Forecasting U.S. Inflation: A Look Beyond the Conditional Mean

Most studies on U.S. inflation forecasting have focused on predicting the mean inflation using time series and Phillips Curve (PC) models. The findings indicate that using real economic indicators (such as unemployment or the output gap) improve out-of-sample forecasting performance during the late 1970s and the first half of the 1980s. But after 1985, PC based forecasts do not lead to forecasting gains vis-á-vis time series models (autoregressive and random walk models) when the latter models became a lot harder to outperform. Our study examines whether indicators of economic activity carry relevant information about the dynamics of higher moments of inflation, and hence help improve the accuracy of the distribution of inflation. We forecast (out-of-sample) the distribution of inflation for 6 and 12 months ahead for the period 1985:1 to 2007:12 and evaluate the performance of the various models in two sub-samples (1985:1 to 1995:12 and 1996:1 to 2007:12). Our results show that for the core inflation measures and PCE conditioning the dynamics of the predictive distribution on the leading indicators provides more accurate forecasts relative to time series models.