1
UNIVERSITEIT VAN AMSTERDAM
MSc Business Economics, Finance track
MASTER THESIS
“NATIONAL CULTURE AND THE CAPITAL STRUCTURE”
N.A. Velner, 6073344
July, 2014
2
Abstract. Using Hofstede’s six dimensions and the World Governance Indices as proxies for national culture, this study analyses whether there exists a relationship between national culture and firms’ capital structure. The study uses a sample of 53 countries around the world and compares these to the U.S.. The goal of the study is to address ambiguity surrounding the concepts national culture and capital structure and specifically provide more understanding on the determinants of capital structure. Results suggest that national culture affects firms’ book leverage ratio. The effects are small, but remain statistically significant even after controlling for other determinants of the leverage ratio. The highest effects are observed using the index comprised of Hofstede’s six dimensions. Based on this same index, results suggest that firms in countries approximated to be culturally similar to the U.S. have increasing book leverage ratios. The effect diminishes and even becomes negative the higher countries’ cultural deviation from the U.S.. Moreover, again based on this same index, reducing the sample to solely G-7 countries shows that for (developed) countries that work together closely the national culture factor has less of an effect on the book leverage ratio. However, this cannot be confirmed using the other national culture proxies.
3
Table of contents Title page………. 1 Abstract………... 2 Table of contents.……… 3 1. Introduction……… 5 2. Literature review………... 82.1 Cross-country capital structure……… 8
2.1.1 Classical theories.……….. 8 2.1.2 Firm-specific determinants……… 9 2.1.3 Country-specific determinants……….. 10 2.2 National culture……….. 11 3. Empirical methodology………. 13 3.1 Hypotheses………. 13 3.2 Sample……… 17 3.3 Model specification……… 18 3.4 Robustness checks………. 19 4. Data……… 20 4.1 Dependent variable……… 20
4.2 Independent variable of interest……… 21
4.2.1 Hofstede’s (2010) dimensions………. 21
4.2.2 Worldwide Governance Indices (2014) ………. 23
4.3 Control variables……….. 25 4.3.1 Firm-specific……… 25 4.3.2 Country-specific………. 26 4.4 Summary statistics………... 27 5. Results……… 30 5.1 Main results………. 30
5.1.1 Regression model one; Hofstede index……….. 30
5.1.2 Regression model two; Individual Hofstede dimensions……….. 34
5.1.3 Regression model three; WGI index……….. 37
5.2 Results by countries’ cultural distance from U.S……… 39
4
5.4 Limitations……… 44 6. Conclusion………. 45 7. References………. 48 Appendix A……… 55 Appendix B……….... 665
1. Introduction
According to Modigliani and Miller (1958) in perfect capital markets the capital structure is irrelevant to the value of the firm. However, these perfect capital markets do not exist and there may be an incentive to take on more debt due to for example the tax shield. In reality firms show different capital structures, especially when comparing firms cross-border. Whether this is because of the increased financial risk that comes with debt financing or more behavioural aspects play a role remains unclear. Based on extant empirical studies there exists no clear underpinning of what affects capital structure and no optimal leverage ratio is yet defined.
Traditional theories suggest firms select their capital structures based on a balance between the costs and benefits of leverage financing. Aside from these traditional theories, other prevailing theories such as Baker and Wurgler’s (2002) market timing are also present. Baker and Wurgler (2002) find fluctuations in equity market value have an impact on firms’ capital structure for at least a decade. This theory seems to have substantial explanatory power, but can yet again not fully explain the capital structure. Further, Lemon et al. (2008) argue that capital structures remain fairly stable over long periods of time, because the variation in firms’ leverage ratios is driven by an unobserved time-invariant effect. Titman and Wessels, already in 1988, argue it might be that the explanation lies in firm attributes that are expressed in terms of abstract concepts that are not directly observable. This study takes this into account and tests whether national cultures play a statistically significant role in the determination of the capital structure.
Previous literature on capital structure is mostly based on firms in the United States (U.S.) (Chui et al., 2002) and regarding cross-border researches; mostly institutional or economic differences are compared. Rajan and Zingales (1995) for example compare capital structures of firms residing in G-7 countries based on institutional differences and find that at the aggregate firm leverage looks similar across these countries, but a thorough look suggests that the observed correlations are still unsolved. Only a few recently hand out cultural differences as an important determinant. Chui et al. (2002) state “Culture matters because it affects management’s perception of the cost and risk related to debt finance and agency problems in each country”. However, most papers use indirect measures for
6
national culture due to its ambiguity in interpretation (Chang et al., 2012). Consider Stulz and Williamson (2003) who study differences in culture measured by differences in religion and language when examining investor protection across countries. These kinds of proxies reflect only “partial aspects of national cultures, and might not capture the multidimensional effects of these cultures (Chang et al., 2012)”. This study contributes to the existing literature because of the lack of consensus that the existing literature provides in the fields, especially that of capital structure. Furthermore, it contributes to the growing work into the fields of both capital structure and national culture.
This study tests whether national cultural differences, mainly measured by Hofstede’s (2010) six cultural dimensions, affect firms’ corporate capital structure. The research question therefore states: Do national cultural differences affect firms’ capital structures? In order to test whether this is the case the developed model in the study investigates the relationship between two proxies of national culture and the book leverage ratio. The first proxy is an index comprised of Hofstede’s (2010) six national culture dimensions. The dimensions are also used as regression inputs separately. Earlier research did not use all six of Hofstede’s (2010) dimensions (including Chang et al., 2012; Gleason et al., 2000; Kearney et al., 2012; Kwok and Tadesse, 2006) and brought these dimensions into relation with the capital structure. In this sense the study is unique in its variable choice on national culture. The second proxy contains an index sore based on the World Governance Indices (WGIs) (2014). Both proxies are referenced to the U.S., generating index variables that represent countries’ cultural distance from the leading economy of the world. The book leverage ratio is defined as the total book value of debt divided by the total book value of assets, where most variables are found in Standard and Poor’s annual COMPUSTAT database (Compustat). The study thus analyses whether differences in firms’ book leverage ratios across countries can be explained by differences in national cultures as they deviate from the U.S.. The study is based on 53 economies besides the U.S., selected based on Hofstede’s (2010) study. The aim of the study is to provide more insight and evidence in how capital structure is formed. Results show national culture weakly affects the book leverage ratio. The coefficients are statistically significant, even after controlling for other determinants. The effects are highest using the index comprised of Hofstede’s six national culture dimensions. Furthermore, results
7
suggest that firms in countries approximated to be culturally similar to the U.S. have increasing book leverage ratios. The effect diminishes and even becomes negative the higher countries’ cultural deviation from the U.S.. What causes this effect remains unclear, but results indicate the dimension uncertainty avoidance index plays an important role. Also, when reducing the sample to solely G-7 countries the national culture factor appears to be a weaker determinant of the leverage ratio. This suggests that for (developed) countries that work together closely financial structures are comparable. However, this cannot be confirmed using the other national culture proxies.
The topic may be relevant to firms operating in a volatile global economic environment, where economic turmoil can create opportunities regarding mergers and acquisitions (M&As) (ITAP, 2014). Various studies find a relationship between cultural differences and M&As (Chakrabarti et al., 2009; Kogut and Singh, 1988; Li et al., 2011; Stahl and Voigt, 2008). Knowledge about the effect of cultural dimensions on capital structure may enhance the value-creation of these M&As. Furthermore, in the light that Europe is working together closer and closer and capital structure might mitigate agency conflicts (Chang et al., 2012), a higher understanding of the topics is preferred and possibly hints where firms’ power of control lies. Extant literature shows that national culture is linked to managerial behaviour and corporate financial decisions (Chang and Noorbakhsh, 2009; Doney et al., 1998; Han et al., 2010). Furthermore, since a lot of the governance systems around the (western) world are based on U.S. systems, for investors and policy makers it would be interesting to see whether cultural and corporate governance deviations from the U.S. lead to differences in financial structure.
The remaining study is structured as follows. Chapter 2 contains the literature review. Section 2.1 focuses on literature behind the capital structure. It starts with some previous findings relevant for this study and continues with classical capital structure theories and possible determinants. Section 2.2 focuses on literature behind the concept national culture. Chapter 3 contains the empirical methodology and builds further on the literature review by constructing two hypotheses. Following the hypotheses, the chapter comments on the used sample, regression model specification and the robustness checks. Chapter 4 contains a description of the data used, with special attention to the Hofstede (2010) indices. It also contains a section summary statistics where different correlation tests
8
are performed. Chapter 5 contains the results of the study. Section 5.1 shows the main results, presented following the three different regression models. Section 5.2 comments on regression results regarding the effect various cultural distances from the U.S. have on firms’ leverage ratios. Section 5.3 and 5.4 show the results of the robustness checks and the limitations of the study respectively. The study ends with the conclusion and future research recommendations.
2. Literature review
2.1 Cross-country capital structure
This study builds on the literature both of capital structure and the role of national culture in finance. On the determinants of capital structure extensive literature exists, starting with the traditional propositions of Modigliani and Miller (1958) about perfect capital markets, the trade-off theory, and the agency theory. Also Stewart Meyer’s (1984) pecking order theory, where managers prefer internal sources of debt before equity financing, is used much. Booth et al. (2001) provide evidence that choices on capital structure in emerging markets are affected by the same variables as in developed countries, which is relevant for this study since the study is based on firms around the world with varying economic conditions. Booth et al. (2001) conclude country-factors are at work, but also state the influence of other characteristics cannot be excluded. Fan et al. (2011) show that the capital structures among developed countries are at the aggregate dissimilar and the leverage ratio increases when a country has poor public governance. Frank and Goyal (2003) find that debt financing does not dominate equity financing in publicly traded U.S. firms, whereas this would be predicted by the pecking order theory.
2.1.1 Classical theories
Modigliani and Miller (1958) showed in perfect capital markets the value of the firms does not depend on these firms’ capital structure. Regardless the amounts of debt, the cash flows generated by projects remain the same, and therefore the value of the firm does not change when the capital structure changes. However, perfect capital markets do not exist and in practice other theories are also at work. The static trade-off theory yields an optimal leverage ratio by finding a balance between the costs and
9
the benefits of leverage (Myers, 1984), such as benefits created by the interest tax shield, financial distress costs and agency costs. Thus, this trade-off theory suggests that an optimal leverage ratio exists, but varies by firm. The static trade-off theory works to some extent, but empirical evidence suggests yet again it cannot fully explain the capital structures of firms (Myers, 1984; Graham and Harvey, 2001). Myers (1984) therefore put forward the pecking order theory. This theory states managers have a preference regarding financing methods and follow a so called ‘pecking order’; first use internal funds, otherwise debt, and if debt is not available use equity. This order is based on the notion that when managers perceive firms’ equity to be undervalued, managers would not want to issue it and rather use other ways of financing. One could assume this effect to be highest when managers have important private information about the firm. The concept is therefore closely linked to asymmetric information and Baker and Wurgler’s (2002) market timing theory. Empirical evidence exists supporting this theory, but again it appears to be not the sole determinant (Fama and French, 2002; Frank and Goyal, 2003; Graham and Harvey, 2001). Furthermore, the leverage ratio may also be influenced by the agency theory. When ownership and control of firms are separated and no perfect alignment of interests between shareholders and managers exists, large free cash flows increase the power of the CEO. This creates room for managerial discretion and entrenchment. Jensen (1986) hypothesized that large free cash flows induce managers to engage in wasteful spending, e.g. spend money on value-destroying investments. By taking on debt firms commit to making interest payments, thereby reducing the free cash flows available for managers to waste. This is supported by empirical evidence. For instance Berger et al. (1997) find entrenched managers use less debt in their capital structures. Thus, based on this theory debt (maturity) selection can be an effective mechanism to mitigate agency conflicts. Furthermore, Lemmon et al. (2008) find that leverage ratios and factors driving cross-sectional variation in firms’ leverage ratios are stable over long periods of time.
2.1.2 Firm-specific determinants
Numerous papers document a relationship between capital structure and specific firm characteristics. Most prominent are the relationships regarding firms’ profitability, growth, asset tangibility and firm size. Starting with profitability, on the one hand this concept is associated with the possible retained
10
earnings of the firm, and may thus correspond with less leverage under the pecking order theory where firms prefer internal funds over external funds due to asymmetric information (Baker and Wurgler, 2002). On the other hand, profitable firms that have large free cash flows have to face agency problems. To address these problems effectively, governance mechanisms lead to more leverage (Jensen, 1986; Baker and Wurgler, 2002). These views can be extended looking at firms’ growth. A firm facing high growth may have more positive Net Present Value (NPV) projects to engage in, therefore less free cash flows available, and thus has to deal with less agency problems. Hence, the agency problem expects there to be a negative relationship between growth and the leverage ratio. Firms with high growth rates are associated with more debt in the light of the pecking order theory, because this theory states that when internal funds are low firms prefer to issue debt over equity. Asset tangibility is expected to be positively related with the debt ratio. Tangible assets may be used as collateral, and consequently firms showing a high proportion of tangible assets on their balance sheets are able to increase their borrowings (Chui et al, 2002; Antoniou et al., 2008). Previous research on size documents a positive relationship to capital structure based on firms’ diversification (Kayo and Kimura, 2011). Larger firms may be more diversified, making them less prone to bankruptcy risk (Titman and Wessels, 1988). Moreover, larger firms can spread their cost of debt over a larger amount of debt that they were able to issue by being more transparent (Byoun, 2008; Kayo and Kimura, 2011). However, small firms may have higher debt ratios because they experience a higher cost of equity, e.g. pay more for issuing new equity (Titman and Wessels, 1988; Antoniou et al., 2008). Thus a negative sign on size invigorates on the view of information asymmetry. Li et al. (2011) document on a positive relationship between asset maturity and the leverage ratio.
2.1.3 Country-specific determinants
Country characteristics also prove to be related to firms’ capital structures. Antoniou et al. (2008) provide evidence that leverage is affected by market conditions and argue that the capital structure of a firm is heavily influenced by the investor protection of the country it operates in. Consider countries were investor protection is low, based on agency theories managers would there prefer to hold cash instead of distributing it to shareholders (Kayo and Kimura, 2011). La Porta et al. (1997; 1998) show
11
that countries with poor investor protections have less developed markets, e.g. have significantly smaller debt and equity markets (Chui et al., 2002) and common law countries protect more shareholder rights than civil law countries, with French civil countries being the worst. Fan et al, (2011) find that more corrupt countries and those with weaker laws tend to use more debt, especially short-term debt. Furthermore, firms in market-based economies tend to have a less concentrated ownership structures compared to firms in bank-based economies (Antoniou et al. 2008; Kayo and Kimura, 2011). National culture appears to be significant in bank-based systems, whereas it does not matter in a market-based economy where there is a better investor protection system (Chang et al., 2012). Moreover, countries with a higher legal protection to minority shareholders tend to be more market-oriented (Demirgüc-Kunt and Levine, 1999; Kwok and Tadesse, 2006). Relating this to agency theories again, one can argue that firms in market-based economies have higher leverage ratios. Furthermore, Kayo and Kimura (2011) argue in the light of De Jong et al. (2008) that when “the bond market in a given country is highly developed, issuing and trading bonds becomes easier and firm leverage tends to be higher”. In contrast, when “the stock market is developed, firm leverage is lower because the higher supply of equity funds decreases the cost of equity (Kayo and Kimura, 2011)”. Li et al. (1998) document a positive relationship between GDP growth per capita and the leverage ratio, which supports the idea that when a country is developed, now in terms of GDP growth, debt tends to be less risky, and thereby increases the likelihood of debt financing.
To conclude, effects of certain variables on firms’ capital structures are ambiguous. Moreover, none of the theories has proven to fully explain the (cross-national) differences in capital structures. Therefore more literature in this field is preferred.
2.2 National culture
The concept of culture is subjective, and a clear understanding lacks. Hofstede (2014) refers to culture as “the collective mental programming of the human mind which distinguishes one group of people from another and influences patterns of thinking which are reflected in the meaning people attach to various aspects of life”. Hofstede (2010) has created six cultural dimensions, talked about more
12
extensively in the variable description in chapter 4. Although many researchers have shown validity of Hofstede’s dimensions, critique is also present. Section 4.2.1 variable description elaborates on some of this critique. Other national culture examples include Inglehart (1997), Trompenaars (Trompenaars and Hampden-Turner, 1998) and the GLOBE-project (Chhokar et al., 2007). Another much used proxy for national culture was created by Schwartz (1994; 1999).1
National culture has been used previously to explain certain finance related questions, such as IPO under-pricing (Costa et al., 2012), cross-border acquisitions (Chakrabarti et al., 2009) and financial systems (Kwok and Tadesse, 2006). The latter found that countries with higher uncertainty avoidance index, one of Hofstede’s national culture dimensions, are more likely to have a bank-based system. Also, countries that score high on Hofstede’s uncertainty avoidance index, masculinity versus femininity and pragmatic versus normative in general have higher cash holdings (Chang and Noorbakhsk, 2009). Recently, a few papers have brought the concept national culture into relation with capital structure. Gleason et al. (2000) show that capital structures vary by cultural cluster. Also, Chui et al. (2002) find high scores on conservatism and mastery lead to lower corporate debt ratios based on Schwartz’ measure of national culture, even when there is controlled for various other capital structure determinants. Chang et al. (2012) focus on three of Hofstede’s national culture dimensions (uncertainty avoidance index, masculinity versus femininity and long-term orientation) and use short-term debt as dependent variable. Based on corporate governance theories they find that national culture influences the overall debt maturity negatively. Kearney et al. (2012) find a relationship between national culture and capital structure in small and medium sized firms. The Hofstede (2010)
1 Schwartz (1994) created six value types that could be classified into two cultural dimensions that encompasses two poles; autonomy versus conservatism and hierarchy and mastery versus egalitarian and harmony with nature (Chui et al., 2002). Even though studies have shown considerable convergence between the two measures and Schwartz’ later-created method was found to mostly agree with Hofstede’s (1980) conclusions rather than to contradict them (Smith and Bond, 1998; Oudenhoven, 2001), some researchers argued in favour of Schwartz’ measure. For example, Ng et al. (2006) found that in trade cultural distance scores derived from both national culture proxies showed the same sign and direction, but only Schwartz’ measure appeared to be statistically significant. Therefore they assume Schwarz’ measure to be superior. However, other researchers are indifferent or choose for Hofstede’s national culture measure. For example, Drogendijk and Slangen (2006) disagree with Ng et al. (2006). Moreover, Hofstede (1984) himself presented an example of synergy between different cross-cultural studies to prove validity of his dimensions.
13
dimensions uncertainty avoidance index and individualism versus collectivism are positively related to the amount of small-term debt and negatively related to the amount of long-term debt. Furthermore, a negative relationship was found between the dimension power distance index and short- and long-term debt in the capital structure.
3. Empirical methodology
3.1 Hypotheses
The adjusted Kogut and Singh (1988) index comprised of Hofstede’s (2010) measure for national culture and the book leverage ratio are expected to be correlated, even when there has been controlled for the various firm-specific and country-specific control variables. Various earlier studies document relationships between the different concepts and therefore the main hypothesis of this study states that national culture, measured by the adjusted Kogut and Singh (1988) index comprised of Hofstede’s (2010) six national cultural dimensions, affects firms’ capital structure.
H1: National culture has a statistically significant effect on firms’ capital structure.
Power distance index. This dimension relates to the expectation and acceptance about how power is distributed (un)equally, where a low score on this dimension indicates members of the society place greater value on power equality and interpersonal trust (Gleason et al., 2000; Hofstede, 2014). This study follows the link made by Gleason et al. (2000) based on Dahlstrom and Nygaard’s (1996) findings. Dahlstrom and Nygaard (1996) state that “investment in interpersonal trust enhances firms’ performance in Poland and Norway, where the interpersonal trust has emerged from the legal environments the firms are surrounded with and the extent of inter-firm managerial control (Dahlstrom and Nygaard, 1996)”. Recall that firm performance is related to firm value and capital structure, so one could hypothesize that the national culture dimension power distance index can be linked to capital structure as well. Based on the theories behind the firm-specific variables growth, profitability and size, for this national culture dimension solely a statistically significant relationship is expected with the capital structure, e.g. the coefficient sign on this national culture dimension could be positive or negative. Kearney et al. (2012) showed a negative relationship between this dimension and the
14
short- and long-term debt in the capital structure for small and medium sized firms. As explanation they put forward that when a country has a low power distance score, e.g. puts more value on equality, there may exist a more ‘consultative role’ with financial institutions and firms are therefore able to increase their borrowings.
Individualism versus collectivism. Regarding this national culture dimension ambiguous results are predicted. This dimension may be negatively related to the leverage ratio based on managers’ concern for maximizing their own reputation and success. One could argue that managers in countries exhibiting a high individualism score tend to look more after their own interest and try to enhance their own reputation (Gleason et al., 2000; Kearney et al., 2012). To maximize their own success they may be more likely to choose lower debt ratios. To the contrary, this dimension may be positively related to the leverage ratio based on the pecking order theory. Antonczyk and Salzmann (2013) document that one could link the dimension to managers’ overconfidence and overoptimism, where optimistic managers overestimate firms’ profitability and prefer debt to equity as they perceive their firms’ equity to be undervalued and their future cash flows will be sufficient to cover debt interest payments. Thus, based on the pecking order theory these managers are more likely to choose high debt ratios (Baker and Wurgler, 2002). This might be true especially for the benchmark country the U.S., where the society is perceived to be very individualistic with a Hofstede (2010) dimension score of 91.
Masculinity versus femininity. As Gleason et al. (2000) mentioned, not much direct empirical evidence is available showing how this national culture dimension exactly relates to firms’ capital structures. However, according to Breuer and Salzmann (2008) some researchers link this dimension to Schwartz’ (1994; 1999) national culture dimension mastery because both dimensions refer to assertiveness, activity and ambition of the members in the culture. Even though Breuer and Salzmann (2008) find a correlation of only .15 between the two dimensions, one could predict the following. Chui et al. (2002) note that when managers are more masculine they are eager to show their superior ability and performance to shareholders and investors. This may on the one hand imply firms use aggressive policies where managers do not want to be bonded by debt covenants or have to worry
15
about bankruptcy, hence these firms are expected to have less debt financing (Chui et al. 2002). On the other hand, to get recognition managers may want to signal the firm is in a good state, has good operating prospectus and therefore is able to pay interest on the debt, e.g. there may exist also a positive relationship with firms’ capital structures.
Uncertainty avoidance index. This national culture dimension refers to the extent members of a society relate to uncertain or unknown situations, where a high score means that a country has a low tolerance towards these uncertain and ambiguous situations and members tend to be more risk-averse (Chang et al., 2012). Firms in countries with a low tolerance for uncertainty may find financing with debt too risky, because debt increases their exposure to bankruptcy risk. They therefore prefer equity financing over debt financing. Thus, one could hypothesize that firms showing a high avoidance of uncertainty tend to show lower debt ratios. Alternatively, one could argue that firms take on more debt because of reduced borrower adverse selection and moral hazard problems. When a society is less tolerant towards uncertainty it is generally more rule-oriented, does not accept change easy and takes less risk (Chang et al., 2012). This may translate in firms exhibiting more complete accounting disclosures, reducing the borrower financial risk, making debt more appealing.
Pragmatic versus normative and indulgence versus restraint. On the last two dimensions not much direct empirical evidence exists towards the leverage ratio. For example, Chang et al. (2012) show that three of Hofstede’s (2010) indices, including the dimension pragmatic versus normative, are negatively related to the overall debt maturity in a country. The other two dimensions showing this negative relationship are the masculinity versus femininity and the uncertainty avoidance index. Nevertheless, as shown in section 4.2.1 the two variables pragmatic versus normative and indulgence versus restraint should be included in the regression models and also play am important role in the uniqueness of this study as mentioned in section 2.3.
To proxy national culture this study makes, aside from the Hofstede measures, also use of the WGIs varying by country. Empirical evidence suggests that national culture influences the corporate governance structures of firms and countries, which can subsequently be linked to corporate capital
16
structures based on the behaviour and decision-making of managers. Put differently, extant empirical evidence suggests managerial behaviour and discretion influence corporate decision-making and capital structure and depends (partly) on firms’ governance structures. These corporate governance structures depend in turn on the national cultures where the firms reside.The second hypothesis states that national culture measured using corporate governance structures varying by country affects firms’ capital structure.
H2: National culture, measured by a proxy for corporate governance structures, has a statistically significant effect on firms’ capital structure.
National cultural differences have proven to be a determinant in financial decision-making cross-country (Chang et al., 2012) and numerous management and psychology studies show that managers’ behaviour is closely linked to the (organizational) culture these managers are exposed to (Tsui et al., 2006). This is supported by empirical evidence showing corporate governance structures vary by country, and may thus be influenced by national culture subsequently managers’ behaviour. Consider the following studies. Holderness (2009) comments on the widely held idea that firms’ ownership structures are more concentrated in countries with weak investor protection laws. Becht and Roëll (1999) find that amongst various countries the concentration of shareholder voting power differs significantly. Haniffa and Cooke (2002) provide evidence that different environmental factors, including national culture proxies, influence disclosure practices. Buck and Shahrim (2005) illustrate based on a German sample regulatory and firm-level governance changes are subject to changes in national culture.
Corporate governance structures influence managerial behaviour and thereby corporate decision-making and the leverage ratio. As no perfect alignment of interest between shareholders and agents exists, the decision-making is influenced by the extent of managerial discretion and entrenchment. Recall Jensen’s (1986) hypothesis, related to these agency conflicts, that argues companies that hoard cash levels beyond their operational requirements and financing needs will destroy value by overinvestment. Poorly governed firms may provide managers with opportunities to
17
exercise their discretion (Chang et al., 2012). One could hereby think of firms with a large board comprised of insiders or firms showing high ownership diffusion. These self-centered managers will deviate from optimal leverage and debt maturity, and a firm value-maximizing level of debt is not reached (Chang et al., 2012). This is supported by various other studies, such as the following. Kester (1986) shows a relationship between leverage on the one side and ownership and governance structures on the other side based on a sample that encompasses firms in the U.S. and Japan. Berger et al. (1997) show that entrenched managers use less debt in their leverage ratio, supported with the notion that leverage is lower when the Chief Executive Officer (CEO) has a long tenure in office, has weak stock and compensation incentives and does not face string monitoring. Bennedsen et al. (2007) provide evidence that board size can be related to firms’ performance in small and medium sized firms. Fracassi and Tate (2012) show that including board members with CEO and management ties reduces firm value, especially in the absence of other governance mechanisms to substitute for board oversight. A way to deal with this managerial discretion is to alter the corporate governance structure, and for example increase the leverage ratio.
3.2 Sample
For the primary sample selection, countries are included on which Hofstede (2010) provides cultural scores on all six dimensions. For these countries data was extracted from the other data sources to compute the leverage ratios and other variables. The time period used is eleven years, from January 2002 till December 2012. Industries are determined using SIC codes. In line with previous research (Berger et al. 1997; Baker and Wurgler, 2002; Leary and Roberts, 2014), financial firms are excluded from the sample, because restrictions exist regarding the leverage ratios of these firms. Excluding these financial companies entails excluding all firms with a SIC code between 6000 and 6999 (Baker and Wurgler, 2002). Moreover, Leary and Roberts (2014) exclude utilities as well as government entities to maintain consistency with previous empirical studies and to avoid capital structures dictated by regulatory considerations. This means for this study firms with SIC codes between 4900 and 4999 and greater than or equal to 9000 are excluded as well. Furthermore, firm-year observations for which book leverage exceeds 1.0 are removed from the sample, since 1.0 is the upper limit of book leverage
18
(Baker and Wurgler, 2002). Note that the sample is comprised of publicly listed companies. This is the case because the Compustat database from Standard and Poor’s contains fundamental and market information on solely active and non-active publicly held companies, and thus does not show the complete firm collection of the countries. Furthermore, apart from the independent variables of interest, data is winsorized at the 1% level prior to performing the regressions. The independent variables of interest are not winsorized because these variables are already within a certain range and therefore cannot have outliers. Moreover, Leone et al. (2013) document that of a large sample of studies only 33% winsorize both the dependent and independent variables and that winsorizing each variable (including the independent variable of interest) has only a modest impact on parameter estimates compared to not winsorizing or truncating the data. This all leaves the three main regression models with at least 29172 observations across 5418 firms in 53 countries aside from the observations in the U.S..
3.3 Model specification
For the main results this study makes use of three multivariate panel data Ordinary Least Squares (OLS) regression models. These OLS regression models are shown below, where i represents the company by firm identifier and t the fiscal year. In the first regression model, the independent variable of interest represents the outcomes of the adjusted version of the Kogut and Singh (1988) index where Hofstede’s (2010) six national culture dimensions were used as inputs. This independent variable remains constant over the firms and years, but differs across countries. This also holds for the independent variables of interest in the second model, where the independent variables of interest now represent the individual Hofstede’s (2010) national culture dimensions. Note that solely four out of the six national culture dimensions are used in the regression model, since some of the culture dimensions appear to be correlated. A more detailed description of the correlations between the various cultural dimensions is shown in section 4.4 summary statistics. In the third model, the independent variable of interest represents the outcomes of the adjusted Kogut and Singh (1988) index comprised of the scores following the WGIs (2014). This independent variable differs across countries and years. There exists a time lag. Culture changes are long-term changes, and even for short-term changes it might take time
19
before these are fully integrated. Especially when it comes to the leverage ratio. Furthermore, recall the findings by Lemmon et al. (2008) about the relatively stable capital structures over long periods of time. The leverage ratio at time t thus is affected by previous national culture, which the model restates by using the subscript t-1 for the variable representing national culture. This also holds for the control variables, apart from the dummies. Since the proxy for national culture in the first regression model only changes across country, not across year, for this variable in the regression it is not necessary to take the lagged version. The Hausman test provided evidence that fixed effects over random effects are preferred in the regression models. However, random effects are used because the key explanatory variables do not have much time variation. In addition, various other explanatory variables are included to control for this. Data is analysed using STATA and all standard errors are clustered.
3.4 Robustness checks
Two robustness checks are performed. First, as already stated earlier, market leverage is used as an alternative measure for capital structure. It will test whether accounting differences play a significant role in the results obtained.
The second one includes reducing the sample to solely the G-7 countries; Canada, France, Germany, Italy, Japan, the U.K. and the U.S.. These countries represent the leaders of the developed countries and share a lot of businesses (NOS, 2014). It would therefore be interesting to see whether cultural differences in this subsample still have a statistically significant impact on firms’ capital structures. Recall Rajan and Zingales (1995) found that the leverage ratios of G-7 countries look fairly similar and the differences that do exist cannot be explained by solely institutional differences. Therefore, one could hypothesize that national culture is in absolute effect a weaker determinant of the leverage ratio in the subsample consisting of solely G-7 countries compared to the full sample.
20
4. Data
4.1 Dependent variable
The dependent variable equals the book leverage ratio. This study defines the book leverage ratio as total book value of debt divided by total book value of assets (Compustat annual data item 6). The majority of the variables are determined following the procedures of Baker and Wurgler (2002), which also holds for the construction of the control variables as is shown later in the study. The total book value of debt is defined as total assets minus book equity, where book equity is defined as total assets less total liabilities (item 181) and preferred stock (item 10) plus deferred taxes (item 35) and convertible debt (item 79). When the preferred stock is missing it will be replaced by the redemption value of preferred stock (item 56). As a proxy for market leverage, book debt is divided by the result of total assets minus book equity plus market equity. Market equity equals common shares outstanding (item 25) times price (item 199). Market leverage will be used in a robustness check as an alternative measure for capital structure, as was explained in section 3.4. See Table A-I in Appendix A for a complete list of variable definitions and their corresponding data sources. Table A-II in Appendix A lists the Compustat annual items corresponding to the used variables.
There are numerous reasons whether to use book leverage or market leverage as dependent variable. According to Barclay et al. (2006), book leverage is a better measure than market leverage because it captures the value of assets in place and not the growth options reflected by current market values. However, book equity may be negative and the correlation between book value and market value may be weak when firms are still small (Welch, 2004). If this is the case, market leverage might be a more appropriate measure, because market value is closer to the intrinsic value and it provides a better perspective of the potential for future leverage (Kayo and Kimura, 2011). This study uses the book leverage as standard, as it solely wants to catch the value of assets in place and the firms in the sample are large with a total assets mean of $8,484M.
21
4.2 Independent variable of interest
4.2.1 Hofstede’s (2010) dimensions
Hofstede’s (2010) relative six national culture dimensions that distinguish countries from each other are used as independent variable of interest. These measures of Hofstede (2010) have been used in many researches, from human social science to healthcare and business. The measures have been created in 1980, the year Hofstede published the book Culture’s Consequences.2
Systematic differences in survey questions were observed and values that emerged from surveyed preferences for certain states grouped themselves statistically into four clusters, now referred to as the first four of Hostede’s (2010) six dimensions (Hofstede, 2014). Hofstede (2010) stated that he was influenced by his own Dutch culture leading to biased ‘Western questionnaires’. 3 More research resulted into adding the fifth and sixth dimensions. Moreover, the additional research extended the studied number of countries to 93. Inclusion of the last two dimensions is thus needed for a more reliable and comprehensive view of national culture.
Important to understand is that Hofstede (2010) quantified national culture by giving countries scores along the six dimensions. These dimensions reflect the national cultures and not the organizational cultures. Hofstede (2014) argues on his website, “national cultures belong to anthropology and are given facts for organization management, whereas organizational cultures belong to sociology and are somewhat manageable”. The six cultural dimensions are the following: “1) Power distance index (PDI): Expectation and acceptance of less powerful members of organizations and institutions about how power is distributed (un)equally, 2) Individualism versus collectivism (IDV): Degree of independence a society maintains among its members, 3) Masculinity
2 This book combined Hofstede’s personal experiences with statistical analysis of two unique databases, and commented on systematic differences between the countries in these two databases. The first database contained answers of matched employee samples from forty different countries to attitude survey questions, whereas the second database was comprised of answers to some of the same questions from fifteen countries and from a variety of companies and industries (Hofstede, 2014). 3
Michael Bond found a solution to this ‘Western bias’ problem by creating the Chinese Value Survey
(CVS). The results revealed the existence of a fifth cultural dimension: Long-term versus short-term, later also called Pragmatic versus normative (PRA). Misho Minkov unravelled from the World Value Survey based on Bond’s measure a dimension which he called Indulgence versus restraint (IND).22
versus femininity (MAS): Refers to the emotional roles between genders and the motivation of people; wanting to be the best or liking what another does, 4) Uncertainty avoidance index (UAI): Deals with a society’s tolerance for uncertainty and ambiguity and thus expresses the extent members of culture feel threatened by ambiguous or unknown situations and have created beliefs to avoid these, 5) Pragmatic versus normative (PRA): Degree of society-orientation when members relate to the fact that things around cannot be explained, where long-term orientation fosters pragmatic virtues towards future rewards and short-term orientation fosters virtues related to the past and present, 6) Indulgence versus restraint (IND): Extent members of the society try to control their desires and impulses (Hofstede, 2014)”.
Even though the measures of Hofstede (2010) comprise a comprehensive database, critique on the measures is also present. First, one can argue about its validity on a more recent sample. According to Kirkman et al. (2006) countries in earlier studies were selected to be maximally different on cultural value (Chang and Noorbakhsh, 2009), leading to reasonable variances and cultural differences in line with Hofstede (1980). Furthermore, Hofstede (2010) has revised the measures and included two additional dimensions. Hofstede (2014) states that the relative scores have been proven to be quite stable over time and thus can be used for a recent sample of firms as well. Second, the measures are based on one single company’s data and have therefore been criticized for being too simplistic and failing to capture within-country cultural heterogeneity (Kirkman et al., 2006; Chang and Noorbakhsh, 2009). However, as Chang and Noorbakhsh (2009) suggest, one could interpret this as clarity and prefer the measures due to its parsimony and resonance with managers (Chang and Noorbakhsh, 2009). Third, as previously stated Hofstede’s measures are relative and are not to be interpreted as reality. Hofstede (2010) emphasizes that when a national culture shifts, the shift tends to be global or continent-wide and solely the relative country scores between the countries hold where societies are compared to other societies. The dimensions are orthogonal and without comparison a country score is meaningless (Hofstede, 2010).
To address all previous mentioned problems, the six dimensions are used as input in an adjusted version of the Kogut and Singh (1988) index as measure for national culture. The index
23
provides scores representing countries’ cultural deviation from a reference country. Despite some criticism, such as provided by Shenkar (2001), this method is commonly adopted by previous empirical research. The Kogut and Singh (1988) method is stated in Equation (3) in Appendix A, whereas the adjusted algebraic version is stated in equation (1) and is as follows.
∑ ( )
,
( 1 )where is the cultural distance of the jth country and the ith cultural dimension from the corresponding cultural dimension score of the U.S., represents the index for the ith cultural dimension and the jth country. represents the cultural dimension score for the country U.S. and represents the variance of the index of the ith dimension. There has been chosen for the country U.S. as benchmark, because for the used time period it is considered to be the ‘leading’ economy in the world. For example, from the year 2002 till 2012 the U.S. were the world’s largest economy measured by GDP (Bergman, 2014). This means that the higher the index score, the more this country is perceived to behave culturally differently compared to the U.S, which in turn indicates the country behaves culturally differently compared to the leading economy in the world. The study thus analyses the dynamics of capital structure cross-country as the countries deviate in national culture from the U.S. Recall that the research question is: Do national cultural differences affect firms’ capital structures? In light of the previously explained method for measuring national culture, the answer to this question is established by testing whether countries that have a higher cultural distance compared to the leading economy of the U.S. on average helps explain whether firms choose for different capital structures. There has been made use of the natural logarithm of this index variable to increase the fit of the model. Data is extracted from the website of Hofstede (2014).
4.2.2 Worldwide Governance Indices (2014)
The Worldwide Governance Indicators project reports aggregate governance indicators for two hundred fifteen economies from the year 1996 up till now (WGI, 2014). The indicators represent the views of a large number of enterprises, citizens and expert survey respondents in industrial and developing countries and are created for the following six dimensions of governance: “1) Voice and
24
accountability: Extent to which citizens are able to participate in selecting government, freedom of expression, association and media, 2) Political stability and absence of violence: Likelihood government will be destabilized or overthrown by unconstitutional or violent means, 3) Government effectiveness: Quality of public services, civil service and the degree of its independence from political pressures, policy formulation and implementation and the credibility of the government’s commitment to such policies, 4) Regulatory quality: Ability of government to formulate and implement sound policies and regulations that permit and promote private sector development, 5) Rule of law: Extent to which agents have confidence in and abide by the rules of society, and in particular the quality of contract enforcement, property rights, the police and courts and the likelihood of crime and violence, 6) Control of corruption (WGI, 2014)”.
Each of the scores on the corporate governance dimensions is constructed by averaging together data from the individual sources that correspond to the concepts of governance. The following method was used. First, data from the individual data sources was assigned to the six aggregate dimensions and subsequently rescaled. Second, a weighted average of the individual indicators for each dimension was constructed using an Unobserved Components Model (UCM), resulting into estimates of governance that are a weighted average of the data from each source with weights reflecting the pattern of correlation among data sources (WGI, 2014). A greater weight has been given to data sources that tend to be more strongly correlated with each other (WGI, 2014). The ultimate measures of governance generated by the UCM have a mean zero, a standard deviation of 1 and go from approximately -2.5 to 2.5. Important is that the WGI (2014) notes that year-to-year periods are difficult to measure with any kind of data and differences are typically small. This strengthens Hofstede’s (2010) notion on the applicability of his measures over years. Higher WGI dimensions scores correspond to better governance. Also for this WGI variable the adjusted version of the Kogut and Singh (1988) index shown in Equation (1) is used to create a measure for national culture. The benchmark country is again the U.S., leading to index values that represent the deviation of countries’ corporate governance structure from the U.S.. The higher the index score, the higher the
25
governance deviation. There has been made use of the natural logarithm of the index variable to improve the fit of the model.
4.3 Control variables
The control variables are based on prior literature and are separated into two groups; firm-level control variables and country-level control variables. On the determinants of capital structure a lot of literature exists, but Kayo and Kimura (2011) state that the majority of leverage variance is due to the firm and time level. As mentioned earlier, Lemmon et al. (2008) found that the majority of variation in the leverage ratio is determined by time-invariant factors. Moreover, in their study the adjusted R-squareds increase when including time-fixed effects. This study includes time-, country- and industry-fixed effects. The country-industry-fixed effects are used as alternative for the country-specific control variables, as well as supplementary excluding the country-specific control variable GDP growth. See Table A-III in Appendix A for a clear overview of the control variables’ predicted coefficient signs.
4.3.1 Firm-specific
Kayo and Kimura (2011) derive from their results that firm-specific factors are the main drivers of variances in leverage. In line with other previous empirical research, as was commented about in section 2.1, they provide the following five firm-level determinants of capital structure: growth opportunities, profitability, distance from bankruptcy, size and tangibility. Firm growth is measured by market-to-book value and equals book equity plus market equity all divided by total assets. The proxy for firm profitability equals the Return On Assets (ROA) and is calculated by dividing Earnings Before Interest and Taxes (EBIT) (item 13) by total assets. For firm size, the natural logarithm of net sales will be used, Log(Net Sales) (item 12). The logarithmic rescaling on net sales makes the firm sales data fit better to the regression models. A precise proxy for firm bankruptcy is not included, because it will be correlated with the size of the firms. Recall from section 2.1 that larger firms may be less prone to bankruptcy risk and therefore are able to issue more debt, consequently enjoy a lower average cost of debt. Firm asset tangibility is defined as the ratio of fixed assets to total assets. Note that the fixed assets are determined as total assets minus current assets. The annual data item
26
corresponding to current assets (AC) is 4. In addition, Li et al. (2011) also include firm asset maturity as a control variable, measured by the following formula.
( ) ( ) ( ) ( )
,
( 2 )where the Cost Of Goods Sold (COGS) annual data item is 41 and the annual ite for depreciation is 14. Furthermore, SIC numbers will be used to control for industry variation in leverage. First, the regressions are performed using the four-digit SIC numbers. Second, the regressions are performed using solely the first digit of the four-digit SIC numbers. The data is available on Standard and Poor’s Compustat.
4.3.2 Country-specific
In line with previous research there will be controlled for the impact of legal system and thereby investor protection using variables that indicate countries’ legal origin and financial system. For the legal origin dummies indicating French civil law, English common law, German civil law or Scandinavian law are included. The dummy equals 1 for countries with a French civil law, 2 for countries with an English common law and 3 for countries with a German civil law. Data was found in La Porta (1988). Furthermore, a dummy indicating countries’ financial system, bank-centered or market-centered, is included. The dummy equals 0 for countries with a market-based economy and 1 for countries with a bank-centered economy. For this variable data came from Kwok and Tadesse (2006) and was supplemented with data from Demirgüc-Kunt and Levine (1999) provided by the World Bank. Besides these variables, Kayo and Kimura (2011) state that stock- and bond market development and GDP-growth also affect capital structure and should therefore be included as country-specific control variables as well. Country stock market development is measured as the ratio of stock market capitalization to GDP divided by 100, and was found in the Global Financial Development Data provided by the World Bank. Country bond market development is measured as the ratio of private and public bond market capitalization to GDP divided by 100, as suggested by De Jong et al (2008). The necessary data was found in the Financial Development and Structure Dataset of the World Bank. Country GDP growth divided by 100 was included to control for the varying
27
economic conditions of the countries and was obtained from the World Development Indicators of the World Bank.
4.4 Summary statistics
A summary of the statistics can be found in Table A-IV and Table A-V in Appendix A. Table A-IIII Panel A lists the statistics of the financing leverage ratios found by country. Panel B in this same table reports on the means of the independent variables of interest national culture by country. The first column shows the statistics of the index based on Hofstede’s measure of national culture and the second column shows the statistics of the index based on the WGIs. Table A-V in Appendix A shows an overview of the statistics of all the other firm- and country-specific control variables. The total amount of leverage ratio observations equals 46183. The means of the leverage ratios differs by country and has a total average of 43.09. The total average mean corresponding to the index comprised of Hofstede’s national culture measure is .81 and the total average corresponding to the means of the index variable comprised of the WGIs is .32. Remember that the higher the index variable values for Hofstede’s national culture measure and the WGIs, the more a country is perceived to behave culturally differently compared to the U.S. and thereby is expected to affect firms’ capital structure. A quick comparison between the countries reveals several unsurprising differences. Consider the countries Russia and South Korea with the highest Hofstede index scores of 4.73 and 4.76 respectively, compared to for example Australia with the very low Hofstede index score of .03. Also the scores for Great Britain appear to be quite low, suggesting that their economy is not much culturally different compared to that of the U.S.. The index comprised of Hofstede’s measure of national culture shows a value of .27, whereas the index comprised of the WGIs shows a value of .04. Moreover, at first glance it appears both measures of national culture show similar outliers. This adds to the validity and representativeness of the culture measures. The benchmark country the U.S. shows a leverage ratio of 44.02 (not shown in the table) based on 369 U.S. firm observations, only .93 points higher compared to the average of 43.0. This slightly higher leverage-financing ratio for the U.S. only weakly supports the behavioural pecking order theory behind the dimension individualism versus collectivism as was stated in section 3.3.
28
Multi-collinearity might have been an issue. A pairwise correlation test has therefore been carried out to test whether this was present. Results are shown in Table A-VI in Appendix A and indicate, except for two variables, no reason for concern. Variables showing a pairwise correlation of .700 or higher are assumed to be correlating. The country-specific variable public bond market development shows a correlation of .7201 with the Hofstede national culture variable and the country-specific dummy variable indicating the financial systems of the countries correlates with the Hofstede national culture variable (.7589) and the variable stock market development (-.7334). Noteworthy are the two different high correlations observed concerning the dummy variable indicating the financial system of the countries, especially the correlation with the Hofstede national culture variable. This positive correlation suggests a link between countries’ cultural distance from the U.S. and their financial system, where a high distance relates to a bank-based system. The positivity of the coefficient is not surprising, since the U.S. is considered to have a market-based financial system. A possible explanation for this correlation may be found in the single dimension uncertainty avoidance index. A bank-based system may be perceived more ‘safe’ and a market-based system more ‘speculative’ (Kwok and Tadesse, 2006; Chang et al., 2012). Recall that Kwok and Tadesse (2006) found that countries with higher uncertainty avoidance index are more likely to have a bank-based system. Thus, in this study the variable uncertainty avoidance index, referring to people’s tolerance to risk, may be positively related to the dummy indicating the financial systems of the countries. This is supported by the high positive correlation of .7812 between the dimension uncertainty avoidance index and the dummy indicating the countries’ financial system (correlations between individual dimensions and the control variables are not shown in the table). However, some other cultural dimensions also show a correlation with this dummy variable and thus the proposed relationship cannot be statistically confirmed. Even so, the observed high pairwise correlations in this study do not immediately indicate reason for concern. The variables do not show high correlations with the dependent variable and as can be seen later in the study the regression models have also been performed without the country public bond market development variable and the dummy variable indicating the financial system of the countries. F-tests are carried out to see whether there exists statistical evidence the variables do not belong to the model as well. Moreover, in regression model
29
two the variable uncertainty avoidance index is excluded from the regression model due to possible correlation with other dimensions. So, in the regression model where the individual Hofstede (2010) dimensions are used as inputs the proposed pairwise correlations do not lead to regression validity problems. Furthermore, the variable firm size shows a correlation of .5329 with the book leverage ratio and a correlation of .5244 with the variable firm profitability. Even though these correlations are not strikingly high, based on a lower correlation threshold of .500 the variable firm size is correlated with a regression determinant and the dependent variable book leverage ratio and may therefore induce bias in the coefficient regression results. Therefore, the regression models will also be performed without this variable firm size.
Further, notice the positive relationship between the independent variable of interest comprised of Hofstede’s proxy for national culture and the book leverage ratio. The value of .4142 appears to be statistically significant at the highest 1% level. Remarkable is that this correlation is, apart from the correlation with the variable firm size, highest in absolute value. This suggests that national culture based on Hoftede’s measure is an important and unique determinant regarding the firms’ capital structure. The independent variable of interest comprised of the WGIs shows a very weak negative relationship of -.0028 towards the book leverage ratio, but is not statistically significant. Thus, from a first quick view based on the pairwise correlation matrix and the Hofstede national culture measure the more culturally different the countries are from the U.S., the higher the leverage ratio.
Table A-VII in Appendix A provides the pairwise correlation matrix between Hofstede’s (2010) six national culture dimensions. The correlations are all significant at the highest significance level of 1%. Based on the .700 pairwise correlation threshold four dimensions do not correlate with each other. These four variables include the power distance index, individualism versus collectivism, masculinity versus femininity and pragmatic versus normative. Regression model two uses these four individual national culture dimensions as independent variables of interest, instead of using the averaged index measure of national culture. Note that this study does not exclude the possibility that the dimension variables are endogenous. As stated earlier, national culture is a subjective topic and
30
therefore quantifying it into separate non-correlating dimensions may be hard. The fact that based on the pairwise correlation matrix two out of the six dimensions are eliminated due to high correlation with other dimensions other supports this possibility. Moreover, the regressions have also been carried out without the two variables masculinity versus femininity and pragmatic versus normative, leaving the model with solely two out of the six Hofstede (2010) national culture dimensions. The variable masculinity versus femininity was left out because extant literature about the variable and its relationship with the capital structure is based on Schwartz’ (1994; 1999) measure of national culture instead of Hofstede’s (2010) measure. The variable pragmatic versus normative was left out because there does not exist direct empirical evidence regarding the relationship between the variable and the capital structure. Section 3.1 hypotheses elaborates on the predicted relationships of these two variables with the capital structure. For the individual dimension variables the raw country scores have been used, so there has not been made use of the adjusted version of the Kogut and Singh (1988) index. This provides the opportunity to check for a relationship between national culture and capital structure without the use of an average referenced index score and thereby the possibility that some very important dimensions are limited in their effect.
5. Results
5.1 Main results
5.1.1 Regression model one; Hofstede index
The first model regresses the book leverage ratio against the outcomes of the adjusted Kogut and Singh (1988) index comprised of Hofstede’s measure of national culture and all the control variables. The regression results of this model are presented in Table B-I in Appendix B. Column (1) reports the univariate model, which has an overall R-squared that equals .172. This relatively low fit of the regression model indicates no reason for concern, because it is in line with previous research on capital structure, such as Baker and Wurgler (2002). Moreover, Lemmon et al. (2008) note that the adjusted R-squareds from traditional leverage regressions ranges between 18% and 29%, whereas it increases when performing the regression with time-fixed effects. The results show a very modest
