Retained Government Ownership and the Cost of Debt Across Countries in the European Union
Ginka Borisova[(]
Finance Division
Michael F. Price College of Business
The University of Oklahoma
Norman, OK 73019
First Version: January 15th, 2006
This version August 30th, 2006
Abstract
This paper explores whether retained government ownership affects the cost of debt for privatized companies in the European Union. The sample includes both fully and partially privatized companies to test the effect of state ownership on the companies’ cost of debt. I perform the analysis in a cross-sectional setting, after controlling for both bond-specific and firm-specific variables. Decreases in the percentage of government ownership in a firm lead to a slightly higher cost of debt (one basis point on average). EMU companies have a lower cost of debt compared to non-EMU firms. Also, bond-specific variables are more successful in predicting the variation in the cost of debt than firm-specific variables.
JEL Classification: G32
Keywords: Privatization, Government, Bonds, Cost of Debt
I. Introduction
Privatization, the process of government divestiture of its holdings in state-owned enterprises (SOEs), has been increasingly explored by scholars and policy makers in order to determine its effect on both individual companies and on countries as a whole. Although divergent opinions exist on the benefits of privatization, empirical research has proved that companies experience improvement in profitability, efficiency and resource allocation following privatization (Megginson et. al. (1994)).
Megginson and Netter (2001) summarize the various areas of research in the field of privatization that have been implemented over time. These include comparisons between public and private ownership, studies of the effects of privatization in developed or developing countries or both, and measurement of initial and long-term returns for investors participating in the privatization process. The authors conclude that privatization has significantly influenced the role of government in the past two decades. Empirical research has suggested that other reforms undertaken by governments “coupled with privatization” lead to improved efficiencies of state-owned firms. Furthermore, the authors suggest that reduction in state ownership leads to improvements in corporate governance.
The benefits of privatization are confirmed by Claessens and Djankov (1998) who find that companies that have gone through privatization “have five times higher total factor productivity growth” compared to SOEs regardless of the privatization method, the presence of corruption, and the origin of the legal system. Perotti and van Oijen (1999) conclude that successful privatization programs are perceived by investors as political risk reducing events and therefore lead to substantial expansion and development of the local capital markets. The relation between share issue privatizations (SIPs) and market development has been further developed by Boutchkova and Megginson (2000), who conclude that privatization programs significantly increase the number of both individual and institutional investors, and thereby the volume of trading and development of capital and particularly stock markets.
While previous research establishes an obvious pattern of overall improvements in the operations of formerly government-owned companies, to the best of my knowledge, no paper directly explores the impact of privatization on the companies’ cost of debt. The purpose of this study is to explore how the level of government ownership affects publicly-traded debt. On the one hand, government ownership may provide an implicit guarantee to bondholders that the company will not default on its payments. Furthermore, the government usually owns nationally important companies that are deemed “too significant to fail”, and thus bankruptcy is not very probable. Thus, one could argue that SOEs may have a lower cost of debt. On the other hand, the government’s presence in a particular company may hinder the company from borrowing funds from the capital markets due to higher uncertainty about the company’s future. This could be the case if the government uses its ownership to achieve some political goals, which could benefit the nation as a whole but hurt the investors in that company (Megginson (2005)). For example, governments often maintain excess employment. Therefore, it could be the case that bondholders require a higher rate of return on their investment in SOEs.
The cost of debt after privatization is an important issue to investigate because if government divestiture leads to companies being able to borrow money from the capital markets at lower rates, they will be able to pursue more value-enhancing projects and thus bring about economic growth. The question is important to both the corporate world and regulatory bodies, since it could prove to be another benefit of privatization that governments can use in achieving their country goals and policies (Jones et. al. (1999)). Megginson et al. (1994) note that “SOEs traditionally have extremely high debt levels”. The same study, together with D’Souza and Megginson (1999) for privatization during the 1990s, finds that most companies experience a decline in their leverage ratios following privatization. Seventy percent of the companies in the sample experience this decline in the first study, compared to sixty seven percent in the second study. One of the goals of the current paper is to shed light on this trend by exploring the relationship between retained government ownership and the cost of debt for companies that have been privatized by the state.
Furthermore, Benerji and Errunza (2005) suggest that efficient investors will only choose to invest in SOEs that are fully privatized. The authors argue that when governments face the adverse selection problem, i.e. the inability to infer the “quality” of potential buyers, they should sell 100 percent of the SOE and provide direct subsidies to cover a portion of the interest costs that the company could incur if it issues new debt. Thus, the post-privatization probability of default will be reduced. The current study empirically tests how retained government ownership affects the cost of debt for both fully and partially privatized SOEs.
This issue is explored by analyzing bond-specific and firm-specific data from fourteen European Union (EU) countries[1]. The main finding of this paper is that the cost of debt is slightly lower when the government retained ownership is higher, other things being equal. Although statistically significant, this result is not economically significant since on average across models, a decrease in government ownership by one percentage point leads to an increase in the credit spread by one basis point. This result is robust to changes in the main model, and it holds true for the full sample, for EMU companies only, and when banking firms are excluded from the samples.
Consistent with previous research, such as Fama and French (1993), the analysis shows that bond-specific variables explain more of the variation in credit spreads than firm-specific variables. Furthermore, companies in the full sample that are listed as American Depository Receipts (ADRs) and have a retained government ownership of less than 10.71 percent have a higher cost of debt than similar non-ADRs; for only EMU companies, this value is 8.69 percent.
The rest of the paper proceeds as follows. Section II describes the data collection and methodology. Section III outlines the hypothesis to be tested. Section IV provides the estimation results, and Section V concludes.
II. Data and Methodology
The companies included in the sample are from the Privatization Barometer (PB) database[2]. This database contains the 25 countries in the EU, twelve of which are members of the European Monetary Union (EMU). The PB database provides information on the percentage of retained government ownership and the date when the company was initially privatized. The collected data are for 60 companies from 14 countries and 6 industries, according to PB industry classification. Table 1 provides the sample description.
The analysis primarily employs a cross-sectional OLS regression to explore the relationship between retained government ownership and the cost of debt for these companies. To control for heteroskedasticity in the data, robust standard errors are used. The bond-specific variables are collected from the Bloomberg system, and the firm-specific variables are taken from the Thomson One Database. The starting year for the data collection is 1992, following the adoption of Maastricht Agreement, which laid out the conditions of the EMU. The credit spread, the difference between the current yield to maturity (YTM) of the corporate bond under review and a government bond of the respective country, is obtained from Bloomberg by finding the interpolated point on the Treasury curve that most closely matches the corporate bond’s maturity. If the bond is Euro-denominated and issued by a company headquartered in the EMU, the credit spread is recorded as the difference between the current YTM on the bond and the most closely-matched (by maturity) sovereign bond of the same country. If the bond is Euro-denominated and issued by a company not headquartered in the EMU, the credit spread is obtained as the difference between the current YTM on the bond and the most closely-matched (by maturity) German government bond, since this is the benchmark for the EMU countries according to Bloomberg. If the bond is not Euro-denominated, regardless of the issuing company, the credit spread is obtained as the difference between the current YTM on the bond and the most closely-matched (by maturity) government bond of the country in whose currency the bond is denominated. This credit spread represents the dependent variable, cost of debt, in the model. Since the bonds used in the analysis are issued over the last decade, it is difficult to control for market conditions over time in a cross-sectional analysis. Therefore, current yield to maturity, rather than yield at issuance, is used in order to control for the fact that the latter could be affected by the market conditions at the time of the issuance. The credit spread data are collected as of November 23rd 2005, which is a Wednesday, in order to avoid any possible weekend and holiday effects. The credit spreads are recorded in basis points.
The analysis includes only bonds that do not have any convertible, callable, putable, or sinking fund features. Only fixed coupon bonds are used. Bonds could be denominated in any currency, but Euro-denominated bonds are prevalent; therefore, a dummy variable equal to one if the bond is Euro-denominated is used in the analysis as a control variable. If the bond appears as “not priced” in Bloomberg, it is excluded from the sample since it is impossible to find the interpolated point on the Treasury curve. Following Avramov et. al. (2004), negative credit spreads are excluded from the sample.
The explanatory variables of primary interest are the percentage of retained government ownership, which is obtained from the PB database, and the strength of the companies’ relationship with the government. The strength is proxied by a dummy variable, which takes a value of 1 if the government still owns a proportion of the company and 0 if it is fully privatized. This partially privatized indicator will allow the analysis to discover whether there is a significant effect on the cost of debt for government-owned companies once the government partially divests, or only after the company is completely “free” of direct government influence. The signs of the beta coefficients of these two variables are the primary subjects of this study.
The analysis also includes a series of other independent variables, which could explain the variation in the cost of debt. The variable credit rating is obtained from the Bloomberg system, and while several rating agencies’ scores are available, Moody’s is the most widespread. Therefore, if a particular bond issue does not have a Moody’s rating, it is excluded from the sample. Since the ratings are reported in terms of alphabet letters, an ordinal scale is formed with the best credit quality assigned the highest number and so on (Datta et. al. (1999), Krishnan et. al. (2005)). Following the existing literature, the natural logarithm of the credit rating is used in the regressions due to the possible nonlinearity of the relationship with the credit spread. The expected sign of the estimated beta coefficient is negative – the higher the credit rating, the lower the spread. The credit rating is used as a proxy for the probability of default, as is the leverage of the companies under review. This leverage variable is defined following Krishnan et. al. (2005), and the expected relationship is positive: the more debt the company has, the higher the spread (Collin-Dufresne et. al. (2001), Krishnan et. al (2005)).
Maturity is an important determinant of bond credit spreads. Van Landschoot (2004), in a paper that is the European counterpart of Collin-Dufresne et. al. (2001), empirically confirms that maturity plays a role, although less significant than bond rating, in determining the relationship between financial and macroeconomic news and credit spread changes. The expected relationship is positive: the longer the time to maturity, the higher the spread. This is due to the higher rate of return required by investors for parting with their money over a long period of time, during which many more events may occur relative to short-term bonds. These events include both macroeconomic and firm-specific changes that might increase the opportunity cost of investors or decrease the probability of getting back the money that they lent to the company.
Another control variable is the age of the bond. According to Houweling et. al. (2003), the age of the bond determined as the “time between the issue date and quote date” is one of the two most important variables they test in determining corporate bond liquidity for the Euro corporate bond market. The authors concur with previous research that liquidity premium is a priced factor[3]. The argument is that due to the higher transaction costs (e.g. bid-ask spreads) investors will incur if they try to reverse their position in illiquid bonds, they need to be compensated by a higher return. The authors claim that as the bonds becomes older, investors await their maturity and trade them less frequently. Therefore, the authors conclude that older bonds have higher required rates of return. On the other hand, it can also be noted that as the bond becomes older, its maturity approaches, and consequently there is less uncertainty associated with it. Therefore, age could have a negative impact on the credit spread, and the sign of the beta coefficient for this variable is an empirical question.
The size of the companies, measured as log of total assets, and the market to book ratio are also included as explanatory variables. Fama and French (1993) state that these measurements seem to explain the common variation in bond returns. The hypothesized relationship for both variables is negative. Companies with a high market to book ratio are usually growth companies that have good prospects for the future but not very many assets available to collateralize their debts. Therefore, they will borrow less, and thus should have lower spreads. By the same token, capital structure literature finds that generally large companies borrow less, and consequently they also should have lower credit spreads (Datta et. al. (1999)). However, one could argue that growth companies would have a higher cost of debt due to these same future growth options and the lack of tangible assets. In this case a high market to book ratio could have a positive impact on the credit spread.