Petros Dellaportas (University College London)

We deal with the problem of identifying and predicting common risk factors that drive the probabilities of jumps in stock daily returns. The stochastic volatility model combined with Poisson-driven jumps is used for modelling the time evolution of the returns. To capture the dependence of the jumps over time and across stocks we model the unobserved intensities of the Poisson process by using a dynamic factor model. We develop an efficient Markov chain Monte Carlo (MCMC) algorithm to estimate the parameters and latent states of the proposed model. We compare its predictive performance with the corresponding performance of existing models. Our methodology is tested on simulated data and is applied on the daily returns of the 600 stocks of the STOXX Europe 600 Index, observed over the period of 2007-2014.