Subscribe to our "Population in Perspective" Substack Here!
A new study shows how digital trace data can be combined with traditional surveys to produce more detailed estimates of migrant populations by age and sex. The research, published in the Journal of the Royal Statistical Society Series A, was co-authored by former Leverhulme Centre for Demographic Science researcher Francesco Rampazzo and LCDS researcher Jakub Bijak, alongside Agnese Vitali, Ingmar Weber and Emilio Zagheni.
Accurate information on the age and sex of migrant populations is essential for population projections and for planning public services. However, conventional surveys can struggle to capture some migrant groups, particularly younger people and those living in temporary or non-standard accommodation.
The researchers developed a hierarchical Bayesian model that combines data from the UK Labour Force Survey with estimates from Facebook’s Advertising Platform. They applied the approach to the ten largest European migrant groups living in the UK in 2018 and 2019.
The two sources offer different strengths. The Labour Force Survey provides a broad picture of the population but can be affected by small sample sizes and may underrepresent migrants with unstable housing or living arrangements. Facebook data can provide more timely and detailed information, particularly about younger adults, but are shaped by patterns of platform use and changes in how the company classifies and reports its users.
By bringing these sources together, the model can compensate for some of their individual limitations.
The results revealed three broad migrant age profiles. Western and Southern European migrants were generally younger, with the largest populations concentrated between ages 20 and 29. Migrants from Central and Eastern Europe had slightly older working-age profiles, concentrated primarily between ages 25 and 34. Irish migrants had a markedly older age structure, reflecting the long history of migration between Ireland and the UK.
The researchers compared their 2019 estimates with age and sex profiles from the 2021 Census for England and Wales. Their model produced estimates that were substantially closer to the Census benchmark than an alternative model used for comparison, although differences in geography and timing mean that the Census cannot provide a direct like-for-like test.
“No single dataset gives us a perfect picture of migration. Surveys and digital platforms each miss different parts of the population, but by making those limitations explicit and combining the sources carefully, we can produce estimates that are both more detailed and more credible,” said Francesco Rampazzo.
Although the study uses Facebook data, the framework is not tied to one platform. It could also incorporate age-structured information from other digital platforms, mobile-phone data or administrative records, provided that researchers account for differences in coverage and data quality.
The authors stress that digital data should not replace official statistics. Instead, it can complement surveys and administrative records, helping researchers produce more timely estimates while recognising that patterns of platform use and data availability can change.
The findings demonstrate how integrating traditional and digital sources could improve migration statistics and support more effective population projections, policy decisions and service planning.
Read the full paper: Migrant age profiles reconciling digital trace and survey data: an example of the United Kingdom in 2018 and 2019, published in the Journal of the Royal Statistical Society Series A: Statistics in Society.
