New insights into the selection process of Mexican migrants.What can we learn from discrepancies between intentions to migrate and actual moves to the U.S.?

Comparing intentions to migrate and actual migration of Mexicans, I intend to assess the impact of unexpected shocks and misevaluated costs on the materialization of migration plans. I show that both sets of reasons may explain discrepancies between intentions and subsequent actions without denying the rationality of intentions by resorting to the theoretical framework of the Roy model. I use intention and migration data from the Mexican Family Life Survey, together with precipitations monthly series, hurricane and crime data to represent different sets of shocks. Correlations between intentions and migration on the one hand, and between intentions and individual labor market characteristics show that intentions are not devoid of informational content. Then, modelling intentions and migration with a bivariate probit, I find that shocks, and in particular rainfall and hurricanes, affect the probability to migrate conditional on initial intentions. The key finding is nonetheless the much lower propensity for women to migrate abroad conditional on intentions, which suggests that women incur specic costs or constraints misestimated at the intention stage. Alternative explanations, such as gendered preferences are discussed, but convergent empirical evidence suggest that women are more constrained than men on the international migration market. Moreover the data suggest that migrants are positively selected with respect to their unobserved characteristics whereas those with intention to migrate abroad are negatively selected. The shift in selection between the two stages of the migration process may be due to the cost reducing effect of individuals' unobserved characteristics that explain their higher local wages. Mots-clés : Migration, Roy model, migrant self-selection, subjective data, shocks, Mexico