Trust and reciprocity: implications of game triads and social contexts

Abstract

Trust and reciprocity have been observed in experiments with simple stylized games. Such characteristics of social preferences have been observed to vary with procedures that alter the social environment in an experiment, such as single blind or double blind payoff protocols. This paper reports an experiment on the effects of a change in the social context of an experiment on trust and reciprocity. The strong social context introduces a stylized version of a characteristic of everyday life in large cities: a player in one game knows that other games lie ahead but does not know precisely what those games will turn out to be nor with whom they will be played.

Keywords:

• experiments
• trust
• reciprocity
• social context

Acknowledgements

Helpful comments and suggestions were received from the editors and two anonymous referees. Financial support was provided by the National Science Foundation (grant numbers DUE-0622534 and IIS-0630805).

Notes

1. This is not an example of the type of ‘indirect reciprocity’ modeled in Nowak & Sigmund (1998a, 1998b). In their model of repeated play, agents are rewarded for having a reputation for generous behavior and penalized for having a reputation for ungenerous behavior. In contrast, in the driving example the drivers do not appear to be rewarding other drivers only because of their personal reputations.

2. The triadic design for the investment game, and the way it decomposes trust, reciprocity, and altruism, are explained in more detail in Cox (2004). Methodological issues concerning the triadic design and across-subjects, double-blind protocols are discussed in Cox, Sadiraj & Sadiraj (2008).

3. This design feature means that there is no ‘social history’ in the way there is in the Berg et al. (1995) experiment.

4. The group decision making data from the second task are reported in Cox (2002), as are data on gender effects.

5. Single blind and double blind payoff protocols have been shown to produce different behavior in fairness games (Hoffman et al., 1994; Hoffman, McCabe, & Smith, 1996; Cox & Deck, 2005, 2006; Cox, Sadiraj, and Sadiraj, 2008).

6. As noted above, Berg, Dickhaut & McCabe (1995) used a double blind payoff protocol.

7. Treatments A_WC and A_SC are, respectively, the weak social context and strong social context investment games reported in this paper. Treatments NH and SH are, respectively, the no history and social history investment games reported by Berg, Dickhaut & McCabe (1995).

8. Conover (1980, pp. 368–373) explains the Smirnov test for comparing two empirical distributions and the Kolmogorov test (Conover, 1980, pp. 346–356) for comparing an empirical distribution with a theoretical distribution. The Smirnov test is reported here. Both types of tests are sometimes referred to as ‘Kolmogorov–Smirnov’ tests.

9. The tests reported in Table 3 do not use data for four of the subjects in sessions with treatment B_SC. Questionnaire responses revealed that three subjects were confused. Data for these three subjects (identified by their payoff key codes) are not used. One other subject was a repeat participant. The payoff key number of this subject was identified after the end of the experiment and the data are not used.