Robust Nonparametric Estimation of Efficiency and Technical Change in U.S. Commercial Banking

This paper examines the performance of the U.S.~commercial banking industry over 1984--2002. Rather than measuring performance relative to the unknown (and difficult-to-estimate) boundary of the production set, performance for a given bank is measured relative to {\it expected} maximum output among $m$ banks using no more of each input than the given bank. This approach permits fully non-parametric estimation with $\sqrt{n}$-consistency, avoiding the usual curse of dimensionality that plagues traditional non-parametric efficiency estimators. The resulting estimates are robust with respect to outliers and noise in the data