Quantifying market risk: : - An evaluation of VaR methodologies in thebanking sector

Abstract: In this study some of the most commonly used methods by banks whenestimating the Value-at-risk (VaR) is evaluated. To find the best models forestimating the VaR, the performance of these methods is assessed on the basisof Kupiec’s unconditional coverage test of statistical accuracy. Data from fourworld indexes over an 8-year period are used in the calculations. I find thatthe more newly developed methods; Filtered Historical Simulation andVolatility-Weighted Historical Simulation provide the most accurateestimates. Further, estimates using GARCH volatility seem to have providedbetter estimates than EWMA.