本研究討論CARR(Conditional Auto-Regressive Range)模型的經濟涵義及其性質,並以台灣發行量加權股價指數做為主要的研究對象,在週資料與日資料的基礎上,分別進行CARR模型及GARCH模型在波動性預測能力之比較。實證結果顯示,不管是樣本內及樣本外,在週資料的預測評比上皆得到CARR模型優於GARCH模型的結果,此與Chou(2003)利用S&P500指數所進行的研究結論具有一致性,而為了強化CARR模型的一致性,本文亦針對台灣店頭市場交易指數資料,進行頑強性驗證,二者皆支持CARR模型的可適用性。同時,股票市場中常見的槓桿效果也在本文的實證研究下獲得證實。
ARCH/GARCH family models have become popular in forecasting volatilities since the 1980's. In this paper we compare the empirical performance of the CARR model by Chou (2003) with the GARCH model. The CARR model effectively provides a dynamic structure for the range data which is more informative than conventional standpoint. Using the Taiwan Stock Exchange Capitalization Weighted Stock Index, the CARR model outperforms than GARCH model both in in-sample and out-of-sample forecasts of weekly stock market volatilities. Our results are consistent with that of Chou (2003) where the CARR model has better forecast abilities than the GARCH model based on S&P500 Index data. We also find significant evidence of the existence of a leverage effect in the Taiwan stock market.