A Framework for Estimating Migrant Stocks Using Digital Traces and Survey Data: An Application in the United Kingdom

Abstract

Abstract An accurate estimation of international migration is hampered by a lack of timely and comprehensive data, and by the use of different definitions and measures of migration in different countries. In an effort to address this situation, we complement traditional data sources for the United Kingdom with social media data: our aim is to understand whether information from digital traces can help measure international migration. The Bayesian framework proposed is used to combine data from the Labour Force Survey (LFS) and the Facebook Advertising Platform to study the number of European migrants in the United Kingdom, with the aim of producing more accurate estimates of the numbers of European migrants. The overarching model is divided into a Theory-Based Model of migration and a Measurement Error Model. We review the quality of the LFS and Facebook data, paying particular attention to the biases of these sources. The results indicate visible yet uncertain differences between model estimates using the Bayesian framework and individual sources. Sensitivity analysis techniques are used to evaluate the quality of the model. The advantages and limitations of this approach, which can be applied in other contexts, are discussed. We cannot necessarily trust any individual source, but combining them through modeling offers valuable insights.