Board and ownership characteristics across legal families

As said, we start by grouping countries according to the tradition of their legal system (following LLSV, 1996) to see whether we observe substantial differences across families and test if the dissimilarities are statistically significant.

Table 8 shows the descriptive statistics of our ownership and financial variables sorted by legal family. For most variables, we find the English and German families at both ends of the spectrum, while the French and Scandinavian systems would typically lie somewhere in between. The exception would be total assets: while the biggest banks are found in the English law countries, it is Scandinavia who has the smallest, which could perhaps be explained by the smaller country size.

This way, we can see that the banks in the English family present the most dispersed ownership structure (with approximately 13% of the shares being closely held in the average bank), are the best performing ones according to Tobin’s Q (1.079) and return on assets (2.032), and finally, and after Scandinavia (with 20%), have the lowest capital ratio (slightly below 21%), showing also a similar capital structure to French banks (slightly above 21%).

As opposed to this, the German family presents the most concentrated ownership structure (with approximately 66% of closely held shares in the average bank), has the

lowest values for both for Tobin’s Q (1.025) and return on assets (0.866) and shows a very different capital ratio than the other three families of nearly 32%.

Table 9 displays the descriptive statistics of the board variables sorted by legal family. When making comparisons of the variables across systems. we should keep in mind, however, that the way we defined our board variables already removes some of the systemic differences, i.e. the existence of two-tier boards versus unitary boards, by only considering the supervisory board when the board is two-tiered, a circumstance predominant in the German family, possible in the French and Scandinavian families, and inexistent in countries of English law origin.

If we start by looking at the size of the board, we see that banks in the French family, with an average of 14.375 members, have the largest boards; at the other end of the scale we would find the Scandinavian banks with 8.179 members on average.

Board members in the English law countries enjoy the highest average number of positions held (2.88). They are closely followed by directors from the French family (2.82). The lowest values of this variable are observed in the Scandinavian (1.88) and German families (2.437). This could be at least partly explained by the prevalence of employee representatives in the boards in those countries, a type of board member that is included in our variable and it is not likely to serve in other boards or hold relevant positions in other companies.

Concerning the degree of director independence in the board, the German family presents the highest proportion of non-executives sitting on the board (0.754, this makes sense if we think we are only including here the supervisory boards) and the French family has the lowest (0.601). In all cases, boards are on average dominated by non-executive directors. Contrarily, if we use our other measure of independence,

Scandinavian boards would be on average dominated by insiders (the ratio of outside directors is 0.49), and the most independent boards are to be found now in the English family, with an outside directors ratio of 0.64.

The existence of political directors is most frequent in the English family banks (a ratio of 0.056), and least common in the French (0.018).

In order to find out whether these differences in the board and ownership variables across the legal families are statistically significant, we follow the same procedure as in the previous section (see results on Table 10). The ANOVA results (in Panel A of Table 10) confirm the existence of differences across the legal families, even after controlling for bank size. Since the Bartlett’s test makes us doubt that equal variances are present in our data, we run again the Kruskal-Wallis test, that further confirms the statistically significance of the differences between the legal systems (in Panel B of Table 10).

However, we might wonder what the actual power of the legal families is compared to that of nationality in explaining these differences in the corporate governance variables. Table 11 provides the coefficients of determination corresponding to the ANOVA tests of the board and ownership variables across countries and legal families (F-values previously shown in Table 6 and Table 10), as well as the ratios between these two measures of goodness-of-fit. The results presented here indicate that, while nationality is a better determinant of the governance characteristics of banks (accounting for up to 47% of the variation in the case of percentage of closely held shares, when looking at the R-squared), the legal family stands also for a relatively large part of the variability of the variables, particularly in the case of blockholder ownership and board size, where it explains the main portion of the nation effect.

To confirm the validity of these results, we conducted two additional robustness checks. First, we designed a nested ANOVA model to test the effects of the four different types of legal origin on the governance variables, where nationality was nested with legal origin. The results are reported in Table 12. As expected, nationality comes up as the most important factor, but the legal origin still shows up significant to describe blockholder ownership and board size. Second, we ran further ANOVA tests within each legal family sub-sample to test the significance of the country dummies. Aware of the heteroscedasticity problem in our data, we perform Kruskal-Wallis tests to ensure the soundness of the findings. The combined results from these two tests are reported in Table 13 and show that, while the differences across countries are still significant in the Scandinavian family and for some of the board variables in the countries of French origin, once the legal tradition is accounted for we are not able to observe the existence of a nation effect in the English and German families, nor in the case of the ownership structure and ratio of political directors in the French group. These results provide extra support to the hypothesis of the relevance of the legal systems to explain international differences in corporate governance.

Therefore, given that the legal families seem to capture an important part of the nation effect observed in board and ownership structures, we will include them in the following regressions analysis, as a way of controlling for the different institutional environments.