A0357
Title: Causal analysis of cointegrated systems: Model manifestations of hierarchical properties
Authors: Emanuele Lopetuso - University of Reading (Italy) [presenting]
Abstract: The VAR in its cointegrated version offers an effective approach to model dynamics generated by steady state violations. However, the existing econometric literature scarcely assessed the duality between model characteristics and causal attributes. Building upon the recent preliminary investigations concerning causality within cointegrated systems and partially observed models, a novel configuration of the vector error correction model is proposed. This arrangement is able to emphasize the long-run causal structure even when the model user fails to identify the whole set of variables involved in the data-generating process. The presented configuration produces a submodel whose mathematical features are connected with a counterfactual understanding. This dichotomy allows the translation of model constraints into causal assertions, enabling the application of standard testing procedures. The evaluation of hypotheses aids the model user in the process of identifying the long-term causal structure, making a significant contribution to the field of causal inference in econometrics. Novel concepts such as causal exogeneity and causal rank are also introduced in order to facilitate and support the development of a causal language specifically tailored for cointegrated systems.