Abstract
We study the real-time predictive content of crude oil prices for U.S. real GDP growth through a pseudo out-of-sample (OOS) forecasting exercise. Comparing our benchmark model “without oil” against alternatives “with oil,” we strongly reject the null hypothesis of no OOS population-level predictability from oil prices to GDP at the longer forecast horizon we consider. This examination of the global OOS relative performance of the models we consider is robust to use of ex post revised data. But when we focus on the forecasting models’ local relative performance, we observe strong differences across use of real-time and ex post revised data.