We exploit the fact that generosity and trustworthiness are highly correlated and the former can thus be a sign of the latter. Subjects decide between a generous and a mean split in a dictator game. Some of them are informed from the start that afterwards they will participate in a trust game and that their choice in the dictator game may matter; others are not informed in advance. In the trust game, before trusters decide whether or not to trust, some trustees can reveal (or conceal) only their true choice in the dictator game, while others can say to trusters, truthfully or otherwise, what they chose. We find that a generous choice made naturally by uninformed trustees and reliably revealed is more effective in persuading trusters to trust than a generous choice that could be strategic or a lie. Moreover, we find that, when they can, mean subjects lie and go on to be untrustworthy.

Laborexperiment

Laboratory experiment

In total, 215 subjects participated in our computerized laboratory experiment. Subjects were Oxford University students, 46% were female and they were 23.4 (sd = 4.91) years old on average.

All participants in our experiment were recruited from the subject pool maintained by the Centre for Experimental Social Sciences (CESS) in Oxford. A random sample of participants was drawn from this subject pool and people included in this sample were invited via e-mail to participate in the experiment.