Comment on "What Drives Bank Competition? Some International Evidence"

The question of bank competition is vitally important for a number of reasons. The essential role of bank credit and other financial services as an input in the production of most other goods and services places banks in a unique and influential position, such that any allocative inefficiency or other market distortions in banking are almost certain to be felt throughout the economy. Moreover, recent history has provided numerous instances where the textbook paradigm of atomistic competition has proven inadequate as a policy guide for efficient banking—either because there are simply too few banks to rely on sheer numbers as a guarantee of vigorous competition, as in Canada; or because of evidence that there can be such a thing as "too much competition" in banking, as suggested by several studies; or because of instances where nearly competitive pricing has been observed in markets containing only one or two banks; or because, conversely, substantially noncompetitive pricing has sometimes been deduced in banking products—such as credit cards—with thousands of suppliers.

Thus, the study of bank competition remains an urgent field of research. Claessens and Laeven (2004, this issue of JMCB) make two key contributions here. First, they extend a proven empirical method to an unprecedentedly large and varied cross-country sample. Second, they offer the logical and policy-relevant additional step of seeking to identify factors associated with variations in measured conduct. This step is needed to assess the empirical validity of the traditional structure-conduct-performance paradigm in their sample and is especially important where that paradigm is found lacking as a predictor of bank conduct. Indeed, although the authors find evidence that contestability is significantly related to more competitive bank conduct, they find no evidence that market concentration per se is reliably associated with anticompetitive conduct. Clearly, such a finding—which is consistent with a growing body of empirical research over the past decade or two—suggests that competition policy in the financial sector needs to be adequately sophisticated and comprehensive to reflect that reality.

As cogently explained by the authors, these results speak for themselves once you understand the model. Therefore, in the following remarks, I shall focus primarily on the model—providing additional background and pointing out further strengths and limitations of the empirical method.

1. The Empirical Model

As noted in the article, the Rosse-Panzar approach used here is one of two important methods of measuring competition in the manner of the "new empirical industrial organization" (NEIO) literature. Both methods, the Rosse-Panzar reduced-form revenue model and the Bresnahan-Lau markup model, can be formally derived from profit-maximizing equilibrium conditions. This is their great analytical strength and their primary advantage over earlier, heuristic approaches. Not surprisingly, since both models follow directly from first-order conditions, their test statistics are systematically related to each other, as well as to additional measures of competition such as the Lerner index, under certain conditions (Shaffer 1983).

As in most situations, the choice of technique involves tradeoffs. The Bresnahan-Lau model does not require firm-specific data, and its estimated conduct parameter maps directly into any static or dynamic oligopoly equilibrium concept (see, e.g., Worthington 1990) as well as providing an estimate of the percentage distortion in output quantities relative to the competitive benchmark. It can even cope with monopsony power without revising the model or data (Shaffer 1999). Perhaps its foremost advantage is a known, easily implemented set of conditions sufficient to establish econometric identification of the conduct parameter (Lau 1982). On the other hand, the Bresnahan-Lau model can exhibit an anticompetitive bias if the sample fails to span complete markets (see Note 7 of Shaffer 2001). This drawback may cause serious problems in cross-country samples, if the researchers assume that a small, open economy constitutes a single market and fail to correct for cross-border competition in estimating market demand.

The Rosse-Panzar test, by contrast, is robust to the extent of the market. Indeed, because the core model involves...