1. Introduction
Direct democracy has not only become a prominent instrument for making policy decisions, but is also increasingly used to ratify constitutions, constitutional amendments, and other revisions to a polity's fundamental law.Footnote 1 Some democratic theorists see this development as critical for both legitimizing the choices of a framework for governing, as well as a mechanism for constraining the ability of elites to design constitutions in their favor.Footnote 2 The prominence and importance of this practice is emphasized by Jeffrey Lenowitz, who calls the ratification referendum a “common, recommended and consequential constitution-making procedure.”Footnote 3 Yet, many advanced democracies, including the United States, France, and Germany, do not require the direct ratification by the electorate of constitutional amendments. In fact, less than half of existing constitutions that specify an amendment process mandate a popular ratification requirement as well.Footnote 4 Despite this procedure's prominence, surprisingly few studies have investigated the factors that account for this institutional variation across democracies.Footnote 5
We contribute to our understanding of this constitutional development by studying the uneven adoption of mandatory ratification referendums across American states during the antebellum period (1789–1860). Independence from Great Britain in 1776 brought not only the need for constitutional frameworks for self-government, but also for procedures by which these could be reformed. In particular, it was quickly recognized that constitutional law should not be made via ordinary statute in the state legislatures.Footnote 6 As a result, constitutional conventions—temporary legislative bodies elected solely to draft a new constitution or modify an existing one—were quickly adopted across all states. While this mechanism became the standard constitution-making process for more than a quarter century, a period of widespread constitutional reforms in the decades following the War of 1812 brought demands to institutionalize greater popular input in the constitution-making process.Footnote 7 Reformers demanded the ratification by voters as necessary for legitimizing constitutional changes and as a check on the power of elites in state-level representative bodies.Footnote 8 Soon after, the ratification of conventions via referendum became mandatory across the North. Although referendums also came to be widely used in the South, their use was not institutionalized. Instead, a doctrine of “absolute sovereignty” of conventions emerged in which conventions had the discretion of whether to submit constitutional changes to voters for ratification.Footnote 9
What explains this uneven adoption of mandatory ratification across otherwise similar political systems? We begin by relying on insights from works analyzing the political consequences of direct democracy.Footnote 10 For one, direct democracy is seen as a mechanism that can offset the power of elites. It allows citizens to directly participate in the policymaking process and bypass representative institutions, which may be biased in favor of special interests.Footnote 11 Because the outcomes of representative and direct democracy may differ, ex post ratification via referendum serves as a veto player in the policymaking process.Footnote 12 In terms of the constitution-making process, mandatory ratification serves as a “downstream constraint” on the power of assemblies and therefore conditions the behavior of delegates when crafting the state's fundamental laws.Footnote 13 By removing the discretion of whether to submit constitutional changes for ratification, the electorates of Northern states could veto constitutional revisions that deviated significantly from the preferences of the state's median voter.Footnote 14 Hence, this institutional innovation provided the electorates of Northern states an important constraint on the ability of local elites to otherwise structure the state's constitutions in their favor.
Using these insights, we provide a logic for why Southern states adopted the first constitution-making innovation, the constitutional convention, but not the second, the mandatory use of ratification referendums. Unlike the legislatures of most Northern states, in which apportionment was based on population, Southern legislatures tended to systematically overrepresent high slave-population-share regions (henceforth, slave-share regions).Footnote 15 The shift from representative to direct democracy would therefore have favored the poorer and less slave-dependent regions. We argue that Southern elites rejected the adoption of mandatory referendums because this practice would have threatened their ability to write and implement constitutional law that furthered pro-slavery interests. In particular, slaveholders sought to retain their ability to maintain the overrepresentation of higher slave-share districts (which was seen as crucial for protecting their interests, especially from redistributive taxation).Footnote 16 Hence, the discretionary—as opposed to mandatory—use of referendums allowed the slaveholding elite to bypass the electorate when constitutional law divided voters along slave-dependency lines.
We use various pieces of evidence to substantiate our argument. First, we use an original data set of legislative representation during the antebellum period to show that Southern state legislatures, and therefore constitutional conventions, were significantly more malapportioned than those in Northern states. We then use this data set of within-state inequality in legislative representation between 1790 and 1860 to demonstrate that the slave-intensive districts were systematically overrepresented in these states’ legislatures. Next, we investigate several cases of political reform between 1829 and 1844. These cases show that opposition to mandatory referendums was associated with legislative overrepresentation, and when biases to representation were reformed, the opposition to requiring voter ratification declined. Our cases include the Virginia convention of 1829, where delegates from the lower slave-share “highland” regions of the state tried and failed to reform the representation in the General Assembly. Key to our argument, these delegates then tried to have the issue of future apportionment decided via statewide referendum. When the higher slaveholding districts voted down this proposal, the delegates from lower slave-share districts proposed that changes to apportionment made by the General Assembly should require voter ratification. We contrast this with the 1834 convention held in Tennessee, where reforms to apportionment occurred at the same time that the state adopted mandatory ratification of constitutional revisions. We argue that the removal of the bias to representation favoring higher slave-share districts had the effect of reducing the benefits to discretionary referendums and therefore also reducing elites’ opposition to making voter approval mandatory.
We also explore the generalizability of our argument by showing how reforms to apportionment in the relatively undemocratic nonslave states of New Jersey and Rhode Island were linked with reforms to the constitutional amendment process. In each state, reforms to the system of representation enacted by conventions held in the 1840s removed the control of rural elites over state politics. In each convention, the adoption of mandatory referendums was met by little opposition. Last, we use the contentious fights over apportionment in Georgia in the 1830s to show why Southern elites tried to block the mandatory referendums in favor of a norm where its use was discretionary. In this case, despite the rejection by voters of two proposals to change the state's system of apportionment to one that favored the higher slave-dependent districts, elites chose to change the constitution without seeking voter approval. When the shift from representative to direct democracy would favor low slave-share regions, Southern elites blocked mandatory referendums to retain the ability to forgo voter approval to further the constitutional protection of slave property.
We conclude by investigating how this divergence between regions in the adoption of the mandatory referendum influenced federal politics in the late antebellum period. In particular, we analyze the strategic use of discretionary referendums during the secession movement of 1860 and 1861, perhaps the most consequential political crisis in American history. In all the states leading this movement, the decision to secede was made in conventions, each of which chose against submitting their ordinances for popular ratification. We argue that these referendums were strategically blocked in order to facilitate secession by bypassing a direct vote of the statewide electorate on this issue. Using an original data set of roll call votes in these conventions, we show that the likelihood of delegates opposing motions conditioning the unilateral secession of their state from the Union to the ratification of voters was positively associated with the slave labor incidence of their constituencies.
Our argument and evidence are relevant to various literatures. First, while there is extensive literature on how distortions to representation influence the allocation of public resourcesFootnote 17 and the relationship between partisan votes and seats,Footnote 18 to our knowledge no previous work has analyzed its relationship with instruments of direct democracy. We therefore expand the scope of these works, demonstrating that biases to representation, such as malapportionment, can play a role in constitutional outcomes. We also contribute to studies on how elites exploit institutional biases. Most of these focus on how multiple pivots can help preserve status quo policies and do not consider how formal biases interact with the strategic use of constitutional procedures.Footnote 19 Similarly, our arguments contribute to a literature exploring how elites design and employ constitutional procedures to preserve political inequality.Footnote 20 Last, this also contributes to our understanding of how slaveholders dominated Southern politics. While the importance of malapportionment in amplifying slaveholders’ power is well documented, we are among the first to show how this source of power influenced other institutional developments.
2. The North-South Divergence
Modern constitution making in the West likely began with the resolutions in May 1776 from the Continental Congress, the de facto governing council of the British North American colonies, to the various colonial legislatures to create sovereign frameworks of government.Footnote 21 The process by which each state pursued this varied by both region and circumstance. The Southern slave states largely created their first constitutions using extralegal meetings of the various colonial legislatures.Footnote 22 In the Northern states threatened militarily by the British, such as New Jersey and New York, colonial elites quickly wrote constitutions with little popular input.Footnote 23 Connecticut and Rhode Island simply continued with their relatively undemocratic colonial charters. In Pennsylvania, on the other hand, a group of pro-independence radicals seized power, abolished the existing colonial legislature, and minimized the existing elites’ control of politics by, among other means, removing the property requirement for suffrage.Footnote 24
The first state constitutions that emerged out of the exigencies of the American Revolution led to immediate demands for reform. The debates on how this should occur resulted in the widespread acceptance that there was a fundamental distinction between constitutional and statutory law and that the former should not occur in the regular legislature.Footnote 25 In the words of one scholar, “The notion that a legislature, even if a ‘full and free representation of the people,’ might lack sufficient authority to act for the people reflected a recognition, present from the outset, that constitutions differed from ordinary statutes and that greater popular input and control were required for their adoption.”Footnote 26 To wit, when the Massachusetts state legislature, the General Court, proposed to draft a constitution in 1778, one town's petition against this stated, “because a Constitution alterable by the Supreme Legislative is no Security at all to the Subject against any Encroachment of the Governing part on any, or on all of their Rights and Privileges.”Footnote 27 The General Court responded to this outcry by calling for a constitutional convention in 1779 whose delegation would be elected with the sole purpose of creating the state's basic law. By the 1780s, this institutional innovation was adopted across the states as the primary mechanism for creating and revising state constitutions.Footnote 28
In addition to the constitutional convention, the Massachusetts General Court employed two additional innovations to the process of creating and amending constitutional law. First, a referendum was held in 1779 on whether a constitutional convention should be called. Second, and arguably more importantly, the General Court required that the convention's constitution be referred back to the voters for popular ratification. While the use of conventions was institutionalized across all states, these mechanisms of receiving both ex ante and ex post direct popular approval spread more slowly. A case in point can be seen by the reforms to Pennsylvania's 1776 constitution, which Pennsylvania elites saw as too democratic.Footnote 29 When the state's eastern elite regained control of the legislature in the mid-1780s, they called for a convention to be held without seeking public approval. This convention, which implemented the separation-of-powers system that eventually became the standard American constitutional framework, chose against seeking voter approval for these revisions.Footnote 30 This process for constitution making was not unusual. Other than Massachusetts (1780) and New Hampshire (1784), no state convention, in either the North or the South, sought the voters’ approval until 1818.Footnote 31
With the popular ratification of constitutions produced by conventions in Connecticut (1818), Maine (1819), and New York (1821), the procedure for framing state constitutions in the North never deviated from the two-step process of electing delegates to a special convention that required ex post approval via a referendum of the state's electorate. Specifically, after 1818, no Northern state convention promulgated a new constitution, whether producing its initial framework or replacing a previous one, without being ratified by referendum. Furthermore, all new constitutions created after 1818 included a provision requiring any revisions to be submitted to the electorate for ratification.Footnote 32
In the 1820s, the product of Southern state conventions also began to be submitted to the state's electorate for ratification. Yet, Southern states deviated in terms of requiring voter approval and instead developed a doctrine that William Freehling describes as the “Southern gospel of a state convention's absolute sovereignty.”Footnote 33 A fundamental pillar of this doctrine is that state sovereignty resides solely with a constitutional convention, and therefore the convention alone retained the discretion of whether their output should be submitted to the voters for ratification. To wit, while conventions in Virginia (1829), Georgia (1832, 1839), Tennessee (1834), and North Carolina (1835) submitted their output for popular ratification, conventions in Delaware (1831), Mississippi (1832), and Arkansas (1836) chose against doing so.Footnote 34 During the period between 1818 and 1860 when all Northern conventions sought voter ratification, fourteen of twenty-two conventions in the South did so.
This divergence between Northern and Southern states can be seen in Figure 1. The upper panel shows the share of constitutional conventions held in the South (left) and North (right) in which the body's ex post results were submitted for popular ratification by different time periods. The first column shows the total number of conventions held (light bar) in the slave states between 1776 and 1820, which is overlaid by the number of these which were submitted to the voters (dark bar). The second bar in each panel reports this for 1821 until 1860. This figure shows how postconvention referendums went from rarely used in the North to being universal by the early 1820s. In the South, we see that their use becomes similarly widespread, although without the same level of ubiquity.
Conventions and Referendums by Region, 1776–1860.
3. The Importance of Discretionary Referendums
Why did the Southern states deviate from their Northern brethren in not requiring referendums to ratify the product of constitutional conventions? This is a puzzle in part because constitutional development across these regions generally occurred in lockstep.Footnote 35 The rapid adoption and spread of the mandatory referendums across the Northern states following the War of 1812 occurred in a period of important and widespread democratic reforms. In particular, most of the remaining economic restrictions to suffrage on adult white males were removed across both regions.Footnote 36 Hence, in terms of patterns of state-level constitutional development, the South's failure to adopt the mandatory postconvention referendum was the exception rather than the rule.
We contend that the Southern doctrine of a convention's absolute sovereignty (i.e., the use of referendums are discretionary rather than mandatory) emerged solely in the South to allow conventions the discretion to bypass the state electorate when writing constitutional law that would divide voters along local slave-dependency lines. Representation in most Southern state legislatures was systematically biased in favor of the higher slave-dependency regions of their states. Because convention delegates were elected from the existing legislative electoral districts, the preferences of the median delegate could therefore deviate substantially from the preferences of the statewide electorate (especially on issues regarding slavery). This meant that the shift from representative to direct democracy would favor the lower slave-share regions. Hence, we argue that it was the existing distortions to representation—and not the direct effect of slavery—that incentivized Southern elites to block the adoption of this constitutional procedure. The importance of apportionment also explains why most Southern states adopted many of the key democratizing reforms of the period (e.g., adult white male suffrage, elected judges), but blocked this pillar of constitutional reform in the North.
In turn, we argue that the slaveholders resisted reform to apportionment versus other democratic reforms, as this was seen as a key mechanism for protecting their investments in slavery. While this typically involved preserving the status quo, large population shifts and other events meant that revising the constitutional rules of apportionment was occasionally necessary. Blocking the adoption of mandatory referendums allowed Southern elites the discretion to manipulate state constitutions (e.g., revise electoral rules) without seeking approval from the statewide electorate.
If, however, the bias to apportionment in favor of higher slaveholding districts was reformed, this should have lowered the bias in favor of slave interests in state representative institutions vis-a-vis the statewide electorate. Hence, we argue that reforms to apportionment should have caused opposition to mandatory referendums to decline. Furthermore, the removal of slaveholders’ de jure control of the legislature substantially altered the incentives of slaveholders with respect to constitutional design. Instead of a relatively easy amendment process, slaveholders should have instead sought constitutional protections to slavery combined with an arduous constitutional revision process.
4. Political Inequality in Southern Legislatures
We begin substantiating our argument that Southern legislatures, and therefore Southern constitutional conventions, were significantly more malapportioned than those in the Northern states. This demonstrates that the shift from representative to direct democracy was likely more consequential in the South, and therefore more likely to be opposed by those residents who were favored by this bias. We then show that this legislative malapportionment heavily favored higher slave-dependent constituencies. This evidence is critical to our claim that the shift from representative to direct democracy would favor the systematically underrepresented low slave-dependent districts within Southern states.
To support these claims, we constructed a measure of county-level representation based on the number of members to a state's upper and lower houses each county elected between 1790 and 1860. This information was coded from the relevant statutes or constitutional provisions specifying the apportionment for each chamber of each legislature for each decade from 1790 to 1860. We used a measure of representation that is relative to the “fair” level of each particular state: the Relative Representation Index (RRI).Footnote 37 The RRI creates a common metric across legislatures by normalizing the representation of each county by the fair weight specific to each state. Voters in electoral districts with an RRI of less (more) than 1 were underrepresented (overrepresented) in their legislature. A value close to 1 corresponds to a level of representation consistent with the “one person, one vote” principle. From each Census between 1790 and 1860, we calculated each county's adult white male population, which we used as a proxy for the eligible-voter population. As each state legislature was bicameral, we took the average across chambers.
The difference between regions in inequality of legislative representation is depicted in Figure 2. Each panel plots the empirical distribution of the RRI score for Northern and Southern counties separately for three periods between 1800 and 1860. In each period, it is evident that Northern counties are more concentrated around an RRI of 1 (i.e., the “fair” level of representation) with a smaller right skew. By 1830, both the mean and the variance of RRI in the Northern counties are significantly smaller than the ones of Southern counties.Footnote 38 As shown in panel 3, the RRI scores in the Northern states remain on average significantly smaller and with less variance in 1860. The three panels demonstrate the persistence of relatively high RRI scores in the South, indicating greater within-state inequality in legislative representation.
Distribution of RRI by Region, 1800–1860.
We next demonstrate that this Southern bias in legislative representation favored high slave-share districts within each state. In Table 1, we present a series of models predicting RRI using the slave-population share of counties for each Southern state and for each Census decade from 1790 to 1860.Footnote 39 Each column reports this estimate for the corresponding Census decade. In each model, we control for total county population (logged), which is the primary factor associated with legislative malapportionment.Footnote 40 All demographic data were taken from the relevant decennial Census.
County RRI and Slave-Population Share by State Decade
Notes. Each column reports a county-level OLS (least ordinary squares) estimate for slave-population share in which the dependent variable is a county's mean relative representation (RRI) in the state legislature. Robust standard errors are shown in parentheses. Each column reports the Census year of the model and includes total county population for the corresponding Census decade. For each state-decade estimate, the number of counties is reported in Appendix.
**/**/* indicates the level of significance at the 0.01/0.05/0.1 level, respectively.
In total, there are eighty-two individual state-decade models, and the relationship between county slave share and RRI is statistically positive (p < 0.05) in sixty-four of them. In addition to a few that barely miss being significant at conventional levels, most insignificant models occur either early in statehood (e.g., AL, KY, MO, MS) or, most importantly to our arguments, following reforms to apportionment (e.g., TN after 1834). In sum, with the exception of a few states (e.g., TN, TX), in most Southern states there is strong evidence that the shift from representative to direct democracy would strongly benefit the lower slave-share districts.
5. The Incomplete Diffusion of the Mandatory Ratification
We argue that biases to apportionment and opposition to the adoption of the mandatory referendum are linked. When the system of representation deviated significantly from a “one (adult white) man, one vote” basis, the beneficiaries sought to preserve the ability to revise a state's fundamental law without seeking statewide voter approval. If apportionment was reformed to be based on the voting population (e.g., not include slaves), the opposition to mandatory ratification would fall.
In this section, we explore four cases of constitutional reform to support each part of our argument. The first and second are the constitutional conventions in Virginia (1829–1830) and Tennessee (1834), which we use to show that opposition to adopting mandatory referendums worked through apportionment and not directly through slavery. In both states, the calling of conventions was the result of persistent demands to reform the undemocratic elements of their initial constitutions. Virginia's convention resulted in the maintenance of a biased system of representation that preserved the long-existing structural majorities for the highly enslaved coastal areas. Tennessee's convention, on the other hand, made numerous democratizing reforms, including the adoption of a system of representation based on the “one (adult white) man, one vote” principle. Unlike Virginia, we show there was little opposition to reforming the amendment process to include a mandatory referendum. Third, we provide external validity for our claim that opposition to mandatory referendums was influenced by the extent of malapportionment by exploring constitutional reform in the relatively less democratic Northern states of New Jersey and Rhode Island. As with Tennessee, we show that reforms to representation, which diminished rural elites’ control of the legislature, were accompanied by near unanimous support for mandatory referendums to approve constitutional revisions. Our fourth case demonstrates why Southern elites opposed the adoption of mandatory referendums. Specifically, we use the various attempts to reform the system of apportionment in Georgia in the 1830s, each of which were rejected in statewide referendums. Georgia elites then chose to reform the state's system of representation, which was reformed to further protect slave interests, without seeking voter approval.
5.1 Apportionment and Opposition to Referendums: Virginia's Convention of 1829–1830
Virginia's fixed “county” system of representation had long been a source of deep resentment in the rapidly growing (and largely nonslaveholding) western portions of the state.Footnote 41 In particular, unmet demands for greater state involvement in internal improvements, banking, and education were widely attributed to the systemic inequalities entrenched by the state's 1776 constitution.Footnote 42 Years of pressure from not only the Appalachian counties but also the rising and underrepresented urban centers, such as Richmond, led to the legislature consenting to a referendum on whether to hold a convention.Footnote 43 The voter's approval led to the holding of a convention in the fall of 1829, whose debates about the propriety of majority rule were made by some of the country's most distinguished political figures. The ninety-six-member delegation (four delegates elected from each of the state's twenty-four senatorial districts) included two former presidents (James Madison and James Monroe), one future president (John Tyler), the future Speaker of the U.S. House of Representatives (John Jones), approximately twenty-five past or future members of Congress, and the sitting chief justice of the U.S. Supreme Court (John Marshall, as well as future justice, Philip Barbour).
Along with other democratizing reforms (e.g., lowering economic restrictions to suffrage, reform of the judicial system), the primary debates in the convention revolved around whether to apportion representation in the state legislature according to white population. Slaveholders saw preserving the system that guaranteed majorities for the eastern part of the state, regardless of white-population movements, as critical for limiting the ability of low slave-share regions within their states to increase taxation on slave property. This logic was expressed in a speech to the convention by Abel Upshur, a coastal slaveholder and future U.S. secretary of state:
It is contended by our opponents, that the proper basis of representation in the General Assembly, is white population alone, because this principle results necessarily form the right which the majority possess, to rule the minority.…There are two kinds of majority. There is a majority in interest, as well as a majority in number … that those who have the greatest stake in the Government shall have the greatest share of power in the administration of it.… Gentlemen tell us that our alarms are unfounded; that even if we should give them power to tax us at their will and pleasure, there is no danger that they will ever abuse it … [t]hey tell us that there is no disposition among them, to practice injustice towards their eastern brethren. Sir, I do firmly believe it … But who can answer for the generations that are to come?… I think, Sir, it must be manifest by this time … that property is entitled to protection, and that our property imperiously demands that kind of protection which flows from the possession of power. Gentlemen admit that our property (slaves) is peculiar, and that it requires protection, but they deny to it the power to protect itself.Footnote 44
We now use roll calls from the convention to show that the proportion of a district's population who were slaves in 1830 is strongly inversely related to a delegate's support for reforming to a white-population basis of representation and requiring statewide ratification of changes to the constitutional basis of apportionment. Specifically, for each roll call we estimate a logit model where the dependent variable is an indicator for either a yea (1) or a nay (0) on each motion. The estimate for each model is presented in Figure 3. Descriptions of each roll call and vote tally are presented in Appendix Table A2.
Roll Calls, Virginia Convention, 1829–1830.
The first roll call analyzed is an amendment to the Committee of the Whole's report to apportion representation by fixed and unchanging districts that gave majorities to the eastern section of the state. This amendment, which called for apportionment to be “based on white population exclusively,” failed by only four votes (forty-six yeas to fifty nays).Footnote 45 As seen in model 1 of Figure 3, district slave share is strongly negatively correlated with support for this amendment.
Upon losing this fight, western delegates sought to resolve the issue of future apportionment. While eastern delegates claimed that the issue had been settled “ad indefinitum,”Footnote 46 western delegates proposed numerous motions seeking to put the issue of the constitutional basis of apportionment directly to the voters via referendum.Footnote 47 Yet, each of these efforts were narrowly defeated by delegates from the eastern counties. Instead, Madison proposed that a separate provision be included in the constitution that specified that reapportionment could occur in 1841 (and once every ten years thereafter) if agreed to by both chambers of the legislature. A delegate offered an amendment that stated that any changes to apportionment made by the legislature had to be submitted to the state's voters for approval. This amendment was again narrowly defeated. As shown in model 2 of Figure 3, opposition to this amendment is highly correlated to the importance of slavery to each delegate's district. This is expected, as every delegate who opposed (supported) a white basis of apportionment also voted against (for) the amendment requiring voter ratification of changes to the system of representation. This demonstrates that delegates on both sides saw the shift from representative to direct democracy as benefiting the lower slave-dependent regions.
Despite the adoption of some democratizing reforms (e.g., reduction in suffrage restrictions), one western delegate claimed that the “obnoxious arrangements as to the representation” would cause him to oppose passage of the overall constitution.Footnote 48 Indeed, the final vote on the overall constitution was fifty-five in favor to forty opposed, with only one delegate from the western part of the state voting in favor. In model 3 of Figure 3, we show that overall support for the final constitution was highly correlated to district slave share.
5.2 Why Adopt Mandatory Referendums? Tennessee's Convention of 1834
While the roll calls in the Virginia convention demonstrate the tight link between preserving the pro-slavery bias to legislative representation and blocking mandatory referendums, a convention held soon after in Tennessee shows how opposition to mandatory referendums can decline once the bias to apportionment is removed. In 1834, a convention in Tennessee met to write a new constitution to replace their initial framework of 1796. As with most other Southern states, apportionment in the initial constitution overrepresented higher slave districts; in Tennessee's case, this was accomplished by including adult black men (i.e., taxable inhabitants) in the population counts for reapportionment. Yet, one unique provision of the original constitution led to strong pressures for reform. The 1796 compact mandated that taxation of land must be equal by acre (and not by value). While adopted to induce migration into the sparsely populated frontier state, it resulted in great inequities in the incidence of taxation.Footnote 49
The need to reform this regressive constraint on taxation led to a convention that adopted one of the more democratic constitutions in the South.Footnote 50 Yet, as we show in Figure 4 using roll call votes from this convention, opposition to reforming the basis of apportionment and requiring ratification of constitutional changes was similarly conditioned by district-level dependency on slave labor. As with Virginia, we located the relevant roll calls in the convention's journal.Footnote 51 We also use logit analysis to predict support for each resolution using the slave-population share of each delegate's district.
Roll Calls, Tennessee Convention, 1834.
First, model 1 in Figure 4 shows that support for the provision that taxes on property should now be based solely on value is negatively correlated with district slave share of the delegates.Footnote 52 This is unsurprising, as this reform would increase taxation on the more valuable land that was conducive to the large-scale use of slave labor. The inability to defeat this provision (39 Y, 19 N) signaled the weakness of the higher slaveholding districts in the convention. We next analyze this relationship for the vote on adopting a qualified voters basis of apportionment.Footnote 53 Given that Tennessee had no economic restrictions to voting, the passage of this reform (33 Y, 24 N) would increase the legislative representation of the lower slave-share counties.Footnote 54 Model 2 in Figure 4 shows that opposition to this resolution was negatively correlated with the delegates’ district slave share. We argue that once the bias to representation is reformed, the value of retaining legislative discretion of whether to submit constitutional changes for ratification declines. Indeed, this is precisely what occurred. The vote on whether to require voter ratification of constitutional revisions was passed by a wide margin (46 Y, 9 N).Footnote 55
Why did reform prevail in Tennessee where it failed in Virginia (not to mention most other slave states)? One idiosyncratic reason for this success can be explained by the fact that the unique tax provision in the original constitution led to intense demands for reform. This can be seen in the composition of the elected delegation. In contrast to the storied convention in Virginia, voters in Tennessee selected a delegation that featured relatively few prominent federal or even statewide political figures.Footnote 56 Perhaps more importantly, Tennessee was simply less dependent on slave labor. When sorting districts by the proportion of the population who were slaves, the district slave share of the median delegate was 19 percent. By comparison, the district slave share of the median delegate in the Virginia convention was 41 percent. Thus, the economic geography of slavery in Tennessee, combined with demands from the electorate for reform to the restriction on taxation, resulted in a delegation that could overcome slaveholder opposition to these democratizing reforms.
Finally, the success of reform in Tennessee could plausibly be explained by two key protections won by slaveholders. There was significant pressure in the convention to consider abolishing slavery in the state, and a committee was created to consider the matter.Footnote 57 Critically for slaveholders, the Committee on Emancipation chose to not take up the matter in this convention.Footnote 58 Model 3 shows the relationship between district slave share and the delegates’ support for a resolution to concur with the committee's decision.Footnote 59 The second (and enduring) critical protection was a provision of the constitution that stipulated that the legislature could not emancipate slaves without receiving the consent of the slaves’ owners. In a tight vote (30 Y, 27 N), slaveholders won this key protection.Footnote 60 The strongly positive relationship between support for this provision and district slave share is shown in model 4 (Figure 4). As a result, and despite the many democratizing reforms enacted by the convention, the final vote in favor the new constitution showed none of the divisions along slave-dependency lines observed in Virginia. In fact, it was passed nearly unanimously (55 Y, 3 N).Footnote 61
This narrowly won constitutional protection for slavery combined with the loss of political control over the state legislature meant that slaveholders in Tennessee wanted to ensure that the constitution was difficult to amend.Footnote 62 For instance, unlike the 1796 constitution, the new constitution specified no procedure for calling a new convention. Moreover, the process for amending the constitution was arduous. Not only did the voters now have to ratify any revisions, but it also had to first be approved by two-thirds of each chamber across two successive legislative sessions (the second session occurring after the election of new members).Footnote 63
We can see the divergence in strategy among slaveholding elites when looking at the initial constitutions of Florida and Texas, the final two slave states admitted to the Union (each in 1845). In Florida's convention, slaveholders managed to have slaves included in the population counts for apportionment for both legislative chambers.Footnote 64 Consistent with our arguments, the constitution included a procedure for calling a convention and did not include mandatory ratification of any constitutional revisions.Footnote 65 In Texas, delegates from higher slave-share counties also tried to implement a biased system.Footnote 66 After failing to implement a system of legislative representation biased in favor of higher slaveholding districts, delegates from higher slaveholding districts sought and received constitutional protections for slavery. For one, a provision that was nearly identical to that in Tennessee's constitution—in which the legislature could not emancipate slaves without the owner's consent—was included in the constitution.Footnote 67 The convention also chose to not include a procedure for calling a new convention and instead implemented an arduous amendment process that included voter ratification. As in Tennessee, if slaveholders could not control the legislative process, they sought constitutional protections of slavery combined with high institutional impediments to changing the state's fundamental law. A consequence of this divergent strategy can be seen in the number of amendments to the constitution of each state. While the Florida constitution was amended thirteen times between 1845 and 1860, the constitutions of Tennessee (1834) and Texas (1845) were amended a combined one time.Footnote 68
5.3 Mandatory Referendums and Representation in the North
We demonstrated in Figure 2 that within-state inequality in representation was on average lower in the Northern states.Footnote 69 An implication of our argument is that the few Northern states with systematic biases to representation should have been similarly less likely to adopt mandatory referendums. Consistent with this claim, in the few early Northern states that were slow to adopt mandatory ratification, such as New Jersey, Rhode Island, and Vermont, each maintained a highly contentious system of representation that was malapportioned in favor of rural elites.Footnote 70 Furthermore, we argue that reforms to this bias should see a decline in the opposition to mandatory referendums. We explore this claim looking at the state conventions in New Jersey and Rhode Island, respectively, which reformed their basis of legislative representation in the 1840s.
The initial post-colonial constitution of both New Jersey and Rhode Island contained a number of undemocratic elements that would lead to intense pressure for reform. For instance, the original constitution of New Jersey, which was framed over five days in 1776 while the state faced imminent invasion by British forces, was widely seen as less democratic than the initial constitutions of its neighbors, New York and especially Pennsylvania. The result was a governing framework that was similar to those created in the Southern states (e.g., the governor was elected by the legislature combined with a basis of apportionment that gave equal representation to each county regardless of population).Footnote 71 Similarly, Rhode Island maintained its colonial charter following independence, which entrenched a system of extreme malapportionment and high economic restrictions to suffrage. This de jure inequality effectively allowed rural elites to control state politics in both rapidly urbanizing states. Over time, resentment toward each state's undemocratic institutions increased and, in the case of Rhode Island, even culminated in the so-called Dorr Rebellion, where dissidents tried to reform the state's constitution by calling and holding an extralegal convention and legislature.Footnote 72
These pressures, which caused neighboring states to reform decades earlier, eventually led to conventions being held in New Jersey (1844) and Rhode Island (1842), respectively. Each resulted in the adoption of many of the democratizing reforms passed previously in other Northern states, including a provision mandating that all constitutional revisions be ratified by the state's voters. Of particular relevance to our argument, the lower house in each state would now be apportioned based on population. Yet, in each convention, rural elites were able to preserve the preexisting one-senator-per-unit (county in NJ and the town in RI) basis in the upper house. Debates from the Rhode Island convention indicate that this was a key compromise that allowed the other reforms to be passed.Footnote 73 Furthermore, in each of these states, the process for constitutional reform was specified to occur solely in the state legislature, where the maintenance of the biased system in one chamber allowed rural elites to block threatening reforms.
As with Tennessee, reforms to apportionment occurred jointly with the adoption of mandatory ratification of constitutional revisions. Furthermore, these procedures were also adopted with little opposition in the conventions. This is consistent with our claim that opposition to mandatory referendums declined with the ability of elites to manipulate the constitution in their favor. Furthermore, in each case, the elites favored by the pre-reform basis of representation (rural elites in NJ and RI, and slaveholders in TN) won key constitutional protections that allowed them to block policies antagonistic to their interests. Yet, in each case, the reforms also diminished their ability to control the legislature and with it the constitutional revision process (either in the legislature or via a convention). Thus, and unlike the case we present next (Georgia), preserving the ability to choose to circumvent the statewide electorate was no longer as salient in these states.
5.4. The Value of Discretionary Referendums: Georgia's Failure to Reform in the 1830s
We argue that when legislative representation was biased in favor of higher slave-share districts, the possibility to choose whether to seek statewide voter approval allowed Southern elites to write favorable constitutional laws that might be opposed by a majority of the electorate. The value of this lack of mandatory approval when the preferences of convention majorities conflicted with those of the statewide electorate can be seen in the series of conventions held in Georgia over reforming the state's legislative apportionment.
As with each original Southern state, Georgia's constitution of 1798 enshrined a basis of representation that was constructed to preserve legislative majorities for high slave-share districts, a source of political inequality that provoked repeated calls for reform.Footnote 74 While the lower house was based on population, which specifically included slaves in the counts, it was the choice of a one-senator-per-county basis that caused the system to eventually require constitutional revision. Due to the forced removal of Native Americans, especially in the 1820s, large swaths of the northern and western portion of the state opened for settlement.Footnote 75 This necessitated the creation of administrative districts (counties), which threatened the Senate majorities of the eastern slave-dependent districts. Increasing calls for reforms led to the holding of conventions in 1832 and 1839. When, according to one historian, the system of representation proposed by the 1832 convention “did little towards equalizing representation.… Citizens of middle Georgia who had worked hardest for reapportionment of representatives now worked hardest to defeat the proposed amendments (in a post-convention referendum), and in this they were successful.”Footnote 76 A second convention in 1839 similarly failed to equalize legislative representation. As explained by one delegate, the convention tried to revise the constitution to ensure that “the minority of the people govern the many—the very distinguishing characteristic and essence of aristocracy.”Footnote 77 After opposition to the amendments in the low slave-share portions of the state led to its defeat in a referendum, Georgia's governor said, “this is the second time that the people have refused to sanction the proceedings of conventions, held to reform the constitution. In both instances … by the belief that the amendments offered for their approval were intended for sectional purposes.”Footnote 78
Following this defeat, Georgia elites chose to amend the basis of apportionment in the constitution without submitting it for popular ratification.Footnote 79 Specifically, the one-senator-per-county basis was revised to a system of districting that would better protect higher slave-share counties. Georgia's constitution would subsequently be amended ten more times before 1860 without receiving voter ratification, including further constitutional revisions in the 1850s to the system of representation.Footnote 80 This ability to persistently manipulate the state's constitution without voter consent, as similarly exemplified in Florida, would no longer have been possible in the North due to the institutionalization of the mandatory referendum.
6. Conventions, Referendums, and Federal Politics
We conclude by exploring how discretionary referendums were used strategically during the regional crisis of the 1850s and ultimately the Southern secession movement of 1860–61. We first show how Southern elites employed the language of a “convention's absolute sovereignty” (i.e., discretionary referendums) to justify the choice to not submit Kansas's pro-slavery “Lecompton” constitution for voter approval. We then also show how this was similarly utilized during the subsequent secession crisis. Following a brief overview of the process by which each state chose to secede, we provide evidence that, as in the Virginia convention of 1829, support for using referendums largely split along slave-dependency lines. Using the same approach, we demonstrate that delegates from high slave-share districts successfully rejected the proposal to submit the ordinances of secession to the electorates for ratification. This supports our claim that on questions pertaining to slavery, Southern elites saw the use of statewide referendums as harming their interests.
6.1. Bringing Kansas into the Union
The political importance of the divergence between regions is revealed in the debates regarding the admission of Kansas to the Union in 1857–58. This episode is particularly relevant because it shows how the Southern doctrine of a “convention's absolute sovereignty” was used by slaveholders in Kansas to bypass the opposition to bringing the territory into the Union as a slave state. A sectional conflict occurred when a convention requested admission to the Union without submitting the proposed constitution to the territory's voters for ratification.Footnote 81 When the appointed territorial governor, Robert Walker, opposed the admission without voter ratification, Mississippi congressman and future delegate to his state's secession convention, Lucius Lamar, invoked the Southern doctrine in a letter to the U.S. Secretary of the Treasury Howell Cobb: “our objection goes no farther than to Walker's threat (to refuse admission without a referendum) to make such course a sine qua non of admission as a state to the Union.” In the same letter, Lamar went on to claim that it was a “shameless abandonment of our right … [to] oppose the admission of a slave state, merely because her constitution was not submitted to the people.”Footnote 82
On the question of the “propriety of submitting the constitution for ratification,” Alexander Stephens, a congressman from Georgia and future vice president of the Confederacy, also invoked the doctrine in a letter to his constituents: “I have nothing to say, because … it is the right of the convention to do it or not, as they please.” Yet, he also revealed the strategic importance of not submitting the constitution for ratification by saying, “If they do thus conform [to Walker's demand of submitting], the question [of Kansas’s status as a slave state] will most probably be ended.”Footnote 83 Cobb, a future secessionist and the first president of the Provisional Confederate Congress, wrote privately to Stephens that “refusing to submit the constitution to the people for ratification … will produce the most dangerous crisis we have yet had on the Kansas question. The reply can be made with overwhelming power that the refusal to submit was the result alone of a fear that a majority would condemn it.”Footnote 84
The ensuing debates in Congress on the Kansas issue revealed the clear divide between free and slave states over the necessity to ratify the eventual constitution. In his famous “Cotton Is King” speech to the Senate on March 4, 1858, South Carolina Senator James Hammond began by stating that the free states would admit Kansas if “this constitution embodied the will of the people of Kansas … the only question is, how that will is to be ascertained, and upon that point, we differ. In my opinion the will of the people of Kansas is to be sought in the act of her lawful convention elected to form a constitution, and nowhere else.”Footnote 85 New York Senator Preston King responded that if Kansas were admitted, “the sovereign right of the people is denied, and the sovereign right of a convention is affirmed.… [The question] is no less vital to the people of every State now in the Union, because it is the question where sovereignty resides, whether in representative bodies, or in the people.”Footnote 86 Stephen Mallory, senator from Florida, challenged King's premise by arguing that the free states’ objections “could be summed up in this: the constitution framed at Lecompton was not submitted to the vote of the people, and it does not reflect the will of the people of Kansas. Let me ask, sir, who are the people of Kansas? The senator from NY says that the people means the majority. I deny any such doctrine.”Footnote 87
6.2. The Secession Crisis of 1860–61
The failure to bring Kansas into the Union as a slave state contributed to rapidly deteriorating sectional tensions over slavery. When Abraham Lincoln was elected president solely with electoral votes from Northern states, most slave states quickly convened special sessions of their legislatures to debate their response to Lincoln's election and the possibility of leaving the Union. Pro-secession legislators considered two mechanisms by which a state could declare its independence. The first was simply to secede via statute in the state legislature, an expedient and more certain path preferred by many secessionists.Footnote 88 This option was defeated in part because it would have violated the long-held precedent of using an elected convention to modify a state's fundamental law. Alexander Stephens explained this in a speech to the legislature of Georgia in November of 1860: “the Legislature is not the proper body to sever our Federal Relations.… Sovereignty is not in the Legislature.… I am for … calling together an untrammeled convention, and presenting all the questions to them whether they will go out of the Union.… [I] know of no way to decide great questions affecting fundamental laws except by representatives of the people.”Footnote 89
The second primary debate was whether the electorates in each state would be called upon to sanction the holding of a “secession convention” or ratify a convention's decision to secede. Harboring doubts about the popularity of their cause, secessionists in each state opposed any form of direct vote on secession. This was candidly expressed by Alfred Aldrich, a South Carolina legislator, in a letter to U.S. Senator James Hammond: “If the question of secession must be referred back to the people, then it will be an utter failure … [I] do not believe the common people understand it, in fact, I know they will not understand it.”Footnote 90 Another state legislator from South Carolina wrote to Congressman Milledge Bonham about avoiding a referendum: “[secessionists] know that if the people decide against them, it will be decisive against separate state action for all time to come.”Footnote 91 By contrast, those opposed to secession demanded to hold a statewide vote on whether to have a convention or to place limits on the powers of such a body by requiring posterior ratification of the voters.Footnote 92
In each of the first six states to secede (South Carolina, Mississippi, Florida, Alabama, Georgia, and Louisiana), secessionists were successful in passing bills calling for a convention without seeking voter approval. More importantly, these legislatures abided by the doctrine of a convention's absolute sovereignty and placed no “upstream constraints” on its scope or power or “downstream” constraints by requiring postconvention ratification.
Between the passage of this legislation and the elections for delegates, two factions emerged. Immediate secessionists advocated for their state to choose to unilaterally and immediately secede in the state's convention. An opposing faction, which became known as cooperationists, was a coalition of unionists, moderates, and pro-slavery supporters who, at a bare minimum, opposed unilateral secession. While cooperationists advocated for holding a Southern-wide convention, among other stances, a unifying position among this heterogeneous group was that any decision by a convention would only take effect if voters ratified it via referendum.Footnote 93 In each Lower South convention, secessionists composed a majority of delegates and successfully blocked each motion, including the proposal to submit the decision for voter ratification, passing ordinances that immediately and unilaterally removed their state from the Union.
In the remaining slave states, anti-secessionist state legislators were successful in requiring direct voter input on secession. Voters in these states decided directly via referendum on whether to hold a convention and even to subject any secession resolution to a posterior ratification. For example, in North Carolina and Tennessee, voters initially rejected the holding of a convention. This defeat was anticipated by a secessionist leader in North Carolina in a letter to one of the state's congressmen: “You cannot unite the masses of any Southern state much less those of N.C. against the Union & in favor [of] slavery alone.”Footnote 94 In Virginia, the legislature called for a convention but also held a referendum on whether secession would require a posterior popular ratification. The inclusion of this provision caused a “heated debate” and was strongly opposed by secessionists in the legislature.Footnote 95 Voters later approved by a wide margin the need for postconvention ratification. The view among secessionists that referendums would harm the prospects of the movement (and the similar view among cooperationists) is consistent with our argument that slaveholders opposed mandatory referendums because the shift to direct democracy was politically harmful for their interests. These Upper South states only seceded after conflict began with the firing on Fort Sumter in Charleston and Lincoln's subsequent call on April 15, 1861, for troops to suppress the movement.
6.3 Opposition to Popular Ratification of Secession Ordinance and Slavery
We now use roll call votes from the Lower South conventions to analyze the relationship between the choice to oppose requiring popular ratification of secession and the slave labor incidence of local constituencies. We show that the delegates from higher slave-share districts were crucial to defeating the drive to condition secession to ratification by voters.
Our measure of support for the use of direct democracy comes from key roll call votes recorded in the various “secession” conventions. Namely, we coded all available votes on resolutions to submit any secession proposal to the voters for ratification. Of the six states that seceded prior to the formation of the Confederacy in February, 1861, four states recorded the roll calls on whether to require voter ratification (Alabama, Florida, Louisiana, and Mississippi).Footnote 96 The precise resolutions, their description, and the final vote tally in each is presented in Appendix Table A2.
As we did previously, our main explanatory variable is the slave-population share of each delegate's district in 1860. Unlike earlier Censuses, the 1860 Census allows us to control for a rich set of demographic and economic factors that could influence the choice of whether to submit the secession ordinance for ratification. Specifically, we control for total population, the level of urbanization, population density, the Gini coefficient of land inequality, the density of railway coverage (miles of railroad per square mile), and the share of the labor force employed in manufacturing. All variables are from 1860, and their definitions and sources are described in Appendix Table A3. Most importantly, the 1860 slave schedules allow us to control for the number of slaves each delegate owned in 1860. This should help capture a delegate's personal economic stake in slavery. Each record was located using Ancestry.com, which allowed us to identify the information using the name of each delegate.
We again used logit analysis to investigate the relationship between the delegate-level support for requiring voter ratification of secession and the slave dependence of their constituents. For each state, the dependent variable is a binary indicator of whether the delegate voted to submit the ordinance for ratification. Figure 5 summarizes our estimates. In the left panel, where we estimate the relationship without controls, we plot the corresponding marginal effect (and its 95 percent confidence interval) of district slave-population share on the likelihood of opposing popular ratification in each of the four conventions. The right panel shows this same model with the controls described above included.
Slavery and Roll Call Opposition to Popular Ratification.
In Figure 5, the estimate on slave share for all conventions is positive and highly statistically significant with controls (right column) and without controls (left column). The magnitude of these effects is also large. For instance, the estimated effect for Alabama (marginal effect equal to 0.80, standard error of 0.2), is highly significant and implies that a delegate from a county 1 standard deviation above the mean slave share is expected to have around 0.17 more chance of voting against ratification than a delegate from the average county in this state (which has a slave share of 38 percent). The point estimates for the other states are remarkably similar and precisely estimated.
This is consistent with our findings from earlier conventions and our overall argument that slaveholders saw the shift from representative to direct democracy as harming their interests. In Appendix Figure A1, we compared this opposition to popular ratification with the support for other pro-secession motions in these conventions and found that support for secession is similarly conditioned by the dependency on slavery of each delegate's constituents.Footnote 97 This is consistent with our claim that both supporters and opponents of secession, who largely split along slave-dependency lines, saw statewide referendums as harmful to the movement's electoral prospects.
7. Conclusion
In this article, we examined the conditions under which democracies adopt the use of direct democracy mechanisms to ratify constitutional laws. We explored this issue by studying the puzzle of why regions in the United States adopted constitutional conventions during the antebellum period, but only the Northern states subsequently adopted mandatory postconvention referendums. We argued that long-running distortions to representation in favor of slave-dependent regions in the South meant that the shift from representative to direct democracy would favored lower slave-share districts. Instead, the discretionary referendum, as embodied by the Southern doctrine of a “state convention's absolute sovereignty,” were adopted allowing the slaveholding elite to modify state constitutions without seeking voter ratification.
We provided evidence for our argument using an original data set measuring county-level inequality in legislative representation. We showed that not only were Southern legislative bodies on average more malapportioned than those in Northern states, but that the bias in representation was highly correlated with local slave-population incidence. We supplemented this evidence with four cases of state-level constitutional reform between 1829 and 1844. Each shows that opposition to mandatory ratification of constitutional revisions was related to the extent of malapportionment present. If biases to representation were reformed, then opposition to requiring voter ratification declined. Yet, if Southern elites could retain de jure control of state legislatures, they preferred to retain the option to bypass state electorates when constitutional revisions divided voters along interests over slavery.
We explored the implications of this institutional divergence in the events leading to the Civil War. In particular, we emphasized the sectional conflict over the admission of Kansas in the late 1850s and the lack of popular ratification of all the Lower South secession conventions during the 1860–61 crisis. While it is difficult to evaluate whether slaveholders correctly feared majority opposition to unilateral and immediate state secession, the existing electoral returns for the convention delegates suggest that in the larger Lower South states, support for immediate secession was far from overwhelming.Footnote 98 In fact, the closeness of the overall votes combined with the desire to convince the public that secession was overwhelmingly popular led the governors of each state to refuse—despite many demands to do so—to publish the election results.Footnote 99
The patterns established during the antebellum period persisted into the twentieth century. While no Northern state in the postwar period failed to submit a convention's product for voter ratification, Southern states continued to use referendums inconsistently. Most notoriously, the conventions in Mississippi (1890), South Carolina (1895), Louisiana (1898), and Virginia (1901), which enacted suffrage restrictions, such as literacy tests, that largely disenfranchised African Americans until the Voting Rights Act of 1965, were not submitted for voter ratification.Footnote 100 Consistent with our argument, malapportionment remained one of the key institutions amplifying the political power of rural elites in the postbellum period.Footnote 101
Our argument has a number of important generalizable implications. For one, distortions to representation, including malapportionment, are a prevalent feature of the world's legislatures.Footnote 102 Our argument suggests that the winners of these biases, typically rural interests, will oppose the adoption of mandatory referendums. Yet, it is these polities where the adoption of this institutional veto point might be most consequential for constitutional design. A related implication is regarding the mechanisms by which de jure political inequality is maintained over time. The mandatory referendum provides a constraint on the ability of elites to write a state's fundamental law in its favor. To wit, all U.S. states now require statewide ratification of any constitutional revisions. Yet, the federal constitution was not submitted for voter ratification, and it contained many institutions that were biased in favor of particular interests (e.g., equal apportionment in the Senate, the electoral college, the three-fifths clause). Furthermore, instead of facilitating the revision process, the constitution was made incredibly difficult to amend and included no mechanism for receiving direct nationwide popular input.
DATA APPENDIX
Constitutional Conventions, 1776–1861
Description of Roll Calls
Data Description and Sources
Roll-Call Opposition to Voter Ratification and Support for Secession.
