Skip to content

Regression Discontinuity Design

Recover a causal effect from a threshold rule by comparing units just above and just below a sharp cutoff on a continuous running variable, where they are comparable in expectation, so any jump in the outcome at exactly the cutoff is attributable to the treatment rather than to selection.

Core Idea

Regression discontinuity design (RDD) is a quasi-experimental technique exploiting settings where treatment is assigned by a sharp cutoff on a continuous running variable — a test score, an age, an income line, a vote share. Because tiny fluctuations near the cutoff are uncontrolled by subjects, units on either side are comparable in expectation, so any jump in the outcome at the cutoff is attributable to treatment. Sharp RDD assigns treatment deterministically; fuzzy RDD only shifts its probability, making the cutoff an instrument that recovers a local effect for compliers.

Scope of Application

RDD's home is empirical causal inference on administratively-assigned treatments; it travels intact wherever a continuous running variable carries a sharp cutoff, only the score's referent changing.

  • Education-policy evaluation — the founding turf: scholarship, school-entry, and grade-retention cutoffs.
  • Health policy — the Medicare-at-65 threshold, BMI eligibility lines.
  • Labour and welfare policy — benefit-eligibility ages and means-tested income lines.
  • Crime and sentencing — mandatory-minimum thresholds on drug quantity or prior convictions.
  • Election studies — close-race vote-share discontinuities around 50%.

Clarity

Naming RDD makes legible a class of causal opportunities that correlational analysis renders invisible: sharp administrative thresholds become identification strategies hiding in plain sight. It relocates a confounded pooled comparison to the neighbourhood of the cutoff, licensing a local, answerable question — at the threshold, is the outcome smooth or does it jump? Its second clarity is that the identifying assumption is visible and testable, and it sharpens two distinctions: sharp versus fuzzy (which sets the estimand) and local versus global (the effect is credible exactly where it is narrow).

Manages Complexity

Recovering a causal effect from observational data is generally intractable — the treated and untreated differ on everything that drove treatment. RDD collapses this to a single local question by exploiting one structural feature: a sharp cutoff. It is governed by a short, decidable checklist — is there a running variable with a cutoff; do covariates balance; does the McCrary density test show no bunching; is assignment sharp or fuzzy — from which both the validity and the meaning of the estimate read off along a clean branch structure, including its deliberately narrow scope.

Abstract Reasoning

RDD licenses spotting found randomization in a threshold rule (reasoning from an administrative cutoff to a local natural experiment), and converting the identifying assumption into testable diagnostics (the McCrary manipulation test and covariate-balance checks, so credibility is earned by surviving checks). A classification move sets the estimand by reading the treatment-probability step (sharp ATE versus fuzzy complier-LATE). A scope-bounding move keeps internal and external validity apart, reporting the effect together with the narrow population it is identified for, with bandwidth the explicit bias-precision knob.

Knowledge Transfer

What transfers is a research design, not a world-pattern, and its within-domain reach is wide and literal because its precondition is exact: a continuous running variable with a sharp cutoff. The same checklist and vocabulary (running variable, cutoff, bunching, sharp/fuzzy, bandwidth) run untranslated across education, health, labour, crime, election, regulatory, and development studies. It does not export to natural science as such. Where a cross-domain lesson is wanted, the methodological move rides on the natural-experiment / instrumental-variables parent, and the threshold-discontinuity world-pattern recurs in control engineering and physics under their own frameworks — not as RDD traveling.

Relationships to Other Abstractions

Local relationship map for Regression Discontinuity DesignParents appear above the current abstraction, mutual partners to the right, and children below. Node labels state whether each abstraction is prime or domain-specific; colors identify relation types.RegressionDiscontinuity DesignDOMAINDomain-specific abstraction: Natural Experiment — is a kind ofNaturalExperimentDOMAIN

Current abstraction Regression Discontinuity Design Domain-specific

Parents (1) — more general patterns this builds on

  • Regression Discontinuity Design is a kind of Natural Experiment Domain-specific

    Regression discontinuity design is a natural experiment specialized to found as-if-random assignment at a sharp cutoff on a continuous running variable.

Hierarchy paths (12) — routes to 8 parentless roots

Neighborhood in Abstraction Space

Regression Discontinuity Design sits in a sparse region of the domain-specific corpus (89th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (309 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-07-12