Dynamic Inferences in Political Science with Vector Autoregressive (VAR) Models

• In Event: Models for TSCS, Time, and Duration Data

Abstract

Recent articles in political science have highlighted the importance of testing for cointegration in the context of time-series dynamics,. These studies also emphasize the Error Correction Model usage when there is cointegration. However, one of the aspects that are rarely discussed is the assumptions of exogeneity, a necessary condition for using these models to draw inferences about the causal mechanism driving the relationship between the explanatory to dependent variables in an unidirectional manner. When these conditions do not hold, researchers should test their hypotheses in the context of a system of vector autoregressions. In this paper, I introduce a step-by-step guide to help researchers choose the most appropriate models considering the evidence on the degree of exogeneity observed. Finally, I present the usefulness of this approach with the application in a recent study, in which such procedures were not performed, I show the changes in results interpretation after the application of a VECM.

Authors

• Lorena G. Barberia, University of São Paulo

• Maria Leticia Claro de Faria Oliveira, University of Sao Paulo

• Natalia Moreira