Corporate and Individual Income Tax Rates, and PricewaterhouseCoopers’ Worldwide Summary online.

9 _{As other studies (e.g. LLSV (2000); and Bartram et al. (2010)) do the cross section random effect test, initially I}

also used random effects specification. However, the obtained results were counter-intuitive. The coefficients on the dividend tax advantage variable in the different samples were negative and significant in most of the cases.

Furthermore, the civil law and low protection dummy gave opposite results. Therefore, I use the cross section fixed effect test, which is more consistent.

19

models, and thus these models are more likely to represent the data in a realistic way11, I have chosen the fixed effect model estimation. Fixed effects regression is unable to estimate separate effects of the stable variables (i.e. country’s firm law – civil/common; country’s level of shareholder protection – low/high). Even though fixed effects models will not estimate such time-invariant covariances it is possible to estimate the interaction effect between time-invariant variables and other variables (e.g. civil law dummy and sales growth variable).

Additionally, I control for effects between years using year dummies for the years 2009-2012, with 2008 as the base year. Therefore, by adding dummies, I estimate the pure effect of the independent variables, as the dummy variables are absorbing the effects particular to each year. To analyze dividend payouts of companies several regression equations are set up. For these regressions, I use the cross section fixed effects model. The main equation, where civil law dummy (or low protection dummy) is used as proxy for shareholder protection, can be stated as follows: (1)

Where is the dependent variable for firm i in period t. It is the ratio of the total cash dividend paid divided by: (a) the net cash from operations; (b) net income before extraordinary dividends; (c) net sales; or (d) total assets. The independent variables are: (1) - the companies’ annual growth in sales over the period 2008 to 2012; (2) - the interaction between the sales growth and the civil law dummy (or the low protection dummy); (3) - the dividend tax advantage parameter of retained earnings for firms i in the countries; (4) - the year dummy for 2009; (5) - the year dummy for 2010; (6) - the year dummy for 2011; (7) - the year dummy for 2012; and (8) - the error term.

The interaction variable has the same effect as dividing the sample in two subsamples (civil

law and common law) and determining if there is significant difference in variable coefficients between two commercial codes/country’s firm laws (two levels of shareholder protection – low/high protection). A significant change in the variable coefficients between the two legal systems (two shareholder protection levels), would be indicated by a significant interaction variable. However, in the interaction models the assumption on this specification is that the estimated intercept is the same for both civil and common law. Therefore, I additionally analyse if there is a difference within the civil law dummy, by splitting the sample. Accordingly, the sample is split on civil law dummy (into common and civil law countries). As

11 _{In fact, for linear models it has been shown that random effects estimators are a special case of fixed effects}

20

the split is made on variable necessary to construct the interaction, I will not be using the interaction effect between civil law dummy and sales growth for this part as shown in equation 2:

(2)

In order to be able to validate my results and draw sound conclusions, a robustness check is performed, were, additionally, I control for share repurchase. Consequently, equation 1, for both proxies of shareholder protection (i.e. civil law dummy and low protection), changes as follows:

(3)

Where , are share repurchases.

Accordingly, I also control for the share repurchases when performing the analysis within civil law countries. Therefore, the equation 2 is adjusted as follows:

(4)

Further, it is of the interest to this study to investigate12 if there is a difference between high and low protection countries within the civil law countries. I believe that the obtained results may provide valuable insight for future research. The analysis is based on the following equation:

(5)

Where, is the interaction effect between low protection dummy and sales growth.

As in the previous cases, I also control for the share repurchases in order to check if the obtained results are robust, using the following equation:

(6)

12

21

If the results of the regression analysis, based on all of the above regressions, suggest/indicate that one of the alternative hypotheses is valid/true, then based on these results the null hypothesis (H20) will be

accepted or rejected.

Subsequent part presents the main results of the analysis.

IV. Results

This section is divided in three subsections, the univariate analysis, the regression analyses and the robustness check of the results. The first subsection presents the univariate analysis for the sample of companies analyzed. I will present the most important results of the study and provide answer to the research questions. The last subsection presents the regressions on a cross section of companies that control for the share repurchases and discusses the robustness of my results.

4.1 Univariate analysis

22

The Data

Panel A divides countries in civil and common law and presents means and medians per country. The variables are defined in table I in Appendix B. Panel B presents the test of company medians (Mann-Whitney-Wilcoxon test adjusted for ties) for civil versus common law countries whereas Panel C shows the test of company means (T-test) for civil versus common law countries.

Country N Low Div/CF Div/Earn Div/Sales Div/TA SG Div Tax

Protection Advantage

Panel A: Medians

Civil law countries

Argentina 8 1 0.190 0.295 0.020 0.020 0.140 0.662 Austria 17 1 0.200 0.410 0.020 0.020 0.060 0.800 Belgium 21 1 0.240 0.500 0.040 0.020 0.060 0.750 Denmark 20 0 0.190 0.320 0.020 0.025 0.040 0.778 Egypt 29 1 0.650 0.770 0.150 0.090 0.090 1.000 Finland 34 0 0.450 0.650 0.040 0.040 0.030 1.082 France 108 0 0.220 0.350 0.020 0.020 0.050 0.944 Germany 98 0 0.260 0.420 0.020 0.020 0.060 0.800 Hungary 4 1 0.110 0.145 0.025 0.015 0.000 0.809 Indonesia 40 0 0.340 0.440 0.050 0.050 0.190 0.921 Italy 34 1 0.260 0.455 0.030 0.020 0.070 0.891 Japan 818 0 0.140 0.300 0.010 0.010 0.120 0.923 Luxembourg 8 0 0.290 0.485 0.055 0.035 0.065 0.726 Mexico 18 1 0.225 0.305 0.030 0.030 0.090 1.000 Netherlands 27 1 0.230 0.410 0.020 0.020 0.070 1.021 Norway 17 0 0.250 0.450 0.050 0.030 0.080 0.774 Poland 25 1 0.340 0.420 0.040 0.030 0.100 0.901 Philippines 20 0 0.345 0.525 0.105 0.050 0.120 0.846 Portugal 11 1 0.180 0.560 0.020 0.020 0.040 0.813 Republic of Korea 191 0 0.130 0.190 0.010 0.010 0.100 0.970 (South Korea) Spain 30 0 0.285 0.510 0.040 0.030 0.030 0.834 Sweden 57 0 0.380 0.470 0.030 0.040 0.110 0.757 Switzerland 55 1 0.270 0.410 0.030 0.030 0.070 1.315 Taiwan 335 1 0.450 0.680 0.050 0.050 0.050 0.600 Turkey 24 1 0.415 0.530 0.060 0.055 0.085 1.009

Civil Law Country Mean 82 1 0.282 0.440 0.039 0.031 0.077 0.877

Civil Law Country Median 27 1 0.260 0.440 0.030 0.030 0.070 0.846

Common law countries

Australia 126 0 0.440 0.600 0.050 0.050 0.190 0.885 Bangladesh 5 1 0.170 0.180 0.030 0.020 0.130 0.800 Canada 92 0 0.230 0.470 0.040 0.030 0.155 1.083 Hong Kong 30 0 0.265 0.425 0.075 0.030 0.085 1.000 India 324 0 0.180 0.220 0.020 0.020 0.140 1.040 Ireland 10 0 0.210 0.330 0.015 0.020 0.105 1.050 Malaysia 161 0 0.300 0.420 0.040 0.030 0.110 1.000 New Zealand 37 0 0.380 0.730 0.050 0.040 0.110 0.956 Nigeria 6 0 0.355 0.575 0.060 0.090 0.080 0.923 Pakistan 25 0 0.310 0.310 0.040 0.030 0.100 0.900 Singapore 69 0 0.390 0.470 0.050 0.050 0.090 1.000 South Africa 60 0 0.355 0.420 0.030 0.040 0.160 0.996 Thailand 81 0 0.380 0.550 0.060 0.050 0.110 1.279 United Kingdom 221 0 0.270 0.410 0.030 0.030 0.070 0.650 United States 529 1 0.230 0.380 0.030 0.020 0.050 0.883

Common Law Country Mean 118 0 0.298 0.433 0.041 0.037 0.112 0.963

Common law Country Median 69 0 0.300 0.420 0.040 0.030 0.110 0.996

Sample Country Mean _{96 0 0.288 0.437 0.040 0.033 0.090 0.909}

23

Difference between Civil versus

Common Law 5.960*** 1.806* 11.666*** 10.634*** 6.811*** 10.130***

Panel C: Test of Company Means (T-test)

Difference between Civil versus

Common Law -3.386*** 0.241 -7.631*** -6.224*** -10.166*** -12.391***

*** and * indicate significance at the 1 and 10 percent levels, respectively.

Based on the low investor protection dummy, presented in the second column in the above table, it can be established that common law countries have better shareholder protection when compared to civil law countries. Next four columns present the medians by country of the four different payout ratios, used in this study. Based on the sample split on civil law dummy (into civil law countries and common low countries), presented in the first column in the above table, and the dividend payout ratios (except dividend-to earnings ratio), it can be established that common law countries have better shareholder protection when compared to civil law countries. The difference in medians and means are highly significant at the 1 percent significance level for all three dividend payout ratios. This result confirms the first prediction of the outcome model that states that dividends are the outcome of an effective legal protection of shareholders.13 Contradicting results for the dividend-to-earnings ratio may indicate the difference in country’s accounting standards as the vast majority of the civil law countries are complying with IFRS accounting standards. However, the compliance of the countries from the common law subsample is mixed, both IFRS and GAAP. The difference in median for this ratio is significant at the 10 percent significance level but the difference in mean is not significant. Relying on the results of the univariate tests for the three dividend payout ratios (cash flow, dividend-to sales, dividend-to-total assets), I reject the null hypothesis (H10), which states that companies with high shareholder

protection have similar dividend payouts as companies with low shareholder protection.

The country median and average values of the dividend-to-cash-flow and dividend-to-earnings of approximately 29 (and 27) and 44 (and 42) percent, for both civil and common law countries, suggest that a large share of the cash flows and earnings are distributed to the shareholders. The sample country mean and median of the dividend-to-sales and dividends-to-total-assets of around 4.0 (and 3.5) and 3.0 (and 3.3) percent respectively, reveal that around four percent of sales and three percent of the total-assets are paid out as dividends for each country.

Annual sales growth further indicates a difference between civil and common law countries. The difference in country median and mean are significant at the 1 percent confidence level, which means annual sales of the companies in common law countries grow at a higher pace compared to civil law countries.

13

24

Finally, the difference in country median and mean of the dividend tax advantage are highly significant at the 1 percent significance level. The values of the tax advantage variable indicate that the tax advantage is higher for companies in common law countries. A higher dividend tax advantage ratio denotes that the amount paid in dividends has higher value in common law than in civil law countries. In most countries, the difference between the tax treatment of dividends and retained earnings is relatively small. In countries, such as Argentina, Taiwan and the United Kingdom, the tax advantage of the dividends is relatively low compared to other countries.14

4.2 Regressions analysis

4.2.1 Regressions

Table III presents the results of the regressions on the sample of 3,825 companies in 40 countries.I use the cross section fixed effect regression and control for the dividend tax advantage that is specific for each country. I report the results for all four dividend payout ratios. I use dummies to proxy for the quality of legal protection of investors, namely the civil law and the low protection dummy. In Table III, the regressions distinguish between three different samples: (1) the combined subsamples of companies in civil and common law countries; (2) the combined subsamples of companies in low and high shareholder protection countries; and (3) the civil law subsample, which includes companies in countries with low and high shareholder protection countries.

14

25

Regression Results

Regressions with cross section fixed effects per firm. The dependent variables are the 2008-2012 values of the following four ratios: (1) dividend-to-cash flow, (2) dividend-to-earnings, (3) dividend-to-sales and (4) dividend-to-total assets. The independent variables are: (1) Sales growth, the company’s annual growth in sales; (2) IA_SG_Civil, the interaction between sales growth (SG) and civil law dummy; (3) IA_SG_Low, the interaction between sales growth and low protection dummy (4) Tax advantage of retained earnings; t-values are shown in parenthesis. Presented results account for the dividend tax advantage and year effect through dummy variables.

Dividend-to-CF Dividend-to-Earnings

Common +

Civil Low + High

Civil: low + high

Common +

Civil Low + High

Civil: low + high Sales growth 0.0948 0.053 0.017* -0.733** -0.114 0.006 (0.376) (0.751) (1.656) (-2.201) (-1.157) (0.232) IA_SG_Civil -0.063 0.859** (-0.257) (2.222) IA_SG_Low -0.084 0.107 0.677 0.849 (-0.643) (0.713) (0.930) (0.876) Tax Advantage 1.5222** 1.524*** 1.875 12.750 12.774 1.133** (2.5232) (2.691) (1.497) (1.099) (1.061) (2.423) Constant -0.967* -0.963* -1.211 -10.610 -10.701 0.042 (-1.720) (-1.850) (-1.110) (-1.013) (-0.985) (0.110) F-statistic 1.003 1.003 1.015 1.010 1.006 1.028 N 3825 3825 2049 3825 3825 2049

Dividend-to-Sales Dividend-to-Total Assets

Common +

Civil Low + High

Civil: low + high

Common +

Civil Low + High

Civil: low + high Sales growth -0.013*** -0.010** -0.009** -0.005*** -0.001*** -0.001** (-7.659) (-3.042) (-2.505) (-9.070) (-2.902) (-2.047) IA_SG_Civil 0.004 0.070*** (0.994) (2.750) IA_SG_Low 0.001 0.005 0.013 0.024* (0.123) (0.399) (1.061) (1.709) Tax Advantage 0.028 0.029* 0.083*** 0.037** 0.036** 0.103*** (1.509) (1.678) (4.272) (2.222) (2.440) (4.148) Constant 0.024 0.023 -0.030* 0.008 0.008 -0.054** (1.424) (1.540) (-1.724) (0.521) (0.559) (-2.471) F-statistic 19.664*** 19.646*** 17.706*** 14.403*** 14.461*** 16.903*** N 3825 3825 2049 3825 3825 2049

***, ** and * indicate significance at the 1, 5 and 10 percent level, respectively.

4.2.1.1 Sales growth

26

dividend payouts. For companies in the civil law subsample, there are three significant results of a different sign (i.e. a positive sign for the cash-flow and a negative sign for the dividend-to-sales and dividend-to-total-assets payout measures). Therefore, the relationship is ambiguous and should be interpreted with caution.

4.2.1.2 Interaction effect between sales growth and dummies

The most interesting results from the regression are shown with the interaction variable in Table III. Of the civil law dummy and sales growth (SG) variable only two are significant for the dividend-to-earnings and dividend-to-total-assets payout measures. Compared to the negative coefficient on the SG variable, the positive sign on coefficients the interaction variable provides interesting insight. My interpretation of the statistically significant variable is that sales growth (investment opportunities) in civil law countries has higher impact on the dividend payouts. The interacted coefficients on the sales growth variable are of the higher magnitude than without interaction, indicating an increased emphasis on the variable in the civil law countries. Furthermore, a positive coefficient on the interaction variable indicates that for the civil law countries, positive sales growth has favorable impact on the dividend payouts. Thus, civil law firms payout more if sales growth increases.

For the civil law subsample the interaction effect between sales growth and the low protection dummy is also positive and significant, but only for the dividend-to-total-assets payout measure. This suggests that sales growth for the low protection countries within civil law countries has higher impact on the dividend payouts. As in the previous case, a positive sign indicates that for the low protection countries within civil law, positive sales growth has positive impact on dividend payouts. Given the results obtained on the interaction variable between sales growth and civil law dummy (low protection dummy) I reject my null hypothesis (H20), which states that there is no difference in sensitivity of

dividend payouts to sales growth between companies in low and high shareholder protection countries.15

4.2.1.3 Tax advantage

Table III shows that for dividend-to-cash-flow and dividend-to-total-assets, the sign on the tax variable is positive and statistically significant for companies in common and civil law countries and low and high shareholder protection countries. For the civil law subsample of companies, the tax variable is also positive and significant for dividend-to-earnings and dividend-to-sales payout measures. Positive

15

27

coefficients, suggest that a higher dividend tax advantage encourages dividend payouts and can be interpreted as support for the traditional view (i.e. dividend taxes discourage dividend payouts).

4.2.2 Common/civil regressions

Further, I analyse in more detail the difference between countries with different legal origin (see Table IV). In this table I focus on the differences between common and civil law countries, as the differences between high and low shareholder protection countries are less pronounced.

Table IV

Regression Results

Regressions with cross-section fixed effects per firm. The dependent variables are the 2008-2012 values of the following four ratios: (1) dividend-to-cash-flow, (2) dividend-to-earnings, (3) dividend-to-sales and (4) dividend-to-total-assets. The independent variables are: (1) Sales growth, the company’s annual growth in sales; (2) IA_SG_Civil, the interaction between sales growth (SG) and civil law dummy; (3) Tax advantage of retained earnings; t-values are shown in parenthesis. Presented results account for the dividend tax advantage and year effect through dummy variables.

Dividend-to-CF Dividend-to-Earnings

Civil Common

Common +

Civil Civil Common

Common + Civil Sales growth 0.027 0.041 0.0948 0.082 -0.368* -0.733** (1.386) (0.177) (0.376) (0.571) (-1.940) (-2.201) IA_SG_Civil -0.063 0.859** (-0.257) (2.222) Tax advantage 1.875 1.248 1.522** 1.132*** 21.258 12.750 (1.499) (0.715) (2.532) (2.717) (1.393) (1.099) Constant -1.210 -0.767 -0.967* 0.063 -19.446 -10.610 (-1.111) (-0.459) (-1.720) (0.185) (-1.348) (-1.013) F-statistic 1.015 1.000 1.003 1.026 1.005 1.010 N 2049 1776 3825 2049 1776 3825

Dividend-to-Sales Dividend-to-Total Assets

Civil Common

Common +

Civil Civil Common

Common + Civil Sales growth -0.009** -0.012*** -0.013*** 0.001 -0.005*** -0.005*** (-2.210) (-6.151) (-7.659) (0.589) (-15.302) (-9.070) IA_SG_Civil 0.004 0.070*** (0.994) (2.750) Tax advantage 0.083*** 0.010 0.028 0.103*** 0.012 0.037** (4.275) (1.040) (1.509) (4.085) (1.482) (2.222) Constant -0.030* 0.049*** 0.024 -0.053** 0.035*** 0.008 (-1.730) (5.257) (1.424) (-2.438) (4.377) (0.521) F-statistic 17.670*** 20.596*** 19.664*** 16.193*** 13.326*** 14.403*** N 2049 1776 3825 2049 1776 3825

28

4.2.2.1 Sales growth

For both the common law subsample and the entire sample in Table IV, the coefficients on sales growth are negative and statistically significant. This is true for three out of the four dividend payout measures (i.e. dividend-to-earnings, dividend-to-sales and dividend-to-total-assets). A negative coefficient on sales growth implies that for common law countries, companies with high growth opportunities pay lower dividends compared to similar companies with low growth opportunities. These results are consistent with the outcome agency model of dividends that suggests that high growth companies should pay less dividends compared to low growth companies in particular countries with high shareholder protection, assumedly because shareholders expect to extract higher dividend payouts in the future.

4.2.2.2 Interaction effect between sales growth and civil law dummy

In Table IV, the coefficients on the interaction between sales growth (SG) and the civil law dummy are positive and statistically significant only for the entire sample for the dividend-to-earnings and dividend-to-total-assets payout measures. Ofcourse, no interaction effects are shown for the subsamples of civil and common law countries, as it is not possible to have the interaction effect for the subsample that is created on the split of the civil law dummy. My interpretation of the statistically significant variable is that sales growth (investment opportunities) in civil law countries has higher impact on the dividend payouts. The interacted coefficients on the sales growth variable are of the higher magnitude than without interaction, indicating an increased emphasis on the variable in the civil law countries. Furthermore, a positive coefficient on the interaction variable indicates that for the civil law countries, positive sales growth has favorable impact on the dividend payouts. Thus, civil law firms payout more if sales growth increases.

4.2.2.3 Tax advantage

29

4.2.3 First robustness check

Subsequently, Table V presents summary of the robustness check results of the regression analysis for the combined subsamples of companies in common and civil law, low and high shareholder protection countries and for companies in the civil law subsample. The table shows the impact of sales growth (SG) and the interaction effect between sales growth and the civil law or low protection dummy on the dividend payout policy in the presence of an additional control variable, share repurchases.

Table V

Regression Results

Regressions with cross section fixed effects per firm. The dependent variables are the 2008-2012 values of the following four ratios: (1) dividend-to-cash flow, (2) dividend-to-earnings, (3) dividend-to-sales and (4) dividend-to-total assets. The independent variables are: (1) Sales growth, the company’s annual growth in sales; (2) IA_SG_Civil, the interaction between sales growth (SG) and civil law dummy; (3) IA_SG_Low, the interaction between sales growth and low protection dummy (4) Tax advantage of retained earnings; t-values are shown in parenthesis. Presented results account for the dividend tax advantage, share repurchases and year effect through dummy variables.

Dividend-to-CF Dividend-to-Earnings

Common +

Civil Low + High

Civil: low + high Common + Civil Low + High Civil: low + high Sales growth 0.098 0.168 0.011 -0.751** -0.494 0.414 (0.341) (0.549) (0.190) (-2.091) (-1.481) (0.822) IA_SG_Civil 0.143 1.909*** (0.458) (3.872) IA_SG_Low -0.051 0.383*** 1.300** 1.014 (-0.138) (4.581) (2.407) (1.165) Tax Advantage 1.130 1.143 1.546* 14.445 14.344 1.427*** 1.604 (1.46) (1.797) (1.113) (1.093) (2.967) Share repurchase -1.450E-09 -1.580E-09 -9.59E-09 -2.240E-08 -3.150E-08 -1.22E-07**

(-0.096) (-0.104) (-1.042) (-0.517) (-0.930) (-1.977)

Constant -0.647 0.168 -0.980 -12.319 -12.174 -0.230

(-0.981) (0.549) (1.279) (-1.045) (-1.021) (-0.583)

F-statistic 0.915 0.915 0.966 0.920 0.919 0.926

N 3340 3340 1622 3340 3340 1622

Dividend-to-Sales Dividend-to-Total Assets

Common +

Civil Low + High

Civil: low + high Common + Civil Low + High Civil: low + high SG -0.013*** -0.011*** -0.010*** -0.005*** -0.005*** -0.005* (-12.165) (-8.007) (-3.665) (-4.798) (-3.829) (-1.777) IA_SG_Civil 0.009 0.019** (1.579) (2.358) IA_SG_Low 0.004 0.012 0.019* 0.035*** (0.583) (1.386) (1.912) (3.790) Tax Advantage 0.036** 0.036** 0.089*** 0.0455*** 0.044*** 0.106*** (1.971) (2.097) (5.547) (2.586) (2.597) (4.873) Share repurchase -8.880E-10*** -9.240E-10*** -2.57E-09*** -9.280E-11 -2.060E-10 -5.24E-10 (-2.592) (-2.883) (-4.210) (-0.621) (-1.247) (-1.287)

Constant 0.018 0.0180 -0.036*** -0.001 0.002 -0.057***

(1.061) (1.161) (-2.581) (-0.053) (0.101) (-3.025) F-statistic 19.723*** 19.691*** 17.178*** 14.092*** 14.082*** 16.069***

N 3340 3340 1622 3340 3340 1622

30

4.2.3.1 Sales growth

In Table V, for the dividend-to-sales and dividend-to-total-assets measures, for the combined sample of companies in common and civil law countries, low and high shareholder protection countries and for the civil law subsample, there is a negative effect of sales growth on dividend payouts. For dividend-to-earnings measure, the coefficient on sales growth is also negative and statistically significant for the companies in civil and common law countries. Thus, previously obtained results are robust in the presence of the additional control variable (share repurchases). In the three different samples, the coefficients on sales growth in Table III are also negative and significant at the 1 or 5 percent significance level for dividends-to-sales and dividends-to-total-assets. For the civil law subsample, the coefficient on sales growth for dividend-to-cash-flow is still positive but not significant.

4.2.3.2 Interaction effect between sales growth and dummies

In the presence of share repurchase, as in the case when this variable was not controlled for in Table III, only two of the interaction coefficients of civil law dummy and sales growth (SG) variable are positive and significant, for the dividend-to-earnings and dividend-to-total-assets payout measures. The positive sign of statistically significant interaction variable in Table V shows that in civil law countries, sales growth (investment opportunities) has higher impact on the dividend payouts. Thus, civil law firms payout more if sales growth increases. The interacted coefficients on the sales growth variable are of the higher magnitude than without interaction, indicating an increased emphasis on the variable in the civil law countries. Furthermore, a positive coefficient on the interaction variable indicates that for the civil law countries, positive sales growth has favorable impact on the dividend payouts.

The interaction effect between sales growth and the low protection dummy is also positive and significant, for the low and high protection sample (dividend-to-earnings and dividend-to-total-assets measures) and civil law subsample (dividend-to-total-assets payout measure). A statistically significant positive coefficient suggests that sales growth for the low protection countries (low protection countries within civil law countries) has a positive and higher impact on the dividend payouts.

4.2.3.3 Tax advantage