A nonparametric analysis of efficiency and productivity in large U.S. banks

Date of Completion

January 1997


Economics, General|Economics, Commerce-Business|Business Administration, Banking




This dissertation examines the productivity growth, scale efficiency, size efficiency, and the efficiency dynamics of large U.S. commercial banks between 1984 to 1990, using the nonparametric method of Data Envelopment Analysis and statistical and econometric tools.^ We measure productivity growth using Malmquist productivity indexes, and isolate the contribution of technical change, efficiency change, and scale efficiency change to productivity growth. We find that large commercial banks experienced productivity increases at about 3.35% per year on average during this period. Further, productivity expanded between 1984 and 1988, but declined thereafter. Our second stage regressions suggest that commercial banks with more assets or more specialization in outputs experienced higher productivity growth.^ Next we address the issue of optimal size of banks. We examine the scale efficiency of large U.S. banks employing two kinds of significance tests--a parametric test developed by Banker (1993) and the nonparametric Wilcoxon's signed rank sum test that is suitably constructed to test if the level of scale efficiency falls below unity by more than a prespecified degree. The statistical tests reveal significant deviations from constant returns to scale suggesting that it may be worthwhile to correct for scale inefficiencies even after technical inefficiencies are accommodated. We then focus on the banks that are operating under decreasing returns to scale to examine their size efficiency--i.e., whether they should be reorganized as several smaller banks in order to achieve the most efficient utilization of resources. While majority of the banks are not "too large" according to this criterion, there are many instances where restructuring of the banks into multiple units is recommended.^ Finally, we examine the dynamics of efficiency of the banks during this period, within the framework of Concordance analysis, Kolmogorov-Smirnov tests, Convergence analysis, and Markov-Chain analysis. The analyses suggest that while the efficiency rankings of the banks in each year are overall stable, it does accommodate considerable mobility. The transitional probability matrix is stationary for the period 1987 to 1990 though not for the entire period 1984 to 1990, indicating that by 1987 the large banks have adapted reasonably to the changed environment. ^