Markov Chain Monte Carlo methods for Generalized Stochastic Volatility Models

This paper is concerned with Markov chain Monte Carlo based Bayesian inference in generalized models of stochastic volatility defined by heavy-tailed student-t distributions (with unknown degrees of freedom) and covariate effects in the observation and volatility equations. A simple, fast and highly efficient algorithm, that builds on Kim, Shephard and Chib (1998), is developed for estimating these models. Computation of the likelihood function by a particle filter is considered as are methods for constructing diagnostic measures and the model marginal likelihood. The techniques are applied in detail to daily returns on the S&P 500 index and to weekly changes in the short-term interest rate.