Forecasting Correlation and Covariance with a Range-Based Dynamic Conditional Correlation Model
This paper proposes a range-based Dynamic Conditional Correlation (DCC) model, which is an extension of Engle’s (2002a) DCC model. The efficiency of the range data in volatility estimation is documented in Parkinson (1980), Alizadeh, Brandt, and Diebold (2002), Brandt and Jones (2002), and Chou (2004a, b), among others. It is hence natural to consider the implication of this result in the estimation of multivariate GARCH models. In the DCC model, the conditional correlation coefficients are estimated by a dynamic model for the product of the pair-wise return series with each normalized by their conditional standard deviations. The conditional standard deviation is calculated by using a univariate GARCH for the return series.We use the Conditional Autoregressive Range (CARR) model of Chou (2004a), as an alternative to the univariate GARCH in the DCC first-step estimation. We therefore construct a range-based DCC model. The substantial gain in efficiency in the volatility estimation can induce an efficiency gain in the estimation of the series of the time-varying correlation coefficient and covariance. For comparison we estimate the generalized return-based DCC model as a benchmark to gain insights into the difference of these methods. We use three data sets for empirical analyses: the stock indices of S&P500 and Nasdaq, and the 10-year Treasury bond yield. Both in-sample and out-of-sample results indicate that our argument is supported in terms of the precision in estimating and forecasting the correlation and covariance matrices.
DCC CARR range dynamic correlation covariance volatility
Ray Y Chou Nathan Liu Chun-Chou Wu
Institute of Economics, Academia Sinica Department of Management Science, National Chiao-Tung University Department of International Trade, Chung Yuan Christian University
国际会议
昆明
英文
2005-07-05（万方平台首次上网日期，不代表论文的发表时间）