Forecasting Unemployment Rates in USA using Box-Jenkins Methodology

Abstract

Unemployment, as a measure of market conditions, appears as an economic problem in every society and is a phenomenon with considerable negative social consequences. A low rate of unemployment is one of the main objectives for governmental macroeconomic policy. The main aim of this project is to identify the most appropriate forecasting model, i.e. the seasonal autoregressive integrated moving average (SARIMA), autoregressive conditional heteroskedasticity (ARCH) and the generalized autoregressive conditional heteroskedasticity (GARCH).  Using one or a combination of these models could provide the best forecast for US unemployment. Applying monthly data to the US unemployment rate from January 1955 to July 2017 proved that the SARIMA(1,1,2)(1,1,1)12-GARCH(1,1) is the best model to project US unemployment. Finally, this project evaluates the forecasting performance of the model using forecast accuracy criteria, such as the Root Mean Square Error (RMSE), Mean Absolute Percent Error (MAPE) and Theil's inequality coefficient.Keywords: Unemployment Rates, Seasonal Time Series, SARIMA-CARCH ModelJEL classifications: C53, E27