Title: Regimes in stochastic volatility
Authors: Alessandro Rossi - European Commission, Joint Research Centre (Italy) [presenting]
Christophe Planas - The European Commission (Italy)
Eduardo Rossi - University of Pavia (Italy)
Abstract: The regime-switching behavior in the stock market volatility is analyzed via a Markov switching specification of the log-stochastic volatility. The model is based both on realized measures and daily returns. The formers can biased due to the presence of microstructure noise but are more informative regarding the true latent integrated volatility whereas the latter even if less subject to microstructure noise have less information on the true volatility. The model is parameterized such that it can be considered as a dynamic mixture model, i.e. a conditionally Gaussian state space model. Dynamic mixture models typically include a continuous unobserved state vector, some discrete latent variables that control discontinuities or change-points, plus the model parameters. The results show that allowing for Markov-switching specification in the stochastic volatility process improves the fitting properties as long as the model's forecasting performance.