Semiparametric GAS volatility models with fat-tailed asymmetric kernels

Hudson Torrent (UFRGS)

Abstract: Models for time varying volatility are very popular in statistics and economics as parsimonious descriptions of a range of empirical stylized facts. Blasques, Ji and Lucas (2016) [BJL] provide a new semiparametric model for time-varying volatility in which the form of the error distribution is directly linked to the volatility dynamics using the generalized autoregressive score (GAS) dynamics. BJL model estimates the conditional density of the innovations non-parametrically, and uses the estimate to construct both the semiparametric maximum likelihood estimator, as well as the dynamics of the time varying volatility using the score of the density estimate. This paper contributes to literature investigating the use of fat-tailed and asymmetric kernels in the BJL estimation methodology.

Blasques F., Ji J., Lucas A. 2016. Semiparametric score driven volatility models. Computational Statistics and Data Analysis 100, 58-69.