Endogeneity and sample selection in a model for remittances

We estimate a remittance model in which we address endogeneity and reverse causality relationships between immigrants' remittances, pre-transfer income and consumption. In order to take into account the fact that a large share of individuals do not remit, instrumental variable variants of the double-hurdle and Heckit selection models are proposed and estimated by Limited Information ML; semiparametric extensions are considered as robustness checks. Our results for a sample of recent immigrants to Australia show that endogeneity is substantial and that estimates obtained by the methods previously employed in the literature may be misleading if given a behavioral interpretation.