The Mexican 3x1 Program for Migrants is a matching fund scheme that seeks to direct the money sent by hometown associations abroad (collective remittances) to productive uses. The federal, state and municipal governments contribute to the program multiplying by three the contributions sent by migrants abroad. Using municipal level data on program participation for the period 2002-2006, we evaluate the program’s capacity to target the poorest municipalities. Since migration has a nonlinear relationship with poverty and marginality, a program that unconditionally responds to project initiatives from migrant organizations is bound to be regressive due to self-selection bias. Indeed, poorer municipalities are less likely to participate and they also receive lower amounts and fewer projects than relatively richer localities. Moreover, we find evidence of a partisan bias: states and municipalities ruled by the PAN are significantly more likely to participate in the program, and electoral support for the PAN is associated with more funds or projects awarded. We argue that substantial changes should be implemented for the program to be a truly progressive poverty reduction tool.