Governing what we don’t know: Uncertainty, source adjudication, and the accountability of national statistics offices in contested data environments

Abstract

National statistics offices are increasingly asked to be more than producers of their own observations: data stewards, integrators of administrative, citizen-generated, satellite, and AI-derived data, and custodians of public trust. This paper takes up the resulting question of institutional identity: what an office is once it publishes figures built from data it neither collected nor can fully verify independently. We argue that two decisions hidden from users now merit recognition as constitutive of the office's role, alongside its established responsibilities: the treatment of structural uncertainty (what the data cannot identify, as distinct from sampling error), and the allocation of adjudication authority (who, or what rule, decides the published value when sources conflict). We formalise their common logic as a judgment-boundary principle and operationalise it through a product-level accountability layer that jointly records evidential status and decision authority. The layer is a disciplined implementation of existing metadata governance, not a categorically new instrument. Written from inside a sub-national reform process in a resource-constrained federal statistical system, the paper develops the stakes for fragile and conflict-affected statistical systems, many, though not all, in the Global South, and confronts the political economy of disclosing what an office does not know.