Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
We forecast US inflation with a new Keynesian Phillips curve (NKPC) in the frequency domain. Our method consists of decomposing the time series of inflation and its NKPC predictors into several frequency bands, forecasting separately each frequency component of inflation, and then summing up those forecasts to obtain the forecast for aggregate inflation. We find that (i) accurately forecasting the low frequency of inflation is, on average, crucial to successfully forecast inflation; (ii) our NKPC low‐frequency forecast model consistently and significantly outperforms the time‐series NKPC and standard benchmark models; (iii) the low frequencies of inflation expectations and unemployment are the key predictors; and (iv) optimally switching on / off the forecasts of each frequency components of inflation at each period allows to outstandingly track inflation and show that all frequencies of inflation matter.