Abstract The conditional equality and egalitarian equivalence criteria were proposed by Fleurbaey (Fairness, responsibility, and welfare, Oxford University Press, Oxford, 2008) to provide better foundations to interpersonal comparisons in the context of heterogeneous preferences and multidimensional welfare. The first implementations of the egalitarian equivalence criterion follow an approach where the preferences are captured at the group level (based on socio-demographic variables) rather than at the individual level. Our contribution is to extend these models by using information on individual preferences, derived from the potential discrepancy between the group level optimal choice and the revealed choice of the individuals. We implement and compare the conditional equality and egalitarian equivalence criteria on a 2004 US microeconomic dataset and find that these criteria are relatively consistent in the identification of the worst-off. We also show that up to 18 % of the worst-off are no longer categorized as worst-off when the empirical approach accounts for individual preferences.