Comparison of VaR Methods : The Case of Indian Equities

Abstract

Different approaches to calculate VaR are based on different assumptions. This study dealt with a comparative evaluation of four Value-at-Risk models namely, historical VaR, normal VaR, GARCH (1,1) VaR, and volatility weighted historical simulation (VWHS) VaR in terms of their prediction accuracy for an active portfolio of Indian equities. Daily NAVs of 34 Indian equity growth mutual fund schemes for a period of 10 years were used to calculate 95% VaR and backtest the results using Kupiec's POF test for all four VaR models. To identify the better performing VaR methods accurately, the analysis was performed in two phases : pre-crisis analysis and post crisis analysis. We concluded that there was a significant (insignificant) difference in performance of different VaR models if market conditions during VaR calculation and VaR backtesting periods were in contrast (congruence) to each other. The study found VWHS to be a better methodology for measuring VaR of an active portfolio of Indian equity stocks in both phases of the analysis. The results are relevant for traders & retail and institutional investors who hold stocks of Indian companies in their portfolio and need to calculate VaR as a measure of market risk for their positions.