Which definition of migration better fits Facebook ‘expats’? A response using Mexican census data

Background: Data from social media have emerged as an auxiliary source for real-time information on migrant populations. Facebook users’ tagged ‘expat’ data – an ‘expat’ being someone who lived in country x but now lives in country y – has been used to estimate immigrants and its quality assessment has relied on household surveys and UNDESA migration estimates. Objective: Using the census as the gold standard and six definitions of migration, we examine differences between the 2020 Mexican Census and Facebook data by national origin, age, and sex. We also examine internet penetration among migrants. Methods: We estimate migration stocks by sex, age, and country of origin for nine Latin American countries, using six definitions of migration available within the census. To evaluate biases of Facebook data, we estimate a series of linear regression models on migrant stocks where our key independent variable is ‘expat,’ and we control for age, sex, and origin, as well as internet penetration rate. Results: Findings suggest that Facebook data are only associated with the definition that identifies recent immigrants according to country of residence five years prior to the census. Facebook’s ‘expat’ variable is indeed capturing recent immigrants that resided in a given country, and not migrants by other definitions using country of birth and country of prior residence, as in long-term migrants or returnees. Contribution: The contributions of this paper are fourfold. In our analysis we (1) evaluate Facebook data quality using the census as the gold standard, (2) compare Facebook data to different migrant definitions, (3) include data on migrant populations’ internet access, and (4) include differences by sex and age.