European Corporate Governance Network

Comment on La Porta, Lopez-de-Silanes & Shleifer, Corporate Ownership Around the World by Roberta Romano (Yale)

The authors have to be enthusiastically commended for the careful empirical research that went into this paper (and their other papers in this series of comparative corporate governance studies); empirical work is generally tedious and difficult, but the effort to determine ownership structures for the firms in this sample was prodigious! They have my (our) great admiration and gratitude for the valuable contribution they have made to the corporate governance literature. My comments are loosely organised around five topics: one substantive difference regarding the interpretation, or framing they offer for the issuance of outside equity finance; some questions concerning the accounting for cross-holdings and corporate groups; some questions raised by the data on family ownership; questions concerning some of the variables chosen in Table VIII to test alternative explanations of the data; and finally, a brief reflection on where to go next with this data set.

I. Whose capital is sought, controlling or non-controlling investors'?

La Porta et al. explain the relationship between outside ownership and shareholders' legal protection as dependent upon the interest of controlling shareholders' in not being expropriated themselves by other large shareholders upon going public. In my judgement, they have got the issue backwards. Controlling shareholders do not give up control when they go public (that is they rarely, if ever, retain positions that do not guarantee them working control), and they typically do not sell off their holdings piece meal so as to reach a non-controlling position (where expropriation would be possible), they sell out entirely. The relevant interests concerning legal protections are the outside (public) investors. Investors are not interested in providing outside equity capital if their investments will not be protected from expropriation by the insiders--to put it another way, there is a connection between concentration and legal protection because insiders will be able to sell significant equity amounts at a favourable cost of capita only if their firms fall under an adequate protection regime (or if they can provide other guarantees to outside investors against expropriation). (The latter was the old bank story; La Porta et al.'s data suggest that such a story is incorrect, as they do not find much bank involvement in firms in low protection countries with or without controlling shareholders.)

The incorrect framing of the issue does not significantly affect the value of the paper because it involves a background question that provides the occasion for the paper's inquiry, which is a positive exploration of global ownership patterns. One place where I thought it seems to get in the way was in the conclusion against convergence on p. 37 concerning the "puzzle" that controlling shareholders are not interested in receiving enhanced dividend rights that come from improving minority protection and hence oppose reform. There is no puzzle, however, because this is not a relevant comparison for these shareholders (that is, controlling shareholders will never be in the minority, so the traded-off benefit of higher dividends have no bearing on their choices directly). Moreover, the sale of control blocks carries a premium in high protection states; without evidence that control premiums are lower in the US than other nations (in the low protection class), the emphasised trade-off is not particularly useful in predicting anything about convergence. The relevant question for the controlling shareholder is whether she will need to issue more equity capital, and thus will lose out substantially on proceeds from what would otherwise be higher valued shares. If domestic firms need capital, and debt is not an option, in the global market for equity capital, outside investors' demands for adequate legal protection before investing their dollars may well produce convergence in corporate law rules (particularly as nations become aware of capital outflows on this dimension).

II. Accounting for cross-shareholdings or corporate groups

While I recognise the data limitations that restrict the control group definitions to large ownership blocks, there is a fundamental problem with the method of accounting for cross-ownership that La Porta et al. don't fully acknowledge. Virtually all commentators characterise the corporate governance system of Japan as a system of cross holdings or corporate groups. But in the paper's classification schemes, Japanese firms do not appear as crossly held or owned by other widely-held corporations (0 in Table IV, the same proportion as US and UK firms!). There is something wrong with a classification system that is so at odds with what we know. It is simply not picking up the common understanding of what is meant by referring to a corporate governance system as a system of corporate groups or cross-holding. Although the 10% ownership measure hints at the difference between US and Japanese corporate structures, La Porta et al. surprisingly conclude that (p.21) these patterns look like US patterns. I do not have a good suggestion for how the authors can improve their classification system, as it appears that they cannot obtain information on blocks smaller than 10%, the size of most Keiretsu firms' holdings given statutory restrictions; but given the results on Japan I would focus the discussion on the results for the 10% rather than the current focus on 20% control definitions (i.e., present table IV, etc. using 10% figures). It would at least be reassuring if the authors could give us some idea of the magnitude of the problem--is there only an underidentification problem for Japanese firms, perhaps because of distinctive legal limitations on corporate equity blocks, or is there a more widespread underreporting problem? (For instance, is the 5% reported for Korean firms, another country widely characterised as having a group-based governance system, an understatement of the extent of cross-holdings there as well?).

The classification choice for Allianz Insurance (fig. 6) presents a similar concern--I would have thought this to be a good example of cross-sharing by a corporation, rather than of a widely-held firm (Can/do the managers of Allianz ignore the managers of its corporate investor?) Classifying it as widely held is troubling, as its operating incentives are surely different from those of the widely held US firms with no such owners (as surely are those of members of Keiretsu). If we are interested in ownership patterns it is because we think they have some impact on corporate behavior--collapsing such significant differences makes little sense at this stage in the research process.

Consider further the classification accorded Electrabel (fig.10), which raises a slightly different question. Something important seems to be missing from a classification scheme that does not distinguish a firm with only one 20% shareholder from a firm like Electrabel with two shareholders holding over 20%. That is, concentration of ownership would seem to be an important question concerning corporate governance (managers' incentives, controlling shareholders' ability to expropriate other shareholders' wealth, etc.). The datum that there are few cases of multiple controlling shareholders serves to highlight this point--we would not expect a rational investor to want to hold a big block in a firm in which she does not also obtain control with the block. And the data largely confirm this intuition (table VII). Consequently, firms that have multiple controlling owners are distinctive and should be separately counted. It would be interesting to know whether some systems are more likely to have multiple block owners than others (are they more present in low or high legal protection regimes).

In their defence La Porta et al. might contend that by classifying ownership as they do they are simply biasing things in favour of Berle and Means' picture of the corporation, but the purpose (or I should say contribution) of their paper is not to test whether Berle and Means' characterisation of the US corporation in the 1930s is correct globally. No one today (nor Berle and Means when they wrote their book), perceives the US pattern of dispersed shareholdings as a universal paradigm. The paper's contribution is to provide detailed evidence of the pattern of controlling ownership worldwide--the general fact that in most countries public firms have controlling shareholders has been well known for a long time. From this perspective, providing more detailed information concerning ownership concentration would be very valuable, as would devising a better means of capturing group connections. Perhaps further interesting ownership patterns across low and high protection regimes would be uncovered if such data were available.

III. Family ownership results

There are several interesting results surrounding the prevalence of family control detailed in the paper, which deserve further exploration; I will only touch upon two. First, a fascinating finding is that between 25 to 35% of these families do not appear to hold management positions in the firms (this is fascinating because my prior was that this would be a null set). Perhaps if slightly lower level executive positions (such as corporate presidents or vice presidents) had been included in the survey this percentage would substantially decline. But assuming that expansion would not bring the proportion down to 0 (I assume it would be 0 if directorships were the test of management), the presence of such firms raise many interesting organisational questions. For instance, is there any relation between ownership share and management position? Are firms with no controlling shareholder-managers more likely to be controlled through pyramids (wherein the controlling shareholders have management positions in the intermediaries rather than the sample firms); do the families in non-management shareholder firms hold smaller control blocks (i.e., is there a positive linear relation between ownership and position) and if so, is this a function of greater or lesser ability of smaller blockholders to extract disproportionate rents; or more broadly, are the benefits accruing to controlling shareholders different in these non-manager firms (i.e., does the dividend study include these firms--I'd like to know if they pay out more dividends than firms in which the family members are managers); are these older companies so the families are third or fourth generation past founders; are, accordingly, these firms' managers "freer" than other firms, that is, do they behave more like the managers of widely dispersed firms than managers who are also shareholders; etc.

The list of questions is endless so I'll stop here and simply note that given the authors' aim of showing the conflation of ownership and control in most corporations, it would be helpful if they offered some explanation of this peculiar finding (even without pursuing the additional data that I've suggested would be interesting to examine). The significant difference in management positions across legal protection regimes is consistent with their explanation of ownership as a function of shareholder protection: the holding of a management position should be less likely in a low protection state as presumably the controlling shareholder is better able to exploit the minority in such states and thus may not need a management position to be able to do so.) But to ultimately convince readers of this hypothesis, a better theory of what are the private benefits of control, and how are they realised, needs to be developed than offered in the paper or than exists at present in the literature.

Second, it is interesting that families are more likely to use pyramids than complex voting structures to exercise control (table IV). La Porta et al. view these data, particularly the significant difference in low protection countries, as consistent with a theory of pyramids identified with Wolfenzon, as enabling controlling shareholders to exploit outsiders regarding the gains from new ventures (p.24). I do not think these data support such an explanation. The firms in this sample that are part of the pyramid are publicly traded firms (private firms are excluded from the definition); is there any evidence that these companies are engaging in new ventures? For the public to be cut out most effectively, the controlling shareholders should use the dividends earned from the public company at the bottom to finance new ventures in private entities up the chain. This is not apparently the case in the pyramids identified in the paper. The evidence suggests to me an alternative more traditional explanation: controlling families are interested in leveraging their cash positions (which is not obtained as effectively with dual class stock as opposed to pyramid structures): they use public dollars through the layers to increase the return on their equity investment at the bottom of the chain. Again, connecting the dividend paper with these firm level ownership patterns would be interesting--do the family-controlled pyramid firms pay out higher dividends than those in which control is exerted through disproportionate voting rights? Such evidence could shed light on whether my leverage hypothesis is correct, compared to the new venture expropriation one (where presumably dividends should not be shared with the public).

Finally, these data raise an interesting issue on the relation between legal protections and corporate structure: could the lower level of corporate than individual ownership found in the study be related to the significance of fiduciary problems from partial ownership depending on the level of shareholder protection. Corporations can use integration (100% owned subs) to avoid such problems in a way that families, by definition, cannot. To get at this issue, we would need some sort of count on the undertaking of business through wholly-owned subsidiaries or divisions (acquisitions might be one way to go); the prediction consistent with the paper's hypothesis would be that the count should be higher in countries with high, rather than low shareholder protection (and family ownership should correspondingly be less in nations or industries where corporate integration is easier to achieve).

IV. Table VIII (Alternative Hypotheses) Variables

In Table VIII the authors consider whether there are other explanations besides the level of legal protection for shareholders that can explain the ownership patterns they find. The table does a good job at rejecting the alternatives, but I had a few suggestions for refining the variable choices: although they are clever, I do not think the proxies selected to test the alternatives are the most persuasive available.

First, I do not think that the comparison in panel A across common and civil law origin countries resolves the endogeneity question. This divide in legal regimes was in place at the beginning of the 20th century (and of course the 19th century for the nations in the sample that have existed that long), but the content of these regimes has changed significantly over the past 90 years (at least for the regime with which I am familiar, the US) and if Berle and Means were correct, the ownership patterns of corporate enterprise have changed considerably over the period as well. To determine whether the civil law countries' low level protection depends exogenously on origin or endogenously on vested political interests opposing change, we ultimately need to know more about the political process--but more realistic empirical variables to get at the question would be either construction of an index of changes in the substantive regime, or, were the data available, observed ownership patterns from a much earlier year. If US Firms 90 years ago were as concentrated as civil law country firms are today and the law was as or more pro-shareholder then than today (e.g., pre-emptive rights were mandatory not optional, self-interested transactions were automatically void, etc.), or if common and civil law regimes changed substantively as frequently over time and in the same direction toward increased protection, than this would weaken the hypothesis that the legal rules determine ownership, rather than it going the other way around (or the view, which I do not espouse, that there is no connection between these two variables).

It would also be useful, given the reliance placed upon the distinction, to think more about why the distinction between a common law and civil law system should matter with respect to shareholder protection? Why should one system (over time) have fewer protections than the other? One possibility is the importance of courts in corporate law, and for shareholder rights in particular. If it is difficult to write down rules guiding conduct in the fiduciary context given the residual nature of the equity claim (as Williamson, 1984 has convincingly argued), then the common law system of case-by-case adjudication, of necessity will better protect shareholders than a civil law system. This feature of the regimes is not, however, captured by the protection index La Porta et al. use (they look at statutory content). Of course, there is a problem with their legal protection index that I and others have noted to them in their earlier work and I have chosen not to discuss in this paper--the enabling nature of corporate law, and the federal system in the US, makes reliance on statutory features misleading (except for the right to bring a shareholder suit) in determining what is the level of protection afforded a particular firm. For example, they treat pre-emptive rights and cumulative voting as shareholder protections that are part of the US legal regime; but virtually no large US Corporation is subject to either rule (the default was changed by the beginning of the 20th century to opt-in rather than opt-out defaults, and old firms that had not opted out eliminated the provisions with the change, while new firms generally do not opt in.)

Another possibility is federalism or political decentralization--the common law countries tend to be federal systems, or systems with decentralised governments (US, Canada, Australia), and among the civil law countries, the better ranking ones, such as Germany, have decentralised political systems compared to the worst, France, the epitome of a centralised state. This political variable no doubt is correlated with the size of the private economy, as political decentralisation is typically accompanied by less government intervention in markets. This regime characteristic presumably is correlated with the size of the stock market, which relates to another explanation considered by La Porta et al. to which I will return shortly in considering the liquidity variable.

Second, I also have some concern regarding the measures used to test the debt financing alternative. The first measure, restrictions on banks' equity investments, does not go to the question, which is whether corporations rely on bank debt instead of equity for capital (i.e., the relevant issue is whether banks have power over the debt, rather than the equity, market). In my opinion, the results in panel B are irrelevant to this hypothesis. The debt/GDP ratio is a better measure, but it is extremely crude (given the immense problem that individual bank debt is included in the calculation). I would prefer, as a test of this hypothesis, a variable more directly connected to the sample firms--their debt-equity ratios. If their hypothesis is correct, that the legal rules are what matters and not the financing regime, then we should find no or a negative correlation between the firms' leverage, ownership and the level of protection.

Third, the cross-holdings test (p. 32) seems misdirected or confused. The basis for the test is the observation that Japanese firms adopted cross-holdings in order to prevent takeovers; the test then compares restrictions on cross-holdings with dispersion of ownership. But since Japan restricts cross-holdings, and Japanese firms have no cross-holdings by this sample's construction, whatever is being tested has little connection with/cannot confirm the underlying hypothesis. A better test might be the presence or absence of statutory restrictions on takeovers, or the number of acquisitive transactions, in relation to cross-holdings or the other ownership variables: the hypothesis would be that there should be more controlling shareholders in countries where firms are more likely to be subject to takeovers.

Moreover, the liquidity measure (GDP per capita), that is a better test of this alternative hypothesis than the cross holding variable, does not seem to be a particularly useful variable; it is, to say the least, a stretch to consider economic development a proxy for stock market liquidity. Indeed, we know a priori that the most developed economies are widely divergent on more conventional market liquidity measures (such as market cap as a percentage of GDP). More direct measures of liquidity seem to be readily available. For instance, why not use market capitalisation as a percent of GDP or the other variables on domestic firms and IPOs used in an earlier La Porta et al. paper (1998 J.Fin.)? Or trading volumes on the national stock exchanges? Or, more specifically, the per capital market capitalisation of the firms in this paper's sample? All of these measures are more closely related to liquidity than economic development. If the economic development variable is useful because firms' longevity is the issue (a justification for the measure offered on p. 32, as proxying for controlling shareholders' interest in selling out), then there is an obvious far better measure: the length of the corporate life of the firms in the sample.

V. Where to next?

Let me conclude with some suggestions concerning other questions to consider exploring with this terrific data set. The most obvious though most difficult question to pursue is the reason why people are interested in comparative corporate governance in the first place: Are there performance differences across ownership type (and country type) in the firms in this sample? This would get at the issue of the extent to which controlling shareholders can expropriate the minority effectively (perhaps we would get better theories of private control if it was measured): are the returns to non-controlling (controlling) shares comparable across high and low protection regimes, in relation to ownership structure? Some subsidiary issues, raised by the discussion in the paper would also be useful to pursue. For example, are there differences in acquisitive activity by country (high protection regime firms acquiring low protection regime firms) (see e.g., p. 36 concerning cross-country investments), or ownership type (less or more diversifying mergers, for instance, by family-controlled firms, going to the question of managerialist-motivated mergers discussed in the US literature); as well as the cross-sectional questions involving family-ownership firms that I mentioned earlier.

A more readily available agenda would be to examine more carefully cross-country ownership patterns within the high and low legal protection groups. Is it random whether a country has, for instance, more family-held firms than other ownership patterns, given a level of protection? Does the legal protection mix vary across the low-protection countries that have more family-controlled firms compared to those with more state-owned firms? Does the enforcement/corruption variable differentiate these countries (i.e., does state ownership predominate in countries where property rights are less likely to be enforced, so that even families are deterred from making significant capital investments)? Along those lines, is there a size difference as well (are the family held firms in the state-dominated countries smaller than in countries with greater legal protection?). This could be a way to distinguish between whether the inadequacy of legal protections, or of enforcement mechanisms, stymies growth.

To conclude, then, there is much more to be mined from the data set that La Porta et al. have carefully constructed for this paper--the opportunities are exceedingly rich and cannot begin to be enumerated here. I look forward to reading many more papers from the authors on comparative corporate governance, as I know, that like this paper, we will learn a great deal from every one of them.

Last updated: April 29, 1999