U.S. preclinical research consumes ~$28B/year on irreproducible work

Freedman, Cockburn, and Simcoe 2015 estimated that preclinical research in the United States alone consumes approximately $28 billion per year on work that cannot be reproduced.

The estimate is methodologically derived (not measured directly) by combining published reproducibility-failure rates with NIH expenditure data. The number is contestable in its precision, but its order of magnitude has held up across subsequent reanalyses, and the methodology is transparent enough to permit replacement of any input.

The relevance to C-0002 is the cost dimension: the 73-93% non-verifiability rate is a condition, not yet a cost; this is one of the earliest serious estimates of what the condition costs the U.S. preclinical sector specifically when reproducibility failures are priced. The aggregate scales with sector size, and Section 4.1 of the paper tracks the same pattern through reproducibility-failure measurements at smaller corpora (Amgen 11%, Molecular Brain 97%, Stern et al. retraction costs).

This Evidence node primarily supports the cost-side argument that C-0001's architectural framing is worth taking seriously — that the carrying cost is large, not negligible.