Time-varying persistence in US inflation

Abstract

The persistence property of inflation is an important issue for not only economists, but, especially for central banks, given that the degree of inflation persistence determines the extent to which central banks can control inflation. Further, not only is the level of inflation persistence that is important in economic analyses, but also the question of whether the persistence varies over time, for instance, across business cycle phases, is equally pertinent, since assuming constant persistence across states of the economy, is sure to lead to misguided policy decisions. Against this backdrop, we extend the literature on long-memory models of inflation persistence for the US economy over the monthly period of 1876:2-2014:5, by developing an autoregressive fractionally integrated moving average-generalized autoregressive conditional heteroskedastic (ARFIMA-GARCH) model, with a time-varying memory coefficient which varies across expansions and recessions. In sum, we find that, inflation persistence does vary across recessions and expansions, with it being significantly higher in the former than in the latter. As an aside, we also show that, persistence of inflation volatility however, is higher during expansions than in recessions. Understandably, our results have important policy implications.