Forecasting US Inflation in Real Time

We perform a real-time forecasting exercise for US inflation, investigating whether and how additional information--additional macroeconomic variables, expert judgment, or forecast combination--can improve accuracy robustness. In our analysis we consider the pre-pandemic period including Global Financial Crisis following expansion--the longest on record--featuring unemployment that fell to rate not seen nearly sixty years. Distinguishing features of study include use published Federal Reserve Board staff forecasts contained in Tealbooks focus performance before, during, after Crisis, with relevance also current crisis beyond. We find while simple models remain hard beat, information can forecasts, especially post-crisis period. Our results show (1) combination approaches over simpler robustify against bad particularly relevant feature environment; (2) aggregating inflation components compared aggregate directly; (3) judgmental which likely incorporate larger more timely datasets, provide improved at short horizons.