Fugitive Statistics: A Verification-Chain Protocol for Preserving Policy Numbers That Exist Only Inside Documents

A growing share of policy-critical statistics is published only as tables embedded in advisory-committee documents — ad hoc tabulations produced for a single meeting, carried exclusively by PDF, with no versioned machine-readable release, no persistent identifier, and no documented reproduction path. We term these fugitive statistics, by analogy with the fugitive documents of library science — government publications that fall within the scope of a depository system yet escape its collection net. Fugitive statistics are load-bearing: they inform national workforce projections, capacity planning, and crisis-response policy, yet they are exposed to link rot, silent revision, format death, and provenance loss. Crucially, ex-post policy evaluation is possible only if the numbers used ex-ante survive: Japan’s Ministry of Health, Labour and Welfare (MHLW) could score its 7th nursing workforce projection against realized outcomes (deviations of 2.98% and 2.36% for demand and supply) only because the original projection documents remained retrievable. We present a lightweight verification-chain protocol — capture, SHA-256 hashing, independent witnessing via the Internet Archive, semantic extraction with cell-level provenance, bitemporal cataloguing, DOI deposit, and revision monitoring — and demonstrate it end-to-end on the document set of the 4th meeting of MHLW’s 2040 nursing workforce committee, in which prefecture-supplied municipal microdata were re-tabulated into secondary-medical-area statistics that exist nowhere else. All nine artifacts were captured, hashed, and independently witnessed (8/8 PDFs byte-identical to Internet Archive copies), and the embedded tables extracted under two-tier quality control (row-sum checks passed 60/60; double-entry cross-verification matched 916/916 cells), yielding a DOI-bearing dataset within one operator-day on commodity hardware. We position preservation as the precondition of accountability, and fugitive statistics as a measurable site of institutional debt: the divergence between the statistics a policy decision consumed and the statistics later available to evaluate it.