Résumé
Reputation mechanisms are mainly based on information sharing by traders about private trading experience. Each trader can therefore rely on his own past experience as a trader and on other traders’ past experience. The former is the direct component of the reputation mechanism and the latter the indirect component (Bolton et al., 2004a and Bolton et al., 2004b). We design an experiment for isolating the direct component of the reputation system and studying its effect on the level of trust and reciprocity in a population where agents play both roles (trustor and trustee). Our experiment consists on three treatments of a finitely repeated investment game (Berg, Dickhaut, & McCabe, 1995). In the reference treatment there is no reputation mechanism at all, in treatment 1 trustees can build up a direct reputation, and in treatment two players can build up a direct reputation for both roles. We find that trustees’ direct reputation has a positive effect on reciprocity, but does not affect the average trust in the population. Trust is significantly higher only when players can build up a reputation in both roles. We show that the increase in trust is mainly linked to the formation of mutual trust–reciprocity relations.