Frank Schorfheide, University of Pennsylvania
"Forecasting with a Panel Tobit Model"
Laura Liu Federal Reserve Board, Hyungsik Roger Moon. University of Southern California, and Yonsei Frank Schorfheide, University of Pennsylvania CEPR, NBER, and PIER.
Preliminary Version: March 4, 2018
We use a dynamic panel Tobit model to generate point and density forecasts for a large cross-section of short time series of censored observations. Our fully Bayesian approach allows us to flexibly estimate the cross-sectional distribution of heterogeneous coefficients and then implicitly use this distribution to construct Bayes forecasts for the individual time series. We consider versions of the model with homoskedastic and heteroskedastic innovations. In Monte Carlo experiments and the empirical application to loan charge-off rates of small banks, we compare Bayesian point, interval, and density forecasts obtained from various versions of the panel Tobit model. In our empirical application the assumption of homoskedasticity yields more accurate point forecasts, whereas interval and density forecasts are more accurate if computed under the assumption of heteroskedastic innovations. Local house prices and unemployment rates do not appear to be significant determinants of charge-off rates