Beyond methodology: A thick description of public trust in official statistics in a low-income data ecosystem — evidence from Uganda during the COVID-19 pandemic

Trust is widely acknowledged as the cornerstone of official statistics, yet most measurement frameworks in use today were developed for mature statistical systems in high-income, high-trust societies. Recent quantitative work has linked democratic political institutions to statistical capacity, but qualitative evidence from low-income African data ecosystems remains scarce. This paper reports a qualitative study of trust in official statistics conducted in Uganda in 2021–2022, during the COVID-19 pandemic. Funded by PARIS21 and implemented by Humanitarian OpenStreetMap Team (HOT) / OpenStreetMap Uganda, the study used key informant interviews and focus group discussions with producers, users, and lay publics, analysed using Geertz's thick description approach with reflexive thematic analysis. Five well-documented dimensions — methodological soundness, institutional mandate, granularity, accessibility, and comprehensibility — emerged as expected. Four additional, under-theorised dimensions, the principal contribution of this paper, also surfaced: (i) lived-experience verification, in which citizens validate statistics against what they see on the ground and treat divergence as evidence of fabrication; (ii) the “eating our money” narrative, in which government data collection is read as extractive rather than civic; (iii) messenger legitimacy, in which a statistic's credibility depends less on its producer than on who announces it, with cultural and religious authorities trusted more than state spokespersons; and (iv) trust-by-substitution, in which the National Statistical Office functions as the benchmark of last resort. While each draws on prior work in source credibility, the political economy of the African state, and institutional triangulation, their specific configuration in official statistics has not been systematically articulated. We further document an extractive-data paradox whereby the non-payment of respondents is locally re-coded as evidence that government does not value citizens’ information. We propose four corresponding additions to existing trust-measurement instruments, with practical implications for NSOs, development partners, and the wider data ecosystem. Findings are framed as analytic propositions warranting empirical verification in other comparable low-income data ecosystems rather than as generalisations.