August 20, 2020 | Press Release
Migration: New Method Successfully Combines Social Media Data with Traditional Sources
Migration to the United States: Data from Facebook improves timely estimates. © iStockphoto.com/belterz
In their recently published paper Emilio Zagheni and colleagues present a statistical framework to combine social media data with traditional survey data. They produce timely ‘nowcasts’ of migrant stocks by state in the United States.
“The goal of the paper was to develop methods to improve short-term migration forecasts using social media data“, says Monica Alexander, member of the research team and guest researcher at the Max Planck Institute for Demographic Research (MPIDR) in Rostock, Germany.
In particular, the researchers introduce a new statistical method for combining demographic data from social media sources with data from traditional demographic sources, such as a representative survey or census. Their paper was recently published in Population Research and Policy Review.
Download figure (PNG File, 474 kB)
Data from social media has notable strengths, in that it is based on large sample sizes and essentially available in real time. But there is also a drawback since this data is in general not representative of the broader population. The research team took this into account in its statistical framework.
The researchers used data from Facebook’s Advertising Platform to build up a large database on timely information on migrants in the US. They collected these data themselves over a period of two years. In addition, they used data from the American Community Survey, a national-representative survey in the US.
“We show that the proposed model improves prediction of short-term trends when compared to viable alternatives“, say Emilio Zagheni, Director at MPIDR. Monica Alexander is convinced that their model has even more advantages: “We are particularly excited because we believe the general statistical framework we developed could be extended to include information from other sources, or to measure other population processes.“
The researchers provide all their data and code to help other researchers explore and adapt the method for their own purposes on Github.
Alexander, M., Polimis, K., Zagheni, E.: Combining Social Media and Survey Data to Nowcast Migrant Stocks in the United States. Popul Res Policy Rev (2020). DOI: 0.1007/s11113-020-09599-3