This article evaluates the use of Facebook’s advertising platform to estimate numbers of Venezuelan migrants, including their spatial distribution in Latin America and socioeconomic profiles.
Facebook’s advertising platform allows advertisers to target advertisements based on Facebook user attributes, including self-reported attributes (e.g., age, gender, education, and relationship status), inferred attributes (e.g., interests, political orientation, cultural affinities) and automatically extracted information (e.g., device and connection type). Facebook provides the advertiser with an estimate of the number of monthly active users (MAUs) matching specified targeting criteria, available free of charge through the Facebook Graph API.
The authors compare “the number of Facebook users aged 13 and above who used to live in Venezuela” with the most recent official estimates of Venezuelan refugees and migrants. Official estimates at national level come from the Regional Inter-Agency Coordination Platform for Refugees and Migrants from Venezuela (R4V). For Colombia, subnational figures for refugees and migrants are drawn from the Registro Administrativo de Migrantes Venezolanos (RAMV). While there are no official sub-national estimates of Venezuelan migrants in other Latin American host countries, the authors use Facebook advertising data to estimate numbers of Venezuelans across provinces (“provincias” in Peru and Ecuador) or states (“estados” in Brazil). They show how Facebook data can be used at higher spatial resolutions, to estimate numbers of Venezuelan migrants in various parts of a city (in 1 km radius circles). The authors demonstrate how Facebook data can also provide information on the socioeconomic profiles of Venezuelan migrants.
Results:
- Facebook data produces a spatial distribution of Venezuelan refugees and migrants across different countries that closely corresponds to official data from R4V, although estimates produced with Facebook’s data are much higher than official estimates. In January 2019, R4V reported a total of 2.7 million Venezuelan nationals living in 17 countries, whereas there were an estimated 3.2 million Facebook users who previously lived in Venezuela in these countries.
- At the subnational level in Colombia, the correlation of Facebook and RAMV data is weaker, although both data sources produce the same relative rankings of host departments. Facebook estimates more Venezuelan migrants than RAMV for several departments away from the Venezuela-Colombia border. The largest difference is in the capital district of Bogotá, where Facebook data indicates 490,000 Facebook users who previously lived in Venezuela, while RAMV reported only 43,000 Venezuelan refugees and migrants. This difference may be due to the small number of registration stations in Bogotá.
- Facebook data can provide sub-national estimates of Venezuelan migrants in countries where this data does not exist. In Brazil for example, Facebook data suggests that 75 percent of Facebook users who previously lived in Venezuela are in two states on the border with Venezuela (Roraima and Amazonas), whereas the more prosperous states of Sao Paulo and Rio de Janeiro are home to 12 percent and 5 percent, respectively.
- Facebook data can provide indicative socioeconomic information about the Venezuelan population in Latin American countries, for example their self-reported education level and their inferred average income per capita.
The authors conclude that estimates based on Facebook advertising data are a useful proxy for the number of Venezuelan refugees and migrants. They note that Facebook’s advertising data is potentially valuable due to its low latency (days not months), low acquisition cost, high spatial resolution, and ability to disaggregate by socioeconomic status, and remote sensing capabilities. There are however several limitations of Facebook data including the dependence on Facebook’s algorithm for identifying users’ previous countries of residence, lack historic data, use and potential removal of fake accounts and bots, and risk of excluding disadvantaged people without access to digital technology.