VAR Models as Structural Approximations

This paper presents a way of estimating how accurate VAR models are likely to be for answering structural questions. Data are generated from a dynamic deterministic solution of a structural model; a VAR model is estimated using a subset of these data; and the properties of the VAR model are compared to the properties of the structural model. This procedure has the advantage of eliminating the effects of error terms, since the data are generated for a deterministic simulation. The results show that the VAR models do not seem to be good structural approximations.