Aggregate-level tests of models of other-regarding preferences (see, for example, Fehr and Schmidt, 2006) essentially compare the distribution of choices across different experiments that were run with different samples and check for consistency with the model.