---
title: "Does an independent source corroborate the NYC school-nudge RCT's zero-rejection deployment result (25 treatment / 28 control effective applicants, 98% acceptance)?"
type: "question"
status: "open"
date_raised: "2026-10-11T00:00:00.000Z"
tags: ["matching-markets","school-choice","nyc","randomized-controlled-trial","corroboration","single-source-concentration","arxiv"]
---


The inbox capture at `10-inbox/raw/2026-10-07-outside-arxiv261008275-personalized-recommendations-without-inducing-congestion-mitigating.md` draws four distinct claims from a single unrefereed preprint — Chiang, Peng, Lichtenstein, McDaniel, O'Neil, Thomas, Wright, Kleinberg, Tardos & Garg, "Personalized Recommendations Without Inducing Congestion," arXiv:2610.08275. `sources.md`'s single-source concentration cap permits at most three claim-notes resting on one uncorroborated primary before the next finding must route here instead (Cali's 2026-10-08 review of this capture names this exact claim as the one to hold back). Three claims were promoted to `30-notes/`: [[claim-recommendation-induced-congestion-erodes-offer-likelihood-guarantees]], [[claim-balanced-capacitated-nudging-guarantees-safe-and-near-optimal-matches]], and [[claim-nyc-school-nudge-rct-significant-ranking-suggestive-matching-increase]].

The fourth finding, held back by the cap: among treatment applicants who effectively applied to a program they were nudged to (25 such applicants), none were rejected; among control applicants who effectively applied to a program they would counterfactually have been recommended (28 such applicants), exactly one was rejected — an overall 98% acceptance rate, which the paper presents as deployment-scale validation of its central safety guarantee. Source: §6.2 / §7, direct quote in the capture: "No treatment students were rejected from a program they were nudged to (out of 25 effective applicants, i.e., not counting applicants who listed a nudged program but were accepted to another program they ranked higher), and 1 control student was rejected from a program they would have been nudged to (out of 28 effective applicants), for an overall 98% acceptance rate."

What's needed for corroboration, per `sources.md`'s definition (a different author group and a different venue — not a revised version or press summary of the same preprint): either the ACM EC 2026 published/peer-reviewed version of this same paper (if its reported numbers hold up under review), or an independent replication/audit of the 2025-26 NYC deployment data. The paper's own pre-registration (AEA RCT Registry, trial 17146, https://www.socialscienceregistry.org/trials/17146) is itself a lead worth checking first — it may fix the sample-size and outcome-measure commitments in a way that lets a future session assess whether 25/28 "effective applicants" matches what was pre-registered, which is adjacent to but not the same as independent corroboration.

Candidate next move: when a session with fetch tools next touches this topic, check whether the ACM EC 2026 proceedings version has posted, and separately pull the AEA RCT Registry record.

## Progress log

- 2026-10-14 (promotion of `10-inbox/raw/2026-10-12-does-an-independent-source-corroborate-the-nyc-school.md`): checked and still open — **partially answered, not resolved.** The ACM EC 2026 proceedings version has posted (accepted, item 36 on the official program) under an unchanged author roster, and its camera-ready PDF restates the headline finding near-verbatim — internal consistency through peer review, not independent corroboration. No outside source (Cornell's own research-communications office, education journalism, an academic citation) was found discussing the figure at all; the preprint is only six days old, which is the likeliest explanation for silence rather than any judgment on the result. Recorded as [[claim-no-independent-source-corroborates-nyc-nudge-rct-zero-rejection-result]]. Still open: the held-back zero-rejection/98%-acceptance figure itself remains unpromoted pending independent corroboration.
  - Metadata note (not a vault claim, logged here per the sourcing floor's metadata rule): the EC26 program page's roster quote spells the fifth author "Deja Thomas" and credits her to "Cornell Tech," while the paper's own body spells her "Dejaunique Thomas" and credits "New York City Public Schools." The 2026-10-12 capture treated the two rosters as identical; they agree on everyone else and disagree only on this one name/affiliation pair. Worth a clean re-check (the paper's front matter, a later proceedings metadata record) before citing either spelling as authoritative.
  - Unresolved leads carried forward: AEA RCT Registry trial 17146 still 500s on direct fetch (candidate for `sources.md`'s known-blocked-routes list); news.cornell.edu's article page still 403s to `archive_page` (its "59% more likely to apply" figure remains `[unverified-quote]`); Chalkbeat NYC not yet site-searched directly for this deployment; Peng & Ryu's *Nature Cities* undermatching paper not yet checked for whether it discusses this deployment as related work; the claimed EC26 "Best Paper with Student Lead Author" award not yet checked against the EC26 awards page.