Abstract/Summary

The appropriate definition, role and use of exogeneity in economic analysis has been an area of considerable contention. The empirical resolution of these disputes has proven to be challenging due to the incongruity between the econometric tools, which rely on statistical considerations, and the theoretical perspectives that emphasize causal relationships. The findings of this thesis aim to reconcile these two perspectives by investigating whether causal features manifest in model coefficients. Drawing on the properties of the graphical representation of the latent causal structure, the first chapter introduces a causal definition of exogeneity. Exploiting state-of-the-art tools in the field of cointegration analysis, we investigate the interplay between causal exogeneity and the restrictions it imposes on models, without requiring that all relevant variables are observed. The duality between causal exogeneity and the restrictions on model parameters lays the foundation for the development of testing procedures tailored to facilitate causal logic. The second chapter identifies an additional duality between model restrictions and causal properties and leverages it to propose an alternative testing technique. The chapter also evaluates the performance of the two tests using simulated data and demonstrates their practical utility through an empirical illustration on monetary variables. The third chapter examines the causal interpretation of weak exogeneity and cautions against associating it with causality. The chapter's findings are particularly pertinent in the context of cointegrated systems, where weak exogeneity is associated with constraints on the adjustment matrix. This coupling has led to misinterpretations, since variables unaffected by steady-state violations are associated with non-causality, and the absence of weak exogeneity is incorrectly perceived as a characteristic of causally affected variables. We aim to demonstrate that this duality is not automatic and clarify under what circumstances weak exogeneity can or cannot serve as a proxy for causal inference.