Replication Data for: Promises and Limits of Using Targeted Social Media Advertising to Sample Global Migrant Populations: Nigerians at Home and Abroad

Survey research on migrants is notoriously challenging, especially if the goal is to collect data across a range of countries. Social-networking sites’ ability to micro-target advertisements to migrant communities combined with their near-global reach makes them an attractive option. Yet there is little rigorous evaluation of the quality of data thus collected – especially for populations from developing countries. We compare samples of Nigerian emigrants in Canada and Italy and Nigerians (at home) in Nigeria recruited through targeted advertising on Facebook and Instagram to population estimates. We find our samples contain varying degrees of bias in the case age and gender, and systematically miss those with little formal education. How much this affects our samples’ representativeness varies across contexts: discrepancies are much smaller for emigrant populations in Canada than in Italy and much larger in Nigeria, where a large share of the population has little formal education and limited literacy. Post-stratifying each sample on age, gender, and education does not ameliorate bias on other variables such as ethnicity, religion, period of migration, or political attitudes. We discuss the potential and limitations of social-media driven sampling and highlight key considerations for implementing it to collect multi-sited data on migrants. For Peer Review

针对移民群体的调查研究向来以难度极高著称，若目标是在多国范围内开展数据采集则更是如此。社交平台具备针对移民社群精准定向投放广告的能力，加之其近乎全域的触达范围，使其成为颇具吸引力的数据采集方案。然而目前针对此类采集所得数据的质量开展的严谨评估少之又少，针对发展中国家群体的相关评估尤为匮乏。本研究将通过Facebook及Instagram定向广告招募的加拿大、意大利境内尼日利亚移民样本，以及尼日利亚国内的尼日利亚民众样本，与官方人口统计估计值展开对比分析。研究发现，我们的样本在年龄与性别维度均存在不同程度的偏差，且系统性地遗漏了受正规教育程度较低的群体。此类偏差对样本代表性的影响程度因场景而异：加拿大境内移民群体样本的偏差远小于意大利，而尼日利亚国内样本的偏差则更为显著——该国大量人口受正规教育程度较低且读写能力有限。即便对各样本按年龄、性别与教育程度开展事后分层校准，也无法缓解族群、宗教信仰、移民时长或政治态度等其他维度的偏差。本研究探讨了社交媒体驱动型抽样方法的应用潜力与局限，并针对利用该方法开展移民多地点数据采集的关键注意事项展开重点阐述。供同行评议