Source
Prashant Garg and Thiemo Fetzer, “Causal Claims in Economics” (OSF 2024; arXiv v2, February 2026), LIT-079, read in full in NOTE-083. The claim is the paper’s §3 reading of its own series: “This upward trend indicates that economic papers have increasingly incorporated causal inference methods to substantiate their claims over the past three decades.”
What this asserts
That the rise from 7.7% (1990) to 31.7% (2020) in the share of extracted edges labelled design-backed is a rise in what economics papers actually claim on design-based evidence, by field roughly as the paper reports (Health, Urban and Development highest after 2000, Macro lowest), and not an artefact of how a language model reads older and newer papers.
What the record holds toward it
THEORY-674 (Active) holds the series as the extraction’s output. Its direction is robust to the edge-overlap threshold; its level is not; and the per-decade slope falls from +0.0187 to +0.0011 once method mix and graph volume are controlled. Paper-level method labels agree with Goldsmith-Pinkham (2024) on the field ordering. No edge was checked against a human label, and the paper-level RCT label reaches F1 about 0.03.
What would confirm or reject it
The promote_when, carried over unchanged from THEORY-674: an edge-level audit against human labels on a sample drawn by decade, field and method. High and stable precision and recall for the causal label would confirm it; low precision on 1980s and 1990s papers, or recall rising sharply with time, would show the rise is in the extractor.
What this does not say
- That design-based claims are correct, or better identified than the claims they replaced; the pipeline records what papers claim.
- That causal claims are rewarded with top-five publication; that is a conditional association in the paper and is not part of this claim.