Earnings distribution of Cuban immigrants in the USA: evidence from quantile regression with sample selection

This article analyses the conditional earnings distribution for Cuban immigrants in the USA considering Buchinsky sample selection in a quantile regression model. The test proposed by Huber and Melly to test the independence between error terms and regressors (conditional on the selection probability) is also considered. This is the first attempt in the migration literature to use quantile regression with sample selection. The data used come from the US American Community Survey. The results show that the hypothesis of conditional independence is not rejected, and increments in earnings associated with the usual socioeconomic characteristics in labour studies vary between the cohorts considered. The main conclusions are that a decline in returns from education may be a sign that a high level of education no longer provides a competitive advantage and that being a black person is associated with substantially lower earnings regardless of the individuals' position in the earnings distribution. This may explain why, historically, comparatively fewer black Cubans have made the decision to emigrate to the USA because of a lack of economic incentives.