Wavelet based approach for exchange rate portfolio value at risk estimation

With the accelerating level of global integration, the volatilities across exchange markets are co-moving with higher level of fluctuation as well as more complicated dynamic inter correlations, which are key to the deeper understanding and proper measurement of risk. Thus, we propose the multivariate wavelet based Value at Risk estimation algorithm to account for the multiscale dynamic correlation characteristics as the new stylized facts. The multivariate wavelet analysis unveils the time varying correlations over different time horizons, which corresponds to the regime switching across different time horizons. The incorporation of this information during the modeling process leads to the advanced semi parametric model, consisting of mixture of models of different specifications and parameters at different scales. Empirical studies using the proposed approach to investigate the Chinese RMB and Europe Euro as the closely related exchange markets have shown evidence of multiscale dynamic correlation characteristics. The proposed approach has demonstrated the improved performance in the VaR forecasting exercise.