An International Comparison of Capital Structure and Debt Maturity Choices

Transcription

1 An International Comparison of Capital Structure and Debt Maturity Choices Joseph P.H. Fan Sheridan Titman School of Business and Management McCombs School of Business Hong Kong University of Science and University of Texas at Austin Technology Austin, TX Clear Water Bay, Kowloon USA Hong Kong and Garry Twite Australian Graduate School of Management University of New South Wales Syndey, NSW 2052 Australia February 2004

2 Abstract This study examines the capital structure and debt maturity choices of firms in a crosssection of 39 developed and developing countries, focusing on the effect of the countries public policies and institutional structures. Our analysis suggests that a country s taxation and inflation policies as well as its legal institutions have an important effect on how its corporations are financed. Specifically, the preferential tax treatment of dividends is associated with lower leverage, high inflation is associated with lower leverage and the use of short-term debt and corruption increases financial leverage and reduces the maturity structure of the debt. However, controlling for corruption, the legal system per se i.e., whether the country adopts a common law system only influences the debt maturity decision. In addition, the financial institutions that supply capital influence the type of capital that is used. In particular, countries with large amounts of bank deposits tend to have shorter debt maturities and countries with a greater presence of other institutional investors are less levered and have longer debt maturities. We also find that the presence of information intermediaries is associated with lower leverage and the use of trade credit, highlighting their role in facilitating information dissemination. Finally, union bargaining power is unrelated to leverage, but is associated with the greater use of long-term debt. 2

3 1. Introduction Corporate financing choices are likely to be determined by a combination of factors that are related to the characteristics of the firm as well as to the institutional environment. Although most studies examine corporate financing choices within individual countries, and thus focus on the importance of firm characteristics, 1 there is a growing literature [Booth, Aivazian, Demirguc-Kunt, and Maksimovic, 2001; Claessens, Djankov and Nenova, 2001; Demirguc-Kunt and Maksimovic (1996, 1998, 1999); Giannetti, 2003] that examines capital structure choices across countries and considers the effects of institutional differences. This study builds on this recent literature by examining a larger panel of countries and firms and by considering a number of important determinants of financial structure that were not previously explored in this literature. To understand our motivation, it is useful to illustrate the importance of country factors relative to industry factors in determining capital structure. A regression of firm leverage, measured as the book value of debt over the market value of the firm, on firmspecific variables, industry dummy variables and country dummy variables has an adjusted R-square of When the regression is run with all variables except for country dummies, the adjusted R-square is substantially reduced to However, in the regression that includes all variables except for industry dummies, the adjusted R-square is reduced to only When these same regressions are run with debt maturity, 1 Examples of empirical studies examining the association between firm characteristics and capital structure within specific countries include Titman and Wessels [1988] U.S., Campbell and Hamao [1995] Japanand Gatward and Sharpe [1996] Australia. Barclay and Smith [1995], Stohs and Mauer [1996] and Guedes and Opler [1996] examine the association between firm characteristics and debt maturity in the U.S. Gatward and Sharpe [1996] undertake a similar study of debt maturity in Australia. 1

4 measured as the book value of long-term to total debt, as the dependent variable, the R- square is When the regression is run with all variables except for country dummies, the R-square is substantially reduced to However, in the regression that includes all variables except for industry dummies, the R-square is only slightly reduced to Our interpretation of these regressions is that location, i.e., the country in which the firm resides, is a more important determinant of how it is financed than its industry affiliation. To examine why this is the case, we examine the extent to which crosscountry differences in capital structures can be explained by differences in economic development, inflation and taxation policies as well as by legal and financial institutions. In comparison to previous work on this topic, we examine a broader set of countries providing more cross-sectional dispersion in the explanatory variables. In addition, we examine a broader class of explanatory variables and employ an estimation procedure that enables us to examine the influence of these country specific explanatory variables after controlling for cross-country differences in firm- and industry-level characteristics. In contrast to Booth, Aivazian, Demirguc-Kunt, and Maksimovic [2001] and Giannetti [2003] we find that taxes, inflation and the suppliers of capital have an important influence on capital structure choices. In particular, firms use less debt when dividends are preferentially taxed and leverage is lower and firms use more short-term debt when inflation is higher. We also find that firms in countries with large amounts of bank deposits tend to have shorter maturity debt, while firms in countries with more life insurance assets tend to have longer maturity debt. 2

5 In addition, consistent with Demirguc-Kunt, and Maksimovic [1999], we find that the quality of the legal system (i.e., the efficiency and integrity of the legal system) influences capital structure choice. Demirguc-Kunt and Maksimovic [1999] include a measure of the efficiency of the legal system using the extent to which it is utilized to resolve conflicts. They find longer debt maturity in countries where the legal system is used more to resolve conflicts. In contrast, we focus on the integrity of the legal system using an index of country corruption, and find a strong association between corruption and capital structure choice. In particular, firms in countries that are viewed as more corrupt tend to be more levered and use more short-term debt. However, after controlling for corruption, we find that the legal system per se, i.e., common versus civil law, plays a less important role, influencing only (and somewhat weakly) debt maturity. We also examine variables that have not been considered in prior work. Specifically, given the prominence of asymmetric information in models of capital structure we include variables that measure the presence of two information intermediaries, auditors and analysts. The evidence provides weak support for the idea that information intermediaries facilitate the use of equity and long-term debt. In addition, motivated by the association between union power and leverage in the United States (i.e., Bronars and Deere, 1991) we examine whether the legal protection of unions is associated with differences in capital structure. We fail to find a significant relation between the leverage ratio and the legal protection of unions, but find that the firms in countries which have stronger legal protection of unions tend to have longer term debt. The paper is organized as follows. Section 2 discusses the association between country level institutional factors and financial choices. Section 3 describes the sample 3

6 and methodology, and report basic statistics. Section 4 carries out the first-stage procedure to examine firm level determinants of capital structure and debt maturity. Section 5 executes the second-stage procedure that identifies country level determinants of these financing decisions. Section 6 draws some conclusions. 2. Institutional factors and corporate financial policy This section discusses how institutional differences between countries influence how firms are financed The Legal System and Corruption Incentive problems - conflicts of interest between corporate insiders (managers and/or majority shareholders) and external investors - are an important factor that shapes corporate policy and productivity. As pointed out by La Porta, Lopez-de-Silanes, Shleifer, and Vishny [1998], one of the principal remedies to these problems is legal protection, which consists of both the content of the laws and the quality of their enforcement. The extent to which contracts are enforceable is likely to be an important determinant of capital structure. For example, when enforcement is weak, instruments that allow insiders less discretion are likely to dominate. La Porta, Lopez-de-Silanes, Shleifer, and Vishny [1998] find significant variation in the extent of legal protection of external investors across both developed and developing countries, and argue that legal systems based on common law offers outside investors (debt and equity) better protection than those based on civil law. All else equal, this suggests that we would expect common 4

7 law countries to use more outside equity and longer-term debt. To test for these relationships we define a dummy variable that takes a value of one if the country s legal system is based on common law and zero otherwise. Another important characteristic of the legal system is the quality of legal enforcement. This is determined by the efficiency and integrity of the legal system its strength and impartiality. 2 We focus on the integrity of the system using an index of country corruption, the Corruption Perception Index. This index prepared by Transparency International reflects the extent to which corruption is perceived to exist among public officials and politicians, where corruption is defined as the abuse of public office for private gain. We reverse the index, which ranges from 0 to 10, with larger value indicating more severe corruption. In the context of the quality of legal enforcement, the index proxies for the threat of all or part of investor rights being expropriated. When the legal system has less integrity we expect that debt will be used relatively more than equity; likewise, shorter-term debt will be used relatively more than long-term debt. The contractual structure of debt limits the potential for the expropriation of investor rights; at least relative to the more opportunistic environment provided by equity. 3 One also expects that shorter-term debt will be employed relatively more than longer-term debt when the legal system has less integrity since shorter maturity limits the potential to expropriate creditor rights. 2 Demirguc-Kunt and Maksimovic [1999] measure the efficiency of the legal system using an index of the extent to which it is utilized to resolve conflicts. 3 Smith and Warner [1979]. 5

8 2.2. Tax System The tax system in general, and specifically the tax treatment of interest and dividend payments, is an important factor that influences the firm s capital structure choice. 4 We observe three main categories of tax regimes operating globally. The first is the classical tax system in which dividend payments are taxed at both the corporate and personal levels and interest payments are tax-deductible corporate expenses that are only taxed at the personal level. The classical tax system exists in Brazil, Chile, China, Hong Kong, India, Indonesia, Israel, Japan, Korea, Malaysia, Netherlands, Pakistan, Peru, Philippines, Singapore, South Africa, Switzerland and USA. The second is the dividend relief tax system, which differs from the classical tax system in respect to the taxing of dividend payments. Dividend relief takes one of two forms. Dividend payments are not taxed at the corporate level, but remain taxed at the same rate as interest payments at the personal level. Alternatively, dividend payments are taxed at the corporate level, but taxed at a reduced rate at the personal level. A dividend relief tax system exists in Austria, Belgium, Denmark, Greece, Portugal, Sweden, Thailand and Turkey. Third is the dividend imputation tax system where firms can deduct interest payments at the corporate tax rate and domestic corporate taxes paid are distributed to taxable resident shareholders as a tax credit with dividend payments. Interest payments remain a tax-deductible corporate expense, only taxed at the personal level. Dividend imputation systems are in place in Australia, Canada, Finland, France, Germany, Ireland, Italy, Mexico, New Zealand, Norway, Spain, Taiwan and United Kingdom. The proportion of corporate tax available as a tax credit under these imputation systems varies 4 See Graham [2003] for a review of the literature on influence of taxes on capital structure choice. 6

9 from country to country. Only in Australia, Germany, Italy, New Zealand and Norway is the full amount of the corporate tax paid distributed as a tax credit. The tax benefit of debt financing is expected to vary across the different tax systems, which in turn affects capital structure choice. All else equal, we expect that debt will be used less in countries adopting the dividend imputation or relief tax systems than others that adopt the classical tax system. Furthermore, we expect the difference is most significant for the countries that adopt dividend imputation. To test for this relationship we define two dummy variables, a dividend relief dummy that takes a value of one if the country adopts a dividend relief tax system and zero otherwise and dividend imputation dummy that takes a value of one if the country adopts a dividend imputation tax system and zero otherwise Investor preferences Financial economists have typically viewed the capital structure problem from the perspective of firms that face competitive and complete financial markets that offer debt and equity capital at competitive risk-adjusted rates. An exception is the Miller [1977] model where individual capital structure choices by firms are irrelevant, but the aggregate debt ratio in the economy is determined by investor preferences for holding debt versus equity securities. While these preferences are determined by taxes in Miller s model, one can more generally discuss how preferences for holding various debt and equity instruments affect the capital structure choice. 5 For example, the insurance sector is bigger in some countries than in others. Since insurance companies have a comparative advantage in holding longer-term 5 See Titman [2002] for a discussion of the effect of investor preferences on capital structure choices. 7

10 securities, since they have long-term obligations, one might expect to see relatively less short-term debt in countries with larger insurance industries. Similarly, one might expect to see relatively less reliance on short-term debt in countries with more institutional investors, like pension funds and mutual funds. In contrast, banks, because of their monitoring capabilities, may have a comparative advantage holding short-term debt. Hence, one might expect to see firms being financed with more short-term debt in countries where the banking sector has access to more funds. To proxy for the supply of funds in these sectors, we include the following variables in our regressions. We use deposits (or liquid liabilities)/gdp to measure the amount of funds that can be provided by the banking sector, and life insurance penetration and total institutional investment/gdp to measure the amount of funds that can be provided by other institutional investors. Our analysis assumes that the amount of funds that flow to these sectors can be viewed as exogenous. Specifically, we are assuming that capital structure choices do not directly affect the amount of capital that is provided by the individual sectors Liquidity Institutional investors tend to prefer holding securities that are more liquid. This suggests that in those countries with more institutional investors, firms are likely to tilt their capital structures towards the more liquid sources of financing. We use stock 6 One can plausibly argue that the financing needs of corporations affect the funds that are available to the different investor sectors. Suppose, for example, that the need for monitoring declines, making bank loans somewhat less attractive to long-term bonds. On the margin, this would increase the interest rate on longterm bonds, making it more attractive for households to invest in fixed income mutual funds rather than bank deposits. While this creates a potential endogeniety problem, this problem should be mitigated by our sample that includes a relatively small number of industries that are present in most of the countries we examine. 8

11 turnover as our proxy for the liquidity of the equity market and conjecture that firms in countries with more institutional investors will be financed with more equity if the stock market has greater turnover. We conjecture that when the size of the institutional market is relatively small, liquidity is less important. To capture this effect we interact turnover and stock market capitalization/gdp with the size of the institutional market. We do not have a direct proxy for the liquidity of the bond market, but conjecture that if the government bond market is larger, the market for corporate bonds will also be more active and liquid. 2.5 Information intermediaries The existence of asymmetric information is likely to tilt capital structures toward a higher use of debt, especially, shorter-term debt. 7 Hence, financial structure is likely to be shaped by the nature of corporate information, which is in turn affected by institutions that collect and disseminate information. In this respect, auditors and equity analysts play a key role. The role of the auditor in certifying the firm s accounts is to enhance the credibility of public information while mitigating the importance of private information. 8 Likewise the existence of equity analysts facilitates the revelation and dissemination of private information. 9 This suggests that markets characterized by a strong audit function 7 See Flannery [1986] and Diamond [1993]. 8 Fan and Wong [2002a] report that the credibility of accounting information in emerging markets is generally poor due to agency problem and protection of proprietary information. They find evidence that high quality external auditors are more likely employed when the agency problem is more severe, supporting the certification role of these auditors. Choi and Wong [2002] provide international evidence that high quality auditors are more likely employed by large firms in economies with weak legal enforcement, again supporting the certification role of external auditors. 9 DeFond and Hung [2001] present cross-country evidence that financial analysts provide information that helps investors overcome earnings opacity caused by institutional factors. 9

12 and equity analyst following will have lower leverage and longer maturity debt than markets with weaker audit and analyst roles. To test for these relationships we use Big-five auditors market share and the average number of equity analysts following a firm as measures of the level of information intermediary activity. 2.6 Union Power There is a growing literature [Bronars and Deere, 1991; Dasgupta and Sengupta, 1993; Perotti and Spier, 1993] that suggests that higher leverage can affect labor market outcomes to a firm s advantage, when such outcomes involve bargaining between the firm and unions. The literature suggests two potential links between leverage and labor market outcomes. First, by increasing leverage, the firm can reduce the funds a union can extract without bankrupting the firm [Dasgupta and Sengupta, 1993]. Second, the more highly levered firm can credibly threaten not to undertake new investment unless the union concedes to wage reductions [Perotti and Spier, 1993]. In both circumstances, debt is chosen when it provides a bargaining advantage, that is, as the bargaining power of unions becomes bigger, the bargaining advantage of debt becomes more important. Bronars and Deere [1991] in a study of U.S. firms, find that firms do respond to the threat of unionization by increasing their debt-equity ratio. One characteristic of union bargaining power is the extent of legal protection of unionization within the labor force. Strong legal protection increases the union s bargaining power in wage negotiations. We measure this using an index of the level of 10

13 statutory protection provided to unions. 10 The index is one when employers have the legal duty to bargain with unions, union negotiated terms are extended to all workers by law and the law allows closed shops; and zero when none of these conditions apply. 3. Data and methodology This section describes the sample and presents the country and industry patterns of capital and debt maturity structures. It then introduces the empirical procedure employed in this study Sample selection The primary source of our firm-level data is Worldscope, which contains financial data on companies from a wide range of industries in over 50 countries. We restrict the sample to those firms listed on the stock market of the country in which it is domiciled. We limit our analysis to a sample of representative industries to mitigate problems that can arise from the fact that the firms in different countries may exhibit different financial structures because their industrial mix is different. We select firms in eleven industries: business services (SIC 73), chemical (SIC 28), communication (SIC 48), construction (SIC 15, 16, 17), food & beverage (SIC 20), metal fabrication (SIC 33, 34), resource (SIC 10, 12), newspapers (SIC 27), paper & pulp (SIC 26), wholesale (SIC 50, 51), and retail (SIC 52 59). These eleven industries were selected because they have a presence in a large number of countries and there are also differences between these industries that are likely to lead to cross-industry differences in capital structure. For 10 The index is taken from Botero, Djankov, La Porta and Shleifer [2003]. 11

14 example, the high level of tangible assets in metal fabrication firms as opposed to the lower levels in business services firms. Our analysis covers the period of 1991 through If a firm in a given year has less than two other peer firms in the same industry and country, we exclude it from the sample. This ensures sufficient observations in any given industry, country, and year. We further exclude firm-year observations with missing financial data that is required for the firm-level analysis. For any one firm it is not necessary that data be available for all years in the sample period. The final sample consists of 5,344 firms from 39 countries, totaling 51,657 firm-years. Table 1 provides a description of the sample. The sample covers a broad cross section of developed and developing countries with every continent represented. Most of the countries have at least some observations in each year. Exceptions include those countries that did not have stock markets until the 1990s. The representativeness of the sample firms varies across countries in terms of number and/or market capitalization. In some countries almost 50% of the listed firms are included in our sample while in others only 2% of the total listed firms are included. The level of coverage depends in part on the presence of firms from the eleven industries we examine and may also reflect the fact that Worldscope has uneven coverage of firms across the countries. 11 As can be seen from the last two columns of Table 1, the countries with low data coverage tend to be developing economies. [Table 1 about here] We report the sample coverage by country and industry sector in Appendix 1. It shows a dispersed sample of firm-year observations across countries and industries. 11 Worldscope biases the sample towards large listed companies in each country. This can be seen by comparing the number and value coverage of the sample in the last two columns of Table 1: value coverage is generally higher than number coverage across the countries. 12

15 3.2. Country and industry patterns To gain a basic idea about how capital and maturity structures differ across countries, we compute the median and mean leverage and maturity structure by country. As can be seen in Appendix 2 and Figure 1, developing economies occupy both ends of the leverage spectrum. The highest five leverage ratios are observed in South Korea, Thailand, Indonesia, India and Brazil, while the lowest five are observed in Greece, South Africa, Turkey, Israel and United Kingdom. Developing economies seem to dominate the higher range, while developed economies tend to be at the lower range. The median leverage ratio for the developing economies in the sample is 0.32, 12 while for the developed economies the median leverage ratio is The middle range of the leverage spectrum is mixed with both developing and developed economies. [Figure 1 about here] Figure 2 presents the median maturity structure by country. It is clear from the figure that debt obligations have longer maturities in more developed economies. The five countries with the highest long-term debt ratios are Canada, Sweden, New Zealand, United States and Australia; the lowest five median long-term debt ratios are observed in Greece, China, Singapore, Thailand and Malaysia. 13 The median long-term debt ratio for the developing economies in the sample is 0.35, while for the developed economies the median long-term debt ratio is [Figure 2 about here] 12 Countries within the sample classified as developing are Brazil, Chile, China, India, Indonesia, Malaysia, Mexico, Pakistan, Peru, Philippines, Taiwan, Thailand, Turkey and South Africa. 13 This parallels the findings of Demirguc-Kunt and Maksimovic [1999] for an early sample period,

16 Table 2 presents the mean and median leverage and maturity structure by industry sector. The paper & pulp industry has the highest mean leverage, followed by construction, metal, wholesale, retail, food & beverage, chemical, resource, communication, newspapers, and business services. The mean fraction of long-term debt to total debt is the highest in the communication industry, followed by newspapers, resources, retail, paper and pulp, chemical, metal, food and beverage, business services, wholesale, and construction. [Table 2 about here] 3.3. Two-stage procedure We employ a two-stage procedure to test our hypotheses. In the first stage, we identify variations in leverage and maturity structures of firms that cannot be explained by firm- or industry-level characteristics but are related to country factors. In the second stage, we utilise the country-level estimates derived in the first stage. We examine whether country-level variations in leverage and maturity structure can be explained by the institutional factors discussed in the previous section. The two-stage procedure can be described by the following regression equations: DP it = β D + β D + β F [1] j t j t k t k t l t l it β = β C [2] j t m m t where the dependent variable in the first stage, DP for firm i is: it (i) leverage in period t, defined as the proportion of total debt to market value of the firm (total debt/market value). To allow for differences in the structure of 14

17 short-term debt across countries, we consider two alternative definitions of total debt (a) the book value of current and long-term interest bearing debt and (b) the book value of current and long-term interest bearing debt plus trade credit. Market value of the firm is defined to be the market value of common equity plus book value of preferred stock plus total debt, or (ii) debt maturity in period t. Debt maturity is measured as the proportion of long-term debt to total debt (long-term debt/total debt). We include dummy variables to examine the impact of country and industry on leverage and debt maturity. For each period t, j Dt is a vector of j country dummy variables and k D t is a vector of k industry dummy variables. l F it is a vector of l firm characteristics used in prior time series cross-sectional empirical studies as explanatory variables of capital structure and debt maturity choices. In the second stage we examine the extent to which the vector of coefficients on the country dummy j β t is explained by country specific factors. m C t is a vector of m country characteristic explanatory variables reflecting the impact of the institutional environment. 4. Firm level analysis The first-stage regressions include a set of firm level variables that capture factors that are known to affect leverage and maturity structure. 14 These variables include a firm s effective corporate tax rate, measured by income tax paid over earnings before 14 The set of variables is consistent with those identified by Guedes and Opler [1996] and Titman and Wessels [1988] in the context of the U.S. and Rajan and Zingales [1995] for a sample of developed countries. 15

18 interest and taxes, operating risk (absolute value of the change in profitability from period t-1 to period t) and asset tangibility (fixed assets over total assets), firm size (natural logarithm of total assets), the market-to-book ratio (market value of equity over book value of equity), and profitability (net income over total assets). In addition, asset maturity (gross property, plant and equipment over total assets times gross property, plant and equipment over depreciation) is included as a determinant of maturity structure at the firm level. The mean values of the explanatory variables are summarized in Appendix 3. The first-stage regressions are performed year-by-year using the ordinary least square method with the intercept suppressed, because of the specification of the dummy variables. In addition to the stated independent variables, dummy variables for the countries and the industry sectors are also included. However, to gain an overall idea of how leverage and maturity structures are related to the firm characteristics, we present in Table 3 the regression results of the pooled time-series and cross-sectional sample, and include year dummy variables. The overall results are consistent with those in prior research and generally support existing capital structure theories. In contrast to the predictions of theory, we find that leverage is negatively related to income tax over net income. 15 Although this result is consistent with the prior research [Booth, Aivazian, Demirguc-Kunt, and Maksimovic, 2001], it is inconsistent with the tax shield effect of debt. The inconsistency may arise from measurement bias in the expected tax level. One source of the bias is that actual taxes paid include tax losses carried forward from previous periods; hence, it may not properly reflect expected future tax liabilities. As we 15 This result is robust to an alternative specification of the tax variable, where the values of the ratio of income tax over net income lie within the range zero to one. 16

19 discuss below, we do find evidence that taxes influence capital structure choices in our cross-country analysis. [Table 3 about here] Leverage is positively related to asset tangibility and negatively related to the absolute value of annual changes in return on assets, consistent with the effects of financial distress. Also consistent with the prior studies, leverage is negatively related to profitability, which has generally been interpreted as supporting the pecking order theory [Myers and Majluf, 1984]. Leverage is also positively related to firm size and negatively related to the market-to-book ratio, which is consistent with the prior literature. Trade credit has been found to be an important source of financing in economies with underdeveloped financial intermediaries [Demirguc-Kunt and Maksimovic, 2001; Fisman and Love, 2003]. To permit an investigation of the impact of the level of trade credit, Table 3 includes results where the total debt includes trade credit (measured as accounts payable). The results change very little. Appendix 4 reports the results of the pooled time-series and cross-sectional regression for leverage by country. Consistent with Rajan and Zingales [1995] and Booth, Aivazian, Demirguc-Kunt, and Maksimovic [2001] we find that leverage is negatively related to both profitability and the market-to-book ratio in all countries. The variation in legal and tax systems across the 39 countries suggests that the negative relationship between profitability and leverage cannot be explained by the inability of firms to distribute cash flow to shareholders. With the exception of 4 out of 39 countries, leverage is positively related to firm size. While the overall relationship is weak, we find cross-country variation in the sign of the estimated coefficients for income tax over net 17

20 income, asset tangibility and absolute value of annual change in return on assets. Income tax is negatively related to leverage in 30 out of 39 countries, asset tangibility is positively related to leverage in 25 out of 39 countries and the absolute value of the annual change in return on assets is negatively related to leverage in 22 out of 39 countries. Table 3 also present the results of the maturity structure regression. Consistent with prior research [Barclay and Smith, 1995; Stohs and Mauer, 1996; Guedes and Opler, 1996; Demirguc-Kunt and Maksimovic, 1999], long-term debt is used more in firms with longer asset maturity, greater asset tangibility, lower volatility, larger size and higher profits. The market-to-book ratio is positive in these regressions but is insignificant. The results are similar with the inclusion of trade credit. However, with the inclusion of trade credit, the use of long-term debt is negatively related to the market-to-book ratio, but insignificantly related to profitability. Appendix 5 reports the results of the pooled time-series and cross-sectional regression for maturity structure by country. Consistent with Demirguc-Kunt, and Maksimovic [1999], for all countries size is positively related to debt maturity structure, while asset tangibility is positively related to debt maturity structure in 32 out of 39 countries. While the overall relationship is weak, we find cross-country variation in the sign of the estimated coefficients for income tax over net income, absolute value of annual change in return on assets, profitability, the market-to-book ratio and asset maturity. Income tax is negatively related to debt maturity structure in 24 out of 39 countries. Absolute value of annual change in return on assets is positively related to debt maturity 18

21 structure in 22 out of 39 countries. Profitability is positively related to debt maturity structure in 26 out of 39 countries. The market-to-book ratio is positively related to debt maturity structure in 21 out of 39 countries and asset maturity is positively related to debt maturity structure in 24 out of 39 countries. The estimated coefficients of the country dummy variables for the pooled timeseries and cross-sectional sample are presented in Appendix 6. With respect to both leverage and maturity structure, all the estimated coefficients are positive and significant. 5. Country level analysis This section examines whether country-level variations in leverage and maturity structure can be explained by the institutional differences across countries. We begin by presenting summary statistics of the country variables Summary statistics In addition to the independent variables described in Section 3, we include inflation and a developed economy dummy variable that takes a value of one if the country is classified as a developed economy according to the World Bank classification based on countries gross national income levels. 16 Inflation is added because debt contracts are generally nominal contracts and high inflation, which is generally associated with high uncertainty about future inflation, is likely to tilt lenders away from long-term debt. A developed economy dummy variable is included because it may pick up an element of financial development that is not already captured by our other variables. 16 The set of independent variables are defined in Appendix 7, along with their data sources. 19

22 To enable us to consider the interaction between market liquidity and the supply of funds, we include a dummy variable which takes a value of one if both the variable measuring market liquidity and the variable measuring the level of funds available to institutional investors are in the top 50th percentile of the sample and zero otherwise. The measures of market liquidity we use are stock turnover, stock market capitalization/gdp and government bonds capitalization/gdp. The measures of the level of funds available to institutional investors we use are life insurance penetration, and total institutional investment/gdp. As we mentioned earlier, we include this interaction variable because we believe that liquidity is only important if institutional investors, which are concerned about liquidity, provide a significant amount of funds. Summary statistics for these country level variables are presented in Table 4. The statistics reveal cross sectional variation in these variables. Except for the common law, developed economy, union bargaining power and tax system variables that are constant across time, all of the independent variables have varying annual observations. 17 The institutional investment variables have fewer observations because they are only available for OECD countries. [Table 4 about here] In addition, for each country, we compute the time-series median values of the estimated coefficients for the country dummy and the explanatory variables. The results are summarized in Appendix 8. The appendix reveals variations in these median values consistent with country level factors influencing leverage and debt maturity. For 17 The corruption index prior to 1995 is taken as the composite level, because compatible annual data is not available prior to

23 instance, the corruption level is high for Asian and Latin American countries (high leverage and short term debt) and low for North American and Scandinavian countries (low leverage and long term debt). Equity and bond market development are positively correlated, the correlation coefficient is (not reported). As anticipated, institutional investor participation is higher in developed countries Univariate correlations To gain a basic understanding of how capital and maturity structures are influenced by these variables, we compute the Pearson correlation coefficient between the country level variables and the estimated coefficients of the country dummies from the leverage and debt maturity regressions. The results, reported in Table 5, suggest that the legal system, the tax system, the suppliers of funds, union bargaining power and the information environment potentially influence the capital structure choice. In particular, low levels of corruption along with a common law legal system are associated with lower leverage ratios and a greater use of long-term rather than short-term debt, suggesting that the legal environment influences capital structure choices. In addition, firms in countries with high levels of union power tend to use longer term debt. Firms in countries with dividend imputation systems tend to have lower leverage ratios. In addition, Big-five auditor market share and analyst activity tends to be negatively correlated with the use of short-term debt as well as with the leverage ratio. Finally, firms in countries with more institutional investors tend to use significantly less short-term debt, and are less levered. [Table 5 about here] 21

24 The macro-variables, economic development and the inflation rate, are not highly correlated with the leverage ratio, but do seem to be highly correlated with maturity structure. Specifically, maturity structure is positively correlated with economic development and is negatively correlated with the inflation rate. The former correlation is probably due to the fact that countries with developed economies tend to have more efficient legal systems encouraging the use of long-term debt and the latter correlation is probably due to the problems associated with long-term debt in an environment with high and uncertain inflation. Economic development is also positively correlated with country level leverage. In addition, with liquid bond and equity markets, institutional investment is positively correlated with maturity structure Country-level leverage regressions This sub-section reports cross-country regressions that explain the variation in the coefficient of the country dummies that are estimated in our first stage regressions. The second-stage regressions are performed on the pooled country-level data using the ordinary least square method with heteroscedastic / autocorrelation corrected (HAC) errors [Andrew, 1991]. The HAC procedure accounts for the potential heteroscedasticity and auto-correlation in the data by deriving the t-statistics of estimated coefficients from Generalized Methods of Moments (GMM) standard errors corrected for heteroscedasticity and auto-correlation. The dependent variables in the regressions (Equation [2]) are the estimated coefficients for the country dummy variables from the first-stage year-by-year firm level regressions. Essentially, these estimates provide the 22

25 mean leverage and debt maturity ratios in each country adjusted for differences in industry and firm characteristics. [Table 6 about here] Table 6, which present the results of the regression of our country-level leverage dummies on institutional variables, reveal that capital structure is significantly related to a number of institutional and macro variables. 18 For the full sample, the first column in Table 6 reports the regression results for the case where the leverage ratio does not include trade credit and the second column reports the regression results for the case where trade credit is included. The results reported in the two columns are very similar. The regressions reveal that country-level leverage is positively related to economic development and negatively related to inflation. In addition, weak legal system integrity (measured by the corruption index) is associated with higher leverage; however, whether or not a country has a common law system does not appear to be important in this multiple regression. 19 Further, we do not find a significant association between leverage and union bargaining power (measured by the level of statutory protection of unions). Consistent with prior research, 20 leverage is lower in countries with a dividend 18 The results are similar across sub periods 1991 through 1996 and 1997 through 2000, and are robust to the use of alternative proxies for the country level variables - country s legal system, corruption, taxation, financial market development and institutional investment activity. 19 A possible explanation for the insignificant result is that the corruption index substitutes for the common law variable. However, the correlation between the two variables (-0.199) is statistically insignificant. Demirguc-Kunt and Maksimovic [1999] include a measure of the extent to which the legal system is utilized to resolve conflicts. They report that this measure is highly correlated with the common law variable (0.699). The corruption index reflects the extent to which corruption is perceived to exist among public officials and politicians, where corruption is defined as the abuse of public office for private gain. In this context the index proxies for the threat of expropriation of the property rights of investors. 20 For example, Twite [2001] finds that the introduction of a dividend imputation tax system into Australia led to a reduction in leverage. 23

26 imputation tax system and stock market turnover is positively related to leverage. 21 However, we do not find a significant relation between leverage and the development of the banking system, 22 but do find that leverage is negatively related to the size of the government bond market. This latter finding suggests that there may be a fixed demand for fixed-income securities and government bonds squeeze out corporate bond issues. 23 The size of the life insurance industry is insignificant, but the variable that interacts the size of the life insurance industry with the size of the government bond market is significantly positively related to the leverage ratio, indicating that life insurance firms may be more likely to buy corporate bonds if there is also an active government bond market. The presence of Big-five auditors is associated with lower leverage, consistent with their information role that facilitates equity financing. However, the similar effect is not found for equity analysts. 24 Table 6 also presents regression results for the sub-sample of OECD countries. For these countries we have more detailed information about the assets of institutional investors. From Appendix 8, OECD countries are Australia, Austria, Belgium, Canada, Switzerland, Germany, Denmark, Spain, Finland, France, United Kingdom, Greece, Italy, Japan, Korea, Mexico, Netherlands, Norway, Portugal, Sweden, Turkey and USA. With the exception of Mexico and Turkey these are classified as developed economies. The 21 Demirguc-Kunt and Maksimovic [1999] report a similar result. 22 There exists a potential for reverse causality associated with the measure for bank (deposits/gdp), stock market (stock market turnover) and bond market (government bond capitalization/gdp) size include in the country level leverage regression. The results are unchanged if these variables are excluded from the regression. 23 In theory, government bonds can either help or hurt the market for corporate bonds. If the market for fixed income securities is fixed, government bonds will squeeze out the corporate bonds. However, it is also likely that a larger government bond market will promote the development of the fixed income sector, which can, in turn, promote the development of the corporate bond market. 24 The results are consistent with those reported in Table 6 when estimated for each of the eleven industry sectors. 24

27 overall results are consistent with those reported for the full sample. We find that leverage is lower in countries with dividend relief or dividend imputation tax systems. The presence of equity analysts is significantly negatively related to leverage in OECD countries. This result, together with the insignificant relationship identified in the full sample, suggests that analyst activity play an important information role in developed economies whose equity markets are generally more developed. By contrast, our evidence suggests that equity analysts do not perform a similar role in developing economies whose equity markets are generally less developed. As reported in Appendix 8 as well as in Chang, Khanna, and Palepu [2000], the less developed markets attract a smaller number of analysts. On the other hand, the market share of Big-five auditors is insignificantly related to leverage in OECD countries. This result, together with the significantly negative relation identified in the full sample, suggests that the information role of Big-five auditors is generally important in developing economies and is generally less so in developed economies. This may reflect the fact that developed economies have more substitute information and governance mechanisms than do developing economies. Consistently with this argument, Fan and Wong (2002b) find Big-five auditors play an important information and governance role in mitigating agency problems in East Asia, a result not found in prior studies that focus on developed economies. In addition, we find that in OECD countries the presence of institutional investors, i.e., insurance companies, pension funds, and mutual funds, is associated with the greater use of equity. 25

28 5.4. Maturity structure regressions We repeat the regression analysis for maturity structure. Table 7 presents the regression results based on the full sample of countries. 25 Debt maturity, measured by the ratio of long-term to total debt, is strongly negatively related to corruption and weakly positively related to the common law dummy variable. Together the results suggest that the integrity of a country s legal system is important, and perhaps more important than the legal system per se, in shaping firms financing choices. In contrast to the insignificant result reported for leverage, debt maturity is positively related to the level of union bargaining power. In addition, debt maturity is positively related to the degree of life insurance penetration in a country, consistent with their interest in holding long-term securities. Debt maturity is negatively associated with stock market capitalization and unrelated to both stock market turnover and inflation. In contrast to the negative but insignificant result reported by Demirguc-Kunt and Maksimovic [1999], debt maturity is significantly negatively associated with the size of the banking sector. This may in part be explained by differences the variables used to proxy for the size of the banking sector. We use deposits (or liquid liabilities)/gdp a measure of the amount of funds that can be provided by the banking sector, whereas, Demirguc-Kunt and Maksimovic [1999] use domestic bank assets/gdp, which measures the funds that were provided. Finally, debt maturity is not significantly related to the level of economic development, the size of the government bond market, 26 the activities of Big-five auditors 25 The results are robust to the use of alternative proxies for the country level variables - country s legal system, corruption, taxation, financial market development and institutional investment activity. Alternative proxies leave unaffected other estimated coefficients. 26 There exists a potential for reverse causality associated with the measure for bank (deposits/gdp), stock market (stock market turnover) and bond market (government bond capitalization/gdp) size include in the country level debt maturity regression. The results are unchanged if these variables are excluded from the regression. 26