NBER paper: a third of significant development economics findings may be false
gleech · x · 2026-11-10
A new NBER working paper by Qingyuan Chai, Kevin Lang, and Liang Zhong analyzes 745 primary-hypothesis tests from 190 papers in leading economics journals and questions the statistical credibility of RCTs, finding that nearly a third of statistically significant results in development economics may be false discoveries. Quote-tweeters pair it with the classic "Laws of Evaluation" — the Iron Law (expected net impact of large-scale social programs is zero) and the Stainless Steel Law (the better designed the evaluation, the more likely the estimated impact is zero) — noting the new findings echo decades-old pessimism about program evaluation.
More from Fun
- Even power users spend weekends decoding frontier model usage quotas — menhguin · 2026-11-10
- Seinfeld AI skit: George is now enjoying the "Chinese" models too — banteg · 2026-11-10
- AI-generated character Ananya goes viral with "GOD IS A WOMAN" meme — ashishllm · 2026-11-10
- YouTube served his video to viewers who 'like watching horrific car crashes' — billyjhowell · 2026-11-10
- AI game engine spawn7 shows first frames, promising 10x GTA V-sized worlds that load in a second — TAbrodi · 2026-11-10
- Parody S-1 filing mocks Anthropic: 'exclusionary coalition' exposure could sink model commercialization — OwariDa · 2026-11-10