The implications of latent technology regimes for competition and efficiency in banking

Banks continue to di er in many ways, for instance with respect to business models, growth strategies, or nancial health. Neglecting these di erences confuses ine ciency with heterogeneity while sub-sample estimation prohibits e ciency comparisons across di erent samples. We use a latent class stochastic frontier model to estimate simultaneously multiple technology regimes and group membership probabilities. The latter are conditioned on six bank traits of German banks and we identify four signi cantly di erent technology regimes. Only small, retail focused banks exhibit cost ine ciencies, which are 5.4% on average and thus substantially lower compared to previous studies. We use technology regime speci c cost parameters to measure competition with Lerner indices. Large, national universal banks and the smallest, most specialized banks exhibit the lowest level of competition. In turn, medium sized universal banks are both e cient and exhibit the lowest Lerner margins between 1994 and 2004.