Clustering Heterogeneous Commercial Bank Lending Models

Abstract

This research aims to uncover commercial bank business models through cluster analysis. Detailed regulatory snapshots of bank balance sheet data are used to derive a small set of bank lending features for clustering. These snapshots can be considered chronologically to better understand trends in the banking industry. For instance, the number of commercial banks has dramatically declined from 1980 to 2013 (the window of analysis). In addition, commercial bank business models have become more uniform or homogeneous with respect to lending activities. Characterizing banking business models is part of a larger research initiative focusing on agent-based modeling and simulation of the Federal funds market, as well as the detailed inter-bank lending market. Understanding banking business models is key to populating the agent ecosystem, implementing better behavior models and interpreting the simulation results. Unsupervised learning methods such as cluster analysis are a natural fit for this business model discovery challenge.