Persistence and Instability in Disaggregate Inflation Information
Speaker
Abstract
| This paper examines the evolving predictive densities for inflation persistence. We use Bayesian Model Averaging to construct predictive densities of the features of interest - in this case the sum of the autoregressive coefficients (SARC) measure of US inflation persistence - allowing at the same time for breaks of unknown number and timing. We show that a number of disaggregate series exhibit multiple breaks in the means and/or the variance. Since the early 1990s, there is considerable evidence in support of bimodal predictive densities for US inflation persistence. Hence, the average level of persistence of inflation exhibited by the disaggregate series misses an important feature of the inflationary process. Given the bimodal nature of inflation persistence across disaggregates; the uncertainty about inflation persistence is greater than indicated by a Gaussian aggregate inflation model. From our disaggregate perspective, the debate typified by the exchange between Cogley-Sargent and Stock in the NBER Macro Annual focuses on the wrong summary statistic. The persistence of aggregate inflation may or may not have changed. The disaggregate densities point towards a recent divergence in the modal behavior of certain disaggregate. |
| Contact information: |
| Erik Kole |
