Abstract

Regression Discontinuity Designs in Social Science Research: Causal Inference of Cutoff-based Programs

Although the randomized control trial (RCT) provides the most straightforward route to causal inference, practical and ethical concerns can make its implementation in evaluation contexts challenging. For example, a treatment or service is offered only to those most in need, based on whether their value for a program measure (e.g., income level, reading score) falls above or below a certain threshold or cut-point. Cutoff-based assignment yields a regression discontinuity design (RDD) that facilitates unbiased causal inference at the cutoff as the process of assignment to treatment conditions is completely known and can be accounted for in the impact model.

This presentation will provide an introduction to the practical application of regression discontinuity designs in evaluating non-random, cutoff-based programs and policies in social science research. The emphasis will be on foundational RDD concepts, RDD modeling assumptions, correctly building RDD statistical models and conducting and interpreting analyses (i.e., parametric, nonparametric, graphical analyses). Empirical examples will be provided to illustrate the application of the method.