Glossary
Regression Discontinuity Design (RDD)
A quasi-experimental method that exploits a threshold or cutoff in an assignment variable to estimate causal effects. Units just above and just below the cutoff are compared, under the assumption that they are similar in all respects except treatment status. RDD is considered ...
Definition
A quasi-experimental method that exploits a threshold or cutoff in an assignment variable to estimate causal effects. Units just above and just below the cutoff are compared, under the assumption that they are similar in all respects except treatment status. RDD is considered one of the most credible non-experimental designs and is widely used to evaluate educational policies, electoral outcomes, and programme eligibility rules.
Why It Matters
RDD approximates a randomised experiment around the cutoff, because units just above and just below the threshold are essentially identical except for treatment assignment. This local randomisation provides internally valid causal estimates without requiring an actual experiment, making RDD a gold-standard design in policy evaluation.
Example
A scholarship programme awards funding to students with exam scores above 75. Researchers compare outcomes for students scoring just above 75 (treated) with those scoring just below 75 (untreated). Because the two groups are nearly identical in ability, the difference in college graduation rates can be attributed to the scholarship with high credibility.
Related Terms
Software Notes
- SPSS: Not built-in; use the
R pluginwith therdrobustpackage - R:
rdrobustpackage; e.g.,rdrobust(y, x, c = 75)for a cutoff at 75 - Stata:
rdrobustcommand (install viassc install rdrobust);rdplotfor visualisation