£1(£5) or Nothing in Dictator Games: Unexpected Differences

We conducted an online Dictator Game experiment (N = 1,195) to test three hypotheses about the role of monetary incentives in prosocial behavior. First, we examined whether real incentives reduce the dispersion of responses compared to hypothetical ones. Surprisingly, we found the opposite: hypothetical responses were less dispersed, with choices clustering around the egalitarian split. This pattern held in a replication (N = 308) with higher stakes (£5), offering no support for the first hypothesis. Second, we tested whether real incentives—by involving actual monetary consequences—lead to more selfish decisions, as they are expected to reveal true preferences. With £1 stakes, no significant differences emerged across conditions. However, when the stake was increased to £5, participants became more selfish under real incentives, supporting the second hypothesis only when the amount at stake is substantial. Third, we explored whether probabilistic payments differ behaviorally from certain ones. At low stakes, probabilistic incentives resembled real ones. But with higher stakes, real and probabilistic outcomes diverged, suggesting participants respond to expected value only when it is meaningful. Finally, in a separate study (N = 299), we found that many participants misunderstood hypothetical-payment instructions. Only explicit phrasing eliminated this confusion, underscoring the importance of precise wording in experimental design.