How representative was our migration data?

Ensuring sample representativeness is critical for the validity of any project relying on data-driven insights. This report aims to evaluate the representativity of the data collected for DYNAMIG in Work Package 3 using social media to target individuals.

We examine how well our potential sample reflects the target population. Representativity is essential to minimize bias and ensure the validity of conclusions and recommendations derived from the data. Thus, we compare the sample characteristics of people having internet access to those having no access using existing survey data from Gallup World Poll and Afrobarometer.

This report involves four key steps. First, we briefly lay out the methods used to obtain our own survey data, including which part of the population was targeted. Second, we obtain and analyze official statistics of our target countries Kenya, Morocco, Nigeria, Tunisia, and Senegal. Third, we assess whether relying solely on individuals with internet access (and therefore social media users) introduces significant selection bias, based on existing survey data. Fourth, we discuss the implications, such as selection effects, of using Meta as a recruitment platform based on 12,300 observations from our most recent survey in Kenya and Nigeria. Using official statistics, we document macro-level differences between the five countries focused on in the DYNAMIG project.

Our descriptive analysis of existing survey data highlights selection patterns and therefore potential bias when relying on internet-based sampling such as via social media. Compared to the general population, individuals with internet access are younger, more educated, more urban, more likely to be male, and more optimistic about economic prospects. They 7 also report higher migration aspirations, which is especially important in the context of the DYNAMIG project, which aims to strengthen the understanding of how people make decisions on whether and how to migrate.

We also document selection patterns in the preliminary survey data from Kenya and Nigeria, which we are gathering via the social media platform Meta. Since selection produces samples that overrepresent individuals with high migration aspirations, they provide a valid basis for migration research and data-driven policy advice. However, extrapolation to the general population should be undertaken with caution.