Bank-level estimates of market power

The aim of this study is to provide an empirical methodology for the estimation of market power of individual banks. The new method employs the well-known model of Panzar and Rosse (1987) and proposes its estimation using the local regression technique. Local regression yields coefficient estimates equal to the number of observations and, thus, market power is estimated for each bank at each point in time. In addition, a number of restrictive assumptions regarding the properties of the production function of banks are relaxed. A panel of banks from transition countries that has been recently employed by Delis (2010) to obtain market power estimates using the Panzar and Rosse model at the country level is used for comparative purposes. We find that country averages of the bank-level results exhibit a very close relationship with standard, industry-level Panzar-Rosse estimates. However, the empirical results suggest that many banks in countries with fairly competitive banking systems deviate significantly from the country averages and that market power varies substantially across banks in each country.