Comments

Arturo Bris: The impact of financial liberalization on growth is an empirically challenging question. The economic theory predicts an unequivocally positive effect of liberalization on growth. Identifying such an effect, however, is not so simple. First, financial liberalization policies are commonly coupled with other changes in the functioning and regulation of the financial system. A good example of this is Italy, where a general overhaul of the securities markets in 1991 brought in new legislation on takeovers, insider trading, market liberalization, derivatives trading, and trading systems, among others. Isolating the effects of liberalization is a hard task in such an environment. Second, as the authors clearly state in the article, it is difficult to argue that the causality between liberalization and growth is unidirectional. Any cross-sectional regression would suffer an endogeneity problem.

Galindo, Micco, and Ordoñez circumvent these difficulties by using the methodology pioneered by Rajan and Zingales. ¹ The explicit assumption here is that financial liberalization fosters more growth in industries that are more financially dependent. By regressing a measure of industry growth on the product of a financial liberalization index and a measure of industry's financial development, the authors conclude that more liberalized countries grow faster. Moreover, because they use a panel of industry-country-year observations, they can control for fixed effects—like the effects of other regulatory changes.

My comments are organized as follows: I begin with some methodological issues and then comment on the interpretation of the results.

Methodological Issues

An observation in this article is an industry-country-year. This is how the authors are able to control for other institutional changes at the country level. In equation 2, for instance, there are twenty-eight countries times twenty years (approximately), which equals 560 country-year dummies that perfectly capture the effect of specific policy measures different from financial liberalization. To avoid multicollinearity, the regression must not include any other country-year specific variable (such as, say, the inflation rate), and the financial liberalization index must therefore be interacted with a measure of industry-specific financial dependence. This is an extremely intelligent approach. The only industry-specific variables used in the estimation, however, are the index of the industry's financial dependence and industry fixed effects. The index of financial dependence is taken from Rajan and Zingales; it is therefore estimated with U.S. data. ² The industry fixed effects, in turn, are, by definition, equal across countries. Consequently, the regressions in the paper do not have any country-industry specific control. What if the semiconductor industry grows consistently more in Taiwan than it does in France? Or, since the paper and pulp industry is an export sector in Finland, should one expect the effect of financial liberalization on the industry's growth to be stronger in Finland than in the United States?

A second issue has to do with the interaction with the corporate governance variables. The authors argue that because the correlation between the corporate governance variables and the financial liberalization index is high, it is sensible that both variables become jointly insignificant at explaining growth. This is why, they say, individual t statistics are not reliable, and they instead use a test of joint significance. The test of joint significance rejects the hypothesis that both coefficients are zero. However, it is not convincing evidence that the corporate governance measures are significant alone. Table 3 then splits the sample depending on whether the industry (country) has an index of investor protection higher or lower than the sample median. As the authors recognize, the evidence is not conclusive.

Finally, a comment on the analysis of financial liberalization and the efficiency of the domestic financial system. The regression in table 5 uses credit to the private sector—a measure of financial development—and the financial liberalization index, as independent variables. The issue of multicollinearity is important here, but it is not acknowledged. My suggestion is to instrument either one, or else to orthogonalize the financial development index with respect to the financial liberalization...