Date of Award
Master of Science in Statistics
Computer Science and Statistics
Predicting the unemployment rate is one of the most important applications for economists and policymakers (Golan, 2002). In this thesis, the focus is on the seasonally adjusted U.S. national unemployment rate (UR). The goal is to introduce the seasonally adjusted job openings (JOB) from Bureau of Labor Statistic for UR forecasting.
In order to forecast UR, firstly, an integrated autoregressive moving average model (ARIMA) is constructed as a benchmark mode. For a better comparison, a well known leading indicator - the seasonally adjusted initial claim for unemployment insurance (IC), released by the U.S. Department of Labor, is also included. By using JOB and IC as external variables, integrated autoregressive moving average with external variable(s) models (ARIMAX) are successfully constructed. Multivariate vector autoregressive models (VAR) are also well constructed for UR, JOB and IC. Akaike Information Criterion (AIC), Schwarz-Bayesian Criterion (BIC) and Hannan-Quinn Criterion (HQ) are applied for models selection.
In the out-of-sample comparison, both rolling forecasts and recursive forecasts are chosen. Mean absolute forecast error (MAFE) and mean square forecast error (MSFE) are calculated, along with Diebold-Mariano (DM) test for models comparison. The results show that the JOB related models have much better forecasting power than the benchmark model and the IC related models, which demonstrate that JOB index can be used as one of the leading indicators to improve UR forecasting.
Huang, Xinkai, "Forecasting the US Unemployment Rate with Job Openings Index" (2015). Open Access Master's Theses. Paper 699.