Identification of Banking Stock Risk Factors through Stochastic Search Variable Selection in a CoVaR Models Based on Quantile Regression and Quantile Autoregressive

Abstract

The stability of the banking sector is crucial for maintaining economic balance, particularly in Indonesia where banks play a central role in the financial system. Conventional risk measures such as Value-at-Risk (VaR) mainly capture individual bank risk and are limited in assessing systemic risk arising from interbank spillovers. This study proposes an integrated systemic risk framework that combines Quantile Autoregressive (QAR) based VaR estimation with Conditional Value-at-Risk (CoVaR) derived from quantile regression, while incorporating Stochastic Search Variable Selection (SSVS) to identify key risk factors. The QAR approach accommodates asymmetry and heavy-tailed characteristics of bank return distributions, whereas CoVaR measures the conditional impact of bank distress on the overall financial system.  The SSVS is implemented within a Bayesian framework to select significant market and macroeconomic variables based on posterior inclusion probabilities. Model performance is evaluated using the Kupiec Proportion of Failures (POF) test. The results show that QAR-based VaR effectively captures tail risk at the 5% and 1% quantiles. CoVaR estimates reveal heterogeneity in systemic risk exposure, with medium-sized and digital banks exhibiting greater sensitivity to systemic stress than large banks. Overall, the CoVaR–SSVS model demonstrates superior validation performance and estimation stability compared to the conventional CoVaR approach.