Many socially desirable actions are subject to risk and occur in situations where the others are not anonymous. Assessing whether lower subject-subject anonymity affects behavior when outcomes are risky is likely important but has not been studied in depth so far. Herein, we provide evidence on this issue. In a series of allocation tasks, all of them variations of the dictator game, we systematically vary the party who is exposed to risk and manipulaterecipient anonymity by reducing the social and/or moral distance between the two parties. We propose a model that extends previous work by allowing not only for ex ante and ex post fairness but also for altruism. The model is consistent with observed behavior. In particular, a reduction in social and moral distance significantly increases the likelihood of equal split and more than equal split choices.