Dimitris Korobilis

My PhD thesis deals with model selection issues arising in multivariate Macroeconometric models. In my first paper, I develop automatic Bayesian model selection/averaging methods in a large forecasting problem using 124 predictors in 8-variable VARs and FAVARs. Using state-space methods, my second paper incorporates dynamic model averaging (model averaging at each time period) in order to get dynamic probabilities of inclusion of predictors in models with time-varying parameters. The third paper extends the so-called Factor-Augmented VAR model, allowing for time-variation in all model parameters, and uses an automatic Bayesian test to check the amount of time-variation supported empirically by the data. The fourth paper develops Bayesian automatic model selection algorithms in general nonlinear VAR specifications, and I specifically apply them to the time-varying parameters VAR (TVP-VAR) model case, which is currently a very popular model in macroeconomics.

I am a fellow of the Rimini Center for Economic Analysis.