A New Approach for Forecasting the Price Range with Financial Interval Valued Time Series Data

Abstract

This paper proposes an intervalbased methodology to model and forecast the price range or rangebased volatility process of financial asset prices. Comparing with the existing volatility models, the proposed model utilizes more information contained in the interval time series than using the range information only or modeling the high and low price processes separately. An empirical study of the U.S. stock market daily data shows that the proposed intervalbased model produces more accurate range forecasts than the classic pointbased linear models for range process, in terms of both insample and outofsample forecasts. The statistical tests show that the forecasting advantages of the intervalbased model are statistically significant in most cases. In addition, some stability tests have been conducted for ascertaining the advantages of the intervalbased model through different sample windows and forecasting periods, which reveals similar results. This study provides a new intervalbased perspective for volatility modeling and forecasting of financial time series data.