The applied cointegration analysis for the open economy: A critical review

Most applied cointegration investigations for the open macro economy rely on error correction models to infer causality, predictability, market efficiency, dominance, and market segmentation. The error correction model is well defined only when cointegration is due to "simultaneous common factors." When common factors are not explicitly described as in a "latent common factor" model, however, error correction models are misspecified. Researchers should therefore be careful in using the error correction model for cointegrated time series to ensure that they are indeed generated from simultaneous common factor models. Analysts should investigate the exact nature of both long-run and short-run relationships by presenting a full-fledged simultaneous equations model. Without such an explicit simultaneous equations model, the presence of cointegration will only uncover the existence of a long-run relationship, but not causality, predictability, market efficiency, dominance, or market segmentation. A "best practice" will be prescribed for the proper use and interpretation of cointegration application.