Predicting Inflation Without Running Predictive Regressions

Abstract

Inflation rates are highly persistent and extremely difficult to predict. Most statistical predictions based on predictive regressions fail to outperform the simple assumption of random walk in out-of-sample testing. The poor out-of-sample performance is a common feature of predictive regressions on highly persistent time series. This paper proposes a new approach for inflation forecasting that does not specify or estimate any predictive regressions, but rather starts by estimating a contemporaneous relation between inflation rate and a short-term interest rate, and then relies on the forward interest rate curve to predict future interest rates and accordingly inflation rates over both short and long horizons. Historical analysis with the US inflation series shows that this approach can outperform random walk, out-of-sample, by 30-50% over horizons as far as three to five years.