Laboratory

Migration and Mobility

At a Glance Projects Publications Team

Project

Combining Digital Trace Data and Representative Surveys to Estimate and Predict Migration Stocks and Flows

Maciej Danko, Emilio Zagheni, Jisu Kim; in Collaboration with Francesco Rampazzo (University of Oxford, United Kingdom), Ingmar Weber (Saarland University, Saarbrücken, Germany), Lee Fiorio, Jonathan Wakefield (both: University of Washington, Seattle, USA), Jakub Bijak (University of Southampton, United Kingdom), Agnese Vitali (University of Trento, Italy), Yuan Hsiao (University of Washington, Seattle, USA), Monica Alexander (University of Toronto, Canada)

Detailed Description

Reliable and timely estimates of migration flows and stocks are needed to guide policy decisions and to improve our understanding of migration processes. However, obtaining timely and fine-grained estimates remains an elusive goal.

Digital trace data can be used to improve our understanding of migration processes. However, samples from these data are typically not representative of the general population and are often biased. Thus, the crucial question is how to combine the representativeness of traditional sources with the fine-grained geographic and temporal information of digital trace data. We propose Bayesian statistical frameworks that involve combining digital and survey data for migration estimation by accounting for the bias structure for each data source. If the bias has a structure over time and space that can be statistically modeled, we can combine different sources of data for prediction. Different versions of the same principle are adapted to different contexts.

In a first step, we assume that the estimates from representative surveys are unbiased, whereas the estimates from digital data are biased. We then jointly model the two types of estimates using a Bayesian hierarchical model that accounts for spatial and temporal effects. The method is used to nowcast internal migration flows in the United States by combining information from the American Community Survey (ACS) and geolocated Twitter data. With a different but related model, ACS data as well as Facebook data for advertisers are used to generate nowcasts of stocks of international migrants in the United States.

In a second step, we account for the possibility that traditional data sources are also biased. We use a Bayesian hierarchical model that builds on the so-called “Integrated Model of European Migration” in order to combine data from the Labor Force Survey and Facebook data for advertisers in the European context. The aim is to produce an estimate of the number of migrants after assessing the limitations of each data source. In addition, we propose a model to disaggregate the estimates by age and sex. The method can be used to nowcast migration stocks and augment traditional data sources with digital traces, especially when data from surveys or registers lack sufficient quality or the needed granularity.

Third, this project directly evaluates the reliability of migration indicators based on Meta platform audience estimates (e.g., Facebook and Instagram), which classify users by their previous and current country of residence. We ran a targeted online survey (January–March 2025) of Meta users identified as formerly living in Poland and currently residing in the United Kingdom or Germany. We then compared their self-reported migration histories with Meta’s aggregated user estimates. The study also assesses how well these classifications match the UN definition of long-term migrants, and identifies potential biases such as differences in user activity, VPN-related location mismatches, multiple accounts, and ad delivery expansion. The results provide empirical estimates of bias and uncertainty to support more reliable use of digital trace data in migration research and statistical modeling of migration.

The infographic describes the structure of a framework to estimate migrant stocks using digital traces and survey data

Based on the paper by Francesco Rampazzo, Jakub Bijak, Agnese Vitali, Ingmar Weber, Emilio Zagheni: A Framework for Estimating Migrant Stocks Using Digital Traces and Survey Data: An Application in the United Kingdom. Demography 1 December 2021. © MPIDR

Publications

Dańko, M. J.; Rampazzo, F.; Donzowa, J.; Kim, J.; Zagheni, E.:

Assessing bias in migration stock estimates from Facebook and Instagram using a social media-based survey

MPIDR Technical Report TR-2026-001. (2026)    

Rampazzo, F.; Bijak, J.; Vitali, A.; Weber, I. G.; Zagheni, E.:

Assessing timely migration trends through digital traces: a case study of the UK before Brexit

International Migration Review 59:1, 119–140. (2025)    

Yildiz, D.; Wiśniowski, A.; Abel, G. J.; Weber, I. G.; Zagheni, E.; Gendronneau, C.; Hoorens, S.:

Integrating traditional and social media data to predict bilateral migrant stocks in the European Union

International Migration Review 59:1, 90–118. (2025)    

Hsiao, Y.; Fiorio, L.; Wakefield, J.; Zagheni, E.:

Modeling the bias of digital data: an approach to combining digital with official statistics to estimate and predict migration trends

Sociological Methods and Research 53:4, 1905–1943. (2024)    

Alexander, M. J.; Polimis, K.; Zagheni, E.:

Combining social media and survey data to nowcast migrant stocks in the United States

Population Research and Policy Review 41:1, 1–28. (2022)    

Rampazzo, F.; Bijak, J.; Vitali, A.; Weber, I. G.; Zagheni, E.:

A framework for estimating migrant stocks using digital traces and survey data: an application in the United Kingdom

Demography 85:6, 2193–2218. (2021)    

Alexander, M. J.; Polimis, K.; Zagheni, E.:

Combining social media and survey data to nowcast migrant stocks in the United States

arXiv e-prints 2003.02895. unpublished. (2020)    

Hsiao, Y.; Fiorio, L.; Wakefield, J.; Zagheni, E.:

Modeling the bias of digital data: an approach to combining digital and survey data to estimate and predict migration trends

MPIDR Working Paper WP-2020-019. (2020)