EMPIRICAL RESULTS FROM VAR PREDICTION USING PEARSON?S TYPE IV DISTRIBUTION

Two most important characteristics of equity returns time series data are volatility clustering and non-normality. GARCH model has been widely used to forecast dynamic volatilities and hence has been used for value-at-risk (VaR) estimation. (Bhattacharyya et al 2008) has developed a new VaR estimation model for equity return time series using a combination of the Pearson?s Type IV distribution and the GARCH(1,1) approach which showed superior predictive abilities. This new model was tested on indices of eighteen countries [3] on daily return up to March 1st, 2005. In this project, we replicate the results in [3], and test the model for its predictive power over a more volatile period (i.e. 350 trading days prior to July 18th, 2008). We backtest the validity of the VaR estimations and compare the predictive power of this model over both of the above time periods on indices of eight countries. We discover that the Pearson?s type IV model still remains a good predictive ability during the more volatile period