Regression Discontinuity (RD) Design Fundamentals

Flashcards covering the key concepts, assumptions, and best practices of Regression Discontinuity (RD) designs as outlined in the lecture notes.

Overlap

In RD designs, unlike standard regression analysis, there is no overlap in treated and control units because individuals with different values of $D$, the treatment, have different values of the covariate by construction.

Conditional Independence Assumption (Sharp RD)

An assumption that holds automatically in a sharp RD design because treatment assignment is determined solely by the cutoff value of the running variable.

Fuzzy RD Design

An RD design in which the cutoff value of the running variable determines the probability of treatment.

Key Identifying Assumption of RD

The assumption that the average potential outcomes are continuous through the cutoff.

Sharp RD Treatment Effect Identification

Under its assumptions, a sharp RD design identifies an average treatment effect measured at the cutoff, defined as $\delta = E(y_{1i} - y_{0i} | RunningVariable_i = 0)$.

Binned Scatter Plot

A plot of the outcome and the running variable used as the basis for an RD analysis.

RD Regression Specification

A model that includes a low-order polynomial in the running variable and an interaction of the running variable with the treatment indicator.

Running Variable Manipulation

Evidence of this is checked by observing the distribution of the running variable; it should be smooth through the cutoff without bunching.

Baseline Covariates Analysis

An RD analysis of these variables should show no evidence of discontinuities among them, and their inclusion in the regression model should not affect the estimated treatment effect.