To meet increasing user demands for granular demographic and socio-economic indicators under tightening budgets, national statistical offices have to continuously engage in methodology research and development, including harnessing big data, satellite imagery, and transactional sources to improve or redesign data collection instruments. AI can help with that ongoing effort. This paper focuses on using AI-generated algorithms to test, verify, and illustrate a new statistical methodology before it is trusted for production. Trust in the use of AI in official statistics has many dimensions; we deliberately restrict this paper to two pillars: an independent statistical verification step before the algorithm is trusted for production; and disciplined protection of respondent confidentiality throughout development and testing. We illustrate this process of building trust, drawing on the author’s own experience directing AI to build and run the algorithm for a Mini-Max Hierarchical Bayes (HB) sampling method. This paper also gives a checklist that any national statistical office can apply when evaluating an AI-assisted or algorithmic method for production use. The protocol and checklist are cross-checked against two established reference points for methodological practice in official statistics: the UN Fundamental Principles of Official Statistics and the HLG-MOS Quality Framework for Statistical Algorithms.
This paper evaluates a reproducible probabilistic forecast reconciliation pipeline for official tourism statistics based on Eurostat data. The study does not propose a new reconciliation estimator; rather, it documents how established tools—covariance-weighted reconciliation (MinT-W), nonnegative post-processing with re-coherence (NonNeg), prequential evaluation, and proper scoring rules—can be operationalized in a transparent, file-backed statistical production pipeline. The focus is on coherence, nonnegativity, rolling-origin prequential evaluation, and reproducibility in an official-statistics setting. Using Eurostat's monthly tourism-nights series (tour_occ_nim) for a two-level retained-country hierarchy of 26 member states and their aggregate—constructed as the sum of these 26 series, and not the published Eurostat EU27_2020 aggregate— we construct rolling-origin probabilistic forecasts at horizons h ∈ { 1 , 3 , 6 , 12 } months, reconcile them with MinT-W and NonNeg, and assess performance using the Continuous Ranked Probability Score (CRPS), Logarithmic Score (LogS), and Diebold–Mariano tests with Holm adjustment. All forecasts and evaluation outputs are generated by a deterministic pipeline with fixed random seeds, single-threaded numerical routines, and explicit validation checks, so that every table and figure can be regenerated exactly from the accompanying reproduction package.
Official statistics increasingly draws on non-traditional sources, such as social-media posts and other web-scraped text, analysed with AI language models for nowcasting, indicator production, and crisis monitoring. We show that such pipelines can manufacture false statistics with no misinformation actor involved: ordinary, invisible data-processing choices are enough. Using standard, publicly available zero-shot classifiers, we give fully reproducible demonstrations on conflict-related messages. A routine “N comments” interface string captured by scrapers can collapse a model's judgement that a message concerns armed conflict from near-certain to below the retention threshold, while a benign sentence of equal length does not; the effect is large on a widely used model yet negligible on others. Four equally defensible phrasings of a single analytical question yield almost disjoint datasets from identical inputs, and swapping the classifier alone moves an unambiguous conflict report from retained to discarded. These distortions are silent and idiosyncratic: pipelines return confident, plausible numbers, and the direction and size of error depend on the specific model, prompt, and preprocessing. We argue that interpretation artifacts are a distinct threat to data quality and public trust, and propose a short, demonstrated audit that statistical agencies can adopt.
Urbanization is a major driver of economic development but also intensifies pressure on land resources, making sustainable urban growth an important challenge for rapidly developing regions. This study proposes an integrated framework for assessing Sustainable Development Goal (SDG) Indicator 11.3.1 (Land Use Efficiency, LUE) by combining multi-source satellite imagery, machine learning, the Degree of Urbanization (DEGURBA) framework, and official population statistics. Bali Province, Indonesia, was selected as the study area due to its rapid tourism-driven urbanization and strategic role as one of the country's metropolitan regions. Multi-temporal land cover maps for 2010, 2015, and 2020 were generated using the Hist-Gradient Boosting (HGB) classifier, which achieved overall accuracies of 80.9%, 84.8%, and 85.9%, respectively. The resulting built-up maps were integrated with gridded population data to derive Land Consumption Rate (LCR), Population Growth Rate (PGR), and LUE at the subdistrict level based on the DEGURBA classification. The results indicate that built-up area expanded by approximately 71.2% during 2010–2020, while the overall urban hierarchy remained relatively stable, suggesting that urban growth primarily occurred through the densification and outward expansion of existing urban centers. Furthermore, the analysis demonstrates that LUE should be interpreted together with its constituent indicators, as exceptionally high or negative LUE values were largely associated with variations in population growth rather than built-up area change. The proposed framework provides a reproducible and spatially explicit approach for integrating Earth observation data with official statistics to support SDG 11.3.1 monitoring and evidence-based urban planning in Indonesia.
Integration of data from mobile network operators (MNOs) with suitable data from non-MNO sources is by and large necessary to serve the defined interests of official statistics. In this paper we summarise key insights from the ESSnet project Mobile Network Operator Methods for Integrating New Data Sources (MNO-MINDS), distinguish and explain the central ideas of possible statistical approaches, discuss how some existing methods may be extended or improved, and highlight certain specific issues that may not have received adequate attention previously.
Māori, the Indigenous people of Aotearoa New Zealand, view health as holistic; ill health, like Coronavirus disease 2019 (COVID-19), therefore has implications far beyond physical health. To improve service delivery to Māori, it is essential to understand their perspectives. This article shares findings from the Kia Kitea ai te Ora Life-Changing COVID-19 research project which provides insight into Māori views on COVID-19 infection and Māori experiences of COVID-19-related support services in Aotearoa New Zealand. Specifically the impact of COVID-19 on Ngāti Maniapoto, a Māori iwi (tribe), whose traditional lands are located in the central-west of the North Island of New Zealand. This kaupapa Māori mixed-methods project involved a survey and interviews to gather Indigenous knowledge on the views of tribal members who experienced COVID-19 between March 2021 and March 2023. With a focus on the lived experience narrative of COVID-19 and the associated post-COVID-19 environment. The findings of this research identified and confirmed opportunities for improvement in responses to COVID-19 requiring further investigation. Findings from the participants’ lived experiences also offer opportunities to promote Indigenous-led action concerning health inequities to mitigate the repercussions of the COVID-19 for Indigenous communities.
Administrative datasets, characterized by large scale and high volume, often exhibit data quality issues, including reporting delays, missing entries, and irregular reporting. Outlier detection as a precursor to imputation is a critical step for reliably identifying missing reports and ensuring accurate downstream estimates. This paper proposes a novel two-step method for detecting irregularities in administrative reported data by combining clustering with robust outlier detection using the median and the median absolute deviation. Through 10,000 simulations across 10 distinct scenarios, we evaluate the proposed method against established approaches, including the mean ± SD, boxplot, and ratio-to-median methods. The results show that the proposed method matches or outperforms traditional one-step methods in eight scenarios and ranks second best in the remaining two. We further apply the method to monthly counts from 136 NIBRS agencies, examining the frequency and severity of flagged observations, comparing results with the ratio-to-median approach, and assessing sensitivity to the primary tuning parameter k . Overall, the proposed framework provides a robust and interpretable approach to identifying reporting irregularities in administrative data, with clear implications for improving data quality and downstream statistical estimation.
The 2023 Uruguayan Census yielded an estimated total population of 3,444,451 at an estimated undercoverage of 10.3%. How a national statistical office corrects such differential undercoverage is inseparable from what that office is: a question of institutional capacity and data governance as much as of statistical method. Administrative records recover counts but not selection bias, leaving enumerated-microdata estimates biased. We treat enumerated households as a non-probability sample and weight them with a doubly robust estimator combining a response-propensity model (web linkage rate as contact proxy) with calibration to census demographic totals, at three million records. We then offer a framework to guide statistical offices in choosing adjustments given their registers and paradata. We argue that this choice is not merely technical but institutional: it depends on an office's capacity to integrate administrative registers, to capture and retain paradata, and to field a powered post-enumeration survey, together with the governance conditions (legal access to data, interoperability, autonomy, and public trust) that make such capacities possible. Read this way, the Uruguayan experience clarifies what a national statistical office must be able to do to correct differential undercoverage, and what it implies for offices, particularly across South America, that operate without those conditions.
National statistical systems are central to how states recognise populations, define public issues, allocate resources, monitor inequities and report progress. For Indigenous Peoples, official statistics shape whether collective and individual rights are visible, actionable and accountable within public policy. Yet Indigenous Peoples remain invisible, misclassified, externally defined or deficit-framed in many official statistical systems. This article argues that identification, disaggregation and technical quality are necessary but insufficient foundations for statistics that uphold Indigenous rights. Disaggregation can support visibility, but visibility without Indigenous authority can reproduce surveillance, extraction, statistical misrecognition and harm. This article develops the Indigenous Statistics Governance Framework as a rights-based analytic framework for assessing how Indigenous Data Governance can be embedded within official statistical systems. The framework is grounded in international human rights instruments, Indigenous Data Sovereignty and Indigenous Data Governance scholarship, official statistics standards, data stewardship principles and practice examples drawn primarily from Canada, Australia, Aotearoa New Zealand and the United States. It identifies four interdependent system functions: Recognition, Stewardship, Authority and Accountability. Recognition concerns whether Indigenous Peoples are made visible in ways that are accurate, meaningful and governed. Stewardship concerns whether Indigenous data are cared for safely and responsibly across the statistical lifecycle. Authority concerns whether Indigenous Peoples have decision-making power over statistical priorities, concepts, classifications, interpretation, access and use. Accountability concerns whether statistics support obligations, implementation, equity, redress and institutional performance. The framework is offered as a transferable analytic framework, not a universal implementation model. Its wider application requires contextual adaptation, particularly in statistical systems where Indigenous Peoples are not legally recognised, where identity is captured through alternative classifications, or where Indigenous governance institutions are absent, emergent or contested. The article reframes official statistics as rights-enabling and rights-implicating infrastructure and provides a practical structure for assessing whether Indigenous Data Governance is embedded in statistical practice.
Official statistics provide an essential basis for monitoring gender equality, evaluating public policies, and supporting evidence-based decision-making. Using official gender statistics from Eurostat, this study comparatively assesses gender equality across 31 European countries through six indicators covering employment, care-related labor force inactivity, political representation, women in senior management, early school leaving, and tertiary educational attainment. Country performance was evaluated using an integrated multi-criteria decision-making framework combining Equal Weighting, Entropy, Criteria Importance Through Intercriteria Correlation (CRITIC), TOPSIS, MOOSRA, MABAC, Preference Selection Index (PSI), Range of Value (ROV), Weighted Sum Model (WSM), Weighted Product Model (WPM), Weighted Aggregated Sum Product Assessment (WASPAS), and Borda approaches. The findings reveal substantial disparities across Europe. Nordic countries consistently achieve the highest performance, reflecting stronger institutional support, lower care-related employment barriers, and higher levels of women's representation. In contrast, several Southern and Eastern European countries continue to experience multidimensional disadvantages. The results demonstrate that gender equality extends beyond labor market participation and should be understood through the combined effects of education, leadership, political representation, and care responsibilities. By integrating official statistics with an objective multi-criteria evaluation framework, the study offers a transparent and replicable approach for comparative gender equality assessment and provides evidence to support gender-sensitive policymaking and the monitoring of progress toward Sustainable Development Goal 5.
Meaningful Citizen Data integration on Official Statistics requires a shift from viewing citizen data as a supplemental input toward recognising co-production as a legitimate mode of statistical work. The article proposes a structured model for integration, examines the institutional, ethical, and methodological conditions necessary for sustainability, and considers the risks involved. National Statistical Offices increasingly confront persistent data gaps and rising expectations for timely and disaggregated information and new demand from SDG requirement. Traditional sources such as censuses, surveys, and administrative registers struggle to keep pace with demands for responsiveness, granularity, and representation. Citizen data produced by civil society organisations, community actors, and local monitoring initiatives has emerged as a promising yet underutilised resource. Building on the Copenhagen Framework on Citizen Data endorsed by the United Nations Statistical Commission in 2025, this article advances a conceptual and policy-oriented framework for integrating citizen data into official statistics.
A crucial question faced by statistical authorities is how the use of different data sources, namely surveys and administrative files, affects the compilation of specific statistics. This paper analyzes aggregated data from Greek structural business statistics (2014–2018), comprising 8 indexes across 140 business branches originated from two sources, survey and administrative files. Analysis reveals whether computed indexes differ by data source, repeated for each index. The structure of the data is that of repeated measures. The statistical testing was performed using two parametric statistical tests, the two-way repeated measures ANOVA and the linear mixed models, as well as the respective bootstrap tests (wild bootstrap for the linear mixed models). The consistency of the parametric and bootstrap tests was first assessed on simulated data, using the data setting of the real data but determining different scenarios for the dependence of the statistic on each factor. The simulation study concluded that, even for strong deviations of the data from normality, the parametric and bootstrap tests result to the correct test decisions. Real data analysis confirms agreement between parametric and bootstrap tests, with two indexes showing statistically significant data source effects. The findings suggest that the data source may have an impact on the derived statistics.
The World Programme for the Census of Agriculture 2030 (WCA 2030) marks a paradigm shift in how countries design and implement agricultural censuses. It explicitly encourages the integration of Earth Observation (EO) and geospatial data to enhance efficiency, accuracy, and comparability. This paper presents a methodological synthesis of how EO can be embedded across the census cycle—from the preparation of geospatial reference layers and georeferencing of holdings to validation and area estimation. Drawing on lessons from FAO's EOSTAT programme, the UN Handbook of Remote Sensing for Agricultural Statistics, and innovative examples such as Brazilian Institute of Geography and Statistics's (IBGE) AI-based parcel delineation in Brazil, this article illustrates practical pathways for operationalization. The analysis emphasizes institutional readiness, quality assurance, and emerging AI-based approaches that enable scalable, cost-effective census operations aligned with WCA 2030 standards.
International organizations including the Bank for International Settlements (BIS) have adopted SDMx (The standard for Statistical Data and Metadata) as the standard for exchanging official statistics. Trust in published data is essential for evidence-based policymaking. This paper shows how binding each SDMx dataset to its source using blockchain technology can enhance confidence in official statistics. We present a proof of concept implemented on the XRP Ledger (XRPL) and contribute, as an integrated whole, (i) an SDMx-native canonicalization and per-< Series > hashing pipeline, (ii) a domain-separated Merkle aggregation scheme for batched anchoring, (iii) a self-contained, identity-bound verification artefact in which the SDMx message itself carries both the ordered Merkle leaves and a W3C Verifiable Credential signed by a publisher identity key cryptographically bound to the publisher’s XRPL address via an on-chain attestation registry, so any consumer can re-derive the anchored root and verify the publisher’s identity from the file alone plus a single ledger lookup, (iv) an open-source XRPL-based reference implementation, and (v) a cost model that captures the batch-size / latency / fee trade-off and is solved for an economically optimal batch size. The system enables near-real-time data verification, provides cryptographic integrity guarantees, and establishes a foundation for future extensions, including zero-knowledge proofs and automated verification by AI agents. Measurements on the prototype show median publication latency of 3–5 s and verification latency of 1–2 s under the controlled test conditions described in Section 7. The approach is data-format-agnostic and can be extended to other structured statistical or regulatory formats.
National statistical offices face an intensifying dilemma: rising demand for domain-level estimates under fixed or shrinking budgets. This paper presents a practical two-stage strategy for minimising survey sample size while preserving pre-defined precision targets for all target variables across all geographic domains simultaneously. Stage 1 applies Bethel allocation, which finds the globally minimum sample satisfying all coefficient-of-variation (CV) constraints at once. Stage 2 asks whether this Bethel sample can be reduced further via Hierarchical Bayes (HB) small area modelling; a nested sub-sampling algorithm with four eligibility gates identifies the largest achievable reduction. Applied to a synthetic labour-force population ( N = 1,000,000 ; H = 100 strata; D = 10 domains), the strategy reduces the required sample from 91,308 to 18,262 — an 80% reduction — while meeting all national and domain CV targets. A Monte Carlo study ( B = 1,000 ) confirms that CV targets are passed in more than 95% of replications for all three target variables, and credible-interval coverage is close to the nominal 95%. Four sensitivity scenarios varying auxiliary-variable strength, domain heterogeneity, and event rarity confirm that HB achieves an 85% reduction in each case, with the Bethel baseline scaling appropriately to the difficulty of each setting. The principal trade-off is a shift from design-based to model-based inference, whose risks and mitigants are discussed explicitly.
The Istat modernisation programme was focused on the centralisation of cross-cutting functions like data collection and methodology, the development of the Integrated System of Statistical Registers (ISSR) as the foundation of statistical production, and the exploitation of new data sources. The modernisation made it necessary to update the existing quality assurance system. Consequently, Istat defined a five-year quality strategy in 2020-2021, endorsed by top management. Its implementation is coordinated by the Quality Committee and supported by the Quality Manager. The strategy focuses on quality assessment and is differentiated by statistical process type, having regard to the varying availability of quality assurance tools. A checklist was applied to traditional processes, leading either to an internal conformity label or to improvement actions, and was complemented by an internal audit programme covering three processes per year. For the ISSR, an ad hoc quality framework, based on GSBPM and GSIM metadata standards and specific quality indicators, was developed for documentation and monitoring purposes and is currently being implemented. For statistics based on new data sources, efforts focus on identifying the main quality issues and the assessment methodology. The paper highlights achievements and lessons learned and outlines directions for the next quality strategy.
National Statistical Offices are increasingly expected to expand their role beyond traditional statistical production to act as central stewards of national data ecosystems. This expansion creates a fundamental institutional tension: organisations established to guarantee professional independence and confidentiality are now being asked to promote data sharing, reuse, and policy analytics. This article examines Lithuania as a critical case of this transformation. In 2023, Statistics Lithuania was legally reconstituted as the State Data Agency (SDA), receiving a mandate that integrates official statistics with nationwide data governance. Drawing on legal analysis, institutional documentation, and operational evidence, the study explores how Lithuania has centralised administrative data flows, established a national data lake, and developed secure analytical environments to support ministries and municipalities. Concrete applications in health surveillance, social policy, and municipal planning demonstrate the benefits of this model, including reduced respondent burden, more timely indicators, and enhanced evidence-based decision-making. At the same time, the Lithuanian experience reveals significant governance risks related to confidentiality, politicisation, and technological dependence. This paper argues that NSO-led data agencies can strengthen public sector capacity only if robust legal safeguards, transparent access procedures, and ethical oversight evolve alongside technological systems. Lithuania thus provides both inspiration and caution for countries seeking to reposition their statistical authorities at the centre of the digital state.
