Comment on 'Giving According to GARP : An Experimental Test of the Consistency of Preferences for Altruism'

We present clean experimental evidence that a methodological confound was introduced by Andreoni and Miller (2002) that leads to diametrically opposed conclusions regarding comparisons of preferences between categories of fellow human beings distinguished by gender or age. Our study is a warning not to run the Andreoni-Miller tests of the consistency of altruistic preferences on the basis of their ‘interactive’ experimental protocol, and in particular not to perform preference estimations with the purpose of said comparisons. It is more interpretable and controllable to run the traditional, standard dictator game