This project explores the use of mobile phone data to supplement national migration statistics from population censuses. Usually, population censuses are conducted quite infrequently – mobile phone data, on the other hand, are collected with every action (calls, messages, data exchanges) of a user within his or her mobile network. Therefore, it could be a useful data source to complement census-based national migration statistics, especially since mobile phone penetration is increasingly high in most world regions, including low- and middle-income countries. To investigate this potential, the research team analysed a dataset of 72 billion anonymized call detail records (CDRs) in Namibia from October 2010 to April 2014. First, they explored how to derive and model migration estimates from CDRs at subnational and annual scales; and second, how the precision and accuracy compares to census-derived migration statistics. Furthermore, the team investigated whether this novel data source could reveal year-by-year migration patterns that so far have remained unmeasurable when using traditional census-based migration statistics.
The results of this study highlight that migration estimates derived from mobile phone data constitutes a promising avenue for complementing traditional national migration statistics, especially since they provide for a higher frequency and spatial resolution. Not only does the research team show that estimates of migration produced through CDRs are as accurate as the census data-derived estimates; these data also offer additional benefits in terms of up-to-date statistics and an improved understanding changing of patterns of annual internal migration due to its high temporal resolution of longitudinal data. In the case of Namibia, for instance, the CDR data precisely reflected seasonal temporary movement, such as holiday-related travel, in addition to the place of residence. This project can potentially be replicated in any regional context that is covered by the services of a mobile network operator. In particular, this might be a relevant option for regions with a low and irregular frequency of conducting population censuses.
(Picture: © UNHCR, Frederic Noy)