Forming appropriate peer groups for bank research: a cluster analysis of bank financial statements

Abstract

Choosing appropriate peer groups for commercial banks is important to investors comparing bank performance, for regulators evaluating safety and soundness, for bank management looking at merger alternatives or relative performance, and for bank researchers testing hypotheses and making policy judgments about the banking system. We use commercial bank financial statements with common size variables as the inputs to a cluster analysis model to identify clusters or groups of banks with financial structures that are relatively homogeneous within groups and distinct across groups. Managerial strategies and idiosyncrasies, local and global economic conditions, and the regulatory environment shape bank financial statements, and financial statements should reflect the financial and operational differences across banks. Using year-end data from 2014, we cluster 6444 banks into several such groups. Our results show that bank clusters are formed largely around loan types, funding differences, and management’s strategic choices. We compare the ability of bank clusters and bank size to explain several widely used measures of bank performance and risk in additional years. These bank clusters are shown to have substantially greater explanatory power in regression models when compared to groupings based on bank size in several different years.