On Forecasting Economic Time Series: A Comparative Study

Since the appearance of the book by Box and Jenkins in 1970, the use of auto regressive integrated moving average model has become widespread in analyzing and forecasting economic time series data. Sims (1980) proposed a vector auto regressive model (VAR) model approach as an alternative to the conventional strategy for constructing macro- economic model. Trevor and Thorp (1988) emphasized that estimating a V AR model using non-stationary data may result in unstable econometric relationships and suggested using Bayesian V AR model to accommodate non- stationarity. In this paper we have compared the forecasting performance of these models using Australian macro- economic time series data.