752 JOURNAL OF POLITICAL ECONOMY gations.For example,circa 1050 the Muslim ruler of Sicily imposed a 10 percent tariff (instead of the 5 percent tariff specified in the Islamic law)on goods imported to Sicily by Jewish traders.The trad- ers responded by imposing an embargo and sending their goods to the rival trade center,Tunisia.The embargo was effective,and after a year the Sicilian ruler relented and removed the tariff(David Kauf- mann Collection,Hungarian Academy of Science,Budapest,docu- ment no.22,pt.a,lines 29-31;pt.b,lines 3-5;Gil [1983,pp.97- 106];Taylor-Schechter Collection,University Library,Cambridge, document no.10 J 12,folio 26,p.a,lines 18-20;Michael [1965, 2:85]) The examples above suggest that a multilateral reputation mechanism might be able to surmount the commitment problem without the aid of any formal organization.In each case,merchants imposed a collective punishment on the city that included participation by merchants who had not been directly injured.Several of the cited offenses were offenses against an entire group of merchants.In medieval trade,however,a city could also discriminate among merchants,abus- ing or not protecting them selectively.For example,a city could con- fiscate the belongings of some traders or withhold legal protection from them without directly harming other alien merchants.Indeed, the Sicilian rulers increased the tariff only to Jewish traders;and during two attacks on the Genoese quarter in Constantinople,other Italian merchants were not harmed.This suggests two interconnected reasons why,without a supporting organization,a multilateral repu- tation mechanism may be insufficient to surmount the commitment problem at the efficient level of trade.The first involves contract ambiguities and asymmetric information,whereas the second reflects the distinct incentives among different traders generated by a multi- lateral response. Long-distance premodern trade took place in a highly complex and uncertain environment.Unanticipated events and multiple interpre- tations of existing agreements were always possible under these cir- cumstances,implying that the definition of a"contract violation"was often ambiguous.Information asymmetry,slow communication,and different interpretations of facts among merchants imply that without an organization that coordinates responses,it was not likely that all the merchants would respond to the abuse of any group of mer- chants.As demonstrated formally in Section II,if the fraction of merchants who detect and react to an abuse against any group of merchants is only proportionate to the number abused,then a multi- lateral reputation mechanism is ineffective at the efficient volume of trade.It is ineffective for the same reason that a bilateral reputation mechanism is ineffective:a threat by a group of marginal traders to752 JOURNAL OF POLITICAL ECONOMY gations. For example, circa 1050 the Muslim ruler of Sicily imposed a 10 percent tariff (instead of the 5 percent tariff specified in the Islamic law) on goods imported to Sicily by Jewish traders. The trad￾ers responded by imposing an embargo and sending their goods to the rival trade center, Tunisia. The embargo was effective, and after a year the Sicilian ruler relented and removed the tariff (David Kauf￾mann Collection, Hungarian Academy of Science, Budapest, docu￾ment no. 22, pt. a, lines 29-31; pt. b, lines 3-5; Gil [1983, pp. 97- 106]; Taylor-Schechter Collection, University Library, Cambridge, document no. 10 J 12, folio 26, p. a, lines 18-20; Michael [1965, 2:85]). The examples above suggest that a multilateral reputation mechanism might be able to surmount the commitment problem without the aid of any formal organization. In each case, merchants imposed a collective punishment on the city that included participation by merchants who had not been directly injured. Several of the cited offenses were offenses against an entire group of merchants. In medieval trade, however, a city could also discriminate among merchants, abus￾ing or not protecting them selectively. For example, a city could con￾fiscate the belongings of some traders or withhold legal protection from them without directly harming other alien merchants. Indeed, the Sicilian rulers increased the tariff only to Jewish traders; and during two attacks on the Genoese quarter in Constantinople, other Italian merchants were not harmed. This suggests two interconnected reasons why, without a supporting organization, a multilateral repu￾tation mechanism may be insufficient to surmount the commitment problem at the efficient level of trade. The first involves contract ambiguities and asymmetric information, whereas the second reflects the distinct incentives among different traders generated by a multi￾lateral response. Long-distance premodern trade took place in a highly complex and uncertain environment. Unanticipated events and multiple interpre￾tations of existing agreements were always possible under these cir￾cumstances, implying that the definition of a "contract violation" was often ambiguous. Information asymmetry, slow communication, and different interpretations of facts among merchants imply that without an organization that coordinates responses, it was not likely that all the merchants would respond to the abuse of any group of mer￾chants. As demonstrated formally in Section II, if the fraction of merchants who detect and react to an abuse against any group of merchants is only proportionate to the number abused, then a multi￾lateral reputation mechanism is ineffective at the efficient volume of trade. It is ineffective for the same reason that a bilateral reputation mechanism is ineffective: a threat by a group of marginal traders to