674 International Organization hindering the benefits conferred to producers in other areas,and the costs of these measures are born generally by all consumers.In such cases,legislatures are ex- pected to generate a high degree of unanimity in the form of a universal logroll.17 Each member of the legislature will propose generous protective measures for indus- tries in his or her own district without accounting for the costs they impose on con- sumers elsewhere.To gain support for these measures,each member will vote in favor of similar measures proposed by other legislators.If members can vote indefi- nitely on a sequence of such proposals,a universal logroll is supported as an equilib- rium by each legislator's belief that a vote against another's proposal would induce others to retaliate by offering an amendment to withdraw protection from the defec- tor's district.18 The result of such unchecked logrolling is that district-specific ben- efits(in this case,protective measures)are oversupplied:the costs to consumers are such that all legislators are worse off than they were before the bill was passed. This lesson was brought home to Congress,so the argument goes,by the infamous Smoot-Hawley Act of 1930 in which logrolling appeared to run wild.20 Congress supposedly resolved to mend its ways and delegate control of policy to an agent who would act more responsibly:the president.Having a national constituency,the presi- dent is expected to internalize the negative (cross-district)externalities to consumers when setting tariff rates and adopt lower,more efficient,levels of protection.The most sophisticated version of this argument is provided by Susanne Lohmann and Sharyn O'Halloran.21 They develop a model of the delegation of trade policymaking authority in which Congress can choose between full delegation,partial delegation (which imposes a congressional veto on the president's proposal),and zero delega- tion.In the absence of partisan considerations,they expect legislators will prefer the more efficient outcomes provided by full delegation to the president.If the president is expected to be partisan in setting tariffs,however,favoring legislators in the presi- dent's own party over others,full delegation is only expected when the president's party has majority control since,under divided government,the congressional major- ity has an incentive to constrain the president.It fits,according to this view,that the RTAA provided full delegation to the president under unified Democratic govern- ment in the 1930s,allowing for rapid trade liberalization,and that greater restraints on presidential tariff cutting were imposed in later years under divided rule. How convincing is this anti-logrolling story?A debilitating weakness in this ac- count of the RTAA,as Karen Schneitz has pointed out,is that the congressional voting records indicate precious little learning on the part of legislators.Among members of Congress voting on both bills,almost all those who voted for Smoot- Hawley in 1930 voted against the RTAA in 1934:225 representatives voted on both bills in the House,for instance,and only 9 of these changed their views in the liberal 17.Weingast 1979. 18.Lohmann and O'Halloran 1994,605. 19.Weingast,Shepsle,and Johnsen 1981. 20.The Senate alone made 1,253 amendments to the original House bill,and duties on over twenty thousand items were altered.Pastor 1980,77-78. 21.Lohmann and O'Halloran 1994.674 International Organization hindering the benefits conferred to producers in other areas, and the costs of these measures are born generally by all consumers. In such cases, legislatures are ex￾pected to generate a high degree of unanimity in the form of a universal logroll.17 Each member of the legislature will propose generous protective measures for indus￾tries in his or her own district without accounting for the costs they impose on con￾sumers elsewhere. To gain support for these measures, each member will vote in favor of similar measures proposed by other legislators. If members can vote indefi￾nitely on a sequence of such proposals, a universal logroll is supported as an equilib￾rium by each legislator's belief that a vote against another's proposal would induce others to retaliate by offering an amendment to withdraw protection from the defec￾tor's district.18 The result of such unchecked logrolling is that district-specific ben￾efits (in this case, protective measures) are oversupplied: the costs to consumers are such that all legislators are worse off than they were before the bill was passed.19 This lesson was brought home to Congress, so the argument goes, by the infamous Smoot-Hawley Act of 1930 in which logrolling appeared to run wild.20 Congress supposedly resolved to mend its ways and delegate control of policy to an agent who would act more responsibly: the president. Having a national constituency, the presi￾dent is expected to internalize the negative (cross-district) externalities to consumers when setting tariff rates and adopt lower, more efficient, levels of protection. The most sophisticated version of this argument is provided by Susanne Lohmann and Sharyn O'Hall~ran.~~ They develop a model of the delegation of trade policymaking authority in which Congress can choose between full delegation, partial delegation (which imposes a congressional veto on the president's proposal), and zero delega￾tion. In the absence of partisan considerations, they expect legislators will prefer the more efficient outcomes provided by full delegation to the president. If the president is expected to be partisan in setting tariffs, however, favoring legislators in the presi￾dent's own party over others, full delegation is only expected when the president's party has majority control since, under divided government, the congressional major￾ity has an incentive to constrain the president. It fits, according to this view, that the RTAA provided full delegation to the president under unified Democratic govern￾ment in the 1930s, allowing for rapid trade liberalization, and that greater restraints on presidential tariff cutting were imposed in later years under divided rule. How convincing is this anti-logrolling story? A debilitating weakness in this ac￾count of the RTAA, as Karen Schneitz has pointed out, is that the congressional voting records indicate precious little learning on the part of legislators. Among members of Congress voting on both bills, almost all those who voted for Smoot￾Hawley in 1930 voted against the RTAA in 1934: 225 representatives voted on both bills in the House, for instance, and only 9 of these changed their views in the liberal 17. Weingast 1979. 18. Lohmann and O'Halloran 1994,605. 19. Weingast, Shepsle, and Johnsen 1981. 20. The Senate alone made 1,253 amendments to the original House bill, and duties on over twenty thousand items were altered. Pastor 1980,77-78. 21. Lohmann and O'Halloran 1994