Multidimensional Surveillance of the Indian Banking System: A Cluster Approach

Abstract

This paper develops a multidimensional risk ranking model for bank supervision, using the k-means cluster approach. It employs a combination of size, balance sheet, and market-based indicators for predicting idiosyncratic and systemic risk. The risk rankings are benchmarked to a long-run threshold, which regulators may wish to target for the resolution of financial crises. When the tool is applied to data on Indian banks between 2005 and 2023, several important results emerge. The effectiveness of different signals depends on the nature of the impending financial stress event. Market-based indicators are better at predicting external shocks, while markers related to asset size forecast credit booms and busts with greater efficiency. Moreover, the model is superior to heuristic regulatory measures of bank-specific distress and its resolution. The framework is also able to distinguish between the risk performance of public and private sector banks, during a period that spans the global financial crisis (GFC), the non-performing asset (NPA) crisis in India, and the COVID pandemic. Private banks exhibit better risk profiles during the GFC and NPA crises. This study emphasizes a multifaceted approach to bank supervision, in order to capture the heterogeneity of financial crises.