Kandori (1992) "Social Norms and Community Enforcement" RES, 59

The paper considers self-enforcing mechanisms in the situation where agents change their partners over time. Technically, the model in the paper is an extension of the theory of repeated games to the case of matching games. As main results, the following two are shown.

1) Community can sustain cooperation even when each agent knows nothing more than his personal experience.
2) Community can realize any mutually beneficial outcomes when each agent carries a label which is revised in a systematic way. That is, Folk Theorem holds.

As a motivation of the research, he mentions the benefit of specialization. After introducing theoretical achievements in personal enforcement mechanisms, i.e. Folk Theorem in the repeated game literature, he says as follows.

However, many important transactions are infrequent in nature. As economic historians argue, the division of labor and specialization are important driving forces of economic progress. Potential gains are larger in diverse transactions with different specialists than with fixed partners. Therefore, the control of incentives in such an infrequent trade is of vital importance to understand the organization of economic transactions.

He refers to two papers which initiated this line of research.

The attempt to generalize the Folk Theorem of repeated games to the case of matching games was initiated by Milgrom, North and Weingast (1990) and Okuno-Fujiwara and Postlewaite (1989). The former analyzed concrete examples of information transmission mechanisms and the latter introduced the notion of local information processing. Both of them, however, mainly deal with the infinite population case to avoid potentially complicated problems of incentives on off-equilibrium paths. Our paper shows that such problems can be resolved in a simple way if the stage game satisfies weak condition. Equilibria constructed in our paper work for any population size and any matching rule, and are robust to changes in information structures.

What a strong result he derived!! Although he does not stress the results given in Section 3 "Folk Theorem under Public Observability", I think Proposition 2 is very interesting. It is easy to understand that Folk Theorem holds if all the other players get into punishment phase after some player deviates, which is stated as Proposition 1. However, if we restrict our attention to such a situation where only the deviator is to be punished and innocent pairs are to play the originally prescribed actions, to show Folk Theorem is not straight forward. To be more precise, to check the incentives for innocent players in off-equilibrium path where community is highly populated with "guilty" agents is involved.
Introducing some "forgiveness" in the social norm, the author elegantly shows this problem can be avoided which leads to Proposition 2.

Interesting Papers in References

Harrington (1989) "Cooperation in Social Settings" mimeo
Section 7 of the above paper was revised and available as the following.
Harrington (1995) "Cooperation in a One-Shot Prisoners' Dilemma" GEB, 8
Milgrom, North and Weingast (1990) "The Role of Institutions in the Revival of Trade: The Law Merchant, Private Judges, and the Champagne Fairs" Economic Inquiry, 25
Okuno-Fujiwara and Polstlewaite (1995) "Social Norms in Random Matching Game" GEB, 9
Rubinstein and Wolinsky (1990) "Decentralized Trading, Strategic Behavior and the Walrasian Outcome" RES, 57