Roberto Rigobon: Alberto Moel's paper studies a fascinating question: what is the impact on the development of a local market when domestic firms decide to issue ADRs? From the theoretical point of view, the answer to this question provides evidence on how markets evolve. Alberto's results do not support the innovation spiral advanced by Merton, but rather favor an alternative explanation that, I argue below, is consistent with so-called flight-to-quality effects. I applaud Alberto's effort to provide a first pass to this engaging question.
The goal of the paper is to determine whether listing an ADR improves the conditions for the other local firms. Alberto studies the impact on four dimensions: the degree of transparency (or openness); the possibility for foreigners to invest locally; overall liquidity; and the growth of the market. Surprisingly, his results indicate that the effects are mixed. Listing a new firm improves only the degree of transparency and deteriorates the other three measures. My prior assumptions were in line with the innovation spiral theory, which states that the competition for domestic assets by foreigners should encourage the development of local markets. Reading Alberto's paper has changed my view of the problem. This is not a reflection of how weak my prior assumptions were, but of how robust Alberto's evidence is.
Because this is a first pass to this question, it is subject to several critiques. I concentrate my comments on methodological issues and provide avenues for improving some of the results. Therefore, my comments are far more negative than my overall reaction to the paper.
The comments are organized as follows. First, I discuss some empirical issues with regard to the regressions Alberto runs. Except for the liquidity measures, the regressions should be taken cautiously. I include some possible solutions for the problems. Second, I discuss an alternative interpretation to his results and offer additional dimensions to explore in the future. Finally, I conclude. [End Page 258]
The results of the paper are based on three separate regressions: how transparency rules are affected by ADR listing, how market growth variables are influenced by the listings, and how liquidity changes with the issue of an ADR. Of these three regressions (or measures), the liquidity results are the most interesting and robust. The other two regressions are relatively weaker, and their conclusions should be taken cautiously.
For example, the results of the openness (or transparency) regressions could be explained by endogenous bias problems. In the growth regressions, the definition of the variables could create spurious correlations that have not been fully addressed in the implementation. These problems are not present in the liquidity regressions (or at least they are not very impor-tant). Hence, the liquidity regressions are more convincing than the other two.
In this section I discuss the problems with the openness and growth results, and offer some corrections that might solve them or diminish their impact.
Transparency and Openness Regressions
Alberto essentially estimates the following specification
where LISTNUM is substituted by other measures of the intensity of ADRs in some of the regressions.
Alberto finds that listing is associated with an improvement in accounting standards. As even he argues, however, the decision to issue an ADR is endogenous. This result could thus be driven by reverse causality. In particular, assume that underdeveloped markets have the advantage that domestic firms have to invest relatively low effort to comply with accounting standards. The disadvantage is the limited and perhaps costly access to capital. If firms anticipate that transparency rules will be improved in the future, the advantage of issuing in local markets is reduced. Thus more firms issue ADRs in anticipation of these changes in the local regulation. In this context, more ADRs are associated with improvements in accounting standards mainly because the openness drives the listing decision. [End Page 259]
This implies that in the previous regression, a positive β1 could be the result of reverse causality. Alberto tries to deal with this complaint by showing the results of Granger causality tests. However, given that in this case it is difficult to assume that the residuals are not serially...