FORECASTING UNEMPLOYMENT RATE IN SOUTH AFRICA WITH UNEXPECTED EVENTS USING ROBUST ESTIMATORS

Abstract

The purpose of the study is to build the time series model and forecast the unemployment rate in South Africa in the presence of the unexpected events or contamination of data using robust estimators. Robust estimators deal with outliers (unexpected events) when identifying the orders, with a view to estimating the parameters of the time series models. Often, time series data are contaminated with anomalies or outliers. The standard methods of parameter estimation such as maximum likelihood (ML), least squares (LS) and method moments (MM) are sensitive to outliers. The quarterly unemployment time series data over time of January 2010 through December 2020 is used. Outliers are identified, and not removed and an ARIMA (1, 1, 1) model is found to be the best suitable model for the unemployment series. An accuracy of the forecast is measured by the standard methods, such as the RMSE, MAPE, and MAE. The forecasting results for the unemployment rate show that the robust estimators perform better in the presence of outliers.