Forecasting with a panel Tobit model

We use a dynamic panel Tobit model with heteroskedasticity to generate forecasts for large cross‐section of short time series censored observations. Our fully Bayesian approach allows us flexibly estimate the cross‐sectional distribution heterogeneous coefficients and then implicitly this as prior construct Bayes individual series. In addition density forecasts, we set that explicitly target average coverage probability cross‐section. present novel application in which forecast bank‐level loan charge‐off rates small banks.