A Spatial Analysis of Delegate Voting at the Constitutional
Convention∗
Jac C. Heckelman
Department of Economics
Wake Forest University
122 Carswell Hall
Winston–Salem, NC 27109
[email protected]
Keith L. Dougherty
Department of Political Science
University of Georgia
Baldwin Hall
Athens, GA 30602
[email protected]
Abstract
Past analyses of the U.S.Constitutional Convention have relied strictly on votes
recorded for the states or a relatively small number of roll calls. We construct a
new dataset covering delegate votes on over 600 roll calls. We utilize the data
in several ways. First, we use the votes to estimate a single dimensional position for the delegates to determine overall voting patterns. Next, we explain
these positions using a variety of delegate and constituent variables in a double–
censored heteroskedastic Tobit model. Finally, we suggest a method to identify
potential state and floor medians to predict equilibrium outcomes.
∗ Funding for this research was supported by the National Science Foundation, SES–0752098. We thank Paul Carlsen, Rocky Cole,
Rachel Columb, George Cone, Christopher Cotter, Brandon Kliewer, Yao Kang, Monica Petrescu, and Rebecca Sherman for research
assistance. We also appreciate comments from participants at West Virginia University, Dept of Economics seminar; the Southern
Economics Association conference, Topics in Cliometrics session; the 2011 Public Choice meetings; and the 2010 Midwest Political
Science Association meetings.
1
Introduction
The Constitutional Convention of 1787 has inspired a variety of claims. To some, the delegates acted as
disinterested patriots guided by principle (Bowen 1966; Morgan 1993). To others, the Constitution was an
economic document designed to benefit the interests of some groups at the expense of others (Beard ([1913]
2004). It is not difficult to see how a multitude of ideas, interests, and sociological factors may have affected
delegate voting at the convention. What is more elusive is to try to determine which ideas, interests, and
sociological factors were the most resonant.
Although a variety of assertions have been made, few have been rigorously tested. As historian James
Hutson puts it (1987a, 411), “evidence has not been sufficiently evaluated, claims have not been confirmed,
theses have not been tested. The result is a profusion of unsubstantiated assertions and contradictory hypotheses that frustrates the attempt to describe the course of events at the Constitutional Convention.” One
major impediment to studying delegate behavior comes from the lack of data on delegate votes. Because
the delegates voted in state blocs and wanted to maintain secrecy to promote candid discussion, the convention journal and Madison’s notes recorded the majority vote of the state blocs, but rarely recorded the
votes of individual delegates. As Gibson (2007) duly notes, this has seriously hampered the ability for
scholars to analyze questions regarding the voting behavior of individual delegates.
In this paper, we attempt to determine if various delegate attributes can systematically explain their
voting behavior. We face two difficulties. First, as noted above, actual votes are not known and must be
inferred. Second, we need a mechanism for grouping substantive motions together.
Previous empirical studies of delegate voting have “overcome” these problems by limiting themselves
to a small subset of motions. For example, McGuire and Ohsfeldt (1984, 1986), McGuire (2003), and Heckelman and Dougherty (2007) analyze up to 16 specific motions originally chosen by McDonald (1958). They
classify delegate positions on these motions by whether or not the position is pro–national. Dougherty and
Heckelman (2008) infer delegate votes for nine motions related to slavery and classify delegate votes by
whether or not they voted on the pro–slavery side of the issue.
Each of these studies provide new insights into delegate voting at the convention, but they are limited
to specific issues and may not generalize to over–arching themes. Indeed, some of the conclusions regard-
1
ing delegate motivations contrast with each other, possibly because different attributes motivate different
issues. Thus, a broad consensus regarding delegate motivations across the convention has not been found.
Even if we were to address all issues simultaneously, it is not clear how to combine separate votes
into a single group. For example, pro–national and pro–slavery positions do not always align themselves
consistently – particularly across a large number of issues. As a case in point, McGuire and Ohsfeldt (1986)
could not pool all of McDonald’s 16 motions because they could not determine a clear pro–national position
on all 16 motions. We want to analyze delegate positions on hundreds of motions without relying on these
type of characterizations.
We address the problem by first inferring delegate votes based on statements made by delegates in debate for all substantive motions at the convention. Rather than pooling votes directly for empirical analysis,
we use spatial modeling techniques to estimate a single dimensional position for each delegate. Such procedures identify delegates who tend to vote similarly without relying on the content of a motion (such as
determining if a ‘yea’ vote on a specific motion would be considered “pro–national” or “pro–slavery,” etc.).
Next, we try to explain delegate positions using a variety of delegate and constituency variables that
reflect various claims about the ideas, interests, and demographics of the delegates that appear in the literature.
Our regressions reveal that the relative positions of the delegates are explained primarily by whether
or not they were anti–Federalists, the number of years they served in legislative, judicial, and executive
service, if they owned slaves or private securities, and the extent to which their state complied with federal
requisitions during the confederation. A less robust effect stems from the region of the country from which
a delegate resides. Some insignificant factors include whether a delegate held public securities and the
population of his state. These results help show the important effects of a strong central government,
political experience, and commitments made by a delegate’s state on their voting behavior. Their personal
interests in the new sets of institutions, as measured by slave and private security ownership, also appear
to have affected their voting positions.
To further illustrate the usefulness of our study, we make out–of–sample predictions for the location of
delegates for whom we have too few inferred votes to be included in our initial spatial estimates. We then
use the full set of spatial estimates to determine the median delegate for each state and the median state on
2
the floor. We find that median delegates from Georgia and South Carolina were typically pivotal, with the
particular state median determined by whether or not New York or New Hampshire attended. Because the
medians of Georgia and South Carolina held similar positions, the distinction between these two positions
may have had little effect on the final form of the Constitution. Thus, despite the very different positions
held by the median delegates from New York and New Hampshire, the departure of the New York delegates and later arrival of the New Hampshire delegates would have a smaller effect than many scholars
might presume.
2
Inferring Delegate Votes
Because delegate votes were not recorded at the convention, we first inferred delegate votes for each of
the 620 non–procedural motions at the convention (only 569 of which were numbered in the journal)1 in
three steps.2 We assigned a delegate the vote recorded for his state if the state only had two delegates in
attendance.3 By the rules of the convention, the position of each state (yea, nay, divided) was determined by
a simple majority of the state’s delegates. Hence, if there were only two delegates in attendance from a state
and the state voted yea or nay, we inferred that both delegates voted the same as the vote recorded for their
state. For example, on vote 387 to prohibit the states from issuing paper currency, Georgia is recorded as a
yea. Thus, Abraham Baldwin and William Few were both coded yea because they were the only members
from Georgia present.
Additional delegate votes were then inferred using statements made by the delegates in debate as
recorded in the notes of James Madison, Robert Yates, Rufus King, and others (collected in Farrand 1966).
We also used statements from personal manuscripts and speeches published in Farrand’s (1966) volume
1 Procedural motions included motions to adjourn, commit, postpone, and reconsider. Some of the motions to postpone were
actually substitute amendments. We treated such votes as non–procedural motions and attempted to code them.
2 Our procedure is briefly outlined here. Additional details and comparisons to other smaller datasets are developed by Dougherty
et al. (2012).
3 Our attendance records are based on Farrand (1966, 3: 586-590), with updates from Hutson (1987b) and Collier (1971) – the latter
was used to find more precise dates for the attendance of William Samuel Johnson and Roger Sherman when the Committee of Detail
met. We also examined the Journals of the Continental Congress to determine if delegates were meeting in Congress when they were
allegedly at the Constitutional Convention; we looked at the minutes of the General Assembly of Pennsylvania and the minutes of
the Supreme Executive Council of Pennsylvania to see if any of the Pennsylvania delegates were conducting state business during a
Convention meeting; and we consulted biographies for members from New Jersey and Delaware to see if we could attain more precise
times of departure for specific delegates. The additional sources provided no new information.
3
3 or the supplement (Hutson 1987b) if they could be tied to a particular roll call on a particular day. A
statement was applied to a vote only if it came from the same debate.4 For example, Nathaniel Gorham, a
member of the Massachusetts delegation, was inferred as voting nay on vote 387 because in response to the
motion to add an absolute prohibition of state currency to Article XII (of the resolves of the Committee of
Detail), Gorham said the purpose of the clause was already protected in Article XIII which required the consent of the national legislature for the states to issue currency. He further claimed “an absolute prohibition
would rouse the most desperate opposition from its partizans” (Farrand 1966, 2: 439). George Read, from
Delaware, strongly opposed paper currency two weeks earlier, but he was not coded on vote 387 because
his statement was not directly related to that vote.
After the positions of the delegates were recovered, attendance records were again consulted to determine whether additional delegate votes could be inferred from the state’s vote and the fact that each state’s
vote was determined by a majority of its delegates. For example, Massachusetts was recorded as a yea
on vote 387. Because Caleb Strong was absent and Gorham was coded as nay, the two remaining Massachusetts delegates, Elbridge Gerry and Rufus King, must have voted yea in order for a majority of the
Massachusetts delegation to have voted yea.
These three coding steps produced 5,121 vote inferences, by far the largest data set assembled on constitutional votes.5 This results in an average of 93 vote inferences per delegate. Baldwin has the most codes
with 451 inferences, while William C. Houston, who left the convention one week after it began, has no
codes.
Of the 620 substantive motions, only 398 had at least one delegate on both sides of the issue. W–
nominate, the spatial estimation technique described below, requires at least one delegate on each side
to be included in the analysis because one-sided roll calls provide no information about relative distances.6
4 Procedurally, we considered a debate started when the issue was first raised (by formal motion or the item was reconsidered) and
ended when the substance of the issue was voted upon and a new issue was raised. This allowed for the inclusion of cases where
discussion continued immediately after a vote and cases where an issue was raised, debated, and voted upon at a later date. In the
vast majority of cases, “debates” lasted less than a day.
5 In constract, the McGuire and Ohsfeldt (1984, 1986) data set contains 848 inferences. However, many of their inferences are for
delegates who were not in attendance on a particular vote and some of their votes were filled in using assumptions based only on
delegate support or opposition to the final Constitution. For further examination of the McGuire and Ohsfeldt inferences and their
methodology, see Heckelman and Dougherty (2007).
6 There were actually 399 roll calls with at least one delegate on each side of an issue. However, vote 168 was dropped from the w–
nominate estimates because we required delegates to have a minimum of ten votes to be included in the estimation. James McClurg,
the sole dissenter on vote 168, only had six inferences. It should not be assumed that the excluded motions were unanimous because
observations may have been missing on the dissenting side of such issues.
4
Among these 398 roll calls, we were able to record a yea or nay position for an average of 10.1 delegates per
roll call. Thus our included sample contains 4,102 usable observations.
2.1
Can Votes be Inferred from Statements?
The second step in our coding process implicitly equates a position stated by a delegate with his vote on
the issue. In other words, it assumes no strategic “talking.”7 This assumption is reasonable for two reasons.
First, because votes were taken verbally and sequentially, delegates who stated positions contrary to their
vote could incur fairly high costs. Other delegates would notice their inconsistencies and might discount
their statements on future votes. Second, in cases where historians believe there was a vote trade (such as
preventing Congress from prohibiting the importation of slaves until 1808 traded for rejection of a 2/3rds
requirement to pass navigation acts) delegates who were alleged parties to the vote trade always stated
positions consistent with the vote trade they were about to cast or they remained silent. Thus, there is good
reason to expect strategic “talking” to be rare.
To test this conjecture, we compared “known” votes of the delegates to votes inferred from delegate
statements. One set of “known” votes are those with no more than two delegates from a state in attendance.
For each of these cases, if a state voted yea or nay (as opposed to divided) both delegates must have voted
the same way to be consistent with the position recorded for their state. Following the methods described
previously, we asked a coder to infer votes in such cases based on statements made in debate. We then
compared these codes to the “known” votes of the delegates.
Of these 61 cases, 58 were in agreement (95%). Furthermore, only three statements came from days
other than the vote. In all three of these cases, the statement was coded consistent with the delegate’s vote.
This experiment suggests that the process used to infer votes provides a fairly accurate assessment of actual
votes. In other words, cases of strategic talking (where a delegate talks one way but votes another) appears
to have been rare.
7 Strategic “talking” differs from strategic “voting.” A delegate “talks” strategically (our terminology) if he makes a statement in
favor of a proposal (status quo) but votes in favor of the status quo (proposal). Strategic “voting” refers to differences between a
delegate’s vote and his preferences, which typically occurs if an individual votes against his preferences at a particular point in an
agenda to attain a more preferred outcome at the end of the agenda. As Londregan (1999) points out, strategic “voting” appears to
have been mitigated at the convention because the agenda was not fixed and there was an explicit rule that adopted clauses could be
revisited at later dates (Farrand 1966, 1:17).
5
3
Estimation of spatial positions
After inferring the delegate votes, we then estimate a single dimensional mapping8 of the delegates using
a 55 by 398 (delegate by motion) matrix and w–nominate.9 W–nominate is a parametric scaling procedure
that explicitly assumes each delegate votes for the alternative closest to themselves based on a probabilistic
utility function. The parameters of the function and the spatial coordinates of the delegates can be estimated
from binary choice information such as observed votes. In one dimension, voter ideal points (i.e., delegate
coordinates) are restricted to the [–1, 1] interval, similar to the [0, 100] interval used by ADA or other interest
groups to rank members of Congress. One notable difference between w–nominate and ADA scores is that
ADA scores (and other interest group scores) require the liberal side of each issue (or the side favored by the
interest group) to be identified a priori and for that position to be more extreme than any of the voters. The
resulting scale then measures a conservative to liberal spectrum (or whatever the relevant interest group
represents). W–nominate does not use a priori assessments of motions. Hence, it can scale votes that appear
to have little in common on the surface.10
We excluded delegates with 10 or fewer inferred votes in our w–nominate estimation and used James
Wilson, a Federalist, as the right restriction. Of the 55 delegates at the convention, 42 had enough inferred
votes to be included in the analysis. Among them, five had between 10 and 20 inferred votes and five others
had more than 200 inferred votes.11 Figure 1 displays the nominate scores and bootstrapped standard errors
created from this procedure.
Two observations are worth noting. First, the scale appears to reflect localism on the left and nationalism
on the right. This represents the classic dichotomy between individuals who favor decentralization and
those who favor centralization. Three of the six delegates identified by Riker (1987) as Anti–Federalists are
at the left endpoint, with another, Yates (NY), very close to the left extreme (– 0.95). The Anti–Federalist
furthest to the right is John Mercer, who arrived late, departed within eleven days and still remains left
8 Poole and Rosenthal (1997) find that one–dimension is usually appropriate for Congressional roll–call votes over the entire history of Congress. We are able to correctly classify 81% of the convention votes using one dimension, roughly equivalent to their
classification rates for House (83%) and Senate (80%) votes.
9 W–nominate is available at <http://pooleandrosenthal.com/wnominate.asp>.
10 Poole and Rosenthal (1997) show that w–nominate scores are highly correlated with ADA scores in modern Congresses. A detailed
description of w–nominate can be found in Poole and Rosenthal (1997, pp. 249–251).
11 We were able to include at least three delegates from each state (except New Hampshire which only sent two). Specifically, our
w–nominate estimates include 3 out of 3 delegates from Connecticut (3/3 CT), 4/5 DE, 3/4 GA, 4/4 MA, 5/5 MD, 3/5 NC, 2/2 NH,
3/5 NJ, 3/3 NY, 3/8 PA, 4/4 SC, and 5/7 VA.
6
of center (– 0.56).12 At the same time, some of the convention’s most ardent Federalists, Wilson, Madison,
Read, and King are on the far right of the scale. Alexander Hamilton (0.87) is also toward the far right.
However, there are exceptions to this interpretation. Edmund Randolph, the delegate who presented the
Virginia Plan, is on the left (– 0.53) even though he showed signs of being an “ardent nationalist ... during
the convention” (Riker 1987, 14).13 Many of the more moderate delegates like Mason and Sherman are near
the center of the scale – though no one has previously argued for an exact location. Thus, while almost all
of the expected positions are consistent with the literature, many other delegates positions are revealed for
the first time. This gives us greater confidence in our estimates while filling in some historical gaps.
Second, the bootstrapped standard errors are largest for delegates who spoke the least, such as Jacob
Broom (DE), Jonathon Dayton (NJ), and Richard Dobbs Spaight (NC). We were less capable of inferring
votes for such delegates and it is not surprising that their bootstrapped standard errors are larger than the
standard error for delegates such as James Madison (VA), who spoke often.14 Although the standard errors
from w–nominate are often ignored in empirical analysis, we incorporate them directly into the empirical
model as detailed in the next section.
4
Econometric Methodology
Having estimated spatial positions for most delegates, we next seek to explain overall voting patterns at
the convention. Our empirical model is
Yi = Xi B + ui
(1)
where Yi is delegate i’s ideology, Xi is a matrix of potential determinants (including a constant term), B is a
vector of parameters to estimate, and ui is a random error term such that ui ∼ N (0, v2 ). We do not directly
observe ideology. Rather our observed proxies for delegate ideology are generated by w–nominate. Let yi
12 The
remaining Anti–Federalist, Alexander Martin of North Carolina, did not have enough inferred votes for inclusion.
did not sign the Constitution because he opposed the method of ratification, opposed a single executive, and wanted a
more republican form of government. However, he switched sides at the Virginia ratifying convention and became an advocate of the
federal constitution. His w–nominate score seems to reflect such indecisiveness.
14 The correlation between the number of inferred votes and the estimated standard error of the w–nominate ideology score is –0.69.
We easily reject the null hypothesis of no correlation, distributed as t(40) = –6.01, with p–value = 0.00.
13 Randolph
7
represent the w–nominate scores, where
yi = Yi + zi
(2)
and zi represents measurement error associated with the nominate procedure such that zi ∼ N (0, wi2 ).
Substituting (2) into (1) suggests that our regression actually takes the form
y i = X i B + ei
where ei = ui + zi , such that ei ∼ N (0, s2i ). Assuming Corr (ui , zi ) = 0, then si =
(3)
q
v2 + wi2 . This derivation
suggests that our model is heteroskedastic because the dependent variable is an estimate. In standard linear
models, heteroskedasticity associated with measurement error of the dependent variable leads only to biased standard errors, but the coefficient estimates are still consistent (Lewis and Linzer 2005).15 Estimation
of robust standard errors is a standard solution to the problem.
However, ideal points estimated by w–nominate are limited to the [–1, 1] scale. As shown in Figure 1,
five delegates have a w–nominate score at the left endpoint, and five delegates are at the right endpoint.
Thus, a non–trivial portion of our data is censored. The censored nature of our data is represented by
yi =








−1,
Yi + zi ,







1,
Yi + zi ≤ −1
–1 < Yi + zi < 1
(4)
Yi + zi ≥ 1.
For this reason, Tobit estimation is appropriate, treating the data as both top and bottom censored. In Tobit
models heteroskedasticity leads to inconsistent, not merely inefficient, parameter estimates (Wooldridge
2004). Thus, applying robust standard errors is not sufficient in our case. Instead, we estimate the heteroskedastic error term, si , as part of the Tobit routine.
Using our censored values from (4), the likelihood function of the double–censored heteroskedastic
15 Lewis and Linzer (2005) report through Monte Carlo simulation that jackknifed standard errors outperform White’s adjustment
in small samples although they advocate primarily for a two–step FGLS procedure. However, they focus exclusively on linear models
and do not consider the effects from censored data where heteroskedasticity leads to inconsistent parameter estimates (Wooldridge
2004).
8
Tobit takes the form
L=
∏
{i |−1<yi <1}
[si−1 f ((yi − Xi B)si−1 )]
∏
{i |y=−1}
[ F ((−1 − Xi B)si−1 )]
∏
{ i | y i =1}
[1 − F ((1 − Xi B)si−1 )]
(5)
where f (•) is the standard normal density function, F (•) is the standard normal cumulative distribution
q
v2 + wi2 from above.16 Output from w–nominate generates both nominate scores
function, and si =
(yi ) and bootstrapped standard errors (wi ), which we use in the estimation of (5). Maximizing the log of
L produces estimates for B̂ and v̂, from which ŝi can be computed if desired. Our initial interest is in the
estimation of B̂, the unconditional marginal effects of each variable on latent ideology from (1). As described
in a later section, we can also use this method to make out of sample predictions on the conditional expected
value of ŷi for delegates missing from the w–nominate sample.
5
Explaining delegate ideology
We now attempt to explain overall voting patterns at the Constitutional Convention using (5) as our likelihood function and the single dimensional nominate scores from Figure 1 as our dependent variable. Our
explanatory variables, motivated by the literature, are described below.
The most obvious explanatory variable is a dummy for whether the delegate was an Anti–Federalist.
Just as party affiliation explains most of the variation in contemporary legislatures (Peltzman 1984; Poole
and Rosenthal 1997; Monroe et al. 2008), whether or not a delegate was an “Anti–Federalist” might explain
a large amount of the variation in delegate positions at the Convention. As described earlier, Figure 1 shows
that almost all Anti–Federalists are on the far left. Yet, as obviously important as such a label might be, it
fails to explain any of the variation among those identified as Anti–Federalists nor any of the variation
among those not identified as Anti–Federalists. Hence, additional variables should provide some further
explanatory power.17
16 See Maddala (1983, p.161) for additional details on the double–censored homoskedastic Tobit model. As noted by Greene (2007),
the heteroskedastic model for a single-censored Tobit can be derived by simply replacing the homoskedastic error term with a heteroskedastic error term in the Tobit likelihood function. We apply the same logic to the double-censored Tobit model employed here.
17 Studies of contemporary legislative bodies which use nominate or ADA scores as the dependent variable will typically include a
variety of constituency (state or district) characteristics and the representative’s party but no other personal characteristics (Peltzman
1984; Goff and Grier 1993; Jung et al. 1994) except for race or gender when that is of specific interest (e.g. Poole and Rosenthal 1997;
9
Political experience is a good candidate. Jensen (1950) argues that one of the factors which affected
whether a delegate was localist or a nationalist was their political experience. Delegates who worked in local politics were beholden to local constituents and became more localist. Those who spent years in national
politics saw the problems of a weak confederation and learned to favor stronger national institutions (also
see Main 1973 and Ferguson 1969). To capture this idea, we include the number of years of past political
experience, and consider separate breakdowns for federal versus state, and also breakdowns by legislative,
executive, or judicial branches, separately.
Another explanation for delegate positions may be slavery. In recent works, Kaminski (1995) and Beeman (2009) imply that slavery was a dominant issue at the convention. Heckelman and Dougherty (2010)
further show that slave ownership affected whether a delegate supported the final version of the Constitution. Hence, we include a dummy for whether a delegate owned slaves as an explanatory variable.18
Grubb (2003, 2006) further argues that delegates with financial interests in the Bank of North America
(BNA) favored granting the BNA a monopoly on currency issues and blocking the ability of states to issue
their own currencies. Heckelman and Dougherty (2010) extend the argument by suggesting that this applies to stockholders of state banks as well because state bank notes were in competition with the bills of
credit issued by state governments. In contrast, Wittman (1995) argues that the people later identified by
Grubb held such beliefs for more disinterested reasons. According to Wittman, delegates with a financial
interest in the BNA wanted to erect a set of institutions that would protect sound financial interests, assure
the payment of current debts, and tax at rates necessary to pay federal expenditures without having to
accrue additional debt. Much earlier, Beard ([1913] 2004) argued that those who held securities were part
of an economic class that derived advantages from a new constitution distinct from those who primarily
owned real estate. He believed these two coalitions battled over key clauses at the convention. Despite
differences regarding delegate motivation, each of these studies suggest that delegate behavior may be
affected by the ownership of securities. To capture the economic interests of the delegates, we include a
dummy for whether the delegate owned public securities and a dummy for whether they owned private,
Griffen and Newman 2007). Because all delegates were white males, race and gender cannot be used in the current study.
18 The specific number of slaves owned by a delegate at the time of the convention is unclear. For example, McDonald (1958) reports
that William Blount owns 30 slaves, whereas McGuire (2003) has Blount owning 80 slaves. The 1790 census corroborates McDonald
but only for Blount’s Pitt county estate, and records Blount as having an additional 22 slaves in his Tyrell county estate. We use a
dummy variable because it may be a more accurate, though less precise, measure.
10
bank securities.19
In addition, delegates may have acted as representatives of their states. Various scholars have noted
differences between the large and small states over issues of apportionment (Jillson and Anderson 1978;
Slez and Martin 2007). To represent this concept, we include the total population of each state as an additional independent variable.20 It is important to keep in mind, however, that two small states, Georgia and
North Carolina, voted with the large state coalition during the apportionment debate because they anticipated considerable population growth (Jillson and Anderson 1978). Hence, population at the time of the
convention may not adequately capture state interests in apportionment, even though it may control for
other interests and abilities related to state size.
We also include a variable measuring the compliance of a delegate’s state with federal requisitions prior
to the Convention. The Continental Congress raised funds, soldiers, and supplies by requisitioning the
states according to a pre–specified apportionment. Without any enforcement power it was up to the states
to voluntarily comply. Dougherty (2001) argues this was a major failing of the Articles and a justification
for the new constitution.21 Delegates from states with greater compliance may have been frustrated with
states free-riding on requisitions and more cognizant of the need for a strong federal government with the
power to tax independently.
Finally, Elazar (1972, 103–112) argues that political culture emanated from the three colonial regions of
New England, the Mid–Atlantic, and the South. Each of these regions had economic, ethnic, and religious
differences based on the people that colonized them. New England maintained a characteristically moral
culture, the Mid–Atlantic observed private and individualistic values, and the South maintained a landed
gentry reminiscent of the old world (also see Davis 1977). To determine the effect of regional differences,
we include dummy variables for delegates from the South (up to Virginia) and Mid–Atlantic (Maryland
19 The specific value of securities held, or the value relative to a delegate’s personal wealth, may better capture the intensity of a
delegate’s interest. However, McDonald (1958) criticizes Beard’s estimates for delegate bond holdings and created alternative values
based on records housed at the National Archives. In our visits to the National Archives we found McDonald missed some bond sales
to certain delegates (typically when they purchased bonds in other states) or sometimes counted purchases made after the convention.
Unfortunately, some of the state records are no longer available and/or too damaged to make a complete accounting for each delegate
so we instead use a simple dummy rather than use the specific values created by Beard or McDonald.
20 Total population and total white population in a state are highly correlated at 0.91. Although we report estimates using total
population, we get similar results using either measure.
21 Congleton (2011, 542) claims “Governance under the Articles of Confederation proved adequate for a time of war . . . However, it
was widely regarded to be too weak to advance national interests during the time of peace after the war had been won in 1783.” For
contrasting comparisons of the Articles and the new Constitution, see Congleton (2011, 538–544) and Dougherty (2001, ch: 4, 7).
11
to New York), leaving New England as the omitted category. These variables could represent either the
cultural preferences of the individual delegate or his perceived view of the preferences of his constituents.
Descriptive statistics for all variables are presented in Table 1. Data sources are listed in the appendix. A
proto–typical delegate in our sample (for which we have a w–nominate score) was a 43 year old Federalist
with 12 years of past political experience (8 years of legislative experience, 4.3 years of experience in the
judicial branch, and 3.2 years of experience in the executive branch).22 In addition, the delegate would have
been in service at the federal level for 3.8 years, mostly in the Continental Congress, and 10.4 years at the
state or local level. He was likely to have held public but not private securities, and he did not personally
own slaves. He came from a state of roughly 340,000 persons, 3/4ths of whom were white, and 20% of
which were slaves. His state sent only one–quarter of the money requested by the Continental Congress,
and he was more likely to have lived in the Mid–Atlantic than in the South or New England.
6
Regression Estimates
Each of the variables from the upper portion of Table 1 are included in our base specification, except for
the political variables which have substantial overlap. Specifically, we initially include the total years of
past political experience, then present a separate regression which includes state and federal experience,
followed by a third regression which breaks down political experience by branch.
Coefficient estimates for the base specification appear in Table 2, column (I). As presented in (5) and
(1), these estimates represent the unconditional marginal effect on the latent ideology of interest rather
than the same effect on the observed dependent variable (i.e. the nominate scores). The estimates suggest
that, ceteris paribus, Anti–Federalists and those from the mid–Atlantic region tended to hold ideological
positions significantly to the left of other delegates. Even though four of the five Anti–Federalists were
from mid–Atlantic states, the high level of statistical significance for both coefficients suggest they capture
independent effects.
We also find that owning private securities and/or slaves tends to move a delegate in a pro-national
direction (i.e., to the right). This might be surprising because the delegates who owned slaves eventually
22 We do not double count the total number of years for a delegate who simultaneously serves in separate branches. Hence, the
number of years of political experience is often less than the sum of the number of years for the separate branches.
12
lost their slave–holding rights because of a stronger federal government. At the time of the convention,
only Massachusetts had abolished slavery, although Connecticut and Pennsylvania had passed legislation
for the gradual emancipation of slaves.23 On the other hand, Beard ([1913] 2004) argued that slave owners
favored a strong national government because they wanted the ability to call on federal support to suppress
potential slave revolts. Our result is consistent with Beard’s theory about slavery although it does not
address his rationale directly.
Yet, Beard’s primary focus was on bondholders. He thought holders of state and continental securities
were “immediately concerned in the establishment of a stable national government” (p. 32). We do not
find ownership of public securities to be a significant factor, and thus our findings can only provide limited
support for Beard’s thesis.
We do find, however, that private securities had a positive and significant effect. Owning private securities tended to make a delegate more nationalistic, consistent with the rent-seeking theory of Grubb (2003,
2006) and good governance theory of Wittman (1995). Both argued that those who owned private bank
stock supported a strong national government, but for reasons different from each other and from Beard.
In addition, we find that greater compliance from a delegate’s state pushed delegates further to the
right. This is consistent with the argument that delegates from states in greater compliance with federal
requisitions were more cognizant of the need for a strong federal government. They may have been more
concerned about free–riding on common concerns than delegates from states that complied less.
Our estimates also indicate that neither the number of years of previous political experience nor the size
of the state were significant factors in explaining delegate voting patterns. The latter result is inconsistent
with the notion of state size having a dominant effect across the convention, but, as noted previously,
current population totals may be a weak proxy for membership in the large state or small state coalitions.
Although some of our variables were insignificant, we are not suggesting they were never important on
any individual votes. For example, McGuire and Ohsfeldt (1986) find ownership of public securities to be
significantly related to yea or nay votes on four of the fourteen votes they analyze, including a prohibition
on export tariffs. Our results simply suggest that owning public securities did not have a significant impact
23 Although most states allowed slavery, all states except for Georgia had effectively banned the importation of slaves. South Carolina imposed a three year ban in 1787 as a precaution against slave revolts and North Carolina enacted a “prohibitive duty” (DuBois
([1896] 1969: 223–9).
13
on delegate voting patterns overall.
Given that our nominate scores appear to represent a distinction between localists on the left and nationalists on the right, in column (II) we replaced the number of years of political experience with separate
variables indicating the number of years of federal or state/local experience. Estimated coefficients for
the two variables are of opposite sign as expected but the former is not statistically significant. The more
years a delegate served in state and local politics, the more localist the delegate tended to vote – though
this variable is only significant at the .10 level. The importance of owning private securities, which just
barely meets the 10% error criterion in column (I), now falls slightly under that threshold in column (II).
The estimated marginal impacts suggest 30 years of state or local political experience were needed to fully
offset the nationalist ideological effect from owning slaves. Thus the slave owner William Houstoun (GA),
who had no state/local political experience, and the non–slave owning Benjamin Franklin (PA), who had
31 years of state/local experience, would be expected to hold roughly similar ideological positions at the
convention if they came from the same state.24 Similarly, Gerry (MA), the lone anti–Federalist from outside
the mid–Atlantic region, would need 34 years of additional state/local experience compared to the other
anti–Federalists to match their ideological positions, if their other state and personal characteristics were
identical.
Except for Franklin, Langdon, and both Morrises, previous federal experience comes almost exclusively
from service in the Continental Congress. In contrast, only about half of the delegates with state or local
experience earned their years in a state or local legislature. Thus, federal experience may be capturing
legislative experience rather than experience in national politics, per se.25
To test this notion directly, we separated years of political experience into three variables, presented in
column (III): previous years of legislative service, previous years of executive service, and previous years
of judicial service. More years of service in the legislature pushes a delegate further to the right (as did
federal experience, insignificantly) and more experience in executive or judicial work moves a delegate to
the left (as did state/local experience). The level of statistical significance for each of the branch coefficients
is greater than for either of the federal or state/local coefficients in column (II). Note the magnitude of the
24 Neither
delegate was an Anti–Federalist, and the other significant determinants are all state–level variables.
a check, we replaced the federal politics variable with the number of years of federal experience in the Continental Congress.
This variable also had the effect of pushing delegates to the right and was significant at the 6% level. In addition, the state/local
politics variable increased in significance to better than 5%.
25 As
14
legislative and judicial effects are the same but in opposite directions. Hence, if their other characteristics
were the same, we would predict McHenry’s five years in the legislature and two years in the judiciary
would put him in a similar ideological position as Davie’s three years of exclusively legislative experience.
It is not clear why these political variables have their estimated signs. One explanation stems from
the fact that the Articles of Confederation provided no federal judiciary and almost no federal executive.
All federal business was conducted through a federal legislature. Although a more nationalist constitution
might separate and strengthen the three branches, the greatest increases in strength occured in the executive
and judiciary (Congleton 2011, 543). Delegates with many years of local judicial experience knew local
laws and may have worried that a federal judiciary would require them to learn new legal precedents and
potentially overrule their prior (or future) local and state court decisions through a federal appeals process.
In the same vein, delegates with years of executive service (almost entirely at the state or local level) might
view the federal government as a competing power that could infringe on state sovereignty and undermine
their authority as an executive. In contrast, delegates with years of legislative experience may have seen the
failures of the confederation first hand (both at the state and federal level) and felt relatively unthreatened
by a stronger federal government which already had a legislature.26
Separating the branches in column (III) also returns private securities to statistical significance (at an
increased magnitued and a greater level of significance) while eliminating the significance of the regional
effect. Everything else equal, owning private securities pushes a delegate 0.5 units to the right. Owning
slaves pushes him 0.6 units to the right. Owning both would still not fully offset the leftward effect of being
an Anti–Federalist.
For each percent that a state that was in compliance with federal requisitions, a delegate tended to move
0.02 units to the right. For comparative purposes, consider delegates from New York, a state which had the
highest level of compliance at 64%, but none of its delegates owned slaves. Every delegate from Virginia
owned slaves but came from a state with only 38% compliance with federal requisitions. Our estimates
suggest that these delegates would hold similar ideological positions if their other characteristics were held
constant. The sign and significance of the remaining variables were not affected significantly by separating
26 This argument does not contradict the traditional view that the federal legislature was the strongest of the three branches. Arguing
that the federal judiciary and executive may have increased their strength more than the federal legislature increased its strength does
not imply that the federal legislature was not the most powerful branch in the new constitution.
15
political experience into three branches.
We also ran several additional regressions, which are not presented in the table. Elkins and McKitrick
(1961) suggest that younger delegates were more likely to support the Constitution than older delegates
because they were eager to rectify the nation’s problems by altering the design of government institutions.
Including a variable for the delegate’s age produced a negative coefficient that was never statistically significant.
Several pundits claim that religion affected political attitudes at the convention. Bradford (1982) lists
Congregationalists, Diests, Dutch Reformed, Episcopalians, Lutherans, Methodists, Presbyterians, Quakers, and Roman Catholics as the denominations represented at the convention. We tried various meaningful combinations of these sects. The Puritans (which we define as one of the four Calvinist religions:
Congregationalist, Dutch Reformed, Methodist, or Presbyterian) stood out as a useful category because
they emigrated to America to avoid religious persecution and seemed generally opposed to strong central
authority. Their unique combination of heightened political awareness and austere lifestyle set them apart
from other religions at the convention, and may have affected their preference for localism over nationalism. A second scheme we considered grouped delegates that practiced a hierarchical religion (Episcopalian
or Roman Catholic) into one category and delegates that practiced a communitarian religion (Congregationalists, Diests, or Quakers) into another. Finally, we considered the importance of being from a religion
that actively supported the abolition of slavery during the period of our study (Methodists or Quakers).
Various combinations of these variables were added to the specification reported in column (III). None
were significant.
We also considered various ways of measuring federal debt held by constituents in a delegate’s state.
The total federal debt in a state included Loan Office Certificates (federal bonds), IOUs for conscripted
goods, and salaries owed to veterans. None of these measures were significant individually or in combination, on a total per capita or per white male only basis. The number of slaves per capita (or per white
male only) was also found to be insignificant.27 Finally, we considered an alternative definition of region
(North vs. South) with the North including the six states north of Delaware. This regional dummy was not
27 We note that when controlling for population as we do in our regressions, the only variation in the per capita variables for debt
and slaves would come from the numerators and thus be similar to including the raw totals rather than their relative values. In this
case, the variables might be representing a size rather than intensity effect. Dropping the state population variable did not affect the
insignificance of the debt and slave per capita variables.
16
significant.
Finally, to gauge the effect of our heteroskedastic correction, we also generated estimates from homoskedastic double–censored Tobit regressions for the specifications reported in columns (I) – (III). In each
case, the signs and significance of the variables remain the same, but the magnitudes for the coefficients are
generally larger, suggesting that ignoring the measurement error associated with the w–nominate scores
would over–estimate the size of the marginal effects.
Our analysis therefore suggests a delegate’s ideology was significantly impacted by whether he was
an Anti–Federalist, the number of his prior years in executive, judicial, or legislative work, if he owned
private securities or slaves, and the extent of his state’s compliance with federal requisitions. Ownership
of public securities and the population size, level of debt, and proportion of slaves in the delegate’s state
were not significant, nor was the delegate’s age or religious background. Being from the mid–Atlantic may
have affected his ideology but the effect was not robust, and an alternative regional definition was never
significant. Thus, we expect delegates on the left of the ideological spectrum to be Anti–Federalists with
several years of executive and judicial experience, who came from a state in the mid–Atlantic which hardly
complied with federal requisitions, such as New Jersey. Delegates on the right would tend to be Federalists
who served many years in the legislature, owned slaves and bank securities, and represented a state such
as Pennsylvania, which had largely complied with federal requisitions.
7
Estimating Spatial Positions for Missing Delegates
One of the useful implications of our regression analysis is that it can be used to predict the spatial location
of delegates who were excluded from the w–nominate procedure because they had less than ten inferred
votes. In the next section, we present a potential application of the data by identifying each state median
position. In the simplified methodology we use, every delegate needs to be included.
The predicted values for the dependent variable of a heteroskedastic Tobit model censored at –1 and 1
are calculated as
17
b + ŝi [ f (h−1i ) − f (h1i )][ F (h1i ) − F (h−1i )]−1 ] − F (h−1i ) + F (1 − h1i )
ŷi = [ F (h1i ) − F (h−1i )][ Xi B
(6)
b)ŝi .28 It can be shown that the predicted values of this representation fall within [–1,
where hki = (k − Xi B
1] and are thus comparable to the w–nominate scores for the other delegates.
We have all of the explanatory variables for the delegate locations we wish to predict, but because the
delegates were not included in the w–nominate procedure, we are missing values for their measurement
q
error (wi ) which we would need to compute ŝi = v̂2 + wi2 .29 We consider several measures of wi based on
the standard errors estimated by w–nominate for the other delegates. We considered using the minimum,
mean, and maximum values of the observed measurement error from the delegates with nominate scores as
the measurement error for delegates without nominate scores. As can be observed from Figure 1, the range
of potential standard errors from w–nominate estimates is quite varied, from a low of 0.055 (Luther Martin)
to a high of 0.323 (William Davie). Figure 2 depicts the predicted nominate score for missing delegates
using these three types of errors and the specification in Table 2, column (III). For comparative purposes,
the w–nominate scores of the other delegates are also included. As can be seen in that table, the three
different measures of wi appear to have little affect on the predicted values for the 13 missing delegates.
Predicted spatial positions for the missing delegates are spread throughout the distribution, with four
delegates placed left of –0.45, six delegates placed right of 0.45, and three in between. This suggests that
our missing delegates are fairly randomly distributed over the interval, allaying concerns about potential
sample selection bias in our regressions reported earlier. As the figure shows, distinctions between the three
types of measurement error have little impact on the predicted positions and never reverse the rank order
of delegates from the same state.
28 This formulation is derived from Maddala (1983, p.161) with substitutions for the values of the upper and lower limits and the
heteroskedastic error term.
29 Conversely, v̂ is estimated as part of Equation (5).
18
8
Median Voters
With a complete list of nominate scores we can estimate the location of the median delegate from each
state and consequently the median delegates of all the state medians.30 For this exercise, the maximum
measurement error of the delegates estimated by w–nominate (0.332) is used to predict the locations of the
13 “missing” delegates. Because the standard error from nominate is highly correlated with the number
of inferred votes, and our missing delegates each have fewer than 10 inferred votes, we would expect
the missing delegates to have relatively high standard errors on their nominate scores. Hence, predicting
positions based on the maximum measurement error appears most appropriate.31
Figure 3 presents the position of each state, as determined by its median delegate (or co-median delegates in the case of even sized delegations).32 Scholars have long recognized Pennsylvania’s voting record
to be the most nationalistic of the states (McDonald 1958; Jillson 1981; Gibson 2007). In constrast, New
York and New Jersey are generally thought to support strident localist positions. The other states are less
clear-cut, although Massachusetts and Virginia are often depicted as relatively nationalistic as well (the latter in part due to the prominence of James Madison and the Virginia Plan). Figure 3 is consistent with this
conventional wisdom and also depicts where the other states fall in between.
It is interesting to note that the location of a state’s median is not always near all of the state’s delegates.
Massachusetts was a pro-national state at the convention, but Elbridge Gerry was in a localist position,
as one might expect. New York was a localist state, because of the position of John Lansing and Robert
Yates, while the third member of the New York delegation, Alexander Hamilton, was one of the most
outspoken nationalists near the other extreme. And despite Franklin’s relatively strong left position and
Ingersoll’s centrist ideology, these two delegates would have been routinely overruled by the other six
30 Here, and in the rest of this section, we apply various extensions of the Median Voter Theorem (Black 1948). According to the
Median Voter Theorem if all issues can be represented on a single–dimension, voting is pair–wise under majority rule, and preferences
are single–peaked, then the unique equilibrium outcome will be at the spatial position of the median voter. Moreover, a corollary
shows that if voter utility functions are also symmetric around the single peak, then any alternative closer to the median voter can
defeat any alternative farther away.
31 If we used instead the mean measurement error, then the state median positions for New Jersey, Delaware, Virginia, and Pennsylvania would each move slightly away from center (New Jersey to the left and the others more right), with Virginia’s median delegate
position falling between the two median delegates from Massachusetts. However, this would not appreciably affect the expected
outcomes under full attendance described below, because the median states (GA and SC) would be unaffected.
32 We also considered alternative specifications for predicting the positions of the thirteen missing delegates. However, this rarely
changed the identity or position of the median delegate in a state, and when it did never moved the new state median to the other side
of the median state’s position so our interpretation of potential floor equilbria remained fixed. Yet due to some potential movement of
the other states, Figure 3 is meant to be illustrative, not definitive.
19
Pennsylvania delegates who were much further to the right. Thus our mapping is consistent with historical
understandings of both delegate and state positions simultaneously, giving us even greater confidence in
our estimates.
An implication of delegates voting in state blocs by majority rule indicates the median delegate of state
j would pivot state j and the median of the state medians would pivot the floor.33 In the case of Figure
3, Baldwin (GA) would be the left floor pivot and John Rutledge (SC) would be the right floor pivot. All
points in between would be in equilibrium.
Contrast this with what would occur if votes were tallied by individuals rather than by states. If votes
were tallied by individuals and all 55 delegates were present at the same time (which never occurred),
Jared Ingersoll (PA) would be the median delegate, even though Davie (NC) is closer to the center of the
scale. This suggests that if delegates voted as independent individuals, the clauses in the Constitution
might gravitate toward Ingersoll’s position, rather than the position of Baldwin (GA) and Rutledge (SC).
The result would be a more localist document than that expected from bloc voting.
34
Following Dougherty and Heckelman (2006), we can also break down bloc voting at the convention into
three periods based on the attendance of delegates from New York and New Hampshire. From May 30–July
10, delegates from New York attended the convention but New Hampshire did not. During this period, our
estimates suggest Georgia was generally the median state with Baldwin and Few holding nearly identical
co–median positions. Growing discouraged with the proceedings, both John Lansing and Yates left the
convention after the July 10 meeting, leaving New York with less than the minimum number of delegates
required by their state. From July 11 to July 22 only ten states were represented, making South Carolina a
co–median with Georgia, with Few (GA) becoming the left pivot and Charles Cotesworth Pinckney (SC) the
new right pivot. This subtle shift allowed for winning proposals to become slightly more nationalistic. On
July 23, Nicolas Gilman and John Langdon of New Hampshire arrived for their first vote. Because they were
33 A delegate is pivotal on the floor if he can change the assembly’s choice by changing his vote in the event of a close vote. In single
dimensional voting with single peaked and symmetric preferences, the pivotal voter will always be in the winning coalition. Any
status quo will be an equilibrium when it is not to the left or right of all pivotal voters. Moreover, for any out-of-equilibrium status
quo, the set of alternatives preferred by the pivot nearest to the status quo is one-to-one to the set of alternatives that can pass under
bloc voting.
34 The argument should not be pushed too far. The overall median is highly dependent on attendance for any particular vote, and
the position of the identified overall median sometimes fluctuated radically. Furthermore, attendance may have been endogenous to
the voting rule in place. It is possible that other elected delegates who never attended may have felt more obligation to attend if it had
been decided that voting would not be based on state blocs. Depending on their positions, the floor median might have been far from
Ingersoll.
20
positioned to the right of both floor pivots, South Carolina solely controlled the floor pivots with Charles
Cotesworth Pinckney and John Rutledge in the left and right pivotal positions. The convention would have
become even slightly more nationalistic but probably not by a substantial margin. These descriptions are
summarized in Table 3.
However, there were always at least some absences on any given roll call vote, which could alter the
position of a state’s median and the floor median. Fore example, Houston (NJ) and George Wythe (VA) permanently returned home after the first week. Their absences made Brearly, William Paterson, and William
Livingston co–medians for New Jersey and John Blair the sole median for Virginia. These changes would
move New Jersey further left and Virginia further right, but keep them on opposite sides of Georgia and
South Carolina. Thus, their absences by themselves would not alter the equilibrium.
A cursory view of attendance records suggests that the floor medians were relatively stable, with either
a delegate from Georgia or South Carolina typically representing the median state(s). The four delegates
from South Carolina never missed a day, and with the exception of William Pierce, each of the Georgia
delegates held very similar positions. If the Articles of Confederation were a localist document, perhaps
in the spatial vicinity of its author John Dickinson (–0.80), then the final version of the Constitution might
have gained nationalist strength because Georgia and South Carolina favored more centralization. It did not
become too centralized, by adopting something similar to Hamilton’s Plan because the pivotal delegates
from Georgia and South Carolina were more moderate than Hamilton and others on the right extreme.
This might explain why the Constitution was more nationalistic than the Articles of Confederation, but still
moderate among the delegations attending.
9
Conclusion
The Constitutional Convention provides a rich story about nation building. Although every good story is
worth retelling, our intent has not been to retell the story of how the Constitution was created. Instead, we
have brought new data to bear on the age old question of the motivation of the framers at the Constitutional
Convention. Previous studies have tried to address this question by examining delegate positions on a
handful of roll call votes. Since the sample is usually related to finance issues (McGuire and Ohsfeldt 1984,
21
1986), or perhaps votes on issues related to slavery (Dougherty and Heckelman 2008), it is not surprising
that measures of financial interests, or ownership of slaves, are important factors in explaining delegate
behavior in those studies. More comprehensive studies of the Constitutional Convention have relied on the
votes recorded for the states. Those studies generally find voting behavior consistent with state interests,
such as differences between the large states and small states (Jillson and Anderson 1978; Slez and Martin
2007) or regional differences between northern and southern interests (Davis 1977).
It would not be surprising if delegates responded to state interests because delegates were elected by
their legislatures to represent their state. Nevertheless, the notion that delegates were primarily influenced
by state interests cannot be evaluated by solely examining state votes. Variation within each state delegation
would be hopelessly lost if only state level data were examined.
By making an effort to infer delegate votes on all substantive motions at the convention, we have been
able to estimate latent ideological positions for the delegates across the convention. This dimension does
not show delegates clumped together based on the securities held within their state nor does it show one of
the geographic regions grouped on an extreme. Instead, latent interests at the convention appear to reflect
differences between localists and nationalists. These differences often varied within the same state. Prototypical nationalists, such as Alexander Hamilton, were near the right extreme, while Hamilton’s only codelegates, John Lansing and Robert Yates, were estimated near the left extreme. Delegates from Delaware
ranged from John Dickinson (-.80) on the left to George Read (1.0) on the right, and other states such as Virginia and Pennsylvania were less unified than many may have believed. As McGuire (2003: 50–51) states,
“the major question at Philadelphia was whether there would be a strengthened, more powerful central
government with state sovereignty or a strong, nationally oriented central government with little or no
state sovereignty.” Finding that the issue of centralization was the primary dimension at the convention is
consistent with the traditional accounts of Jensen (1950) and Ferguson (1969), who thought the Convention
was pushed by people who favored stronger national government. It also provides evidence against the
claims of Crosskey and Jeffrey (1981), Davis (1977), and Marks (1973), who claim the primary motivation
for the new Constitution was to regulate commerce, address sectional conflicts, or allay problems of foreign
affairs, respectively.
However, there is more to our story than different tastes for centralization. Other factors covaried with
22
the localist-nationalist dimension. With other factors controlled, the type of political experience (judicial,
executive, or legislative), ownership of slaves and/or private bank securities, and the extent of home state
compliance with federal requisitions seem to be major determinants of the spatial positions of the delegates.
Legislative experience and greater compliance with federal requisitions by a delegate’s state tended to
move a delegate toward a nationalist position, perhaps because these factors gave them greater faith in
federal decision making, centralization, and a document that would virtually eliminate state free–riding.
Owning slaves and bank securities had a similar effect, but for different reasons. Greater experience in
the judiciary and executive tended to give delegates a more localist view. The latter qualities may reflect
stronger interests in state sovereignty, free from federal interference.
Again, most of these factors vary by delegate, not by state. The population of each state, state slave
holdings, debt held within each state, and to a lesser extent region had limited or no affect. Finding that
delegates voted primarily according to delegate level factors raises questions about the principle-agent relationships they maintained with their legislatures and whether delegates freely pursued their own interests
when voting on clauses or acted on their best judgment of principle.
As an ancillary result, we have also shown that delegates from Georgia and South Carolina held median
positions across the convention. Because the median delegates from South Carolina were to the right of the
median delegates from Georgia, the departure of New York and later arrival of New Hampshire pushed
the convention further right over time – toward greater nationalism. This might explain why a number of
the more nationalistic clauses in the Constitution, such as the national supremacy clause and the necessary
and proper clause, took hold late in the convention. Because Abraham Baldwin, William Few, Charles
Cotesworth Pinckney and John Rutledge were floor medians from Georgia and South Carolina respectively,
perhaps greater emphasis should be placed on their role than the traditional literature suggests. The final
form of the Constitution may reflect the preferences of these delegates more than it reflects the preferences
of someone like James Madison, who was allegedly the framer of the Constitution but too extreme to control
its creation.
23
Appendix: Data Sources for Independent Variables
Age, years in politics, and years in specific offices were gathered from three sources: the American
National Biography, Oxford University Press, 2000; “America’s Founding Fathers: Delegates to the
Constitutional Convention,” National Archives
<http://www.archives.gov/exhibits/charters/constitution_founding_fathers.html>; and the
“Biographical Directory of the United States Congress” <http://bioguide.congress.gov/>.
Anti–Federalists came from Riker (1982).
Religion came from Bradford (1982).
Public and private securities owned by the delegates, as well as their slave holdings, were made available
by McGuire. See McGuire (2003).
Debt per capita came from the American State Papers, Volume 1 (Finance): 231.
State compliance with requisitions for money (1784–1789) came from Dougherty (2001, 95).
Population figures came from the Department of Commerce and Labor (1909) A Century of Population
Growth from the First Census of the United States to the Twelfth, 1790–1900. We use the state boundaries
that existed at the time of the convention. Hence, Maine is included with Massachusetts, Kentucky is
included with Virginia, and Tennessee (also known as the Southwest Territory) is included with
North Carolina.
The Mid–Atlantic region includes New York, New Jersey, Pennsylvania, Delaware, and Maryland. States
north and south of this region were coded as New England and the South, respectively.
24
References
Barrow, Clyde W. 2000. More Than a Historian: The Political and Economic Thought of Charles A. Beard.
New Brunswick: Transaction Publishers.
Beard, Charles A. [1913] 2004. An Economic Interpretation of the Constitution of the United States. New
York: Dover Publications.
Beeman, Richard. 2009. Plain, Honest Men. New York: Random House.
Black, Duncan. 1948. On the Rationale of Group Decision–making. Journal of Political Economy 56(1):
23–34.
Bowen, Catherine Drinker. 1966. Miracle At Philadelphia: The Story of the Constitutional Convention.
Boston: Little Brown and Company.
Bradford, M. E. 1982. A Worthy Company: Brief Lives of the Framers of the United States Constitution.
Marlborough, New Hampshire: Plymouth Rock Foundation.
Collier, Christopher. 1971. Roger Sherman’s Connecticut, Yankee Politics and the American Revolution
(1st ed.). Middletown: Wesleyan University Press.
Congleton, Roger. 2011. Perfecting Parliament: Constitutional Reform, Liberalism, and the Rise of Western
Democracy. New York: Cambridge University Press.
Crosskey, William Winston, and William Jeffrey, Jr. 1981. Politics and the Constitution in the History of the
United States, vol. 3: The Political Background of a Federal Convention. Chicago: University of
Chicago Press.
Davis, Joseph. 1977. Sectionalism in American Politics, 1774-1787. Madison: the University of Wisconsin
Press.
Dougherty, Keith L. 2001. Collective Action under the Articles of Confederation. New York: Cambridge
University Press.
Dougherty, Keith L. and Jac C. Heckelman. 2006. “A Pivotal Voter from a Pivotal State: Roger Sherman at
the Constitutional Convention,” American Political Science Review 100(2): 297–302.
Dougherty, Keith L. and Jac C. Heckelman. 2008. “Voting on Slavery at the Constitutional Convention,”
Public Choice 136(3–4): 293–313.
25
Dougherty, Keith L., Jac C. Heckelman, Paul Carlsen, and David A. Gelman. 2012. “A New Dataset of
Delegate Positions on 620 Substantive Roll Calls at the U.S. Constitutional Convention,” CCRG
Working Paper.
DuBois, W. E. B. [1896] 1969. The Suppression of the African Slave–trade to the United States of America,
1638–1870. Baton Rouge: Louisiana State University Press.
Elezar, Daniel J. 1972. American Federalism: A View from the States (2nd ed.). New York: Crowell.
Elkins, Stanley and Eric McKitrick. 1961. “The Founding Fathers: Young Men of the Revolution,” Political
Science Quarterly 76(2): 181–216.
Farrand, Max, ed. 1936-1987. The Records of the Federal Convention of 1787. 4 vols. New Haven: Yale
University Press.
Ferguson, E. James. 1969. “The Nationalists of 1781-1783 and the Economic Interpretation of the
Constitution,” The Journal of American History 56(2): 241–261.
Gibson, Alan. 2007. Understanding the Founding: The Crucial Questions. Lawrence, KS: University Press
of Kansas.
Goff, Brian L. and Kevn B. Grier. 1993. “On the (Mis)measurement of Legislator Ideology and Shirking,”
Public Choice 76(1–2): 5–20.
Greene, William H. 2007. Econometric Analysis (7th ed.). Upper Saddle River, NJ: Prentice Hall.
Griffen, John D. and Brian Newman. 2007. “The Unequal Representation of Latinos and Whites,” Journal
of Politics 69(4): 1032–46.
Grubb, Farley W. 2003. “Creating the U.S. Dollar Currency Union, 1748–1811: A Quest for Monetary
Stability or a Usurpation of State Sovereignty for Personal Gain?” American Economic Review 93(5):
1778–1798.
Grubb, Farley W. 2006. “The U.S. Constitution and Monetary Powers: An Analysis of the 1787
Constitutional Convention and How a Constitutional Transformation of the Nation’s Monetary
System Emerge,” Financial History Review 13(1):4 3–71.
Heckelman, Jac C. and Keith L. Dougherty. 2007. “An Economic Interpretation of the Constitutional
Convention of 1787 Revisited,” Journal of Economic History 67(4): 829–48.
Heckelman, Jac C. and Keith L. Dougherty. 2010. “Personalty Interests at the Constitutional Convention:
26
New Tests of the Beard Thesis,” Cliometrica 4(2): 207–228.
Hutson, James H. 1987a. “Riddles of the Federal Constitutional Convention,” William and Mary Quarterly
44(3): 441–423.
Hutson, James H. ed. 1987b. Supplement to Max Farrand’s The Records of the Federal Convention of 1787.
New Haven: Yale University Press.
Jillson, Calvin. 1981. “Constitution-Making: Alignment and Realignment in the Federal Convention of
1787,” American Political Science Review 75(3): 598–612.
Jillson, Calvin and Thorton Anderson. 1978. “Voting Bloc Analysis in the Constitutional Convention:
Implications for an Interpretation of the Connecticut Compromise,” Western Political Quarterly 31(4):
535–547.
Jensen, Merrill. 1950. The New Nation: A History of the United States during the Confederation,
1781-1789. New York: Knopf.
Jung, G-Ryong, Lawrence W. Kenny, and John R. Lott, Jr. 1994. “An Explanation for Why Senators from
the Same State Vote Differently So Frequently,” Journal of Public Economics 54(1):65–96.
Kaminski, John P., Ed. 1995. A necessary evil? Slavery and the debate over the Constitution. Madison,
Wisconsin: Madison House.
Lewis, Jeffrey B. and Drew A. Linzer. 2005. “Estimating Regression Models in Which the Dependent
Variable Is Based on Estimates,” Political Analysis 13: 345–364.
Londregan, John. 1999. “Deliberation and Voting at the Federal Convention of 1787,” University of
California, Los Angeles (mimeo).
Maddala, G. S. 1983. Limited-dependent and Qualitative Variables in Econometrics. Cambridge:
Cambridge University Press.
Main, Jackson Turner. 1973. Political Parties before the Constitution. Chapel Hill: University of North
Carolina Press.
Marks, Frederick W. 1973. Independence on Trial: Foreign Affairs and the Making of the Constitution.
Baton Rouge: Louisiana State University Press.
McDonald, Forrest. 1958. We the People: The Economic Origins of the Constitution. Chicago: University
of Chicago Press.
27
McGuire, Robert. 2003. To Form a More Perfect Union. New York: Oxford University Press.
McGuire, Robert and Robert L. Ohsfeldt. 1984. “Economic Interests and the American Constitution: a
Quantitative Rehabilitation of Charles A. Beard,” Journal of Economic History 44: 509–519.
McGuire, Robert and Robert L. Ohsfeldt. 1986. “An Economic Model of Voting Behavior over Specific
Issues at the Constitutional Convention of 1787,” Journal of Economic History 46: 79–111.
Monroe, Nathan W., Jason M. Roberts, and David W. Rohde. 2008. “Introduction: Assessing the Impact of
Parties in the U.S. Senate,” in Nathan W. Monroe, Jason M. Roberts, and David W. Rohde (eds.) Why
Not Parties? Party Effects in the United States Senate. Chicago: University of Chicago Press, pp. 1–19.
Morgan, Edmund. 1993. The Birth of the Republic, 1763-89. Chicago: University Of Chicago Press.
Peltzman, Sam. 1984. “Constituent Interest and Congressional Voting,” Journal of Law and Economics
27(1): 181–210.
Poole, Keith. 2000. “Non-parametric Unfolding of Binary Choice Data,” Political Analysis 8(3): 211–37.
Poole, Keith and Howard Rosenthal. 1997. Congress: A Political-Economic History of Roll Call Voting.
New York: Oxford University Press.
Riker, William. 1987. “The Lessons of 1787,” Public Choice 55(1–2): 5–34.
Slez, Adam and John L. Martin. 2007. “Political Action and Party Formation in the United States
Constitutional Convention,” American Sociological Review 72: 42–67.
Wittman, Donald A. (1995) The Myth of Democratic Failure: Why Political Institutions are Efficient.
University of Chicago Press, Chicago
Wooldridge, Jeffrey M. 2004. Econometric Analysis of Cross Section and Panel Data. Cambridge, MA: MIT
Press.
28
Table 1: Descriptive Statistics
mean std dev min
w–nominate score (dependent variable)
0.02
0.7
–1
max
1
Primary explanatory variables
Anti–Federalist (dummy)
previous years in politics
previous years in federal
previous years in state/local
previous years in Continental Congress
previous years in executive
previous years in legislature
previous years in judicial
slave owner (dummy)
private securities (dummy)
public securities (dummy)
total population (thousands)
state compliance for requisitions (ratio)
mid–Atlantic region (dummy)
southern region (dummy)
0.1
12.1
3.8
10.4
2.8
3.2
8.0
4.3
0.4
0.2
0.6
341.8
26.9
0.4
0.3
0.3
10.0
5.3
9.4
2.4
7.7
6.6
6.8
0.5
0.4
0.5
217.3
20.0
0.5
0.5
0
1
0
0
0
0
0
0
0
0
0
59.1
0
0
0
1
51
33
37
9
48
30
28
1
1
1
821.2
64
1
1
Additional explanatory variables
slaves in state (percentage of total population)
total debt in state (per capita)
loan office debt (per white male)
white population (thousands)
age
Puritan religion (dummy)
communitarian religion (dummy)
hierarchical religion (dummy)
19.1
7.5
5.6
262.0
43.1
0.4
0.3
0.6
16.3
4.3
4.8
156.8
12.2
0.5
0.4
0.5
0
1.8
0.7
46.3
26
0
0
0
43
15.0
18.1
503.2
81
1
1
1
29
Table 2: Determinants of Delegate Ideology
(I)
(II)
(III)
constant
0.053
0.90
–0.434
(0.17)
(0.29)
(–1.30)
Anti–Federalist (dummy)
–1.612***
–1.606***
–1.689***
(–4.44)
(–4.54)
(–4.90)
previous years in politics
–0.014
(–1.38)
previous years in federal
0.012
(0.60)
previous years in state/local
–0.021*
(–1.91)
previous years in legislature
0.041***
(2.12)
previous years in executive
–0.033**
(–2.08)
previous years in judicial
–0.041***
(–2.79)
slave owner (dummy)
0.553*
0.624**
0.613**
(1.79)
(2.03)
(2.14)
private securities (dummy)
0.433*
0.396
0.540**
(1.60)
(1.49)
(2.13)
public securities (dummy)
0.087
0.080
0.180
(0.44)
(0.42)
(0.96)
total population (thousands)
–0.001
–0.001
–0.0005
(–1.48)
(–1.55)
(–0.72)
state compliance for requisitions (ratio)
0.025***
0.025***
0.021***
(3.28)
(3.34)
(2.83)
mid–Atlantic region (dummy)
–0.727**
–0.765***
–0.396
(–2.52)
(–2.71)
(–1.35)
south region (dummy)
–0.144
–0.197
–0.150
(–0.43)
(–0.60)
(–0.49)
Log likelihood
–35.92
–35.11
–32.30
Notes: N = 42. Estimates are derived from heteroskedastic double–censored Tobit regression with data
censored at –1 and 1. Associated z–statistics are in parentheses.
* significant at < .10, ** significant at < .05, *** significant at < .01
Table 3: Pivotal Delegates
# of
Left Pivot
Right Pivot
Dates
Roll Calls Delegate
Position Delegate
Position
May 30 – July 10
163 Baldwin (GA)
0.273
Few (GA)
0.274
July 11 – July 22
88 Few (GA)
0.274
CC Pickney (SC)
0.366
July 23 – Sept 17
544 CC Pickney (SC)
0.366
Rutledge (SC)
0.393
Notes. Pivots were identified assuming full attendance by all delegates. All positions between the left and
right pivots (inclusive) are in equilibrium.
30
Figure 1: Delegate Nominate Scores and Bootstrapped Standard Errors
31
Note: * anti-Federalist
Figure 2: Predicted Nominate Scores
32
Notes: Positions for delegates marked by * are conditional expected values from double-censored Tobit model: Table 2, column
(III) specification. See Equation (6) for the formula. Positions for all other delegates are determined by w–nominate.
Figure 3: Median Delegates from each State
Note: The figure depicts the median members of each state assuming full attendance. Medians are determined by delegate nominate scores, or predicted nominate scores using the maximum measurement error
from Figure 2.
33