Abstract Mixed results have been documented for the hedging performance of multivariate generalized autoregressive conditional heteroskedasticity (GARCH) models. This paper investigates this issue by evaluating representative models (Baba−Engle−Kraft−Kroner [BEKK], generalized orthogonal GARCH [GO], and dynamic conditional correlation [DCC]) that use different approaches to construct the conditional covariance matrix. Using daily data on equity index spot and futures, it is found that the GARCH hedging models have different levels of effectiveness, from both risk-minimizing and utility-maximizing standpoints. Overall, the empirical evidence favors the GO model for hedge ratio estimation, because this specification enables the modeling of the return process with flexible dynamics.

中文翻译：

---

评估多元GARCH模型的对冲表现

摘要多元广义自回归条件异方差（GARCH）模型的套期保值性能已被证明是混合结果。本文通过评估使用不同方法构造条件协方差矩阵的代表性模型（Baba-Engle-Kraft-Kroner [BEKK]，广义正交GARCH [GO]和动态条件相关[DCC]）来研究此问题。使用有关股指现货和期货的每日数据，发现从风险最小化和效用最大化的角度来看，GARCH套期保值模型具有不同的有效性水平。总体而言，经验证据更倾向于使用GO模型进行套期保值比率估计，因为该规范使建模过程具有灵活的动态性成为可能。