As the dominant provider of payments services, the efficiency with which the Federal Reserve provides such services is an important public policy issue. This paper examines the productivity of Federal Reserve check-processing offices during 1980–1999 using non-parametric estimation methods and newly developed methods for non-parametric inference and hypothesis testing. The results support prior studies that found little initial improvement in the Fed's efficiency with the imposition of pricing for Federal Reserve services in 1982. However, we find that median productivity improved substantially during the 1990s, and the dispersion of productivity across Fed offices declined.