Asymmetry dynamic volatility forecast evaluations using interday and intraday data

Abstract

The accuracy of financial time series forecasts often rely on the model precision and the availability of actual observations for forecast evaluations. This study aimed to tackle these issues in order to obtain a suitable asymmetric time-varying volatility model that outperformed in the forecast evaluations based on interday and intraday data. The model precision was examined based on the most appropriate time-varying volatility representation under the autoregressive conditional heteroscedascity framework. For forecast precision, the evaluations were conducted under three loss functions using the volatility proxies and realized volatility. The empirical studies were implemented on two major financial markets and the estimated results are applied in quantifying their market risks. Empirical results indicated that Zakoian model provided the best in-sample forecasts whereas DGE on the other hand indicated better out-of-sample forecasts. For the type of volatility proxy selection, the implementation of intraday data in the latent volatility indicated significant improvement in all the time horizon forecasts.