Hybrid LASSO-Quantile Regression and Support Vector Regression for Estimating Conditional Value-at-Risk of Banking Stock Returns in Indonesia

Fluctuations in banking stock prices in Indonesia in recent years have emphasized the importance of comprehensive risk analysis. This tudy examines the individual and systemic risks of 15 Indonesian banking stocks from June 2018 to January 2025, considering key events such as the COVID-19 pandemic and global trade tensions. Stocks such as BBCA and MEGA exhibit wide price ranges, reflecting their large market capitalization. Risk estimation is conducted using Value-at-Risk (VaR) and Conditional Value-at-Risk (CoVaR) with the Hybrid LASSO-Quantile Regression-Support Vector Regression (LASSO-QR-SVR) approach. VaR is calculated based on stock returns, allowing risk comparisons across banks on a standardized scale. The LASSO-QR model is applied to select relevant variables that influence each bank's VaR. Results show that stocks such as ARTO and BBHI exhibit higher potential losses, as evidenced by larger return variability. Selected variables are then used as inputs for CoVaR modeling via Support Vector Regression (SVR) with a rolling window to capture daily systemic risk dynamics. The accuracy of the risk estimates is validated through backtesting using Expected Shortfall (ES) and Kupiec Test, that providing a more comprehensive assessment of extreme risk prediction accuracy. This hybrid approach offers valuable insights for systemic risk management.