How well do risk measurement models estimate VaR during good and bad times? Evidence from the Korean stock market

This paper presents empirical evidence on the performance of a number of selected risk measurement models for measuring the value-at-risk in the South Korean stock market with regard to their ability to consistently furnish accurate estimated VaR risk measure during good and bad times. The soundness of model performance in the accuracy of estimated VaRs are compared using risk management loss function and likelihood ratio tests for coverage probability to ascertain which of the models can accurately capture market risk over varying market conditions. The results indicate that the Equally Weighted Moving Average model and the RiskMetrics model captures and thus furnish unsteady but accurate estimated VaRs of the index return series, especially during crisis periods. The findings also demonstrate that these models can deliver more accurate estimated VaRs than the widely employed GARCH and historical simulation specifications.