Granger Causality and the Times Series Analysis of Political Relationships

Abstract

Political scientists have employed causal modeling techniques for quite some time. In most cases, however, they have relied on methods of a static, cross-sectional variety. Time series methods have been disseminated and applied, but these have not been expressly causal. The problem of assessing the causal character or direction of political relationships over time has not been addressed. Granger (1969) has suggested a notion of causality that is applicable in longitudinal analysis. The study and application of Granger's idea of causation now is widespread in economics and business. This discussion evaluates the usefulness of applying Granger causality in the study of political relationships. Granger's definition of causality is introduced, several popular procedures for assessing Granger causality are reviewed, and certain problems surrounding interpretation of those empirical tests are elucidated. The virtues and pitfalls of Granger causal analysis then are illustrated in an analysis of an important topic in international political economy and also in a reanalysis of a recent Granger causal investigation of arms races.