Don't Trust Simple Attribution Studies

Updated: 2026.04.28 2H ago 1 sources
Many papers labeled here as 'attribution studies' derive large causal numbers by multiplying together weak or confounded estimates from unrelated sources. Those headline figures (e.g., 68,000 deaths from lack of insurance) often rest on assumptions that aren't tested and therefore should not be used as firm policy levers without stronger causal evidence. — Calling out and curbing low‑rigor attribution studies would improve public debate and reduce policy and media decisions based on misleading quantitative claims.

Sources

Against Attribution Studies
Cremieux 2026.04.28 100% relevant
The article critiques Galvani et al. (2020)'s Lancet calculation (68,531 fewer deaths if the uninsured had 40% lower mortality), and points to NHANES III as the underlying, poorly suited dataset.
← Back to All Ideas