An Analytical Approach to Assess Spatial Interplays among Migration Drivers: The Case Study of Ghana

Daniela Ghio, University of Catania
Sarah Hoyos-Hoyos, Toronto Metropolitan University
Robert McLeman, Wilfrid Laurier University
Gavin Liu, Toronto Metropolitan University
Emmanuel Kyeremeh, Toronto Metropolitan University
Gabby Resch, Ontario Tech University
Ali Mazalek , Toronto Metropolitan University

Scholarly research has investigated how slow-onset climate factors are associated with population vulnerability for the creation or exacerbation of economic and violent conflicts. Yet, the lack of data makes difficult the measurement of interplays among contextual factors and population dynamics. Taking advantage of newly available net migration datasets, which provide estimates of migration changes at fine spatial resolution, we define a three-step methodological strategy. First, we use the Geographically-Weighted-Regression to assess spatial variations of the estimated local parameter effects by geographical unit. Second, we adopt the Multiscale-Geographically-Weighted-Regression, recently applied in different fields of social science research, to process flexibility in spatial heterogeneity at different geographic scales. Third, we adopt the variation of forest models, the Geographically-Weighted-Random-Forest-Regression, which disaggregates the traditional forest model into multiple local models to highlighting local variations. Our empirical analysis focuses on Ghana to explore how precipitation patterns, extreme temperatures, conflicts, and artisanal (small-scale-company) gold mining have influenced migration patterns over 30 years, from 1985 to 2014. By offering a comparative perspective to interpret results, our study contributes to assess at what extent geospatial regression techniques and machine learning models complement each other to depict spatial-temporal variations of the relationship among environmental conditions, conflicts and migration patterns.

Keywords: Population, Environment, and Climate Change, Data and Methods, Comparative methods

See extended abstract.