Long-run exclusion and the determination of cointegrating rank: Monte Carlo evidence

Abstract: This note investigates long-run exclusion in a cointegrated vector autoregressive (VAR) model from the viewpoint of finite-sample statistical inference. Monte Carlo experiments show that, in various circumstances, a mis-specified partial VAR model, which is justified by the existence of a long-run excluded variable, can lead to better finite-sample inference for cointegrating rank than a fully specified VAR model. Implications of long-run exclusion for econometric modelling are then considered based on the Monte Carlo study.