Disaggregated Approach to Measuring Core Inflation

To distinguish inflation signal from transient noise, monetary policymakers have long used core inflation measures. Using disaggregate CPI data for Korea, this paper reviews extant measures of core inflation and documents several important empirical features of the measures. Our theoretical analysis demonstrates that the stylized facts on the extant measures are not compatible with a single stochastic trend, and our empirical findings strongly support this view. Motivated by price divergence, we model disaggregate prices in multiple-component structure and find there are four persistent components together with a group of diverging items. Having identified distinct common components, we employ a new core inflation measure based on a limited influence estimator for each convergence club. The new core inflation dominates the extant measures in its ability to account for central tendency of price distribution and for generating low variance of price changes. In addition, it forecasts the underlying trend of headline CPI inflation more accurately than the extant core inflation indicators do.