A comparative approach of stochastic frontier analysis and data envelopment analysis estimators: evidence from banking system

Purpose The consistency of stochastic frontier analysis (SFA) and data envelopment analysis (DEA) cost efficiency measures using a sample of 650 commercial and domestic banks in the United States is investigated based on cluster analysis while accounting for the yearly variation in banks. Design/methodology/approach Due to the importance of efficiency measures for policy and managerial decision-making, the cost efficiency measures of SFA and DEA estimators are examined according to four criteria: levels, rankings, stability over time and stability over clustering groups. In this paper, we present two clustering methods, Gap Statistic and Dindex, that involve SFA and DEA cost efficiency measures. The clustering approach creates homogeneous groups of banks offering a similar mix of efficiency levels. Hence, each evaluated bank knows the cluster to which it belongs. Furthermore, this paper provides nonparametric statistical tests of SFA and DEA cost efficiency measures estimated with and without a clustering approach. Findings The results suggest that the clustering approach plays a considerable role in the rankings of US banks. Furthermore, the average SFA and DEA cost efficiency measures over time of the homogeneous US banks are substantially higher than those of the heterogeneous US banks. Originality/value This research is the first to provide comparative efficiency measures needed for desirable policy conclusions of heterogeneous and homogeneous US banks.