Regression Discontinuity Design (RDD)
Regression Discontinuity Design · Also known as: RDD, regression discontinuity design, sharp RDD, fuzzy RDD, Regresyon Süreksizliği (RDD)
Regression Discontinuity Design is a quasi-experimental method that identifies a causal effect by locally comparing units just above and just below a cutoff on a continuous assignment (running) variable. Formalised for applied work by Imbens and Lemieux (2008) and developed as a practical framework by Cattaneo, Idrobo, and Titiunik (2020), it estimates a local average treatment effect (LATE) at the threshold.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
+17 more
When to use it
Use RDD when treatment is assigned by a known rule that crosses a sharp cutoff on a continuous running variable, and you have a reasonably large sample (at least about 200 observations, with enough mass near the threshold). It relies on units being unable to precisely manipulate the running variable around the cutoff (checked with a McCrary density test) and on potential outcomes being continuous in the running variable at the threshold. Prefer local linear or low-order polynomial fits with an MSE-optimal bandwidth; global high-order polynomials are discouraged. It is less suitable for small samples or when individuals can sort themselves across the cutoff.
Strengths & limitations
- Delivers a transparent, credible causal estimate near the cutoff under weak, partly testable assumptions.
- Manipulation can be probed directly with the McCrary density test, and the design can be visualised with a clear discontinuity plot.
- Approximates a local randomised experiment: units just above and just below the threshold are comparable by construction.
- Identifies only a local average treatment effect (LATE) at the cutoff, which may not generalise to units far from the threshold.
- Needs a large sample with enough observations near the cutoff; with n below about 200 the bandwidth becomes too narrow and the estimate unreliable.
- If units can precisely manipulate the running variable to land on the favourable side, the identification strategy breaks down entirely.
Frequently asked
What is the difference between sharp and fuzzy RDD?
In a sharp design crossing the cutoff perfectly determines treatment, so everyone above the threshold is treated. In a fuzzy design crossing the cutoff only changes the probability of treatment; the discontinuity in treatment take-up is then used as an instrument to recover the effect.
Why does the McCrary test matter?
RDD assumes units cannot precisely manipulate the running variable to land on the favourable side of the cutoff. The McCrary density test checks for a jump in the density of the running variable at the threshold; a significant jump signals sorting and invalidates the design.
How is the bandwidth chosen?
The estimate uses only observations near the cutoff, and the bandwidth sets how near. It should be picked with an MSE-optimal selector rather than by eye, balancing the bias from including far-away points against the variance from using too few observations.
Does RDD give an effect for everyone?
No. RDD identifies a local average treatment effect at the cutoff. It is most informative about units near the threshold and may not generalise to those far above or below it.
Sources
- Imbens, G. W., & Lemieux, T. (2008). Regression Discontinuity Designs: A Guide to Practice. Journal of Econometrics, 142(2), 615-635. DOI: 10.1016/j.jeconom.2007.05.001 ↗
- Cattaneo, M. D., Idrobo, N., & Titiunik, R. (2020). A Practical Introduction to Regression Discontinuity Designs: Foundations. Cambridge University Press. ISBN: 978-1108710206
How to cite this page
ScholarGate. (2026, June 1). Regression Discontinuity Design. ScholarGate. https://scholargate.app/en/causal-inference/regression-discontinuity
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Interrupted Time SeriesCausal inference↔ compare
- Matching MethodsCausal inference↔ compare
- OLS RegressionEconometrics↔ compare
- Propensity Score MatchingResearch Statistics↔ compare
- Two-Stage Least Squares (2SLS)Causal inference↔ compare