Learning from each other: causal inference and American political development
- PDF / 588,299 Bytes
- 7 Pages / 439.37 x 666.142 pts Page_size
- 100 Downloads / 221 Views
Learning from each other: causal inference and American political development Jeffery A. Jenkins1 · Nolan McCarty2 · Charles Stewart III3 Received: 25 September 2019 / Accepted: 28 September 2019 © Springer Science+Business Media, LLC, part of Springer Nature 2019
Abstract Within political science, a movement focused on increasing the credibility of causal inferences (CIs) has gained considerable traction in recent years. While CI has been incorporated extensively into most disciplinary subfields, it has not been applied often in the study of American political development (APD). This special issue considers ways in which scholars of CI and APD can engage in mutually beneficial ways to produce better overall research. As the contributions to the symposium demonstrate, clear scientific gains are to be had from greater CI–APD engagement. Keywords Causal inference · American political development · Gains from engagement JEL Classification C18 · N41 · N42
1 Introduction In recent years, a trend toward research based on more careful and explicit “causal inference” (CI) has spread throughout political science and other social science disciplines. The casual inference movement stresses the development of research designs that produce relationships among variables that can credibly be interpreted as causal. While the CI movement has had its greatest impact within the discipline’s methodology community, it also has had a profound effect on the applied (quantitative) subfields of American politics, comparative politics, and international relations.
* Jeffery A. Jenkins [email protected] Nolan McCarty [email protected] Charles Stewart III [email protected] 1
Price School of Public Policy, University of Southern California, Los Angeles, USA
2
Department of Politics, Princeton University, Princeton, USA
3
Department of Political Science, Massachusetts Institute of Technology, Cambridge, USA
13
Vol.:(0123456789)
Public Choice
The most common approach to CI within political science is based on the potential outcomes framework (Rubin 1974, 1977, 2006; Holland 1986). That approach is predicated on a hypothetical comparison of an outcome Y associated with a treatment T with the counterfactual outcome absent the treatment. The central problem, of course, is that for any observation, it is impossible to observe both the outcome and the counterfactual. But under certain assumptions, the causal effect of T can be estimated as the observed difference in the mean values of the observed outcomes of the treated and untreated units. Roughly speaking, those conditions are met when the treatment T is randomized across units and the treated units have no spillover effects on the untreated units. For that reason, many scholars refer to the randomized controlled trial (RCT), wherein the researcher can control the randomization and minimize spillovers, as the gold standard for causal inference.1 But for many questions in political science the RCT standard may be out of reach. Either practical or ethical considerations pre
Data Loading...
