Using Conditional Autoregressive Range Model to Forecast Volatility of the Stock Indices

Abstract

This paper compares the forecasting performance of the conditional autoregressive range (CARR) model with the commonly adopted GARCH model. Two major stock indices, FTSE 100 and Nikkei 225, are studies using the daily range data and daily close price data over the period 1990 to 2000. Our results suggest that improvements of the overall estimation are achieved when the CARR models are used. Moreover, we find that the CARR model gives better volatility forecasts than GARCH, as it can catch the extra informational contents of the intra-daily price variations. Finally, we also find that the inclusion of the lagged return and the lagged trading volume can significantly improve the forecasting ability of the CARR models. Our empirical results also significantly suggest the existence of a leverage effect in the U.K. and Japanese stock markets.