Estimating Irregular Migrant Stocks Using Social Media Data and Machine Learning

Irregular migration is a sensitive policy issue around the world, often exacerbated by the lack of data. In data scarce contexts, triangulation relying on multiple sources and data points is the best strategy to arrive at evidence-based decisions. We want to contribute to this process by introducing a novel method for estimating the population size of migrants lacking legal residence, combining official statistics on international migrant stock, global social media data (Facebook), and predictive analytics through machine learning algorithms. We attain estimates of the size of the irregular migrant population for the 40 largest destination countries and, for some cases, the composition of the irregular migrant population by world regions of origin and destination. The estimates we obtain are consistent with alternative benchmarks in countries with available estimates. Inconsistencies can be explained by data limitations. We discuss potential limitations of the approach and broader implications for policy and migration research.