Modelling of Structural Changes in Demand for Money Cointegration Relations

In this paper the multivariate cointegration technique coupled with a smooth nonlinear trend of time is applied to model the demand for money. Unmodelled gradual structural changes in the cointegration parameters affect the specification of the cointegration relations so that the number of cointegrating vectors found by linear methodology is smaller than suggested by the economic theory. Here the demand for broad money in Finland during 1980–1996 is analysed. It turns out that, if the cointegrated VAR model is extended with a suitable nonlinear deterministic trend of time related to the intercept term, then the missing cointegration relation between broad money and the scale variable is found and the cointegration space can then be identified.