Slavery in the Constitution: Why the Lower South Occasionally Succeeded at the Constitutional Convention

Abstract

The Lower South’s successes at the Constitutional Convention of 1787 contributed to a constitution that prohibited federal interference with the slave trade until 1808, guaranteed fugitive slaves would be returned to their masters, and prevented export tariffs. We explain why the Lower South occasionally succeeded on sectional issues at the convention using a multiple-dimensional model of sincere voting, estimated using a new dataset of delegate votes, multiple imputation, and optimal classification. We argue that mixing sectional issues with powers of the federal government made the Lower South more mainstream and helped it gain support from various Northern delegations. We test this relationship using regression analysis and apply it to two substantive issues where the Lower South succeeded. The result is a largely new account of how slavery became encoded in the Constitution.

Keywords

U.S. Constitutional Convention Lower South slavery

Introduction

In 1786, tensions between the North and South came to a head when Secretary of Foreign Affairs John Jay asked Congress to forgo navigation of the Mississippi River for a period of twenty-five to thirty years to close a deal with Spain that protected Northern fishing. Jay had been instructed to secure both objectives, but he requested to forgo the Southern part of the deal to attain the Northern part. After a bitter debate, Congress repealed Jay’s earlier instructions to protect the Mississippi River by a vote of seven states to five. All seven Northern states voted in favor of the proposal and all five Southern states, Maryland southward, voted against.¹ The vote illustrated why the South could not win purely sectional votes in the Congress of the Confederation.

Jillson and Wilson’s (1994) multidimensional scaling of the Congress of the Confederation illustrates the polarization in Congress at the time. Sectional issues were so dominant that Congress divided itself into two disjoint clusters in 1786, a Northern cluster and a Southern one, along a sectional dimension of voting.

Shortly thereafter, the Lower South, South Carolina and Georgia,² was surprisingly successful at the Constitutional Convention. The convention agreed to prohibit federal interference with the slave trade until 1808, to guarantee fugitive slaves would be returned to their masters, and to prevent export tariffs.³ It also provided a louder voice for Southerners by counting three-fifths of slaves in the apportionment of the House of Representatives. The latter agitated Northerners and led extreme New England Federalists, such as Timothy Pickering and William Plummer, to propose New England’s secession from the union.

Counting Delaware northward as the North, Northern states outnumbered Southern states eight to five in the confederation, making it particularly surprising that the Lower South partly succeeded at the convention.⁴ Why does the Constitution have so many clauses that seem to protect slavery? What caused the Lower South to succeed on some votes at the convention and to fail on others?

The standard historical account is that these clauses resulted from vote trades (Lynd 1966; McDonald 1965; Rossiter 1966; Stewart 2008; Van Cleve 2010). For the purposes of this paper, it is important to differentiate a vote trade from a compromise. A Vote Trade occurs if one person votes for (or against) another person’s issue in exchange for the other person’s vote for (or against) their issue. For example, delegates from the Lower South allegedly traded their demand for a two-thirds majority requirement for navigation acts in exchange for New England’s support for a clause protecting the slave trade until 1808. A compromise settles a dispute by mutual concessions. For example, Southerners may have wanted slaves counted the same as whites in the apportionment of the legislature and Northerners may have wanted slaves wholly excluded in the apportionment. Agreeing to count them as three-fifths of whites is a potential compromise. Most historians claim that delegates engaged in vote trades as well as compromises. Our position is that much can be explained with compromises alone. We argue that the Lower South succeeded on sectional issues when a clause appealed to delegate preferences for federalism. Because federalism was not an issue in the policy votes of the Congress of the Confederation but it was an issue in the constitutional votes of the convention, the Lower South occasionally succeeded at the convention while it consistently failed in Congress.

Our analysis follows the spirit of Jenkins and Sala (1998) and Jenkins and Morris (2006), who argue that sincere behavior⁵ can help explain historical events.

We proceed by introducing institutional features of the Constitutional Convention and the literature on why the Lower South occasionally succeeded. We then develop a theory of sincere voting by estimating a two-dimensional map of the preferences of delegates at the convention using a relatively new dataset of delegate votes, multiple imputation, and optimal classification. To the best of our knowledge, this paper is the first to use these data to provide a multiple-dimensional scaling of the convention, which we find reflects preferences for federalism and sectional issues.

To test whether the specific mix of federalism and sectionalism affected the Lower South’s successes, we then run two regressions. The first analyzes the relationship between the Lower South winning and the trajectory of the vote.⁶ It suggests the Lower South was more likely to be on the winning side of a vote if the vote was on an angle that protected states’ rights or the large state coalition. The Lower South often failed as it did in the Congress of the Confederation if the vote purely appealed to the sectional dimension. The second regression tests the relationship between the Lower South’s success and its centrality on the issue. It suggests the Lower South’s delegates were more likely to win when the mix of federalism and sectionalism made them closer to the center of a vote than when the mix made them farther away.

We then examine two case studies that show how our theory explains the Lower South’s failures and successes on votes related to apportionment and international trade. We conclude that the Constitution’s slavery clauses could have resulted from sincere voting, without the North surrendering their alleged opposition to slavery.

There is a renewed interest in early American politics (Clinton and Meirowitz 2004; Dougherty and Heckelman 2006; Jillson 2008; Pope and Treier 2011). In addition to helping us understand how the Constitution acquired some of its most detestable clauses, this paper provides a framework for understanding why numerical minorities occasionally succeed.

Institutional Rules

Voting in the Congress of the Confederation (1775–1789) and the Constitutional Convention (1787) was conducted by state blocs, with each state delegation casting one vote. The size of each delegation varied from two to eight members depending upon the number of delegates appointed by each state. The yea or nay position of each state was determined by a majority of its delegates. In the event of a tie, the state’s vote was recorded as divided.

Unlike bloc voting in the Congress of the Confederation, a motion passed at the Constitutional Convention if more state delegations voted yea than nay and a quorum of seven delegations were present. In the Congress of the Confederation, minor issues passed with seven affirmative votes (a majority of the thirteen states in the union) and major issues passed with nine affirmative votes (three-fourths of the states in the union). There were no rights of first recognition or other mechanisms that would allow a single delegate, or group of delegates, to control the agenda. Any delegate could propose a motion or propose to table a motion at any time.

With decentralized rules such as these and the North consistently holding a majority of state votes in both assemblies, it is surprising that the Lower South succeeded on sectional issues.

Literature

With the Lower South accomplishing little in the Congress of the Confederation and surprisingly more at the convention, some scholars have concluded that there must have been a “corrupt bargain” (McDonald 1965, 182). Otherwise, Northern states would have asserted their numerical supremacy and rid the Constitution of slavery clauses. The vote trade most commonly described in the literature is one in which New England delegates, particularly those from Connecticut, voted for a twenty-year extension of the slave trade, vote 368, in exchange for the Lower South’s opposition to a clause requiring a two-thirds majority in both houses of Congress to pass navigation acts, vote 399 (McDonald 1965; Rossiter 1966; Stewart 2008; Van Cleve 2010).⁷

Perhaps the greatest evidence in favor of such a trade comes from James Madison himself. Madison wrote in a footnote to vote 399 that “An understanding on the two subjects of navigation and slavery, had taken place between those parts of the Union, which explains the vote on the Motion depending, as well as the language of Gen Pinckney & others” (Farrand 1966, 449). Charles Cotesworth Pinckney (SC), a.k.a. General Pinckney, spoke against the two-thirds requirement arguing that the South had an interest in being united with the “strong Eastern States” and thought it was “proper that no fetters should be imposed on the power of making commercial regulations” (Farrand 1966, 449–50). Other delegates claimed there must have been a trade as well.

Although several contemporary scholars still believe that there was a vote trade (Klarman 2016; Stewart 2008; Van Cleve 2010), others have been more skeptical (Fehrenbacher 2001; Hutson 1987). For example, Fehrenbacher (2001) finds it strange that once the convention secured the slave trade clause, they did not dispense with the other part of the deal for another four days. When they did, “only the South Carolina delegation acted as though it were participating in an intersectional bargain on commerce and the slave trade” (Fehrenbacher 2001, 35). He notes that if we were to take the New England delegates at their word, then they would have voted for the clause to hold the union together—that is, to compromise, not to trade. We further note that seven states voted against the two-thirds requirement while four states voted for it, implying that a vote trade was not necessary for the North to carry its position on both issues. We conduct an empirical test for a trade of these votes in the online supplement. Although we find no evidence for the trade, our test is not dispositive. We merely introduce it to pave the way for a story of sincere voting that is closer to Fehrenbacher’s (2001) argument than others.

A Spatial Map of the Convention

Our explanation for why slavery clauses made it into the Constitution is that sectional issues mixed with federal issues at the convention, changing the coalitions supporting sectional clauses.

To explain the Lower South’s successes, we need a map of delegate preferences at the convention, which we create using a 55 by 397 matrix of delegate votes (Dougherty and Heckelman 2012).⁸ The vote matrix includes 4,096 yea or nay positions across all substantive roll calls at the convention, where the position of at least one delegate could be inferred on both sides of the issue.

Data on delegate votes come from Dougherty and Heckelman (2012). Because delegates wanted candor, the convention journal and Madison’s notes recorded the vote of the state blocs, but rarely recorded the votes of individual delegates. Hence, Dougherty and Heckelman (2012) inferred delegate votes using a three-step process. First, by the rules of the convention, the position of each state (yea, nay, or divided) was determined by a simple majority of the state’s delegates. Hence, if there were only two delegates attending from a state and the state voted yea (resp. nay), then Dougherty and Heckelman coded both delegates yea (resp. nay). Second, additional delegate positions were inferred using statements made by delegates in the notes of Madison, Robert Yates, and others (Farrand 1966). Personal manuscripts and secondary sources were used at this stage only if the statement could be tied to a particular roll call. For example, Luther Martin (MD) was coded as voting no on vote 368, because he explicitly said he voted against the clause in a letter to his general assembly. He also made statements consistent with that coding during the convention’s debates (Farrand 1966, 364, 211–12). Third, after the positions of the delegates were recovered, attendance records were consulted to determine whether additional delegate votes could be inferred from the majority rule requirement for each delegation. For example, Maryland was recorded as a yea on vote 368. Because John Mercer was absent and Luther Martin was coded as nay, the three remaining Maryland delegates, Daniel Carroll, Daniel of St. Thomas Jenifer, and James McHenry, must have voted yea in order for a majority of the Maryland delegation to vote yea.⁹

We use these inferred votes and optimal classification (Poole 2000 and 2005) to map delegate positions.¹⁰ Optimal classification is a nonparametric, unfolding procedure that derives an optimal cut line¹¹ for each vote, then optimally classifies voters in the regions formed by the cut lines. The process is then iterated until the number of classification errors are minimized—that is, it minimizes the number of times a delegate votes yea (resp. nay) but his ideal point is on the nay side (resp. yea side) of a cut line. The resultant scaling places those who voted similarly more closely together than those who voted dissimilarly. Unlike interest group scores, such as Americans for Democratic Action (ADA) scores, there is nothing in the procedure that defines the recovered dimensions ex ante. Instead, the substantive content of each dimension must be interpreted ex post.¹²

To prevent delegates with only a few votes from affecting the location of the cut lines, we excluded delegates with 10 or fewer inferred votes from the optimal classification routine. Of the 55 delegates at the convention, 42 had more than 10 inferred votes. Among them, five had between 12 and 20 votes inferred and five others had more than 200 votes inferred. Unlike the Aldrich-McKelvey (1977) procedure, optimal classification can handle large amounts of missing data. Poole’s (2000) simulations show that with 70 percent missing data, optimal classification can place legislators in positions that correctly classify 86 percent of the votes in two dimensions.

With many missing observations in the roll call matrix, a scree plot of the double-centered agreement score matrix does not help us determine the appropriate number of dimensions (Poole 2005, 151). Instead, we determine the appropriate number of dimensions using the percentage of votes correctly classified and the aggregate proportional reduction in error (APRE). We conclude that two dimensions adequately scale the votes. Two dimensions correctly classify 92.6 percent of the 4,096 choices. That is, it correctly puts the delegates on the yea or nay side of a cut line for 92.6 percent of their inferred yea or nay votes. One dimension correctly classifies 84.9 percent of the votes while three dimensions correctly classify 96.8 percent (little more than two).¹³

The estimated locations of the forty-two delegates are depicted by the solid markers in Figure 1. Gray triangles indicate Southern delegates while blue circles indicate Northern ones. One insight of our scaling is that it provides a glimpse at the latent issues of the convention. Scholars have hypothesized a variety of dimensions for voting at the Constitutional Convention including issues related to apportionment, localism-nationalism, and separation of powers (Jillson 2008; Pope and Treier 2012). Any combination of these dimensions, or another, could be recovered.

Figure 1.

Delegates at the Constitutional Convention.

The first dimension of our scaling depicts the classic dichotomy over federalism between those who wanted a stronger national government, on the right, and those who wanted to protect states’ rights, on the left. To evaluate our claim, we used three sources that identify delegates as localists or nationalists. Jillson and Wilson (1994) and Henderson (1974) label politicians localists or nationalists based on voting behavior in the Congress of the Confederation 1781–1783. Main (1973) labels politicians localists or cosmopolitan based on the votes they cast in their state legislatures, 1780–1788. The three sets of authors label twelve of the delegates exactly the same way. The locations of these twelve delegates are consistent with a localist–nationalist interpretation of the scale. The three delegates identified as localists (Gerry, Luther Martin, and Mercer) are all to the left of the nine delegates identified as nationalists (Carroll, Hugh Williamson, William Few, Nathaniel Gorham, Gouverneur Morris, Hamilton, Charles Cotesworth Pinckney, Madison, and James Wilson). Furthermore, our first dimension correlates with Heckelman and Dougherty’s (2013) single dimensional scaling at 0.89, which they describe as measuring localism–nationalism for similar reasons. Delegates who were known for their localist stances at the Constitutional Convention, such as John Lansing, Luther Martin, and Robert Yates, are near the left extreme, while some of the convention’s most ardent nationalists, Hamilton, Madison, and Wilson, are toward the right extreme. Because other issues correlate with this dimension, the label should not be interpreted as measuring differences solely over centralization.¹⁴

The second dimension reflects sectional issues, with the North at the top of the figure and the South at the bottom. The pattern is easily discernible from the blue circles for Northern delegates and the gray triangles for Southern delegates. The description of this dimension is reinforced by examining sectional unity votes. A sectional unity vote occurs whenever a majority of Southern states vote against a majority of Northern states. Focusing on those votes alone, we created a sectional unity score for each delegate, which is the percentage of times the delegate votes with his section (i.e., North or South) on a sectional unity vote. The twelve delegates with the highest sectional unity scores are all placed on the correct side of the x axis in Figure 1—Northern delegates above and Southern delegates below.¹⁵ Furthermore, twenty-three of the twenty-nine voters with at least ten sectional unity votes were on the correct side of the x axis. This includes Nicholas Gilman, John Langdon, Sherman, and Carroll who were on the wrong side but close. As nothing prohibits delegates from overlapping near the center, it is fairly clear that the second dimension picks up a sectional divide.¹⁶

Further evidence that the dimensions capture differences over federalism and sectionalism are provided in the online supplement. Our method for placing the thirteen excluded delegates in the space (indicated by hollow markers in Figure 1) is also described in the online supplement. As a robustness check, we re-conducted our analysis without the excluded delegates and attained similar results.

Theory: Mixing Sectionalism with Federalism

A quick glance at Figure 1 suggests that delegates like William Davie (NC) and Carroll (MD) are near the center of the space. This makes them mainstream on almost any vote. Other delegates like John Dickinson (DE) and Charles Cotesworth Pinckney (SC) are not near the center of the space, but they could be near the center of a vote depending on the vote’s trajectory. In this sense, they are mainstream or extreme, depending on the mix of the two dimensions.

Our theory is that the Lower South was more likely to succeed when the vote was on a trajectory that made it more mainstream—sectional issues that protected states rights or advanced the large state coalition. To establish this we need a measure of centrality that varies based on attendance and the trajectory of the vote. We will use it to determine the mix of sectionalism and federalism that made the Lower South most mainstream, then show that these trajectories made the Lower South more successful in a regression analysis.

Bloc Median Lines

Because delegates voted in state blocs, we measure centrality as the distance to the vote’s bloc median line. For any status quo q and proposal p, the bloc median line (1) is perpendicular to line segment $\bar{q p}$ and (2) demarcates the space such that no more than a majority of the state delegations are on either side. Davis et al. (1972) and Godfrey, Grofman, and Feld (2011) describe the idea for votes tallied by individuals.

In two dimensions, a bloc median line can be found by projecting ideal points onto any line parallel to $\bar{q p}$ (e.g., the normal vector line), calculating the median of each state’s projection, and determining the median of the state medians.¹⁷

Figure 2 provides an example with nine voters divided into three delegations. A, B, and C are delegates from state₁; D, E, and F are delegates from state₂; and G, H, and I are delegates from state₃. Given the $\bar{q p}$ trajectory, B is the median of state₁, F is the median of state₂, and G is the median of state₃. The bloc median line, M, intersects the median of these medians, F. If votes were tallied by individuals, the median line would intersect A.

Figure 2.

A bloc median line for three hypothetical state delegations.

With perfect spatial voting,¹⁸ the delegate(s) on the bloc median line must be on the winning side of the vote. Furthermore, delegates closer to the bloc median line are more likely to be on the winning side than delegates farther away (see the online supplement for explanation).

It is important to keep in mind that the bloc median line depends on the angle of $\bar{q p}$ . A status quo on such a line is typically not in equilibrium because it is vulnerable to proposals made on different trajectories, consistent with McKelvey’s (1976) well-known theorem.

Issue Combinations and Centrality

The next step in our argument is to show that the Lower South tended to be more mainstream (i.e., closer to the bloc median line) when a vote advanced Southern interests while protecting states’ rights or advancing the interests of the large state coalition (Massachusetts, Pennsylvania, and all states south of Maryland).

Figure 3 illustrates two extremes for all fifty-three delegates with the delegates from South Carolina and Georgia marked in red and green, respectively.¹⁹ The two frames contain projections from two different normal vectors (i.e., different $\bar{q p}$ trajectories). If all fifty-three delegates voted and the vote was on a 45^○ line, as depicted in frame A, then the delegates from Georgia and South Carolina would be closer to the bloc median line. They would also be near the center of the voters. Contrast this with frame B. If everyone voted and the vote was on a −45^○ line, then the delegates from Georgia and South Carolina would be farther from the bloc median line and more extreme. This explains how an actor can be mainstream on one vote and extreme on another, given fixed ideal points in two dimensions.

Figure 3.

Trajectory of the vote and the extremism of the lower south: (A) The lower south as centrists and (B) the lower south as extremists.

Votes on a purely sectional trajectory near 90^○ would make delegates from the Lower South extreme. Such votes dominated the Congress of the Confederation, albeit with different delegates, which explains why the Lower South repeatedly failed in that body.

The next step in our argument is to determine the angles which made the Lower South more mainstream in our observed votes—accounting for changes in attendance. Figure 4A shows the relationship between the observed distance of the delegates from South Carolina and Georgia to the bloc median line and the observed angle of the vote. Red markers are for South Carolina and green markers are for Georgia, with a median spline through each to highlight trends. It is clear from this figure that delegates from the Lower South were generally closer to the center of a vote on angles between 0^○ and 50^○, heavily mixing sectionalism with federalism. We might call those angles sweet for the Lower South. Delegates from the Lower South were generally far from the center of a vote on angles less than −50^○ or greater than 80^○, which we might call bitter for the Lower South. The bitter angles were more purely sectional than the sweet ones.

Figure 4.

The centrality of four states and the angle of the vote: (A) Lower south and (B) New Jersey and Delaware.

The middle of the Lower South’s sweet spot is roughly 25^○. On this angle, proposals would either be to the right and slightly up from the status quo or to the left and slightly down from the status quo. The three largest states, Massachusetts, Pennsylvania, and Virginia, would be projected to the right of the Lower South and a few small Northern states, such as New Jersey and Connecticut, would be projected to the left. Moving right along the 25^○ trajectory would advance the interests of the large state coalition. For example, once the House was apportioned by the Three-Fifths Clause, requiring salaries and appropriations to originate in the House advanced the interests of the large state coalition. Moving left along the 25^○ trajectory would advance the interests of the South while protecting state rights. For example, prohibitions on federal export taxes, which New Hampshire and Connecticut supported, protected states rights. This is why we conclude that the Lower South was mainstream when a vote protected state rights or it appealed to the large state coalition.

To show that “sweet angles” are not the same for all delegations, consider a similar measure for New Jersey and Delaware in Figure 4B. New Jersey and Delaware voted together 72 percent of the time. Three lessons can be learned from frame B. First, New Jersey and Delaware were typically closer to the bloc median line than South Carolina and Georgia, as indicated by the smaller y-values in frame B than in frame A. Second, different states were more mainstream on different angles. South Carolina and Georgia were more mainstream for trajectories near 25^○, while New Jersey was more mainstream for votes closer to −90^○ or 90^○, purely sectional angles. Third, not all states had clear sweet spots. For example, Delaware is roughly 0.3 units away from the bloc median line regardless of the angle of the vote. Hence, while many states have sweet spots (i.e., mixtures of federalism and sectionalism that make them mainstream), the location of sweet spots varies by two-state coalition and those coalitions need not have the big swings in centrality that affected the Lower South.

Regression Analysis

The relationship between centrality and success can be tested using two logit regressions, each of which focuses solely on cases where South Carolina and Georgia voted together. The dependent variable for both is a dichotomous variable indicating whether the Lower South was on the winning side of the vote.²⁰ In the first regression, this variable is regressed on the difference between the trajectory of the vote and the middle of the Lower South’s sweet spot, 25^○, with several controls. If our hypothesis is correct, the Lower South is more likely to be on the winning side if it is closer to the middle of its sweet spot. The results of this analysis, with Huber-White robust standard errors, are reported in Table 1, column (1). We describe the results for our control variables first.

Table 1.

The Probability of Success and the Mixture of Issues.

	(1)SC-GA wins	(2)SC-GA wins
Degrees from 25^○	−0.017** (0.007)
SC-GA distance from the bloc median line		−3.616** (1.147)
SC-GA motioned	−2.085** (0.509)	−2.027** (0.491)
Northern state margin	−0.153(0.338)	−0.265(0.362)
Debate length	0.024(0.022)	0.027(0.023)
Sectional unity vote	−1.749** (0.519)	−1.787** (0.536)
│State vote margin│	0.174* (0.071)	0.157* (0.072)
Constant	2.416** (0.782)	3.270** (0.991)
Observations	278	278
Log-likelihood	−92.37	−89.49

Robust standard errors appear in parentheses.

p < .05. **p < .01.

SC-GA motioned is a dummy variable indicating whether the motion was made explicitly by a Lower South delegate.²¹ We include it to control for the possibility that delegates from the Lower South might propose more controversial clauses than delegates from other states, such as slavery protections. The coefficient for this variable is negative and significant, suggesting that everything else equal a Lower South motion was less likely to succeed than a non-Lower South motion, consistent with our hypothesis.

Northern state margin is the number of Northern states attending minus the number of Southern states attending. If voting was sectional, delegates from the Lower South should be less likely to win as this margin increases. Contrary to our expectations, however, the coefficient is negative and insignificant, suggesting that Northern delegates did not vote cohesively against the Lower South.

Debate length measures the length of the debate, coded as the total number of delegates who spoke on an issue between the time the issue was raised and the time a vote was taken and a new proposal raised (coded using Farrand 1966). We include it to control for contentiousness. We expect more contentious issues to require longer debate, making it more difficult for any region to be on the winning side. In our case, the coefficient is positive but insignificant, providing no indication that longer debate affected the region’s success.

Sectional unity is a dummy variable indicating whether a majority of Southern states voted on one side of an issue while a majority of Northern states voted on the other side. Presumably, sectional unity votes (not to be confused with votes along the sectional dimension) are more likely to go against the Lower South, because the North almost always had more delegations on the floor. The coefficient for this variable is negative and significant, suggesting that the Lower South was more likely to lose votes that pitted Northerners against Southerners.

State vote margin is the absolute difference between the number of state yea votes and the number of state nay votes on a roll call. This variable controls for the fact that any state is more likely to be on the winning side of a lopsided vote than a close vote. Unsurprisingly, the effect is positive and significant suggesting that everything else equal a state is more likely to win if the vote margin is large.

Of course the variable of primary interest is degrees from 25^○. The negative and significant coefficient suggests that delegates from the Lower South were more likely to be on the winning side of a vote if the vote was on an angle that mixed federalism and sectionalism in a way that made the Lower South more mainstream. With other independent variables held at their means (or modes for dichotomous variables), moving from a purely sectional vote of 90^○ to the Lower South’s sweet spot of 25^○ increases the probability the Lower South would win by 8.4 percentage points. Considering that the variable ranges from −90^○ (exclusive) to 90^○ (inclusive), this is a modest but noticeable effect.

Table 1, column (2) shows results for a model with degrees from 25^○ replaced by the average distance of the Lower South delegates to the bloc median line. As spatial voting models suggest, the Lower South should be on the winning side more often if it is closer to the middle of a vote.²² Because delegates from the Lower South were closer to the center of a vote on angles closer to 25^○, distance provides an explanation for why the mix of federalism and sectionalism led to Lower South successes.²³ For the reasons described in the previous section, shorter distances reflect more mainstream stances on the issue given the trajectory of the vote, the delegates who attended, and bloc voting.

The coefficient is negative and significant, corroborating our hypothesis. With other variables held at their means, the predicted probability of the Lower South winning would increase by 16 percentage points if distance shrank from its largest value of 0.84 to its mean of 0.79. If all of the Lower South delegates were on the bloc median line, we predict they would win 98 percent of the time.

Combined with observations from the previous section, our results suggest that delegates from the Lower South were more likely to be on the winning side of a Lower South vote when federalism mixed with sectionalism in a way that made the Lower South more mainstream. To convince readers that success of the Lower South was truly related to federalism, we repeat the analysis for three dimensions in the online supplement.

Two Case Studies

Several issues were important to the Lower South, such as maximizing the number of slaves used in the apportionment of the legislature, protecting the slave trade, maintaining control of commercial treaties, and preventing export tariffs. However, the Lower South succeeded on only some of these issues. For example, it tended to be less successful on apportionment than it was on international trade.²⁴

Part of the reason why delegates from the Lower South were more successful on international trade may have been the centrality of their preferences on those votes. On average, delegates from the Lower South were 0.51 units away from the bloc median line on apportionment votes and 0.26 units away from the bloc median lines on international trade votes (i.e., 49% closer on international trade). The closer distance may explain why it was more successful on international trade.

Legislative Apportionment

Although some readers may think that the Three-Fifths Compromise was invented late in the convention to overcome some impasse between the North and South, it was initially proposed twelve days after voting began. When the idea was first raised, the convention was trying to outline an “equitable ratio” of apportionment. John Rutledge and Pierce Butler of South Carolina proposed using the quota of contribution from each state. Perhaps cognizant of how much support their proposal might gain and wanting to establish a different principle, Wilson (PA) quickly interrupted and proposed,

the whole number of white & other free Citizens & inhabitants of every age sex & condition including those bound to servitude for a term of years and three-fifths of all other persons not comprehended in the foregoing description except Indians not paying taxes, in each State (emphasis added, Farrand 1966, 201).

His proposal passed nine states to two, with only New Jersey and Delaware opposed (vote 39). Wilson likely chose three-fifths as the initial ratio because ten states had agreed to the same ratio for apportioning requisitions in 1783 (Farrand 1966, 205). Wilson was no friend of slavery and thought the institution would soon be extinguished through the normal course of politics (Ballingrud and Dougherty 2018; Van Cleve 2010). Vote 39 was on a mixed trajectory of 67^○, reasonably close to the Lower South’s sweet spot, reflecting both a sectional division in the vote and an appeal to the large state coalition. If popular apportionment was not included, Massachusetts and Pennsylvania would have opposed the apportionment and the motion would have failed.

Southern delegates would try to improve their position from this mark, and Northern delegates would try to curb it back, but ultimately their attempts to expand or contract the Three-Fifths Clause floundered.

The clause was revisited on July 11, when Charles Cotesworth Pinckney and Pierce Butler, co-delegates from South Carolina, proposed to strike out three-fifths and replace it with an apportionment that treated “Blacks” equal to “Whites” (their terminology).²⁵ This did not mean that they were ready to give blacks the right to vote. It meant that they wanted to count blacks as 5/5ths in the apportionment rather than 3/5ths which would increase the proportion of seats allocated to their state. The trajectory of the vote was −52^○ putting South Carolina and Georgia on the extreme. South Carolina and Georgia were the farthest states from the bloc median line at 0.79 and 0.72 units, respectively, compared to the average state distance of 0.32 units. As it turns out, both Northern and Southern delegates spoke against the proposal and vote 132 failed three states to seven.

Four votes later, the convention tried to formally agree to the Three-Fifths Clause. This motion failed four states to six. Clearly, the convention was willing to let the Three-Fifths Clause on the table to establish population as the principle of apportionment, but they were not prepared to concede to the three-fifths ratio yet. Again, this vote was on a trajectory of −25^○, making delegates from South Carolina and Georgia the farthest from the center of the vote. King (MA) opposed the motion because “the admission of [blacks] along with Whites at all, would excite great discontents among the States having no slaves” (Farrand 1966, 586). Nine votes later, Charles Pinckney tried to improve the representation of the Lower South by again proposing to rate blacks equal to whites. Again, the trajectory of the vote made delegates from South Carolina and Georgia extreme. Without a principle to base the exception upon, even Southern delegates would not give into the demands of the Lower South. George Mason (VA) “could not agree to the motion, notwithstanding it was favorable to Virga. Because he thought it unjust” (Farrand 1966, 581). The vote failed, with only South Carolina and Georgia voting in favor of it.

Ironically, the convention’s experience with apportionment was not one where Northern delegates were openly cajoled into accommodating the Lower South. Southern interests were advanced with Wilson’s initial proposal, then subsequent votes that went against their interests failed. The remaining attempts to strengthen the clause were thwarted partly because preferences for those votes varied on the sectional dimension alone.

Eventually, the convention accepted the three-fifths ratio in a form that was almost identical to its initial proposal. For this reason, the story of the Three-Fifths Compromise is not one of gradually succumbing to the will of the South. It is one of initially proposing something that appealed to a winning coalition, then finding little room for amendment.²⁶

Export Taxes

If votes on apportionment highlight some of the Lower South’s failures, then votes on international trade, particularly export taxes, should mark some of its successes. It is natural to presume that delegates from the Lower South were more successful on international trade because those issues were more unanimous with less sectional voting. However, this was not the case. The average vote margin across state blocs was smaller on international trade than on apportionment, making international trade the more divisive issue. Furthermore, a majority of Northern states were pitted against a majority of Southern states on a greater percentage of international trade votes than they were on apportionment votes.

The convention first addressed export taxes when it considered Article VII of the report of the Committee of Detail on August 16. Section 1 of the report gave Congress the power to lay and collect duties, imposts, and excises, but not export taxes.

Shortly thereafter, Madison (VA) proposed that export taxes should be allowed if two-thirds of each house gave their consent (vote 335, Farrand 1966, 363). Although an outright prohibition of export taxes was dear to many delegates from the Lower South, the underlying issue was largely about the power of the national government, putting the vote on a 28^○ trajectory, with South Carolina and Georgia the most mainstream states. Although Madison, a Southerner, favored the exception, delegates from the Lower South did not. Abraham Baldwin (GA) was on the bloc median line. He and Few (GA) voted against it, as did the South Carolina delegation. The measure was struck down in a narrow five to six vote. Delegates from the Lower South were on the winning side of a sectional vote because their preferences were mainstream.

The next vote was to agree that “no tax or duty shall be laid by the Legislature on articles exported from any State,” as proposed in the report of the Committee of Detail. The proposal was on an angle of 33^○, which again favored delegates from the Lower South. This time Rutledge (SC) was on the bloc median line. After hearing the proposal, Gerry (MA) stood up and proclaimed he “was strenuously opposed to the power over exports. We have given [the general government] more power already than we know how will be exercised” (Farrand 1966, 362). Gerry’s argument was not about sectional imbalance. It was about the excessive power of the federal government, which might explain why Massachusetts switched sides and voted with the South in a 7-4 vote. Again, the vote was sectional, it protected Southern interests, but delegates from the Lower South succeeded on this vote because it also limited powers of the federal government which enough Northern delegations favored.²⁷

Votes over import duties and supermajority rules for treaties proceeded similarly.²⁸ They further show that delegates from the Lower South needed federalism to help them appease enough Northern delegates.

Conclusion

The effect of the Three-Fifths Clause on the apportionment of the U.S. House of Representatives is inescapable. Rhode Island had 15,000 more free citizens than Georgia in the first census, but Georgia was entitled to one more seat than Rhode Island because it also had 29,000 slaves. Similarly, New Jersey had 30,000 more free citizens than South Carolina, but South Carolina was entitled to one more seat than New Jersey because it had 100,000 slaves (Amar 2006). The clause gave the South relative parity with the North, helping to maintain human bondage until the Civil War. The extra seats in the House allowed Missouri to be admitted as a slave state in 1820 and ensured enactment of the gag rule in 1840, which smothered antislavery petitions.

Did the original Constitution include provisions like the Three-Fifths Clause because the North traded votes with the Lower South? While there is no way to know what happened behind the scenes during the summer of 1787, there is a rational scenario that differs from the common theme of a “corrupt bargain” between Northern delegates and the Lower South. That scenario made interfering with slavery a lower priority for delegates than nationhood, federalism, and issues of representation. Assuming that all delegates voted sincerely, we have shown that the Lower South succeeded when a vote protected states rights or appealed to the large-state coalition. In this sense, it is possible that delegates voted sincerely and mixing federalism with sectionalism allowed slavery into the Constitution.

It is natural to think that a region, party, or ethnic group that controls a majority of seats must control voting. For example, it is natural to think that Northern states would have always won sectional votes because the North controlled a majority of state delegations. Similarly, between 1955 and 1980 Democrats controlled both houses of the U.S. Congress and might have been expected to control legislation during those years. However, numerical minorities can win majority rule votes if they appeal to the interests of enough members of the predominant coalition. The Lower South got what they preferred on sectional issues at the Convention when a vote appealed to at least a few Northern states. Similarly, the Republican Conservative Coalition won several votes in Congress between 1955 and 1963 when their ideas appealed to Southern Democrats, most notably on issues of segregation. Mixing preferences on a primary dimension with preferences on a second dimension can help regional, partisan, or demographic minorities help form new alliances, even if those alliances are only temporary.

Our study suggests that when politics is multidimensional, then these types of minorities can win by appealing to higher dimensions. Such groups may not win on all issues, but they can be in the center of a vote even if they are not in the center of the preference distribution. In this way, they might affect policy the same way that the Lower South affected the Constitution.

Footnotes

Acknowledgements

The author would like to thank Paul Carlsen, Jac Heckelman, and Jeremy Pope for useful comments; Jamie Carson for help with the broader implications; Jamie Monogan for statistical advice; and Robert Cooper for research assistance. The author would also like to give special thanks to Ryan Bakker for coaching him through MICE and developing the Bayesian IRT (item response theory) model described in the online supplement, note 6.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Funding for part of this research was provided by the National Science Foundation, SES-0752098.

Notes

Supplemental Material

Supplemental materials for this article are available with the manuscript on the Political Research Quarterly (PRQ) website. Data for this project are available at .

References

Aldrich

John H.

McKelvey

Richard D

. 1977. “A Method of Scaling with Applications to the 1968 and 1972 Presidential Elections.” American Political Science Review 71 (1): 111–30.

Amar

Akhil Reed

. 2006. America’s Constitution: A Biography. New York: Random House.

Ballingrud

Gordon

Dougherty

Keith L

. 2018. “Coalitional Instability and the Three-Fifths Compromise.” American Journal of Political Science 62 (4): 861–72.

Bateman

David A.

Lapinski

John

. 2016. “Ideal points and American Political Development: Beyond DW-NOMINATE.” Studies in American Political Development 30 (2): 147–71.

Bilder

Mary Sarah

. 2015. Madison’s Hand: Revising the Constitutional Convention. Cambridge: Harvard University Press.

Brown

Ralph H.

Whitaker

J. Russell

. 1948. Historical Geography of the United States. New York: Harcourt, Brace and World.

Clinton

Joshua D.

Meirowitz

Adam

. 2004. “Testing Explanations of Strategic Voting in Legislatures: A Reexamination of the Compromise of 1790.” American Journal of Political Science 48 (4): 675–89.

Davis

Otto A.

DeGroot

Morris H.

Hinich

Melvin J

. 1972. “Social Preference Orderings and Majority Rule.” Econometrica 40 (1): 147–57.

Dougherty

Keith L.

Heckelman

Jac C

. 2006. “A Pivotal Voter from a Pivotal State: Roger Sherman at the Constitutional Convention.” American Political Science Review 100 (2): 297–302.

10.

Dougherty

Keith L.

Heckelman

Jac C

. 2008. “Voting on Slavery at the Constitutional Convention.” Public Choice 136 (3–4): 293–313.

11.

Dougherty

Keith L.

Heckelman

Jac C

. 2012. Delegate Positions on All Substantive Roll Calls at the United States Constitutional Convention, 1787 [Computer File]. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor]. doi:10.3886/ICPSR33865.v1.

12.

Farrand

Max

. 1966. The Records of the Federal Convention of 1787, vols. 1, 2, and 3. Westford: Yale University Press.

13.

Fehrenbacher

Don

. 2001. The Slaveholding Republic. New York: Oxford University Press.

14.

Godfrey

Joseph

Grofman

Bernard

Feld

Scott L

. 2011. “Applications of Shapley-Owen Values and the Spatial Copeland Winner.” Political Analysis 119 (4): 306–24.

15.

Heckelman

Jac C.

Dougherty

Keith L

. 2013. “A Spatial Analysis of Delegate Voting at the Constitutional Convention.” Journal of Economic History 73 (2): 407–44.

16.

Henderson

H. James

. 1974. Party Politics in the Continental Congress. New York: McGraw-Hill.

17.

Hutson

James E.

1987. “Riddles of the Federal Constitutional Convention.” William and Mary Quarterly 44 (3): 411–23.

18.

Jenkins

Jeffery A.

Morris

Irwin L

. 2006. “Running to Lose? John C. Breckinridge and the Presidential Election of 1860.” Electoral Studies 25 (2): 306–28.

19.

Jenkins

Jeffery A.

Sala

Brian R

. 1998. “The Spatial Theory of Voting and the Presidential Election of 1824.” American Journal of Political Science 42 (4): 1157–79.

20.

Jillson

Calvin

. 2008. Constitution Making: Conflict and Consensus in the Federal Convention of 1787. New York: Agathon.

21.

Jillson

Calvin

Wilson

Rick

. 1994. Congressional Dynamics: Structure, Coordination, and Choice in the First American Congress, 1774-1789. Palo Alto: Stanford University Press.

22.

Klarman

Michael J.

2016. The Framers’ Coup: The Making of the United States Constitution. New York: Oxford University Press.

23.

Lynd

Staughton

. 1966. “The Compromise of 1787.” Political Science Quarterly 81 (2): 225–50.

24.

Main

Jackson Turner

. 1973. Political Parties before the Constitution. Chapel Hill: The University of North Carolina Press.

25.

McDonald

Forrest

. 1965. E Pluribus Unum. Boston: Houghton Mifflin.

26.

McKelvey

Richard D.

1976. “Intransitivities in Multidimensional Voting Models and Some Implications for Agenda Control.” Journal of Economic Theory 12:472–82.

27.

National Archives. 2018. “Elbridge Gerry, Massachusetts.” October 30. https://www.archives.gov/founding-docs/founding-fathers-massachusetts#gerry.

28.

Poole

Keith

. 2000. “Non-Parametric Unfolding of Binary Choice Data.” Political Analysis 8 (3): 211–37.

29.

Poole

Keith T.

2005. Spatial Models of Parliamentary Voting. New York: Cambridge University Press.

30.

Pope

Jeremy

Treier

Shawn

. 2011. “Reconsidering the Great Compromise at the Federal Convention of 1787: Deliberation and Agenda Effects on the Senate and Slavery.” American Journal of Political Science 55 (2): 289–306.

31.

Pope

Jeremy

Treier

Shawn

. 2012. “Mapping Dimensions of Conflict at the Federal Convention of 1787.” Legislative Studies Quarterly 37 (2): 145–74.

32.

Rossiter

Clinton L.

1966. 1787: The Grand Convention. New York: Macmillan.

33.

Stewart

David.

2008. The Summer of 1787: The Men Who Invented the Constitution. New York: Simon & Schuster.

34.

U.S. Bureau of the Census. 1989. Historical Statistics of the United States, Colonial Times to 1970. Washington, DC: U.S. Department of Commerce.

35.

Van Cleve, George William.

2010. A Slaveholders’ Union: Slavery, Politics, and the Constitution in the Early American Republic. Chicago: The University of Chicago Press.