Abstract
This paper compares the estimates of the two most widely used non-structural models for market power measurement in banking, namely the conduct parameter method and the revenue test, as applied to three panels of Greek, Latvian and Spanish banks over the period 1993-2004. We also propose a dynamic reformulation of these models within a panel data context, in order to address possible statistical problems associated with the dynamic nature of bank-level data. The results suggest that both static methods provide lower estimates of market power relative to their dynamic counterparts. Therefore, the inclusion of some dynamics in the models, even though it increased estimation complexity, helped to reveal some collusive behavior of banks. © 2008 Blackwell Publishing Ltd.