Deterministic vs. Stochastic Altruism

We report experimental evidence from a 3-player Dictator Game in which dictators decide over the distribution of probabilities of winning a fixed, indivisible monetary prize. This evidence is compared with (i) a control treatment in which money is perfectly divisible and dictators allocate shares of the prize across the group members and with (ii) two novel “hybrid” protocols by which a fraction of the prize is allocated deterministically, and the remainder by way of a lottery. Dictators’ decisions are framed within a (suitably modified version of) Saito (2013)’s model of distributional justice, also controlling for (own payoff) risk aversion. This allows us to disentangle -both structurally and experimentally- motives for equality of opportunity from motives for equality of outcomes. Our structural estimates suggest that the former have a greater impact on subjects’ distributional choices (favoring altruism in the stochastic frame), whereas subjects’ revealed risk aversion lowers the opportunity cost of giving (favoring altruism in the deterministic frame). In our data, those conflicting effects cancel each other, yielding a no significant difference in giving between the stochastic and deterministic setups.