Cointegration implications of linear rational expectation models

This paper derives the cointegration spaces that are implied by linear rational expectations models when data are I(1). The cointegration implications are easy to calculate and can be readily applied to test if the models are consistent with the long-run properties of the data. However, the restrictions on cointegration only form a subset of all the cross-equation restrictions that the models place on data. The approach is particularly useful in separating potentially data-consistent models from the remaining models within a large model family. Moreover, the approach provides useful information on the empirical shock structure of the data.