A Local Examination for Persistence in Exclusions-from-Core Measures of Inflation Using Real-Time Data

Using parametric and nonparametric methods, inflation persistence is examined through the relationship between exclusions-from-core inflation and total inflation for two sample periods and in five in-sample forecast horizons ranging from one quarter to three years over fifty vintages of real-time data in two measures of inflation: personal consumption expenditure and the consumer price index. Unbiasedness is examined at the aggregate and local levels. A local nonparametric hypothesis test for unbiasedness is developed and proposed for testing the local conditional nonparametric regression estimates, which can be vastly different from the aggregated nonparametric model. This paper finds that the nonparametric model outperforms the parametric model for both data samples and for all five in-sample forecast horizons.