DataTerminology.comNEW
A quasi-experimental design that identifies causal effects at a cutoff in an assignment variable by comparing units just above and below.
DESCRIPTION
Regression Discontinuity exploits a threshold rule assigning treatment when a running variable crosses a cutoff. Under continuity assumptions, the outcome jump at the cutoff identifies a local causal effect. Designs can be sharp (deterministic treatment) or fuzzy (probabilistic). Careful bandwidth selection, local polynomial fitting, and robustness checks are essential.
ALTERNATIVE DEFINITIONS
Wikipedia
The regression discontinuity design is a quasi-experimental pretest–posttest design that elicits a local average treatment effect by assigning a cutoff or threshold for receiving an intervention.
Emphasizes threshold-based assignment and local effect.
Cattaneo, Idrobo, Titiunik
RD designs identify causal effects at cutoff points using local polynomial estimators and robust inference tools.
Highlights local estimation and robust tools for RD analysis.
RELATED TERMS
Sharp RD Fuzzy RD Bandwidth Selection Local Linear Regression Continuity Assumption Running Variable Threshold Causal Inference
COMPETING TERMS
Difference-in-Differences Interrupted Time Series
KNOWLEDGE AREAS
TARGET AUDIENCE
Data Scientist Information Analyst AI Engineer
EXTERNAL SOURCES
- Wikipedia - Regression discontinuity design
- rdpackages - RD theory, software, and tutorials
- Scott Cunningham - Practical RDD guide with examples
TOP QUESTIONS
What is Regression Discontinuity Design (RDD)?
RDD estimates local treatment effects by comparing observations just above and below a threshold in a running variable.
Category: Definition
What assumptions does RDD require?
Continuity of potential outcomes at the cutoff and no precise manipulation of the running variable around the threshold.
Category: Requirements
What’s the difference between Sharp and Fuzzy RDD?
Sharp assigns treatment deterministically at the cutoff; Fuzzy uses the cutoff to shift treatment probability, requiring IV-like estimation.
Category: Definition
How do you estimate effects in RDD?
Local linear or polynomial regressions near the cutoff with careful bandwidth selection and robust standard errors.
Category: Practicality
What are common mistakes in RDD?
Using global polynomials, ignoring bandwidth sensitivity, or not testing for bunching/manipulation.
Category: Mistakes
What is the main risk in RDD studies?
Manipulation of the running variable near the cutoff can invalidate identification.
Category: Risks
When to use RDD over DiD?
When a credible threshold rule exists and local effects near the cutoff are of interest.
Category: Alternatives
How local are RDD estimates?
They identify effects at the cutoff; external validity beyond that region may be limited.
Category: Opinion
What diagnostics support RDD credibility?
McCrary density test for manipulation, covariate balance at the cutoff, and bandwidth robustness checks.
Category: Practicality