Two Measures of Core Inflation: A Comparison

Abstract

Trimmed-mean personal consumption expenditure (PCE) inflation does not clearly dominate PCE inflation excluding food and energy in real-time forecasting of headline PCE inflation. However, trimmed-mean inflation is the superior communications and policy tool because it has been a less-­biased real-time estimator of headline inflation and because it more successfully filters out headline inflation's transitory variation, leaving only cyclical and trend components.

Introduction

Most central banks describe their price stability goals in terms of the behavior of an all-items price index. For example, each January since 2012, the U.S. Federal Reserve's Federal Open Market Committee (FOMC) has reaffirmed its judgment that "inflation at the rate of 2 percent, as measured by the annual change in the price index for personal consumption expenditures [PCE], is most consistent over the longer run with the Federal Reserve's statutory mandate." Nevertheless, "core" inflation measures often play a prominent role in the forecasting frameworks used by policymakers, as well as in their narrative accounts of realized inflation outcomes. The economic projections published four times per year by the FOMC, for example, include policymakers' projections both for all-items PCE inflation and for PCE inflation excluding food and energy (ex-food-and-energy PCE inflation). Monetary policy statements by the Reserve Banks of New Zealand and Australia often make explicit reference to measures of core inflation, and the Bank of Canada monitors core inflation measures as an "operational guide to help the Bank achieve the total CPI [consumer price index] target" (Bank of Canada, 2018).

Broadly, the motivation for monitoring a core inflation measure is that headline inflation is subject to large transitory shocks unrelated to changes in cyclical inflation pressures or the public's confidence in the central bank's commitment to long-run price stability. This transitory variation potentially complicates both inflation forecasting and policymaking.

As regards forecasting, stripping statistical noise from the left-hand side of an equation prior to its estimation yields more-precise coefficient estimates, increasing the chances that the fitted equation will produce superior forecasts. In the current context, there is an advantage to forecasting headline inflation using an equation estimated with core inflation on its left-hand side, to the extent that the difference between headline and core inflation can't be forecasted.

As regards policymaking, there are both theoretical and practical reasons for thinking that central banks should react more strongly to that portion of headline inflation that is core inflation than to that portion of headline inflation that is not core inflation. According to theory, policy should stabilize "sticky" prices. To the extent that deviations of headline inflation from core are driven by volatile, flexible-price components of the headline index, therefore, those deviations should receive reduced weight in the policy reaction function. Prac­­tically, monetary policy affects inflation with a substantial lag, so a strong policy response to transitory inflation movements risks having policy provide stimulus when underlying cyclical inflation pressures are waxing or having policy apply restraint when underlying cyclical pressures are waning.

Policymakers also sometimes draw inferences about slack from the behavior of inflation. If inflation remains low or fails to increase despite historically low unemployment, it is tempting to infer that the natural rate of unemployment must have declined. Such inferences are reliable, however, only to the extent that there is a tight empirical link between inflation and labor-market slack. As discussed above, in a forecasting context, one will more accurately identify any such link if one strips idiosyncratic noise from headline inflation before estimation.

In this article, we compare two alternative measures of core PCE inflation: ex-food-and-energy PCE inflation and trimmed-mean PCE inflation. Given the trimmed mean's relatively late introduction (in 2005), it is only recently that we have a sufficiently long history to perform a real-time comparison between the two core measures. 

We start with a brief history of core inflation. Then, we examine whether either ex-food-and-energy PCE inflation or trimmed-mean PCE inflation is useful in real-time forecasting of headline inflation and whether either core measure has a reliably tight empirical link to slack. We discover that both trimmed-mean and ex-food-and-energy PCE inflation are useful in real-time forecasting of headline PCE inflation and that neither core measure has a strong, consistent forecasting advantage over the other. However, the trimmed-mean measure has exhibited less real-time bias than the ex-food-and-energy measure and is more tightly linked to labor-market slack than is either headline inflation or the conventional core measure. For those reasons, trimmed-mean should arguably receive greater weight than headline and ex-food-and-energy PCE inflation in policy discussion and Federal Reserve communications.