Estimating portfolio value-at-risk via dynamic conditional correlation MGARCH model – an empirical study on foreign exchange rates

Abstract

This study compares efficiencies of five Generalised Autoregressive Conditional Heteroskedasticity (GARCH) models in terms of value at risk (VaR) backtesting on the number of prediction failures and the average deviation between VaR and realized return series. Unlike the previous literature which presumes constant correlation coefficients, a new model proposed by Engle (2002, the DCC model) is applied to highlight time-varying conditional correlations amongst positions, which is essential for portfolio risk management. From the empirical studies of exchange rates data including the US Dollar to British Pound, Japanese Yen and Euro Dollar, we find that the DCC model produces least prediction failures.