Permanent-transitory decompositions and traditional measures of core inflation

The aim of this paper is twofold. First, after defining core inflation, in line with the recent literature, as the long-run component of the rate of change of the Consumer Price Index (CPI), we address the general problem of establishing a set of conditions under which, given the assumption of inflation as an I(1) process, it is possible to identify an indicator of trend inflation. We show that some criteria recently proposed in the literature [Economics Letters 75 (2002) 17] are unnecessarily restrictive. Second, an empirical investigation, for the period 1983-2001, reveals that these conditions are met in the US economy both for the traditional measure of core inflation, which is obtained by subtracting food and energy components from CPI, and for the weighted median of CPI proposed by [Monetary Policy, University of Chicago Press for NBER, (1994) 261]. These conclusions hold for a measure of core inflation based on the year-on-year rate of change of the core index and thus we provide some motivation for this choice.