We develop a method for constructing prediction intervals for a nonstationary variable, such as GDP. The method uses a factor augmented regression [FAR] model. The predictors in the model includes a small number of factors generated to extract most of the information in a set of panel data on a...