AI
An AI-Powered Multiverse Analysis of Empirical Finance Research: Evidence from Reused Natural Experiments
Publication: cs.AI updates on arXiv.org
Author: Ruijiang Gao, Steven Chong Xiao
Source published: 20/11/2026, 05:00:00 UTC
Briefing created: 20/11/2026, 05:26:14 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 20/11/2026, 05:03:24 UTC.
Briefings
An AI agent analyzed 324 empirical finance papers reusing 13 natural experiments, rebuilding data across 62,409 specifications to measure robustness. The study found 42% of main estimates reproduced under the papers' own specs, with 25% maintaining significance across all grids. Regression discontinuity claims showed higher robustness than difference-in-differences, while first-order effects and top journal claims were more robust. The study demonstrates feasibility of large-scale multiverse analysis using agentic coding.
Key points
- AI agents analyzed 324 empirical finance papers reusing 13 natural experiments.
- Across 62,409 estimated defensible specifications, 25% of claims kept published significance.
Context
Empirical finance research frequently reuses natural experiments to study other outcomes, raising concerns about multiple testing and publication bias. This analysis quantifies how much published claims depend on analysis choices.
Limitations
Evidence is limited to 324 papers from arXiv; no additional background or source-supported context provided. Findings represent observed robustness patterns rather than forecasts or causal claims.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 20/11/2026, 05:26:14 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeiegjkrjy6yxfx57qpj4lkzpnta3efmawlazefi4wjf5c7ns2rf2my