The structural approach to time series modeling uses economic theory to model the relationship among the variables of interest. Unfortunately, economic theory is often not rich enough to provide a dynamic specification that identifies all of these relationships. Furthermore, estimation and inference are complicated by the fact that endogenous variables may appear on both the left and right sides of equations.

These problems lead to alternative, non-structural approaches to modeling the relationship among several variables. This section describes the estimation and analysis of vector autoregression (VAR) models.

We also describe tools for testing the presence of cointegrating relationships among several non-stationary variables.