Concepts, Methods and Techniques in
Political Science, Public Policy
and International Relations

TRACK: Causality and experiments
week 1: 15-19 january 2024

Quantitative Analysis for Public Policy

Bruno Cautrès, Sciences Po

Teaching Assistant: Lucas Herzog

Course Description

In the modern world of public policy analysis the “impact evaluation” perspective is the dominant question: how to measure, to estimate and to infer the causal effect of a public policy on an “outcome”? This question plays a fundamental role in modern government and in international organizations on today. Participants will learn the major concepts and tools used by public policy specialists for “causal reasoning” and the quantitative evaluation of policies. Among the concepts covered by the course are “counterfactuals”, “potential outcomes”, “treatment effects”, “before/after effects”. Participants will learn the basics of the so-called “Rubin model” and how statistical models can help when the data are coming from non-experimental frameworks, a situation quite frequent in the real world. The regression-based methods can be used to develop a formal framework for quasi-experimental reasoning and to address major public policy questions related to such diverse issues as education, public health, social policies.

The beginnings of the course focuses on the role of quantitative methods in public policy analysis and evaluation and theoretical considerations behind (correlation versus causality; the potential outcomes and the Rubin causal model; experiments and quasi-experiments versus observations; internal/external validity). After this, the week is more applied with practical examples and replication of published papers. The objective is in showing how regression-based methods provide an empirical framework to test for “causal effect” when data come from observations rather than experiments. The key point will be to learn the participants why and how the regression-based methods are used to identify the so-called “treatment effect”, in other words to prove that the policy has or not the expected effects. Methods like the “comparison of means across groups” (“treated group/control group“), difference in differences”, “instrumental variable” or “regression discontinuity design” will be covered and presented in easy formalization, with practical lab sessions. The course is introductory/intermediate

Dates - This course runs January 15-19, 2024.

what we will learn


Basic knowledge in research methods and introductory statistics.