Uncovering the Skewness News Impact Curve

Abstract:

We propose and evaluate a flexible method to model the dynamics of conditional skewness. The method uses the partially nonparametric model of Engle & Ng (1993, The Journal of Finance 48, 1749{1778) who uncover the news impact curve (NIC) for volatility. The model is estimated and analyzed on series of daily returns on major stock indexes. We find that past returns may impact skewness in the way that sharply differs from those proposed in earlier literature. In particular, the NIC for skewness is nonlinear and non-monotonic. We also run simulation experiments to examine how well the estimation procedure identifies parameters of the NIC. Finally, we show that parametric models for skewness typically used in the literature are unable to capture the skewness dynamics found in our empirical study.