Creditor protection and debt levels of foreign affiliates of U.S. multinational corporations

(1)

Creditor protection and debt levels of foreign affiliates of U.S.

multinational corporations

University of Amsterdam Amsterdam Business School MSc Finance, Corporate Finance track

Master thesis

Supervised by dr. J.E. Ligterink Lotte Coppelmans Student no. 11407905

(2)

2 | P a g e

Abstract

This research analyses the relation between creditor protection and corporate financing, by assessing leverage ratios and internal capital markets of foreign affiliates of U.S. multinational corporations. Creditor rights evolve slowly over time. However, developments of individual elements of creditor law are absent, and legal rules do not converge. Creditor protection influences foreign affiliates’ capital structures and their composition of debt. A one-unit increase in debtor control laws is associated with approximately 13 percent higher leverage ratios, with borrowing of affiliates in civil law countries being particularly sensitive to creditor law. Better protection of creditor interests increases the availability of debt, this research shows that affiliates of multinational companies are willing to adopt the elevated credit supply. Additionally, foreign affiliates reduce financing from parent sources in countries with strong creditor rights.

Statement of originality

This document is written by Lotte Coppelmans who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3 | P a g e

Table of main figures

Figure 1: Graphical representation of average aggregate creditor protection over time across all

countries ... 30

Figure 2: Graphical representation of average sub-index scores over time across all countries ... 30

Figure 3: Graphical representation of aggregate creditor protection over time per legal origin ... 31

Figure 4: The relationship between aggregate creditor protection and affiliate leverage, 2005 ... 35

Figure 5: The relationship between aggregate creditor protection and external borrowing, 2005 ... 45

Figure 6: The relationship between aggregate creditor protection and parent borrowing, 2005 ... 45

Table of main tables

Table 1: Descriptive statistics of affiliates in all countries for the years 1995 – 2005 ... 27

Table 2: Descriptive statistics of affiliates in all countries for the years 1995 – 2005 ... 29

Table 3: Time trend in creditor protection index and components of creditor protection ... 33

Table 4: The impact of creditor protection scores on multinational affiliate leverage ... 37

Table 5: The dynamic impact of creditor protection on leverage of affiliates in common and civil law countries ... 41

Table 6: The impact of creditor protection on the composition of affiliate leverage ... 47

(5)

5 | P a g e

1. Introduction

Legal creditor protection varies substantially between countries and appears to determine the strength of local financial markets (La Porta, et al., 2008; Djankov, et al., 2007; Deakin, et al., 2016). Emerging countries often have illiquid and extremely shallow credit markets as a result of limited legal creditor rights, whereas other countries provide legal protections which foster credit supply (Desai, et al., 2007). Market imperfections are an important aspect of a firm’s desired debt structure (Titman, 2002), and the common perspective is that creditor rights influence corporate financing choices across countries. U.S. multinational companies and their foreign affiliates have access to multiple sources of leverage (Doukas & Pantzalis, 2003). Foreign affiliates with globally operating parents have an advantage over domestic firms, since they can finance their investments with either external or internal funds, whichever yields most cost efficiencies (Froot & Hines Jr., 1995). Intercompany financing is by definition internal within the organization and therefore terms and conditions can be more flexible compared to external financing (Sermeus, 2016). Internal credit markets allow subsidiaries to overcome expensive external capital from shallow credit markets caused by underperforming legal mechanisms (Desai, et al., 2004). Various laws and regulation are created to enforce protection of creditors’ commitments and to stimulate credit supply (La Porta, et al., 2008). This thesis sheds light on legal reforms and their impact on corporate financing decisions. Laws, regulation and the quality of enforcement appear to be important determinants of corporate financing and may be linked to total, as well as external and intercompany, leverage ratios of foreign affiliates. The master thesis intends to answer the following question: “To what extent do improvements in legal creditor rights of affiliates’ host countries change the appetite for intercompany leverage of affiliates of U.S. multinational corporations?”

Besides less strict covenants and lower borrowing costs (Houston, et al., 2010), legal origin matters for companies’ ability to raise capital in various countries and accordingly to their capital structures (Noe, 1998; Desai, et al., 2004). As a result of strong creditor rights, lenders are willing to supply additional funds at more favourable terms (Qian & Strahan, 2007). Companies will increase external leverage to benefit from the elevated supply (Faulkender & Petersen, 2006; Acharya, et al., 2009). On the contrary, stronger enforcement of regulation and stricter creditor rights may result in a loss of management control which increases incumbent managers’ preference for internal funds (Cho, et al., 2014). Governments intend to strengthen financial development (Deakin, et al., 2009), which is amongst other factors achieved by providing sufficient creditor protection. Strict creditor legislation is on the agenda of many large corporations and local governments (Deakin, et al., 2009), it is yet unclear whether interventions and reforms produce proposed effects. This research tries to illustrate the effectiveness of legal protection and the workings of internal credit markets.

Multinational corporations are exposed to distinct legal systems around the world (Desai, et al., 2004), which makes it possible to identify the sensitivity of financing decisions to the set of local

(6)

6 | P a g e

commercial laws. Analysing the leverage choices of affiliates of multinational companies ensures clean estimates of the mechanism by which underperforming creditor protection impacts financing decisions. Due to difficulties in accessing firm-specific data, this research focuses on country-wide patterns in the flow of credit between U.S. corporations and their affiliates abroad. The sample contains financial information of majority owned non-bank foreign affiliates of non-bank U.S. multinational companies collected by the Bureau of Economic Analysis. Data is disaggregated per country for affiliates operating in 21 nations around the globe and covers the period 1995 to 2005.

The research first examines variation in types of legal protection. Through simple averaging, yearly averages for the aggregate creditor protection index per origin of law are calculated for the common law family, French-, German-, and Scandinavian civil law countries. The creditor index consists of various forms by which creditors are protected by law. The category debtor control laws measures how lenders can impose restrictions on debtor actions. The second category is labelled creditor contract laws and implies how the facilitation of secured credit is organized. The third category insolvency procedures captures creditor protection during corporate bankruptcy proceedings. Means of the aggregate creditor protection index, the three components of creditor protection and yearly averages per legal origin are graphed over time. A time series regression is proposed to assess stability of law by statistically testing whether creditor protection laws in 21 major countries, and the U.S., evolved or converged over the years 1995 to 2005. The research continues by examining to what extent affiliate leverage is affected in response to increased costs and lacking credit supply, imposed by weak creditor jurisdiction. Panel data regressions describe how legal transitions change leverage ratios for foreign affiliates of U.S. multinational corporations. The model builds on the framework applied by Desai, Fritz Foley and Hines Jr. (2004). Adoption of the aggregate creditor protection index delivers imprecise results, therefore the categories that bundle correlated legal variables are proposed as key independent variables. Additional longitudinal regressions explore the sensitivity of external leverage and leverage from parent sources to changes in creditor protection. Both panel data models allow to control for interacted year and country fixed effects, therefore controlling for any variation in jurisdictions over time and across nations. The research concludes by estimating whether multinational companies use internal credit to substitute for costly external funds. Since the decision to raise external debt and borrow from parent sources is simultaneously determined, an ordinary least squares regression produces biased estimates due to endogeneity issues. Instrumental variables models are constructed in order to mitigate reverse causality problems.

The main findings of this master thesis are as follows. First, the research provides no concrete evidence in favour of the proposition that creditor protection in civil law countries underperforms common law countries. Especially German civil law countries facilitate strong legal protection to their creditors compared to other legal families. Regressions provide evidence at the 10 percent significance level that creditor protection evolves over time across all countries, which is against the stability of law hypothesis of Djankov et al. (2007). However, a yearly estimate of 0.05 implies an increase of

(7)

7 | P a g e

aggregate creditor protection at slow pace. Development in the regulatory environment is not driven by transition of one legal family, except for reforms in insolvency procedures which appear to be most pronounced in French civil law systems. This delivers the implication that civil law legislation does not converge to common law standards. Second, there is evidence that foreign affiliates of U.S. multinational companies amend their overall leverage ratios in response to changes in creditor protection. A one-unit change in debtor control laws is associated with approximately 13 percent greater affiliate leverage as fraction of total assets. Findings support the view that corporations adopt additional credit provided by lenders. The effect of credit contract laws on affiliate leverage is weak, and often statistically insignificant. Insolvency procedures are no determinant of subsidiary leverage ratios. The model further examines whether legal protection has a distinct impact on affiliate leverage depending on legal origin of the affiliates’ host country. Credit financing of subsidiaries in civil law countries is in particular sensitive to local creditor regulation, yet to a lesser extent than the aggregate sample. It takes a year for leverage ratios of subsidiaries in civil law countries to adjust subsequent a change in debtor control laws. Third, the composition of affiliate leverage is influenced by creditor protection strength in the affiliates’ host country. Affiliates in countries with poor debtor control laws and less creditor rights in bankruptcy proceedings borrow more internally. The effect of local creditor rights on external sources of leverage is less clear. A short-term increase in external borrowing of approximately 7 percent is offset by a negative delayed estimate of similar magnitude. Greater political risk, local tax and inflation rate regimes also appear to influence affiliates’ leverage. Host country characteristics are particularly important for affiliates in common law countries. Instrumental variable analyses allow to identify the extent to which affiliates substitute internal leverage from their U.S. parents for leverage from external sources. The research finds no significant estimates for any of the instruments used. This implies that subsidiaries do not substitute for costly or absent external borrowing if local legal systems underperform. Multinational companies therefore ignore to utilize advantages of internal credit markets which are not available to domestic companies.

The research connects individual strands of a broader set of academic literature and empirical articles. Early works of La Porta et al. (1997 & 1998) argue that creditor protection differs across nations based on the jurisdiction they adopted. By constructing binary variables, they capture the quality of law per classification of legal origin. La Porta et al. refer to distinctions between English common and French or German civil law. Armour et al. (2009) extend this methodology by incorporating more detailed and accurate measurement of the legal mechanisms. Although previous studies vary in the perspective which legal family provides better creditor protection, the comprehensive view that strong legal support has led to expanded credit market development is shared among practitioners1. Few studies consider the link between creditor rights and corporate leverage. Fan et al. (2011) focus on the effect of institutional factors on capital structure, yet neglect to cover

(8)

8 | P a g e

creditor protection. Bea and Goyal (2009) provide evidence that better creditor rights are related to longer loan maturities, greater loan sizes and reduces loan spreads. Desai et al. (2004) test whether local creditor rights and capital market conditions affect capital structures of foreign affiliates and the composition of affiliate debt. They find that subsidiaries alter their leverage in response to better creditor protection and less shallow credit markets in the host country. However, their dataset dates back to the 1980s and employs the La Porta et al. creditor protection index that excludes significant variables of interest.

This research will add to existing literature a review of credit market conditions based on the contemporary and more extensive CBR Extended Creditor Protection Index by taking a corporate financing perspective. It interacts the development of credit market conditions to country-level debt structures of U.S. multinational companies’ affiliates. The set of literature that clarifies how legal protection is related to company leverage is thin. This study tries to asses to what extent foreign affiliates have appetite to adopt elevated credit supply. Previous attempts to study foreign affiliates’ capital structures faced substantial challenges (Desai, et al., 2004). Detailed research on the performance of internal credit markets is scarce, and cross-country studies raise issues associated with inaccurate measurement of financial information across countries. All information this sample contains is filed through U.S. parent companies so that reporting adopts widely accepted U.S. accounting guidelines. As a result, it is not required to create difficult assumptions generally necessitated when financial information is collected from various nations. Another advantage of the information collected by the Bureau of Economic Analysis is the distinction between current liabilities and long-term debt from U.S. parent companies and external parties, so that the impact of local creditor rights on the leverage composition of foreign affiliates can be examined. The research further contributes consideration of a broad set of countries, including countries in Latin American and Asia. Coverage of more countries generates results for various levels of legal systems and contrasting degrees of economic development.

The next section reviews existing literature and recent studies on legal origin, types of creditor protection, as well as the effect of creditor rights on affiliate leverage and the composition of debt. Section 3 discusses the types of data and regression methods used. Section 4 describes leximetric legal and affiliate information, and summarizes the variables employed in subsequent analyses. Section 5 examines transition of creditor protection across legal origins, the determinants of affiliate leverage and the workings of internal credit markets including the substitution of costly external financing by internal funds. Section 6 presents a set of robustness checks and section 7 concludes.

(9)

9 | P a g e

2. Literature review

Credit market conditions vary per country based on differences in jurisdictions. La Porta, Lopez-de-Silanes, Shleifer and Vishny (1997 & 1998) were amongst the first to describe the diversity in creditor protection based on origin of law. The degree of protection likely influences creditors’ appetite to lend funds. Consequently, creditor protection may affect the amount of leverage domestic and multinational companies can take on the external market. Moreover, affiliates’ preferences for parent company debt are influenced by the strength of the local credit market. As a result, the leverage composition of affiliates illuminates the impact of credit market conditions on capital structure and illustrates the importance of the composition of debt to overcome costly external financing. This section first considers legal origin theory and outlines the motive of differing jurisdictions. Next, it describes ways by which creditors can receive legal protection. It will elaborate on the literature that links creditor rights to corporations’ capital structures and the composition of debt. The section concludes by summarizing empirical evidence from related studies.

2.1 Theory of legal origin

Although no two countries’ laws are precisely identical, commercial laws in different countries are derived from few legal families and divided among two traditions. The main legal families are Romanian civil law and Anglo-American common law (La Porta, et al., 1997; La Porta, et al., 1998; Djankov, et al., 2003). By conquest of territories both civil law and common law have expanded to other countries around the world (La Porta, et al., 1998).

Common law has English origin and the tradition is adopted mainly by British colonies (La Porta, et al., 1998). In this sample, the United Kingdom, the United States, Canada, India, Malaysia and South Africa follow English common law (Djankov, et al., 2007). Scholars classify English law as adaptive due to the construction of new and improved rules (Botero, et al., 2004). Judges develop and interpret the law and employ jurisdiction on a case by case basis which enhances flexibility of the legal system to economic development (Deakin, et al., 2016). As a result, commercial law in common law countries is more adaptable to evolving circumstances. Common law judges are effective in the protection of individual property rights from appropriation of the state due to their independence from legislature (Rajan & Zingales, 2003; Beck, et al., 2003).

Civil law is derived from Roman law and is the oldest and most universally dispersed. This legal tradition is subdivided into three major families: French civil law, German civil law and Scandinavian civil law (La Porta, et al., 1998). Besides France itself, countries that adopt French civil law are those conquered by Napoleon (Djankov, et al., 2007). In this sample nine countries follow French civil law: France, Italy, the Netherlands, Spain and Turkey have jurisdictions based on French origins. Along the Spanish sovereignty in Latin America, French civil law is also spread across Brazil, Argentina, Mexico, Chile (La Porta, et al., 1998). German commercial law is adopted by Japan and Switzerland

(10)

10 | P a g e

(La Porta, et al., 1998). The Czech Republic has returned to its pre-war system after the collapse of communism and is therefore classified to the German civil tradition as well (Djankov, et al., 2007). Scandinavian civil law is the least spread around the world and followed solely by Nordic countries. Nevertheless, scholars describe Scandinavian jurisdiction as distinct from other legal families (La Porta, et al., 1998; La Porta, et al., 2008). In this sample only Sweden belongs to the Scandinavian legal origin. Judges that practice civil law are restricted by clear-cut statutes and codes which comes at the cost of adaptability and flexibility (Botero, et al., 2004). Civil law is thus characterized by excessive rigidity and regulation does not change regularly (Deakin, et al., 2016). Legislature and executive management seize control over legal institutions in civil law jurisdictions (Rajan & Zingales, 2003), which means that civil law judges are less successful in protecting individual property rights (Beck, et al., 2003). Communist or transition countries belong to none of the traditional legal families (Djankov, et al., 2007). Transition legal systems in this sample are classified as a separate category labelled “others” and which incorporates the countries Russia and China.

The extent to which legal families exert influence over countries’ jurisdictions has been questioned (Deakin, et al., 2016) because laws have evolved since their origination. Nevertheless, individual laws across nations can be compared as laws continue to reflect the influence of the original traditions (Botero, et al., 2004; Djankov, et al., 2007; La Porta, et al., 2008). Yet a large body of literature exists with respect to the variation in law. Since governments intend to strengthen financial development (Deakin, et al., 2009), which is amongst other factors achieved by providing sufficient creditor protection, it is expected that countries with civil law origin have reformed creditor rights to converge to common law standards. Yet common law countries, that achieved a high level of creditor protection in the past (La Porta, et al., 1997; La Porta, et al., 1998; Djankov, et al., 2007), are expected to show less severe strengthening of creditor laws (Deakin, et al., 2016). Furthermore, numerous examples of cross-fertilization between legal traditions exist, such as a reform of the German insolvency procedures being influenced by the U.S. Bankruptcy Code (Eidenmüller, 2006).

2.2 Types of creditor protection

Based on the variation of legal origin across countries, La Porta et al. (1997 & 1998) were amongst the first to develop a dataset to measure diversity in creditor protection. Measurement of creditor rights is complex due to different classes of creditors that have specific interests and a distinct collection of rights. All lenders have the incentive to undertake actions so that their own interests are served even if this deploys interests of creditors as a group (Baird & Jackson, 1984). A senior debtholder, for example, is favoured by immediate liquidation during corporate default since his rights may decline if the company continues. On the contrary, a junior creditor gains from waiting to cease the company in order to avoid distribution of assets today, because they receive payment after senior creditors. Bankruptcy codes are designed to instruct lenders to function collectively instead of acting individually not in the interest of creditors as a group (Baird & Jackson, 1984).

(11)

11 | P a g e

Debtor control laws protect creditors when a company is going concern by limiting default risk. It takes into account actions available to creditors in order to limit managements’ and shareholders’ transactions and operations (Armour, et al., 2009). Certain activities by management and owners may dispossess creditors of assets of the company (Deakin, et al., 2016). Provisions that limit debtor actions are related to the minimum capital required for establishment of a private company, dividend restrictions and directors’ duties to creditors. Payment of excessive dividends to shareholders may deprive creditors from access to company assets. Dividend restrictions include restrictions on non-cash forms of dividends, such as share repurchases and undervalued transactions between companies and their shareholders (Armour, et al., 2009). Limiting debtor actions in the period just before investors may realize the company is insolvent protects the position of creditors already before the company files for bankruptcy. Especially unsecured creditors are protected by possessing a duty on directors’ actions to act in creditors’ interest just before bankruptcy (Deakin, et al., 2016). However, too much protection is not desirable either since this may possess a limit on management’s ability to take risks (Armour, et al., 2009).

Credit contract laws facilitate the protection of secured credit to preserve creditors’ interests. Credit contract laws capture the degree to which non-possessory security interests may take over borrowers’ assets and the extent to which these interests must be registered (Armour, et al., 2009). Additionally, credit contract enforcement assesses whether a creditor must go to court to enforce secured interest when the debtor is in default (Armour, et al., 2009). Creditors’ interests are better secured if the facilitation of secured credit is well registered and easily enforced.

Insolvency procedures regulate corporate reorganizations and liquidations. Creditor rights in corporate bankruptcy proceedings are important to advance the position of creditors when a debtor files for bankruptcy. Bankruptcy may be used as a threat against investors when a debtor is able to commence for bankruptcy unilaterally, in absence of a requirement that the debtor is insolvent (Armour, et al., 2009). Creditors are better protected when a debtor is obliged to commence bankruptcy, based on the invoke of a single creditor that the debtor is insolvent by some benchmark, since this compels payment. Creditor protection is optimal when a debtor commences bankruptcy if the company is balance sheet insolvent (Armour, et al., 2009). If the debtor has the possibility to rehabilitation, stay of secured creditors is desirable to impede creditors’ ability to liquidate their collateral (Armour, et al., 2009). Furthermore, the position of creditors is better protected if they are able to vote on the outcome of bankruptcy proceedings, considering that more decision making control at the hands of courts or borrowers weakens creditors’ positions. From an efficiency point of view and the standpoint of maximizing realization values, allocating control rights to the residual claimant is preferred over mechanisms that allocate voting rights to all creditor classes or a particular creditor class (Armour, et al., 2009). Additionally, subordination of secured claimants is an important driver of protecting creditors’ interest in corporate bankruptcy (Deakin, et al., 2016). The expected value of secured creditors’ rights is reduced if secured claimants are subordinated to certain types of preferred

(12)

12 | P a g e

claims. Lenders may be allowed to substitute one security interest for another. The greater the range of subordinated claims, the more pervasive is the reduction in expected value. The rights of creditors are best protected if no type of security interest is subordinated to any preferred claim (Armour, et al., 2009).

Different national pathways in the degree of creditor protection are attributed to diversity in legal families. Since judges in common law countries are able to establish new laws (Botero, et al., 2004), new types of collateral may be created. This enhances the development of improved remedies in favour of creditor interests (Deakin, et al., 2016). In countries that follow English legal traditions, the interests of creditors are in general placed ahead of those of management2_{. In the French legal system,}

companies have a public interest and commercial judges serve multiple interests instead of protecting private property rights (Deakin, et al., 2016). French civil law is therefore characterized by minimizing creditors’ interests relative to a wider company perspective where French civil law standards appear to perform worse in terms of creditor protection compared to other legal origins (Djankov, et al., 2007; Deakin, et al., 2016). The German legal family, although mainly influenced by French civil law, recognizes creditors’ individual property rights and henceforth approaches common law standards (Deakin, et al., 2016). The development of German legal rules towards common law shows that jurisdictions change over time (Djankov, et al., 2007) and that some attributes of law evolve into examples for other jurisdictions (Deakin, et al., 2016). Nonetheless, analogous to a wider set of laws and regulations, creditor protection rules are rigid and are not expected to transform much over time. The inefficiency of legal rules, however, remarks the necessity of reforms which have occurred in dozens of countries in the last decade (La Porta, et al., 2008). Creditor rights have been strengthened by the belief that this would result in more liquid credit markets, and to foster development of bank and private credit supply (Deakin, et al., 2009). The first hypothesis of this research states that civil law countries provide worse protection to creditors compared to common law standards and that legal rules in poor legal families are converging over time. Especially jurisdictions that follow French traditions underperform in terms of guarding creditors’ interests.

2.3 Effect of creditor protection on company leverage and the composition of debt

Transfer of risk arises when shareholders diminish bondholders’ wealth by increasing the riskiness of projects. Equity claimants secure most of the projects’ payoff, whereas bondholders bear the costs of the high risk project (Jensen & Meckling, 1976). Risk-shifting is a significant problem in corporate finance (Desai, et al., 2004). Debt holders typically request strict covenants and urge strong monitoring to protect their wealth from risk-shifting. Especially when a firm’s executives act in their own interest or solely the interest of shareholders, rights attached to debt become crucial (La Porta, et al., 1998). These rights are determined by the legal rules and quality of enforcement of the country in

(13)

13 | P a g e

which debt is issued. A secured creditor in Germany may be treated differently from secured creditors in India in the event of default, based on the variation in jurisdiction. It may be that their collateral is not worth the same in both cases, other things equal (La Porta, et al., 1998).

Differences in legal origin have an effect on the size and depth of credit markets (La Porta, et al., 1997), such that countries with weak creditor protection experience substantially smaller debt markets. Civil law countries that forfeit strong creditor rights suffer from less investor activity and hence have smaller credit markets (La Porta, et al., 1997; La Porta, et al., 1998; La Porta, et al., 2008). Laws, regulation and the quality of enforcement appear to be important determinants of corporate financing. The set of creditor laws and regulation provide a legal solution to agency problems. Protection of creditors’ interests determines their appetite to finance firms (La Porta, et al., 1998). Better legal protection enables creditors to offer companies external financing at better terms and conditions (La Porta, et al., 1997). Besides less strict covenants and lower borrowing costs, (Houston, et al., 2010) legal origin matters for companies’ ability to raise capital in various countries and accordingly to their capital structures (Noe, 1998; Desai, et al., 2004).

More creditor rights increase the power creditors have over companies, which substantially reduces the risks they are exposed to. If more funds are supplied because the attractiveness to lend increases, companies will increase overall leverage to benefit from the elevated supply (Faulkender & Petersen, 2006; Acharya, et al., 2009). On the contrary, stronger enforcement of regulation and stricter creditor rights transfer decision making capacity from management to creditors. From management’s perspective, this lowers the attractiveness of credit (Houston, et al., 2010). Consequently, if management aims at avoiding the loss of control, managers will have a stronger preference for sources of capital that do not shift decision making capacity. The second hypothesis states that the strength of creditor rights affects affiliate leverage. The sign and magnitude of the relation is still ambiguous; creditor protection may have a positive or negative correlation with total leverage. The relation depends on the willingness of executive management and shareholders to discard control as well as the extent to which creditors supply additional financing at low interest rates. Stronger creditor protection increases lenders’ willingness to provide credit, thereby increasing the supply of funds (La Porta, et al., 2008). Furthermore, raised legal protection results in lower interest rates, requirement of less collateral and longer loan maturities (Bae & Goyal, 2009). Companies may adopt the elevated leverage levels at supportive terms. In contrast, legal protection can have a negative effect on leverage. Some nations’ bankruptcy codes are not protective for debtors, for example in countries where management positions can be replaced by creditors during periods of distress. As a result, self-interested management and shareholders aim at avoidance of losing control and may deter the use of leverage (Cho, et al., 2014).

Local jurisdictions determine affiliates’ preference for the composition of leverage. In the event of bankruptcy, affiliates of multinational companies generally must follow local bankruptcy laws of the host country (Desai, et al., 2004). Therefore, multinational firms can take the advantage of

(14)

14 | P a g e

borrowing across a variety of legal institutions. In countries where creditors face less rights in the event of default, interest rates should be higher to compensate the creditor for the risks of being less protected (Desai, et al., 2004). Instead of being the least protected during bankruptcy, creditors can impose actions to prevent financial distress if such laws are enforced. For example, renegotiation with creditors is beneficial for a financially distressed firm, but less advantageous for creditors since it lowers the firm’s incentive to avoid bankruptcy (Desai, et al., 2004). Less protection not only drives up borrowing costs, it also diminishes incentives to lend credit as well and hence reduces the availability of credit (La Porta, et al., 1997; Beck, et al., 2003; Deakin, et al., 2016).

The presence of agency problems increases the benefits of internal capital markets, since internal capital markets are less prone to agency problems (Stein, 1997). If weak creditor protection negatively influences the cost of external financing, affiliates of multinational companies have an advantage over domestic firms. So too, absence of external credit may reduce domestic companies’ ability to fund projects, whereas the continuation of affiliates’ projects is less threatened. Affiliates can supply capital from parent companies to overcome expensive external funds and other shortcomings associated with underperforming host country credit markets (Desai, et al., 2004). The third hypothesis states that the degree of creditor protection in an affiliate’s host country determines the composition of leverage of the affiliate. Creditor rights negatively affect affiliates’ preference for leverage from their U.S. parent sources. If creditors’ interests are not well protected, external leverage is scarce and costly. Affiliates of multinational parents can rely on internal sources of debt in response to the shortcoming of local credit markets.

Besides legal protection of creditors, alternative institutional factors are determinants of affiliates’ leverage structures and the decision on parent or external leverage as well. As a result of tax deductibility of interest expenses, tax policies in the affiliates’ host country provide incentives to use debt rather than equity. The greater corporate taxes, the larger the incentive to finance investments with leverage (Desai, et al., 2004). Multinational firms can benefit from interest deductibility in high-tax countries (Hines Jr. & Rice, 1994), by raising external leverage and reallocating credit across business segments. Other factors that shape the ability of affiliates to raise funds are variation in financial development (Rajan & Zingales, 2003) and inflation policies of affiliates’ host countries. During periods of high inflation, the real cost of debt diminishes (Modigliani, 1982) and consequently higher rates of inflation are associated with highly leveraged foreign affiliates (Desai, et al., 2008). In order to hedge political risks such as expropriation of assets, multinational corporations can increase leverage of their foreign affiliates (Kesternich & Schnitzer, 2009). Currency depreciations may constrain domestic and multinational companies in their investment opportunities. Internal credit markets provide additional funds to foreign affiliates in order to overcome negative consequences of currency depreciations. Countries with economic instability often impose short-term capital controls to foster growth and exposure to worldwide capital markets (Desai, et al., 2008). However, capital

(15)

15 | P a g e

controls increase spread on loans and thereby diminish subsidiaries’ incentives to raise external leverage. The effects of capital controls are similar to the incentives of tax rates (Desai, et al., 2008). 2.4 Empirical evidence from existing research

La Porta et al. (1997 & 1998) show that commercial laws vary substantially between legal families. French civil law countries severely underperform in terms of protecting security interests of creditors. The effect of creditor protection on corporate finance has been deliberated by La Porta, Lopes-de-Silanes, Shleifer and Vishny (1997 & 1998) in their later work. The authors show that legal origins have an effect on external finance, and that companies with presence in countries that offer better legal protection should rely more on debt finance. The original study on legal traditions and their effect on private credit has been stretched by many academics, among which Djankov, McLiesh and Schleifer (2007). Their period of study is characterized by few changes in insolvency laws since their index varies relatively little over time. Nevertheless, Djankov et al. (2007) note the period was actually one of many changes in bankruptcy regulation. This result suggests that the index constructed by La Porta et al., and used by academics such as Djankov and colleagues, excludes significant variables of interest. Evidence by Djankov, McLiesh and Schleifer (2007) further suggests that better creditor rights have a positive relation with aggregate private credit to gross domestic product, and that the effect is most pronounced in richer countries. Deakin, Mollica and Sarkar (2016) show that different aspects of creditor protection laws have opposite effects on the expansion of national credit for four major OECD countries. They examine national pathways in levels of creditor protection and law reforms for four OECD countries, namely France, Germany, the United Kingdom and the United States, and show that civil law countries have developed strengthening debtor control laws over a period of thirty years. The study conducted by Deakin et al. (2016) focuses solely on wealthy nations with well-developed juridical systems and deep credit markets. This thesis incorporates a wider set of countries, including countries in Latin American and Asia. The coverage of more countries generates results for various levels of legal systems and contrasting degrees of economic development.

The effect of creditor rights on total company leverage is still ambiguous. Although the U.S. credit market is very liquid, many foreign credit markets are less deep. Titman (2002) shows that small credit markets are less liquid and lack efficiency, which is associated with lower supply of funds. As a result, borrowers may be constrained in their ability to increase leverage (Faulkender & Petersen, 2006). Yet Ghoul et al. (2014) provide evidence that managers prevent giving up control and refrain from long-term debt if creditor rights are strong.

The influence creditor protection laws have on internal credit allocation of multinational corporations and their foreign subsidiaries is less deliberated in corporate finance literature. Faulkender & Petersen (2006) provide evidence that agency costs are passed on to the borrower and reflected by increased interest rates. Consequently, it is less attractive for companies to raise external leverage in underperforming credit markets. There is evidence that affiliates are funded with more

(16)

16 | P a g e

internal debt in countries with weak creditor rights. Desai et al. (2004) suggest that affiliates of U.S. multinational companies use parent debt as substitute for expensive external leverage. Their findings further indicate that interest rates are significantly higher in nations with weak creditor protection and shallow credit markets. Hubbard and Palia (1999) show that conglomerates opportunistically employ internal credit markets to overcome costly external financing.

Cheng and Shiu (2007) find important effects of country specific factors other than legal mechanisms predicting leverage. It is difficult to empirically estimate the sensitivity of corporate indebtedness to local taxes due to a lack of variation in corporate tax rates. Some empirical studies find no evidence for the positive association between corporate taxes and company leverage. Desai, Fritz Foley and Hines Jr. (2004) employ local tax rates for various countries and find incentives for multinationals to alter leverage if local taxes rise, which are greatest for parent borrowing. Desai and colleagues thoroughly study capital structures of multinationals and their foreign subsidiaries. In a later study they consider leverage levels of domestic firms and foreign affiliates in areas with sharp currency depreciations, finding benefits of internal credit to compensate financing constraints (Desai, et al., 2008b). Another study finds that multinational corporations with affiliates in countries imposing capital controls employ less leverage as a result of higher borrowing rates (Desai, et al., 2004b). Evidence of shallow financial markets discouraging growth is present at firm level (Demirguc-Kunt & Maksimovic, 1998) and country level (King & Levine, 1993). Noveas (1998) studies capital structures of foreign subsidiaries in Brazil and finds greater indebtedness for foreign affiliates compared to Brazilian counterparts, based the level of political risk. Rajan and Zingales (1995) focus on determinants of capital structure in industrialized countries, finding similar ways by which institutional factors appear to have an influence on capital structures.

Existing work primarily adopts the creditor protection index defined by La Porta et al. (1997 & 1998) and refrains from analysing the more detailed database that is constructed by the CBR at the University of Cambridge. So too, existing work is less detailed in explaining to what extent protection of creditor rights affects capital structure choices, and the use of internal debt as substitute for external funds. This research will add a review of credit market conditions based on the CBR Extended Creditor Protection Index to current literature and interacts the development of credit market conditions to country-level debt structures of U.S. multinational companies’ affiliates. This papers aims to extent current corporate finance literature, and deliberates on the relation between creditor protection and affiliate leverage. Furthermore, the paper analyses the composition of affiliate leverage in response to variations in host country creditor laws. The following three hypotheses are tested, which have been briefly introduced in this section. First, civil law countries provide worse protection to creditors compared to common law standards and legal rules in poor legal families are converging over time. Second, the strength of the affiliates’ host country legal system determines the level of affiliate leverage. Last, this research hypothesizes that the degree of creditor protection in an affiliate’s

(17)

17 | P a g e

host country negatively affects the composition of affiliate leverage with reference to leverage from parent sources.

3. Methodology

3.1 Pathways of legal origins

The set of rules that protect creditors’ interests reveal clear pathways to industrial developments and political influences, and shape regulatory responses to evolving business circumstances (Pistor, 2005). The Extended Creditor Protection Index provided by the CBR at the University of Cambridge provides longitudinal data on legal development for 25 countries over the period 1995 to 2005. A visual representation of creditor law evolvement at legal origin level is designed to test whether creditor rights differ between legal traditions. Through simple averaging, yearly averages for aggregate creditor protection per origin of law are calculated for the common law family, French-, German-, and Scandinavian civil law countries and the transition group classified as others. These data are plotted across a number of graphs.

Although the visual representation provides a broad image of legal origin pathways, graphical evidence does not provide sufficient statistical evidence. Time series regressions are built in order to isolate more carefully the pathway of the aggregate creditor protection index, and three sub-indices, over time and to examine whether creditor laws prevail a sharp trend. Armour et al. (2009) divide the creditor protection index into three sub-indices, based on various ways in which creditors’ interests are guaranteed. The following basic time series regression is constructed with time as independent variable:

(1) creditor protection indext= a1*constant + β1*timet

The regression is examined for the total sample including all countries over the period 1995 to 2005, and replicated for two subsamples based on legal origin. The dependent variable is the yearly average aggregate creditor protection index. In order to distinguish trends in ways by which creditor rights are protected by law, additional regressions are run with the yearly averages of the sub-indices debtor control laws, credit contract laws, and insolvency procedures as dependent variables. Based on relevant literature, it is not yet clear whether time has a positive or insignificant impact on creditor laws. A positive coefficient implies strengthening of the aggregate index over time and is in line with evidence of Deakin et al. (2016). In contrast, an insignificant coefficient is associated with stable creditor rights over time, as found by Djankov et al. (2007). Interpretation of estimates for the three sub-indices is similar. To test whether the pathway for French civil law countries varies from evolvement of rules in the common law family, results are revisited for French civil law countries and common law countries separately. French civil and English common law nations are divided into two subsamples. Should legal rights in French civil law traditions converge towards common law

(18)

18 | P a g e

standards, as hypothesized by Deakin et al. (2016), the coefficient for countries in the first group exceeds the time impact for the second group. Regressions for subsamples are constructed with the aggregate creditor protection, as well as with three sub-indices, as dependent variable to test whether convergence of law is responsive to different components of legal protection. All regressions are corrected for heteroscedasticity using robust standard errors.

3.2 Impact of creditor protection on affiliate leverage

Multinational companies and its subsidiaries face differing legal regimes, which allows assessment of the impact of creditor rights on financing choices (Desai, et al., 2004). This research further examines whether affiliates’ total leverage levels respond to developments in domestic credit market conditions using panel data regressions. Tests that build a regression which includes all ten individual components of creditor protection allow to examine the impact of each variable of creditor protection separately. However, inclusion of each creditor law separately may introduce multicollinearity issues since the variables are highly correlated with one another. Performing the regression analysis by adopting the aggregate creditor protection index as independent variable mitigates the multicollinearity problem but reduces the precision of results drastically. Categories that bundle correlated creditor protection variables mitigate the multicollinearity dilemma, while accuracy of the results is ensured.

The CBR dataset pursues to capture the complexity of creditor legislation by creating three categories: debtor control laws, credit contract laws and insolvency procedures (Deakin, et al., 2009). The sub-indices indicate various forms by which creditors are protected by law. The category debtor control laws measures, amongst others, the minimum capital level for a private company, normalized across (0,1) where a minimum capital level of €25,000 yields a score of 1. Moreover, the sub-index includes scores based on the level of dividend restrictions and the strength of directors’ duties to creditors. The second category is labelled creditor contract laws and implies how the facilitation of secured credit is organized. This category includes measures for security interests taken over debtor’s assets, measures to which extent non-possessory security interests are registered and a measure for the power of enforcement of security interests. The third sub-index, insolvency proceedings, reflects the stay of secured creditors in rehabilitation, whether an individual creditor may commence bankruptcy proceedings, the ability of creditors to vote on bankruptcy proceedings and the subordination of secured claimants.

Laws are subject to many interpretations which makes it difficult to capture legal origins in leximetric coding. Academics’ judgements have an impact on the selection of variables, the choice of correct weights of indicators and the aggregation of scores (Armour, et al., 2009). The index of La Porta et al. uses binary coding, a method that fails to take into account the ambiguity and complexity in the interpretation of rules of law. The CBR methodology addresses some of the objections raised by opponents of legal origin theory, as it aims to capture more of the complexity of jurisdictions (Deakin, et al., 2016). The CBR Extended Creditor Protection Index uses intermediate coding, which means

(19)

19 | P a g e

each individual variable can obtain a score between 0 and 1 (Armour, et al., 2009). A low score indicates weak protection on this individual aspect of creditor protection, and a high score stands for maximum creditor protection offered by means of the separate indicator. The widely applied creditor rights index constructed by La Porta et al. (1997 & 1998) consists of four measures and focuses solely on corporate reorganization or liquidation procedures. The index discards other means that determine the association between companies and creditors, such as minimum capital requirement and capital maintenance rules (Deakin, et al., 2016). The CBR dataset conquers a larger range of legal information, since it contains ten variables in separate areas of creditor protection, and is therefore more detailed than the dataset constructed by La Porta et al. (1997 & 1998).

The categories debtor control laws and credit contract laws both contain three individual variables, therefore these sub-indices range between 0 and 3. The sub-index insolvency proceedings incorporates four variables that measure creditor rights, hence it can obtain a total score between 0 and 4. The following regression attempts to examine the relation between affiliate leverage and creditor rights using three sub-indices of creditor protection:

(2) total leverage ratioi,t = a1*constant + β1*debtor control lawsi,t + β2*credit contract lawsi,t + β3*insolvency proceduresi,t + δ1*weak creditor rightsi,t + δ2*tax ratei,t + δ3*tangible assetsi,t + δ4*log of salesi,t + δ5*political

riski,t+ δ6*inflation ratei,t+ δ7*sales growthi,t+ time fixed effects + country fixed effects

The dependent variable is the total leverage ratio aggregated for all affiliates per country. This measure is defined as current liabilities and long-term debt over total assets, per country, for foreign affiliates of U.S. multinationals over the years 1995 to 2005. Financial affiliate data disaggregated per country is obtained from the Bureau of Economic Analysis’ Annual U.S. Direct Investment Abroad Surveys. Affiliates have access to a larger pool of debt if supply of debt increases due to, for example, better covenants available to investors, stricter dividend restrictions or more weight on creditors’ votes in bankruptcy. Given that companies increase their debt levels to benefit from the elevated supply (Faulkender & Petersen, 2006), the coefficients of the sub-indices are positive. If a negative effect on leverage is estimated, as indicated by negative coefficients, stricter creditor rights may result in a reduced demand for credit by management, because they fear a loss of control (Acharya, et al., 2009).

Factors that are expected to determine total leverage, such as the affiliate’s host country tax rate and the rate of inflation, are included as control factors (Desai, et al., 2004). Subsidiaries’ incentives for leverage grow as host county tax rates increase due to the tax deductibility of interest payments (Graham, 1999). The relation between local monetary policy and leverage is less certain. A sharp increase of local GDP diminishes the real cost of debt, thereby increasing the benefit of debt (Modigliani, 1982). Instead of the positive association hypothesized by Modigliani, inflation devalues outstanding debt (Djankov, et al., 2007) so that companies are less likely to alter leverage.

(20)

20 | P a g e

Additionally, the regression controls for various degrees of political risk as well, since a rise in foreign political risk increases affiliates’ appetite for leverage (Desai, et al., 2008).

Other subsidiary characteristics that are expected to affect leverage are affiliate size, outlook of sales growth and asset tangibility as shown by Rajan & Zingales (1995). If an affiliate has a large portion of tangible assets, these may serve as collateral which diminishes the agency cost of debt and increases lenders’ willingness to supply loans. The effect of firm size on leverage is more ambiguous. Larger affiliates tend to be more diversified and have a lower probability of bankruptcy, but simultaneously are more subject to information asymmetries which increases the preference for equity. Desai et al. (2004) find that growth opportunities have a small but significant effect on leverage, which may indicate that the debt-overhang problem is diminished for subsidiaries of multinational companies.

Although control factors for legal origin have no significant influence on private credit beyond main estimates (Djankov, et al., 2007), relatively weak creditor rights may have an effect on leverage in addition to other regressors. Countries with weak legal protection have less deep credit markets (La Porta, et al., 1997), which influences affiliates’ ability to increase leverage. In order to control for this effect, regressions include a dummy variable for weak creditor protection. Countries with weak creditor rights are countries with aggregate creditor protection below the sample’s mean.

A regression that includes lagged variables for the leverage ratio and three sub-indices captures the dynamic aspect of the relation between affiliate leverage and local set of laws. The current leverage ratio is a function of past leverage ratios and of current and past values of creditor protection sub-indices since lenders’ responses to changing creditor rights may be delayed (Deakin, et al., 2016). Accordingly, it takes some time for affiliates to adjust leverage ratios if credit supply does not increase simultaneous to regulatory reforms. The number of appropriate lags is selected by general to specific model selection criteria. With backward elimination, estimates with a low p-value are step by step omitted until all variables are significant. Inclusion of multiple variables that are highly correlated increases the probability of multicollinearity. Therefore, the default significance level is altered to 25 percent in order to decide whether lagged variables are included or removed from the dynamic model (Montgomery, et al., 2015). F-tests are conducted to test whether coefficients are jointly significant. The standard and lagged regression are revisited for common law and civil law countries separately, in order to acquire the strength of creditor rights on leverage ratios for affiliates’ host countries with different legal traditions. Different results across the sub-samples may suggest that the legal mechanism enforcing creditor rights is more important in civil or common law countries, or that results may even be driven by one single legal family.

A Hausman test is conducted to indicate which panel data technique provides the most accurate analysis of the longitudinal data, based on the autocorrelation of errors between countries. Output is presented in Panel A of Table A1 in the appendix. The null of no autocorrelation is rejected, hence regressions are constructed using fixed effects. This method is suitable to examine changes in affiliate

(21)

21 | P a g e

debt composition within a country caused by movements in creditor protection laws. Time fixed effects are included to control for time-invariant differences between countries. The Im-Perasan-Shin unit root test, that allows the dataset to be unbalanced, is conducted to test for non-stationarity of the data. The result is presented in Panel A of Table A2 in the appendix. Leverage is stationary because the null of a unit root in all panels is rejected. The ordinary least squares methodology for panel data is appropriate, since it is unlikely that affiliates’ capital structure decisions influence the host country’s creditor protection policy. Tests in section 5 attempt to validate this statement. The fixed effects model controls for variations between countries that are constant over time, such that the estimates are not biased because of omitted time-invariant variables. Inclusion of time fixed effects further mitigates biased estimates because of time varying factors. So too, the methodology is in line with the research by Desai, Foley and Hines (2004), who adopt a similar set of regressions although for affiliate-year observations.

Several limitations to this approach should be noted. Not all the countries in this dataset have a high level of respect for commercial laws and rules, nonetheless a separate control factor for differences in the extent of legal enforcement is omitted. This research has a corporate finance vision and focusses to a lesser extent on legal features. Furthermore, this research includes a small number of countries so that it is not sufficient to cluster standard errors and statistical power is limited. In order to improve accuracy of estimates, and correct for heteroscedasticity, robust standard errors are used. Every country-year observation is handled as an independent observation in order to limit the reduction of the sample size and enhance statistical power. Additionally, detailed and confidential affiliate information is only available to special sworn BEA employees, the data available to the wider audience does not enclose information on affiliate parent companies. It is therefore not possible to identify the borrowing behaviour of foreign subsidiaries of the same parent companies, and estimates that control for effects that are universal to all affiliates of the same U.S. parent company cannot be obtained. As a result, parent-specific considerations may bias the estimates of this research. Besides purely long-term debt, the current measure of affiliate leverage includes current liabilities as well. Affiliate information provided by BEA does not separate the accounts current liabilities and long-term debt, thereby a clearer estimate excluding unrelated working capital cannot be obtained. Limited variation in creditor protection index possesses additional limitations.

3.3 Composition of affiliate leverage

The research continues by examining to what extent affiliate leverage from external sources declines in response to increased costs, and lacking credit supply, imposed by credit market underdevelopment because of weak jurisdiction. External debt is costly in countries where creditor rights are fragile and that lack credit market depth (Desai, et al., 2004). If creditors’ security interests are weakly protected against default, lending becomes unattractive and the supply of credit declines (Houston, et al., 2010).

(22)

22 | P a g e

As a result, companies may become capital constrained and have to pass on profitable investments. The third hypothesis states that affiliates will rely more on the sources of their U.S. parents to overcome the increased costs and limited supply associated with shallow and underperforming creditor rights. The third hypothesis is tested using either affiliates’ parent borrowing as a fraction of total assets or the ratio of affiliates’ external borrowing to total assets on a country-year basis as dependent variable, and explanatory and control variables that are defined in the regressions below.

(3) external borrowingi,t = a1*constant + β1*debtor control lawsi,t + β2*credit contract lawsi,t + β3*insolvency proceduresi,t + δ1*weak creditor rightsi,t + δ2*tax ratei,t + δ3*tangible assetsi,t + δ4*log of salesi,t + δ5*political

riski,t+ δ6*inflation ratei,t + δ7*sales growthi,t + time fixed effects + country random effects

(4) parent borrowingi,t = a1*constant + β1*debtor control lawsi,t + β2*credit contract lawsi,t + β3*insolvency proceduresi,t + δ1*weak creditor rightsi,t + δ2*tax ratei,t + δ3*tangible assetsi,t + δ4*log of salesi,t + δ5*political

riski,t+ δ6*inflation ratei,t+ δ7*sales growthi,t + time fixed effects + country fixed effects

External and parent borrowing are defined as current liabilities and long-term debt from U.S. parent companies or external sources, respectively, over total assets. The dependent variables are on affiliate level and disaggregated per country, for affiliates of U.S. multinationals over the years 1995 to 2005. The regressions include key independent variables similar to the regression in paragraph 3.2, and employ country-specific and time-specific effects such that considerations other than creditor protection do not influence results. Positive coefficients imply a positive relation between protection of creditor rights and external or parent borrowing, whereas a negative coefficient implies that better protection of creditor interests lowers either debt from external or parent sources. Control factors comparable to the second regression are included to control for affiliate and country characteristics that may determine the composition of leverage. Following research of Desai et al. (2004), greater political risk is associated with elevated external debt whereas the local inflation rate has a positive impact on external borrowing yet reduces leverage from parent sources. In order to incorporate the delayed response of subsidiary leverage choices to changes in creditor protection laws, regressions 3 and 4 are revisited using lagged variables of the creditor right sub-indices as well as lagged variables of external and parent borrowing.

A Hausman test is conducted based on the autocorrelation of errors between countries for each regression. Test statistics are presented in Panel B and C of Table A1 in the appendix. The null hypothesis of no autocorrelation is rejected for parent borrowing, yet cannot be rejected for external borrowing since legal variation across countries likely influences affiliates’ appetite for external leverage (Houston 2010, Deakin 2016). Based on the Hausman test fixed effects are estimated for regressions including parent borrowing and random effect regressions are constructed if external borrowing is the variable of interest. Time fixed effects are included in all regressions to control for time-invariant differences between countries. The Im-Perasan-Shin unit root test is conducted to test

(23)

23 | P a g e

for non-stationarity of the data, and presented in Panel B and C of Table A2 in the appendix. External borrowing and parent borrowing are both stationary processes since the null of a unit root in all panels is rejected. F-tests and c2_{-tests are conducted to test whether the coefficients of the estimators are}

jointly significant. The ordinary least squares methodology for panel data can be used to construct regressions, since it is unlikely that internal debt levels of U.S. multinational companies and their subsidiaries influence local creditor protection policies. Desai, Fritz-Foley and Hines Jr. (2004) employ a similar framework. Random and fixed effects control for time-invariant differences between countries, and the inclusion of time fixed effects further mitigates potential biases that may be caused by time varying factors.

Unfortunately, the distinction between external and parent borrowing comes at a cost, as this variable is available only for a subset of years which limits the number of observations. Before 1999, the Bureau for Economic Analysis did not publish data on external financing of affiliates. Required information for the construction affiliate leverage composition is therefore collected for a smaller set of country-year observations which limits the statistical power of the results. Three data shortcomings with respect to identification of external and parent leverage exist. The shortcomings make the divergence between external and parent borrowing less distinct, and may result in underestimation of the effects of creditor protection and degree of substitution. The BEA data does not contain information on the level to which parent companies indemnify subsidiary loans. Second, debt is classified as external when a parent lends credit to a multinational bank, which in turn lends it to a subsidiary of the U.S. multinational company via a foreign branch. This structure of debt has substantial parent involvement and it may be questioned if the label external debt is suitable. Third, BEA classifies loans from one affiliate to another affiliate as external leverage, yet the accuracy of this classification is questionable as well.

3.4 Substitutability of parent and external leverage

If affiliates use less external leverage, and rely more on debt from parent sources, when security interests are weakly protected, parent borrowing may serve as substitute for external debt (Desai, et al., 2004). This research tests whether affiliates employ internal credit markets to substitute for costly external funds. Substitutability of parent and external debt can be estimated directly by means of a panel data regression. However, since the decision to raise external debt and borrow from parent sources is simultaneously determined, an ordinary least squares regression produces biased estimates due to endogeneity issues. Creditor protection score and capital market depth, that proxy for legal and credit market conditions, are used as instruments to measure the substitutability of parent borrowing. Affiliates are exposed to different interest rates of external debt in a way that corresponds to measures of creditor protection and capital market depth (Desai, et al., 2004). Sub-indices of creditor protection, separately and combined, instrument for external interest rates and provide clarification as to what extent affiliates alter the composition of leverage in response to differences in debt interest rates.

Creditor protection and debt levels of foreign affiliates of U.S. multinational corporations