Forecasting Financial Volatilities with Extreme Values: The Conditional Autoregressive Range (CARR) Model
提出条件自回归极差模型，用资产价格的高低价差预测波动率，基于S&P500数据证明其预测比标准GARCH模型更精准。
We propose a dynamic model for the high/low range of asset prices within fixed time intervals: the Conditional Autoregressive Range Model (henceforth CARR). The evolution of the conditional range is specified in a fashion similar to the conditional variance models as in GARCH and is very similar to the Autoregressive Conditional Duration (ACD) model of Engle and Russell (1998). Extreme value theories imply that the range is an efficient estimator of the local volatility, e.g., Parkinson (1980). Hence, CARR can be viewed as a model of volatility. Out-of-sample volatility forecasts using the S&P500 index data show that the CARR model does provide sharper volatility estimates compared with a standard GARCH model.