January 13, 2021 | Press Release

New Framework Enables Migration Estimation over Varying Time Scales

With digital trace data it is possible to study migration with a method similar to a survival function. © Addictive Stock / photocase.com

Estimating migration over varying time scales is now possible with a new framework that Emilio Zagheni and colleagues introduce in their recent paper in “Demography.”

The interest in using big data to analyze migration and mobility has never been higher. This has been demonstrated recently by the massive amount of research on COVID-19 which seeks to measure the rate at which people moved around geographically before and after a COVID-19-related lockdown.

The paper by Lee Fiorio, Emilio Zagheni, and an international team, now published in Demography, the popular peer-review academic journal, points to the next step in this kind of research. Emilio Zagheni, the Director of the Max Planck Institute for Demographic Research (MPIDR) in Rostock, Germany, and Lee Fiorio, former guest researcher at MPIDR, not only focused on how often people moved, but also on how often they migrated. “We analyzed how many people changed the city or administrative unit they were living in, over a whole sequence of time intervals,” says Lee Fiorio. Through this kind of analysis, the authors demonstrated the sensitivity of standard migration measures to different temporal definitions.

The researchers used digital trace data, or big data, that contained information about where people were at a specific point in time. This data came in a variety of different formats, like call detail records. This consisted of a record of calls or texts and the time and location from where they were sent.

Since the demographers wanted to analyze migration at the population level, they were not interested in identifying single individuals by name. Instead, they used anonymous data sets consisting of Tweets from 2011 to 2014, a sample of call detail records from a mobile phone provider in Senegal from 2013, and the complete set of geo-tagged check-ins from Gowalla, a defunct location-based social network in the U.S.

How to study migration with a method similar to a survival function

“We have provided a framework to assess the relationships between short-term mobility and long-term relocations. As the concept of a fixed place of residence has become less clearly defined with the rapid expansion of the gig economy and teleworking, we have developed a way of rethinking how to measure migration in the digital and post-COVID age,” says Emilio Zagheni.

The researchers showed that by taking advantage of the temporal specificity in big data, it is possible to study migration with a method similar to a survival function. Following a population from a set reference point, they were able to track the proportion of individuals that changed location with an increase in time intervals. For practical purposes, the method is also useful for harmonization of migration statistics when measures of migration are produced using different definitions by statistical offices in different countries.

A problem with big data is that it is proprietary, and access is limited by the big tech and telecommunications companies that own them. “I would love it if our paper would encourage demographers and other researchers to think big about what can be done with highly granular digital trace data,” says Lee Fiorio. “These kinds of data are becoming increasingly common, and demographers should make sure they are at the forefront of discussions of how to use this type of data for population research in ways that protect privacy.”

Original publication

Fiorio, L., Zagheni, E., Abel, G., Hill, J., Pestre, G., Letouzé, E., Cai, J.: Analyzing the Effect of Time in Migration Measurement Using Geo-referenced Digital Trace Data. Demography. (2021) DOI: 10.1215/00703370-8917630

Authors and Affiliations

Lee Fiorio, University of Washington

Emilio Zagheni, Max Planck Institute for Demographic Research, Rostock

Guy Abel, Asian Demographic Research Institute, Shanghai University 
Wittgenstein Centre (IIASA, VID/ÖAW, WU), International Institute for Applied Systems Analysis

Jonathan Hill, University of Washington

Gabriel Pestre, Data-Pop Alliance

Emmanuel Letouzé, Data-Pop Alliance, Massachusetts Institute of Technology Media Lab 

Jixuan Cai, Department of Geography and Resource Management, The Chinese University of Hong Kong, Wittgenstein Centre (IIASA, VID/ÖAW, WU), International Institute for Applied Systems Analysis