Revisiting forced migration: A machine learning perspective
This study is motivated by an interest in understanding which factors play an important role in forced migration abroad. I use machine learning methods that allow for more effective ways to estimate complex relationships, particularly with highly nonlinear data. There are three key findings of this study. First, I document and test for the importance of the changing nature of conflict in Africa. My estimates show that riots are the most important type of conflict for explaining asylum applications, more important than battles or violence against civilians. Second, this paper is the first macrolevel study to investigate the role of the internet penetration rate in explaining forced migration flows and the first to show the importance of this factor in explaining asylum applications. Third, I find that country fixed effects are of primary importance in explaining forced migration flows. This finding suggests that a considerable variation in forced migration flows still remains to be explained.