Aggregation analysis in empirical multivariate dynamic models*

The paper analyses the aggregation problem when the micro relations consist of multivariate specifications. We focus on a generalization of the model selection criterion proposed in the previous literature for discriminating between aggregate and disaggregate models. In addition, we extend the model selection problem in forecasting aggregate variables out of sample. Both the stationary and cointegrated VAR processes are considered. The paper also introduces a new “test of perfect aggregation” in the multivariate environment. The properties of the test are analysed by Monte Carlo simulation. The results suggest adopting a correction to reduce the small sample bias of the test.