Skip to main content

Effect of a scholarship on academic performance (RDD)

This illustrates RDD: a scholarship is awarded to students with an admission score ≥ cutoff. Comparing students just above and just below the cutoff (as-good-as random) yields a credible causal estimate at the threshold. Figures are illustrative.

Summary: estimate the jump in academic performance (next-year GPA) at the score cutoff = the causal effect of the scholarship at the threshold.


Step 1 — Ideation

  • Question: does receiving the scholarship improve academic performance?

Step 2 — Literature Review

Impact evaluation of financial aid in education; regression discontinuity design.

Step 3 — Data Collection

RoleVariableDescription
Running variablescoreadmission score
Cutoffcutoffscholarship threshold
Treatmentscholarship1 if received (sharp: score ≥ cutoff)
Outcomegpa_nextnext-term/year GPA

Step 4 — Modeling

Choose the Causal inference family → RDD (sharp); declare the running variable, cutoff, bandwidth:

Illustrative results (format — not real results):

Value
GPA jump at cutoff (LATE)+0.18***
Optimal bandwidth±0.5 points
McCrary (p)0.42 (no manipulation of the cutoff)

Sample interpretation: the scholarship raises GPA by ~0.18 at the threshold; a non-rejected McCrary test ⇒ no sign of score manipulation around the cutoff. This effect is local to the cutoff (LATE).

* ── RDD: scholarship effect on GPA ────────────────
* Install: ssc install rdrobust
rdrobust gpa score, c(50)

* ── RDD plot ──────────────────────────────────────
rdplot gpa score, c(50)

* ── Optimal bandwidth selection ───────────────────
rdbwselect gpa score, c(50)

Step 5 — Reporting

Export a report + the RDD plot (scatter + fitted lines on both sides of the cutoff) + replication code.

Video tutorial

Video Tutorial: Guide to running RDD in EcoLab

See also