Modelling internal migration in South Africa

Abstract

The aim of this study was to model internal migration in South Africa using the 2011 Census data. The net-internal migration was modelled in the district municipalities of South Africa using Ordinary Least Squares (OLS) and Geographically Weighted Regression (GWR). In this study, the following global and local modelling techniques were used, Gravity, Poisson, Negative Binomial (NB), Gamma, and GWR model (local model). Poisson and NB failed to fit the migration data, while the Gamma model managed to fit the data reasonably well. The GWR model performed better than OLS regression in modelling net-internal migration in district municipalities of South Africa. The results from these models revealed that there was a strong relationship between internal migration and economic variables, as well as living conditions and demographic variables. The Monte Carlo significance test results showed that the parameters of the white population vary significantly across space. The results of the study signal that the differences in social and economic disparities in the district municipalities of South Africa are the drivers of internal migration.