Business Group A¢ liation, Financial Development and Market Structure: Evidence from Europe Sharon Belenzonyand Tomer Berkovitzz January 23, 2008
Abstract This paper investigates the e¤ect of …nancial development on the incentives to form business groups. We examine how this relation varies across exogenous industry conditions, legal environments, and …rms’ life cycle. Using a comprehensive dataset on group a¢ liation of European …rms, we …nd evidence that business groups substitute imperfect …nancial markets via their internal capital market. Our results highlight the constructive role of business groups in economies with less developed …nancial markets and contribute to the debate on whether business groups should be dismantled. Keywords: business groups, …nancial development, internal capital markets JEL Classi…cation: G32, G18, and O16
Acknowledgement: We express our special gratitude to Patrick Bolton, Francisco PerezGonzalez, Mark Schankerman, and John Van Reenen for numerous helpful discussions. We are also grateful for valuable comments from Nick Bloom, Bjorn Jorgensen, Daniel Paravisini, Tano Santos, Yishay Yafeh and seminar participants at Columbia University (GSB), London School of Economics, and the NBER. We thank Liat Oren for invaluable assistance with the programming of the ownership algorithm and Hadar Gafni for excellent research assistance. All remaining errors are our own. y NBER, Oxford University, and London School of Economics (
[email protected]). z Columbia University, Graduate School of Business (
[email protected]).
I.
Introduction
Business groups have been at the heart of …nancial economics literature in recent years.1 Many studies have documented the ubiquity of business groups (most notably La Porta et al. (1999)) and their e¤ect on …rm performance (e.g. Hoshi et al. (1991), Khanna and Palepu (2000), Gopalan et al. (2007)).2 Yet, there is little evidence on the incentives of …rms to form business groups and, more speci…cally, on how institutional di¤erences between countries may a¤ect these incentives. Le¤ (1978) argues that business groups can be viewed as an organizational response to imperfect or missing markets. Though an appealing argument with important policy implications, it is not straightforward to test Le¤’s argument empirically for several reasons. First, business groups are not only a¤ected by market development, but can also restrain the development of the institutions they mimic –thus, it is hard to infer causality. Second, omitted or latent macro variables can be correlated with both market development and the incentive to join a business group. Finally, most business group a¢ liates are privately-held corporations. Hence, a multinational database with large coverage is needed in order to test this hypothesis. In this paper, we develop a comprehensive database on business group a¢ liation and …nancials for over 80K …rms in 15 European countries to examine the e¤ect of …nancial development on business group a¢ liation. We focus on …nancial development since most studies reporting a positive e¤ect of business group a¢ liation on …rm performance argue that the business group internal capital market is the key mechanism leading to this e¤ect.3 If group internal capital market …lls the void of imperfect …nancial markets, then we would expect …rms in countries with less developed …nancial markets to have a higher incentive to join a business group. We deal with the potential reverse causality and unobserved heterogeneity issues by focusing on the channel through which …nancial development a¤ects business group a¢ liation. Speci…cally, developed …nancial markets reduce the cost of external funding by mitigating moral hazard and adverse selection problems. In countries with less developed …nancial mar1
kets, …rms join together to form business groups which potentially grant them access to a signi…cant internal capital market and reduce their cost of capital. Thus, …nancial development should disproportionately a¤ect the probability of a …rm to join a business group in industries that, for exogenous reasons, rely more on external funding or have higher degree of asymmetric information. Our industry measures are computed based on a sample of US …rms, where business groups are not common (La Porta et al. (1999)), so they are largely immune to endogeneity problems. Since we are only interested in the relative values of the industry measure, it is enough to assume that the ranking of industries according to external dependence and information asymmetry is mostly due to technological di¤erences and, therefore, similar across countries (Rajan and Zingales (1998)). By comparing the di¤erence in group a¢ liation between industries in each county rather than the absolute level of group a¢ liation across countries, we mitigate two common problems in cross-country studies. First, since there are less than two hundred countries in the world and only a relatively small fraction of these countries have available data on ownership structure and …nancials, most cross-country studies run into a degrees of freedom problem. By using the within-country between-industry variation, we increase the number of observations and alleviate this problem. Second, …rms’coverage may vary across countries.4 This problem is mitigated to some extent by our sample inclusion criteria. Yet, our empirical test, which compares the industry distribution of group a¢ liation between countries rather than the absolute percentage of group a¢ liates, alleviates the problem even further. We …nd that the di¤erence in business group a¢ liation between industries with high external dependence (asymmetric information) and low external dependence (asymmetric information) is higher in countries with lower …nancial development. The results suggest that the ex-ante development of …nancial markets facilitates the ex post formation of business groups in sectors that can potentially bene…t more from internal capital markets. This implies that the theory proposed by Le¤ (1978) explains, at lest in part, the incentives to
2
form business groups. The bene…ts associated with the group a¢ liation may vary with the …rm’s life cycle. Younger …rms can bene…t more from the group internal capital market as they are more likely to rely on external funding than older …rms (Rajan and Zingales (1998)). In addition, younger …rms are likely to have a higher degree of asymmetric information with outside investors as they have a shorter track record (e.g. Ritter (1984), Oliner and Rudebusch (1992), Gompers (1995)). Indeed, we …nd that the e¤ect of …nancial development on group a¢ liation is signi…cantly stronger for younger …rms. While we hypothesize that the bene…t from the group internal capital market is higher in countries with less developed …nancial markets, the cost of group a¢ liation may vary across di¤erent legal environments. If, for example, countries with higher …nancial development have tax rules that disincentivize the formation of business groups (e.g. inter-company dividend tax), then our results can be cost-driven. Hence, we investigate the tax regulation in Europe that may a¤ect group a¢ liation. We …nd that rules regarding inter-company dividend tax and consolidated tax returns a¤ect group a¢ liation, but our results are highly signi…cant even after controlling for these tax rules. This paper also contributes to the debate on whether business groups should be dismantled. Morck (2005) claims that business groups in the US were disentangled in the 1930’s by the Roosevelt administration since the administration believed that business groups facilitate governance problems, tax avoidance, market power, and dangerously concentrate political in‡uence, while not providing the economic stability they claim to provide. Almeida and Wolfenzon (2006b) derive a theoretical model that predicts a negative correlation between business groups and …nancial development. Yet, the authors focus mainly on the negative e¤ect that business groups may have on …nancial development. Our results show that business groups also have a bright side at the country level. We …nd evidence supporting Le¤’s (1978) argument that business groups …ll the voids of imperfect markets, as the group internal capital market o¤ers an alternative for imperfect debt and equity markets.
3
The rest of the paper is organized as follows: Section 2 describes the data, Section 3 discusses the theoretical motivation and empirical approach, Section 4 presents the results, Section 5 discusses the policy implications of our results, and Section 6 concludes.
II.
Data and Descriptive Statistics A.
Data
Our newly developed dataset relies on ownership structure and accounting information from Amadeus - a comprehensive pan-European database by Bureau van Dijk Electronic Publishing (BvDEP), which covers both private and public …rms. In this section, we explain our methodology for constructing the data and describe our sample. We de…ne business groups as an organizational form in which at least two legally independent …rms are controlled by the same ultimate owner (Almeida and Wolfenzon (2006a)). In order to fully characterize groups, we determine group a¢ liation status for all …rms in the Amadeus database. For this purpose, we refer to the Amadeus ownership database, which includes detailed information on direct ownership links between European …rms. To ensure all ownership links represent control, we make the following assumptions: for private subsidiaries, we keep only links where the shareholder has at least 50 percent of the voting rights and for public …rms, we keep only links where the shareholder has at least 20 percent of the voting rights.5 These two assumptions leave us with close to one million ownership links. In order to infer group structure from these links, we apply an algorithm that constructs the corporate control chains, and then group together …rms controlled by the same ultimate owner.6 Accounting information is taken from the Amadeus …nancials section. BvDEP has developed a uniform format that maximizes the availability of …nancial items across the various countries’ …ling regulations, balanced with a realistic representation of European company accounts. A key advantage of these data is their large coverage of …rms and unique accounting information on private …rms across a wide size distribution. 4
Countries in our sample di¤er in their reporting requirements for small …rms. For example, small …rms in Great Britain are not obliged to disclose accounting information (including number of employees, sales, or total assets), while French …rms must provide such information regardless of their size. We mitigate the potential bias of voluntary disclosure of …rm …nancials by eliminating all …rms with less than 50 employees (Bloom and Van Reenen (2007)). To ensure that all …rms in our sample have ‘real’economic activity, we also eliminate …rms with less than one million dollars in annual sales. Finally, we drop state-owned …rms from the estimation sample because such …rms are likely to have di¤erent considerations for forming business groups than internal capital markets. We use the clean sample to apply our de…nitions of group a¢ liates and standalone …rms. Hence, a business group is de…ned as an organizational form in which at least two legally independent …rms with ‘real’ economic activity (i.e., …rms in our clean sample) are controlled by the same ultimate owner (individual, family, or widely-held …rm). All other …rms that have complete ownership information in Amadeus are classi…ed as standalones.7 This procedure leaves us with 82,446 …rms, out of which 42,484 …rms are a¢ liated with business groups. B.
Descriptive statistics
Panel A of Table I provides summary statistics for the …rms in our sample. The average …rm has 501 employees (a median of 114) and $132 million in sales (a median of $26 million). Panel B of Table I reports business groups characteristics. Our sample …rms belong to 12,652 di¤erent business groups. The average group has 3.4 a¢ liates, the median is 2 and the 90th percentile is 6 (where each a¢ liate must have at least 50 employees and report over $1 million in annual sales). Groups in our sample are resource abundant - the average group has nearly $1 billion in cash, compared to only $18 million for the average …rm (the distribution of cash ‡ow is highly skewed: the median business group has about 30 million in cash, where the
5
90th percentile group has more than $1.3 billion). [TABLE I ABOUT HERE] Table II reports summary statistics separately for a¢ liates and standalones. The main di¤erence between a¢ liates and standalones is that a¢ liates tend to be larger in terms of the number of employees, sales, …xed assets and total assets. Yet, there is no signi…cant di¤erence in their age, sales growth, and capital to employee ratio. [TABLE II ABOUT HERE] Table III reports the number of standalones and a¢ liates across the 15 countries in our sample. Germany, Great Britain, and France have the largest number of …rms that meet our inclusion criteria. On average, 52 percent of our sample …rms are a¢ liated with business groups. The percentage of a¢ liates is higher in Finland, Denmark, Great Britain, Netherland, and Sweden (around 70 percent in each) and lower in Greece, Austria, and Germany (around 30 percent in each). One has to be cautious when comparing these percentages with previous studies on business groups (e.g. La Porta at al. (1999), Faccio and Lang (2002)) since our sample includes private …rms, where previous studied focused on public …rms. We …nd a much higher percentage of group a¢ liates in Great Britain, for example, compared to previous studies. Finding stronger a¢ liation for private …rms as compared to public …rms may be related to the EU Takeover Directive. According to the directive, which applies only for securities that are traded on regulated markets of the member states (public …rms), minority shareholders have a sale-out right in case of a takeover. This makes control with partial ownership of public …rms more di¢ cult (or more costly) from the perspective of the group’s ultimate owner. Since most ultimate owners magnify their control using partial ownership, business groups are less attractive for listed …rms. [TABLE III ABOUT HERE]
6
III. A.
Empirical Modeling Strategy
Theoretical arguments and econometric approach
Our aim in this paper is to examine the e¤ect of …nancial development on group a¢ liation. More speci…cally, we ask whether business groups substitute under-developed …nancial institutions. Yet, reverse causality between group formation and …nancial development makes it di¢ cult to establish a causal relation. The main issue is that in the absence of developed …nancial markets, business groups may disincentivize the development of …nancial markets, especially if groups provide e¢ cient internal capital markets. To deal with the reverse causality issue, we analyze a key channel through which …nancial development a¤ects group a¢ liation –internal capital markets. If …rms form business groups to create a substitute for the underdeveloped …nancial markets, we would observe higher probability of group a¢ liation among …rms that have higher potential bene…t from the group internal capital market. A straightforward example of companies that would bene…t from the group internal capital market would be companies with higher demand for external …nance. Yet, most …rm-speci…c proxies for external …nancial dependence measure the external funding set in equilibrium rather than the demand for external funding and su¤er from endogeneity problems. Hence, we follow Rajan and Zingales (1998) methodology and rank industries according to the extent that they rely on external funds. We compute the industry measures based on US Compustat …rms. This has several important advantages: (i) Since the US market is one of the most advanced capital markets in the world, large publicly-traded …rms face the least frictions in accessing …nance. This means that the amount of external …nance used by these companies is likely to be a good measure of their demand for external …nance; (ii) Disclosure requirements imply that data on external …nancing are comprehensive; (iii) While using US industry data is rather exogenous to European …rms, it is likely that an industry’s dependence on external funds in the US is a good measure of its dependence in European countries; (iv) Business groups are not
7
common in the US (La Porta et al. (1999)). Thus, the demand of US …rms for external funds is potentially a good proxy for the demand in the absence of business group internal capital markets. The only two assumptions needed are that technological di¤erences explain why some industries rely on external funds more than others and that these di¤erences persist across countries. In addition to external dependence, we also rank industries according to their level of asymmetric information. Because asymmetric information between the …rm and the outside lender is an important determinant of the cost of capital, we expect external funds to be more expensive in industries with higher asymmetric information (Myers and Majluf (1984)). Under the assumption that asymmetric information is mitigated within a business group, since both the lender and the borrower are controlled by the same ultimate owner, we expect that group internal capital markets would be a more attractive source of …nancing in industries with high asymmetric information. Figure 3.1 illustrates our identi…cation strategy. For each industry we compute a measure of external dependence. Based on this measure we allocate industries into quartile (where the lowest quartile is for industries with the lowest measure of external dependence). Then, for each country we compute the di¤erence in the percentage of a¢ liates between the highest and lowest quartiles of external dependence. Countries are ranked according to their level of …nancial development (countries with higher …nancial development are closer to the origin). The emerging pattern supports the hypothesis that groups substitute under-developed …nancial markets. First, groups are more prominent in industries with high external dependence, supporting the internal capital markets hypothesis (on average, industries in the top quartile of external dependence have about 34 percent more a¢ liates than industries in the lowest external dependence quartile). Second, a¢ liation skewness toward industries with high external dependence varies substantially across countries, apparently systematically over the quality of …nancial institution.
8
[FIGURE 1 ABOUT HERE] B.
De…nition of variables
Financial development.— To measure the …nancial developments of the countries in our sample we refer to the world-bank indices developed by Beck, Demirgüç-Kunt and Levine (2000, 2007). For each country we examine two dimensions of …nancial development: stock market and banking system. For the stock market development, we use (i) the ratio between the stock market capitalization and GDP and (ii) ratio of stock market total value traded to GDP. For the banking system, we use (i) the ratio of private credit by deposit money banks and other …nancial institutions to GDP and (ii) the ratio of bank deposits to GDP, where bank deposits are the demand, time, and saving deposits in deposit money banks. Table IV presents the …nancial development measures across countries. It can be seen that while almost all countries in our sample are members of the European Union and have a relatively high per capita income, they di¤er signi…cantly in the development of their stock market and banking system which allow for identi…cation (for example, the ratio of stock market value traded to GDP in Great Britain is 1.9 and in Austria it is 0.09). [TABLE IV ABOUT HERE] External dependence.— We compute two measures of external dependence: External Financial Dependence and External Equity Dependence. The …rst measure, External Financial Dependence, is de…ned as the ratio between capital expenditures minus cash ‡ow from operations and capital expenditures. It measures the fraction of the …rm’s investment that cannot be …nanced through its internal cash ‡ows. The second measure, External Equity Dependence, is de…ned as the net amount of equity issued over capital expenditures. It measures the fraction of the …rm’s investment that is …nanced by issuing equity. The measures are calculated for the period 1980-2004 at the three-digit SIC level. Analyst disagreement.— Our measure of information asymmetry is analyst disagreement about earnings forecasts. Analyst disagreement has been widely used in the literature
9
as a measure of asymmetric information (Krishnaswami and Subramaniam (1999), Sadka and Scherbina (2007), and many others). The idea is that when asymmetric information is high, professional analysts covering a speci…c company are less likely to agree on the future earnings of that company. We obtain data on analyst forecasts from the Institutional Brokers’Estimate System (IBES). Following Diether, Malloy and Scherbina (2002) and Johnson (2004), we de…ne analyst disagreement as the ratio between the standard deviation of the forecasted earnings-per-share (EPS) and the mean of the earnings forecast. We consider all US listed stocks between 1990 and 2004 that have been followed by at least two analysts in the same month of the year. In ranking industries according to the measure of analyst disagreement, we compute the industry mean of this measure over the stocks in our sample. Table V presents key summary statistics for all our industry measures. The ratio of R&D to sales is computed using US Compustat data, because there is no information on R&D expenditures for European …rms. Lerner index, the ratio of capital to employees and labor productivity are calculated using data on British …rms.8 The external dependence measures are calculated based on Compustat …rms from 176 di¤erent industries (three-digit SIC level) in the period 1980-2004. Panel A shows the distribution of the industry variables and Panel B presents the correlation matrix.9 [TABLE V ABOUT HERE] C.
Econometric speci…cation
Our goal is to investigate the cross-sectional determinants of group a¢ liation using …rm level data. The dependent variable is a dummy that receives the value of one for a¢ liates and zero for standalones. We estimate the following Probit speci…cation:
Pr(a¢ liate i ) = (x0i
1
+
2 F inDevc
ExtDepj + 'j +
c)
(1)
Where, xi is a …rm-level controls vector, F inDevc is the …nancial development measure
10
for country c, ExtDepj is a measure of the external dependence or analyst disagreement for industry j, 'j and
c
denote complete sets of industry and country dummies, respectively.
A key advantage for using …rm-level data is that it allows us to include complete sets of country and industry dummies, which substantially mitigates potential biases arising from unobserved heterogeneity. For …rm level controls, we include number of employees and a dummy variable for family ownership. We control for size because larger …rms may have di¤erent external …nancing considerations than smaller ones. Our size measure is the number of employees.10 Family …rms may also be di¤erent than widely-held …rms in the way they react to …nancing considerations. For example, an important cost associated with joining a business group is the loss of control rights. Such cost is likely to be more important for family …rms than for widely-held …rms. We also test speci…cations when controlling for public listed …rms because public …rms are likely to have better access to external funds than private …rms. The public …rm dummy is not signi…cant; therefore, we exclude it from our main speci…cation (only 3 percent of the …rms in our sample are public). Under the hypothesis that business groups substitute for underdeveloped …nancial markets, we expect
2
to be negative. Intuitively,
2
< 0 means that in countries with lower
…nancial development, the di¤erence in group a¢ liation between …rms in industries with higher demand for external …nancing and lower demand for external …nancing is more pronounced. Similar intuition holds for our industry measures of asymmetric information. Unobserved heterogeneity at the industry level may downward-bias
2
even in the pres-
ence of complete industry and country dummies. To mitigate this concern we include a wide set of interactions between alternative industry measures with F inDev (such as, the Lerner index of competition and R&D intensity). Finally, due to the structure of our data, observations can be correlated across blocks of ultimate owners (as we observe multiple a¢ liates of the same business group). Therefore, we always cluster standard errors at the ultimate owner level.
11
IV. A.
Results
Main results
Table VI reports the main estimation results. We examine the interactions between the country …nancial development and the industry measures for external dependence and asymmetric information. Columns 1-6 report the estimation results for the stock market proxies of …nancial development. Consistent with our hypothesis that business groups substitute for underdeveloped …nancial markets, we …nd that the coe¢ cient on the interaction terms is negative and highly signi…cant in all speci…cations. Columns 7-12 report the estimation results for the banking system proxies of …nancial development. In accordance with our previous …ndings, the coe¢ cients on the interaction terms with private credit over GDP are negative and highly signi…cant. However, the interaction terms with bank deposits over GDP are negative, but insigni…cant.11 [TABLE VI ABOUT HERE] To quantify the e¤ect of …nancial development, we classify industries into four quartiles based on their value of external dependence. These quartiles are proxies for exogenous shift in …rms demand for external …nancing. Using the estimation results reported in column 1, we compute the predicted probability of a¢ liation for each …rm in our sample. We examine the di¤erence in a¢ liation probability across countries and industry quartiles. For Great Britain, the average predicted a¢ liation probability in the forth quartile (0.640) and in the lowest quartile (0.637) are almost identical. In other words, moving from the lowest quartile to the highest quartile hardly raises the probability of group a¢ liation for British …rms (only 0.5 percent di¤erence). For Germany, however, similar calculations show that moving from the lowest quartile to the highest quartile raises group a¢ liation by 15 percent. For France and Italy, the numbers are 6.8 percent and 11.1 percent, respectively. Our evidence suggests that since British …rms face highly developed …nancial markets, only few …rms form groups to bene…ts from the group internal capital market. Yet, German,
12
French and Italian …rms face a less developed …nancial market (especially stock market) and, therefore, rely on groups for …nancing. The possible policy implications of these results is that if groups were dismantled, an exogenous rise in demand for external funds (measured by the di¤erence between the highest and lowest quartile of external dependence) would leave 15 percent of German …rms, 7 percent of French …rms and 11 percent of Italian …rms in funding di¢ culties. These …rms, which previously relied on the business group internal capital market, would have to fund their operations using the …nancial markets. To the extent these …nancial markets are underdeveloped, investment would be threatened. B.
Firm’s life cycle
In this section, we test how our results vary with the …rm’s life cycle. There are three main e¤ects that may in‡uence the results. First, young …rms are typically more dependent on external …nancing than older …rms (Rajan and Zingales (1998)). Therefore, we would expect to …nd stronger e¤ect for younger …rms. Second, the age of the …rm is commonly used as proxy for the level of asymmetric information between the …rm and outsiders (e.g. Ritter (1984), Oliner and Rudebusch (1992), Gompers (1995)). Younger …rms have shorter track record by which they could be evaluated. Under the conjecture that asymmetric information is mitigated within a business group, the internal capital market bene…t associated with group a¢ liation should be greater for younger …rms. Finally, the …nancial development of a country may change over time. It is possible that older …rms formed business groups in times where the relative ranking of …nancial development was di¤erent from the one indicated by our measures. If dismantling a business group is costly for the ultimate owner,12 the results of a cross sectional estimation for older …rms may su¤er from attenuation bias. All three conjectures imply that if we split our sample into younger and older …rms, we should observe a stronger e¤ect for younger …rms. In Table VII we estimate our baseline speci…cation separately for young …rms (below our sample median of 17 years) and old …rms. The pattern of results is consistent with our hypotheses and suggests that the negative
13
interaction between the industry variables and the country …nancial development is more evident for young …rms than for older ones. [TABLE VII ABOUT HERE] C.
Legal environment
The cross-country variation in group a¢ liation may also be a¤ected by tax and other legal incentives or disincentives. While our empirical methodology which uses the withincountry between-industry variation mitigates this problem, it can still be the case that more …rms form business groups to bene…t from their internal capital markets in countries where the costs associated with groups are smaller. If these legal costs are positively correlated with …nancial development, we may be capturing the costs rather than …nancial development in our baseline speci…cation. We investigate potential legal costs by investigating the American tax reform of the 1930’s, which Morck (2005) describes as a highly successful strategy to disentangle business groups. The tax reform was initiated since the Roosevelt administration believed that business groups facilitate governance problems, tax avoidance, market power, and politically dangerous concentrated political in‡uence. In response, the American Congress enacted inter-corporate dividend taxes, all but abolished consolidated tax …ling for business groups, eliminated capital gains taxes on liquidated controlled subsidiaries, and banned large pyramidal groups from controlling public utilities companies. The explicit goal of these and other policies was to break up large US business groups. In this section, we examine whether similar tax laws exist in Europe and study their effect on our results. For this purpose, we explored the European laws regarding inter-company dividend tax, consolidated tax returns, and capital gains tax (Table VIII).13 Our primary data sources are the PWC Worldwide Tax Summaries and the EU Parent Subsidiary Directive. Detailed information about our data sources and the construction of these variables is provided in the appendix.
14
[TABLE VIII ABOUT HERE] Overall, there is no major variation in the tax and takeover rules for the countries in our sample since almost all of them (besides Norway and Switzerland) are members of the EU and are subject to the EU directives. Table IX presents the results of our estimation when including the legal variables in the regressions. Our measure for inter-company dividend tax receives a value of one in case of tax credit and zero in case of tax exemption. Our measure for consolidated tax return is the minimum holding percentage required to qualify for a consolidated tax return. Our measure for capital tax gains is the percentage of the total gain that is taxed. Thus, higher measures of inter-company dividend tax and consolidated tax returns should discourage the formation of business groups. Yet, it is not clear what should be the e¤ect of the capital gains tax. On the one hand, reducing the capital gains tax encourages the ultimate owner to realize her pro…ts and eliminates business groups. On the other hand, a lower capital gains tax can also encourage the formation of groups since the ultimate owner knows that there is a lower cost if she wants to disentangle the business group in the future. Consistent with our predictions, the coe¢ cients for the interactions of our industry measures with the inter-company dividend tax and the consolidated tax return measures are negative and signi…cant. The coe¢ cient on the capital gains tax measure is positive, but not always signi…cant. Importantly, when we include the legal measures together with the …nancial development measure, the coe¢ cient on the interaction of …nancial development with the industry measure is negative and signi…cant.14 This means that even when controlling for the cost of group a¢ liation associated with the legal environment, we still …nd that …nancial development negatively a¤ect group a¢ liation. [TABLE IX ABOUT HERE]
15
D.
Robustness checks
Alternative industry rankings.— In Table X, we control for alternative ranking of industries. We include …ve additional industry interaction terms: the Lerner competition index (computed as the industry mean of pro…ts over sales), R&D over sales, capital over employees, labor productivity, and …rm growth. The pattern of results reported in Table VI holds in these speci…cations as well. Table X also presents some evidence suggesting a negative interaction between …nancial development and R&D to sales. Such interaction means that group a¢ liation is skewed more towards R&D intensive industries when …nancial development is low. Thus, conditional on external dependence, the group internal capital markets are more important for …nancing innovation. A potential explanation of this …nding is as following. Innovation is typically associated with high asymmetric information between the entrepreneur and the funding agent. In case asymmetric information is mitigated within business groups, we would expect groups to play an important role in R&D …nancing, especially when outside markets are underdeveloped (Belenzon and Berkovitz (2007)). [TABLE X ABOUT HERE] Controlling for …rm characteristics.— Firms of di¤erent size may have di¤erent external …nancing needs. In case our sample of …rms systematically varies across countries, our results could be biased. For example, British …rms in our sample are larger than German and French Firms (783, 417 and 346 average number of employees, respectively). In case the business group internal capital markets are less signi…cant for larger …rms, this may lead to the observed pattern of results (because Great Britain has highly developed …nancial markets). To control for this potential bias, we extend our baseline speci…cation to include interaction terms between …rm characteristics (such as size) and industry measures. The estimation results are reported in Table XI. The same pattern of results of the negative interaction between external dependence and asymmetric information and …nancial development still holds. 16
[TABLE XI ABOUT HERE]
V.
Policy Implications
It has been postulated in the literature that business groups have diverse e¤ects on the economy (Morck at el. (2005) summarize some of these papers). On the one hand, business groups facilitate governance problems, tax avoidance, market power, and dangerously concentrate political in‡uence. On the other hand, as shown in this paper, business groups may …ll the voids of imperfect markets. Since business group internal capital markets have been shown to have a prominent e¤ect on …rm performance, policy makers may be reluctant to disentangle business groups. This raises an important question –what is the level of …nancial development at which the cost of forming a business group and managing its internal capital markets is equal to the bene…t from having a lower cost of capital? In which countries is the …nancial development high enough, so that the bene…t from internal capital market does not a¤ect …rms’a¢ liation decision? While we do not aim to fully address this question in our empirical framework, we try to obtain a rough estimate based on our results. Figure 1 shows the cross-country percentage di¤erence between the fraction of a¢ liated …rms in industries in the highest quartile and the lowest quartile in term of external …nancial dependence. Countries are ranked according to the …nancial development of their stock market (which has the most signi…cant e¤ect out of all …nancial development measures in our regressions). Figure 1 graphically illustrates the …nancial development threshold from which …rms do not seem to organize in business groups to exploit the bene…ts of the internal capital market. Namely, the point where we do not observe a higher fraction of group a¢ liates in industries that would theoretically bene…t more from having access to the group internal capital market. While in countries with high …nancial development (above the median ratio of Stock Market Value Traded to GDP) are spread rather equally across industries with high and low external dependence, in countries with low …nancial development we observe 17
a substantially higher fraction of group a¢ liates in industries that rely more on external …nancing. The median ratio of Stock Market Value Traded to GDP in our sample, which seems to be the threshold for the e¤ect of stock market development on group a¢ liation, is 0.65. For comparison, this ratio for the US in 2005 is 1.81, well above that threshold. This means that US policy makers, when considering whether to remove the tax barriers to formation of business groups, should take into account that access to internal capital markets would probably not a¤ect the behavior of these …rms, since the US stock market is highly developed. However, dismantling business groups in Italy, for example, where …nancial markets are less developed, might have diverse e¤ects on the economy. On the one hand, Italian …rms that rely on the business group internal capital market to …nance their investments may …nd it hard to raise funds. This may lead to less R&D (Belenzon and Berkovitz (2007)) and lower economic growth. On the other hand, the absence of business groups may stimulate the development of the …nancial markets.
VI.
Conclusion
This paper uses a comprehensive …rm-level database on group a¢ liation in 15 European countries to study the e¤ect of …nancial development on group a¢ liation. Our results indicate that less developed …nancial markets incentivize the formation of business groups. Since correlations from cross-country regressions are hard to interpret in the causal sense, we use exogenous industry measures to investigate the channel through which …nancial development a¤ects group a¢ liation. We …nd that countries with less developed …nancial markets have a disproportionately higher percentage of group a¢ liates in industries with high levels of external dependence and asymmetric information. This implies that …rms in less developed equity and debt markets join business groups to bene…t from their signi…cant internal capital markets. We examine how our results vary with the …rm’s life cycle and legal environment. We 18
…nd that the decision of younger …rms, which typically rely more on external …nding and have higher degree of asymmetric information, to join business groups is a¤ected more by …nancial development compared to mature …rms. This is consistent with our hypothesis that the group internal capital market substitutes for less developed …nancial markets. We also …nd that while the legal environment a¤ects the incentives of …rms to join business groups, including it in our estimations does not a¤ect the robustness of our results. In this paper, we focus on the e¤ect of …nancial development on the incentive to form business groups. Yet, the development of other markets can a¤ect these incentives as well. Future research can use the methodology and data introduced in this paper to study the e¤ect of labor or product market development on group a¢ liation. If, for example, the labor market is underdeveloped, …rms may join business groups in order to gain access to their internal labor market. In addition, while the e¤ect of group internal capital markets on …rm performance has been well documented, future research may investigate the e¤ect of other group internal markets on …rm performance.
19
REFERENCES Almeida, Heitor, and Daniel Wolfenzon, 2006a, A Theory of Pyramidal Ownership and Family Business Groups, Journal of Finance 61, 2637-2680. Almeida, Heitor, and Daniel Wolfenzon, 2006b, Should Business Groups be Dismantled? The Equilibrium Costs of E¢ cient Internal Capital Markets, Journal of Financial Economics 79, 99-144. Beck, Thorsten, Asli Demirgüç-Kunt, and Ross Levine, 2000, A New Database on Financial Development and Structure, World Bank Economic Review 14, 597-605 (updated in 2007). Belenzon, Sharon, and Tomer Berkovitz, 2007, Innovation in Business Groups, Working Paper. Bloom, Nicholas, and John Van Reenen, 2007, Measuring and Explaining Management Practices Across Firms and Countries, The Quarterly Journal of Economics 122(4), 1351-1408. Claessens, Stijn, Simeon Djankov, and Larry H.P. Lang, 2000, The separation of ownership and control in East Asian Corporations, Journal of Financial Economics 58, 81-112. Diether, Karl, Christopher Malloy, and Anna Scherbina, 2002, Di¤erences of Opinion and the Cross-Section of Stock Returns, Journal of Finance 57, 2113–2141. Faccio, Mara, and Larry H.P. Lang, 2002, The Ultimate Ownership of Western European Corporations, Journal of Financial Economics 65, 365-395. Gompers, Paul A., 1995, Optimal Investment, Monitoring, and the Staging of Venture Capital, Journal of Finance 50(5), 1461-1489. Gopalan, Radha, Vikram Nanda, and Amit Seru, 2007, A¢ liated Firms and Financial Support: Evidence from Indian Business Groups, Journal of Financial Economics 86(3), 759795. 20
Hoshi, Takeo, Anil Kashyap, and David Scharfstein, 1991, Corporate Structure, Liquidity, and Investment: Evidence from Japanese Industrial Groups, The Quarterly Journal of Economics 106(1), 33-60. Johnson, Timothy C., 2004, Forecast Dispersion and the Cross-Section of Expected Returns, Journal of Finance 59, 1957–1978. Khanna, Tarun, and Krishna Palepu, 2000, Is Group A¢ liation Pro…table in Emerging Markets? An Analysis of Diversi…ed Indian Business Groups, Journal of Finance 55(2), 867-891. Khanna, Tarun, and Jan W. Rivkin, 2001, Estimating the Performance of Business Groups in Emerging Markets, Strategic Management Journal 22, 45–74. Khanna, Tarun, and Yishay Yafeh, 2007, Business Groups in Emerging Markets: Paragons or Parasites?, Journal of Economic Literature 45, 331-372. Krishnaswami, Sudha, and Venkat Subramaniam, 1999, Information Asymmetry, Valuation, and the Corporate Spin-O¤ Decision, Journal of Financial Economics 53, 73–112. La Porta, Rafael, Florencio Lopez-de-Silanes, and Andrei Shleifer, 1999, Corporate Ownership around the World, Journal of Finance 54(2), 471-517. Le¤, Nathaniel H., 1978, Industrial Organization and Entrepreneurship in the Developing Countries: The Economic Groups, Economic Development and Cultural Change 26, 661–675. Morck, Randall, 2005, How to Eliminate Pyramidal Business Groups: The Double-Taxation of Intercorporate Dividends and Other Incisive Uses of Tax Policy, Tax Policy and the Economy 19, 135-179. Morck, Randall, Daniel Wolfenzon and Bernard Yeung, 2005, Corporate Governance, Economic Entrenchment, and Growth, Journal of Economic Literature 43(3), 655-720.
21
Myers, Stewart and Nicholal S. Majluf, 1984, Corporate Financing and Investment Decisions when Firms have Information that Investors Do Not Have, Journal of Financial Economics 13, 187-221. Oliner, Stephen D., and Glenn D. Rudebusch, 1992, Sources of the Financing Hierarchy for Business Investment, The Review of Economics and Statistics 74(4), 643-654. Rajan, Raghuram, and Luigi Zingales, 1998, Financial Dependence and Growth, American Economic Review 88, 559-586. Ritter, Jay R., 1984, The "hot issue" market of 1980, Journal of Business 57, 215-241. Sadka, Ronnie, and Anna Scherbina, 2007, Analyst Disagreement, Mispricing, and Liquidity, Journal of Finance 62(5), 2367-2403.
22
Financial Development
Ita ly G er m an y N or w ay D en m ar k
G re e ce
Notes: This figure describes the difference in the percentage of affiliates between the highest and lowest quartiles of external financial dependence across countries. Countries are ranked according to their financial development in ascending order. Financial development is based on the Beck et al. (2000, 2007), and is defined as: Stock Market Value Traded / GDP, averaged for 2003-2005. Stock Market Value Traded is total shares traded on the stock market exchange. External Finance Dependence is computed at the three-digit SIC level based on Compustat firms in the period 1980-2004, and is defined as the ratio between capital expenditures minus cash flow from operations and capital expenditures.
-0.1
-0.05
Au s tri a Be l gi um
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
Ir
0.45
nd el a
diff in % of affiliated firms
Fi n l an d Sw e d en Sw i tz e lr a nd
Figure 1: Di¤erences in Group A¢ liation between Industries with Low and High External Financial Dependence across Countries
F
ra n
ce N
nd rla he et
B r ita in G re at
23
Table I - Summary Statistics for Firms and Business Groups PANEL A: FIRM-LEVEL CHARACTERISTICS Variable
# firms
Mean
Std. Dev.
10
Distribution th 50
Sales (`000)
82,446
132,446
1,392,561
7,083
26,330
158,080
Employess
82,446
501
4,562
55
114
600
Total Assets (`000)
60,083
170,025
2,224,822
4,360
19,132
163,032
Age
82,446
24
22
4
17
49
Sales Growth
82,446
0.23
0.32
-0.04
0.14
0.30
Capital (`000)
60,048
85,931
1,159,144
464
4,782
66,355
Capital/Employee (`000)
60,048
118
1,723
5
34
180
Sales/Employee (`000)
82,446
257
186
81
196
525
Cash Flow ('000)
58,446
18,426
775,324
-259
1,331
14,224
90
th
th
90
PANEL B: BUSINESS GROUP CHARACTERISTICS # groups
Mean
Std. Dev.
10
Distribution th 50
# of affiliates
12,652
3.4
5
2
2
6
Sales ('000)
12,652
734,221
4,512,623
20,975
104,043
962,398
Employees
12,652
2,691
14,798
119
438
3,660
Total Assets (000')
10,948
840,202
6,177,487
12,164
70,193
926,185
Capital (millions)
10,916
427,571
3,348,654
2,008
22,174
411,456
Cash flow (`000)
9,760
104,553
2,632,008
785
6,253
86,339
Variable
st
th
Notes: These tables provide summary statistics both at the firm level and at the group level for our sample. The cross-sectional data is based on the 2005 ownership structure and 2004 financials (consistent with our regression estimates). Firms are included in the sample if they report at least one million dollars in annual sales and have at least 50 employees. Business group is defined as an ownership form in which at least two legally independent firms are controlled by the same ultimate owner. Panel A provides information on key firm characteristics. Capital is defined as fixed-assets and cash flow is defined as net income plus depreciation. Age is the number of years since the date of incorporation. Panel B provides information on key business group characteristics.
24
Table II - Firm Characteristics: A¢ liates versus Standalones PANEL A: AFFILIATES CHARACTERISTICS Variable
# firms
Mean
Std. Dev.
10
Distribution th 50
90
Sales (`000)
42,484
218,655
1,926,761
9,807
40,097
274,962
Employess
42,484
801
6,283
61
158
1,005
Total Assets (`000)
35,737
257,395
2,875,696
5,537
26,672
257,784
Age
42,484
25
24
4
17
56
Sales Growth
42,484
0.22
0.32
-0.05
0.15
0.29
Capital (`000)
35,709
130,706
1,496,398
511
6,712
105,871
Capital/Employee (`000)
35,709
137
2,212
4
36
203
Sales/Employee (`000)
42,484
284
196
90
226
574
Cash Flow ('000)
34,676
28,719
1,005,771
-583
1,878
22,234
90
th
th
PANEL B: STANDALONES CHARACTERISTICS Variable
# firms
Mean
Std. Dev.
10
Distribution th 50
Sales (`000)
39,962
40,795
194,628
5,894
17,705
71,479
Employess
39,962
182
876
53
90
286
Total Assets (`000)
24,346
41,777
221,958
3,452
12,535
66,834
Age
39,962
22
19
4
17
44
Sales Growth
39,962
0.24
0.32
-0.02
0.14
0.31
Capital (`000)
24,339
20,239
149,758
424
3,274
26,489
Capital/Employee (`000)
24,339
91
3,996
5
31
148
Sales/Employee (`000)
39,962
228
169
73
173
459
Cash Flow ('000)
23,770
3,409
44,673
-15
902
6,184
th
th
Notes: These tables provide summary statistics for group affiliates (Panel A) and standalones (Panel B) in our sample. The cross-sectional data is based on the 2005 ownership structure and 2004 financials (consistent with our regression estimates). Firms are included in the sample if they report at least one million dollars in annual sales and have at least 50 employees. Business group is defined as an ownership form in which at least two legally independent firms are controlled by the same ultimate owner. Capital is defined as fixed-assets and cash flow is defined as net income plus depreciation. Age is the number of years since the date of incorporation.
25
Table III - A¢ liates and Standalones across Countries Country Austria Belgium Germany Greece Denmark Ireland Finland France Great Britain Italy Netherland Norway Spain Sweden Switzerland Total
Stand Alone 533
Affiliates 209
(72%)
(28%)
805
1,465
(35%)
(65%)
15,798
7,380
(68%)
(32%)
1,257
408
(75%)
(25%)
493
1,032
(32%)
(68%)
142
82
(63%)
(37%)
360
797
(31%)
(69%)
6,296
8,554
(42%)
(58%)
5,941
12,583
(32%)
(68%)
1,025
1,236
(45%)
(55%)
634
1,288
(33%)
(67%)
873
604
(59%)
(41%)
4,555
3,844
(54%)
(46%)
870
2,443
(26%)
(74%)
380
559
(40%)
(60%)
39,962
42,484
(48%)
(52%)
Total 742 2,270 23,178 1,665 1,525 224 1,157 14,850 18,524 2,261 1,922 1,477 8,399 3,313 939 82,446
Notes: This table presents the number of group affiliates versus standalone firms across countries in our sample. The ownership structure is for 2005. Firms are included in the sample if they report at least one million dollars in annual sales and have at least 50 employees. Business group is defined as an ownership form in which at least two legally independent firms are controlled by the same ultimate owner.
26
27
1.117 0.791 1.675 0.712 0.959 1.229 1.605 0.788 1.342 0.925 1.734 0.835 1.255 1.117 1.709 1.186
0.834 0.952 0.513 0.483 0.674 0.971 1.155 0.811 0.811 0.529 1.059 0.517 0.835 0.436 1.344 0.795
Stock Market Value Traded / GDP 0.089 0.219 0.717 1.288 0.717 0.569 1.898 0.235 0.264 0.527 1.146 0.591 1.315 1.304 1.304 0.812
Stock Market Capitalization / GDP 0.249 1.377 0.593 0.978 0.830 0.433 1.262 0.590 0.558 0.434 0.994 0.486 0.868 0.992 2.233 0.858
Notes: This table presents measures of financial developments for the countries in our sample. All measures are based on Beck et al. (2000, 2007), averaged for 2003-2005. Private Credit is the private credit by deposit money banks and other financial institutions. Bank Deposits are the demand, time and saving deposits in deposit money banks. Stock Market Value Traded is total shares traded on the stock market exchange. Stock Market Capitalization is the value of listed shares. All measures all divided by the Gross Domestic Product (GDP).
Country Austria Belgium Denemark Finland France Germany Great Britain Greece Ireland Italy Netherland Norway Spain Sweden Switzerland Avergae
Private Credit / GDP Bank Deposits / GDP
Table IV - Main Country Characteristics
28
176 176 176 176 176 176 176
External Equity Dependence
Analyst Disagreement
Lerner Index
R&D/Sales
Capital/Employees
Labour Productivity
Labour Productivity Growth
0.65 0.34 -0.06 0.14 -0.02 -0.07 -0.09
External Equity Dependence
Analyst Disagreement
Lerner Index
R&D/Sales
Capital/Employees
Labour Productivity
Labour Productivity Growth
0.030
142
39
2.42
0.04
0.72
0.92
0.59
Std. Dev.
-0.04
-0.01
-0.18
0.08
0.02
0.43
1.00
External Equity Dependence
0.15
0.04
-0.01
0.19
0.11
1.00
Analyst Disagreement
st
-0.015
82
7
0
0.03
0.39
0.26
0.50
10
0.04
-0.36
0.49
-0.03
1.00
Lerner Index
0.03
-0.04
0.02
1.00
R&D/Sales
0.014
143
17
0.10
0.05
1.17
1.00
0.87
Distribution th 50
0.16
0.07
1.00
Capital/ Employees
0.047
357
40
0.41
0.12
1.93
2.29
1.62
th
90
-0.11
1.00
Labour Productivity
1.00
Labour Productivity Growth
Notes: These tables report the summary statistics of the key industry variables. External dependence measures and R&D/Sales are computed at the three-digit SIC level based on Compustat firms in the period 1980-2004. Analyst disagreement is computed for all US listed firms in the period 1990-2004 that appear in IBES and are followed by at least two analysts. Lerner Index, Capital/Employees, and productivity measures are computed for British firms in Amadeus. External Finance Dependence is defined as the ratio between capital expenditures minus cash flow from operations and capital expenditures. External Equity Dependence is defined as the net amount of equity issued over capital expenditures. Analyst Disagreement is the ratio between the standard deviation and the mean of the forecasted EPS. Panel A describes the distribution of the industry variables and Panel B shows the correlation matrix between the variables.
1.00
External Finance Dependence
0.015
195
25
0.33
0.07
1.18
1.21
1.01
Mean
PANEL B: CORRELATIONS MATRIX External Finance Dependence
176
# industries
External Finance Dependence
Variable
PANEL A: DISTRIBUTION OF VARIABLES
Table V - Summary Statistics for Industry Variables
29 82,446
0.169 82,446
0.170
Yes
(4)
82,446
0.169
Yes
Yes
(0.026)
0.587***
(0.011)
0.075***
(0.015)
-0.075***
Analyst disagreement
(5)
(6)
82,446
0.170
Yes
Yes
(0.026)
-0.580***
(0.011)
0.076***
(0.024)
-0.122***
External Funds
82,446
0.170
Yes
Yes
(0.026)
-0.580***
(0.011)
0.075***
(0.017)
-0.101***
External Equity
82,446
0.169
Yes
Yes
(0.026)
-0.586***
(0.011)
0.075***
(0.021)
-0.106***
Analyst disagreement
Stock Market Capitalization / GDP
Stock market
(3)
(0.026)
-0.581***
(0.011)
0.076***
(0.030)
-0.086***
82,446
0.169
Yes
Yes
(8)
82,446
0.169
Yes
Yes
(0.026)
-0.581***
(0.011)
0.076***
(0.022)
-0.077***
External Equity
(9)
82,446
0.169
Yes
Yes
(0.026)
-0.587***
(0.010)
0.075***
(0.029)
-0.115***
(11)
(12)
82,446
0.169
Yes
Yes
(0.026)
-0.581***
(0.011)
0.076***
(0.037)
-0.060*
82,446
0.169
Yes
Yes
(0.026)
-0.581***
(0.011)
0.076***
(0.026)
-0.044*
External Equity
82,446
0.169
Yes
Yes
(0.026)
-0.587***
(0.011)
0.075***
(0.036)
-0.083**
Analyst disagreement
Bank Deposits / GDP
(10)
External Funds
Banking
Analyst disagreement
Private Credit / GDP External Funds
(7)
Notes: This table reports the results of Probit regressions that examine the effect of financial development on business group affiliation. The dependent variable is a group affiliation dummy that equals one for firms that are affiliates to groups and equals zero for standalones. The cross-sectional estimation is based on the 2005 ownership structure and 2004 financials. The measures of financial development are based on Beck et al. (2000, 2007), averaged for 2003-2005. Private Credit is the private credit by deposit money banks and other financial institutions. Bank Deposits is the demand, time and saving deposits in deposit money banks. Stock Market Value Traded is total shares traded on the stock market exchange. Stock Market Capitalization is the value of listed shares. External dependence measures are computed at the three-digit SIC level based on Compustat firms in the period 1980-2004. Analyst disagreement is computed for all US listed firms in the period 1990-2004 that appear in IBES and are followed by at least two analysts. External Finance Dependence is defined as the ratio between capital expenditures minus cash flow from operations and capital expenditures. External Equity Dependence is defined as the net amount of equity issued over capital expenditures. Analyst Disagreement is the ratio between the standard deviation and the mean of the forecasted EPS. The dummy for family ownership equals one when the ultimate owner of the firm is a family or an individual and zero otherwise (widely-held). Country and industry fixed effects are included in all regressions. Standard errors (in brackets) are robust to arbitrary heteroskedasticity and allow for serial correlation through clustering by ultimate owner. * significant at 10%; ** significant at 5%; *** significant at 1%.
Observations
2
Yes
Three-digit SIC dummies (176)
(0.026)
Yes
-0.581***
(0.026)
(0.011)
-0.581***
0.075***
(0.011)
(0.012)
0.076***
-0.058***
(0.017)
External Equity
-0.064***
External Funds
Yes
Psedu R
(2)
Stock Market Value Traded / GDP
Country dummies (15)
Dummy for family ownership
Employees ('000)
Financial Development × Industry measure
Industry measure:
Financial Development:
(1)
Dependent variable:Dummy for Group Affiliation. Probit estimation
Table 1: Table VI - Financial Development and Group A¢ liation
30
2
40,474
0.171 40,474
0.171
Yes
(4)
40,474
0.170
Yes
Yes
(0.029)
-0.581***
(0.023)
0.106***
(0.020)
-0.110***
Analyst disagreement
41,972
0.184
Yes
Yes
(0.029)
-0.608***
(0.011)
0.063***
(0.024)
-0.024
External Funds
Stock market
(3)
41,972
0.184
Yes
Yes
(0.029)
-0.609***
(0.011)
0.063***
(0.017)
-0.043***
External Equity
Mature Firms
(5)
41,972
0.184
Yes
Yes
(0.029)
-0.613***
(0.011)
0.063***
(0.021)
-0.028
Analyst disagreement
(6)
40,474
0.170
Yes
Yes
(0.029)
-0.575***
(0.023)
0.106***
(0.039)
-0.174***
External Funds
(7)
40,474
0.170
Yes
Yes
(0.029)
-0.575***
(0.023)
0.106***
(0.028)
-0.113***
External Equity
Young Firms
(8)
40,474
0.169
Yes
Yes
(0.029)
-0.581***
(0.023)
0.106***
(0.036)
-0.145***
(10)
41,972
0.184
Yes
Yes
(0.029)
-0.608***
(0.011)
0.063***
(0.047)
-0.002
External Funds
Banking
Analyst disagreement
(9)
41,972
0.184
Yes
Yes
(0.029)
-0.575***
(0.011)
0.063***
(0.030)
-0.051*
External Equity
Mature Firms
(11)
41,972
0.184
Yes
Yes
(0.029)
-0.613***
(0.011)
0.063***
(0.041)
-0.071*
Analyst disagreement
(12)
Notes: These tables report the results of Probit regressions that examine the effect of financial development on business group affiliation for young (below the age median) and mature firms. Age is defined as the number of years since the date of incorporation. The dependent variable is a group affiliation dummy that equals one for firms that are affiliates to groups and equals zero for standalones. The cross-sectional estimation is based on the 2005 ownership structure and 2004 financials. The measures of financial development are based on Beck et al. (2000, 2007), averaged for 2003-2005. The financial development measure for the stock market is Stock Market Value Traded divided by GDP and for the banking industry it is Private Credit divided by GDP. Stock Market Value Traded is total shares traded on the stock market exchange. Private Credit is the private credit by deposit money banks and other financial institutions. External dependence measures are computed at the three-digit SIC level based on Compustat firms in the period 1980-2004. Analyst disagreement is computed for all US listed firms in the period 1990-2004 that appear in IBES and are followed by at least two analysts. External Finance Dependence is defined as the ratio between capital expenditures minus cash flow from operations and capital expenditures. External Equity Dependence is defined as the net amount of equity issued over capital expenditures. Analyst Disagreement is the ratio between the standard deviation and the mean of the forecasted EPS. The dummy for family ownership equals one when the ultimate owner of the firm is a family or an individual and zero otherwise (widely-held). Country and industry fixed effects are included in all regressions. Standard errors (in brackets) are robust to arbitrary heteroskedasticity and allow for serial correlation through clustering by ultimate owner. * significant at 10%; ** significant at 5%; *** significant at 1%.
Observations
Psedu R
Yes
Three-digit SIC dummies (176)
Yes
(0.029)
Yes
-0.574***
(0.029)
(0.023)
-0.574***
0.106***
(0.023)
(0.016)
0.106***
-0.076***
(0.023)
External Equity
Young Firms
(2)
-0.105***
External Funds
Country dummies (15)
Dummy for family ownership
Employees ('000)
Financial Development × Industry measure
Industry measure:
Financial Development:
(1)
Dependent variable:Dummy for Group Affiliation. Probit estimation
Table VII - Financial Development and Group A¢ liation: Firm’s Life Cycle
Table VIII - Inter-Company Dividend Tax, Capital Gains Tax and Consolidated Tax Returns
Full Exemption Austria Denmark Finland Netherlands Sweden Spain
PANEL A: INTER-COMPANY DIVIDEND TAX Exemption & 5% management fee Belgium France Italy Germany
Tax Credit Greece Ireland UK
PANEL B: MINIMUM HOLDINGS FOR CONSOLIDATED TAX RETURNS No Consolidation 95% 90% 75% 50% Belgium France Finland Ireland Austria Greece Netherlands Sweden UK Germany Spain Italy Denmark PANEL C: CAPITAL GAINS TAX FROM SELLING A SUBSIDIARY Full exemption Austria Finland Ireland Netherlands Belgium Denmark Luxemburg Portugal Sweden
95% exemption Germany
91% exemption Italy
Lower than corporate tax France
No exemption Greece Spain UK
Notes: These tables present the 2006 rules regarding inter-company dividend tax (Panel A), consolidated tax returns (Panel B), and capital gains tax (Panel C) in the EU. Our primary data sources are the PWC Worldwide Tax Summaries, AGN Europe Tax Surveys, Global Legal Group (GLG) International Comparative Legal Guide to Corporate Tax 2006, EU Parent Subsidiary Directive (90/435/EEC), and various European Commission working papers on corporate tax.
31
32
2
81,193
0.169 81,193
0.168
Yes
81,193
0.169
Yes
Yes
(0.026)
-0.584***
(0.011)
0.076***
(0.024)
-0.066***
(0.035)
-0.097***
External Funds
81,193
0.168
Yes
Yes
(0.027)
-0.589***
(0.011)
0.075***
(0.018)
-0.037**
(0.033)
-0.122***
Analyst disagreement
Banking
(4)
81,193
0.169
Yes
Yes
(0.026)
-0.582***
(0.011)
0.076***
(0.055)
-0.156***
(0.029)
-0.153***
External Funds
(7)
81,193
0.168
Yes
Yes
(0.026)
-0.589***
(0.011)
0.075***
(0.048)
-0.078*
(0.025)
-0.135***
Analyst disagreement
81,193
0.168
Yes
Yes
(0.026)
-0.583***
(0.011)
0.076***
(0.055)
-0.247***
(0.046)
-0.020
External Funds
(8)
81,193
0.168
Yes
Yes
(0.026)
-0.589***
(0.011)
0.075***
(0.048)
-0.183***
(0.043)
-0.064***
Analyst disagreement
Banking
Consolidated Tax Returns
(6)
Stock market
(5)
81,193
0.169
Yes
Yes
(0.026)
-0.583***
(0.011)
0.076***
(0.033)
0.093***
(0.028)
-0.125***
External Funds
(11)
81,193
0.168
Yes
Yes
(0.027)
-0.589***
(0.011)
0.075***
(0.029)
0.073***
(0.025)
-0.125***
Analyst disagreement
(0.026)
-0.583***
(0.011)
0.076***
(0.024)
0.011
(0.039)
-0.096***
81,193
0.168
Yes
Yes
(12)
81,193
0.168
Yes
Yes
(0.027)
-0.589***
(0.011)
0.075***
(0.021)
-0.003
(0.036)
-0.117***
Analyst disagreement
Banking External Funds
Capital Gains Tax
(10)
Stock market
(9)
Notes: This table reports the results of Probit regressions that examine the effect of financial development on business group affiliation when controlling for the countries’tax rules. The dependent variable is a group affiliation dummy that equals one for firms that are affiliates to groups and equals zero for standalones. The financial development measure for the stock market is Stock Market Value Traded divided by GDP and for the banking industry it is Private Credit divided by GDP. Stock Market Value Traded is total shares traded on the stock market exchange. Private Credit is the private credit by deposit money banks and other financial institutions. External Finance Dependence is defined as the ratio between capital expenditures minus cash flow from operations and capital expenditures. Analyst Disagreement is the ratio between the standard deviation and the mean of the forecasted EPS. Inter-Company Dividend Tax is equal to zero in case of tax exemption and equals one for tax credit. Consolidation is the minimum holdings required to qualify for consolidated tax return. Capital Gains Tax is percentage on the capital gains that is taxed in case of selling a subsidiary. The dummy for family ownership equals one when the ultimate owner of the firm is a family or an individual and zero otherwise (widely-held). Country and industry fixed effects are included in all regressions. Standard errors (in brackets) are robust to arbitrary heteroskedasticity and allow for serial correlation through clustering by ultimate owner. * significant at 10%; ** significant at 5%; *** significant at 1%.
Observations
Psedu R
Yes
Three-digit SIC dummies (176)
Yes
(0.027)
Yes
-0.589***
(0.026)
(0.011)
-0.584***
0.075***
(0.011)
(0.020)
0.076***
-0.049***
(0.024)
(0.017)
-0.073***
-0.084***
(0.019)
Analyst disagreement
-0.073***
External Funds
(3)
Inter Company Dividend Tax
(2)
Stock market
Country dummies (12)
Dummy for family ownership
Employees ('000)
Legal Measure × Industry measure
Financial Development × Industry measure
Industry measure:
Financial Development:
Legal Measure:
(1)
Dependent variable: Dummy for Group Affiliation. Probit estimation
Table IX - The Cost of Group A¢ liation: The E¤ect of the Legal Environment
33
2
82,446
82,446
0.170
(0.026)
0.170
-0.581***
(0.026)
(0.011)
-0.581***
0.075***
(0.011)
(0.619)
0.076***
1.069*
(0.616)
(0.029)
0.825
0.016
(0.029)
(0.023)
-0.006
-0.009
(0.021)
(0.003)
0.022
0.007**
(0.003)
(0.334)
-0.008
0.087
(0.334)
(0.013)
-0.257
-0.055***
(0.018)
External Equity
-0.056***
External Funds
(4)
82,446
0.169
(0.026)
-0.587***
(0.011)
0.075***
(0.625)
1.094
(0.029)
-0.004
(0.022)
0.009
(0.003)
-0.006**
(0.332)
-0.137
(0.016)
-0.069***
Analyst disagreement
(5)
(6)
82,446
0.170
(0.026)
-0.580***
(0.011)
0.076***
(0.878)
1.114
(0.029)
-0.019
(0.029)
0.013
(0.005)
-0.009**
(0.473)
-0.806*
(0.025)
-0.119***
External Funds
82,446
0.170
(0.026)
-0.580***
(0.011)
0.075***
(0.885)
1.621*
(0.040)
0.024
(0.031)
-0.048
(0.005)
-0.008*
(0.469)
-0.132
(0.018)
-0.108***
External Equity
82,446
0.170
(0.026)
0.856***
(0.011)
0.075***
(0.889)
1.454
(0.040)
-0.014
(0.029)
-0.002
(0.004)
-0.009**
(0.469)
-0.557
(0.022)
-0.102***
Analyst disagreement
Stock Market Capitalization / GDP
Stock market
(3)
Stock Market Value Traded / GDP
(2)
(8)
82,446
0.170
(0.026)
-0.581***
(0.011)
0.076***
(1.159)
-0.001
(0.057)
-0.067
(0.041)
0.107***
(0.006)
-0.009
(0.629)
-0.464
(0.032)
-0.074**
82,446
0.170
(0.026)
-0.581***
(0.011)
0.076***
(1.164)
0.231
(0.058)
-0.045
(0.045)
0.078*
(0.006)
-0.010
(0.629)
-0.107
(0.024)
-0.052**
External Equity
(9)
82,446
0.170
(0.026)
-0.587***
(0.011)
0.075***
(1.178)
0.426
(0.057)
-0.061
(0.043)
0.082**
(0.006)
-0.007
(0.624)
-0.346
(0.029)
-0.106***
(11)
(12)
82,446
0.169
(0.026)
-0.580***
(0.011)
0.076***
1.808 (1.447
(0.072)
-0.075
(0.049)
0.113**
(0.008)
-0.010
(0.755)
-0.096
(0.040)
-0.045
82,446
0.169
(0.026)
-0.580***
(0.011)
0.076***
(1.395)
1.732
(0.070)
-0.068
(0.052)
0.107**
(0.008)
-0.011
(0.727)
-0.089
(0.029)
-0.008
External Equity
82,446
0.170
(0.026)
-0.587***
(0.011)
0.075***
(1.463)
1.922
(0.073)
-0.071
(0.052)
0.094*
(0.008)
-0.009
(0.747)
-0.020
(0.373)
-0.072**
Analyst disagreement
Bank Deposits / GDP
(10)
External Funds
Banking
Analyst disagreement
Private Credit / GDP External Funds
(7)
Notes: This table reports the results of Probit regressions that examine the effect of financial development on business group affiliation when controlling for various industry characteristics. The dependent variable is a group affiliation dummy that equals one for firms that are affiliates to groups and equals zero for standalones. Private Credit is the private credit by deposit money banks and other financial institutions. Bank Deposits is the demand, time and saving deposits in deposit money banks. Stock Market Value Traded is total shares traded on the stock market exchange. Stock Market Capitalization is the value of listed shares. External Finance Dependence is defined as the ratio between capital expenditures minus cash flow from operations and capital expenditures. External Equity Dependence is defined as the net amount of equity issued over capital expenditures. Analyst Disagreement is the ratio between the standard deviation and the mean of the forecasted EPS. The dummy for family ownership equals one when the ultimate owner of the firm is a family or an individual and zero otherwise (widely-held). Country and industry fixed effects are included in all regressions. Standard errors (in brackets) are robust to arbitrary heteroskedasticity and allow for serial correlation through clustering by ultimate owner. All regressions include country and industry dummies. * significant at 10%; ** significant at 5%; *** significant at 1%.
Observations
Psedu R
Dummy for family ownership
Employees ('000)
Financial Development × Productivity Growth
Financial Development × Productivity
Financial Development × Industry Capital/Employee
Financial Development × R&D/Sales
Financial Development × Lerner Index
Financial Development × Industry measure
Industry measure:
Financial Development:
(1)
Dependent variable:Dummy for Group Affiliation. Probit estimation
Table X - Robustness Check: Controlling for Interactions with Industry Characteristics
34
(3)
61,879
61,879
0.158
(0.042)
0.158
-0.529***
(0.049)
(0.032)
-0.510***
-0.122***
(0.035)
(0.015)
-0.109***
-0.037**
(0.017)
(0.075)
-0.003
0.639***
(0.079)
(0.013)
0.640***
0.044***
(0.017)
(0.023)
0.048***
(0.035)
-0.066***
(0.018)
-0.058
-0.016
(0.026)
(0.011)
-0.031
0.026**
(0.011)
(0.042)
0.001
-0.067
(0.059)
(0.011)
(0.015)
-0.089
0.023**
(0.014)
-0.018
-0.044***
(0.019)
External Equity
-0.041**
External Funds
61,879
0.157
(0.047)
-0.500***
(0.037)
-0.143***
(0.027)
-0.060**
(0.079)
0.696***
(0.017)
0.065***
(0.029)
0.037
(0.024)
-0.003
(0.024)
0.064***
(0.053)
-0.123***
(0.138)
-0.013
(0.018)
-0.056***
Analyst disagreement
Stock Market Value Traded / GDP
(2)
(5)
(6)
61,879
0.158
(0.049)
-0.509***
(0.035)
-0.109***
(0.017)
-0.003
(0.079)
0.648***
(0.017)
0.049***
(0.034)
0.057*
(0.026)
-0.031
(0.011)
-0.001
(0.058)
-0.096*
(0.015)
-0.018
(0.033)
-0.129***
External Funds
61,879
0.158
(0.042)
-0.529***
(0.032)
-0.121***
(0.015)
-0.037**
(0.075)
0.645***
(0.013)
0.044***
(0.023)
0.065***
(0.018)
-0.017
(0.011)
0.026**
(0.042)
-0.071*
(0.012)
0.023**
(0.025)
-0.106***
External Equity
61,879
0.157
(0.047)
-0.499***
(0.037)
-0.142***
(0.027)
-0.060**
(0.079)
0.700***
(0.017)
0.065***
(0.029)
0.036
(0.024)
-0.003
(0.025)
0.065***
(0.052)
-0.126**
(0.138)
0.030
(0.030)
-0.101***
Analyst disagreement
Stock Market Capitalization / GDP
(4)
(8)
61,879
0.158
(0.049)
-0.509***
(0.035)
-0.109***
(0.017)
-0.003
(0.080)
0.636***
(0.017)
0.048***
(0.021)
0.058
(0.026)
-0.031
(0.011)
0.001
(0.059)
-0.085
(0.015)
-0.019
(0.031)
-0.068**
61,879
0.158
(0.042)
-0.529***
(0.032)
-0.123***
(0.015)
-0.377**
(0.075)
0.634***
(0.013)
0.044***
(0.023)
0.065
(0.018)
-0.016
(0.011)
0.026**
(0.042)
-0.062
(0.011)
-0.024**
(0.022)
-0.074***
External Equity
(9)
61,879
0.157
(0.048)
-0.499***
(0.037)
-0.145***
(0.028)
-0.061**
(0.079)
0.687***
(0.017)
0.064***
(0.029)
0.036
(0.024)
-0.001
(0.025)
0.066***
(0.053)
-0.116**
(0.138)
0.024
(0.029)
-0.101***
Analyst disagreement
Private Credit / GDP External Funds
(7)
(11)
(12)
61,879
0.158
(0.049)
-0.513***
(0.035)
-0.109***
(0.017)
-0.003
(0.079)
0.640***
(0.017)
0.048***
(0.035)
-0.062*
(0.026)
-0.031
(0.011)
-0.006
(0.059)
-0.089
(0.015)
-0.018
(0.041)
-0.062
External Funds
61,879
0.158
(0.042)
-0.533***
(0.032)
-0.123***
(0.015)
-0.038**
(0.075)
0.637***
(0.013)
0.044***
(0.023)
-0.069***
(0.018)
-0.016
(0.011)
-0.026**
(0.042)
-0.065
(0.012)
-0.023**
(0.029)
-0.063**
External Equity
61,879
0.157
(0.047)
-0.505***
(0.037)
-0.144***
(0.027)
-0.061**
(0.079)
0.694***
(0.017)
0.065***
(0.029)
0.041
(0.024)
-0.002
(0.025)
0.065***
(0.053)
-0.121**
(0.138)
-0.024
(0.038)
-0.092**
Analyst disagreement
Bank Deposits / GDP
(10)
Notes: This table reports the results of Probit regressions that examine the effect of financial development on business group affiliation when controlling for various firm characteristics. The dependent variable is a group affiliation dummy that equals one for firms that are affiliates to groups and equals zero for standalones. Private Credit is the private credit by deposit money banks and other financial institutions. Bank Deposits is the demand, time and saving deposits in deposit money banks. Stock Market Value Traded is total shares traded on the stock market exchange. Stock Market Capitalization is the value of listed shares. External Finance Dependence is defined as the ratio between capital expenditures minus cash flow from operations and capital expenditures. External Equity Dependence is defined as the net amount of equity issued over capital expenditures. Analyst Disagreement is the ratio between the standard deviation and the mean of the forecasted EPS. The dummy for family ownership equals one when the ultimate owner of the firm is a family or an individual and zero otherwise (widely-held). Country and industry fixed effects are included in all regressions. Standard errors (in brackets) are robust to arbitrary heteroskedasticity and allow for serial correlation through clustering by ultimate owner. All regressions include country and industry dummies. * significant at 10%; ** significant at 5%; *** significant at 1%.
Observations
Psedu R
2
Dummy for family ownership
Employment Growth
Capital/Employees
Sales/Employees
Employees ('000)
Family ownership × Industry measure
Employment Growth × Industry measure
Capital/Employees × Industry measure
Sales/Employees × Industry measure
Employees × Industry measure
Financial Development × Industry measure
Industry measure:
Financial Development:
(1)
Dependent variable:Dummy for Group Affiliation. Probit estimation. Firms with more than 50 Employees.
Table XI - Robustness Check: Controlling for Interactions with Firm Characteristics
A A.
Appendix Legal variables
This section provides detailed information about the legal variables used in our estimation (Table IX). Our data sources include the PWC Worldwide Tax Summaries, AGN Europe Tax Surveys, Global Legal Group (GLG) International Comparative Legal Guide to Corporate Tax 2006, EU Parent Subsidiary Directive (90/435/EEC), and various European Commission working papers on corporate tax. (1) Inter-company dividend tax. According to the EU Parent Subsidiary Directive (90/435/EEC), the Member State of the subsidiary paying the dividends may not charge withholding tax on the dividends paid. The Member State of the company receiving the dividends has two options: (a) the dividends received may be exempt from corporation tax; (b) the dividends received are taxed but a tax credit equivalent to the corporation tax paid by the subsidiary on the pro…ts distributed to the parent company is granted. Our prediction is that tax credit discourages the formation of business groups compared to tax exemption. The inter-company dividend laws in the EU as for June 2006 are presented in Panel A of Table VIII. (2) Consolidated tax returns. The option to …le a consolidated tax return is a key incentive for the formation of pyramidal business groups. Consider, for example, a parent …rm with pro…ts that are negatively correlated with those of its subsidiary. If the two …rms can …le consolidated tax returns, then the losses of one …rm o¤set the pro…ts of the other and decrease the total taxable income of the two companies. Hence, our prediction is that higher holding threshold to qualify for consolidated tax returns discourages the formation of business groups. Panel B of Table VIII presents the 2006 laws in the EU regarding consolidated tax returns. It describes the minimum holdings needed to …le consolidated tax returns. (3) Capital gains tax. While the former two laws deal with the incentive to form new
35
business groups or enlarge existing ones, the tax on capital gains refers to the cost of disentangling existing business groups. Reducing the capital gains tax encourages the ultimate owner to realize her pro…ts and eliminates business groups. Yet, it is not clear what would be the prediction in a cross sectional study since a lower capital gains tax can also encourage the formation of groups since the ultimate owner knows that there is a lower cost if she wants to disentangle the business group in the future. Panel C of Table VIII presents the 2006 capital gains tax in the EU and speci…es the percentage of the exemption (if any). B.
Industry variables
We construct the industry measures used in the econometric speci…cations, using data from Compustat, Amadeus, and IBES. The following variables are based on Compustat. They are a weighted average over the period 1980-2004 and are computed at the three-digit US SIC level. External Finance Dependence - this variable is de…ned as the ratio of Capital Expenditures (Compustat #128) minus Cash Flow (#110) to Capital Expenditures. When #110 is missing, Cash Flow is de…ned as the sum of the following Compustat items: #123, #125, #126, #106, #213, and #217. External Equity Dependence - this variable is de…ned as the ratio of the net amount of equity issued (#108 minus #115) to capital expenditures. R&D Intensity - this variable is de…ned as the ratio of R&D expenditures (#46) to sales (#12). Productivity and Lerner index are based on Amadeus and are a weighted average over the period 1995-2004: Productivity is de…ned as the weighted average of the ratio between sales and number of employees. Lerner index - this variable is de…ned as the industry median of the …rm-weighted average of one minus the ratio of pro…ts to sales. Both variables are computed based on UK data. Analyst disagreement variable is based on analyst earning-per-share forecasts from IBES (Institutional Brokers Estimate System) Detail History dataset. This …le contains information about the forecasts of individual analysts including the forecasting date. Following Di-
36
ether et al (2002) and Johnson (2004), we de…ne analyst disagreement as the ratio between the standard deviation of the forecasted earnings-per-share and the mean of the earnings forecast. We consider all US listed stocks between 1990 and 2004 that have been followed by at least two analysts in the same month. In ranking industries according to the measure of analyst disagreement, we compute the industry mean of this measure over the stocks in our sample.
37
Notes 1
See Morck at al. (2005) for a summary of this literature.
2
The ubiquity of business groups was also documented in: Faccio and Lang (2002) on
Western Europe, Claessens et al. (2000) on East Asia, and Khanna and Yafeh (2007) on 12 emerging markets and Japan. The e¤ect of group a¢ liation on di¤erent measures of …rm performance was studied in numerous papers including: Khanna and Palepu (2000) who study the e¤ect on …rm value, Hoshi at al. (1991) on investments, Khanna and Rivkin (2001) on pro…tability, Khanna and Yafeh (2007) and Gopalan et al. (2007) on risk sharing and the probability of default, and Belenzon and Berkovitz (2007) on innovation. 3
For example see Hoshi et al. (1991), Khanna and Palepu (2000), Gopalan et al. (2007),
and Belenzon and Berkovitz (2007). 4
Since our data also includes private …rms, we mitigate the common selection bias of most
studies on business groups which analyze only public …rms (e.g. La Porta et al. (1999) and Faccio and Lang (2002)). 5
For reasons of conservatism and simplicity, we de…ne control of a private …rm as owning
more than 50% of the …rm’s voting rights (excluding non-voting shares). Following previous literature on public …rms (La Porta et al. (1999), Faccio and Lang (2002) and others), which have a more dispersed ownership, we set the threshold for public …rms at 20%. All the results of this paper are robust to di¤erent plausible speci…cations of these thresholds. 6
More details on the algorithm and the name matching procedure are provided in Belenzon
and Berkovitz (2007). 7
Some …rms form fully-owned subsidiaries in other countries for tax purposes (dormant
…rms). Hence, it is important to focus only on …rms with real economic activity for the de…nition of a business group. 8
We compute the competition and productivity measures for British …rms (instead of US
…rms) to exploit the availability of data on private …rms, which result in a more accurate measure of these variables. Yet, we get the same pattern of results when calculating all 38
variables using US data. 9
More information on the construction of the industry variables is provided in the appen-
dix. 10
We use the number of employees rather than total assets since German …rms are not
obliged to disclose their total assets. 11
When comparing the ranking of countries according to stock market versus banking
system measures, we …nd that the main di¤erence is in the ranking of Germany. While most countries have similar relative ranking, Germany is ranked very highly on the banking system metric, but it is ranked much lower on the stock market metric. Since Germany represents nearly a quarter of the observations in our sample, this may explain why the bank deposits measure is insigni…cant. Indeed, if we exclude Germany from the estimation, the coe¢ cient of the bank deposit measure is signi…cant. 12
Indeed, our data suggest that ownership structure is ‘sticky’over time (also claimed by
La Porta et al. (1999)). 13
We did not encounter any explicit restriction on utilities companies as in the US. In
addition, we do not focus on the takeover rule mentioned earlier in this paper and presented in Table 8 since it applies only for listed …rms which are only a small fraction of our sample. 14
We also tested speci…cations with all the legal variables together in the same regressions
and got the same results (excluded from the paper for brevity).
39