Capital structure and the product market: The case of European electricity markets Anna Toldrày

Jerome Reboul June 2010

Abstract We study the interaction between capital structure choices and competition in the product market. We address the problem of reverse causality in this relationship by using exogenous variation in both the potential market structure and balance sheet variables. We use the deregulation of the European electricity industry as an exogenous shock to competition, and legal measures of creditor protection at the country level (from La Porta et al. (1998) and Spamann (2009)) to instrument …rms’ leverage. Using a double di¤erences strategy, we …nd that …rms reduced debt in their capital structures following deregulation, and that where credit market conditions led incumbent …rms to reduce their leverage further: i) pro…t margins diminished less after deregulation and (ii) less entry occurred following deregulation. We conclude that changes in the balance sheet had a strategic e¤ect on the goods market that is consistent with the theoretical predctions of the so-called "strategic theory of debt". Keywords: leverage, competition, reverse causality GREMAQ, University of Toulouse. E-mail: [email protected] y

Kellogg School of Management, Northwestern University. E-mail: [email protected]

1

1

Introduction

Economic theory predicts that when capital markets are perfect, competition ensures that only the most e¢ cient …rms remain in the market. However, economic models show that in the presence of imperfect capital markets e¢ cient …rms may be forced to exit because of the lack of funds. The intuition behind these models is based on predation arguments. For example, in Bolton and Scharfstein (1990) cash-rich …rms (“deep pockets”) are able to drive their e¢ cient but …nancially constrained competitors out of business by reducing their cash-‡ows; this predatory act ensures their poor performance and leads investors to terminate their funds. Thus, …nancing decisions a¤ect …rms’ability to compete in the product market ultimately determining their exit or survival. But if leverage a¤ects …rms’ competitive position, then …rms will choose their capital structures accordingly. The basic idea that …rms make …nancing decisions in order to strategically a¤ect behavior in the product market lies behind the models in the strategic theory of debt. In a seminal paper, Chevalier (1995) conducts a test of the strategic theory of debt. She studies the e¤ects of leverage on competition and exit in an industry that experienced a large number of leveraged buyouts (LBOs). She …nds that LBOs lead to a signi…cant increase in prices and increased exit in the industry. Still, as Zingales (1998) points out, the interpretation of these results is made controversial because the decision to undertake an LBO is not necessarily exogenous to the competitive environment in which the …rm operates. In fact, as argued by the trade-o¤ theory of debt, …rms will respond to shocks in the product market by adjusting their capital structures in order to balance the bene…ts of borrowing against its costs. Hence, to the extent that …rms can anticipate LBO outcomes, it is impossible to distinguish whether these outcomes are the result of increased leverage (trade-o¤ theory) or the reason why the LBO was undertaken in the …rst

2

place (strategic theory of debt). In general, an endogeneity problem arises when the attempt is made to evaluate empirically the interaction between capital structure choices and product market competition, and therefore any empirical test of the strategic debt theory may actually be a test of the trade-o¤ theory of debt. As Zingales (1998) explains, any analysis of the relationship between capital structure at the …rm level and its competitive environment should take into consideration that capital structure choices, market structure, and competitive outcomes are simultaneous. Our main focus in this paper is to provide a clean test of the strategic theory of debt. To address endogeneity issues we use two sources of variation that allow separate identi…cation of the e¤ect of the competitive environment on capital structure decisions, and of the e¤ect of capital structure decisions on the competitive environment. On the one hand, we use the deregulation of the European electricity market that took place after the European Union Directive 96/92, formalized in 1999. Deregulation in our sample of countries stemmed from legislation at the European level and deregulatory measures had to be transposed by member countries by July 2004. We argue that, given that deregulation had to occur during a determined period of time in all European countries (between 1999 and 2004), and even though there was uncertainty regarding the exact year in which it would enter into force, the timing of deregulation is exogenous. We therefore use deregulation as an exogenous source of variation in the competitive environment. On the other hand, we use six country-speci…c indexes that a¤ect the way …rms are …nanced. We obtain an index of anti-directors’rights from Spamann (2010), and we obtain 5 other determinants of …nance from La Porta et al. (1998)1 : an index of creditor protection, an index of accounting standards, an index that measures corruption, an index that measures the risk of repudiation, and an index measuring 1

We use Spamann (2010) anti-director rights index instead of La Porta et al. (1998) similar index because the

former is an revised version of the latter.

3

the e¢ ciency of the judicial system regarding business in a country. We use these variables as a source of exogenous variation in debt levels given a competitive environment. We argue that these variables a¤ect leverage but do not directly a¤ect the competitive outcome. We use a unique hand-collected dataset of electricity …rms that operated in 13 European countries from 1990 to 2007. We proceed in three steps. First, we estimate the e¤ect of deregulation on …rms’ leverage. We …nd that …rms decreased their debt levels after deregulation. This result is in line with the strategic theory of debt. Given the threat of increased competition following deregulation, …rms reduced their leverage in order to toughen competition in the post-deregulation environment2 . But this result can also be explained by the trade-o¤ theory of debt. If deregulation modi…es the costs of borrowing relative to its bene…ts –for example by increasing the risk of …nancial distress due to higher product market competition – …rms should respond by decreasing the debt ratio in their capital structures. To gain insight on the strategic theory of debt, we test the e¤ect of …rms’leverage reductions following deregulation on their price-cost margins. In order to account for endogeneity we use the country-level legal rules regarding investor protection from Spamann (20010) and La Porta et al. (1998). We show …rst that these legal measures determine …rms’…nancing patterns, as would be expected from the trade-o¤ theory of debt, but do not directly a¤ect …rms’ markups. We then instrument …rms’ leverage before and after deregulation and show that lower leverage lead …rms to obtain higher pro…ts after deregulation. This result is in line with the strategic theory of debt if we consider the possibility of entry. Our hypothesis is that if deregulation encourages entry, incumbent …rms may reduce leverage after deregulation in order to look tough to potential rivals 2

Tougher competition is de…ned in the sense of John Sutton (1991): lower debt ratios lead to lower price-cost

margins holding concentration constant.

4

(i.e. they can prey if necessary) and discourage the entry of new …rms. This allows them in the end to preserve their monopolistic rents and remain in the market. Our last test provides further support for the strategic theory of debt. We test whether leverage of incumbents a¤ects entry. We …nd that leverage reductions lead to increases in the market shares of the largest operators in every country. We conclude that, consistent with the strategic theory of debt, …rms decreased their debt levels after deregulation in order to credibly commit to be tougher competitors in the post-deregulation environment; this allowed them to prevent the entrance of new …rms and protect their margins. This paper is part of the literature that relates capital structure and product market competition. The pioneer empirical papers in this area are the works by Phillips (1995) and Chevalier (1995), which analyze the e¤ect of leverage on competition and exit in several industries that experienced leveraged buyouts (LBOs). They …nd that increased leverage increased prices and increased exit in those industries. A later paper by Kovenock and Phillips (1997), also examining the e¤ect of LBO’s but using plant-level data, …nds that leverage a¤ects plant closing and investment decisions in highly concentrated industries. As explained above, the major problem with these papers is that the decision to undertake an LBO may not be an exogenous event. Managers may decide to undertake an LBO precisely anticipating that product market conditions will change. Hence, the causal relationship is hard to establish. Chevalier and Scharfstein (1996) try to address this problem in a way similar to ours, by using time variation in the competitive environment: they show that price-cost margins of supermarket chains that are more …nancially constrained increase more during recessions relative to the price-cost margins of chains that are less …nancially constrained. To the extent that some major recessions are unexpected and modify competition in the product market, CS do rely on a possibly exogenous source of variation in the competitive environment. In a related 5

paper, Zingales (1998) studies the e¤ect of leverage on the survival of …rms in the trucking industry after deregulation. He …nds that highly levered carriers are less likely to survive deregulation. Since his major concern is that capital structure could re‡ect unobserved heterogeneity in e¢ ciency, he controls for this particular endogeneity problem, but he does not try to draw conclusions on the e¤ect of capital structure on market outcomes. Finally, a related paper by Schargrodsky (2002) addresses the question of whether …rms’…nancing choices are modi…ed as a response to competition in the product market. The author exploits the development of mass media to study the e¤ect of increased competition on leverage decisions of US newspaper …rms. He …nds that oligopolies have higher debt ratios than monopolies. In sum, these papers show either how changes in leverage a¤ect competition or how competition induces …rms to modify their …nancial structures. Our contribution is twofold. First, to the best of our knowledge, this is the …rst paper to analyze both of these e¤ects: …rst we consider whether an exogenous change in competition a¤ects …nancial leverage, and then whether the change in leverage allows …rms to sustain their competitive position in the product market. Second, we exploit two sources of exogenous variation to account for the endogeneity problems that appear in the relationship between capital structure and competition. We believe our identi…cation strategy is a progress towards a better understanding of these problems. This paper is organized as follows. In section 2 we describe the deregulation process of the electricity industry in Europe. Section 3 presents the theories of debt and their predictions. Section 4 describes our data. In section 5 we discuss our empirical strategy and the basic result. In section 6 we study the e¤ect of capital structure changes on competition. Section 7 concludes.

6

2

Electricity deregulation in Europe

In Europe the deregulation of electricity markets started with the establishment of the England and Wales Pool market in 1991. The …rst European Union Directive for the deregulation of electricity in continental Europe was established in 1996 with Directive 96/92EC and further amendments were made in 2003 with Directive 2003/54/EC. These directives established common rules for the organization and functioning of the electricity sector, access to the market, and operation of the systems, with the objective of creating a European internal market for electricity. This process o¢ cially started the liberalization of the electricity industry in Europe. This industry was traditionally organized with the vertical integration of all sections in the electricity value chain: Generation - Transmission - Distribution - Supply. Generation consists in the production of electricity. Transmission is done through the electricity highways from the producer to its designated region. Distribution is done by the regional and local electricity networks. Supply is the delivery of electricity to the end customers. In the electricity value chain, the network units (transmission, distribution) are non-competitive and constitute natural monopolies. The objective of the EU Directives is to introduce competition at the commercial units level, i.e. generation and supply. According to these Directives, the key element of the deregulation is to guarantee that market participants compete with each other for market shares in both the wholesale and retail markets by enjoying non-discriminatory access to the still monopolized transmission and distribution units. Such liberalization involves several essential measures: i) a minimum functional and legal unbundling of the previously vertically integrated activities; ii) the creation of an independent national regulator who guarantees third-party network access to the transmission and distribution lines and the introduction of new institutions such as wholesale and retail competitive markets

7

with free entry of generators and suppliers; iii) the opening of the electricity markets to the …nal consumers3 . Unbundling may occur in di¤erent degrees, which from the weakest to the strongest are: administrative unbundling, management unbundling, legal unbundling and ownership unbundling. The EU Directives require legal unbundling of all networks by speci…c dates: unbundling of the transmission unit is required by July 2004, unbundling of the distribution unit is required by July 2007. There is no obligation of ownership unbundling in the EU Directives although some countries have proceeded with that. Network access to industrial consumers must take place by 2004, and to all consumers by 2007. In August 2004, 70% of the European electricity market was open, but the degree of openness ranged from 0 to 100% depending on the country. Finally, in each country, regulatory bodies must be established to implement the EU Directives. These usually act ex-ante, supervising business agreements. However, in two countries, regulatory bodies may only become active ex post, i.e. after complaint. This occurs in Sweden and Finland. Germany has no regulatory authority. It is also the only country where the tari¤s for access to the distribution network are not set by the regulatory authority, but are freely negotiated between the …rm that wishes to enter the market and the …rm that runs the distribution network. Table A1 in the appendix shows the type of unbundling implemented, the degree of openness of each country’s electricity market, and the type of regulatory authority adopted. Below we conduct some analysis on the e¤ect of country-level heterogeneity in the deregulation process on leverage. Deregulation is an exogenous shock to each country’s competitive environment because it 3

The texts of the Directives can be found at http://europa.eu.int/comm/energy/electricity/legislation/index_en

.htm

8

stemmed from legislation at the European level. Indeed, the deregulation of electricity markets is part of the European integration process which aims to guarantee the free movement of goods, freedom to provide services, and freedom of establishment in Europe. All member states were obliged to comply with deregulation by the speci…ed deadlines but heterogeneity exists among European countries regarding the dates by which the electricity directives were transposed into domestic law. Table A2 in the appendix displays the transposition dates for the 13 countries in our sample. These dates coincide with the European Commission’s statement of deregulation in its 2005 benchmark report4 . Whereas some countries like the UK, Denmark, Austria, Finland, and Sweden were the fastest countries to deregulate; other countries like Italy, Greece, France, Luxembourg and Portugal were the slowest to enact deregulation laws. Germany, Belgium and Spain fell in between. Given that EU members enjoyed some discretion in the transposition of deregulatory measures, we may question whether deregulation was an exogenous event. However, we are con…dent that deregulation in our sample is exogenous for the following reasons. First, a number of studies show that delays longer than six months in implementing the EU directives are due to cross-national factors like the political priority assigned to the process, the timing of national elections, the political power in the European Council of Ministers, the length of membership in the EU, or the economic power of the member state; rather than idiosyncratic characteristics related to a particular law5 . A second argument for exogeneity is suggested by data on delays in the transposition of EU Directives by member countries. We obtained information for the 13 countries in our sample for the period 1986 to 2002 from Celex (Sector 7), a database on the legal issues of the European 4

Report on progress in creating the internal gas and electricity market (COM/2005/0568). Technical Annex.

European Commission 2005. 5

See Mbaye (2001), Falkner et al (2005), Konig and Luetgert (2008).

9

Union. Transposition delays vary widely across countries, ranging from 8% of directives delayed in Sweden to 51% of the directives delayed in Portugal. Table A3 in the appendix ranks member states according to the percentage of delays (in the third column) from low to high. The fourth column in the table reports the year of transposition of the EU electricity directives. A comparison of these two columns reveals that the transposition of the electricity directive is no exception: countries that have traditionally had the highest (resp. lowest) percentage of delays in implementing EU directives are also the latest (resp. earliest) ones to implement the electricity directives.

3

Theory and Predictions

The objective of our paper is to test the predictions of the models in the strategic theory of debt. The models by Fudenberg and Tirole (1986) and Bolton and Scharfstein (1990) show that cashrich …rms (“deep-pockets”) drive their high-debt rival competitors out of business (or discourage them to enter the market) by interfering in the uncertainty of future pro…ts of these rivals (FT86), or by reducing the rivals’ cash ‡ow and ensuring their poor performance so that they cannot obtain further …nancing in the capital markets (BS90). The less leveraged is the deep-pockets …rm, the more incentives it has to prey. According to these models, lower debt makes competition tougher. The models by Chevalier and Scharfstein (1996) and Dasgupta and Titman (1998) also show that higher leverage engages …rms in softer competition based on the idea that predatory pricing can take place as a result of capital budgeting decisions: when …rms raise prices they initially realize higher pro…ts, but pro…ts decrease in the future due to the loss of market share (Klemperer, 1987). In both models the incentive to increase or decrease prices depends on the discount rate of future pro…ts. Finally, the model by Faure-Grimaud (2000) makes similar predictions relying on an

10

adverse selection argument. It shows that when there is asymmetric information on ex-post pro…ts, standard debt contracts are not …rst-best because of a dead-weight loss due to the less-than-optimal level of reward/…nancing. Renegotiation of such contracts at the interim stage can force …rms to internalize this e¤ect. Thus …rms with more debt contracts will compete less aggressively in the output market in order to decrease the size of default and improve the probability of getting a good credit record for further …nancing. The model also shows that the e¤ect of …nancing costs due to asymmetric information o¤sets the positive e¤ect of limited liability on output decisions put forward by Brander and Lewis (1986). In short, models in the strategic theory of debt predict that higher leverage induces …rms to engage in softer competition by reducing output or increasing prices in the product market. A recent empirical paper (Abe de Jong et al., 2008) shows that this result is robust to whatever kind of competition is at work. These models do not make clear-cut predictions regarding the e¤ect of deregulation on …rms’ leverage. The reason is that these models do not consider entry. They predict that high debt ratios lead to higher margins because high debt commits …rms to avoid predation. But if entry is possible, then high debt accommodates entry. On the other hand, lower debt may prevent entry because of the predatory threat by incumbents desiring to keep competition low and margins high. Thus, high and low debt ratios may lead to higher margins depending on their e¤ect on entry. We expect low debt ratios after deregulation to prevent entry. We do not have clear expectations as to whether low debt should be associated with high or low margins, at least as long as we are unable to control for entry. In sum, the predictions of the strategic theory of debt are that high debt is associated with high entry and low competition, and that low debt is associated with low entry and low competition. The theory would be rejected if low debt were found to be associated with high entry and low 11

competition. Hence, in order to test the strategic theory of debt in our context, we need to consider leverage, the market outcome, and entry. However, when studying the relationship between capital structure and competition causality may exist both ways. This means that at least until we are able to account for reverse causality we need to take into account the predictions of the trade-o¤ theory of debt. The prediction is that by increasing risk and uncertainty in the industry, deregulation increases …rms’leverage above the desired level and, as a consequence, …rms should respond by lowering their debt. In addition to that, related empirical papers have documented that regulated industries have the highest debt-to-value ratios (Bradley et al. 1984, Bulan and Sanyal 2008), and other papers have provided explanations for that. Taggart (1985) for instance, argues that such high debt levels are due to the safer risk environment created by regulation. Thus, on the basis of the trade-o¤ theory of debt we expect that deregulation will have a signi…cant negative e¤ect on leverage.

4

Data

We built our main dataset from Datastream. We selected all listed …rms whose main activity was electricity generation in the European Union countries for the period between 1990 and 2007. Our dataset includes …rms in thirteen countries: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Italy, Luxembourg, Spain, Sweeden, Portugal, and the United Kingdom. Firms from new member countries are not included in our sample because they were not subject to deregulation directives before entering the EU. Firms in The Netherlands and Ireland were not listed during our sample period. Our sample includes government-owned electricity …rms that became privatized and listed during our sample period (e.g. EDP in Portugal or EDF in France). Since these …rms

12

may bene…t from particular conditions on the credit market, we will use …rm …xed e¤ects in our regressions in order to control in a non-parametric way for all permanent di¤erences these …rms could display relative to other …rms. In addition to generation, many of the …rms in our sample also owned the transmission or distribution grids before starting to deregulate. Use of …rm …xed e¤ects should also take care of di¤erences between these and generation-only …rms. Overall, our sample has 48 to 62 observations per year, except for year 2007 for which we have 11 observations. The number of total usable …rm-year observations is 955. The unbalanced nature of the panel is due to mergers or liquidation of existing …rms, as well as entry. We do not have information covering these movements. We collected country-level information on countries’implementation of the deregulation process (reported in Tables A1 and A2 in the appendix and explained in the previous section). We obtained this information from the European Commission Directorate-General of Transport and Energy. Information at …rm level includes most of the …rms’balance sheet variables and their key accounts reports. Table A4 in the appendix reports summary statistics on total debt, total assets, leverage, and operating pro…t margins for each country before and after deregulation by retaining only the permanent …rms in our sample (i.e. the ones that exist both before and after deregulation). In most countries, the mean size of …rms increases after deregulation, sometimes strongly as in Denmark or Finland. The mean and median leverage fall in half of the countries. Average and median pro…t margins increase in almost all countries. Our dataset also includes the market shares of the largest operators in each country during our sample period which we collected from Eurostat. Summary statistics of this variable are reported in Table A5 in the appendix. Average concentration is very high in Belgium, France, and Greece, where the largest operators had an average market share above 90% during our sample period; followed by Luxembourg’s largest operator which had 80% 13

market share on average during the period. The lowest average market concentrations were in UK, Finland, Austria, and Germany. Finally, we collected country-level information on the legal determinants of …nance which we obtained from the work by La Porta et al. (1998) and Spamann (2010). These authors created several indexes that quantify the extent of investor protection and legal enforcement in each country in our sample. We use six of these indexes as instruments in our regressions. A description of these indexes and summary statistics are provided in tables A6 and A7 in the appendix.

5

Changes in leverage after deregulation

5.1

Empirical Strategy

In this study we use the deregulation of the European electricity industry as an exogenous event that a¤ects the degree of competition to which …rms are exposed. As discussed in section 2, for a country that deregulates, the year of deregulation is an exogenous shock to the competitive environment. Hence, when we are interested in the e¤ect of increased competition on …rms’…nancing choices, we have two potential control groups: i) those …rms in countries that deregulate at the start of our sample period (UK), or are not subject to the European Directive during that period (Portugal); and ii) those …rms in countries that are subject to deregulation but have not yet put it in practice (i.e. all the rest of the sample before the year, speci…c to each country, in which the directive was assimilated in the nation’s body of law). The treated group corresponds to the …rms in the countries that are subject to deregulation, after deregulation was put into force. The structure of our sample makes it possible to use a di¤erence-in-di¤erence strategy. We …rst explore the change in …rms’leverage following deregulation. The fact that deregulation

14

is not simultaneous in each country forces a slight modi…cation in the usual di¤erence-in-di¤erence method. We run the following regression: j yit = Xit + 1dereg(j) + (1dereg(j) 1post(j;t) ) +

t

+

i

+ uit

j For …rm i in country j at time t, the dependent variable yit corresponds to a measure of leverage

and is explained by …rm-speci…c characteristics Xit ; common time e¤ects across Europe heterogeneity

i,

t,

…rm-level

and the treatment variables. The …rst treatment variable (1dereg(j) ) is a dummy

for the potentially treated group, i.e. it takes the value of 1 for …rms that deregulated during the sample period and 0 for …rms that did not. The second treatment variable (1dereg(j)

1post(j;t) ) is

a dummy for the post-treatment in the potentially treated group, i.e. it takes the value of 1 for …rms that deregulated in the years that deregulation took place and 0 otherwise. In our sample deregulation occurs at a di¤erent date in each country. We therefore depart from the standard di¤erence-in-di¤erence setting, where treatment occurs at a …xed date and include the dummies 1post(j;t) which take the value of 0 before the deregulation date of country j; and take the value of 1 after that. Table A2 in the appendix shows the year of deregulation for each country in our study. When there was some ambiguity about the exact timing of deregulation, usually because of a gradual opening process, we chose the last year in which the minimal requirements of the directive were implemented as the year of deregulation. This choice is, of course, somewhat arbitrary, but since our methodology gives weight to long di¤erences, we are con…dent that the results are not sensitive to it. Since identi…cation comes from the di¤erence in the change in y before and after treatment in the treatment and the control group, we include time dummies ( t ) to control for common time e¤ects across Europe (such as the introduction of the Euro). Heterogeneity in debt is widespread

15

and it is likely to be correlated with Xit through the private information of managers regarding their …rms’ prospects, as well as through a number of omitted variables such as age. We control for …rm-level heterogeneity using …rm …xed e¤ects ( i ): Also, following the suggestion by Bertrand et al. (2004), standard errors are robust (sandwich) estimates. uit is clustered at the country level. Finally, as explained in section 2, we believe that our empirical strategy is robust: …rst, deregulation is imposed on all European countries by the European Commission as part of the integration process in Europe; and second, the timing of deregulation in each country is related to speci…c internal constraints, such as countries’political processes, which are unlikely to have a direct e¤ect on …rms’balance sheet determinants. Given our cross-country analysis, we also need to consider the extent of …rms’ cross-border activity. If cross-border electricity ‡ows were substantial, our deregulation measure would measure with error the true exposure to competition of these …rms. Fortunately for our study, congestion of the cross-border interconnectors occurs very frequently. When this happens, price convergence is not possible and the neighbor electricity markets are separated6 . In fact, the DG TREN7 reports that cross-border electricity ‡ows were only 8% in 2000, and increased to only 10.7% …ve years after that8 . Therefore, cross-border integration in the electricity markets during our sample period is weak. In a number of cases, on the other hand, …rms that have not deregulated their industry hold shares in foreign …rms that already operate in a deregulated environment. This could pose a problem for the interpretation of our results; however, although …rms hold more assets at the end than at the beginning of the period, …nancial assets do not explode –which is what we would expect if cross-border participations were widespread. 6

Bosco et al. (2006). "Deregulated wholesale prices in Europe". Mimeo

7

Directorate General of Transport and Energy

8

DG TREN Benchmarking report (2005) - Technical Annex, page 24.

16

We use two alternative measures of leverage, one is the ratio of the book value of a …rm’s debt to the book value of its total assets, the other one is the ratio of the book value of a …rm’s debt to the market value of its assets. Firm characteristics include proxies for growth opportunities (market value of …rm divided by book value of assets, i.e. Tobin’s q), size (log of total assets), a measure of potential collateral (physical assets over total assets), and pro…tability (ratio of operating income to total assets). These control variables are standard in the literature (see Rajan and Zingales, 1995 and Frank and Goyal, 2003) and are meant to capture additional determinants of …rms’ capital structure choices such as agency costs, information asymmetries, and the possibility of …nancial distress. Also, following the common practice in the …eld, we lag these variables once to mitigate potential simultaneity problems. A …rm’s leverage ratio is expected to vary inversely with its growth opportunities. This prediction is due to two well-known theories. As Myers (1977) argues, …rms with risky debt tend to underinvest in value-enhancing projects. Since the cost of underinvestment increases for …rms with higher growth opportunities, such …rms will tend to …nance their projects with equity instead of debt in order to avoid underinvestment. Jensen’s (1986) argument is that debt can reduce the agency costs of free cash ‡ow. These agency costs are less severe for …rms with higher growth opportunities; hence this theory also predicts a negative relationship between leverage and growth opportunities. Size should be positively related to leverage since larger …rms have lower costs of debt and thus higher debt capacity. Collateral should be positively correlated with leverage because …rms with a larger portion of physical assets are able to obtain more debt. Finally, pro…tability should enter the regressions with a negative sign consistently with the pecking order theory (Myers and Majluf, 1984) according to which …rms tend to prioritize internal sources of …nancing over external ones (and after that, to prioritize debt over equity).

17

5.2

Results

The results from our baseline speci…cation are the following:

18

19

. . . . 0.68 955

Growth opportunities

Pro…tability

Size

Collateral

R2

N. observations

670

0.75

.

.

-0.29 (0.15)*

-0.03 (0.01)**

-0.10 (0.04)**

670

0.77

-0.02 (0.08)

0.06 (0.03)*

-0.39 (0.17)**

-0.02 (0.01)

-0.09 (0.04)**

0.63 648

738

.

. 0.56

.

-0.45 (0.17)**

-0.07 (0.02)**

-0.14 (0.06)**

.

.

.

-0.14 (0.05)**

All regressions include …rm and year …xed e¤ects, and clustered errors at the country level.

Robust standard errors in parentheses.

*** Signi…cance at the 1% level, ** at 5% level, * at 10% level.

-0.06 (0.03)*

Deregulation

(5)

(4)

(3)

(1)

(2)

Debt-to-market value of assets

Debt-to-book value of assets

648

0.64

-0.23 (0.06)***

0.07 (0.07)

-0.58 (0.18)***

-0.06 (0.02)**

-0.13 (0.06)**

(6)

Table 1 –Di¤erence-in-di¤erence regression of leverage ratios on deregulation, …rm characteristics and …xed e¤ects

Regressions (1) and (4) omit the covariates and include only …rm and year …xed e¤ects. In the rest of regressions we gradually add the controls described in the previous section including both year and …rm …xed e¤ects. The table also reports the R2 and the number of observations of each regression9 . The estimated e¤ect of deregulation is negative and signi…cant in all speci…cations. After deregulation, the ratio of debt to assets decreases by 10 percentage points and the ratio of debt to enterprise value decreases by 14 percentage points. Firm-level heterogeneity is pervasive. An F-test for individual e¤ects strongly rejects the absence of such heterogeneity. Growth opportunities and pro…tability negatively a¤ect leverage in both speci…cations, as expected. A one standard deviation increase in growth opportunities decreases the debt-to-book value of assets by 10% and the debtto-market value of assets by 16%; and a one standard deviation increase in pro…tability causes a 24% and a 26% decrease in the debt-to-book and debt-to-market value of assets respectively. Size positively a¤ects leverage, also as expected. Collateral is negative and signi…cant in the last regression, which is surprising given the theoretical predictions. Before we turn to the interpretation of the estimated treatment e¤ect, it is important to assess its robustness. Using short di¤erences by taking only the …rst di¤erence in our model gives the same qualitative results, but we lose too much variance and the estimated e¤ect is not signi…cant. Using country-speci…c time trends and their square rather than our set of dummies does not change our results, but the estimated time dummies do not display a linear relationship, so we retain our choice. We have also estimated the above model without imposing the restriction that deregulation should have the same e¤ect across all European countries. We …nd that leverage decreases after 9

The number of observations decreases in columns 2 to 6 for two reasons: i) by construction, lagged variables

demand two lags of the data; ii) Datastream does not report the entreprise value for some …rm year observations.

20

deregulation for the majority of countries, and we report the result of this regression in table A8 in the appendix. We have also exploited the fact that we have data on the various deregulatory measures that countries may use to implement deregulation. The estimated e¤ects of such deregulatory measures signi…cantly a¤ect leverage after deregulation. We report these estimates in tables A9 and A10 in the appendix. A description of the results follows these tables in the appendix. The negative e¤ect of deregulation on leverage is consistent with the strategic theory of debt. But this result can also be explained with the trade-o¤ model of debt if one is willing to accept the idea that the greater competition introduced by deregulation increases uncertainty. If this is the case, …rms rationally decrease leverage after deregulation in order to compensate for the increased probability of …nancial distress. In order to further assess whether the strategic theory of debt is in play, we explore the e¤ect of leverage decreases on competition in the next section.

6

E¤ect of capital structure changes on competition

Models in the strategic theory of debt predict that higher debt ratios make competition softer by committing incumbent …rms not to prey; or, similarly, lower debt ratios engage …rms to be tougher competitors in the product market because deep pockets …rms have higher ability to reduce prices and capture the market shares of other …rms. Thus the predictions of the strategic theory of debt should not be rejected if we …nd that lower leverage is associated with lower margins after deregulation. However, if lowering leverage is a way for incumbent …rms to commit to be tougher competitors post-deregulation, we may also …nd that lower leverage is associated with higher margins precisely because the threat of tougher competition discourages the entry of rival …rms. Thus, the strategic theory of debt should not be rejected in this case either. In sum, the

21

strategic theory of debt makes ambiguous predictions regarding the e¤ects of leverage on price-cost margins because the e¤ect of debt on competition depends upon potential entry. Hence, to test this theory we need to consider, in addition to market outcomes, the e¤ect of a change in leverage on the entry of new …rms. We …rst look at the e¤ect on margins, then on market structure.

6.1

Gross pro…t margins

In this section we test the strategic theory of debt by investigating the e¤ect of leverage reductions on …rms’ pro…t margins. The problem, as previously recognized in attempts to test this theory, is that changes in leverage are not exogenous; on the contrary, in the previous regressions we have shown that the prospect of competition a¤ects …rms’leverage (trade-o¤ theory of debt). To account for that, we adopt an instrumental variables approach, and instrument …rms’debt ratios before and after competition by using the fact that the trade-o¤ theory of debt is relevant. Firms have di¤erent incentives to take on debt before and after deregulation for reasons that are not directly related to competition. Any variable that captures these incentives is potentially a good instrument provided that it can reasonably be assumed exogenous with respect to the competitive outcome. The permanent features of countries’laws regarding the protection of creditors’interests and their enforcement have these characteristics: they a¤ect the way the prospects of greater competition a¤ect changes in capital structures, but they presumably do not a¤ect the competitive outcome itself. Our strategy is to compare the outcome of deregulation in terms of pro…t margins across countries in which …rms undertake di¤erent debt levels, using several legal determinants of …nance as exogenous sources of variation in …rms’…nancial leverage. We obtained the following …ve indexes from the work by La Porta et al. (1998): i) creditor protection, ii) accounting standards, iii) repudiation risk and iv) e¢ ciency of the judicial system, v) corruption. And the following one

22

index from Spamann (2010): anti-directors’rights. The …rst index, ‘creditor protection’, refers to countries’bankruptcy and reorganization laws. It is a sum of four dummy variables that capture the rights of creditors to seize …rm assets, the protection of seniority, the right of the judge to overcome private debt contracts, and the protection of management in bankruptcy. The second index, ‘accounting standards’, relates to company disclosure rules and the measures to make income or assets veri…able in court. These rules are important if for example a bond covenant stipulates immediate repayment when income falls below a certain level. The third index, ‘repudiation risk’, captures the risk of repudiation of contracts by the government for example due to a change of government of budget cutbacks. The fourth, ‘e¢ ciency of the judicial system’, is meant to capture the general e¢ ciency and integrity of a country’s legal environment as it a¤ects business. The …fth index, ‘corruption’measures the level of corruption of the government in a country, for example if there are bribes connected to the granting of loans. Finally, the sixth index, ‘anti-directors’rights’ which aggregates di¤erent shareholders’ rights. A detailed description of these indexes and their values for each country are provided in tables A6 and A7 in the appendix. We run the following instrumental variables regression:

j d yit = Xit + 1dereg(j;t) + leverage it

1

+ (1dereg(j;t) d leverageit )

2

+

t

++

i

+ uit

j where yit is a measure of competition of …rm i in country j at time t, i.e. the log of its price-

cost margin, de…ned as the ratio of its operating income to its net sales (in percentage terms)10 . Leverage levels before and after deregulation (leverageit and 1dereg(j;t) leverageit ) are instrumented by the interaction of the deregulation dummy (1dereg(j;t) ) and the variables that capture cross10

We take the log of operating pro…t margin (instead of the same variable in levels) because the distribution of the

log transformation is closer to a normal distribution.

23

country heterogeneity on the legal determinants of debt. The additional independent variables include, as in our previous regressions, a set of …rm-level covariates, …rm …xed e¤ects and common time e¤ects. Errors are robust and clustered at the country level. The results of our …rst-stage regression are reported in the appendix. The F-statistics displayed in our table below show that our instruments have power: the Cragg-Donald Wald F statistic ranges from 9.8 to 20.3 depending on the speci…cation11 . However, we cannot formally test the exogeneity of instruments since our model is exactly identi…ed. A Hausman test rejects the exogeneity of leverage (prob>chi2 = 0.045), which, given the validity of our instruments, con…rms the relevance of our instrumental variable strategy. If the strategic theory of debt is veri…ed, we expect

2

to be negative: lower debt ratios negatively

in‡uence competition ex-post; and at least part of the negative e¤ect of deregulation on leverage can be attributed to …rms’willingness to mitigate competition. We run six regressions: with di¤erent combinations of the covariates, with simple ols and IV:

11

The Stock and Yogo maximum 5% and 10% IV relative bias are 16.8 and 9.6 respectively.

24

25 . 0.67 . 777

Growth opportunities

R2

F-test strength of instruments

Number of observations

743

20.3

0.62

.

.

.

.

-2.91 (0.91)***

3.87 (0.64)***

1.09 (0.31)***

(IV)

706

.

0.69

.

.

1.42 (0.62)**

0.91 (0.29)***

-1.8 (0.29)***

0.3 (0.21)

0.58 (0.16)***

(OLS)

All regressions include …rm and year …xed e¤ects, and errors are clustered at the country level.

Robust standard errors in parentheses.

674

18.1

0.62

.

.

2.5 (0.98)***

1.21 (0.59)**

-3.06 (0.88)***

4.23 (0.79)***

1.12 (0.29)***

(IV)

Gross pro…t margin

Signi…cance at the 1% level (***), at the 5% level (**), at the 10% level (*).

.

.

Collateral

Size

-1.7 (0.25)***

Dereg*leverage

.

0.45 (0.25)*

Leverage

Pro…tability

0.57 (0.15)***

Deregulation

(OLS)

Gross pro…t margin

Table 2 - E¤ect of the decrease in leverage on …rms’pro…t margins

559

.

0.72

-0.05 (0.07)

-0.06 (0.12)

2.9 (1.1)**

1.38 (0.35)***

-1.88 (0.4)***

0.35 (0.33)

0.61 (0.2)***

(OLS)

527

9.82

0.66

-0.07 (0.13)

-0.11 (0.08)

4.27 (2.01)**

1.99 (0.6)***

-3.58 (1.07)***

4.32 (1.01)***

1.35 (0.37)***

(IV)

Gross pro…t margin

The estimated e¤ect of the interaction term is negative and signi…cant at the 1% level in all regressions. According to this result higher leverage is associated with lower margins in the postderegulation environment, and hence leverage decreases can be attributed at least in part to …rms’ desire to mitigate the decrease in margins. At …rst glance, this result may seem at odds with the predictions of the strategic theory of debt. However, our previous regression does not take into account the possibility of entry of new …rms. But if lower debt ratios make …rms look tougher to competitors, they may discourage entry and in this way help preserve the margins of incumbent …rms. This is what we investigate in the next section. Additionally, two of the four control variables, collateral and pro…tability, are positive and signi…cant. We performed robustness tests using time trends and shorter di¤erences; our results are maintained qualitatively but we lose some signi…cance. Incidentally, a seemingly surprising result is that gross margins are signi…cantly higher after deregulation. This may come from productivity enhancement or rationalization of the industry structure due to higher competition. We do not believe that this correlation questions our interpretation of the gross margin as a measure of the intensity of competition. The reason is that the positive mean e¤ect of deregulation has to be interpreted as the mean e¤ect of deregulation on margins across countries, while we identify the di¤erential e¤ect of debt on the change in pro…t between countries. It is thus reasonable to believe that countries that experienced, everything else equal, a higher rise in margins after deregulation, did not take full advantage of competition. We argue that this is because they encouraged greater debt reduction.

26

6.2

Entry

As we previously explained, another direct way of gauging the e¤ect of changes in leverage on competition is to test whether lower debt discourages the entry of rival …rms. In principle, the best way to do this is to measure entry directly, as in the seminal paper by Chevalier (1995), but unfortunately we do not have such data. We use a measure of market concentration, namely, …rms’ market shares, as our proxy for entry12 . Because our sample only includes the …rms that are listed, we cannot use an internal measure of market share. However, the European Commission asked Eurostat to publish data on …rms’market shares for the period 1999 to 2007. We use this data to test whether leverage had an e¤ect on the market share of historical incumbents after deregulation. We run the following regression:

j d yit = 1dereg(j;t) + leverage it

1

+ (1dereg(j;t) d leverageit )

2

+

t

++

j

+ uit

j where yit is the log of the market share of the largest operator in every country, and leverage is

now the size-weighted mean leverage among the operators of that country, in the spirit of Kovenock and Phillips (1995). As previously, we instrument this size-weighted mean with permanent countrylevel di¤erences related to corporate law and legal enforcement. We include country …xed e¤ects and a time trend in a …rst speci…cation, and country and time …xed e¤ects in a second regression. We run these regressions at the country level. Note that each i is associated with a single j in this regression since we have only the market share of the largest operator in every country. Because of the strong reduction in the sample size, we lose some identifying power. 12

Another widely used measure of concentration is the Her…ndahl-Hirschman Index (HHI). Unfortunately such

data is only available in a recent European Commission report for the years 2003 to 2005 for 5 countries. We have tried to exploit this data but the number of observations is too small to obtain meaningful results.

27

We obtain the following results: Table 3 - E¤ect of leverage reduction on market shares Market share

Market share

Leverage

-1.07 (1.24)

1.48 (1.18)

Dereg*leverage

-0.3 (0.18)*

-0.98 (0.52)**

Year FE

No

Yes

Time trend

Yes

No

Country FE

Yes

Yes

R2

0.83

0.57

N. observations

90

90

** Signi…cance at the 5% level, * at 10% level. Robust standard errors in parentheses. The debt ratio and the market share of the largest operator are negatively and signi…cantly correlated at least at the 5% level. These results support the prediction that higher leverage of incumbents facilitates entry, and hence incumbents’leverage reductions after deregulation can be attributed to incumbents’desire to discourage entry. This result is in line with the predictions of the strategic theory of debt. The Cragg-Donald Wald F statistic for the strength of instruments equals 8 in both speci…cations.

7

Conclusion

We use the deregulation of the European electricity industry as a natural experiment to study the interaction between competition and capital structure choices of …rms. According to the strategic theory of debt, lower leverage makes …rms tougher competitors in the product market. We …nd that 28

…rms decrease their debt levels after deregulation. This result is consistent with the strategic theory of debt. We then study the e¤ect of leverage reductions on competition in the product market. Our hypothesis is that …rms decrease debt in their capital structures as a credible commitment to prey on potential rivals. In order to deal with endogeneity problems we instrument …rms’ debt levels before and after deregulation with four indexes of creditor protection and legal enforcement at the country level. We test the strategic theory of debt by regressing instrumented leverage on pro…t margins and on entry. We …nd that reducing leverage allows …rms, everything else equal, to enjoy higher margins in the post deregulation environment and prevent the entry of new …rms.

References [1] Bertrand, Marianne, Esther Du‡o and Sendhil Mullainathan (2004), "How much should we trust di¤erences-in-di¤erences estimates?", The Quarterly Journal of Economics 119, 249-275. [2] Bolton Patrick, and David Scharfstein (1990), "A Theory of predation based on agency problems in …nancial contracting", The American Economic Review 80, 93-106. [3] Bradley, Michael, Gregg A. Jarrell and E. Han Kim (1984), "On the existence of an optimal capital structure: Theory and evidence", Journal of Finance 39, 857-880. [4] Brander, Tracey, and Lewis (1986), "Capital structure and product market behaviour: The limited liability e¤ect", The American Economic Review 76, 956-970. [5] Bulan, and Sanyal (2008), "Regulatory risk, ,arket risk and capital structure: Evidence from US electric utilities", Mimeo Brandeis University

29

[6] Chevalier, Judith (1995), "Capital structure and product market competition: Empirical evidence from the supermarket industry", The American Economic Review 85, 415-435. [7] Chevalier, Judith and David Scharfstein (1996), "Capitral market imperfections and countercyclical markups: theory and evidence", The American Economic Review 86, 703-725. [8] Dasgupta, Sudipto and Sheridan Titman (1998), "Pricing strategy and …nancial policy", Review of Financial Studies 4, 705-737. [9] De Fraja, Gianni and Clive Stones (2003), "Risk and capital structure in the regulated …rms", Mimeo, University of York. [10] Falkner, Gerda, Miriam Hartlapp, Simone Leiber and Oliver Treib (2005) Complying with Europe: EU Harmonization and Soft Law in the Member States. Cambridge: Cambridge University Press. [11] Faure-Grimaud, Antoine (2000), "Product market competition and optimal debt contracts: The limited liability e¤ect revisited", European Economic Review 44, 1823-1840. [12] Fudenberg, Drew and Jean Tirole (1986), "A ‘signal-jamming’ theory of predation", Rand Journal of Economics 17, 366-376. [13] Harris Milton, and Arthur Raviv (1991), "The theory of capital structure", Journal of Finance 39, 127-145. [14] Jensen, Michael and W. Meckling (1976), "Theory of the …rm: Managerial behavior, agency costs, and capital structure", Journal of Financial Economics 3, 305-360.

30

[15] Jensen, Michael (1986), "Agency costs of free cash ‡ow, corporate …nance and takeovers", The American Economic Review 76, 323-329 [16] Klemperer, Paul (1987), "Markets with consumer switching costs", Quarterly Journal of Economics 102, 375-394. [17] Konig, Thomas and Brooke Luetgert (2008) ‘Troubles with Transposition? Explaining Trends in Member-State Noti…cation and the Delayed Transposition of EU Directives’, British Journal of Political Science 39: 163 –94. [18] Kovenock, and Gordon Phillips (1997), "Capital structure and product market behaviour: An examination of plant exit and investment decisions", Review of Financial Studies, 3, 767-803. [19] La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer and Robert Vishny (1997), "Legal determinants of external …nance", Journal of Finance 52, 1131-1150. [20] La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer and Robert Vishny (1998), "Law and …nance", Journal of Political Economy 106, 1113-1155. [21] Maksimovic, Vojislav (1988), "Capital structure in repeated oligopolies", Rand Journal of Economics 19, 389-407. [22] Mbaye, Heather A. D. (2001) ‘Why National States Comply with Supranational Law: Explaining Implementation Infringements in the European Union, 1972-1993’, European Union Politics 2: 259-81. [23] Myers, Stewart C. (1977), "Determination of corporate borrowing", Journal of Financial Economics 4, 147-175.

31

[24] Phillips, Gordon (1995), "Increased debt and industry product markets", Journal of Financial Economics 37, 189-238. [25] Rajan, Raguram and Luigi Zingales, (1995), "What do we know about capital structure ? Some evidence from international data", The Journal of Finance 50(5), 1421-1460. [26] Schargrodsky (2002), "The e¤ect of product market competition on capital structure: Empricial evidence from the newspaper industry", Working Paper Universidad Torcuato di Tella. [27] Spamann, Holger (2010), "The ‘Antidirector Rights Index’ revisited", Review of Financial Studies 23 (2): 467 - 486. [28] Sutton, John (1991), "Sunk costs and market structure: Price competition, advertising, and the evolution of concentration",. Cambridge, MA: MIT Press. [29] Taggart, Robert A. Jr. (1985), "E¤ects of regulation on utility …nancing: Theory and evidence", Journal of Industrial Economics 33, 257-276. [30] Zingales, Luigi (1998), "Survival of the …ttest or the fattest? Exit and …nancing in the trucking industry", Journal of Finance, 53, 905-938.

8 8.1

Appendix Tables

32

33

Market opening (2004) 100% 80% 100% 100% 37% 100% 34% 66% 57% 45% 100% 100% 100%

Country

Austria

Belgium

Denmark

Finland

France

Germany

Greece

Italy

Luxembourg

Portugal

Spain

Sweden

United Kingdom

ownership

ownership

ownership

legal

management

ownership / legal

management

legal / management

management

ownership

legal

legal

legal

Unbund. transmission

Table A1 –Unbundling, market opennes and regulator

legal

legal

legal

accounts

accounts

legal

accounts

accounts

accounts

accounts

legal

legal

accounts

Unbund. distribution

ex-ante

ex-post

ex-ante

ex-ante

ex-ante

ex-ante

ex-ante

planned

ex-ante

ex-post

ex-ante

ex-ante

ex-ante

Regulator

Table A2 - Electricity transposition dates by country Country

Law and transposition date

Austria

Law of Electricity Supply –1998

Belgium

Law for the Organization of the Electricity Market –April 29, 1999

Denmark

Amendment to Danish Supply Act –1998

Finland

Electricity Market Act –1995

France

Law No. 2000-108 –2000

Germany

Act on the Supply of Electricity and Gas –1999

Greece

Electricity Law –2000

Italy

Law 239/2004 –2004

Luxembourg

Law of 24 July 2000 on the Organization of the Electricity Market –2000

Portugal

Infringement (implemented August 2006)

Spain

Electricity Power Act –Novermber 1997 and amendments 1998

Sweden

Law for the Supply of Electricity –1999

United Kingdom

Electricity Act –1989

Source: European Commission benchmark report 2005

34

Table A3 –EU member states’response to transposition of EU Directives* Country

Number of

Number of

Year of transposition

(ranked high to

directives with one

delayed directives

of EU electricity

low according to

or more delayed

as percentage of

Directive 96/92

value in 2nd column)

measures

total

Sweden

60

8%

1999

Finland

125

17%

1995

Denmark

320

20%

1998

Germany

400

25%

1999

Austria

198

27%

1998

UK

475

30%

1989

France

495

31%

2000

Spain

503

32%

1998

Belgium

539

34%

1999

Italy

615

39%

2004

Greece

625

39%

2000

Luxembourg

639

40%

2000

Portugal

813

51%

2006

Source: Konig and Luetgert (2008) *Period 1986 to 2002. For this period the total number of EU Directives equals 722 for Austria, Finland and Sweden (because they entered the EU in 1995), and it equals 1,592 for the rest of countries. 35

36

37

Finland

Denmark

2300793 0.4013723 2.367868

Total Assets

Total Debt / Total Assets

Pro…t Margin (Log)

0.6618511

Pro…t Margin (Log)

894959.6

0.1817442

Total Debt / Total Assets

Total Debt

1361651

Total Assets

0.412416

Pro…t Margin (Log) 261624.1

0.2099415

Total Debt / Total Assets

Total Debt

11300000

Total Assets

2.188693

Pro…t Margin (Log) 2312533

0.2616662

Total Debt / Total Assets

Total Debt

2980558

Total Assets

Belgium

1187963

Total Debt

Austria

Mean

Variable

Country

Table A4 –Summary statistics of main variables

2.269028

0.4123075

2626440

883049

1.258848

0.1726181

1691271

231024

0.6573455

0.1884964

11500000

2170823

2.225675

0.1890975

1710853

163978

Median

0.1847411

0.0954137

520785.6

191766.7

1.3304

0.0558669

878457.6

245809.7

1.030961

0.0987961

704795.7

956496.2

0.6136152

0.1918215

2591733

1522409

Std. Dev.

Before deregulation

5

5

5

5

10

15

15

15

6

9

9

9

26

27

27

27

N. obs

2.527958

0.30638

13300000

3940175

1.558467

2.621766

0.3018439

15000000

4462362

1.919386

0.2457199

793534.5

4515115 0.2229291

300469

1.483875

0.155735

495824.4

1.471847

0.1507663

19200000

2665108

2844597 19100000

2.347558

2.339848

0.2310319

2657195

3397122 0.3025631

590041

Median

1380166

Mean

0.6356626

0.0618696

5404316

1713453

0.777346

0.1422128

5062557

391557.9

0.3452389

0.0235043

3638759

576249.7

0.5209112

0.1556559

2647157

1510270

Std. Dev.

After deregulation

13

13

14

13

10

14

14

14

5

6

6

6

21

22

22

22

N. obs.

38

1.711782

Pro…t Margin (Log)

Italy

Greece

18400000 0.2847756 2.292797

Total Assets

Total Debt / Total Assets

Pro…t Margin (Log)

3.117933

Pro…t Margin (Log) 6865551

0.2609036

Total Debt / Total Assets

Total Debt

1747253

Total Assets

1.217433

Pro…t Margin (Log) 1188569

0.063985

Total Debt / Total Assets

Total Debt

2225480

Total Assets

167126.1

0.1088544

Total Debt / Total Assets

Total Debt

31300000

Total Assets

Germany

7253506

Total Debt

France

Mean

Variable

Country

2.461675

0.2925542

2141931

612781

3.130275

0.1305814

42437.5

4474

1.113489

0.0102487

962059.5

9701

1.866698

0.0860149

10400000

743119

Median

0.8892154

0.1679732

24300000

9392462

0.243202

0.2718608

3068413

2136432

0.9988723

0.1171637

2461969

300275.3

1.421836

0.1089136

42900000

12000000

Std. Dev.

Before deregulation

Table A4 –Summary statistics of main variables (continued)

32

38

38

38

12

12

12

12

60

98

98

98

34

40

40

40

N. obs

2.417893

0.3810864

21100000

4359939

2.75535

0.342266

6261873

2059883

1.85635

0.1463339

4734455

2.425355

0.3947598

4114146

1520529

2.880689

0.3097846

7736223

1882307

2.070653

0.0757195

986522

44700

1.76815

1.948235 677937.3

0.1188943

10800000

418648.5

Median

0.1675998

43700000

7209536

Mean

0.8670368

0.0978501

37400000

7057078

0.6910734

0.0992147

6054570

2111979

0.9635733

0.1700937

7775698

1501937

0.7379595

0.1836194

63400000

12300000

Std. Dev.

After deregulation

18

18

23

18

12

12

15

12

55

86

95

86

27

30

28

30

N. obs.

39

UK

Sweden

Spain

Luxembourg

Country

0.1004959 2.349377

Total Debt / Total Assets

Pro…t Margin (Log)

. . .

Total Assets

Total Debt / Total Assets

Pro…t Margin (Log)

3.201595

Pro…t Margin (Log) .

0.3393735

Total Debt / Total Assets

Total Debt

20900000

Total Assets

3.084299

Pro…t Margin (Log) 8694998

0.3890678

Total Debt / Total Assets

Total Debt

8426987

Total Assets

3543036

166762.5

Total Assets

Total Debt

14899

Mean

Total Debt

Variable

.

.

.

.

3.235053

0.3612042

13000000

6087500

3.209633

0.3732302

6601114

3093980

2.383878

0.091481

155144.5

17309.5

Median

.

.

.

.

0.1576872

0.1395971

17400000

8024220

0.5475339

0.1292668

7488462

3272695

0.2310167

0.0739599

78700.81

8699.331

Std. Dev.

Before deregulation

Table A4 –Summary statistics of main variables (continued)

.

.

.

.

18

18

18

18

43

43

43

43

20

20

20

20

N. obs

2.371453

0.208542

2444021

630345.8

3.129923

0.3013217

22400000

7985714

2.717457

0.3890929

19400000

2.424803

0.1575242

1142550

162500

3.117507

0.2520944

10700000

2452000

2.8088

0.4079618

14400000

5776846

2.197712

1.939518 7792276

0.1055411

325755

14452

Median

0.0883328

316982.1

16631

Mean

0.6117612

0.2004267

2999802

1019644

0.20279

0.0903744

20700000

9031409

0.5897887

0.1205678

20800000

8177153

0.9527224

0.0434401

254668.7

7781.646

Std. Dev.

After deregulation

127

146

150

146

7

7

7

7

47

48

53

48

14

14

15

14

N. obs.

40

Total

Country 2929593 9673989 0.1941941 2.094401

Total Assets

Total Debt / Total Assets

Pro…t Margin (Log)

Mean

Total Debt

Variable

2.274699

0.1460332

1791755

224680

Median

1.172195

0.1821603

20500000

6498686

Std. Dev.

Before deregulation

Table A4 –Summary statistics of main variables (continued)

266

325

325

325

N. obs

0.1951756 2.387385

2.28705

1584857

291533.5

Median

0.2300815

9617585

2417245

Mean

0.779465

0.1851061

22700000

5521335

Std. Dev.

After deregulation

356

416

442

416

N. obs.

Table A5 –Average market share of the largest generator Country

Mean

Median

Std. Dev.

Austria

29.47

32.6

6.1

Belgium

91.52

92.15

1.9

Denmark

38.79

36

6.2

Finland

24.92

26

1.6

France

90.01

90

1.5

Germany

29.92

28.7

2.3

Greece

97.02

97

2.5

Italy

45.56

45

10.7

Luxembourg

80.9

80.9

0

Portugal

57.84

57.8

3.1

Spain

40.37

41.2

6

Sweden

48.35

48.5

2.32

UK

20.98

21

1.1

Total

40.46

32

23.47

Source: Eurostat

41

42

Accounting Standards

An index aggregating di¤erent creditor rights. The index is formed by adding 1 when: (1)

Creditor Rights

70 percent while …nancial companies represented the remaining 30 percent.

represent a cross-section of various industry groups where industrial companies numbered

special items). A minimum of 3 companies in each country were studied. The companies

statements, balance sheets, funds ‡ow statement, accounting standards, stock data and

omission of 90 items. These items fall into 7 categories (general information, income

Index created by examining and rating companies’1990 annual reports on their inclusion or

the resolution of the reorganization. The index ranges from 0 to 4.

a bankrupt …rm; and (4) the debtor does not retain the administration of its property pending

ranked …rst in the distribution of the proceeds that result from the disposition of the assets of

reorganization petition has been approved (no automatic stay); (3) secured creditors are

reorganization; (2) secured creditors are able to gain possession of their security once the

the country imposes restrictions, such as creditors’consent or minimum dividends to …le for

Description

Variable

Table A6 - Description of instrumental variables

43

Corruption

ICR’s assessment of the “risk of a modi…cation in a contract taking the form of a repudiation,

Repudiation Risk

original range going from 0 to 6).

to 10, with lower scores for higher levels of corruption. (We changed the scale from its

the months of April and October of the monthly index between 1982 and 1995. Scale from 0

export licenses, exchange controls, tax assessment, policy protection, or loans”. Average of

throughout lower levels of government” in the form of “bribes connected with import and

o¢ cials are likely to demand special payments” and “illegal payments are generally expected

ICR’s assessment of the corruption in government. Lower scores indicate “high government

lower scores for higher risks.

of April and October of the monthly index between 1982 and 1995. Scale from 0 to 10, with

government, or a change in government economic and social priorities.” Average of the months

postponement, or scaling down” due to “budget cutbacks, indigenization pressure, a change in

Description

Variable

Table A6 - Description of instrumental variables (continued)

44

waved by a shareholders’vote. The index ranges from 0 to 6.

or equal to 10 percent (the sample median); or (6) shareholders have preemptive rights that can only be

of share capital that entitles a shareholder to call for an Extraordinary Shareholders’Meeting is less than

of directors is allowed; (4) an oppressed minorities mechanism is in place; (5) the minimum percentage

Shareholders’Meeting; (3) cumulative voting or proportional representation of minorities in the board

vote to the …rm; (2) shareholders are not required to deposit their shares prior to the General

rights.” The index is formed by adding 1 when: (1) the country allows shareholders to mail their proxy

Anti-director rights An index aggregating the shareholder rights which we labeled as “anti-director

e¢ ciency levels.

in question”. Average between 1980-1983. Scale from 0 to 10, with lower scores lower

Corporation. It “may be taken to represent investors’assessments of conditions in the country

Source: all indexes are from La Porta et al. (1998) except for the Anti-director Rights index which we took from Spamann (2010)

ICR: International Country Risk Agency

Anti-director Rights

Assessment of the “e¢ ciency and integrity of the legal environment as it a¤ects business,

E¢ ciency of Judicial System

particularly foreign …rms” produced by the country-risk rating agency Business International

Description

Variable

Table A6 - Description of instrumental variables (continued)

45 2 . 1 2 2 4

Italy Luxembourg Portugal Spain Sweden United Kingdom

78

83

64

36

.

62

55

62

69

77

62

61

9.63

9.58

8.40

8.57

.

9.17

6.62

9.77

9.19

9.15

9.31

9.48

9.60

Risk

Repudiation

10

10

6.2

5.5

.

6.75

7

9

8

10

10

9.5

9.5

Judicial System

E¢ ciency of

9.10

10

7.38

7.38

.

6.13

7.27

8.93

9.05

10

10

8.82

8.57

Corruption

4

4

5

3

.

2

3

4

5

4

4

3

4

Rights

Anti-director

we obtained from Spamann (2010).

Source: all indexes come from La Porta et al. (1998) except for the Anti-director Rights index which

1

0

France

Greece

1

Finland

3

3

Denmark

Germany

2

Belgium

54

Standards

Rights 3

Accounting

Creditor

Austria

Country

Table A7 –Instrumental variables. Values by country.

Table A8 –Deregulation e¤ect by country Austria

-0.03 (-0.73)

Belgium

-0.12 (-3.34)***

Denmark

-0.10 (-3.77)***

France

0.02 (0.93)

Finland

-0.34 (-1.27)

Germany

-0.09 (-2.98)***

Italy

-0.11 (-3.5)***

Luxembourg

-0.09 (-2.23)**

Spain

-0.10 (-3.39)***

Sweden

-0.10 (-3.41)***

Greece

-0.28 (-1.13)

R2

0.79

N. observations

699

***Signi…cance at 1%, **at 5% levels t-statistics are in parentheses

46

Table A9 –E¤ect of heterogeneity in the deregulation process Debt-to-book value of assets Deregulation dummy

-0.08 (-2.24)**

Regulator ex-ante

0.044 (1.94)*

Unbund. transmission: legal

-0.02 (-1.90)*

Unbund. transmission: management

0.003 (0.07)

Unbund. distribution: legal

-0.07 (-2.22)**

R2

0.77

N. observations

661

***Signi…cance at 1% level, **at 5% level, *at 1% level t-statistics are in parentheses Table A9 - Discussion This table reports the impact of various deregulatory measures on leverage. As far as unbundling of the transmission business is concerned, the European Directive leaves each country discretion to choose how separation between the transmission and distribution businesses should be made from generation and sales. There is a three level classi…cation which is, from the weakest to the strongest, the following: i) unbundling can be implemented with the separation of management within an entity that is also in the commercial business, ii) it can be implemented with the legal separation of the entity that manages the network and the other …rms that operate in the market or …nally, iii) it can be implemented with separate ownership. Italy has a mixed regime as well as Greece. In Italy ownership separation is partial in the sense that incumbents still hold shares in the transmission business that is nevertheless run as a separate entity. Thus this is a mixed model

47

of ownership and legal separation. In Greece it is part of a group that also includes the historical operator though it is in principle run independently. Thus this is closer to the management type of separation. In the sequel, we classify Italy in the ownership group and Greece in the management group. A similar classi…cation can be made for the distribution business where unbundling can be implemented with i) the separation of the distribution and sales accounts within a single …rm, ii) the legal separation of the two businesses. Countries are almost equally divided along this line: France, Finland, Germany, Austria, Greece and Portugal chose the …rst system, while the others chose the second. As regarding the regulatory body that implements the directives in every country, in the majority of countries these usually act ex ante, supervising business agreements; however, in two countries (Finland and Sweden) regulatory bodies may only become active ex post, after complaint. Germany has no regulatory authority: it is the only country where the tari¤s for access to the distribution network are freely negotiated between the …rm that wishes to enter the market and the …rm that runs the distribution network, as opposed to being set by the regulator. Because of colinearity problems we cannot separately identify the e¤ects of all the deregulatory measures in the same regression; therefore, we have chosen to present only some of them13 . These results are all to be interpreted relative to the mean e¤ect of deregulation. Unbundling the transmission and distribution units through legal separation magni…es the incentives to decrease leverage after deregulation. In so far as a more radical unbundling fosters competition and entry, these results make sense. On the other hand, lower regulatory risk signi…cantly a¤ects the desire to reduce debt: having a regulatory body that acts ex-ante diminishes the desire to reduce leverage after deregulation relative to having one that acts ex-post, which is widely seen as more 13

Additional regressions are available from the authors.

48

uncertain14 . Our usual set of covariates as well as …rm and year …xed e¤ects are also included in the above regression even though we do not report the results. The coe¢ cients and signi…cance of the covariates do not change with respect to our main regression. Robust standard errors are also clustered at the country level. Table A10 –E¤ect of market opening (year 2004) Debt-to-book value of assets Deregulation

0.01 (0.29)

Market opening

-0.13 (-2.57)**

R2

0.77

N. observations

661

** Signi…cance at 5% level t-statistics are in parentheses Table A10 - Discussion The proportion of the electricity market (in terms of annual consumption) that is open to competition varies across countries during our sample period (see table A1). According to the EU Directive, this percentage cannot be lower than 30% but it is in fact higher in most countries. A trouble with this measure though is that most countries have a gradual opening strategy. As 14

Although we can see that …rms’leverage is signi…cantly related to di¤erences in the deregulation process, we did

not attempt to use this heterogeneity and consider deregulation as a single discrete event for most of the paper. We believe this is not a problem for our identi…cation strategy when we explore the e¤ect of deregulation on leverage since what we need is simply that deregulation is associated to greater exposure to competition. However, in the second part of the paper, where we explore the e¤ect of …rms’ balance sheet on the outcome of deregulation, the validity of our strategy hinges on the assumption that the insruments we use for debt are uncorrelated with underlying heterogeneity in the deregulation process. We believe this is a reasonnable assumption.

49

a consequence, even if this gradual opening is known and foreseen ever since its beginning, it is di¢ cult to use a time-varying measure for openness in a short di¤erence approach since changes in the capital structure are forward looking decisions. We have thus run a regression by using countries’market opening as of year 2004. We have run the same regression with countries’market opening as of year 2006 and the results are maintained. The results show that the extent to which the market is open makes a large di¤erence. The more open the market, the larger the debt reduction. This is consistent with our earlier …ndings and with both the predictions of the strategic and trade-o¤ theories of debt. Our usual set of covariates as well as …rm and year …xed e¤ects are also included in the above regression even though we do not report the results. The coe¢ cients and signi…cance of the covariates do not change with respect to our main regression. Robust standard errors are also clustered at the country level.

8.2

First stage regression

We run the following …rst stage regression for our IV strategy:

j yit = Xit + Z j + (1deregj;t

Zj) +

i

+

t

+ uit

j where yit is the ratio of debt to the book value of assets of …rm i, at country j and at time

t and Z j is our set of instruments explained above. We include the usual set of determinants of leverage, …rm …xed e¤ects and time dummies. Errors are robust and clustered at the country level.

50

We obtain the following results for the interaction terms of the instruments: Table A11 –First stage of IV regression Debt to book value of assets

Debt to book value of assets*Dereg

Creditor Rights

-0.04 (0.01)***

0.04 (0.02)***

Accounting Standards

-0.003 (0.002)

0.01 (0.002)***

Repudiation Risk

-0.00006 (0.01)

-0.07 (0.02)***

E¢ ciency Judicial System

0.066 (0.04)*

0.22 (0.04)***

Corruption

-0.06 (0.05)

-0.34 (0.06)***

Anti-director Rights

0.04 (0.03)*

0.17 (0.03)***

N. observations

527

527

R2

0.81

0.86

*** Signi…cance at the 1% level, ** at the 5% level, * at the 1% level Robust standard errors in parentheses Clear patterns emerge regarding the e¤ect of all six indexes on leverage. Moreover the joint F-statistics for the six variables reject the hypothesis that all coe¢ cients are 0 at the 1% level: F-statistics are 31.45 and 24.30 in the …rst and second regressions respectively.

51