MULTIPLE TIME SERIES ANALYSIS OF SINGAPORE'S MACROECONOMIC DATA

Abstract

This is an exercise on the application of the VARMA model to Singapore's macroeconomic data. This model incorporates four series, namely, the gross domestic product, the government consumption expenditure, the gross fixed capital formation and the private consumption expenditure jointly without imposing any restriction on the underlying mechanism of the variables. It has an edge over the univariate model, which is restricted by its assumption that each series is dependent on its lagged value only. It is also free of any simultaneity bias present in the regression analysis when the four variables are determined jointly. The approach used in this study is that proposed by Tiao and Box (1981), which is extension of the modelling techniques developed by Box and Jenkins (1976) in the univariate domain. This method is described in detail for the various stages: the tentative identification of the model, the estimation of the parameters, the refinement of the model and finally the various diagnostic checks to ascertain the adequacy of the specified model. The case study shows that the VARMA model provides a fairly good fit to the macroeconomic time series data. Other than the explicit relationship present in the model, the significant correlation between some of the residuals suggest that that an implicit relationship is also present among some of the variables. These relationships therefore confirm the usefulness of modelling these four variables jointly. Despite the greater loss in the degrees of freedom, the VARMA model exhibits a slight improvement over the ARIMA model as well as the prototype macroeconomic model in terms of its ex-ante forecast. Hence, the VARMA model offers a new direction away from the traditional econometric model used in the analysis of the macroeconomic time series data.