3. Policy Evaluation Designs

Regression Discontinuity

Turn a treatment threshold into a local causal comparison and audit continuity, manipulation and bandwidth

Regression Discontinuity

The cutoff creates the comparison

Northbridge awards an offer when an application score reaches 70. Compare applicants just below and just above 70:

ScoreOffer ruleCompletion rate
68–69no offer0.49
70–71offer0.57

The eight-point gap is credible only if, without the offer, expected completion would change smoothly through 70. RDD does not claim that students with scores 40 and 90 are comparable.

For a sharp design with running variable XX, cutoff cc and treatment D=1(Xc)D=1(X\ge c),

τRD=limxcE[YX=x]limxcE[YX=x].\tau_{RD}=\lim_{x\downarrow c}E[Y\mid X=x]-\lim_{x\uparrow c}E[Y\mid X=x].

This is a local effect at the threshold.

Draw before estimating

  1. plot the raw outcome against the running variable;
  2. show binned means only as a visual summary;
  3. fit separate local lines on each side;
  4. report the bandwidth and polynomial order;
  5. keep the cutoff and data support visible.

A global fifth-order polynomial may fit the sample yet invent extreme curvature near the cutoff. Local linear or local quadratic fits are usually easier to defend. Cattaneo, Idrobo and Titiunik (2020) provide a design-based practical treatment of estimation and diagnostics.

Bandwidth is a bias–variance decision

Suppose the estimated completion effect is:

Window around 70ApplicantsEstimateRobust 95% interval
±22600.09−0.01, 0.19
±56100.070.01, 0.13
±101,1800.120.08, 0.16

The widest window is most precise but asks continuity over a larger range. Do not select the window because it gives the preferred pp-value. Use a principled bandwidth, then show nearby choices as sensitivity checks.

Sharp, fuzzy and kink designs

DesignWhat changes at ccTypical estimand
sharp RDDtreatment status from 0 to 1local treatment effect
fuzzy RDDtreatment probability jumpslocal IV effect for cutoff compliers
regression kinkslope of treatment intensity changeslocal effect per unit of intensity

If offer probability rises from 0.10 to 0.75 at 70 and completion rises from 0.49 to 0.57, the fuzzy Wald estimate is

0.570.490.750.10=0.123.\frac{0.57-0.49}{0.75-0.10}=0.123.

Interpret 12.3 percentage points as a local effect for applicants whose receipt is changed by crossing the cutoff—not an effect for every applicant.

Four threats and their probes

ThreatProbeWhat failure means
precise score manipulationdensity and institutional auditunits around cc may not be comparable
another rule at 70policy inventorydiscontinuity may combine treatments
predetermined covariate jumpbalance plots and placebo outcomescontinuity is doubtful or coding is wrong
functional-form dependencebandwidth/order sensitivityestimate relies on extrapolation

A density test is not a ritual. If assessors can round 69.8 to 70, inspect score production, heaping and appeals. Conversely, a visible density change does not identify its cause without institutional evidence.

What RDD does not identify automatically

  • effects far from the cutoff;
  • effects under a different treatment version;
  • effects when units influence one another across the threshold;
  • a policy expansion that changes capacity or applicant behaviour.

The rdrobust project implements local-polynomial estimation, robust bias correction and companion diagnostics. Software cannot establish that no other rule changes at the cutoff.

Quick check

Scores are recorded only as integers, and 35% of applicants score exactly 70. Is “use a narrower bandwidth” an adequate solution?

Answer
No. The running variable has limited support and a mass point at the cutoff, so arbitrarily narrow continuous-score approximations are not credible. Audit how scores are assigned, report support, consider methods for discrete running variables and make the stronger identifying assumptions explicit.

Next: Matching and Weighting

Copyright © 2026