Forecasting US Inflation Using Bayesian Nonparametric Models

Abstract

The relationship between inflation and predictors such as unemployment is potentially nonlinear with a strength that varies over time, and prediction errors error may be subject to large, asymmetric shocks. Inspired by these concerns, we develop a model for inflation forecasting that is nonparametric both in the conditional mean and in the error using Gaussian and Dirichlet processes, respectively. We discuss how both these features may be important in producing accurate forecasts of inflation. In a forecasting exercise involving CPI inflation, we find that our approach has substantial benefits, both overall and in the left tail, with nonparametric modeling of the conditional mean being of particular importance.

Untitled

“…For the six-month horizon in the middle panel of Figure 4, awhman and ces0600000007 play even more important roles in accounting for the outperformace of the NN forecast visà-vis the AR benchmark starting in 2010, as awhman (ces0600000007) is selected among the top two predictors for seven (six) of the 24-month periods. The importance of awhman for forecasting inflation accords with Clark et al (2022) and Goulet Coulombe (2022), who find that awhman is a leading inflation predictor based on nonparametric Bayesian methods and hemispheric neural networks, respectively. Indeed, Goulet Coulombe (2022) finds that awhman (in combination with other predictors) is particularly pertinent for measuring economic "slack" in a deep learning-based Phillips Curve.…”

Density Forecasts of Inflation: A Quantile Regression Forest Approach

“…For an extensive survey and a systematization of the debate, see Del Negro et al (2020). Several papers in this literature point to a different relationship of inflation with its determinants in high and low inflation regimes (see, for example Akerlof et al, 1996;Costain et al, 2022;Fahr and Smets, 2010;Benigno and Ricci, 2011;Lindé and Trabandt, 2019;Forbes et al, 2021;Clark et al, 2022).…”