Does legal system determine capital structure?

Does legal system determine capital structure?

An Analysis of two common law countries and four civil law countries for 2002-2006

Abstract

Table of content

1. Introduction 4

2. Literature review 6

2.1 Pecking order theory 6

2.2 Trade-off theory 7

2.3 Legal system and capital structure 8

2.4 Related empirical research 10

3. Hypothesis 12

3.1 Hypothesis 12

3.2 Dependent variable 13

3.3 Determinants of capital structure 13

3.4 Control variables 16

4. Data 18

4.1 Data collection 18

4.2 Measurements of the variables 19

4.3 Descriptive statistics 20

4.4 Correlation diagrams 22

5. Methodology 25

5.1 Panel data 25

5.2 Pooled regression model 25

5.3 Fixed effects model 26

5.4 Random effects model 27

5.5 The Regression 28

5.6 The Wald test 29

6. Results 30

6.1 Results for civil law countries and common law countries 30

6.2 Results per country 33

6.3 Results subsamples 35

7. Conclusion 37

1 Introduction

Ever since Miller & Modigliani (1958) wrote their seminal work on capital structure, arguing that under strong assumptions firm value is not affected by the way a firm is financed, many researchers have come up with theoretical explanations for why a firm’s capital structure does affect firm value (Harris and Raviv, 1991; Jensen and Meckling, 1976; and Myers, 1984). In addition, many researchers found empirical evidence that capital structure affects firm value (Demirgüς-Kunt and Maksimovic, 1994; Shyam-Sunder and Myers, 1994; and Titman and Wessels, 1988). Other researchers show that capital structure is not only affected by firm-specific determinants but by country-specific determinants as well (Demirgüς-Kunt and Levine, 1999; Demirgüς-Kunt and Maksimovic, 1998; La Porta et al., 1997; La Porta et al., 1998; and Ryan and Zingales, 1995). In particular, La Porta et al. (1997) state that legal system is an important determinant of capital structure. Legal system may explain differences in capital structure across countries. In this thesis I investigate whether legal system determines capital structure.

In relation to the capital structure of the firm, two commonly cited financing theories emerged, namely the trade-off theory and the pecking order theory, both introduced by Meyers (1984). The trade-off theory assumes that a firm’s optimal capital structure is based on a trade-off between the present value of tax benefits of debt and the costs of debt. The costs of debt consist of higher financial distress and agency problems. On the other hand, the pecking order theory assumes a certain preferred order of financing due to asymmetric information problems between a firm’s management and the investors. Furthermore, this theory suggests that there is no optimal capital structure. The costs and benefits of debt are assumed second order. Changes in debt ratios are driven by the need for external funds, not by any attempt to reach an optimal capital structure (Shyam-Sunder et al., 1999). Both theories have shown to be valuable in explaining capital structure (Beattie et al., 2006; Booth et al., 2001; and Fama and French, 2002).

More precisely, I hypothesise that the characteristics of common law countries create conditions that are in line with the assumptions of the trade-off theory and that the characteristics of civil law countries create conditions that are in line with the assumptions of the pecking order theory. As both financing theories are valid, given the main assumptions of the individual theories, differences in legal systems may explain why in some countries firms tend to adjust their capital structure in line with the pecking order theory and why in other countries firms tend to adjust their capital structure in line with the trade-off theory. This research builds on prior research of Demirgüç-Kunt and Maksimovic (1998) and La Porta et al. (1997). The main hypothesis of this research follows from the combination of the theoretical frameworks and results of these two articles. Firms in civil law countries have more difficulties in finding external financing due to a smaller capital market. They tend to suffer more from asymmetric information problems because of weaker investor protection; hence, firms in civil law countries can be expected to have a very strong preference order with respect to the way of financing. On the other hand, firms in common law countries have easier access to external financing due to a larger capital market. As such, gains from tax benefits of debt play a more dominant role in financing decisions and a trade-off between the benefits of debt and the costs of debt is expected. Differences in legal systems can be linked to the differences in the assumptions of both financing theories.

I aim to jointly test the trade-off theory and pecking order theory. To test this combined theory I investigate two common law countries and four civil law countries, which are all prime examples of their legal system. A panel dataset for the time period 2002-2006 of 2.670 firms and 13.350 firm-year observations is constructed. I apply random effects regressions and perform a Wald test on multiple coefficients to test the main the hypothesis. I find a strong indication that legal system determines which financing theory prevails. As the financing theories are based on assumptions of capital structure, my results therefore indicate that legal system determines capital structure. Most of the results are in line with prior research, which supports my main hypothesis. In addition, the results show that operational cash flow, dividend and growth are unimportant determinants of capital structure. The results also confirm the statement of La Porta et al. (1997) that legal system is an important determinant of capital structure. However, further research should determine whether my indication is justified.

2 Literature review

The two commonly cited financing theories are described in the two following sections because they are the main theories of capital structure choice. These sections are followed by a section about legal systems and capital structure. The three sections form the theoretical foundation of this thesis. The final section discusses the results of empirical research into capital structure over time.

2.1 Pecking order theory

The pecking order theory suggests that there is no optimal capital structure. Firms prefer some finance options above others mainly due to asymmetric information problems between firms and investors. The costs and benefits of debt are assumed second order. Changes in debt ratios are driven by the need for external funds, not by any attempt to reach an optimal capital structure (Shyam-Sunder et al., 1999).

Myers (1984) argues that, under asymmetric information, equity may be mispriced by the market. Managers use private information to issue equity when it is overvalued. Investors are aware of this asymmetric information and will discount the firm’s existing and new equity when new issues are announced. When firms seek finance by issuing equity, underpricing may be so severe that new investors gain more of the projects’ net present value (NPV) than existing shareholders do. This may lead to the “underinvestment problem” because projects are rejected even when they have a positive NPV (Fama and Miller, 1972). Issuing securities that are less likely to be mispriced by the market can reduce the underinvestment problem. Because information asymmetries are not present with a firm’s retained earnings, no undervaluation exists. Therefore, retained earnings are first in the pecking order. When external financing is necessary, debt is less likely to be mispriced than equity due to less asymmetric information problems between managers and debtholders than between managers and equityholders, which justifies the pecking order (Myers and Maljuf, 1984). To minimize asymmetric information costs firms finance projects with retained earnings first, then with debt and last with equity.

2.2 Trade-off theory

In contrast to the pecking order theory, the trade-off theory suggests that there is an optimal capital structure, achieved by a trade-off between the benefits of debt and the costs of debt (Myers, 1984). The benefits of debt are mainly the tax deductibility of interest payments. An increase in a firm’s leverage brings down the tax liability and raises the free cash flow available to investors. The tax shield is therefore an important determinant of leverage for the trade-off theory. The costs of debt exist of higher financial distress as leverage increases and agency problems. Agency problems can be divided into the free cash flow problem, the asset-substitution effect and the debt overhang problem. Agency problems involve asymmetric information, but it is of less relevancy for the trade-off theory than for the pecking order theory (Myers, 1984).

The free cash flow problem arises when a conflict of interest between shareholders and managers exist (Jensen and Meckling, 1976). This conflict of interest arises when managers hold less than 100% of the equity. Managers then bear all the costs of wealth creation for its shareholders but do not earn the entire wealth. This leads to incentives for the management to invest sub optimally, for example by consuming “perquisites” such as corporate jets or “empire building”. This is called the free cash flow problem and leads to value destruction but can be reduced through higher leverage. Holding constant the managers’ absolute investment in the firm, increasing the fraction of the firm financed by debt increases the managers’ share of equity and mitigates the loss in value from the conflict between managers and equityholders. Moreover, in a more levered firm the higher debt level means a higher fixed claim on the firm’s free cash flow and less of the free cash flow is available to invest sub optimally (Jensen, 1986). The mitigation of the conflict constitutes the benefits of debt financing.

The second agency problem, the asset-substitution effect, arises when a conflict of interest between equityholders and debtholders exist (Jensen and Meckling, 1976). This conflict of interest arises when debt contracts give equityholders incentives to invest in risky projects. This is the case when risky projects yield large returns, well above the face value of debt, and equityholders capture most of the gain. The debtholders receive a fixed payment; returns above these payments are directed to the equityholders. If however, the project fails the debtholders bear all the costs. As a result, equityholders benefit from investing in risky projects, even when they are value decreasing. This leads to a decrease in the value of debt. This effect is generally called the asset-substitution effect (Fama and Miller, 1972).

Some of the important determinants of capital structure under the trade-off theory are important determinants of capital structure under the pecking order theory as well. However, in most cases these determinants can be expected to have opposite effects, depending on which theory prevails due to the particular market conditions. An example of such a determinant is profitability; the trade-off theory assumes that high profitability requires a higher tax shield and more debt has to be attracted to hold constant the optimal capital structure level. This implies a positive relationship between profitability and leverage. On the other hand, the pecking order theory argues that high profitability leads to high retained earnings and firms therefore need less external financing. A negative relationship between leverage and profitability is thus to be expected. The financing theories also use different determinants like e.g. capital expenditures and financial slack. The specific relationships of all the variables with leverage will be discussed in more detail in chapter 3.

Summarising, the trade-off theory assumes an optimal capital structure created by a trade-off between the benefits of debt and the costs of debt. As my hypothesis states that legal system determines capital structure legal system is explained in the next section.

2.3 Legal system and capital structure

Laws in different countries are typically not written from scratch but rather implemented voluntarily or otherwise from a few legal families or traditions (Watson, 1974). In general, commercial laws come from two broad traditions; common law, which originates form England, and civil law, which originates from the Roman empire. Civil law is classified in three major families; French, German and Scandinavian. The French and the German civil law traditions, as well as the common law tradition, have spread around the world through a combination of conquest, imperialism, and more subtle imitation; the resulting laws reflect both the influence of their traditional family and the revisions specific to individual countries (Watson, 1974).

raise external financing. However, these rights depend on the legal rules of the jurisdictions in which the securities are issued (La Porta et al., 1998). Being a shareholder in France does not give an investor the same privileges as being a shareholder in the United States (La Porta et al, 1998). As common law countries offer better investor protection it suggests a larger capital market than civil law countries. La Porta et al. (1997) show that countries with more investor protection indeed do have larger capital markets.

Asymmetric information is less of an issue in common law countries as the investors are better protected, therefore firms in common law countries can hold significantly more debt and equity than firms in civil law countries. Firms can optimise their capital structure by making a trade-off between the benefits of debt and the costs of debt. It suggests that the characteristics of common law countries create conditions that are in line with the assumptions of the trade-off theory. But, as civil law countries have weaker investor protection the asymmetric information problem becomes more relevant. Hence, firms in civil law countries have a very strong preference order in the way that they are financed due to asymmetric information costs. Therefore it suggests that the characteristics of civil law countries create conditions that are more in line with the assumptions of the pecking order theory.

Common law countries tend to have market based financial systems and civil law countries are much more likely to have underdeveloped equity markets and are bank based (Demirgüç-Kunt and Levine, 1999). La Porta et al. (1999) state that bank- versus market-centeredness is not an especially useful way to distinguish financial systems. They argue that financing is a set of contracts defined by legal rights and enforcement mechanisms that enhance or reduce financing. From this perspective, a well functioning legal system facilitates the operation of debt and equity markets. It is the overall level and quality of financial services, determined by the legal system that improves the efficient allocation of resources and economic growth (Levine, 2000). La Porta et al. (1999) clearly argue that legal system is a more useful way to distinguish financial systems than focusing on bank based or market based financial system.

2.4 Related empirical research

Beattie et al. (2006) performed a survey of corporate financing decision-making of listed companies in the United Kingdom. They find that half of the responding firms use target debt levels, consistent with the trade-off theory, while 60% of the respondents claim to follow the pecking order theory. These theories are therefore not seen as mutually exclusive. Fama and French (2002) confirm the none mutual exclusivity because their results show that both models explain some of the capital structure and therefore none of the models can be rejected. Also Myers (1984) and Booth et al. (2001) support the none mutual exclusivity. An extensive discussion about which theory has the most explanatory power has not given consensus during the last two decades. Both theories have shown to be valuable in explaining capital structure choice.

Several other papers investigate the effects of country-specific determinants of capital structure as well but focus more on the legal system. Demirgüç-Kunt and Maksimovic (1998) investigate the efficiency of a country’s legal system related to capital structure. They find that the higher the legal system scores on the efficiency index, firms tend to hold more long-term debt. As La Porta et al. (1997) state, common law countries have higher investor protection and firms within these countries hold significantly more debt than firms in civil law countries. These results suggest that the characteristics of common law countries create conditions that are in line with the assumptions of the trade-off theory. Instead of focussing on the efficiency of the legal system La Porta et al. (1997) focus on the type of legal system. They suggest that legal system consists of two types, common law and civil law. Their results show that the laws of common law countries provide much more investor protection in comparison to the laws of civil law countries. Civil law countries therefore have a much smaller capital market, because an investor-unfriendly legal system does not protect investors well and reduces their willingness to finance firms. More difficult access to external financing forces firms in civil law countries to focus more on their retained earnings. This suggests that the characteristics of civil law countries create conditions that are in line with the assumptions of the pecking order theory. La Porta et al. (1997) therefore conclude that not the capital market but the legal system is the most dominant country-specific determinant affecting capital structure decisions. La Porta et al. (1998) argue that a country’s legal system shapes the economic environment and therefore affects capital structure decisions. On the other hand, several other researchers state that the shape of the economic system is largely responsible for differences in capital structure across countries (Booth et al., 2001; and Rajan and Zingales, 1995). Bancel and Mittoo (2004) extend the differentiation of legal system. They include dummy variables in their research for English, French, Scandinavian and German law countries and conclude that legal system influences managerial views on capital structure. Although there is no consensus about the significance and magnitude of country-specific determinants, it seems that legal system affects capital structure decisions.

3 Hypothesis

This chapter describes the hypothesis of my thesis and the underlying relationships of the determinants of capital structure with leverage according to the two financing theories. The control variables are described last.

3.1 Hypothesis

Over time the two financing theories could not be seen as mutually exclusive and have been tested against many potential determinants of capital structure. Demirgüç-Kunt and Maksimovic (1998) and La Porta et al. (1997, 1998) investigate legal system as a determinant of capital structure and find that firms in common law countries hold significantly more debt than firms in civil law countries because common law countries have much better investor protection. The asymmetric information problem is of less relevance in common law countries and the characteristics of common law countries seem to create conditions that are in line with the assumptions of the trade-off theory. For civil law countries the opposite is true, because of the weaker investor protection, asymmetric information is more of an issue. Furthermore, the capital market is much smaller in civil law countries, which makes raising external financing harder for firms in these countries. More difficult access to external financing forces firms in civil law countries to focus more on their retained earnings. Due to asymmetric information costs firms have a very strong preference order of the way of financing. These characteristics seem to create conditions that are in line with the assumptions of the pecking order theory. La Porta et al. (1997) conclude that not the capital market is the most dominant country-specific determinant affecting capital structure decisions but the legal system is. The assumptions of the financing theories linked to the characteristics of the civil law countries and common law countries form the foundation of my hypothesis. The results of related empirical research support the suggestion of my hypothesis.

I hypothesise that legal system determines capital structure. I argue that the characteristics of civil law countries create conditions that are in line with the assumptions of the pecking order theory and that the characteristics of common law countries create conditions that are in line with the assumptions of the trade-off theory. Therefore, my research question is: does legal system determine which financing theory prevails?

3.2 Dependent variable

Leverage is the dependent variable. All the hypothesised relationships of the determinants of capital structure described below are based on a relationship with leverage. Moreover, the financing theories are based on relationships with leverage (Demirgüς-Kunt and Maksimovic, 1994; Demirgüς-Kunt and Maksimovic, 1998; La Porta et al., 1997; and Titman and Wessels, 1988). Leverage is controlled by a few variables, namely GDP, capital market development, financial system and legal system. They are described in the last section. The predicted signs of the relationships between the variables and leverage are summarised in table 1. The relationships of the variables with leverage are described in sections 3.3 and 3.4. Table 1: Expected relations of leverage with country-specific and firm-specific determinants

Determinant Hypothesised signs

Country specific determinant

Dev. of bond market +

Dev. of stock market

-Gross domestic product +

Legal system

+/-Financial system

+/-Trade-off theory Pecking order theory Firm specific determinant (Common law) (Civil law)

Size +

-Profitability +

-Non-debt tax shields

-Tangibility +

Operating cash flow

-Capital expenditures

-Financial slack

-Dividend - +

Growth +/-

+/-Note: hypothesised signs are based on the theory of determinants of capital structure in section 3.3 and 3.4. The measurements of the variables are described in table 3, section 4.2.

3.3 Determinants of capital structure

Size

Larger firms are likely to be more diversified and have lower bankruptcy costs than smaller firms (Titman and Wessels, 1988). Furthermore, they have more assets, which can be offered as collateral, and transaction costs take a smaller proportion of the loan. This suggests that size is positively related to leverage according to the trade-off theory. On the other hand, the pecking order theory suggests a negative relationship between leverage and size. Larger firms have less asymmetric information problems because more information is available to the public (annual reports, quarterly reports, analysts following the firm). Hence, larger firms tend to issue equity sooner as it is less underpriced by the market (Rajan and Zingales, 1995).

Profitability

Many researchers have investigated the relationship between profitability and leverage. However, the sign of the relationship is still inconclusive. With the pecking order theory, a negative relationship is expected because when firms are more profitable their retained earnings will rise. As they prefer internal finance to external financing, less debt is required to make investments. On the other hand, the trade-off theory expects a positive relationship. Firms with higher profits benefit more from tax deductibility of interest payments but this requires a larger tax shield, which requires more debt.

Non-debt tax shields

The trade-off theory is based on a trade-off between the benefits of debt and the costs of debt. The benefits of debt are mainly caused by tax deductibility of interest payments of debt. This holds, however, only when the firm has enough taxable income to justify the amount of debt. The tax advantage of debt decreases when other tax deductions, such as depreciation and amortization, increase because it lowers the taxable income. Therefore a negative relationship is expected between non-debt tax shields and leverage. Bowen et al. (1982) conclude that non-debt tax shields are an important determinant of capital structure.

Tangibility

Operating cash flow and capital expenditures

The pecking order theory suggests that changes in debt ratios are driven by the need for external funds. Firstly, when firms have high dividend payouts and high capital expenditures, debt has to be attracted sooner. On the other hand, when firms have low dividend payouts and low capital expenditures, they tend to work down their debt level (Shyam-Sunder et al., 1999). Firms with high cash outlays and high capital expenditures might not be able to finance their activities internally, as preferred with the pecking order theory, and therefore require external financing in the form of debt. Secondly, firms take into account both current and future finance expenses (Myers, 1977). A high debt ratio obligates the firm to make high interest payments in the future and financial flexibility is constrained. To prevent that future capital expenditures have to be financed with equity a lower debt level is preferred. For both operating cash flow and capital expenditures, a negative relationship with leverage is expected.

Financial slack

The pecking order theory suggests that firms use their retained earnings first when finance is required. Financial slack gives a firm the possibility to make investments with internal finance when they need to. Moreover, it prevents underinvestment, as less debt is required because of a higher proportion of internal finance. A negative relationship between financial slack and leverage is expected.

Dividends

The trade-off theory predicts a negative relationship between leverage and dividend. Firstly, when firms have high dividend payout ratios, a large proportion of income is paid as dividend. Probability of not being able to pay back interest and amortization increases as well as bankruptcy costs. Firms attract less debt because of increased bankruptcy costs. Secondly, the free cash flow problem can be mitigated by dividends instead of debt. By paying out more dividends, the free cash flow available to managers reduces and less debt is available to invest sub optimally.

Growth

Growth influences capital structure, as faster growing firms require more finance. The relationship with leverage is inconclusive despite many researchers investigated the relationship. At first sight a positive relationship is expected, according to the trade-off theory, retained earnings of firms with high growth opportunities increase and firms need to issue more debt in order to sustain the target debt ratio. But a second look shows that growing firms have more agency problems (Myers, 1977). Shareholders tend to invest in very risky projects to increase their returns, but when the project fails the debtholders bear all the costs. This is called the asset-substitution effect, discussed in section 2.2. If the agency cost of debt is significant, faster growing firms tend to have more equity and less debt financing to overcome these costs. This results in a negative relationship between growth and leverage.

For the pecking order theory the relationship is also inconclusive as it is unclear whether the retained earnings of faster growing firms rise or decrease (Myers, 1977). When the retained earnings rise, less external financing is required and a negative relationship is expected. However, when retained earnings decrease, more external financing is required and, according to the pecking order theory, debt is preferred to equity and a positive relationship is expected.

3.4 Control variables

Control variables are included to control for developments outside the hands of firms. Two dummy variables are included to control for financial system and legal system. These are all country-specific determinants of capital structure.

Development of capital markets

To control for developments of the capital market, control variables of the development of the bond market and the development of the equity market are included. It is expected that leverage is positively related to bond market development. Firms have more access to debt under better terms as the bond market increases. For development of the equity market on the other hand, a negative relationship with leverage is expected. Equity markets are larger in countries with better investor protection; asymmetric information problems decrease in these countries because of the higher level of investor protection. As firms have lower cost issuing equity, firms tend to attract equity sooner (Demirgüç-Kunt and Levine, 1999).

Gross Domestic Product

Financial system

Many researchers have investigated the differences in financial systems among countries. There are broadly two classes of financial systems, bank based and market based. Although, there is no consensus about which financial system is optimal it might have some impact on a firm’s capital structure. The difference between the two systems is that market based systems especially focus on the well functioning of equity markets whereas bank based systems especially focus on financial intermediation (Levine, 2000). Therefore, the two different financial systems are expected to have an effect on a firm’s capital structure and a dummy variable for financial system is included.

Legal system

4 Data

This chapter first explains the data collection and the handling of outliers. The measurements of the variables are explained next. The descriptive statistics give some basic information of the variables. Finally, two correlations diagrams are included to indicate relationships between the variables.

4.1 Data collection

To investigate whether legal system determines capital structure I use a dataset of six countries, two common law countries and four civil law countries, over the time period 2002 through 2006. The countries in my sample are all prime examples of their legal system and hence included. I use all listed firms on the countries’ stock exchange; for the United States I use the New York Stock Exchange, for the United Kingdom the London Stock Exchange, for Japan the Tokyo Stock Exchange, for Germany the Frankfurt Stock exchange, for Sweden the Stockholm Stock Exchange and for France the Paris Stock Exchange.

To test financing theories I use determinants of capital structure also used in La Porta et al. (1997), Demirgüς-Kunt and Maksimovic (1994 and 1998) and Titman and Wessels (1988). The information of the firms’ variables is obtained by Datastream and the country-specific information is obtained by the Data Worldbank, see table 3. The measurements of the variables are explained in the next section. Initially, the sample contained 7.821 firms. From the total sample, the financial organisations and firms that did not have a complete record on the variables are excluded. The reason to exclude financial organisations is that they have different levels of leverage compared to non-financial firms. Including these financials would bias the results. The obtained sample is still biased by outliers; hence, they are deleted next. Hereby, the deleted observations are at least three standard deviations larger or smaller than the mean of the concerned variable of that country. This outlier methodology includes 99.6% of the sample but excludes all the extreme observations. In total, 2.670 firms were available resulting in a total of 13.350 firm-year observations. A summary of the data collection is presented in table 2.

Table 2: Collection of firms

Country Initial firms Exlusion financials Missing variables Outliers Firms Observations

4.2 Measurements of the variables

Following other authors of capital structure research I created a list of determinants of capital structure (Demirgüς-Kunt and Maksimovic, 1994 and 1998; La Porta et al., 1997; and Titman and Wessels, 1988). The list I created contains firm-specific determinants of capital structure and country-firm-specific determinants of capital structure, the relationships of these determinants with leverage are already explained in section 3.3 and 3.4. The measurements of the variables are presented in table 3. Most measurements are commonly used in financial research. I will explain the dependent variable leverage and the firm-specific determinant profitability because these variables are often measured differently across capital structure research.

Leverage is measured as total debt divided by total assets. According to Haung et al. (2006) this measurement is a good proxy for changes in leverage ratios within firms. For profitability I use a function of EBITDA divided by total assets. I choose EBITDA instead of net income to exclude accounting differences between countries and different industries in my sample. Firstly, some firms can expense more debt and assets in the form of interest and depreciation than others due to accounting differences. Secondly, the tax rate is not equal among all countries. Thirdly, the depreciation rate of assets is not equal among all the different industries that are included in my sample. Finally, the accounting differences across countries can lead to different net income numbers; the use of EBITDA overcomes this problem because it measures operating income before interest payments, taxes, depreciation and amortization. The measurements of the other variables are explained in table 3.

Table 3: Measurements of determinants of capital structure

Firm specific determinant Abbreviation Function Source

Leverage Lev Total debt over total assets Datastream

Size Size Natural logarithm of sales Datastream

Profitability Prof EBITDA over total assets Datastream

Non-debt tax shields NDT Sales minus EBIT over total assets Datastream

Tangibility Tang PPE over total assets Datastream

Operational cash flow OpCF Net cash flow minus operating activities of one year minus last years', over total assets Datastream Capital expenditures Capex Capital expenditures over total assets Datastream Financial slack FinSlack Difference in working capital of one year over total assets Datastream

Dividend Div Dividend over operating income Datastream

Growth Growth % change of total assets last year Datastream

Country specific determinant

Development of bond market DevBondMarket Market capitalisation of both public and private bond over the country's GDP Data Worldbank Development of stock market DevStockMarket Market capitalisation of equity over the country's GDP Data Worldbank

Gross domestic product GDP % change of GDP last year Data Worldbank

4.3 Descriptive statistics

Table 4 contains the descriptive statistics of the country-specific determinants of capital structure per country. The development of bond markets, development of stock markets and GDP growth are means over 2002 to 2006. The sample contains four civil law countries and two common law countries. The civil law group contains all three major civil law families, French, German and Scandinavian. Japan is included because it is a German derived civil law country that is highly bank based oriented. The common law countries are both market based which is in line with La Porta et al. (1997) who state that countries with better investor protection have higher developed capital markets. However, the civil law countries are divided in their financial system. Sweden is market based whereas the other civil law countries are bank based. Sweden also has a higher developed stock market than the other civil law countries. The common law countries both have highly developed stock markets but the United Kingdom has a small bond market whereas the United States has a large bond market. The civil law countries all have relatively large bond markets. The descriptive statistics of the specific determinants of capital structure are presented in table 4. Table A in the appendix presents the country-specific determinants of capital structure per country per year.

Table 4: Descriptive statistics country-specific determinants

Country Dev. of bond market Dev. of stock market Bank vs market based Legal system GDP growth %

Germany 78,00% 44,49% bank civil 0,90%

France 92,26% 81,53% bank civil 1,73%

Sweden 81,72% 107,86% market civil 3,17%

Japan 178,16% 83,67% bank civil 1,75%

United Kingdom 45,74% 136,34% market common 2,57%

United States 158,14% 132,36% market common 2,75%

Note: time period is 2002-2006.

of bond market of 45.74% for the United Kingdom, which is substantially lower than the 158.14% of the United States. The same weakness is within the civil law countries where Japan has the most firms and the highest development of bond market. But the difference between Japan and the other civil law countries is not as severe as with the common law countries. The development of stock market is high in the common law countries and low in the civil law countries. This difference is in line with the results of La Porta et al. (1997) where common law countries have higher investor protection and higher developed capital markets than civil law countries. Moreover, as the common law countries in the sample are market based their stock market is highly developed. The civil law countries are particular bank based and have relatively higher developed bond markets compared to their stock markets. The descriptive statistics of the variables per country are presented in the appendix table B.

Table 5: Descriptive statistics firm-specific and country-specific determinants

Civil law countries Common law countries

Variable Mean Median Maximum Minimum Std. Dev. Mean Median Maximum Minimum Std. Dev.

Leverage 0.21 0.18 0.94 0.00 0.17 0.24 0.24 0.96 0.00 0.17

Non debt tax shields 1.08 0.97 3.72 -0.24 0.52 1.00 0.87 5.03 -0.38 0.63

Operating cashflow 0.22 0.13 127.69 -82.33 3.78 0.06 0.11 52.77 -61.50 2.47 Profitability 0.09 0.09 0.86 -1.03 0.09 0.12 0.12 0.48 -1.78 0.13 Size 16.59 17.53 22.57 7.16 3.07 13.89 14.28 19.47 4.42 2.41 Tangibility 0.28 0.27 0.79 0.00 0.16 0.35 0.28 1.64 0.00 0.25 Capital expenditures 0.04 0.03 0.33 -0.07 0.03 0.05 0.04 0.32 -0.06 0.04 Dividend 0.15 0.12 11.29 -11.96 0.52 0.15 0.09 4.51 -2.32 0.26 Financial slack 0.19 0.18 1.04 -0.61 0.21 0.17 0.14 1.27 -0.77 0.19 Growth 0.03 0.02 1.42 -0.88 0.14 0.08 0.04 4.96 -0.74 0.24

Development of bond market 1.47 1.69 1.94 0.73 0.46 1.40 1.57 1.67 0.44 0.41 Development of stock market 0.79 0.76 1.49 0.34 0.24 1.33 1.37 1.60 1.07 0.14

Gross domestic product 0.02 0.02 0.04 0.00 0.01 0.03 0.03 0.04 0.02 0.01

Note: Descriptive statistics of firm-specific and country-specific variables of civil law firms (n= 8.130 observations) and common law firms (n=5.220 observations). Time period is 2002-2006.

4.4 Correlation diagrams

Table 6 presents the correlations between the variables for civil law countries. I discuss the most important and interesting correlations. Almost all variables correlate significantly with leverage, except operating cash flow and development of bond market. I expected a positive correlation between development of bond market and leverage because higher development of bond market enables easier access to bank loans for firms. For operating cash flow it indicates that it might not be a good determinant of capital structure choice. Size is positively correlated to leverage, whereas I expected a negative relationship. However, a correlation between two variables is not a sufficient condition to establish a causal relationship in either direction. The causal relationships of the variables with leverage will follow from my results in chapter 6. Financial slack is highly negatively correlated to leverage, which indicates that when a firm has more financial slack it will decrease its debt level. This is in line with my hypothesis that the characteristics of civil law countries create conditions that are in line with the assumptions of the pecking order theory. Tangibility is correlated to capital expenditures, which was expected because when firms spend more on capital expenditures they will receive more tangible assets. Firms tend to grow more when they are highly profitable as they are significantly correlated. As a result it can be said that faster growing firms tend to have higher retained earnings; hence, a negative causal relationship is expected between growth and leverage. The development of stock markets is highly correlated to the development of bond markets and GDP. An explanation can be found in Hackbarth et al. (2004) who state that firms tend to issue more equity in economic “booms”.

Table 7 presents the correlations between the variables for common law countries. Almost all variables correlate significantly with leverage, except operating cash flow and profitability. A positive relationship with leverage was expected for profitability as more profits require a larger tax shield and therefore more debt has to be attracted. On the other hand, non-debt tax shields are highly negatively correlated with leverage, which is in line with my hypothesis. The same holds for tangibility, which is highly positively correlated with leverage. Larger firms tend to be more profitable and spend their profits on capital expenditures to increase their tangibility. The development of stock market however, is uncorrelated with the development of the bond market whereas the civil law countries showed a highly significant

correlation. The GDP is highly correlated with the development of stock market. Again, the findings of Hackbarth et al.

5 Methodology

Several analyses are performed to test the hypothesis including Ordinary Least Squares method, robustness checks are performed to test whether the obtained results are robust cross time and cross section. This chapter will describe the regression models, the regression itself, robustness checks and the Wald test, which is added as an extra test to answer the research question.

5.1 Panel data

To measure whether the dependent variable is explained by several independent variables a regression analysis is applied. The results of this regression help to accept or reject the hypothesis. The regression method is Ordinary Least Squares (OLS). The assumptions of the financing theories are tested by several variables using balanced panel data. Panel data differs from normal time series or cross sectional analyses in a way that it blends the two characteristics together. Whereas time series regressions only measure changes over time and cross sectional regressions only measure relationships between variables at a point in time, panel regressions measure both (Brooks, 2008). This thesis analyses 2.670 firms over a period of five years, resulting in a dataset containing panel data with 13.350 firm-year observations.

Panel data has the advantage that it gives more insight into the development of variables and their relationships over time. Some benefits of using panel data are listed by Brooks (2008) and Hsiao (1985). Firstly, panel data enables to measure a broader range of issues and identify more complex effects, which would not be possible with pure time series or pure cross sectional data alone (Brooks, 2008). Secondly, panel data increases the degrees of freedom and therefore the power of the test increases as well. It also helps to mitigate the problems of multicollinearity among variables that may arise when time series are modelled individually. Hence, the efficiency of the regression is improved (Hsiao, 1985). And finally, panel data can remove the impact of certain forms of omitted variables bias in regression results by structuring the model in an appropriate way (Brooks, 2008).

5.2 Pooled regression model

The simplest way to analyse panel data is to estimate a pooled regression, where all the observations are stacked together in one fixed intercept. The model can be specified as:

it it

it x u

It is similar to an OLS regression and has the same severe limitations. These limitations are caused by the fact that the pooled regression model implicitly assumes homoskedasticity between the variables over time and across all cross sectional units in the sample. This assumption is not made by the fixed effects or random effects models, which are used in this thesis, and therefore overcomes its limitations. Both models are applied in financial research (Brooks, 2008) and will be explained in the following two sections.

5.3 Fixed effects model

The fixed effects model allows an intercept in the regression to differ cross sectional but not over time, while all of the

slope estimates are fixed both cross sectional and over time (Brooks, 2008). Instead of having one intercept

α

for the

regression there are many

α

’s that vary across the firms. The model estimates a constant for every firm that is included.

The result is a different regression line than with the pooled regression because the fixed effects model is able to observe fixed individual effects, which are unobservable by the pooled regression model. The fixed effects model uses dummy variables in the cross section and is therefore also termed the least squares dummy variable (LSDV) approach. This model identifies and controls for omitting variables, differing cross sectional but do not vary over time. The fixed effects model can be specified the same as the pooled regression (5.1) but with the disturbance decomposed into:

it i

it v

u =

µ

+ (5.2)

Where the disturbance term, u_it is decomposed into an individual specific effect, u_i and the ‘remainder disturbance’,

ν

_it

that varies over time and entities.

Rewriting equation 5.1 by substituting u_it for 5.2 leads to:

it i it

it x

y =

α

+

β

+

µ

+

ν

(5.3)

Where

α

is the common intercept. u_i is the individual effect of the disturbance term of each firm and is constant over

time but differs for each firm.

ν

_it is the remaining disturbance.

5.4 Random effects model

The random effects model is an alternative to the fixed effects model when too many variables in the fixed effects model become a problem. Degrees of freedom can be saved as they will be lost in the fixed effects model and the regression should therefore lead to a more efficient estimation. The random effects model is explained by Brooks (2008). The difference with the fixed effects model is that under the random effects model the intercepts for each cross sectional unit

are assumed to arise from a common intercept

α

and a random variable

ε

_i. The common intercept is the same for all

firms and over time, the random variable varies across firms but is constant over time.

ε

_i measures the random deviation

of each firms’ intercept term from the common intercept term

α

.

ε

_i captures the heterogeneity of the cross sectional

dimension where the dummy variables do this in the fixed effects model. The random effects model can be specified as:

it it

it x

y =

α

+

β

+

ω

ω

_it =

ε

_i+

ν

_it (5.4)

Where

α

is the common intercept. x_it is still a 1×k vector of explanatory variables.

ω

_it is the composite error term,

ε

_i

is the random variable and

ν

_it is the remaining disturbance.

The random effects model has one drawback; the new error term

ε

_i requires some restrictive assumptions.

ε

_i has zero

mean, is independent of the individual observation error term v_it, has constant variance

σ

_ε2 and is independent of the

explanatory variables x_it.

The random effects model is preferred, as it should produce a more efficient estimation than the fixed effects model.

However, it is valid only when the composite error term

ω

_it is uncorrelated with all of the explanatory variables. The fixed

effects model on the other hand assumes that the individual specific effect u_i is correlated with all of the explanatory

5.5 The regression

To test my hypothesis two separate regressions are performed, one joint estimate regression and two individual subsamples. The joint estimate regression includes all the observations for both civil law countries and common law countries. Including all observations increases the total number of observations tested and therefore the efficiency and power of the test increases as well. The individual subsample regressions are to check the former results. Both regressions use the random effects model. The joint estimate regression is specified as:

k it k it k it k it k it k

it I Size I of I NDT I Tang I OpCF I

Lev =

α

+ *

β

₁ ( )+ *

β

₂ (Pr )+ *

β

₃ ( )+ *

β

₄ ( )+ *

β

₅ ( )+ *

β

₆ k it k it k it k it k

it I FinSlack I Div I Growth I DevBondMar ket I

Capex ) * 7 ( ) * 8 ( ) * 9 ( ) * 10 ( ) * 11 ( +

β

+

β

+

β

+

β

+

β

it it k it k it k it I GDP I FinSystem I Legal rket DevStockMa )+ *

β

( )+ *

β

( )+ *

β

( )+

ω

( 12 13 14 (5.5)

Where I is the dummy variable legal system, respectively civil law countries and common law countries. k is civil,

common.

ω

is the composite error term. The variables are defined in table 3, section 4.2.

Equation 5.5 does not include the variable financial system for the common law countries because including this variable gives a near singular matrix with the dummy variable legal system because the common law countries are both market based, which leads to the same dummy results as financial system. The dummy variable legal system shows the difference for leverage between civil law countries and common law countries. Equation 5.5 tries to explain the effect of all the variables on leverage. Both the pecking order theory and the trade-off theory assume certain relationships between leverage and the variables. The results show whether the characteristics of civil law countries and common law countries create conditions that are in line with the assumptions of the hypothesised financing theory.

The individual subsample regressions are divided over two groups, one for the civil law countries and one for the common law countries. These results should be similar to the results of the joint estimation regression to be robust. The dummy variable legal system is excluded because the two groups are already divided into civil law countries and common law countries. For 5.7 the dummy variable financial system is also excluded because both common law countries are market based. These two regressions explain the same as equation 5.5.

The civil law subsample regression is specified as:

) ( ) ( ) ( ) ( ) ( ) (Pr ) ( _{2 3 4 5 6 7} 1 it it it it it it it

it Size of NDT Tang OpCF Capex FinSlack

Lev =

α

+

β

+

β

+

β

+

β

+

β

+

β

+

β

+ + + + +

The results of these tests are presented in table 8 section 6.1; many robustness checks are performed in order to show that the results are robust. I performed a fixed effects regression and a random effects regression. A Hausman test for all the regressions is performed to determine which model is the most appropriate. Furthermore, these tests are performed for each country individually and subsample to see whether the regression coefficients are affected by countries with many firms in the sample. Next, subsamples of different time periods are tested to see whether some year(s) are different and last a subsample without the determinant size is tested to see if the results remain constant because size often correlates with profitability. The results of these robustness checks are in table 10 section 6.3.

5.6 The Wald test

Testing the regressions will result in overviews of whether the determinants of capital structure have a significant relationship with leverage and it will give the direction of the relationship. However, using multiple coefficients leads to a problem when not all the coefficients are in line with the assumptions of the suggested financing theory or when they are insignificant. In these cases the research question cannot be answered significantly. In other words, when for example four out of six determinants of the capital structure are significantly in line with the assumptions of the hypothesised financing theory for the common law countries or the civil law countries, it cannot be concluded that legal system determines capital structure. To overcome this problem the Wald test is used. The Wald test tests for multiple coefficients. It tests two sided whether an effect exists or not. It uses a null hypothesis; adding up the coefficients should be equal to zero. Because my hypothesis uses both negative and positive relationships I therefore correct this by reversing the determinants that have a negative hypothesised relationship with leverage. Turning all the coefficients into positive coefficients will result in a null hypothesis equal to zero.

When the null hypothesis is not significantly different from zero the hypothesis that legal system determines capital structure can be rejected. On the other hand, when the null hypothesis is significantly different from zero and the value of the added coefficients is positive it leads to another problem; the result is inconclusive. However, it indicates that legal system determines capital structure but this has not been proven by the test. This drawback of the Wald test is caused by the fact that the Wald test can give different answers to the same question due to the Wald statistic of the coefficients. A large coefficient in line with the assumptions of the hypothesised financing theory can make up for other small coefficients that are not in line with the assumptions of the hypothesised financing theory. Conventional two sided multivariate tests are not designed to test these null hypotheses implied by economic theory.

much-6 Results

This chapter discusses the most important results from the tests that are performed. The joint estimate regression and individual subsamples will be discussed first. Second, the results per country are presented and third, subsamples to check robustness are discussed.

6.1 Main results for civil law countries and common law countries

For the joint estimate regression, presented in table 8, the results look the same for common law countries as for civil law countries. This similarity will be discussed in the next section. The individual subsample regressions, also presented in table 8, lead to almost the same results as the joint estimate regression. A random effects regression is performed in order to include the dummy variables financial system and legal system while the fixed effects regression was recommended by the Hausman test. The fixed effects regression was not able to include these dummy variables, as it would lead to a near singular matrix. However, the results of both regressions are very robust and therefore the random effects model is preferred. The results of the fixed effects model are presented in the appendix table C. Most of the variables seem to be determinants of capital structure, except dividend and growth. These two variables are insignificant in all regressions for both civil law countries and common law countries. Furthermore, operating cash flow does not seem to be a determinant of capital structure as well as it is only significant at the 10% level in the individual subsample for the civil law countries while the joint estimation regression has more explanatory power.

A Wald test is performed to test for all the assumptions of the trade-off theory. The Wald test statistic is highly significant. The results are the same for the individual subsample regression of the common law countries. Therefore the hypothesis that legal system determines capital structure cannot be rejected for the common law countries. The characteristics of the common law countries seem to create conditions that are in line with the assumptions of the trade-off theory.

Focusing on the coefficients of the variables for the civil law countries also leads to an inconclusive result as not all the variables are in line with the assumptions of the hypothesised pecking order theory. Firstly, size has a positive relationship with leverage where a negative relationship was hypothesised by the assumptions of the pecking order theory. The opposite sign may be explained by the fact that tangibility has a high significant coefficient for the civil law countries as well. Large firms tend to have a higher borrowing capacity than small firms because more assets are available as collateral. Rajan and Zingales (1995) investigated the relationship between size and leverage for the G-7 countries. They find an insignificant relationship in France and Italy and a negative relationship for Germany; these countries are all civil law countries. My results deviate from Rajan and Zingales (1995) as I find a significant positive relationship between size and leverage for the civil law countries. This contradiction may be explained by the fact that Rajan and Zingales (1995) find a positive relationship for Japan, which has the most firms of the civil law countries in my sample. Titman and Wessels (1988) also investigated the relationship between size and leverage; they divided leverage into short-term debt and long-term debt where they find a negative relationship for short-term debt and a positive relationship for long-term debt. As larger firms tend to have a high borrowing capacity they therefore hold more long-term debt than smaller firms who prefer short-term debt due to higher issuing costs of long-term debt (Titman and Wessels, 1998). This may explain the positive relationship between size and leverage as my dependent variable is total debt. Secondly, profitability is in line with the assumptions of the pecking order theory, Titman and Wessels (1988) and Rajan and Zingales (1995) confirm the negative relationship between profitability and leverage. Thirdly, operating cash flow has the opposite sign but is insignificant. Fourthly, capital expenditures and financial slack are both in line with the assumptions of the pecking order theory, where financial slack seems to be an important determinant of the pecking order theory having the highest coefficient. The negative relationship between financial slack and leverage holds with prior research of Myers (1984). Fifthly, dividend is insignificant and not in line with the assumptions of the pecking order theory. Moreover, the negative relationship contradicts the results of Baskin (1989) who finds a positive relationship. And finally, growth is insignificant but in line with the assumptions of the pecking order theory because it assumes either a positive or a negative relationship.

The country-specific determinants function as control variables. The negative relationship between development of bond market and leverage is in line with the results of Demirgüç-Kunt and Levine (1999). However, the development of bond market differs from the expectation for the civil law countries explained in section 3.4 as it has a negative relationship with leverage. GDP also has the opposite sign of what was expected; therefore it contradicts the results of Hackbarth et al. (2004). GDP does not seem to be an important determinant of capital structure as it is only significant at the 10% level for the common law countries contradicting the results of Booth et al. (2001) and Fama and French (2002). Legal system significantly affects capital structure confirming the findings of La Porta et al. (1997).

Table 8: Relations between leverage with country-specific and firm-specific determinants

Panel data least squares regression of the civil law and common law countries. A random effects panel least squares regression is performed in order to include the dummy variables financial system and legal system, which is not possible with the fixed effects regression. The random effects regression is performed despite the fact that the Hausman test is significant at the 1% level (Chi-Sq. 657,64). The results however are very robust and including the dummy variables is preferred over the fixed effects model. The joint estimate regression enlarges the total observations and robustness of the results compared to individual subsamples. The individual subsamples are included to check robustness. The dependent variable is total debt over total assets.

The joint estimate regression is specified as:

Levit=α+I*β1k(Sizeit)+I*β2k(Profit)+I*β3k(NDTit)+I*β4k(Tangit)+I*β5k(OpCFit)+I*β6k(Capexit)+I*β7k(FinSlackit)+I*β8k(Divit)+I*β9k(Growthit)

+I*β10k(DevBondMarketit)+I*β11k(DevStockMarketit)+I*β12k(GDPit)+I*β13k(FinSystemit)+I*β14k(Legalit)+ωit (5.5)

Where I is the dummy variable legal system. k is civil, common. α is the common intercept, Xit is a 1*k vector of observations on the explanatory variables, ωit is the composite error term. The variables are defined in table 3, section 4.2.

The common law individual subsample regression is specified as:

Levit=α+β1(Sizeit)+β2(Profit)+β3(NDTit)+β4(Tangit)+β5(OpCFit)+β6(Capexit)+β7(FinSlackit)+β8(Divit)+β9(Growthit)+β10(DevBondMarketit)+

β11(DevStockMarketit)+β12(GDPit)+ωit (5.7)

The civil law individual subsample regression is specified as:

Levit=α+β1(Sizeit)+β2(Profit)+β3(NDTit)+β4(Tangit)+β5(OpCFit)+β6(Capexit)+β7(FinSlackit)+β8(Divit)+β9(Growthit)+β10(DevBondMarketit)+

β11(DevStockMarketit)+β12(GDPit)+β13(FinSystemit)+ωit (5.6)

Determinant Hy pothesised signs

Joint estimate resgression Individual sub samples Common law Civil law Common law Civil law

Country specific determinant 2002-2006 2002-2006 2002-2006 2002-2006

Intercept 0.06** 0.07*** 0.15***

Dev. of bond market + 0.01 -0.09*** 0.02 -0.09***

Dev. of stock market - -0.04*** -0.01** -0.04*** -0.02***

Gross domestic product + -0.42* -0.05 -0.40 -0.05

Financial system +/- 0.01 0.01

Legal system +/- 0.11***

Trade-off theory Pecking order theory Firm specific determinant (Common law) (Civil law)

Size + - 0.01*** 0.01*** 0.01*** 0.01***

Profitability + - -0.11*** -0.10*** -0.11*** -0.10***

Non-debt tax shields - -0.00 -0.02*** -0.01 -0.02***

Tangibility + 0.27*** 0.30*** 0.25*** 0.30***

Operating cash flow - -0.00 0.00 -0.00 0.00*

Capital expenditures - -0.28*** -0.01** -0.30*** -0.09**

6.2 Results per country

To see whether the overall results are different from the results per country and to see differences between the civil law families the regressions are performed for each country individually and for the whole sample. The results are presented in table 9. For the whole sample the variables are basically in line with both financing theories. Size, non-debt tax shields and tangibility are in line with the trade-off theory where size contradicts with the pecking order theory. Profitability, capital expenditures and financial slack are in line with the pecking order theory where profitability contradicts with the trade-off theory. Operating cash flow, dividend and growth are insignificant. The Wald test is performed for both financing theories and both Wald statistics are highly significant. In other words, for the whole sample the Wald test states that the use of both financing theories cannot be rejected. It indicates that the pecking order theory and the trade-off theory are both present in capital structure choice. This result confirms the results of Fama and French (2002) who state that none of the financing theories is mutually exclusive.

For the countries individually, the Wald statistic is insignificant in France, Sweden and the United Kingdom. For these countries it can be concluded the country’s characteristics do not create conditions that are in line with the assumptions of the hypothesised financing theory. However, these three countries have relatively few observations, which can influence the results. Germany, Japan and the United States have significant Wald statistics that do not reject the hypothesis, which leads to the indication that legal system determines capital structure.

The development of bond market has a positive relationship in Japan, the other civil law countries also show a positive relationship but insignificant. This result contradicts the results for civil law countries in table 8. Despite the significant negative relationship for the development of the bond market for the United States and an insignificant positive relationship for the United Kingdom, table 8 shows a positive relationship for the common law countries. The same effect holds for the development of stock market for the United States. It indicates that the conflict of interest between equityholders and debtholders is smaller in the United States compared to the United Kingdom. When the equity market is highly developed firms tend to attract equity sooner because of lower issuing costs. However, because of less asymmetric information problems more debt will be attracted according to the assumptions of the trade-off theory. This might explain the positive relationship between the development of the equity market with leverage for the United States.

The results of Japan and the United States are relatively similar to the results of the joint estimate regression in table 8 for respectively the civil law countries and the common law countries. This similarity will be discussed in the next section of subsamples.

Table 9: Relations leverage with firm-specific and country-specific determinants per country

Panel data least squares regression per country. Random effects is used as the Hausman test is insignificant for all countries. The dependent variable is total debt over total assets. The time period is 2002-2006.

The regression is specified for each country as:

Levit=α+β1(Sizeit)+β2(Profit)+β3(NDTit)+β4(Tangit)+β5(OpCFit)+β6(Capexit)+β7(FinSlackit)+β8(Divit)+β9(Growthit)+β10(DevBondMarketit)+

β11(DevStockMarketit)+β12(GDPit)+ωit (5.7)

Where α is the common intercept, Xit is a 1*k vector of observations on the explanatory variables, ωit is the composite error term. The variables are defined in table 3, section 4.2.

Determinant Whole sample France Germany Sweden Japan United Kingdom United States

Country specific determinant

Intercept 0.17*** 0.02 0.02 -0.00 -0.41*** 0.10 0.34***

Dev. of bond market -0.06*** 0.04 0.05 0,05 0.26*** 0.27 -0.30***

Dev. of stock market -0.03*** -0.04** -0.08 0.01 -0.20*** -0.07 0.13***

Gross domestic product -0.19* -0.55 -0.03 -0.94 -1.9*** -0.09 -1.87***

Financial system -0.02

Legal system -0.05***

Firm specific determinant

Size 0.01*** 0.02*** 0.01*** 0.01 0.03*** 0.00 0.02***

Profitability -0.09*** -0.05 -0.02 -0.07 -0.42*** -0.05** -0.23***

Non-debt tax shields -0.01*** -0.04*** 0.01 0.02 -0.08*** 0.02** -0.02***

Tangibility 0.27*** 0.27*** 0.36*** 0.31*** 0.15*** 0.03 0.30***

Operating cash flow 0.00 0.00 0.00 0.00 0.00 -0.00 -0.00

Capital expenditures -0.16*** 0.08 -0.07 0.24 -0.15*** 0.03 -0.31***

Financial slack -0.18*** -0.10*** -0.16*** -0.14*** -0.34*** -0.22*** -0.05***

Dividend -0.00 -0.01 -0.00 -0.01 0.00 -0.00 -0.00

Growth 0.00 0.02 0.02* 0.04 -0.03*** 0.01 -0.01