Forecasting in a large macroeconomic system

This paper examines the efficiency gains yielded from estimating multiple equation cointegrated systems as compared to their single equation counterparts. In particular, this paper is concerned with the ability of utilizing the cointegrating information to improve forecasting performance. Recently an inability to improve forecasts of real income once money demand error correction terms were introduced has been used to argue that the M2 relationship had broken down during the early 1990s. However, the results suggest that once the underlying responses of variables are more closely investigated, the behaviour of M2 has remained stable.