The influence of institutional factors on capital structure
An Empirical Analysis of the Effect of Institutional Factors on the Leverage of
Organizations in four Euro Countries for 2005-2010
Thesis
Master of Science in Business Administration Specialization: Finance
University of Groningen Faculty of Economics and Business
Sam Knabben S1625853
Abstract
In 2005, the Euro zone adapted the accounting rules of the IFRS, which made cross-national comparison possible. This study examines the influence of institutional factors on firms’ leverage throughout organizations in multiple countries for the timeframe 2005-2010, and finds a significant influence of ownership concentration, marginal tax effect, shareholder rights, and the extent to which countries are bank or market based. Also, this research analyzes the influence of the financial crisis on the effects of the factors, and finds no remarkable differences apart from a decrease in explanatory power of the model.
JEL classification G30, G32
Keywords
Index
1. Introduction and relevance of the subject 5
2. Literature review 7
2.1 Theoretical models 7
2.2 Empirical results 9
2.3 Theoretical explanation of the empirical results 12
2.3.1 Tax 12
2.3.2 Shareholder rights 14
2.3.3 Creditor rights 16
2.3.4 Ownership concentration 17
2.3.5 Market or banked based 19
3. Variable description and hypothesis stating 22
3.1 Leverage 22
3.2 Tax 23
3.3 Shareholder rights 24
3.4 Creditor rights 25
3.5 Ownership concentration 25
3.6 Bank based versus Market based structure 26
3.7 Control variables 26
3.7.1 Tangibility 27
3.7.2 Market to book ratio 27
3.7.3 Size 27
3.7.4 Profitability 27
4. Data description and descriptive statistics 29
4.1 Dependent Variables 31
4.2 Independent variables 31
4.3 Control Variables 33
4.4 Correlation between variables 34
5. Methodology 37
5.1 Regression 37
5.2 Assumptions of Least Square regression 38
5.3 Fixed or Random effects 39
5.4 Robustness 39
5.5 Financial crisis 39
6. Results 40
6.1 Testing for the assumptions for regression 40
6.2 Results of Panel Least Squares Regression 41
6.3 The influence of the financial crisis 46
7. Conclusion 50
8. Limitations 52
Overview of tables and equations
Table 1 Relevant theoretical research on capital structure 9
Table 2 Relevant empirical research on capital structure 21
Equation 1 Marginal tax effect 23
Table 3 Institutional factors and expected relationship with leverage 28
Table 4 Observations 30
Table 5 Descriptive statistics dependent variable 31
Table 6 Descriptive statistics independent variables 32
Table 7 Descriptive statistics control variables 33
Table 8 Correlation matrix all variables 35
Equation 2 Regression 38
Table 9 Least Squares regression assumption tests 41
Table 10 PLS regression results 42
Table 11 PLS regression robustness test 44
Table 12 PLS regression without crisis 46
Table 13 PLS regression robustness test without crisis 47
1. Introduction and relevance of the subject
Capital structure theory, the level of debt and equity used in an organization, has been a highly relevant and well discussed topic in finance literature for decades. It all started in 1958, when Modigliani and Miller stated that, given the investment decisions, the capital structure of a firm does not have any impact on the firm’s value. Therefore, one could argue that debt and equity are interchangeable. However, Modigliani and Miller based there statement on various underlying assumptions. For instance, capital markets are supposed to be efficient, and there is an absence of taxes, bankruptcy costs and asymmetric information. Off course in reality, it is well known that these assumptions do not hold. When Modigliani and Millers assumptions are relaxed, the choice between equity and debt suddenly becomes an important and determining factor in firm value. For instance, an increase in debt financing may
contribute to firm value due to possible tax shield advantages. On the contrary, an increase in the level of debt also comes with additional risk of becoming in financial distress and
ultimately could result in high bankruptcy costs.
A lot is written about the impact of certain variables on the optimal choice of leverage. Since many earlier studies suggest a relationship between capital structure and institutional factors, the overall aim of this research is to investigate the possible influence of the most relevant of these factors provided by literature on the leverage ratio of the firms in the dataset, and ultimately to give an indication on how firms can set their leverage ratio best, given their institutional environment. This results in the following research question:
What is the influence of institutional factors on leverage among four European countries?
In this research, the above question will be answered by testing the influence of the factors suggested by literature, namely marginal tax effect, creditor rights, shareholder rights, market or bank based countries, and ownership concentration. This will be done through panel regressions. Following Rajin and Zingales (1995), the firm specific factors profitability, market to book, size and tangibility will be used as control variables to isolate the effects.
Although there are earlier studies that have tried to examine the influence of these variables, most of them faced an undeniable problem regarding the data. Since all countries had
different accounting standards, it proved to be almost impossible to create a uniform dataset. However, in 2005 the International Financial Reporting Standards (IFRS) were introduced. This made it much easier to create a comparison between the financial data and institutional factors of different countries, and hence, perform a proper research on the influence of institutional factors on the leverage ratio of firms. This paper will focus on only four European countries namely France, Germany, Italy and the United Kingdom. Again this is done for the ease of comparison and to avoid any errors in accounting adjustments, which was one of the drawbacks of previous studies.
2. Literature review
2.1 Theoretical models
As argued in the introduction, the assumptions and constraints of Modigliani and Miller are not supposed to hold. This made capital structure suddenly a relevant factor for firm value, which made the topic much debated. The first stream of research on the capital structure dilemma that followed the initial paper by Modigliani and Miller focused mainly on the United States. The consensus in this U.S- based leverage research provided three general theories about the optimal level op debt, namely the static tradeoff model, the agency theory model, and the pecking order theory. (Myers, 1983, Jensen and Meckling, 1976).
Myers (1983) already states in the title of his paper that the determining of the optimal capital structure is a puzzle. Myers tries to explain the choices in capital structure by describing the costs and benefits of debt financing through the static tradeoff model and the pecking order theory.
As Myers argues, in the static tradeoff model, the optimal level of debt can be seen as the tradeoff of the costs and benefits of borrowing, when holding the assets and investment plans of the firm constant. Generally, this can be regarded as the firm’s attempt to balance the value of interest tax shields against the costs of financial embarrassment and distress or ultimately bankruptcy costs. In essence, the capital structure puzzle than depends on the cost of adjusting the capital structure, the marginal tax effect, and the level of financial distress costs.
This tradeoff theory, Myers continues, is in contrast with the pecking order theory, where firms are expected to base their financing decisions on a pecking order of respectively internal financing, debt financing, and equity financing. The pecking order theory is based on the law of least effort or resistance. When available, internal financing is preferred. If internal
financing is not available, and external financing must be acquired debt is favored over an equity issue. This is because management is reluctant to issue shares, because this brings in external ownership into the organization. Myers (1983) also provides another reason why equity is preferred less. He states that this is because investors feel that managers will only issue new equity in case the firm is overvalued, so that they can benefit from the
The third theory is the agency theory, described by Jensen and Meckling (1976). This theory deals with one of the agency problems, which is the lack of alignment of interests between managers and shareholders, and causes managers to misbehave. As Jensen and Meckling state, debt can be implemented as a tool for easing the conflicts between managers and shareholders, in a way of aligning their interest. For instance, the control rights in a firm that is mismanaged can be taken over by the debt holders. Also, since debt is often senior over equity in case of bankruptcy, shareholders also benefit from correct management. Debt thus reduces this type of agency problem, and more important agency costs, which increases firm value. Then again, if the level of debt becomes too high, other agency problems such as under investment problems or asset substitution can arise. The asset substitution effect states that as the level of debt increases, the management of an organization has a higher incentive to undertake projects that are far more risky, and could even have a negative net present value. Management does this, because in case of success of the project, all the upside comes to the shareholders. In an unsuccessful situation, the downside is for the debt holders. Also, the acceptance of risky and negative NPV projects reduces firm value. The underinvestment problem is the agency problem that in a situation of risky debt, debt holders are accrued with the benefits from a project instead of shareholders. Therefore, management might reject projects with a positive net present value, which of course leads to a situation with a lower firm value than if the positive NPV project was accepted.
Table I
Relevant theoretical research on capital structure
This table gives an overview of past theoretical studies on capital structure that are relevant for this research. The table gives respectively the name of the author, the year in which the original paper was published, the relevant theory, and a short description of the theoretical insight.
Authors Year Theory Description
Modigliani & Miller 1958 Capital structure
Jensen & Meckling 1976 Agency theory
Myers 1983 Static tradeoff model
Myers 1983 Pecking order theory
Under certain assumptions, firm value is unaffected by capital structure
Debt can serve as a tool in agency problems, the difference in interests between managers and shareholders
Optimal level of debt is the
tradeoff between the costs and benefits of borrowing, which are the costs of financial distress and interest tax shields advantages. Firms base their capital structure decisions on a pecking order in which they prefer internal over external financing, and debt over equity
2.2 Empirical results
Recently much is written on what determines capital structure on the international level. Therefore, cross-national comparison is done (Rajan and Zingales, 1995; Cheng and Shiu, 2007; Booth et al, 2001; Vasiliou and Daskalakis, 2009). These studies all make use of an international dataset in order to identify the determinants of capital structure, and their effect on it, in different countries.
Although not very recent, still one of the most relevant studies on capital structure
to book ratio are also acknowledged in the survey of Harris and Raviv (1991), in which they summarize the results of all relevant literature. Although the significance of these variables was proven decades ago, most recent research in the field of capital structure still uses them as control variables.
But perhaps the most valuable contribution of Rajan and Zingales (1995) to the existing capital structure literature was the acknowledgment that apart from determining the role of firm specific factors, an understanding of institutional factors would also be necessary. Such a research might lead to the fundamental determinants of capital structure. Although Rajan and Zingales did not thoroughly investigate the institutional differences between the G7 countries in their study, and what the exact effect on leverage of these differences could be, they suggested that differences in tax code, ownership structure and bankruptcy laws might have their influences in capital structure determination.
Since the paper of Rajan and Zingales (1995), great emphasis is laid on cross national
comparison and especially the differences in institutional factors and their influence on capital structure.
One example of a cross national study that followed the research of Rajan and Zingales is that of Booth et al. (2001). Booth et al. originally were trying to asses whether the capital structure theory is portable across countries that have different institutional settings. More explicit, their goal was to investigate whether the variables that had proven to have their influence on leverage in developed countries such as western Europe countries and the United States, were also significantly relevant for capital structure choices in 10 developing countries, as Brazil, Mexico and India. They find that the variables that are relevant for explaining leverage in the United States and in Europe are also of relevance in developing countries. In particular, Booth et al find evidence for a relationship between leverage and profitability, tangibility and the market to book ratio.
Although Booth et al. found significant results that concur with previous results; they remain skeptic about the outcomes. They state that, although some of the independent variables have the expected sign, their overall impact is low and the signs sometimes vary across countries. In their conclusion, they argue that this might be due to different sample sizes but that it is more likely that the observation implies that significant institutional differences affect the importance of the independent characteristics. According to Booth et al, knowing the country of origin must be considered as at least as important in determining capital structure as
done in terms of empirical research as the quality of the international databases continues to increase.
Both Rajan and Zingales and Booth et al. find that institutional factors might influence capital structure decisions. They remain vague in exactly which characteristics should be accounted for. The following years, research has been done in determining these factors.
2.3 Theoretical explanation of the empirical results
2.3.1 Tax
Antoniou et al. (2008), did research on the determinants of capital structure, including institutional factors. One of those factors is the tax scheme. This idea can be brought back several decades ago. For instance, it can be brought back to the paper about the capital
structure puzzle by Myers (1983). As argued earlier, Myers describes the static tradeoff model where the benefit of debt financing can be described as the value of the interest tax shield. As an increase in the effective corporate tax rate can increase the value of the interest tax shields, the relevant tax rates can well influence the level of leverage.
As Antoniou et al. (2008) put it, gains from borrowing increase with the rate of tax, and therefore one expects a positive relationship between leverage and the effective tax rate. However, it is argued by DeAngelo and Masulis (1980) that tax deductions for depreciation and investment tax credits can be considered as substitutes for tax benefits of debt financing.
Also, Antoniou et al. state that the effect of the tax system on capital structure choice is dependent on the tax policy objectives, for instance viewed as favoring the retention of earnings against dividend payout, which than takes the capital gains tax into consideration.
Another paper that focused on tax decades before the paper of Antoniou et al. is that of Miller (1977). In 1977, Miller issued a paper in which he reacted on the huge stream of capital structure literature that followed his initial study with Modigliani (Modigliani and Miller, 1958). He stated that, in the initial paper, they proved that in case of a full opportunity range available to investors and firms, under the ‘perfect world’ conditions, firm value is
distress. The goal of Miller (1977) was to challenge this new fashionable version of optimal capital structure.
Although Miller acknowledges the influence of agency and bankruptcy costs, he finds that the great emphasis on these topics in the (at that moment) recent discussions is misplaced. As he puts it, the supposed trade-off between tax gains and bankruptcy costs looks suspiciously like the recipe for the fabled horse-and-rabbit stew, which is one horse and one rabbit.
However, Miller does not only challenge the influence of costs of financial distress; he also challenges the influence of tax advantages of debt financing. He refers to a footnote in the original 1958 paper, which says that it should be noted that the tax system also acts in ways that reduce the gains of debt financing. In case a company relies heavy on debt, the company commits itself to paying a substantial part of income in the form of interest payments taxable to the owners through personal income tax. In contrast, a debt free company might reinvest all net income in the business and subject the owners only to the low capital rate.
When the personal income tax is taken into account along with the corporation income tax, Miller states that there is a wide range of values for these rates where the tax shield advantage vanishes entirely or can even turn negative.
Another study that shows that there is more to the influence of tax than just the tax shield advantages is that of Mayer (1990). Mayer’s purpose is to compare the industry financing in eight developed countries. One of his findings is that looking only at the tax shields
advantages of debt financing, of the eight countries in his sample Germany has the highest incentive to use debt financing while in fact they are the second lowest to use debt financing. Therefore Mayer claims tax has no explanatory power. However, Rajan and Zingales (1995) argue, that this statement has to be revised when incorporating personal taxes, as Miller already hinted in 1977. The only assumption that needs to be taken is that investors pay the taxes levied by the city or country where the primary stock exchange is located. To
2.3.2 Shareholder rights
Literature suggests that another institutional factor which could affect the leverage ratio is the level of shareholder rights, also described as investor protection (Cheng and Shiu, 2007). Shareholder rights became a topic of interest during the mid nineties. According to Hart (1995), the defining feature of various securities is the rights that they bring their owners. Hart continues that shareholders receive dividend because these shareholders have the power to vote out the directors who do not want to pay them, and creditors are paid because they have the right to repossess collateral. La Porta et al. (1998) argues that this view is incomplete, since it ignores the fact that these rights are dependent on the jurisdiction and legal system of the country the securities are issued in.
In the late nineties, both economic and legal scholars have started to investigate what the costs and benefits of differences in legal systems in theory could be, especially for investor rights. The paper of La Porta et al. (1998) was of major relevance in exploring this new territory.
Their starting point was to recognize that the law system in different countries is not written from scratch, but evolved out of one of a few legal families or traditions. There can be made a rough distinction between common law (United States and United Kingdom), and civil law, which is derived from Roman law, and can be divided in French, German and Scandinavian civil law. La Porta et al. explains that the common law family is originated from English law, and includes those laws modeled on the English Law. Common law is shaped by judges in the process of resolving specific disputes. In the resolving of these specific disputes they use preceding judicial decisions to base their decision on.
Opposite to common law, La Porta continues, there is civil law. Civil law is the oldest legal tradition as it originates from Roman law, and it has the most influence. Therefore it is spread wide around the world. In contradiction to the reliance of the verdict in preceding cases of common law, civil law uses statutes and comprehensive codes and relies heavily on legal scholars to be the ones that ascertain and also formulate the rules that are used (La Porta et al., 1998).
shareholder rights comes with certain implications, such as an influence on corporate
performance. For instance, Gompers Ishii and Metric (2003) performed a study in which they followed an investment strategy that involved buying firms with the highest shareholder rights and selling firms with the lowest shareholder rights. Their strategy yielded a positive
abnormal return of 8.5 percent over that period.
More recently, literature also suggests an influence of shareholder rights on leverage. Cheng and Shiu (2007) explain that shareholder rights could be positively correlated with the use of equity, since in countries with strong shareholder rights there is a tendency to use equity before debt because there is a higher supply of equity available. With low shareholder rights, investors are expected to buy less equity, and firms will prefer to use more debt.
In their study, Cheng and Shiu attempt to understand if, and to what extent, cross country differences in capital structure can be attributed to shareholder rights. They argue that in countries with high quality shareholder protection the supply of equity is relatively high, and firms will use more equity than debt. Meanwhile, in countries with poor shareholder
protection firms will relatively use more debt than equity. Thus, according to Cheng and Shiu, from this point of view the expected relationship between shareholder rights and leverage is negative. Furthermore, when shareholder rights are low, there is likely to be a divergence of ownership, which could lead to large agency costs. This could be solved by debt financing, which thus also contributes to the hypothesis that shareholder rights have a negative
relationship with leverage.
2.3.3 Creditor rights
Also, much has been written about the possible relationship between creditor rights and capital structure. This can be traced back to the differences between legal families.
Differences between the legal families do not only affect shareholder rights but also have their influence on creditor rights (La Porta et al., 1998). Theories of debt that are based on the transfer of control rights in case of default, especially that of Hart and Moore (1998) and Aghion and Bolton (1992), show that the higher the bargaining power of the creditors is, the more these creditors are willing to extend credit on more favorable terms. These favorable terms are for instance lower interest rates and longer maturities.
In their research of the influence of creditor rights, Quin and Strahan (2007) find results consistent with these theories. They state that when creditor rights are stronger, the willingness to extend credit on favorable terms expands the loan availability. In addition, Quin and Strahan find that weak creditor rights decreases loan availability and, more
important, the ownership concentration of loans drops dramatically. The reason they provide for this, is that diffusion of ownership increases diversification and also raises the cost of restructuring, which than reduces the incentive for borrowers to default strategically. This is consistent with the findings of Bolton and Scharfstein (1996). Quin and Strahan also find that higher creditor rights affect the collateral requirements. Increasing capacity to pledge the assets of a firm, mostly property, plant and equipment, makes collateral more effective, which ultimately creates more creditor protection and in turn enhances loan availability.
In accordance with the suggestions of Aghion and Bolton, Quin and Strahan conclude that strong creditor rights expand loan availability, due to the willingness of creditors to extend credit on favorable terms. For instance, they find a relationship between the creditor rights and the maturities of bank loans and their interest rates. This makes debt much easier to obtain. This could result in a higher use of debt. Therefore, creditor rights are also considered as an institutional factor with possible influence on leverage.
2.3.4 Ownership concentration
A fourth factor considered by literature to have a possible influence on leverage is the level of ownership concentration. Rajan and Zingales (1995), argue that one of the most observed institutional differences among the G7 countries is ownership concentration. For instance, in the United States and in the United Kingdom, ownership is much more dispersed than within countries in continental Europe, as France and Germany. Rajan and Zingales argue that this is due to the use of inter-company crossholdings and pyramiding of ownership.
De Miguel and Pindado (2001) studied the effect of both firm specific characteristics and institutional factors on leverage, in a sample of Spanish firms. One of their considerations regarding institutional factors was the concentration of ownership. De Miguel and Pindado make a clear distinction between patterns of ownership. On the one hand there is the Anglo American pattern where firms have diffused ownership. On the other hand there is the Continental European pattern, in which firms have a relatively high concentration of
ownership. La porta et al (2001), argue that the distinction between these two originated in the different legal families, as did the level of shareholder rights and creditor protection.
theory to some extent, it can be expected that there is less debt needed in order to mitigate this problem than there would be with diffused ownership.
However, theory also predicts positive influences of ownership concentration on leverage. Antoniou et al. (2008), for instance, while acknowledging the negative agency costs effect of high ownership concentration, argue that in a sample of countries with a high level of bank orientation it is likely that some of the large owners are in fact banks. They find that concentrated bank ownership can reduce costs of financial distress, because these firms are very likely to be rescued by the owning bank in case of distress. Obviously, this makes it a lot easier to obtain debt.
Furthermore, Antoniou et al. argue that there is a major downside of equity issuing for the major shareholders, in the form of dilution of shares. High levels of dilution may cause large shareholders to favor debt financing. This point also can contribute to a view wherein
ownership concentration has a positive effect on leverage.
Rajan and Zingales (1995) also found the above described contradicting theories about the influence of ownership concentration on leverage. On the one hand, they argue, when boards of directors contain large shareholders, the extent of agency costs can be reduced which facilitates equity financing. This predicts a negative relationship. However, Rajan and
Zingales also argue that it is likely that to some extent these large shareholders are banks, with a vested interest in forcing firms into borrowing from them, which lead to the expectation of a positive relationship. These contradictions are in line with the findings of De Miguel and Pindado (2001), and Antoniou et al. (2008).
2.3.5 Market or banked based
A last institutional factor debated in relevant literature is to what degree countries are market based or bank based. In finance literature, there is a fierce debate about the consequences of the structure of the financial system in a country. There is a distinction made between two financial structures, namely countries that have a market based structure or countries with a bank based structure. This separation can be traced back to Goldsmith (1969).
Demirguc-Kunt and Levine (1999) explain that within bank based structures, there is a leading role for banks in facilitating savings, the allocation of capital, control of managers’ investment projects and decisions and arranging risk management. They argue that within market based structures, on the contrary, there is a much more important role for the securities market. In such a structure the securities market shares responsibility with banks for
mobilizing savings, assuring corporate control and easing risk management. According to Demirguc-Kunt, analysts debate that markets are more efficient and effective in their role of providing these financial services.
Rajan and Zingales (1995) state that there is no doubt that there are huge differences between countries in the level of power banks have. They argue that the differences of the two
structures is most reflected in the choice between public financing in the form of stocks and bonds, and private financing in the form of bank loans. It might appear that in a bank based country the dominant role of banks assures that there is more debt financing available than in a market based country. However, studies like Diamond (1991) and Sharpe (1990) have emphasized that there can be huge costs to excessive bank debt, and thus debt financing. This might lead to a situation where there is a lot of debt financing available, but firms will only borrow until a certain point.
firms in countries that are more market based, would more easily access these markets to obtain equity funds. Similarly, firms in countries where there is stronger power of the banking sector would have more bank loans, since they are easier to obtain. Hence there might be a relationship between the extent to which countries are market- or bank based and the capital structure of firms in these countries.
Table 2
Relevant empirical research on capital structure
This table gives an overview of past empirical studies on capital structure that are relevant for this research. The table gives respectively the name of the author, the year in which the original paper was published, and a short description of the research and main findings.
Authors Year Research Main Findings
Titman and Wessels
1988 Analyzes the explanatory power of theories of optimal capital structure
Results are not conclusive but serve to document empirical regularities which are consistent with existing theory.
Harris and Raviv
1991 Paper surveys capital structure theories based on different types of models.
Survey shows that theory identified numerous possible determinants for leverage.
Rajan and Zingales
1995 Investigate the determinants of capital structure choice by analyzing the financing decisions of public firms in the major industrialized countries.
Firm leverage in G7 countries is more similar than thought earlier. However, deeper understanding of institutional environment is necesarry.
La Porta et al.
1998 Paper examines legal rules covering protection of shareholders and creditors, and their origin, through research in 49 countries.
Common law countries have the strongest legal protection whereas French civil law countries have the weakest protection.
Demirguc-Kunt and Levine
1999 Authors used newly collected data on a cross-section of up to 150 countries to illustrate how financial systems differ around the world.
Countries in common law tradition tend to be more market based, and civil law countries show more bank based economy and are usually financially underdeveloped.
Booth et al. 2001 Asses whether the capital structure theory is portable across countries that have different institutional settings.
Some modern finance insights are applicable to multiple countries. Deeper understanding of the institutional settings are however necessary.
De Miguel and Pindado
2001 Analyzes the firm characteristics which are determinants of capital structure according to different explanatory theories, and how institutional characteristics affect capital structure.
The evidence obtained confirms the influence of several insitutional factors on leverage.
Ownership concentration seems to be very important
La Porta et al.
2001 Paper describe the differences in laws and the effectiveness of their enforcement across countries.
Legal approach is more fruitful way to understand corporate governance than the classic bank vs market centered view. Gompers
Ishii and Metrick
2003 Construct a “Governance Index” to proxy for the level of shareholder rights at about 1500 large firms during the 1990s.
Firms with stronger shareholder rights had higher firm value,profits and sales growth.
Cheng and Shiu
2007 Investor protection plays an important role in the determinants of capital structure.
Frims in countries with better creditor protection use more leverage, and firms in countries with shareholder protection use equity funding.
Quin and Strahan
2007 The influence of creditor protection on characteristics of bank loans.
Under strong creditor rigts, loans have more concetrated ownership, longer maturity and lower interest rates.
Antoniou et al.
2008 The paper investigates how firms operating in capital market-oriented economies and bank-oriented economies determine their capital structure.
The degree and effectiveness of the leverage determinants are dependent on the legal system and financial
traditions of the specific country. Vasiliou
and Daskalakis
2009 Investigate whether differences in institutional characteristics result in different capital structure determination among countries.
3. Variable description and hypothesis stating
For this research, the dependent variable is leverage. The independent variables are the institutional factors tax, shareholder rights, creditor rights and ownership concentration. For the purpose of isolating the effect of these institutional factors, this research uses the control variables profitability, firm size, tangibility and market to book ratio. Panel data regression models will give an indication of the influence of these variables, where various subtests show whether the models should account for fixed or random effects. As we have seen in the literature review, these are the firm specific factors that have been identified by many studies to have their influence on leverage (i.e. Bradley et al. (1984), Castanias (1983), Long and Malitz (1985), Kester (1986), Marsh (1982) and Titman and Wessels (1988), Harris and Raviv (1991)). In answering the stated research question, the following proxies of the relevant variables will be used.
3.1 Leverage
Rajan and Zingales (1995) state that due to noticeable differences between countries in the composition of their liabilities, any research on the effect of institutional differences on capital structure must start with an appropriate and clearly formulated definition of leverage.
They believe that the broadest definition of firm leverage is seen as the ratio of the market value of the total liabilities to the market value of the total assets. In case of liquidation, this definition acts as a proxy for what is left for shareholders. An argument against this definition is that it does not give any indication whether there is default risk for the firm in the near future because it does not account for the liquidness of the assets. Next to this, Rajan and Zingales argue, total liabilities can also include terms as accounts payable, that in general are more likely to be used for transactions rather than for financing. For instance, a very short term debt which is the result of a transaction where the firm was in the buying end position should not be regarded as debt that is used to finance projects. If this is ignored, leveraged might well be overstated.
(1995) for example explain that an increase in the gross level of trade credit leads to a reduction in leverage, when this definition is used.
Rajan and Zingales conclude that given the problems with the above definitions, the effects of past financing decisions are probably best represented by the ratio of total debt to capital, where capital is defined as total debt plus equity. In order to give an appropriate comparison between the results of this study and that of Rajan and Zingales, this study will use the last definition for the description of leverage. Given the data availability, total debt will be measured by the year end book values derived from the firms annual report, and equity is measured as market capitalization, also on year end basis.
However, to ensure that both the results of this study and the model are robust, the dependent variable will also be measured in another way, which is the ratio of total liabilities to total assets, using year end book values. Although Rajan and Zingales offer criticism to this particular definition, the addition of this alternative dependent variable as a check of robustness results in an increase in the quality of this research.
3.2 Tax
This research will use the tax equation derived by Miller (1977) to calculate the marginal tax effect. This marginal effect rate can best be explained as the gain in firm value of an increase in the value of leverage. Since Miller’s tax equation leans on various different tax rates, the gain of a leverage increase can take both positive and negative numbers. The equation for the marginal tax effect is given in equation 1.
Marginal tax effect (T) = )
) 1 ( ) 1 ( ) 1 ( 1 ( Tpb Tps Tc − − − − (1)
Where Tc is the corporate tax rate, Tps is the personal tax rate on capital gain from equity, and
Tpb is the personal tax rate on capital gain from bonds.
correctly expressed in the equation of Miller, the hypothesis regarding the influence of tax on capital structure is as follows.
Hypothesis 1: There is a positive relationship between the marginal tax effect and the leverage ratio used by firms
3.3 Shareholder rights
This research will follow La Porta et al. (1998) in their measurement of shareholder rights. La Porta et al. measure the level of shareholder rights (in particular how strongly the legal system favors minority shareholders against managers or large dominant shareholders) based on an index, which they call the anti director rights index. This index is based on the following six rights.
(1) The country allows shareholders to mail their proxy vote to the general shareholders meeting, instead of the obligation to show up in person or send an authorized
representative in order to vote.(2) Shareholders are not required to deposit their shares to the company prior to the general shareholders meeting. This prevents shareholders from selling their shares for several days around the time of the shareholder
meeting.(3) Cumulative voting or proportional representation of minorities in the board of directors is allowed, in principle this gives more power for minority shareholders to put their representatives on boards of directors. (4) An oppressed minorities mechanism is in place, for instance the right to challenge the directors’ decisions in court or to force the company to repurchase shares. (5) The minimum percentage of share capital that entitles a shareholder to call for an extraordinary shareholders meeting is less than or equal to 10 percent (the sample median). The higher this percentage, the harder it is for minority shareholders to arrange a meeting in order to challenge or oust the management. (6) Shareholders have preemptive rights to buy new issues of stock that can only be waved by a shareholders vote. The
intention of this right is that it protects shareholders from dilution, whereby the stock is issued to favored investors at a price that is below that of the market.
Given the remarks of Cheng and Shiu (2007) that are put forward in the literature review, the following hypothesis is posed.
Hypothesis 2: Shareholder rights are negatively correlated with leverage.
3.4 Creditor rights
For this research, the level of creditor rights is measured by the same index Quin and Strahan (2007) used, which is in fact the index of La Porta et al (1998). La Porta et al used the
following four creditor rights variables in their analysis.
(1) The country imposes restrictions, such as creditors consent or minimum dividends to file for reorganization, so managers cannot easily escape creditor demands. (2)
Secured creditors are able to gain possession of their security once the reorganization petition has been approved (no automatic stay); (3) Secured creditors are ranked first in the distribution of the proceeds that result from the disposition of the assets of a bankrupt firm, above government and others. (4) The debtor does not retain the administration of its property pending the resolution of the reorganization procedure.
As with the shareholder right index, each of the above conditions that is met results in one point. The level of creditor protection is than the sum of these points.
Taken the findings by Quin and Strahan (2007) that are mentioned in the literature review into account, the following hypothesis is put forward.
Hypothesis 3: Creditor rights are positively correlated with leverage.
3.5 Ownership concentration
Given the fact that the sample of this study contains multiple bank orientated countries, in which high ownership of banks could dramatically reduce the cost of financial distress, there is more reason to assume a positive relationship between ownership concentration and leverage than a negative one. Therefore the hypothesis is as follows.
Hypothesis 4: There is a positive correlation between ownership concentration and leverage.
3.6 Bank based versus Market based structure
As explained in the previous section, the influence of ownership concentration is to some extend intertwined with the structure of a country’s economy, which is either market- or banked based. The variable ‘bank based versus market based structure’ can therefore be considered to correlate with ownership concentration. Though it must be understand correctly that this characteristic also has is own influence. To investigate the influence of the degree that an economy is market or banked based on leverage, this study will use proxies from Demirguc-Kunt and Levine (1999) for the level of market or bank based structures.
The study of Demirguc-Kunt and Levine (1999), examines the financial structure for a cross section of more than 150 countries, using a then recent constructed data set. They determine the degree of market or banked based structures based on their own developed structure index. In their structure index, higher values of structure indicate higher levels of stock market development. Systems of countries with a value above the mean are then classified as being market based systems, where countries that have values below the mean are referred to as bank based systems. For this research, the structure index value of the corresponding
countries is taken, instead of using a dummy for either market or bank structured economies. This is done to allow for differences in various degrees of either one of the named structures. Combining the structure index to the relevant literature that is described, the hypothesis is as follows.
Hypothesis 5: The structure index score is negatively correlated with leverage.
3.7 Control variables
(1984), Long and Malitz (1985), Harris and Raviv (1991) and Rajan and Zingales (1995). There theoretical influence as argued by Rajan and Zingales can be briefly found in the literature review. Their effect will not be further argued here, since their relevance has been proven by earlier research and the background of their effect is beyond the scope of this research. However, the next paragraph does explain exactly how these control variables are measured.
3.7.1 Tangibility
The first control variable is tangibility. The underlying rationale is that tangibility has
influence on the level of a firm’s collateral. Following Rajan and Zingales (1995), tangibility is measured as the ratio of fixed assets to the book value of total assets.
3.7.2 Market to book ratio
A second variable is the market to book ratio, which refers to the way market prices act relative to their book prices. In the research of Rajan and Zingales (1995) this is calculated in the following way. Market-to-book ratio is the book value of assets less the book value of equity plus the market value of equity all divided by the book value of assets (Rajan and Zingales, 1995). However, since this also contains accounts payable, within this research the market to book ratio is measured as the market value of equity divided by the book value of equity.
3.7.3 Size
Rajan and Zingales argue that size can act as a proxy for the inverse probability of default. It will be measured as the natural logarithm of net sales in euros. The variable is measured this way because of the distribution properties.
3.7.4 Profitability
Following the pecking order theory, higher profitability might lead to a reduction in the need for debt, since there is more internal financing available. Following Rajan and Zingales (1995) profitability will be measured by EBITDA divided by the book value of assets.
Table 3
Institutional factors and the expected relationship with leverage
This table gives an overview of the tested institutional factors, the original literature where the data was derived from and the expected relationship of the variables with the leverage ratio.
Institutional factor Source of data Expected relationship
Marginal Tax effect Miller (1977) Positive relationship
Shareholder rights La Porta et al. (1998) Negative relationship
Creditor rights La Porta et al. (1998) Positive relationship
Ownership concentration La Porta et al. (1998) Positive relationship
4. Data description and descriptive statistics
For this study, a Thomson Financial DataStream dataset will be used. The aim is to use a panel containing data about the leverage ratio, size, tangibility, market to book ratio and profitability of all non financial firms in the main indices of France, Germany, the United Kingdom and Italy, from the timeframe 2005 until 2010. The specific requirements of the dataset are that they should contain the book value of the fixed assets, the book value of the total assets, sales, EBITDA, the book values of debt, and the book and market values of equity of all the companies. Al these values are obtained through the use of Thomson Financial DataStream. The methodology necessary to describe the institutional factors is obtained from literature by La Porta et al. (1998) (1999), Miller (1977) and Demirguc-Kunt and Levine (1999). The papers of La Porta et al. and Demirguc-Kunt and Levine are also used for obtaining data about the first years of the data sample. Data of the later years is obtained through literature by Spamann (2010) and Djankov et al. (2007) who both used the exact same methodology as La Porta et al. Although these researches may seem to be outdated, their findings remain highly relevant. Also, many recent studies in the field of capital structure theory use techniques to measure variables which are all in one way or another derived from the methods in these papers. This validates their involvement in this study.
The countries are chosen for multiple reasons. First, they were all included in the research of Rajan and Zingales (1995), which makes it ideal for comparison. Second, they are all big and well developed countries though they have differences in institutional characteristics, mostly because their laws originate from different legal families. As earlier discussed, the timeframe is chosen because of the introduction of the IFRS. The chosen timeframe also includes the recent financial crisis years. During the financial crisis years, the leverage ratio of the
organizations in the sample of this study could well be deviated from their original value. Of course this might be also true for some of the other variables. Therefore, interpreting the results of the research should be done with great caution. To analyze the effect of the financial crisis, the tests are also run on the dataset without the crisis years. This makes it possible to investigate the influence of the financial crisis. Last, all financial firms such as banks or insurance companies will be excluded from the data sample because their leverage ratio is strongly influenced by implicit and explicit investor insurance schemes, like a deposit
which are not strictly comparable to the debt that is issued by non financial organizations. Finally, financial firms are sometimes required to meet minimum capital regulations. Of course, this has a direct influence on capital structure.
Leaving out the financial firms in the sample as well as the observations with any missing values only leaves the outliers to be possibly removed. According to Field & Hole (2003), outliers in a dataset have the property that they can bias the mean, and substantially increase the standard deviation. Zimmerman (1994) also argues that the error variance can be
increased because of outliers and the power of statistical tests is reduced. In general, two types of outliers can occur: outliers as a consequence of data errors, and real outliers. Of course, data errors should always be removed. However, removing legitimate outliers is not per definition the best course of action, as the removal might leads to loss of relevant data. Since the data set in this study does not contain that many legitimate outliers, removing them seams irrelevant since they their small number results in little or no influence on the different statistics.
After the removal of the financial firms and the data errors, 915 observations divided over the four countries from the timeframe 2005-2010 remain. Table 4 shows how the observations are divided between year and country.
Table 4 Observations
The table below gives an overview of how the observations in the dataset used in this study are divided over year and over country.
France Germany Italy United Kingdom Total
2005 33 26 23 69 151 2006 33 26 23 69 151 2007 33 26 23 71 153 2008 34 26 23 71 154 2009 34 26 22 72 154 2010 34 26 21 71 152 Total 201 156 135 423 915
4.1 Dependent Variables
As explained in the variable description section, the dependent variable is the leverage ratio, and will be measured in two different ways for robustness purposes. Table 5 will give an overview about the descriptive statistics of the two dependent variables of this research, also split down on country level.
Table 5
Descriptive statistics dependent variable
This table gives an overview of the descriptive statistics of the dependent variable leverage, measured in two different ways. The table gives respectively values for the average, median, 1stQuartile, 3rd Quartile, the minimum, the maximum and the standard deviation of leverage in total and also per country.
Average Median 1st Quartile
3rd Quartile
Minimum Maximum Standard deviation Leverage (debt / debt + equity) 0.403 0.385 0.248 0.544 0.008 0.954 0.200 France 0.470 0.486 0.325 0.605 0.118 0.940 0.193 Germany 0.476 0.493 0.326 0.616 0.049 0.945 0.211 Italy 0.498 0.492 0.378 0.631 0.086 0.954 0.182 United Kingdom 0.316 0.293 0.202 0.422 0.008 0.782 0.166 Leverage
(deb / total assets) 0.521 0.520 0.406 0.636 0.048 1.184 0.171
France 0.534 0.511 0.426 0.679 0.194 0.886 0.156
Germany 0.530 0.520 0.447 0.600 0.228 0.976 0.150
Italy 0.588 0.596 0.467 0.706 0.281 1.184 0.163
United Kingdom 0.489 0.506 0.372 0.612 0.048 1.023 0.180
4.2 Independent variables
The institutional variables expected to have their influence on leverage are the independent variables of this research. Table 6 gives information about the statistical values of these institutional variables.
Table 6
Descriptive statistics independent variables
This table gives an overview of the descriptive statistics of the independent variables. The table gives respectively values for the average, median, 1stQuartile, 3rd Quartile, the minimum, the maximum and the standard deviation, for the variables Marginal tax effect, shareholder rights, creditor rights, bank versus market based and ownership concentration.
Average Median 1st Quartile
3rd Quartile
Minimum Maximum Standard deviation Marginal tax effect -0,040 0,081 -0,263 0,097 -0,263 0,248 0,215
Shareholder rights 3,608 4,000 3,000 5,000 1,000 5,000 1,422
Creditor rights 2,606 3,000 1,000 4,000 0,000 4,000 1,485
Bank vs Market 0,290 -0,100 -0,170 0,920 -0,570 0,920 0,605
Ownership
concentration 0,329 0,340 0,190 0,480 0,190 0,580 0,149
Interesting to see is that the average value of the marginal tax effect is negative, although the median value is positive. Given the equation derived by Miller (1977), one can thus observe that there is indeed a combination of tax rates for which an increase in leverage leads to a negative effect (marginal tax effect) for firm value. Table 6 also shows that for most statistics presented, the values for both shareholder rights and creditor rights are whole numbers. The underlying reason for the whole values of these institutional factors is the way of
measurement. As explained in the variable description section, this study follows the
methodology of La Porta et al. (1998) which only allows for real numbers. Within shareholder rights, one can observe that the maximum value is 5, which is the maximum number of points that can be scored within the La Porta et al. index. It also can be noticed that the minimum value in creditor rights is zero. This minimum value that can be obtained in the creditor right index of La Porta et al, is the value for France in the years 2005, 2006 and 2007.
In general, it is quite noticeable in Table 6 that all institutional factors have relatively large standard deviations. This can also been observed by the fact that all factors have remarkable relative differences between the maximum and minimum values.
4.3 Control Variables
The data used as input for the control variables is obtained through a Thomson DataStream time series request. Table 7 gives an overview of the descriptive statistics of these variables.
Table 7
Descriptive statistics control variables
This table gives an overview of the descriptive statistics of the control variables. The table gives respectively values for the average, median, 1stQuartile, 3rd Quartile, the minimum, the maximum and the standard deviation, for the variables tangibility, market to book, size and profitability.
Average Median 1st Quartile
3rd Quartile
Minimum Maximum Standard deviation Tangibility 0.634 0.649 0.527 0.765 0.029 0.967 0.170 France 0.602 0.619 0.488 0.711 0.299 0.847 0.141 Germany 0.566 0.586 0.475 0.693 0.029 0.876 0.187 Italy 0.669 0.721 0.557 0.784 0.239 0.956 0.168 United Kingdom 0.663 0.674 0.555 0.800 0.158 0.967 0.167 Market to book 2.285 1.742 1.092 2.820 -7.992 19.843 2.104 France 1.621 1.410 0.897 2.087 0.151 7.781 1.031 Germany 1.708 1.290 0.829 2.106 0.354 7.524 1.333 Italy 1.723 1.586 0.824 2.613 -7.992 5.328 1.500 United Kingdom 2.985 2.287 1.402 3.647 -4.000 19.843 2.600 Size 15.911 15.978 14.899 16.891 10.877 19.332 1.470 France 16.790 16.700 16.273 17.478 14.701 18.893 0.902 Germany 16.851 16.648 16.036 17.882 14.373 18.837 1.121 Italy 15.398 15.313 14.259 16.083 12.536 18.496 1.361 United Kingdom 15.322 15.357 14.400 16.354 10.877 19.332 1.453 Profitability 0.136 0.123 0.089 0.166 -0.769 0.700 0.088 France 0.109 0.099 0.072 0.141 -0.129 0.375 0.063 Germany 0.119 0.113 0.091 0.142 -0.013 0.398 0.061 Italy 0.130 0.124 0.090 0.153 -0.065 0.372 0.073 United Kingdom 0.157 0.141 0.104 0.202 -0.769 0.700 0.104
What can be observed in Table 7 is that for all control variables, the values for the different countries are quite in line with each other. Also, in general the average values and the median values do not show huge differences. The only observations of Table 7 that are quite
average value is 2.285, the maximum value is 19.843 and the minimum value -7.992. These are however two of the legitimate outliers. As argued earlier, they are not removed from the data sample in order to preserve all relevant information. A final remark about the statistics in Table 7 is that profitability shows a negative minimum value for all countries. This implies that each country had at least one observation for which the EBITDA was negative. 4.4 Correlation between variables
In order to perform a proper research, the correlation between the independent, dependent and control variables must be investigated to address possible problems with multicollinearity. Table 8 presents the correlation matrix between the different independent variables.
As expected, there is a considerable amount of correlation between the two measures of leverage used in this study. However, since this correlation is far from perfect, the robustness purposes of adding the second measure of leverage are not jeopardized. The control variables show remarkably less correlation among the variables with values ranging from the smallest value 0.017 (tangibility and size) to a largest absolute value of -0.531 (profitability and leverage (D/D+E)). These correlations are considered to have no significant influence on the outcomes of this research.
Nevertheless, some of the independent variables show correlations that could lead to serious misjudgments of the outcomes of this study. As was already expected in the variable
description section, there is a huge negative correlation between ownership concentration and the bank versus market variable, with a value of -0.929. When the variable bank versus market shows a high value, implying a market based economy, it can be expected that
ownership is dispersed and shows a low value in ownership concentration. This might explain the strong negative correlation between these two independent variables. The other dependent variables also show correlations between each other that should be acknowledged.
In particular, the correlation between the marginal tax effect and the other dependent variables are relatively large. The correlation between creditor rights is and tax is -0.941 and the
Brooks (2008) argues that the above strong correlations might cause multicollinearity problems. According to Grewal et al. (2004), unfortunately many solutions to solve multicollinearity problems are ad hoc. For instance, dropping a variable can solve the
problems but may lead to misspecification errors. Restricting parameters might also help but is not always possible. Grewal et al. continue that solutions are also often impractical. They illustrate this with the example of obtaining additional data that does not suffer from
multicollinearity.
Brooks argues that one way to deal with multicollinearity is to ignore it. This is possible when the model is otherwise statistically adequate. According to Brooks, existence of high
correlation is not always causing significant t-values to be insignificant. Brooks continues by explaining that the presence of near multicollinearity does not influence the properties of the Least Squares estimator. The Least Squares estimator is still consistent, unbiased and
5. Methodology
Whether the previously described theoretical effects of tax, shareholder rights, creditor protection, market or bank based countries and ownership concentration also hold in practice, is investigated through panel data regressions.
The main structure of this regression is a pooled least square regression. First of all, the assumptions for a least square regression are tested. Since this study uses panel data, it is also necessary to determine whether a simple pooled regression is valid, or that models with random or fixed effects are in order. When the appropriate model is determined, the
regressions will be run, together with several robustness tests. These robustness tests are used to investigate to what extent the results are robust on cross sectional level and on cross time level. The panel data equation method that is used differs from cross section equation methods and time series. Where cross sectional methods only regard multiple observations at one point in time and time series analyses specifically looks at the same variable in multiple points in time, panel data combines these two as it overlooks multiple observations in different points in time.
Using this panel data has several advantages over cross-sectional or time-series data. As Hsiao (2006) states, there is more accurate inference of model parameters. As already argued, panel data contains both the cross sectional dimension as the time series dimension.
In general, panel data comes with additional sample variability and has more degrees of freedom than the usual cross-sectional or time series data. Also, Hsiao continues, panel data can also be used in constructing and testing more complicated behavioral hypotheses.
Furthermore, panel data helps controlling the impact of omitted variables. According to Hsiao (2006), the fact that panel data contains information on intertemporal dynamics might help control the impact of both unobserved and missing variables.
5.1 Regression
(2)
In this equation, the L stands for the leverage ratio. The C is the constant factor, and IF are the different tested institutional factors, namely tax, ownership concentration, creditor rights and shareholder rights. This is followed by the control variables, respectively tangibility, firm size, market to book ratio and profitability.
5.2 Assumptions of Least Square regression
When using this type of regression, there are several assumptions that have to me met in order to assure that hypothesis tests regarding the coefficients estimates can be conducted in a valid manner. Before the regression can be computed, these assumptions have to be tested for. These are tests for heteroscedasticity , autocorrelation in the residuals and the assumption of normality.
When the tests for heteroscedasticity or autocorrelation in the residuals show that there is a problem with either of these points within this research, the use of a Pooled Ordinary Least Squares might be inappropriate. This is because Ordinary Least Squares can in such a case be statistically inefficient. Also, they might give inferences that are misleading. If the results show that there is a problem on this level, the Pooled Ordinary Least Squares regression will be replaced by a Generalized Least Squares analysis.
Even if there are no problems with heteroscedasticity or autocorrelation in the residuals, the use of an Ordinary Least Squares regression however might lead to other problems. One of them is that an OLS regression implies a requirement that the model is linear. As Brooks (2008) argues this means that in the most simple bivariate case, the relationship between the variables x and y must be capable of being expressed diagrammatically using a straight line. Since literature does not suggest what kind of relationship the institutional factors have on leverage, the OLS regression assumption tests will be complemented with a Ramsey reset test, in order to test for nonlinearity.
5.3 Fixed or Random effects
Since this study makes use of panel data, Brooks (2008) argues that it is also necessary to test whether the panel estimator approaches ‘fixed effects’ and ‘random effects’ can be employed. As Brooks explains, the simplest types of fixed effects models allow the intercept to differ on a cross-sectional basis while keeping it consistent over time, or the other way around. In a random effects model, each cross-sectional observation has an intercept that is assumed to arise from a common intercept plus a random variable. First, to investigate whether the fixed effects models are necessary or the Pooled Least Squares regression is sufficient; a Likelihood Ratio test will be run. Second, a Hausman test will show if a random effects model is
appropriate, or that the fixed effect model is preferred. 5.4 Robustness
In order to perform a check of robustness of the model, an alternative measure of the dependent variable, the leverage ratio, will also be used in the regression analysis. This is done to check whether the independent variables have the same kind of influences on the debt level when this level is measured in another way. For comparison reasons, this research uses one of the descriptions of leverage also mentioned in Rajan and Zingales (1995), namely the level of total liabilities to total assets.
5.5 Financial crisis
6. Results
This section will provide the results of this research. First of all, the earlier mentioned assumptions for the least squares regressions are tested, and the correct model is determined. After determining the preferred model, the results of the regression and the robustness tests are presented
6.1 Testing for the assumptions for regression
The presence of heteroscedasticity in the data set is tested using White’s test. In order to perform this test, the panel structure of the data needs to be temporarily ignored, and the data needs to be regarded as an unstructured set. For the used data, White’s test gave a F-statistic of 8.299 with a probability of 0.000 which is considerably below the significance level of 0.05 and thus suggest the existence of heteroscedasticity. This problem will be solved by using White’s modified standard error estimates, which are heteroscedasticity consistent.
The appearance of autocorrelation in the residuals is tested by the use of the Durbin Watson test. In order for the Durbin Watson test to be a valid test, the regression must contain a constant term, the regressors must be non-stochastic and there must be no lags in the regression of the dependent variable. For this research, these conditions are met, and
autocorrelation can be tested using the Durbin Watson test. The Durbin Watson value for the regression is 1.164 which gives no clear evidence of any autocorrelation.
Another assumption that needs to be tested for is the test for non-normality in the disturbance tests. This will be tested by the use of the Bera-Jarque test. As Brooks (2008) explains the Bera-Jarque has a chi-square distribution with two degrees of freedom. Its null hypothesis is that the errors are normally distributed. In this study the Bera-Jarque value is 3.551 with a p-value of 0.169. Since the probability is bigger than the significance level of 0.05, there is no reason to reject the assumption of normality.
In order to test for fixed or random effects, a Likelihood Ratio test and a Hausman test are performed. The Likelihood Ratio test has a F-value of 7.995 and a p-value of 0.000, which implies that period-fixed effects should be taken into account and a Pooled Least Square regression is not sufficient anymore. The Hausman test calculates a Chi square statistic of 44.519 and a p-value of 0.001. For this research, the random effects model is not appropriate and the fixed effect model is preferred. To summarize, this research will thus use a Panel Least Squares regression with period fixed effects.
Table 9 below gives an overview of the different values of the assumption test and the possible actions taken because of the result.
Table 9
Least Squares regression assumption tests
This table gives an overview of values of the tests of the different assumptions of the regression, heteroscedasticity, the absence of autocorrelation, normality in the error terms, the test for linearity and the tests for the need for fixed or random effects.
Test Value p- value Implication
White's Test 16.333 0.001
Use of White's heteroskedasticity consistent standard deviation
Durbin Watson 1.164
Bera- Jarque 3.551 0.169
Ramsey RESET 1.677 0.196
Likelihood Ratio 7.995 0.000 Period-fixed effects
Hausman Test 44.519 0.001
6.2 Results of Panel Least Squares Regression
Regression 1 in table 10 shows the influence of the institutional factors on leverage measured as debt / (debt + equity). In table 10 it can be seen that in the full model regression, the effect of creditor and shareholder rights is positive, and the effect of the other institutional variables is negative. However, the effect of creditor rights is not significant at any common level, where the effect of ownership concentration is only significant on the 10% level. The regressions have an average R² of 0,48. The R² of a statistical model is the coefficient of determination and provides information about to what extent the outcome of the dependent variable is explained by the model. In this case, 48 percent of the variability in the data set can be accounted for by the statistical model.
Table 10
Panel Least Square regression results
This table gives the result of the PLS regression with period fixed effects for the years 2005-2010 where leverage is measured as debt/ (debt + equity). *, **, *** indicates that the variable is significant at respectively the 10%, 5% and 1% level. The table gives respectively values for coefficient, significance level and T-statistics.
I II III IV V VI
Marginal tax effect
-0.474 *** (-3.389) 0.130 *** (5.072) Creditor rights 0.016 (1.501) -0.013 *** (-3.521) Shareholder rights 0.013 ** (2.214) -0.021 *** (-5.959) Bank vs Market -0.346 *** (-4.478) -0.068 *** (-8.022) Ownership concentration -0.347 * (-1.897) 0.289 *** (8.705) Tangibility -0.133 *** (-4.612) -0.104 *** (-3.538) -0.115 *** (-3.895) -0.124 *** (-4.317) -0.098 *** (-3.424) -0.102 *** (-3.599) Market to book -0.013 *** (-5.009) -0.016 *** (-5.703) -0.017 *** (-6.067) -0.016 *** (-5.851) -0.014 *** (-5.052) -0.014 *** (-5.021) Size 0.042 *** (11.102) 0.033 *** (8.856) 0.035 *** (9.370) 0.038 *** (10.905) 0.033 *** (9.456) 0.035 *** (10.270) Profitability -0.919 *** (-15.548) -0.881 *** (-14.393) -0.895 *** (-14.530) -0.909 *** (-15.048) -0.876 *** (-14.674) -0.892 *** (-15.065) C 0.085 (0.845) 0.113 * (1.795) 0.121 * (1.835) 0.119 * (1.908) 0.113 * (1.867) -0.031 (-0.507) R² 0.516 0.475 0.468 0.481 0.496 0.502 Adjusted R² 0.508 0.469 0.462 0.475 0.491 0.497 N 915 915 915 915 915 915
