Regression Discontinuity Design in Education Research
Also known as: RDD in education, education RD design, sharp RDD education, score-cutoff design
Regression discontinuity design (RDD) in education research exploits a score-based eligibility cutoff — such as a test score threshold, GPA requirement, or age cutoff — to estimate the causal effect of a program, intervention, or policy on student or school outcomes. Units just below and just above the cutoff are treated as near-randomly assigned, enabling credible causal inference without a randomized trial.
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.
When to use it
Use RDD in education research when assignment to a program, resource, or policy is determined by whether a student, school, or district crosses a known threshold on a measurable running variable. Common examples include: eligibility for remedial programs based on test scores; class size changes triggered by enrollment thresholds (Maimonides' rule); school accountability interventions triggered by performance scores. The method yields a local causal effect — valid only near the cutoff — so it is not appropriate when you need population-average effects, when the running variable is manipulated by agents, or when no natural threshold exists.
Strengths & limitations
- Provides credible local causal identification from observational or administrative data without randomization, exploiting naturally occurring cutoffs.
- The assignment mechanism is transparent and often verifiable from administrative records, making the design easy to communicate to policymakers.
- Robust to unobserved confounders that vary smoothly across the running variable, since these are balanced locally around the cutoff.
- Widely applicable in education settings where score thresholds, age cutoffs, and enrollment caps are routine.
- Placebo and density tests provide formal checks of the identification assumptions, building credibility of the findings.
- Estimates are local — they apply only to units near the cutoff and may not generalize to the broader population of interest.
- Requires a sufficient number of observations near the threshold; sparse data around the cutoff leads to wide confidence intervals.
- Validity collapses if agents manipulate the running variable to just cross the cutoff (e.g., teachers inflating test scores to push students above a threshold).
- Sharp RDD cannot be applied when compliance with the cutoff-based assignment rule is imperfect without moving to the fuzzy RDD framework, which yields a LATE rather than an intent-to-treat effect.
Frequently asked
What makes RDD appropriate for education research specifically?
Education systems routinely use score-based thresholds — exam cutoffs, GPA requirements, enrollment caps, performance targets — to allocate programs and resources. These administrative cutoffs create near-randomized assignment for students or schools just around the threshold, enabling causal inference without a formal experiment.
How do I choose the bandwidth?
Use a data-driven optimal bandwidth selector such as the MSE-optimal or CER-optimal bandwidth from Calonico, Cattaneo, and Titiunik (2014), implemented in software packages like rdrobust in R or Stata. Always report estimates for at least two or three alternative bandwidths to show robustness.
What is the difference between sharp and fuzzy RDD in education?
In a sharp RDD every unit above the cutoff receives treatment and every unit below does not. In a fuzzy RDD, crossing the cutoff only raises the probability of treatment — for example, a test-score threshold that triggers an offer of tutoring that some students decline. Fuzzy RDD estimates a local average treatment effect (LATE) using the cutoff as an instrument for actual take-up.
How do I test whether the running variable has been manipulated?
Apply the McCrary (2008) density test, which checks for a discontinuous jump in the density of the running variable at the cutoff. A significant jump suggests that agents have sorted around the threshold, invalidating the design. Visual inspection of a histogram of the running variable near the cutoff is also informative.
Can I combine RDD with difference-in-differences?
Yes. A difference-in-discontinuities design adds a pre-post dimension to the RDD: it compares the jump at the cutoff after an intervention to the jump before, controlling for any pre-existing discontinuity. This is useful in education when a cutoff existed before and after a policy change.
Sources
- Lee, D. S., & Lemieux, T. (2010). Regression discontinuity designs in economics. Journal of Economic Literature, 48(2), 281-355. DOI: 10.1257/jel.48.2.281 ↗
- Thistlethwaite, D. L., & Campbell, D. T. (1960). Regression-discontinuity analysis: An alternative to the ex post facto experiment. Journal of Educational Psychology, 51(6), 309-317. DOI: 10.1037/h0044319 ↗
How to cite this page
ScholarGate. (2026, June 3). Regression Discontinuity Design in Education Research. ScholarGate. https://scholargate.app/en/causal-inference/regression-discontinuity-design-in-education-research
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.
- Difference-in-DifferencesEconometrics↔ compare
- Fuzzy Regression DiscontinuityCausal inference↔ compare
- Instrumental Variables in Health ResearchHealth Economics↔ compare
- Interrupted Time SeriesCausal inference↔ compare
- Propensity Score MatchingResearch Statistics↔ compare