<![CDATA[This study aims to analyze and compare the accuracy of three Value at Risk (VaR) models in estimating market risk across four systemic banking stocks in Indonesia categorized under the Bank Group based on Core Capital (KBMI) IV, namely BBCA, BBRI, BBNI, and BMRI. A comparative analysis was conducted by dividing the timeframe into two contrasting market conditions: the crisis period driven by the Covid-19 pandemic (2019-2021), representing a Black Swan event, and the stable period (2016-2018). The VaR calculation methods examined include the parametric, non-parametric (Historical Simulation), and semi-parametric (Filtered Historical Simulation) approaches. To evaluate the consistency and robustness of each model against volatility shifts, a formal Backtesting procedure was performed using the Christoffersen Conditional Coverage joint test alongside a Severity analysis (measuring the magnitude of extreme losses). The test results demonstrate that during the stable period, the conventional parametric and non-parametric approaches were efficient for BBCA and BBNI stocks, as they successfully achieved a valid status while minimizing severity values. Conversely, during the Covid-19 crisis period, both traditional models completely failed due to the phenomenon of risk underestimation. Amidst the crisis turbulence, the semi-parametric FHS model proved to be superior and the most robust, where the FHS-Student-t specification successfully secured validity and minimized the severity of extreme losses for BBRI and BBCA stocks, followed by the performance of the FHS-Normal model on the BMRI stock.]]>
