Estimation of Regression Discontinuity and Kink Designs with Multiple Running Variables

Description

The regression discontinuity (RD) design is a popular method used in a wide range of policy evaluations. Its popularity owes to the simplicity, credibility, and wide applicability of its design: individuals are assigned to treatment based on whether the value of their running variable exceeds a known threshold, effectively creating a local randomized experiment in a neighborhood of the cutoff as long as individuals cannot precisely manipulate their running variable. With the growing availability of rich datasets, RD designs with multiple running variables (MRD) have become increasingly common, and yet unlike for single-dimensional RD designs, there is little consensus over how to conduct estimation in such settings. In this webinar, we cover the basics of the MRD design, and discuss a simple estimation method. We also discuss potential applications in healthcare settings, such as studying the effect of Medicaid when eligibility depends on income and age, and health effects of higher education when financial aid eligibility depends on test scores and family income.