We assess the predictive power of a model of other-regarding preferences, inequality aversion, using a within-subjects design. We run four different experiments (ultimatum game, dictator game, sequential prisoner's dilemma and public-good game) with the same sample of subjects. From the data we estimate parameters of aversion to disadvantageous and advantageous inequality. We then use these estimates to test several hypotheses across games. Our data show that results from within-subject tests can differ markedly from aggregate-level analysis. The inequality-aversion model has predictive power at the aggregate level but performs less well at the individual level.