Do Statewide Pre-K Programs Have Persistent Effects? Evidence from North Carolina (with Dodge et al. 2017 and Bai et al. 2020)

1,207,576 (Watts); ~1,004,571 (Dodge); ~900k/grade (Bai)

North Carolina cohorts exposed to the state's More at Four / NC Pre-K program.

DOI CORRECTION NEEDED — the previously recorded DOI (10.1016/j.jpubeco.2023.104908) resolves to an unrelated paper and has been removed. Crossref search could not confidently identify the correct record. Cite by title until re-established. Two-way fixed-effects design on county funding DOSE — a population-level, not participant-level, estimate; the headline SD figure is IV-rescaled from a funding coefficient. This is the WEAKEST of the three modern scaled-pre-K designs.

The strongest counterexample to 'no scaled pre-K shows persistent gains' — that claim is false as stated. But this evidence is population-level, dose-based, IV-rescaled, and internally inconsistent between papers, so it does not override two cleaner randomised designs pointing the other way. Record it as the live dissent, not as the answer.

County-level variation in program funding intensity, not individual assignment.

County-level funding variation; individual selection into the program is not controlled.

The sole design showing durable, growing gains from scaled pre-K — and it is contradicted by TN-VPK (RCT, negative at grade 6) and by Boston (lottery, which rules out effects above ~0.12-0.14 SD). Its two constituent papers also contradict each other.