Academic journal article International Advances in Economic Research

Forecasting with the Nonparametric Exclusion-from-Core Inflation Persistence Model Using Real-Time Data

Academic journal article International Advances in Economic Research

Forecasting with the Nonparametric Exclusion-from-Core Inflation Persistence Model Using Real-Time Data

Article excerpt

Introduction

Financial data and some economic time series data, such as inflation, tend to be from fat-tailed distributions, which should be taken into account in forecasting. Using the kernel-weighted least squares nonparametric model proves to be beneficial since it is more adept at detecting information in the tail regions. Events in fat-tailed distributions have a higher probability of occurrence when compared to thin-tailed distributions. Hence, it is important for policy-makers to have a forecasting tool that is better equipped at detecting events in the tail region, which this paper provides.

In particular, this paper provides the policy-maker with three local nonparametric forecasting models regarding inflation persistence through the use of the nonparametric exclusion-from-core inflation persistence model. Instead of incorporating data based on chronological time, the local nonparametric forecasting models permit the incorporation of datum in the relevant time period. For instance, low inflationary periods a few years apart have more in common in terms of behavior and magnitude than a more contemporaneous time period that could consist of a high inflationary time period.

This paper builds upon previous work done by Tierney (2011, 2012). Tiemey (2012) found that the nonparametric exclusion-from-core inflation persistence model was able to utilize data revisions, which were small in magnitude. These small revisions could very easily be lost in aggregation or in the presence of outliers, which can dominate the parametric exclusion-from-core inflation persistence model. At the local nonparametric level, Tiemey (2011) found that data revisions did produce statistically different model parameters. (1)

The main purpose of this paper is to extend the work of Tierney (2011, 2012) into out-of-sample forecasting at the local nonparametric level by presenting three different forecasting methods. This paper finds that the three local nonparametric methods outperform the parametric and global nonparametric forecasts, which use the average of the local nonparametric estimated coefficients in order to form the forecasts. In addition, this paper shows how nonparametrics can be used to identify underlying problems in the data, which can serve as an early warning system to policy-makers.

The nonparametric exclusion-from-core inflation persistence model is an adaptation of Coogley's (2002) exclusion-from-core inflation persistence model, which permits one to simultaneously analyze the relationship between total and core inflation in a stationary, flexible framework. This flexibility also makes the nonparametric exclusion-from-core inflation persistence model suitable for incorporating and using revisions to real-time data. Real-time data is organized by vintages with each vintage containing a newly released datum, which is the last observation of the vintage, as well as the revision of data that were previously released. Data revisions can be revised up to 3 years after initial release and consist of incorporating new or corrected data (Croushore and Stark 2001, Croushore 2008). Other sources of data revisions are benchmark revisions, which are changes in the data collection methodology (Croushore and Stark 2003).

Elliott (2002) argues for the inclusion of more vintages in order to identify trends when examining real-time data. Keeping this in mind, 62 vintages are examined beginning with vintage V_1996:Q1 and ending with vintage V_2011:Q2 in order to keep the sample analysis the same for the personal consumption expenditure (PCE) and Core PCE. (2) The prefix of "V_" precedes a vintage in order distinguish it from the notation of a given observation.

The real-time U.S. PCE price index is used to measure total inflation and the realtime U.S. core PCE is used to capture inflation trends by removing the volatile components of food and energy. (3) The real-time data of PCE and core PCA were used because this was what the Federal Reserve used to forecast total and core PCE (Croushore 2008). …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.