The predictive content of the output gap for inflation: Resolving in-sample and out-of-sample evidence

This paper evaluates potential explanations for the sometimes poor forecasting performance of the Phillips curve. One explanation is that out-of-sample metrics are noisy or, equivalently, have relatively low power. Another potential explanation is instability in the coefficients of the model. To assess these forces, this paper compares sample forecasting results to results from bootstrap simulations of models that either assume stability or allow breaks in the coefficients of the model. This analysis indicates that a significant portion of the weakness of the out-of-sample evidence is attributable to power limitations. But instabilities in the coefficients on the output gap also play a role.