Quantile regression forecasts of inflation under model uncertainty

This paper examines the performance of Bayesian model averaging (BMA) methods in a quantile regression model for inflation. Different predictors are allowed to affect different quantiles of the dependent variable. Based on real-time quarterly data for the US, we show that quantile regression BMA (QR-BMA) predictive densities are superior and better calibrated compared to those from BMA in the traditional regression model. Additionally, QR-BMA methods compare favorably to popular nonlinear specifications for US inflation