We test the predictions of the Roy-Model about the self-selection of immigrants using an administrative dataset including about 90 % of Italians living abroad. The data comprises 13 countries with substantial differences in inequality and levels of redistribution: Argentina, Australia, Belgium, Brazil, Canada, France, Great-Britain, Germany, The Netherlands, New-Zealand, Switzerland, the US, and Venezuela. Our results confirm the predictions of the model: We find a negative, substantial and significant relationship between the level of redistribution – our indicator for the returns to human capital, measured by the (relative) difference of market and after-tax inequality in the host country in the year of arrival – and immigrants’ individual degree of selection, as well as the likelihood to be positively self-selected. These results hold after including covariates at the individual and country level, as well as controlling for migration costs. Our analysis also shed light on the factors associated with the self-selection of immigrants.
(joint work with Giacomo Corneo)