Abstract

We show that the introduction of a non-contributory pension program (Pension 65) in Peru had remarkably dierent eects for its three main ethnic groups, i.e., Mestizo, Quechua, and Aymara. The Aymara beneficiaries of the program have experienced larger increases in health, respect, and life satisfaction compared to other Peruvians. We argue that preference heterogeneity may explain these difierences. Using a panel life satisfaction regression, we find indeed evidence for some preference heterogeneity between the Aymara and the other ethnic groups that is consistent with the observed dierences. Finally, we turn to the question how the pension program can be evaluated in a robust manner while respecting the preference heterogeneity between the ethnic groups. We propose to use the natural criterion that a program benets a recipient if she is lifted to a higher indifierence curve. We show that the pension program was beneficial for all groups, but that more Aymara beneficiaries were lifted to a higher indifierence curve compared to other Peruvians. Our proposed robust criterion can be useful to evaluate programs in all cases where preference differences matter.