Methods for Producing Economic Data: Are Surveys Becoming Less Reliable? Evidence, Risks, and a Quality-Assurance Framework for Nigeria and Beyond

Economic data shape fiscal and monetary policy, social protection, investment decisions, credit ratings, development finance, and the evaluation of public programmes. Yet the reliability of the data used for these purposes cannot be taken for granted. This article critically examines whether survey-based economic data are becoming less reliable and, more importantly, identifies the conditions under which survey estimates remain fit for purpose. The paper is an integrative methodological review that reorganises and extends the original manuscript through the total survey error perspective, official-statistics quality frameworks, and recent evidence from household-survey practice, with particular attention to Nigeria. The analysis distinguishes sampling error from coverage error, nonresponse, measurement error, interviewer and mode effects, processing and imputation errors, and analytical or reporting distortions. It argues that declining response and changing communication technologies create genuine risks, but that low response rates do not mechanically imply high bias, and surveys should not be contrasted with “statistical data†because surveys are themselves a major source of official statistics. Evidence from Nigeria illustrates both the continuing value of well-designed probability surveys and the importance of mode-specific validation: recent experimental work finds meaningful differences between phone and face-to-face responses across common economic and welfare indicators. The article proposes a data-quality strategy based on probability sampling where feasible, mixed-mode and responsive designs, computer-assisted collection, transparent metadata, validation against administrative or transaction records, selective integration of geospatial and digital data, explicit treatment of uncertainty, and reproducible analysis. The central conclusion is that surveys are not obsolete; rather, single-source, weakly documented and poorly validated data-production systems are increasingly difficult to defend. Reliability is best achieved through transparent, multi-source statistical systems in which surveys remain a core but continuously audited component.