TRANSFER FUNCTION MODELS FOR SOME SINGAPORE ECONOMIC TIME SERIES : AN EMPIRICAL STUDY

Abstract

Our main interest in this academic exercise is to examine the possibility of forecasting the 1985 Singapore recession. For this purpose, transfer function-noise (TFN) models are built for two pairs of local economic time series; and forecasts of gross domestic product (GDP) and general wholesale price index (WPI) are obtained. The two pairs of series are obtained via a shortlisting process involving the National Bureau Of Economic Research Indicator Theory and the Pierce-Haugh Causality Test. For comparison, we also build the Box-Jenkins ARIMA models for GDP and WPI. From our analysis, we found that there is essentially no difference between the TFN model and the ARIMA model for forecasting GDP but the ARIMA model proved to be superior for WPI. Both the TFN model and the ARIMA model forecasted positive growth rate of about 7 per cent for 1985 compared to the actual growth rate of 1.5 per cent. Poor performance of both models could be due to the short data series included in the analysis.