Law and Finance: Common Law and Civil Law Countries ComparedFAn Empirical Critique

Transcription

1 Economica (2008) 75, doi: /j x Law and Finance: Common Law and Civil Law Countries ComparedFAn Empirical Critique By MICHAEL GRAFF University of Queensland and ETH, Zurich Final version received 4 July The theory of law and finance argues that the common law system provides a better framework for financial development and economic growth than the civil law tradition. This paper identifies a number of problems that cast doubt on the soundness of the empirical basis of this literature. However, this analysis supports the idea that the legal tradition is a major factor in shaping corporate law. In particular, while there is not much evidence that common law countries protect financial investors better than civil law countries I find support for the assumption that financial investors are treated differently across legal families. INTRODUCTION The last few years have witnessed the emergence of a considerable body of literature on law and finance that began with two widely cited papers by La Porta, Lopez-de Silanes, Shleifer and Vishny (La Porta et al. (henceforth LLSV) 1997, 1998). This literature of law and finance theory, as it is called by its proponents, is an ambitious and fascinating attempt to combine insights from the theory of corporate finance, institutional economics and legal and economic history, as well as from the recent bulk of studies on the determinants of economic growth, into an encompassing theory of economic development. The theory of law and finance argues that the legal system that countries have inherited from the past is crucial to the analysis of economic growth because of how it variously favoursfor hampersffinancial development. Moreover, financial development is regarded as a major driving force of economic growth; consequently, the legal system is identified as one of the ultimate causes of economic development. The causal chain proposed by this literature is therefore Legal system! financial development! economic growth Drawing on a taxonomy developed by scholars of law and legal history, the theory distinguishes a common law tradition inherited from medieval England and a more diversified civil law tradition that dates back to nineteenth-century codifications in France, Germany and Scandinavia. On this basis, the major finding is that common law provides a better basis for financial development and economic growth than civil law, and that French civil law is worst. In what follows, I shall first summarize the theory s main findings and hypotheses regarding the first link, namely the legal system financial development nexus. While the policy relevance of the theory depends on the existence of both the legal system financial development and the financial development economic growth links, the original contribution of the theory concerns predominantly the first part. This is an important topic in its own right, even if the effect of finance on growth should be less pronounced than is widely believed. 1 After this, I review and re-analyse the data underlying the formulation of the fundamental findings. My findings cast some doubt on the soundness

2 2008] LAW AND FINANCE IN COMMON LAW AND CIVIL LAW COUNTRIES 61 of the empirical basis generally referred to in this literature. In particular, I shall demonstrate that, although investors are treated differently across legal families, the data do not support the claim that common law countries generally protect financial investors better than civil law countries. I. THE LAW AND FINANCE LITERATURE After less than ten years, the finance and law literature has produced its first synthesis. Written by two insiders, Beck and Levine s (2003) contribution to the forthcoming Handbook of New Institutional Economics ( Legal institutions and financial development ) gives an extensive overview of this research programmefits foundations, assumptions, data and main findings. To reproduce the original spirit of the theory, I shall refer closely to this authoritative survey. Beck and Levine (2003, p. 1) argue that the theory of law and finance explains why some countries have well-developed growth-enhancing financial systems, while others do not, and why some countries developed the necessary investor protection laws and contract-enforcement mechanisms to support financial institutions and markets, while others have not. The theory s explanatory power is attributed to two related hypotheses: 1. [I]n countries where legal systems enforce private property rights, support private contractual arrangements, and protect the legal right of investors, savers are more willing to finance firms and financial markets flourish. 2. [T]he different legal traditions that emerged in Europe over previous centuries and were spread internationally through conquest, colonization, and imitation help explain cross-country differences in investor protection, the contracting environment, and financial development today. Two mechanisms are held responsible for these outcomes (Beck and Levine, 2003, p. 2): 1. a political mechanism, which refers to the difference between legal traditions in terms of the priority they attach to private property vis-a` -vis the rights of the State ; 2. an adaptability mechanism, which refers to the formalism that may impair the legal system s capability to minimize the gap between the contracting needs of the economy and the normative status quo. A key argument of this approach is that a major function of property rights consists in ameliorating the inherent risk involved in financial contracts because of informational asymmetry and moral hazard or outright fraud. On this basis, the microeconomic foundation of the theory is the willingness to invest. Accordingly, to the degree that the legal system offers effective protection against the occurrence and, if necessary, the consequences of these types of market failures or deception, financial investors will be more inclined to lend, be it directly on the financial market or via financial intermediaries. With its emphasis on the legal protection of lenders, this approach has roots in both new institutional economics and the theory of corporate finance. The theory is focused mainly on shareholder protection, in particular the legal system s provisions against theft as well as transfer pricing, asset stripping, the hiring of family members and related practicesfor, in more general terms, whether inside managers and controlling shareholders are... in a position to expropriate minority shareholders and creditors

3 62 ECONOMICA [FEBRUARY (Beck and Levine 2003, p. 4). Some, albeit considerably less, attention is devoted to creditors rights. Notwithstanding the differences between equity and debt finance, a unifying notion in the theory of law and finance is the distinction between insiders (stakeholders like the State or workers) and outsiders (shareholders and creditors), including the legal system s support to outsiders. This functions to increase their willingness to invest and so is expected to be beneficial to financial development, whereas a strong position of insiders would be detrimental. The original contribution of the theory to these established ideas and assumptions lies in the way it combines them with its peculiar view of legal history. Let us therefore take a closer look at how the theory deals with the historical legacy of law. The civil law family is traced back to the Roman Empire, the first society with secular and statuary law. Its modern offsprings are divided into French, German and Scandinavian branches. The theory s view of the French legal system is that it evolved as a regionally diverse me lange of customary law, law based on the Justinian texts, and case law. However, by the 18th century, there was a notable deterioration in the integrity and prestige of the judiciary. The Crown sold judgeships to rich families and the judges unabashedly promoted the interests of the elite and impeded progressive reforms. Unsurprisingly, the French Revolution turned its fury on the judiciary and quickly strove to (a) place the State above the courts and (b) eliminate jurisprudence.... Napoleon sought a code that was so clear, complete, and coherent that there would be no need for judges to deliberate publicly about which laws, customs, and past experiences apply to new, evolving situations. Furthermore, this approach required a high degree of procedural formalism to reduce the discretion of judges.... (Beck and Levine 2003, p. 6) In contrast to this, the German legal system is seen as the result of evolution rather than revolution. When Bismarck decided to codify and unify the whole of private law in Germany that led to the adoption of the German civil law in 1900, Germany had a long history of deliberations that illustrated how courts weighted conflicting statutes, resolved ambiguities, and addressed changing situations. Hence there was... a dynamic, common fund of legal principles that then formed the basis for codification in the 19th century. Moreover, in contrast to the revolutionary zeal and antagonism towards judges that shaped the Napoleonic Code, German legal history shed a much more favorable light on jurisprudence and explicitly rejected France s approach.... Whereas the Napoleonic code was designed to be immutable, the Bu rgerliches Gesetzbuch was designed to evolve.... Thus, while codification had a similar role in Germany and France in unifying the country and reasserting the power of the central state, Germany had a very different approach towards jurisprudence. (Beck and Levine 2003, p. 8f.) While the theory it is not very explicit regarding the Scandinavian civil law family, it does stress that, like Germany, Scandinavia rejected the legal traditions brought about by the French Revolution. The other major legal legacy is common law, which is characterized as unique both in terms of (a) the relationship between the State and the Courts and (b) jurisprudence.... English law evolved based on the resolution of specific disputes and increasingly stressed the rights of private property, [and] the courts developed legal rules that treated large estate holders as private property owners and not as tenants of the king. Indeed, the common law at the dawn of the 17th century was principally a law of private property.... In terms of legal formalism, English law typically imposes less rigid and formalistic requirements.... In terms of jurisprudence, the English common law tradition is almost synonymous with

4 2008] LAW AND FINANCE IN COMMON LAW AND CIVIL LAW COUNTRIES 63 judges having broad interpretation powers and with courts melding and creating law as circumstances change... rather than adhering to the logical principles of codified law. (Beck and Levine 2003, p. 9f.) This view of the history of law, combined with the theory s focus on lenders property rights, leads us to conclude that common law is sufficiently flexible to deal adequately with complicated matters like financial contracts that are contingent on a host of foreseeable and unforeseeable states of nature and business. For the civil law family, conclusions are mixed. While the revolutionary fury to wipe out jurisprudence and establish a permanent order of reason through codified, positive law is the antithesis of flexibility, the German and Scandinavian systems are elevated for having rejected the French approach. Of the two mechanisms that translate these generic differences into differential operational growth and development, the theory of the adaptability mechanism predicts a ranking of legal systems in terms of appropriateness to promote financial development in which common law is superior, German and Scandinavian civil law are intermediate, and French civil law is inferior. Regarding the political mechanism, the theory, firmly rooted in contemporary orthodox economics, is entirely predictable in its perception of the State as a menace, a Leviathan always trying to grab an undue share of society s resources and thus interfering with the peaceful activity of private economic agents. As Beck and Levine explain, The political mechanism holds that the civil law has tended to support the rights of the State, rather than private property rights,... with adverse implications for financial development.... A powerful State with a responsive civil law at its disposal will tend to divert the flow of society s resources towards favored ends... which is antithetical to competitive financial markets. (Beck and Levine 2003, pp. 13 f.) The political mechanism hence implies a binary classification of the appropriateness of legal systems to promote financial development: common law is superior to civil law. II. SHOULD WE BELIEVE WHAT THE THEORY OF LAW AND FINANCE IS TELLING US? Taken together, the story told so far is coherent and intuitively plausible, though not in all of its nuances. In particular, the political mechanism implies a ranking of civil law (protecting investors) before common law (giving more power to the state), whereas the adaptability mechanism implies that common law countries (with an inherently flexible legal system) will guarantee the highest level of protection to financial investors, followed by Scandinavian and German origin civil law countries and with French legal origins (with its excessive formalism) yielding the poorest result. Both mechanisms are thus contributing to making common law countries best protective of investors with the difference in adaptability allowing a further ranking within the civil law country group. Note however, that while the adaptability mechanism is developed in a logical sequence of arguments, and therefore intrinsically persuasive, the political mechanism is far less convincing. In particular, the way in which civil law promotes the Leviathan to expand its sphere of influence at the cost of private finance, and thereby disrupting the natural functioning of financial markets, remains unclear. For, unless one assumes that a State that dares give judges a legal code and make them adhere to it without regard to circumstance is also likely to do other nasty things, particularly to financial investors, 2 this argument would surely merit elaboration.

5 64 ECONOMICA [FEBRUARY However, although it may appear to some as a rationale to demonstrate the alleged superiority of the Anglo Saxon model of corporate finance (Singh and Weisse 2001), the law and finance literature is, to a large extent, an empirical approach that offers a neat framework to link empirical phenomena that otherwise would seem unrelatedfand, moreover, to connect these in an important theoretical way so as to link politics, law, finance and economic growth and development. Therefore, given adequate cross-country data, the proposed ranking of countries from different legal origins in terms of investor protection forms a testable hypothesis. Moreover, one of the major achievements of the law and finance literature is the gathering and compiling of a set of international data to perform this test. Let us therefore look at the data presented by LLSV (1998), and referred to in an impressive number of papers. The central point remains that the taxonomy of legal systems and LLSV s comparative assessment of the legal framework for corporate finance and credit markets comprises the very backbone of the theory of law and finance. Practically all subsequent studies, supportive or sceptical, accept this taxonomy as the natural starting point of analysis. Consequently, the validity of the original data that generated this taxonomy is clearly crucial for any degree of confidence we are to have in the research programme as a whole. (a) A closer look at shareholder rights Regarding international comparisons of shareholder rights, LLSV (1998) discuss eight indicators that characterize various aspects of shareholder rights (six of them binary) along with an anti-director rights index which they compute from a selection of six of these variables. In particular, they define (1998, table 1): 1. One share one vote: 1 if the Company Law or Commercial Code (CLCC) of the country requires that ordinary shares carry one vote per share, 0 otherwise. Equivalently, this variable equals 1 if the law prohibits the existence of both multiple voting and non-voting ordinary shares and does not allow firms to set a maximum number of votes per shareholders irrespective of the number of shares owned, and 0 otherwise. 2. Proxy by mail: 1 if the CLCC allows shareholders to mail their proxy vote to the firm; 0 otherwise 3. Shares not blocked: 1 if the CLCC does not allow firms to require that shareholders deposit their shares prior to a General Shareholder Meeting, thus preventing them from selling those shares for a number of days; 0 otherwise. 4. Cumulative voting: 1 if the CLCC allows shareholders to cast all of their votes for one candidate standing for election to the board of directors (cumulative voting), or if the CLCC allows a mechanism of proportional representation in the board by which minority interests may name a proportional number of directors to the board; 0 otherwise. 5. Oppressed minority: 1 if the CLCC grants minority shareholders either a judicial venue to challenge the decisions of management or of the assembly or the right to step out of the company by requiring the company to purchase their shares when they object to certain fundamental changes, such as mergers, assets dispositions and changes in the articles of incorporation; 0 otherwise. (Minority shareholders defined as those owning 10% of share capital or less.)

6 2008] LAW AND FINANCE IN COMMON LAW AND CIVIL LAW COUNTRIES Pre-emptive rights: 1 if shareholders are granted the first opportunity to buy new issues of stock and this right can only be waived by a shareholder vote; 0 otherwise. 7. Extraordinary meeting: minimum percentage of ownership of share capital that entitles a shareholder to call for an extraordinary shareholders meeting. 8. Anti-director rights: index aggregating the shareholder rights, formed by adding 1 if: (i) the country allows the shareholders to mail their proxy to the firm; (ii) shareholders are not required to deposit their shares prior to the General Shareholders Meeting; (iii) cumulative voting or proportional representation of minorities in the board of directors is allowed; (iv) an oppressed minorities mechanism is in place; (v) the minimum percentage of share capital that entitles a shareholder to call for an Extraordinary Shareholders Meeting is less than or equal to 10% (the sample median); and (vi) shareholders have pre-emptive rights that can only be waived by a shareholders vote. 9. Mandatory dividend: equal to the percentage of net income that the CLCC requires firms to distribute as dividends among ordinary shareholders; 0 for countries without such a restriction. These indicators clearly capture interesting features of the shareholder s position in corporate finance. Moreover, the label anti-director rights for the index combining indicators 2 7 is a neat expression of the general focus on outsider v. insider rights and control. Note however that the first indicator is put aside and dubbed a voting right, and the same happens to the eighth, which is called remedial (more on this later). The data assembled by LLSV (1998, Table 2; reproduced in the appendix of this paper) cover a sample of 49 countries over recent years as organized along the legal system dimensions discussed above. Recall that the anti-director rights index is an ordinal variable that by construction ranges from 1 to 6, with higher values implying better shareholder protection. The most important result from this aggregation of the underlying data is that it broadly confirms the predictions of the theory: the common law countries show an average of 4.00, whereas the Scandinavian, French and German civil law countries on average score only 3.00, 2.33 and 2.33, respectively. Accordingly, as predicted, the common law countries appear to offer better shareholder protection on average than the civil law countries. Now the ranking within the civil law family is not quite in line with the theory; that would have the French system perform worst, whereas its score is actually the same as that of the German civil law group. The evidence is nevertheless favourable with respect to the basic distinction of legal tradition. The literature generally refers to this as a sound confirmation of the underlying assumption. In subsequent empirical work by other researchers, the anti-director rights index is therefore frequently used as a comparative measure for investor protection. Yet before we accept that the reported average index scores support the theory, it is essential to know what information is aggregated, and how. Let us now therefore take a closer look at the informational content and country group mean scores of the indicators from LLSV s original listfboth the ones that enter into the index and those that are excluded. First, the one share one vote dummy variable stands for the provision of proportional voting power with respect to claims on dividends. The textbook model of corporate finance assumes this as given. Surprisingly, however, only 17% of the common law countries have legal provisions for this compared with 29% of French and 33% German civil law countries. The four Scandinavian countries are different again, in that none of them scores 1. Taken together, 26% of the civil law countries follow the one share

7 66 ECONOMICA [FEBRUARY one vote principle. Compared with the 17% of the common law countries, if anything, this indicator actually falsifies the theory. 3 Curiously, it does not enter into the index. Second, the proxy by mail dummy variable reflects the ease with which a shareholder can cast a vote. Common law countries score consistently higher on average (39%) than civil law countries (5% for the French, 0% for the German civil law group and 25% for one out of four Scandinavian countries). The idea is that this represents the chance a minority shareholder has to stand up and fight for her rights without, for example, having to travel to Detroit, London or Frankfurt. However, this seems to presuppose the existence of corporate democracy working in the marketplace along the lines of a Swiss village meeting, where communal issues are decided by a majority raising their hands, whereas Licht et al. (2001, p. 26) remind us that minority voting right protection in corporate finance is something distinctly different from the protection of democratic rights in the political sphere. The small public shareholder is rational to be apathetic ; what matters instead is whether large minority shareholders have a voice. These are likely to be institutional investors or individuals from industrial dynasties for whom the proxy by mail facility in LLSV s anti-director rights variable is mostly irrelevant. To the extent that aggregated portfolios are significant players in modern financial capital markets, this variable has more theoretical than practical significance. Third, the shares not blocked before meeting dummy variable also reflects the ease with which a shareholder can cast his vote. The common law countries score consistently higher on this index. Blocking shares is practised in no common law country, whereas it is practised in 12 out of 21 French civil law countries, one out of six German civil law countries and all four Scandinavian countries. Again, the basic idea seems to be that this expresses the ease and effectiveness of participation in the strategic decisions of the joint enterprise. Yet there may be little informational content in this variable, as a minority shareholder is not a Swiss villager, i.e. is not adverse to the requirement that identity be first established (i.e. as an expression of property rights). That the prerequisite to deposit one s shares before being allowed to vote might actually be a protective measure against fraud, thereby effectively protecting rather than excluding minority shareholders, is seemingly lost on La Porta et al. The meaning of this variable remains both ambivalent and uncertain. Fourth, the cumulative voting dummy variable captures provisions for proportional representation rather than majority candidates on companies boards. Yet with respect to this, there is effectively no difference between the major legal system country groups. Apart from the four Scandinavian countries, which do not provide for this, all legal family subgroups score around 30%. Indeed, at 28%, the common law countries are actually at the lower end. Cumulative voting analysis provides only ambiguous support for the theory, if not rejection. Fifth, the oppressed minority dummy variable specifies whether the legal system supports shareholders in a fight against management or majority decisions, or in their wish to exit, by requiring the company to purchase their shares. This seems a sensible and valid approach to pin down the link between law and finance. Indeed, it provides robust support for the theory. Except for Thailand, all common law countries provide for this mechanism, whereas this is provided in only 50% of German-originated legal systems and 29% of French-originated. The Scandinavian countries, which are civil law countries, are special in that this mechanism is not in force in any of them. This is in line with the theory. Sixth, the pre-emptive rights dummy variable reflects whether shareholders can effectively protect themselves against a watering down of their shares during new stock

8 2008] LAW AND FINANCE IN COMMON LAW AND CIVIL LAW COUNTRIES 67 emissions. This variable is appealing, yet it offers little support for the theory. Fully 62% of the French-legal-origin countries protect shareholders in this respect, as do three out of four Scandinavian countries; but only 44% of the common law countries offer this service. The German legal origin countries score even lower (33%). The proportion of common law countries with this provision is on average lower than that of the civil law countries taken together. This is an ambiguous reading at best. Seventh, the extraordinary meeting indicatorfand the last one that enters into the anti-director rights indexfis the only variable originally measured on a numerical scale, reflecting the minimum percentage of ownership to call for a shareholder meeting, a variable that ranges from 1% to 33% in LLSV s 49-country sample. To join this information to the other dummy variables that combine to the anti-director rights index, LLSV effectively dichotomize it by assigning a value of 1 if the requested share is equal or less than the sample median. Since there is some arbitrariness involved in this procedure, I shall return to this point below. Finally, the mandatory dividend indicator quantifies the legally required minimum dividend from net income. This certainly protects the financial entitlement of small shareholders to receive a reliable income from dividends. Interestingly, this provision is known only in countries of French legal origin, which seem to contradict the theory. However, LLSV come to the opposite conclusion: This result is broadly consistent with the rest of our evidence and suggests that mandatory dividends are indeed a remedial legal protection for shareholders who have relatively few other legal rights (Beck and Levine 2003, p. 1132). Yet this interpretation is questionable, as it appears to be an ad hoc rationalization of an unexpected result: the French legal system must produce inferior outcomes; and if not, this is only because it has to resort to a remedy. Nevertheless, this indicator is not analysed further, and LLSV do not seem sufficiently convinced of their interpretation to include (the inverse of) this variable into the anti-director rights index. Instead, they prefer to disregard it. Taken together, what do the proponents of the finance and law view get out of this indicator set? According to LLSV, the differences in the various measures of shareholder rights between different legal families are often significant and almost always significant when common- and civil-law families are compared.... In sum, common-law countries have the relatively strongest and the French-civillaw countries the weakest, protections of shareholders... (LLSV 1998, pp f.) This bold conclusion is reiterated in practically all subsequent papers adhering to this research programme. 4 Yet it is not supported by the underlying data. There are two reasons for this. First, the reported t-tests of the differences of the various shareholder protection dummy variables are flawed, because a t-test is not adequate for a comparison of binary variables. An appropriate parametric test of the difference between two proportions based on a binomial distribution could be performed. 5 Yet we shall leave this aside, as there is no need to worry about significance if the informational content is questionable in the first place. In particular, as argued above, two of the items in the LLSV indicator set for shareholder protectionf proxy by mail and shares not blocked before meeting Fare practically irrelevant and therefore are not valid to measure what they are supposed to. Yet these indicators are included into the anti-director rights index, whereas two relevant and valid indicatorsf one share one vote and mandatory dividend Fare excluded for unconvincing reasons. Second, the dichotomization of extraordinary meeting, a continuous variable that enters into the index, is questionable. Recall that LLSV transform this variable by

9 68 ECONOMICA [FEBRUARY TABLE 1 Mean Values of Different Quantifications of Extraordinary Meeting by Legal Fa mily Legal family EXMREQ EX_LLSV EX_ALT Civil law Common law Total TABLE 2 Correlation of Mandatory Dividend with SRIGHTS Index and Other Indicators Spearman s r Significance (2-tailed) SRIGHTS ONE PROXY BLOCK CUMUL MINOR PRE EXMREQ assigning 1 if the requested share to call in an extraordinary meeting is equal to or less than the sample median, and 0 otherwise. This seems straightforward, but in fact a slightly different dichotomization of the continuous raw data (assigning 1 if the requested share is less than, rather than equal to or less than, the sample median) reverses the result and lets the common law countries appear more protective towards shareholders. Table 1 illustrates this point: the raw data (EXMREQ) do not show a pronounced difference across groups (12% v. 9% required share), and the difference is not statistically significant (p ¼ 0.20). However, LLSV s dichotomization (EX_LLSV) results in 94% of common law v. 68% of civil law countries being shareholder-protective, whereas our alternative median split (EX_ALT), which is preferable since it divides the sample into groups that are closer to each other in numbers, lets the civil law countries outperform the common law countries (35% v. 28%). Hence a median split can produce very different results. 6 Next, let us look at the mandatory dividend variable, which equals zero for all countries except a few countries with French civil law tradition. Recall that LLSV interpret this variable as remedial, claiming that it signals weak shareholder protection. Their data, however, do not confirm this interpretation. As Table 2 shows, the Spearman correlation with the anti-director rights index (SRIGHTS), into which it does not enter, is virtually zero ( 0.07); it is however positively and significantly correlated with the other indicator that LLSV exclude from their overall index, namely one share one vote (ONE). The pairwise correlation with the six indicators that form the index is negative and positive in three cases, respectively. Since two of three negative correlations are with the indicators that we would not consider particularly validfi.e. proxy by mail (PROXY) and shares not blocked before meeting (BLOCK)Fthen, if anything, there is a tendency for mandatory dividend to correlate positively with the valid core of the

10 2008] LAW AND FINANCE IN COMMON LAW AND CIVIL LAW COUNTRIES 69 TABLE 3 Binary Indicators Scoring 1 BY LEGAL FAMILY (% OF COUNTRIES) Legal family ONE PROXY BLOCK CUMUL MINOR PRE EX_ALT MAND Civil law Common Total indicator set, which runs counter to the attempt to describe it as a remedy. The unbiased meaning of mandatory dividend then should be included in the index for shareholder protection with a positive sign. Let us now summarize the group differences of the binary and dichotomized shareholder rights indicators. These findings are mixed, as Table 3 reflects. The civil law countries tend to perform better on one share one vote (ONE), pre-emptive rights (PRE), our preferred dichotomization of extraordinary meeting (EX_ALT) and mandatory dividend (MAND). In addition, the common law countries are on average better with respect to proxy by mail (PROXY), shares not blocked before meeting (BLOCK) and oppressed minority (MINOR). Finally, cumulative voting (CUMUL) does not differ markedly between the large legal families. How do these points taken together affect the index values for the different legal families? Of the indicators that contribute to the alleged supremacy of the common law countries in terms of shareholder protection quantified by the anti-director rights index, two are not valid and should be dropped and a third is dubious because of its dichotomization. Moreover, two variables that would suggest supremacy of the common law countries are disregarded in the construction of the index. To the empirically minded, this is an irritating finding. Yet, referring to the same data-set as LLSV, I shall argue that the supremacy of the common law countries disappears when the set of information is combined in other and perhaps more plausible ways. Table 4 shows the results of some alternative aggregations of the same underlying information: LLSV is the original anti-director rights index. INDEX_1 is an alternative specification where the two indicators of dubious validity ( proxy by mail and shares not blocked ) are dropped from LLSV and extraordinary meeting is dichotomized alternatively (EX_ALT instead of EX_LLSV). INDEX_2 is computed by adding one share one vote to INDEX_1, and INDEX_3 is obtained by adding MAND, which equals 1 when there are provisions for mandatory dividends, to INDEX_2. Obviously, LLSV s key relationship between the two major legal traditions and shareholder protectiveness cannot be reproduced with these alternative indices. Indeed, the more we modify the index according to our theoretical intuition, the closer the group means move towards each other. 7 (b) A closer look at creditor rights Let us now examine the theory s comparative analysis of creditor rights. These are LLSV s (1998, Table 1) definitions of creditor rights indicators: 1. Reorganization: 1 if the reorganization procedure imposes restrictions, such as creditors consent, to file for reorganization; 0 for countries without such restrictions

11 70 ECONOMICA [FEBRUARY 2. No automatic stay: 1 if the reorganization procedure does not impose an automatic stay on the assets of the firm upon filing the reorganization petition; 0 otherwise. 3. Secured first: 1 if secured creditors are ranked first in the distribution of the proceeds from a bankrupt firm; 0 if non-secured creditors, government and workers, come first 4. No management stay: 1 if an official is appointed by the court or by creditors and is responsible for the operation of the business during reorganization; 0 otherwise 5. Creditor rights: index formed by adding the previous four dummy variables 6. Legal reserve: minimum percentage of total share capital mandated by Corporate Law to avoid dissolution of an existing firm; 0 for countries without such restriction The creditor rights indicator set appears intuitively plausible. It is not plagued by conceptual difficulties that, as argued above, make the informational content of some shareholder rights indicators questionable. The creditor rights indicators are directly concerned with how well creditors are protected in the case of bankruptcy, which is the major concern for outside creditors who are entitled to fixed claims otherwise. 8 Before proceeding further, it is in order to note that the theory does not make separate predictions about shareholder and creditor rights, so that common law countries should perform better than civil law countries in both respects. Indeed, comparing the average for the creditor rights index across legal families, 9 LLSV claim that their data support the view that it is not the case that some legal families protect shareholders and others protect creditors (1998, p. 1139). Yet they concede that one possible exception is that German-civil-law countries are protective of secured creditors, though generally not of shareholders... [and T]he United States is actually one of the most anticreditor common-law countries... (p. 1138). LLSV (1998, table 4, reproduced in the appendix of this paper) are certainly understating the anti-creditor orientation of the United States, as it actually ranks lower on their creditor rights index than any of the four legal origin family group means: only Columbia, Mexico, Peru the Philippines and France score worse. In other words, the predictions of the theory do not hold for the German origin country group, which they judge as doing better than expected. The United States in turn, as the herald of property rights and investor protection, nevertheless finds itself by this measure in a group of developing laggards from Latin America, the crony capitalist Philippines and France. What is going on? Readers might find it hard to attribute internal consistency to a theory that can tolerate such manifest empirical contradiction without questioning its premises. To be sure, other proponents of the law and finance view have been more irritated about this than LLSV (see Beck and Levine 2003, pp. 17 ff.). In addition, legal origin seems to have more explanatory power in predicting financial development with respect to stock markets than bank-based financial development (see Bordo and Rousseau 2005). TABLE 4 Alternative Shareholder Rights Indices by Legal Family Legal family LLSV INDEX_1 INDEX_2 INDEX_3 Civil law Common law Total t-test for difference 25 (po0.01) 2.20 (p ¼ 0.15) 0.99 (p ¼ 0.33) 0.18 (p ¼ 0.67)

12 2008] LAW AND FINANCE IN COMMON LAW AND CIVIL LAW COUNTRIES 71 Furthermore, some critics claim that other exogenous variables, such as natural endowments, are better than legal origin in predicting the LLSV creditor rights index (see Acemoglu et al. 2001, 2002), and some proponents of the theory have been quite receptive to this endowment view (see in particular Beck et al. 2003). Accordingly, it seems that the domain of the theory of law and finance is the share market rather than the credit market. However, the generalizations with respect to shareholder protection are built on dubious statistical aggregations. Without theoretical arguments to account for the admitted flaws in explaining creditor protection, the theory rests on shaky empirical foundations. Yet, even if the LLSV indicator set is not delivering convincing support for the theory s predictions regarding investor protectiveness, it might still prove useful to reveal interesting cross-country patterns in the legal framework of finance. Before concluding, then, I shall submit the LLSV data to an explorative statistical reexamination. (c) Further analyses of LLSV s shareholder and creditor rights indicators LLSV s aggregation of data from the indicator set into their creditor and shareholder rights indices entails of a number of arbitrary steps, and some minor modifications yield results that are very different from those originally brought forward in support of the theory. This suggests that the indicator set might reflect not different expressions of the same underlying phenomenon, which could be interpreted as protectiveness towards outside investors, but instead a number of empirically independent characteristics of corporate law. Statistically, this interpretation would imply that the data from which LLSV aggregate their indices are for the most part uncorrelated. Consider the correlation matrices for the shareholder rights and creditor rights indicators (Tables 5 and 6). 10 As Table 5 shows, the correspondence between the shareholder rights indicators is very low. Only three of 28 pairwise binary correlations j are significantly different from zero. Moreover, out of the three significant correlations, one (MAND, ONE) is contrary to LLSV s interpretation of MAND as a remedy (see Section III(a)), so that only two meaningfully significant correlations remain. This indicator set is thus not even close to a one-dimensional structure, so it is hard to attribute a meaning to a sum of its elements. Referring to the theory of law and finance, we would expect that the legal tradition would determine a country s level of protectiveness towards shareholders in general terms, which is clearly not the case if the indicators are uncorrelated. TABLE 5 Correlation Matrix (f) of Shareholder Rights Indicators ONE PROXY BLOCK CUMUL MINOR PRE EX_ALT ONE PROXY 0.26 BLOCK CUMUL MINOR n PRE EX_ALT n 0.15 MAND 0.40 n n Significant at the 0.05 level (two-tailed test).

13 72 ECONOMICA [FEBRUARY TABLE 6 Correlation Matrix (f) of Creditor Rights Indicators REORG AUTOSTAY SECURED NOMANAGE REORG AUTOSTAY 0.45 n SECURED NOMANAGE 0.38 n 0.58 n 0.11 RESERVE 0.29 n n n Significant at the 0.05 level (two-tailed test). The creditor rights indicator set is more coherent, as can be inferred from Table 6. Five out of ten correlations are statistically significant and all have the expected sign, at least if one is willing to accept LLSV s interpretation of legal reserve (RESERVE) as a remedy. In terms of internal consistency, the creditor rights indicator set seems more suitable for aggregation into a one-dimensional index of a country s general protectiveness towards outside investors than the (largely uncorrelated) shareholder rights indicators. Yet the theory performs better in predicting shareholder protection than creditor protection. The data are not consistent. Thus LLSV s aggregation of the data into two dimensions of investor protectiveness does not offer particularly revealing insights into the law and finance nexus. However, this still does not imply that the data-set does not contain any useful information for comparative analyses of corporate law; we should simply dismiss the misleading a priori assumption that the indicators are coherent expressions of an underlying variable that is determining investor protectiveness. Without imposing this restriction, we are free to look at the data from an heuristic perspective to see whether they allow us to identify groups of variables or observations that are similar in certain respects, though not necessarily along the lines suggested by the theory of law and finance. To this end, I now submit the LLSV investor protection indicator set to a hierarchical cluster analysis, a method that is specifically designed to identify groups of observations that are similar in a number of aspects. (d) A cluster analysis Cluster analyses are multivariate procedures that divide a data-set into a number of subgroups (clusters). 11 In general, they refer to measures of similarity or dissimilarity between observations with respect to a set of variables. These are then grouped into clusters of low within-cluster variance and high variance between clusters. In particular, this can be achieved by successively increasing the tolerated level of within-cluster dissimilarity. Starting with the lowest level of aggregation, where only identical observations are clustered together, observations and clusters are merged until the sample is allocated into two groups that constitute the top of the hierarchy. In practical terms, there are: (i) numerous specifications for measuring similarity or dissimilarity; and (ii) various algorithms to refer to these measures in constructing a hierarchy of clusters. Moreover, cluster analyses are an heuristic device to reveal patterns in the data for which a convincing ad hoc interpretation can (or cannot) be found. For our purpose, the interpretation refers to the LLSV data-set characteristics of investor

14 2008] LAW AND FINANCE IN COMMON LAW AND CIVIL LAW COUNTRIES 73 protection that are rooted in the legal tradition of a country. A hierarchical cluster analysis should divide a suitable sample of countries into two top-level clusters corresponding to the two major legal traditions. Such a pattern would not require that the common law countries protect financial investors better than civil law countries; just that they treat them differently. In addition, according to the theory s secondary hypothesis, we might expect a threefold separation of the civil law cluster, corresponding to its Scandinavian, German and French branches. LLSV s indicator set comprises eight shareholder and five creditor rights indicators. Of these, ten are binary and three are continuous. Since different scales require different dissimilarity/similarity measures, we cannot pool these data. The options are either to perform separate analyses, or to dichotomize the continuous variables. Since there are only three continuous variables, I opted for the second alternative. 12 Moreover, there are a number of missing values for Jordan and Venezuela, so that I had to exclude these two countries from the sample. 13 Hence the data for clustering consist of 13 binary variables on investor protection across 47 countries. Now according to the theory of law and finance, the legal tradition should affect both shareholder and creditor protection. On the other hand, the theory makes an empirical distinction between the two, so that I was faced with the decision of whether to perform separate cluster analyses for shareholder and creditor rights or a single analysis comprising both subsets of indicators. To resolve this question I proceeded from general to specific, so as to identify the superior result in reproducing the basic common versus civil law distinction. To assess the similarity or dissimilarity between observations, I refer to the Euclidean distance for binary data, which is defined as the square root of (b þ c)/(a þ b þ c þ d), where b þ c is the number of discordant cases (0,1; 1,0) and a þ d the number of concordant cases (0,0; 1,1) for a pair of observations in contingency tables for all indicators. 14 The clusters are determined by the Ward method (see Ward 1963), which minimizes the within-groups variance. Starting from the lowest level of aggregation, this algorithm successively considers all possible pairings and joins those observations to clusters or merges those clusters to higher-level clusters that result in the minimal increase in total within-groups variance. The focus of this algorithm is thus on the within-group homogeneity rather than on the dissimilarity between clusters, which corresponds well with the assumption of the theory of law and finance that a shared legal tradition will result in similar provisions of corporate law to protect investors rights. With the clustering algorithm determined accordingly, I performed three cluster analyses referring to (i) my entire data matrix, (ii) the 8 47 subset covering LLSV s shareholder protection indicators and (iii) the complementary 5 47 subset for creditor rights. I then evaluated the results in terms of the correspondence of the two clusters on top of the hierarchy with the theory of law and finance s basic legal family distinction. In particular, I define a variable common that equals 1 if a country is classified as belonging to the common law tradition, and 0 if it belongs to the complementary civil law group. I then note its positive binary correlation with two dummy variables for the first two clusters. The results are shown in Table 7. To interpret Table 7, note that it represents the fit of an a priori to the data, predicting that the LLSV indicator set (or subsets of it) will result in a top-level clustering of the 47 countries into two groups corresponding to the two major legal traditions of common law and civil law. This basic distinction is reproduced by the cluster analyses, although not to the same degree with the different indicator sets. I got the best correspondence for the complete indicator set: the correlation of common law membership with cluster no. 2 is In fact, cluster no. 2 unites only common law countries, whereas the

15 74 ECONOMICA [FEBRUARY TABLE 7 Binary Correlation (f) of Common with Clusters 1 and 2 Clusters based on 13 investor rights Clusters based on 8 shareholder rights 0.40 n Clusters based on 5 creditor rights Cluster 1 Cluster n 0.68 n n Significant at the 0.05 level (two-tailed test). complementary group comprises all civil law countries; the only misallocation (to which we shall return) is that five common law countries are clustered together with the civil law countries. The correspondence of a two-cluster distinction to the basic legal traditions drops to 0.68 when clustering according to five creditor rights indicators, and to 0.40 with the eight shareholder rights indicators. Moreover, since lower ordinal numbers are assigned to clusters further up in the hierarchy, i.e. to clusters that are uniting relatively more dissimilar observations, we can infer that the clustering with the full indicator set results in a predominantly common law cluster that joins countries that are less dissimilar than in the corresponding predominantly civil law cluster. The same observation holds for the clustering that is based on the creditor rights indicators alone; but referring only to shareholder rights, the cluster that corresponds more closely to the civil law countries exhibits relatively more similarity. This finding provides further reason to disregard the shareholder rights clustering, since the theory s secondary distinction between three subgroups of the civil law tradition would rather predict a higher within-cluster variance for the non-common-law top-level cluster. Consequently, the full indicator set seems most appropriate. Although its superiority in comparison to the creditor rights indicator set is not overwhelming, the fact that it does not exclude the shareholder rights dimension altogether and refers to more than twice as many indicators adds confidence to the soundness of the classification. Let us now take a closer look at the resulting cluster structure of our preferred solution. The corresponding dendogrammefa graphic representation of the hierarchy of clustersfis reproduced in Figure 1. On the left, all 47 observations are listed. Moving to the right, the tolerated amount of dissimilarity within clusters is increased. This results in progressive aggregation until we end up with just two clusters. The dendogramme therefore provides a comprehensive representation of the selected cluster analysis, showing the final hierarchy as well as the distance at which the aggregations occur. An inspection of Figure 1 confirms that many of the lowest-level aggregations on the far left of the dendogramme intuitively make sense, revealing geographical proximity as well as close historical linkages between countries classified as similar in their provisions regarding shareholder and creditor protection in the late twentieth century. In particular, starting from the left, we can infer that, with a distance of zero, Malaysia and Singapore are identical on the 13 dichotomous indicators, as is true with respect to the United Kingdom and Hong Kong. Now the latter was a UK colony until recently, and Singapore is a city-state in the territory of Malaysia, from which it was segregated in 1965, so that the similarity of their corporate law would not come as a surprise. Moving to the right, the second-lowest levels of dissimilarity are observed between India and Pakistan, Canada and the United States, Argentina and Spain, Austria and Germany, Egypt and Indonesia, France and Mexico, Italy and the Netherlands, Sweden