A Varying Parameter Regression Approach to Investment Modelling in South Africa: Estimation, Stability Testing and Prediction

Abstract

Due to the rapidly changing economic and political environment in South Africa a substantial number of investment models failed one or more tests for structural stability. In this case the model needs to be modified to assimilate the structural change and, hence, it is demonstrated why the Watson/Engle varying parameter regression (VPR) technique, which is based on the recursive application of the Kalman filter, is well suited to predict future values of investment. The forecasting problem is investigated and various criteria of prediction are considered. it is shown that the VPR model can substantially reduce out-of-sample forecasting errors compared to its fixed parameter counterparts.