LAUSR

ABSTRACT This article develops a model that can accurately forecast the volatility of Taiwan stock returns and efficiently estimate value-at-risk (VaR). Because the volatility in the Taiwan stock market has been shown to die down and shift quickly, we find that the model able to outperform others is one that allows the parameters of the volatility models to switch between regimes and conditional volatility to revert quickly to near-normal levels following extremely volatile periods. Compared with nested models, this model has the best performance in terms of the statistical fit of in-sample data and out-of-sample volatility forecasts and VaR estimates.