Microforecasting inflation

Abstract

Forecasting inflation accurately is critical for making many policy decisions. A key modeling choice that economists and policymakers face when forecasting inflation is whether to forecast aggregate inflation directly or its individual components first and then aggregate the results. Another important modeling decision is whether or not to group certain components (for instance, core and noncore components1) and then model them separately. In this article, we present a new disaggregated approach to forecasting inflation. Our focus is on inflation as measured by the Personal Consumption Expenditures (PCE) Price Index from the U.S. Bureau of Economic Analysis (BEA). We developed this new approach primarily because of the widespread heterogeneity evident in the dynamics of inflation both across its components and over time. In what follows, we begin by documenting this heterogeneity in PCE inflation. We then discuss the BEA data we used in our research and explain our method for microforecasting inflation in the PCE components—which we subsequently aggregate to derive a total PCE inflation forecast. Finally, we compare the forecasting accuracy of our novel approach and other methods, including those that forecast aggregate inflation directly. We find that over our sample period and other subperiods, the forecasts produced by our method are more accurate than those produced by the alternative approaches considered here.