The Factor–Spline–GARCH Model for High and Low Frequency Correlations
提出一种新方法，结合因子定价结构与GARCH模型，同时刻画股票相关性的高频和低频成分，改善美国股票相关性的拟合与长期预测。
We propose a new approach to model high and low frequency components of equity correlations. Our framework combines a factor asset pricing structure with other specifications capturing dynamic properties of volatilities and covariances between a single common factor and idiosyncratic returns. High frequency correlations mean revert to slowly varying functions that characterize long-term correlation patterns. We associate such term behavior with low frequency economic variables, including determinants of market and idiosyncratic volatilities. Flexibility in the time-varying level of mean reversion improves both the empirical fit of equity correlations in the United States and correlation forecasts at long horizons.