Most bank efficiency studies that use stochastic frontier analysis (SFA) employ each bank's own implicit input price when estimating efficient frontiers. But the theoretical foundation of most studies is a cost minimisation and/ or profit maximisation problem assuming perfect input markets. At the very least, traditional input price proxies therefore contain substantial measurement error. In this paper, we examine the magnitude and direction of this error in cost and profit efficiency (CE and PE) measurement. We suggest two input market definitions to approximate exogenous input prices alternatively and estimate CE and PE of German banks between 1993 and 2003. Our main findings are threefold. First, after accounting for systematic differences across banks, mean CE is sensitive to alternative input prices. Second, distortions of mean PE due to traditional input prices are small. Third, across CE models small cooperative banks located in large western states are identified as top performers. Large banks and those located in eastern states rank lowest.