Estimation and Forecasting of Stock Volatility with Range-Based Estimatons

摘要

This paper examines the estimation and forecasting performance of range-based volatility estimators for stocks, with two-scales realized volatility as the benchmark. There is evidence that the daily range-based estimators provide an efficient and low-bias alternative to the return-based estimators. These are not downwardly biased in the presence of negative autocorrelation and low liquidity, as generally suspected. The drift is a major cause of the poor performance of Parkinson's estimator. The forecasts of volatility with these estimators are about as efficient as those with the benchmark itself but are more biased. The forecasts based on realized range are only marginally better on the criterion of bias and are about as efficient. Considering their simplicity and lower data requirement, the daily range-based estimators appear to be more desirable. These results are particularly relevant for the option valuation and the risk management of derivative markets.