The Decline of Formal Models and the Rise of Experiments in Political Science in the U.S.: From a Perspective on the Concepts of Causality

Abstract

　This article claims that the recent prevalence of the concept of causality based on potential outcomes explains the decline of formal models and the rise of experiments observed in publication trends in political science in the United States. Observational data analysis, a mainstream approach in political science, necessitated formal models as strong theories to ensure empirical-statistical models to meet the conditions in the classical concept of causality, especially temporal precedence and the exclusion of alternative explanations for cause-effect linkages. Rather than specifying statistical models with the aid of formal models for causal inference, however, designing randomized experiments are the most straight-forward and desirable way of satisfying the conditional independence assumption (CIA) in the recently prevailing concept of causality based on potential outcomes. For this reason, the significance of formal models has been devaluated, and experimental methods have rapidly evolved with technological advances associated with the Internet, and have established its position in political science in the United States.