MERCHANT GUILD 751 a mob destroyed all the Italian quarters in Constantinople during the "Latin massacre"of 1182(Day [1988];for additional examples, see also De Roover [1965,p.61];Lane [1973,p.34];Kedar [1976, p.26ff.]. In light of the theory of repeated games,one might conjecture that a ruler's commitment problem could be solved by a bilateral reputation mechanism in which individual merchants whose person and property were not protected by a local ruler would refuse to return with their goods in the future.The ruler,while perhaps reaping short-run gains from ignoring a merchant's rights,stood to lose the future stream of rents from the cheated merchant's trade.5 As we demonstrate for- mally in Section II,this intuition is misleading.At the level of trade that maximizes the total net value of trade-that is,at the efficient volume of trade-a bilateral reputation mechanism cannot resolve the commitment problem.In our formal theory,the reason is that,at the efficient volume of trade,the value of the stream of future rents collected by the ruler from an individual marginal merchant is almost zero and is therefore smaller than the value of the goods that can be seized or the cost of the services that can be withheld.The same conclusion would hold even at lesser volumes of trade if the fre- quency of visits by an individual trader were low.As long as ruler- merchant relations are governed only by a bilateral reputation mecha- nism,our theory holds that trading volume cannot expand to its efficient level. The preceding discussion and the formal model below allow only one kind of sanction for cheated merchants:the withdrawal of trade. Military action might seem another important alternative.In the late medieval period,however,defensive technology was superior to of- fensive technology,and the costs and risks of offensive military action at distant ports limit its credibility as a sanction for trade violations.? A possible means to increase the punishment is a multilateral re- sponse by all the merchants to transgressions against any subgroup of merchants.Indeed,the history of the relations between trade cen- ters and alien merchants presents several examples of multilateral retaliations against rulers who had reneged on their contractual obli- 6 Clearly,there was a limit to the security a ruler could provide the merchants. Accordingly,we have detailed above instances in which rights were abused in major cities or trade centers in which the relevant ruler had a relatively high level of ability to secure rights. 7 Parker(1988,p.7)comments that "After the proliferation of stone-built castles in western Europe,which began in the eleventh century ..in the military balance between defence and offense,the former had clearly become predominant."This situation changed only during the so-called Military Revolution of the fifteenth century.MERCHANT GUILD 751 a mob destroyed all the Italian quarters in Constantinople during the "Latin massacre" of 1182 (Day [1988]; for additional examples, see also De Roover [1965, p. 61]; Lane [1973, p. 34]; Kedar [1976, p. 26 ff.]). In light of the theory of repeated games, one might conjecture that a ruler's commitment problem could be solved by a bilateral reputation mechanism in which individual merchants whose person and property were not protected by a local ruler would refuse to return with their goods in the future. The ruler, while perhaps reaping short-run gains from ignoring a merchant's rights, stood to lose the future stream of rents from the cheated merchant's trade.6 As we demonstrate for￾mally in Section II, this intuition is misleading. At the level of trade that maximizes the total net value of trade-that is, at the efficient volume of trade-a bilateral reputation mechanism cannot resolve the commitment problem. In our formal theory, the reason is that, at the efficient volume of trade, the value of the stream of future rents collected by the ruler from an individual marginal merchant is almost zero and is therefore smaller than the value of the goods that can be seized or the cost of the services that can be withheld. The same conclusion would hold even at lesser volumes of trade if the fre￾quency of visits by an individual trader were low. As long as ruler￾merchant relations are governed only by a bilateral reputation mecha￾nism, our theory holds that trading volume cannot expand to its efficient level. The preceding discussion and the formal model below allow only one kind of sanction for cheated merchants: the withdrawal of trade. Military action might seem another important alternative. In the late medieval period, however, defensive technology was superior to of￾fensive technology, and the costs and risks of offensive military action at distant ports limit its credibility as a sanction for trade violations.7 A possible means to increase the punishment is a multilateral re￾sponse by all the merchants to transgressions against any subgroup of merchants. Indeed, the history of the relations between trade cen￾ters and alien merchants presents several examples of multilateral retaliations against rulers who had reneged on their contractual obli- 6 Clearly, there was a limit to the security a ruler could provide the merchants. Accordingly, we have detailed above instances in which rights were abused in major cities or trade centers in which the relevant ruler had a relatively high level of ability to secure rights. 7Parker (1988, p. 7) comments that "After the proliferation of stone-built castles in western Europe, which began in the eleventh century . . . in the military balance between defence and offense, the former had clearly become predominant." This situation changed only during the so-called Military Revolution of the fifteenth century