THESE ARTICLES ARE ASKING DIFFERENT QUESTIONS

Dougherty and Heckelman are trying to answer a different question from the question I tried to answer. They are asking a question about the way legislative processes work in general, whereas I am asking a question about the substantive design of American government in particular. They care mainly about roll call voting because roll calls provide the basic data for their statistical analysis of Convention events. These authors examine the Constitutional Convention to find out how the rules can put a particular individual in a pivotal position to affect legislative outcomes. From their point of view, it makes sense to try to compare Sherman's role in the Constitutional Convention and the first U.S. Congress. But I was interested in personal influence at the Constitutional Convention only insofar at it was necessary to understand how the Constitution turned out the way it did, instead of some other way. “Madison's Opponents and Constitutional Design” examines the way the provisions of the initial agenda, the Virginia Plan, evolved into the final document signed September 17. My article simply looks at the ways his opponents defeated substantive provisions that James Madison, the Convention's chief agenda-setter, initially claimed to be indispensable for the Constitution's design. A comprehensive analysis of the Constitution's design requires too much space to include in a journal article (see Robertson 2005b).

Because Dougherty and Heckelman frame their problem in a way that is so different from “Madison's Opponents,” “A Pivotal Voter” has nothing to say about the substance of the final U.S. Constitution. Instead, it focuses on votes and, to a lesser extent, motions. The authors' examination of 569 recorded votes at the Convention reveals only that the states' votes aligned along two dimensions. We do not know anything about the substance of these dimensions, other than the finding that Connecticut occupied the median position on the one dimension throughout the Convention, but other states occupied the median position on the other dimension during different stages of the Convention. The authors' question does not require them to know anything about the substance of these two dimensions. We therefore learn nothing about the way these dimensions can help explain Constitutional design, the puzzle at the heart of “Madison's Opponents.” Dougherty and Heckelman “merely” speculate that the two dimensions roughly correspond to issues of representation and centralization (the authors allude to Jillson 1988, who developed similar dimensions in more detail).

Readers should note well that if one dimension involves representation, then in none of the three periods of the Convention is Connecticut the “pivotal” delegation on representation. This suggestion makes sense only in a technical way. Connecticut surely is pivotal in representation in the common sense of the word, as playing a central role in bringing about an outcome (the sense used by Rodriquez and Weingast [2003] to explain Senator Everett Dirksen's impact on the design of the 1964 Civil Rights Act). After all, the Convention's most famous single decision, the “Great Compromise” on the apportionment of the House and Senate, is widely known as the “Connecticut Compromise.” This widely shared appellation acknowledges Connecticut's impact on the Convention's turning point and its most significant provision regarding representation.

THESE ARTICLES RELY ON DIFFERENT ASSUMPTIONS ABOUT POLITICIANS' PREFERENCES AND VOTES

If “A Pivotal Voter” were asking the same question as “Madison's Opponents,” Dougherty and Heckelman's spatial analysis of Convention votes would offer limited help for explaining the Constitution's design. They assume that if we can merely identify the median voter in this case, we can draw important conclusions about the outcome. But their model allows us to infer very little about the Constitution's specific provisions. The assumptions for the spatial model in “A Pivotal Voter” would allow us to explain the Constitution's design in terms of delegate preferences only if (1) the delegates' votes perfectly represented their preferences and (2) Convention roll call votes provided a sufficient explanation of the Constitution's design. As the authors explain in their appendix, however, the first of these conditions requires them to assume that the delegates' preferences are fixed, and that votes sincerely express these preferences. To the extent that these preferences are not fixed, and votes were not sincere expressions of these preferences, their estimates are subject to error, and if their estimates are subject to error, so is their conclusion that Sherman's influence resulted from the fortunate situation in which he found himself.

But the delegates' preferences were not fixed, and their votes did not perfectly express the framers' design preferences. The assumption that actors have fixed preferences is not reasonable for explaining the details of complicated constitutional choices. This assumption is hard enough to justify when an established legislature produces a complex “super-statute” like a far-reaching civil rights bill or social security program (Eskridge and Ferejohn 2001). It is a much greater problem in a constitutional convention that alters the structure of governance. In a legislature, actors routinely make decisions that directly affect policy outcomes. Legislators can be relatively certain about the way a provision in a tax, highway, or environmental bill will affect their constituents. But in a constitutional convention, delegates are making the rules about the rules for making public policy; most of their decisions affect policy outcomes indirectly and somewhat unpredictably. Constitutional architects are shaping a complicated set of instruments for achieving preferred outcomes, rather than outcomes themselves. Rarely can they be certain the way a given provision, interacting with other provisions, will affect their vital interests. When only about half the votes had been cast in the Constitutional Convention, Madison observed, “[w]hat the effect of A union of these provisions might be, could not be foreseen” (RFC August 14, II, 291). Uncertainty in Philadelphia increased as the weeks wore on.

Because of this uncertainty about outcomes, the delegates were more likely to base design choices on flexible and instrumental preferences than on fixed preferences. The Constitutional Convention records illustrate that the delegates' preferences were dependent on other provisions and likely results. As the delegates generally grew more uncertain of the kinds of outcomes the new government would produce, their preferences for Constitutional design evidently shifted. The delegates were trying to lock in institutional prerogatives that would promote the use of national power to produce outcomes they preferred and to inhibit outcomes they opposed. They came to design structures that would guarantee agents of their favored interests the power to defend against a future use of national authority that could threaten their vital interests (see Moe 1990). Connecticut's delegates, for example, represented a complicated polity with diverse local interests, and its political leaders were keenly attuned to these interests because they had carefully balanced them as they steered their state's political development (Collier 1971, 2003). For them, maintaining Connecticut's relative weight in selecting the new legislature and president took precedence over other design elements such as short legislative terms or protections for state police powers. Sherman in particular practiced the art of the possible, adjusting to changing Convention's dynamics even as he indirectly helped shape the flow of the Convention's politics (Collier and Collier 1986, 127–34). James Madison provides the best example of a delegate whose design preferences shifted in response to Convention developments. The Connecticut Compromise made Congress a less suitable instrument for Madison's nationalist objectives because it put the Senate under the control of the self-serving state governments that in his view created the worst problems for the Confederation. When the Connecticut Compromise was adopted, Madison immediately began to champion an autonomous president as a tribune for national interests against the Senate's parochialism (Robertson 2005b, 153, 215–17).

When so many choices are instrumental, votes may be insincere and strategic, aimed at altering the sequence of decision or increasing influence in a sequence of decisions. Though Dougherty and Heckelman offer reasons that insincere voting was unlikely, there were stronger forces at work that relaxed constraints on disingenuous votes. The fact remains that Connecticut either was not sincere when it helped design the New Jersey Plan in mid-June or was not sincere when it voted against that plan 4 days after it was introduced. Massachusetts voted against a proposal to allocate Senate seats on the basis of population on July 14 (RFC July 14, II, 12), a crucial vote cast against Massachusetts' own interests in apportionment. For a majority of the Massachusetts' delegates at this moment, concessions that could facilitate a national government more capable of governing commerce were preferable to a continued stalemate over Senate apportionment. Delegates operated under considerable uncertainty about which provisions would serve their interests and which would not, which provisions could win majority support in the Convention and which could not, and which provisions would generate external opposition in the ratification process and which would not. During the Convention, as likely outcomes of Constitutional design changed, the delegates adjusted their instrumental preferences to these changes.

Thus, we cannot assume that the delegates' votes faithfully expressed their preferences. It is evident that the delegates brought complicated bundles of preferences to the Convention, that they viewed some preferences as a higher priority than others, and that some preferences were more flexible and contingent than others. Complexity and contingency help account for shifts and inconsistencies during the Convention.

Three votes illustrate why it is impossible to assume that Convention roll calls expressed delegates' true preferences. On May 31, two days after the introduction of the Virginia Plan, the delegates voted overwhelmingly in favor (nine states for, none opposed, and Connecticut divided) of the Plan's provision that the national legislature would have “power in all cases to which the State Legislatures are incompetent.” This vote clearly was contingent on other design choices. Though the South Carolina delegation voted in favor, three of the state's four delegates expressed strong reservations and made it clear that their support depended on “the exact enumeration of power comprehended by this definition” (RFC May 3, I, 53–4). On August 13, when several delegates moved to give the House of Representatives exclusive power to originate money bills, George Washington, renowned as a pillar of integrity, cast an insincere vote in favor of the proposal. Madison noted that Washington “disapproved & till now voted ag[ain]st, the exclusive privilege, he gave up his judgment he said, because it was not of very material weight with him & was made an essential point with others, who if disappointed, might be less cordial in other points of real weight” (August 13, II, 280). In late August, when the issue of admitting new states into the union arose, many state delegations completely reversed their earlier positions on states' rights. They did so to increase the likelihood that the new government would deal with the controversial issue of public lands in a way consistent with their states' interests. Gouverneur Morris of Pennsylvania, previously a supporter of restricting state government obstructionism but representing a state with a large land domain, moved that no new state could be created within the limits of an existing state unless that state's legislature approved. Luther Martin, the Convention's most zealous state's rights advocate on other issues but keenly interested in the national control of western lands, argued forcefully against any state's right to obstruct the creation of new states, because that right would alarm the “limited States” (that is, states with much less land within their borders). Martin articulated the contingent nature of the debate when he observed that “[i]n the beginning, when the rights of the small States were in question, they were phantoms, ideal beings. Now when the Great States were to be affected, political Societies were of a sacred nature” (RFC August 30, II, 461–6).

THESE ARTICLES TAKE DIFFERENT VIEWS OF THE SIGNIFICANCE OF THE SEQUENCE OF DESIGN CHOICES

Because the delegates' preferences were so contingent, understanding the sequence of choices is crucial for understanding the Constitution's design (on sequence, see Fenno 1986; Pierson 2005, 54–78). The delegates agreed to specific arrangements in a specific sequence. Each choice affected likely outcomes of government policy, and therefore the delegates' calculations about the benefits and costs of subsequent design choices. The most influential delegates understood the importance of sequence very explicitly. In a letter he wrote to George Washington before the Convention, Madison emphasized that settling the issue of representation on terms favorable to larger states would make them more willing to vest the national government with stronger powers: “the consideration which particularly urges a change in the representation is that it will obviate the principal objections of the larger States to the necessary concession of power” (PJM 9, 383). Again, Madison's preference for a strong and independent executive is not evident before July 16, but it became very open and forceful after his final defeat on Senate apportionment on that date; this preference was contingent, not hidden. Votes alone do not allow one to appreciate the subsequent impact of the final apportionment of seats in the first House of Representatives. Fortunately, the Convention records include much more evidence than mere vote tallies. One has to examine committee reports, verbal responses to the reports, and the substantive impact of the proposals to understand how this apportionment decision affected the interests of the Carolinas and New York. The evidence strongly suggests that this provision altered the way the Convention unfolded by encouraging the withdrawal of two New York delegates and Southerners' strategic efforts to protect slavery.

Dougherty and Heckelman recognize that the delegates' expressed preferences sometimes were contingent. They acknowledge that delegates occasionally shifted position on a particular question because of cues from an influential delegate, or from “subtle changes in the question, fatigue, or a variety of other factors.” But I think “A Pivotal Voter” understates crucial contingencies at the Convention. For example, there is considerable uncertainty about who actually was casting votes on behalf of a state on a given day. There is no daily record of attendance, and Max Farrand's records provided only a range of days when a delegate seemed to be in Philadelphia, not whether he actually attended and voted. Attendance mattered because it affected crucial votes. It appears that Maryland's Daniel of St. Thomas Jennifer missed the critical July 2 vote on Senate apportionment, for example, resulting in a deadlock, followed by the Connecticut Compromise and the defeat of Madison's original Convention strategy (RFC July 2, I, 150; see III, 188).

THESE ARTICLES TAKE DIFFERENT VIEWS OF ACTORS' INFLUENCE ON CIRCUMSTANCES SUCH AS RULES

Because the delegates' preferences were flexible and sequence had so great an impact on Constitutional design, it was possible for a skilled politician to play a pivotal role in Constitutional design by shaping the rules and sequence of choice even if he was not a “pivotal voter” in a spatial model. A spatial model of roll call votes does not and cannot test whether a median voter actually did anything to influence the outcome. Without some information about the way that delegate took advantage of his opportunity, identifying a statistically derived median voter is a just a curiosity, a statistical product that proves little about influence on substantive outcomes. Convention delegates, like legislators, have to act to take advantage of their opportunities (see Hall 1996). In contrast, a delegate could shape the substance of the Constitution without leaving direct evidence in votes or motions if he shaped the circumstances of choice, including the rules and the sequence of decisions. Rules don't grow on trees. This kind of influence does not require that a politician have extraordinary charisma or prestige. Both Madison and Sherman were experienced, cunning politicians who understood that rules and the sequence of decisions affect political results. They succeeded by showing up, working hard, and manipulating the framework of decision. Madison manipulated the decision framework by shaping the Convention's initial agenda. Sherman manipulated the decision framework by influencing alternative agendas like the New Jersey Plan and the committee reports on legislative representation and the presidency.

Unlike “Madison's Opponents,” “A Pivotal Voter” assumes the rules are given, and it assumes away participants' manipulation of the rules and agendas before and during the Convention. This is an assumption more appropriate for an established legislature like Congress than for an ad hoc convention (but see Schickler 2001). Though they provide evidence that Roger Sherman's influence followed from his pivotal position, Dougherty and Heckelman do not try to examine the possibility that Sherman influenced rules and agendas to put himself into a pivotal position. They cannot show that the either Sherman or Madison failed to manipulate successfully the structure and timing of the Convention's choices. William Riker used the term “heresthetics” to describe such skills of political manipulation (1986, 1996). The Convention records show that the delegates as a group had a very keen understanding of the way the rules and the structure of choices could shape outcomes (Robertson 2005b, 106–16).

Several delegations took steps to manipulate the rules before the Convention began. The Virginians successfully pressed for rules that gave each state an equal vote, but allowed a simple majority of states to decide issues. Consistent with Madison's expectation of six supporting votes (three large states and the three Southern states), Virginia gambled that smaller states could be prevailed upon “in the course of the deliberations, to give up their equality for the sake of an effective Government” (RFC May 28, I, 10–11n3). Connecticut authorized any one—and only one—of its delegates to vote on behalf of the state, strategically increasing the state's opportunity to weigh in on issues even when only a single delegate attended a session in Philadelphia. No other state had the prescience to advantage their delegation in this way (RFC May 30, I, 37; III, 559–86). Delaware's George Read, one of his state's leading politicians, deliberately influenced the Delaware legislature to write credentials mandating that any constitutional change would not alter Delaware's equal vote in Congress, and Delaware's delegates used that provision against Madison's agenda in the early days of the Convention (RFC III, 575n6; Rakove 1996, 60). During the Convention, the delegates evidently understood the importance of heresthetics well enough to discuss at length the rules for ratifying the Constitution. The entire Constitutional Convention, after all, can be viewed as effort by experienced politicians to manipulate rules of American public policymaking to achieve better outcomes than the Confederation government was producing.

This agenda manipulation, though critically important to Constitutional design, can hardly be measured by roll call votes. Dougherty and Heckelman subtly acknowledge this problem when they retreat to evaluating motions. They argue that we can measure personal political influence by examining which delegates made the most motions and had the most success in getting their motions passed. Motions often are part of an effort to change the agenda, though in a much narrower sense than a proposal for an entire new framework for government like the Virginia and New Jersey Plans. But motions also are contingent, and looking at motions apart from other evidence cannot provide much insight into personal power in a setting like a constitutional convention. I think Dougherty and Heckelman's data on motions tends to support my claims about Sherman's political skill. Sherman's success rate on motions is about the same as Madison's. If motions alone are an indicator of influence, Madison does not stand out; yet I am not aware of a reputable scholar who seriously contends that Madison failed to exercise enormous influence at the Convention. Sherman's relatively limited success on motions, then, cannot prove that he lacked personal influence. Both Madison and Sherman tended to advance their arguments strategically by responding to the motions of other delegates. Years later, Sherman showed that he fully understood the tactical utility of inaction during debate when he explained to a young member of the U.S. Congress that “minorities talk; majorities vote” (Collier 1971, 316). Motions actually may indicate political weakness instead of political skill. Perhaps that would explain why three of the delegates who had a better success rate on motions than Sherman or Madison—Edmund Randolph (the delegate whose motions passed more frequently than any other's), Elbridge Gerry and George Mason—were so unhappy with the final Constitution that they refused to sign it.

Note that “A Pivotal Voter” cannot account for Madison's reputation as a delegate whose strong influence on the Constitution is undisputed. It is true that Virginia is a median state on one dimension during the last two periods of the Convention, but Virginia's Shapley-Owens score is lower than that of Connecticut in each period. Madison may or may not have been the pivotal delegate in his state. The authors do not claim that Madison's personal influence is overrated, but I think that is the logical implication of their argument. They suggest as much when they concede that Madison and Sherman “may have influenced the convention in other ways” than motions (italics mine). One can glean delegates' preferences for Constitutional design only by looking at all their recorded behavior in context. The analysis of Constitutional design choices depends on analyzing votes and motions together with speeches, committee reports and other information as part of an evolving sequence of choices contingent on preceding choices.

Note also that when “A Pivotal Voter” locates evidence to demonstrate that Sherman's influence at the Convention was situational rather than personal, it does not use evidence from the Convention at all. Instead, Dougherty and Heckelman argue that because Sherman was not a pivotal voter in the first U.S. House of Representatives (1789–1791), his personal skills were less important at the 1787 Constitutional Convention than his pivotal position. This conclusion does not follow. The Convention and the first Congress are two different decision-making bodies, with different rules, very different substantive agendas, jurisdictions, and participant behaviors. Sherman had to play a pivotal position in the Constitutional Convention to protect his constituents; his behavior in the first U.S. Congress may actually reflect his success in engineering a government design that accomplished that goal. Readers must decide whether Sherman's behavior in the first U.S. Congress really proves anything about “A Pivotal Voter's” central question, whether Sherman's power at the Constitutional Convention was situational or personal.