Migration: an eroding statistical foundation

The production of global migration statis­tics involves a division of labour between national and international institutions. The underlying demographic and social data comes mainly from censuses, population registers and representative large-scale household surveys, in particular specialised migration surveys. At the international level, different organisations perform dis­tinct roles. The population division of the United Nations Department of Economic and Social Affairs (UN DESA) uses national sources to prepare internationally comparable estimates of migrant stocks. The Inter­national Labour Organization (ILO) com­bines these and other sources to estimate the size and composition of the migrant labour force worldwide. The World Bank compiles remittance data and analyses the links between migration and development.

This division of labour leaves the system vulnerable to cuts at several points. A UN survey of national statistical offices in low- and middle-income countries illustrates the impact: 69 percent had experienced fund­ing reductions between January and Decem­ber 2025, affecting an average of four planned or ongoing surveys in each case. The effects extend beyond national statis­tical systems to the international compilation and harmonisation of their data. When national surveys are cut back, international organisations lose up-to-date source data. When their own capacity is reduced, they are also less able to fill gaps, validate national inputs or ensure comparability across coun­tries. Cuts can moreover eliminate coordi­nating structures and specialised surveys.

In preparing the 2024 edition of its International Migrant Stock dataset, the UN DESA population division fully reassessed the estimates for 60 countries and areas on the basis of new empirical data, and extra­polated the 2020 estimates for the remainder. Missing national updates are now exacerbated by processing bottlenecks at the international level. In September 2025, almost a quarter of the positions in the population division were vacant. A sub­sequent internal UN audit found that these vacancies delayed and restricted the divi­sion’s outputs. For the next edition of Inter­national Migrant Stock, it was decided to postpone the updating of age-disaggregated estimates.

These bottlenecks also affect the ILO labour migration statistics, which depend on age-disaggregated UN DESA data. The ILO’s global estimates draw on national sources that must be compiled, checked for comparability and harmonised. Budget cuts have reduced the specialist staff available for this work, making it harder to validate national data and convert it into internationally comparable datasets. At the same time, international migration data from the 2020 census round remains incomplete or insufficiently processed, disseminated and integrated in updated national sampling frames in a number of countries. The win­dow for closing these gaps is narrowing as preparations for the 2030 census round are already under way. If survey cycles are interrupted, reporting and coordination processes with national authorities, that have been built over many years, may be disrupted. Reestablishing them later is both difficult and costly.

Gaps are also emerging in the migration and development data compiled by the World Bank. Although global remittance data continues to be published, important quality assurance, methodological development and international coordination functions were lost when the World Bank-hosted Global Knowledge Partnership on Migration and Development (KNOMAD) ended in 2024. RemitStat, the international working group established to improve remittance statistics, also lost its funding.

The data gap is especially pronounced for migrant workers’ recruitment costs. With KNOMAD’s support, the ILO and the World Bank developed standards and tools for Sustainable Development Goal indicator 10.7.1. The indicator measures the recruit­ment costs borne by workers as a share of their income in the destination country, showing how high recruitment costs impede fair labour migration and limit migration’s contribution to development. The methodology was designed in collaboration with national statistical offices and piloted in several countries. Since KNOMAD ended, however, there has been no comparable mechanism for compiling this data inter­nationally. As a result, global availability of SDG 10.7.1 figures remains extremely limited. The agreed indicator and methodology survive, but the data they require is no longer collected systematically.

Across these cases, essential data series have not disappeared; instead, they are gradually being thinned out. Up-to-date, disaggregated source data is becoming scarcer, quality-assurance capacity is declin­ing, methodological improvements are being postponed and specialised data collec­tions discontinued. The information most at risk is precisely what aggregate figures on migrant stocks do not capture: who migrates, under what conditions, and how the costs of migration and its economic benefits are distributed.

Displacement: funding cuts and a fragmented system

A large proportion of data on forced dis­placement is collected through humanitarian programmes and operations. The United Nations High Commissioner for Refugees (UNHCR) registers refugees and asylum-seekers, while the International Organization for Migration (IOM) records displacement and mobility in its Displacement Tracking Matrix (DTM), usually relying on local data collection teams. These sources are complemented by needs assessments from other humanitarian actors and by official statistics and administrative data. Cuts to projects, field missions or staff therefore translate directly into a reduction in the scope, frequency and depth of data collection.

Funding cuts in recent years now directly affect this data infrastructure. In early 2026, UNHCR and IOM reported reductions of around 40 percent in staff working on data. In some settings, local data collection networks built over many years have been dismantled. Funding shortfalls have forced IOM to scale back DTM data collection sharply: geographical and/or temporal cover­age has been reduced in several countries, and some activities have ceased altogether. In Somalia, for example, Emergency Trend Tracking coverage fell from 25 districts to five in 2025. Individual registration of refu­gees and asylum-seekers, to record identity and basic personal data, has also been severely affected. By the end of 2025, fund­ing and staffing shortages meant that around 934,000 people in West and Central Africa had not been individually registered by UNHCR. Although they may still be included in the aggregate population figures, without individual registration the personal data needed for documentation, protection and needs-based assistance is lacking. More resource-intensive household surveys, protection analyses and repeat surveys of income, education or health are also being reduced. The upshot is a widening gap between what is known about the scale of displacement and what is known about living conditions and protection risks.

The cuts are hitting a data system that was already fragmented. Humanitarian data collection is dispersed across organisations, mandates and project budgets. Even before the current cuts, duplication and institutional rivalry had been flagged as structural weaknesses of the humanitarian data ecosystem. Yet parallel data collection in the same area is not inherently inefficient: a community-level DTM assessment and a representative household survey on income or protection risks answer different questions. Inefficiencies arise when identi­cal or similar information is collected more than once or when projects produce “single-use” datasets that are methodologically or technically difficult to combine with other data and serve no purpose after the project ends.

Institutional misincentives add to the problem. Project funding can encourage organisations to expand their own data products and present them to donors as a distinct service. Research on IOM has described this process as “data entrepreneurship”: data collection evolving from a by-product of operational work into a service in its own right for governments, donors and humanitarian organisations.

Funding cuts thus intensify three existing problems. First, project and organisational incentives can generate new surveys without a systematic assessment of the information that is actually required. Second, in protracted displacement situ­ations project-financed international surveys often continue without being ade­quately integrated into national statistical and monitoring systems. Third, the num­ber of actors involved makes it harder to estab­lish a clear division of labour and reuse existing data. Addressing these prob­lems requires context-specific assessments of data needs, national ownership and ins­titutional coordination.

Consolidation amid funding constraints

Funding cuts have lent additional urgency to wider reform efforts in the UN system and the humanitarian sector. These include initiatives to consolidate data systems at three levels: technical integration through shared infrastructure and standards, institutional coordination across organisations and borders, and organisational consolidation through mergers.

The first level concerns technical integration. UN System Data Commons, launched in September 2026, brings together publicly available data from many UN entities, making it easier to find, link and use. In parallel, the Humanitarian Data Collaborative, pursued under the UN80 initiative, aims to strengthen common platforms as well as shared standards, analytical capacity and funding arrangements. Both initiatives have the potential to reduce fragmentation and improve the use of existing data. A better-integrated data infrastructure would particularly benefit areas of analysis that combine multiple sources, such as human mobility in the context of climate change and natural disasters. Yet infrastructure cannot replace data production. If a survey is cancelled, a time series terminated or a field presence lost, better shared systems cannot recreate the missing information. New digital data sources and AI-supported methods have raised considerable hopes of filling such gaps cheaply and at scale, but these expectation should not be overstated. Digital sources can provide additional information, and AI-supported methods can make it easier to process, link and analyse large datasets, but neither can replace missing primary data. The reliability of AI-supported analysis depends on the avail­ability, quality and comparability of the underlying data.

A second level concerns more effective institutional and cross-border coordination. The route-based approach developed by UNHCR and IOM seeks to connect interventions more closely across countries of ori­gin, transit, asylum and destination, in the process closing information gaps between countries and organisations. This requires the alignment of different mandates, methods and data-protection rules. With resources scarce, however, data collection may concentrate on politically prominent routes and easily measured movements.

The third level is organisational consolidation. Outside the United Nations, the merger of JIPS and IDMC, which was com­pleted in August 2026, brings together global analysis of internal displacement with profiling, methodological development and support for national data sys­tems. The key question is not which struc­tures are merged or cut, but whether spe­cific functions continue to be performed reliably and coordinated effectively.

What Germany and the EU can do

With funding constraints likely to be struc­tural rather than temporary, preserving all existing data activities is not a viable option; nor should consolidation be assumed to generate benefits automatically. Germany and the EU should safeguard core statistical functions, strengthen national and regional structures, and ensure that reforms do not erode the essential functions. Funding deci­sions should start by asking what informa­tion is needed, who is best placed to produce it and how the results can remain usable over time.