What explains the emergence and persistence of democratic institutions? Recent research shows that there are many possible answers to this question. Scholars have linked the probability of democratic transition and consolidation to economic development;Footnote 1 historical path dependence and revolution;Footnote 2 social learning;Footnote 3 and financial openness.Footnote 4 Some have argued that societies with a highly unequal income distribution – which in turn may be driven by differences in initial factor endowments – are less likely to be democratic, as the rich may prefer an authoritarian regime due to redistribution pressures from a low-income median voter.Footnote 5 Democratic institutions will thus emerge and be sustained in instances of relatively low inequality, as the redistribution from rich to poor is minimal.
This article studies one particular political institution – voting rights – to propose an altogether different reason behind political liberalization: elites may use the franchise as a tool to attract migrant labour. Focusing on the political evolution of the thirteen British American colonies from the early seventeenth century until the American Revolution, I show that those colonies which relied on white immigrant workers, rather than slaves, had a more liberal franchise. For example, the Northern colonies had both a more liberal suffrage and a higher proportion of white settlers, while, in the eighteenth century, the Southern colonies combined a restrictive franchise with a large slave population.Footnote 6 These institutional differences are hardly surprising and have been documented by others. In the North, the staple crop (wheat) could be grown efficiently on small family farms, which may have contributed to a more egalitarian income distribution and, therefore, better political institutions.Footnote 7 Alternatively, it is possible that the temperate Northern climate appealed to entire white migrant families, which may have sought to replicate British political institutions, emphasizing private property and controls on government power.Footnote 8
Although such theories are important, they fall short of explaining a different puzzle which emerges from the data analysed in this article. This puzzle lies in the observation that, despite the hot climate which harboured disease and was suitable for growing labour-intensive cash crops (such as tobacco and rice), the Southern colonies actually had a more inclusive suffrage than the North throughout the seventeenth century. During this period, African slaves were not generally available, so Southern landowners recruited European labour through the system of indentured servitude, under which migrants ‘tied’ themselves to a colonial landowner for several years in exchange for a loan covering transport costs from Europe. I argue that Southern elites initially allowed nearly all adult males to vote in order to attract immigrant workers. A liberal suffrage was highly desirable for such migrants, since assemblies in which they were adequately represented guaranteed fair enforcement of indenture contracts. As in the North, representative institutions also prevented the expropriation of land titles and private property of landowners and newly freed servants.Footnote 9
However, the arrival of slavery in the Southern colonies in the late seventeenth and early eighteenth centuries prompted planters to substitute white labourers with cheaper and easily available slaves. Political concessions to white workers were no longer warranted, so Southern elites immediately tightened the suffrage. In contrast, the Northern franchise remained liberal and largely unchanged throughout the colonial period. Black labourers were only of marginal importance, as agricultural work was done on small family farms without the need for much additional labour.
To test the hypothesis about labour market structure as a determinant of the colonial suffrage, I assemble an original dataset covering the thirteen British American colonies from the early seventeenth century to the American Revolution. In order to capture the quality of political institutions, I create an index that aggregates, for each colony-year observation, the presence of various restrictions on the franchise, such as those related to income, land ownership or religion. I measure labour market composition by tracing the spread of slavery via data on each colony’s share of the black population. I use the percentage of a colony’s white landless population to estimate inequality, where a higher percentage of landless people implies higher inequality. In addition, I control for a rich set of additional variables, ranging from urbanization and population density to colonial settlement patterns.
I use four complementary approaches to show that in the thirteen British American colonies, differences in labour markets had a causal impact on the quality of political institutions. First, I construct a new instrument for the extent of slavery, based on each colony’s propensity to produce one of the two labour-intensive crops (tobacco or rice), interacted with the price of these crops. The idea behind this instrument is that when the prices for labour-intensive crops increased, landowners would seek to acquire additional labour, turning to slaves if white labourers were not sufficient or too expensive. Secondly, I use the prices of slaves in the Caribbean – the largest slave market throughout the colonial period, which set prices for the rest of the New World – as an alternative proxy for the prevalence of black workers in the Southern colonies. Thirdly, I control for time-invariant colony characteristics, such as geographical or climate differences, which may affect both labour markets and the suffrage, by including colony fixed effects. Finally, in the online appendix, I supplement the econometric results with a detailed historical discussion of the link between labour markets and the franchise in the colonies, which also includes (fragmentary) migration data.Footnote 10
This article offers several contributions to the literature. By linking the composition of human capital and representative institutions, I not only enrich the inequality–democracy theories developed in previous work, but also suggest a mechanism through which slavery could affect economic and political outcomes over the long term.Footnote 11 Moreover, I break away from the cross-country regression approach on which other similar studies rely and which has been criticized on account of problems of data inconsistency, unit heterogeneity and endogeneity.Footnote 12 The main advantage of my research design stems from my ability to trace the political and economic development of the thirteen colonies from their very settlement until 150 years later. In addition, I observe substantial variation in colonial political institutions as, from the very beginning, each colony was given complete freedom to shape – as well as alter – its representative institutions without any interference from England.
The idea that political institutions, by serving as a commitment device, can credibly constrain the power of elites has been developed in several earlier studies, including for the cases of the North American colonies and states and ancient Greece.Footnote 13 My findings also resonate with those of a small but growing literature which demonstrates that one reason why the American states with a smaller female population enfranchised women earlier was to attract them as settlers.Footnote 14 A separate strand of research has also linked the quality of institutions to migration. Bertocchi and Strozzi as well as Engerman and Sokoloff argue that electoral institutions mattered for attracting European migrants to the New World in the nineteenth century.Footnote 15 Fleck and Hanssen find a correlation between government quality and the opportunities for population exit in cross-country data.Footnote 16
This article enhances such work in at least three important ways. First, my theory applies the institutional commitment mechanism to a new problem: how to prevent self-interested elites from reneging on migration inducements once costly migration has taken place. Secondly, unlike previous work, my argument considers the joint impact of labour market structure and income inequality on representative institutions. Thirdly, a major strength of the article lies in the use of a unique dataset and research setup which makes identification concerns in the empirical analysis less pressing, something which is generally difficult to achieve with similar historical data.
The remainder of the article is organized as follows. The next section outlines the theoretical argument, while the following section discusses the data. The next two sections present the empirical set up and the results, respectively. A section focusing on validity threats follows, and the last section concludes.
Theoretical Argument
Colonial Labour Market Patterns
The primary goal of American settlement was to remedy the twin problems of scarcity of land and abundance of labour facing England in the seventeenth century. While some colonies were governed directly by the Crown, it was more common for either one or several wealthy Englishmen to gain ownership of a colony, which gave them substantial control over political and economic affairs. Migrant labour was essential for the production of high-profit crops such as tobacco, rice and wheat, which were then exported to Europe.Footnote 17 As a result, the successful development of each colony – and the wealth of its proprietors – depended on attracting English migrants.Footnote 18
Despite the importance of immigration, attracting settlers to the colonies was not easy. Even the poorest labourers had alternatives to moving overseas, such as going to a different part of the countryside, to London or to Ireland.Footnote 19 Moreover, the journey across the Atlantic was long, dangerous and expensive, with its price exceeding half a year’s income for a typical English emigrant.Footnote 20 What is more, information about the difficulties of colonial life travelled relatively easily from the colonies to the mainland, as highly literate colonists wrote often, and even the illiterate could hire someone to write for them.Footnote 21
Colonial elites adopted different strategies for populating the colonies. In the North, the temperate climate encouraged a steady inflow of migrant families. Settlers were given small plots of land, which they used to cultivate wheat without the need for additional labour. In contrast, the hot climate and prevalence of disease in the South, along with an agriculture centred around the production of labour-intensive crops (such as tobacco and rice), made it a less attractive area for migrants.
While some slaves found their way to the colonies in the seventeenth century, transported by Dutch privateers or large planters who had migrated from the West Indies, the supply of black labour was generally limited by the high demand for slaves in the Caribbean, as well as by the tight regulation of the slave trade by the English government. Southern elites, therefore, recruited European labour mainly through the system of indentured servitude, under which migrants ‘tied’ themselves to a landowner for several years in exchange for a loan covering transport costs. Master–servant relations were regulated through contracts, which stipulated the length and conditions of the indenture, wages and any dues, such as land, payable once the servant was free. Given that the end-of-contract payments were substantial – sometimes comprising up to 66 per cent of a contract’s price, servants were often able to set up their own farms once their indentures were over.Footnote 22 As a result, the supply of free white labour in the South was largely inelastic, except at very high wage levels. Recruiting workers from other colonies was also costly not only because planter settlements were separated by thick forests, but also because Southern roads, which were built to accommodate the transport of export crops from plantations to the coast, were of very low quality.Footnote 23
Although the indenture system worked well for most of the seventeenth century, the supply of English servants plummeted towards the end of that period as economic conditions in the mother country improved. At the same time, the availability of African slaves rapidly increased at the beginning of the eighteenth century, driven by the rising English share in the slave trade and the low profitability of West Indies sugar production.Footnote 24 While slavery had a limited impact on Northern labour markets as family farms continued to dominate production, in the South it prompted the relatively quick substitution of indentured servants with slaves. As Table A2 shows, in the early eighteenth century, blacks comprised nearly two-thirds of the population of South Carolina, and nearly one-third of that in Virginia.Footnote 25 Even though Southern planters initially preferred white workers, by 1710 they had learned that slaves were just as productive, and ‘sought them avidly every time a slave ship arrived’.Footnote 26 In fact, landowners used slaves not only for agricultural tasks, but also trained blacks for more highly skilled occupations, such as builders and craftsmen.Footnote 27
The Link between Labour Markets and the Suffrage
This subsection argues that one factor which explains why some colonies opted for a liberal franchise was the degree to which their labour markets depended on white migrant workers. When white workers were difficult to obtain – as in the South prior to the arrival of slavery – colonial governments granted the franchise to virtually all males, in order to attract settlers.Footnote 28 In those colonies where white labour was easier to attract – such as in the Northern colonies – additional suffrage regulations, such as the requirement of owning land or property, were introduced. The most restrictive regimes, which required voters to satisfy a combination of requirements, including landholding, property, taxpaying and residency, were adopted in the South once slaves replaced white indentured servants in the late seventeenth and early eighteenth centuries.
Why was a liberal suffrage important for English migrants? Although all colonies were under English rule, representative assemblies, established soon after settlement, enjoyed significant prerogatives including complete control over law making and colonial finances. Therefore, political institutions in which labourers were adequately represented ensured that rules, regulations and policies that were relevant for them could not easily be changed by those in power. In particular, building on seminal work, I argue that representative assemblies served as a commitment device for any promises made to migrants by ruling elites.Footnote 29
In the North, wheat growing took place on small family farms, so an inclusive suffrage guaranteed not only that new settlers obtained the amount of land promised to them by proprietors before arrival, but also that their land, property or profits could not be taken away once they had settled. In the South, the protection of property rights was relevant for those few settlers who had set up their own farms. More importantly, representative institutions had an additional purpose, namely to aid the enforcement of labour contracts on which the indenture system was based. In order to undertake the long and arduous trip across the Atlantic, potential servants needed a guarantee that land owners would not alter contract terms or expropriate their wages once they arrived in the colonies, or once their contracts were over. Since masters often refused to comply with the terms agreed in the contract, colonial assemblies played an active role in devising laws regulating servant–master relations.
Of course, political institutions are only one of the ways credibly to constrain the power of elites. The rule of law – such as an impartial and independent court system – can be used effectively both to resolve contracting disputes, and to force governments to respect private property.Footnote 30 The development of colonial courts, however, tended to lag behind that of political institutions.Footnote 31 While courts existed in some colonies, judges were seldom impartial as they were appointed by large proprietors. Court independence was also a problem, as judges and assemblymen often ruled on disputes in which they themselves were involved.Footnote 32
The Role of Inequality
Building on the idea that representative assemblies force elites to commit credibly to promises made to migrant workers, the previous subsection has developed an argument linking the composition of labour markets and the extent of the colonial franchise. At the same time, the theory presented here also enriches existing inequality–democracy models. According to Boix, elites are more likely to implement an autocratic system when inequality increases, due to redistribution threats from a low-income median voter.Footnote 33 In contrast, Acemoglu and Robinson argue that the inequality–democracy relationship follows an inverted U-shape.Footnote 34 In their story, the negative effect of inequality on democracy working through redistribution is mitigated by the higher likelihood of democratic revolution when inequality is high. Since democratization occurs as a commitment to future redistribution, the stronger the threat of revolution, the more likely elites will be to cede democracy.
While I examine the independent role of revolutionary threats in the penultimate section of this article, it is possible that inequality, through its effect on redistribution, exerts an effect on the franchise that is independent from that of labour markets. A first question to ask is to what extent redistributive transfers mattered in the thirteen colonies. Although an advanced system of taxation and redistribution did not emerge until after the American Revolution, several percentage points of the average colonist’s income went to pay taxes.Footnote 35 Colonial governments used these proceeds to finance in-kind redistribution (rather than pure monetary transfers), such as poor relief, as well as the building of roads, bridges and schools.Footnote 36 The extent to which the labour market structure affects the relationship between inequality and the suffrage depends on the trade-off between the redistributive costs of a particular franchise regime and its benefits in securing migrant labour. On one hand, liberal representative institutions under high inequality are particularly costly for elites, due to the losses associated with relinquishing assembly control to the poor, such as higher levels of redistribution. On the other hand, an inclusive suffrage may also attract white workers, and thus increase elite incomes when labour is difficult to obtain. Although redistribution in the thirteen colonies was non-negligible, labour market issues were of even greater importance, as the survival and progress of each colony depended on attracting a sufficient number of workers. Therefore, I expect that even though inequality may still have a negative effect on the extent of the franchise, labour markets will be the primary determinant of political institutions.
One concern about this theoretical model is the potential relationship between labour markets and inequality. For instance, the spread of slavery in the South in the early eighteenth century not only decreased the importance of white workers, but also may have led to the establishment of large plantations. This could have exacerbated the income differences between large land owners and the rest of the population, leading to a deterioration in the suffrage in the Southern colonies. However, the historical evidence shows that such a scenario is not entirely convincing, as slavery did not lead to a dramatic shift in the Southern income distribution.Footnote 37 I investigate this concern in more detail in the penultimate section of the article.
Testable Predictions
The preceding discussion suggests that differences in labour market organization across colonies, driven by the varying importance of white workers, may have had a direct effect on the extent of the suffrage. Colonies that needed to attract white labourers opted for more democratic institutions, which served as a commitment device for promises made to migrants. However, the previous subsection has also shown the labour–franchise link is moderated by income inequality, as attracting poor labourers by extending the suffrage also imposes redistribution costs which are particularly high when incomes are unequally distributed.
This framework suggests two empirically testable hypotheses:
HYPOTHESIS 1: Liberal representative institutions will be observed in those colonies in which white labourers are important. The substitution of white workers with slaves will lead to a contraction of the franchise.
HYPOTHESIS 2: Higher inequality will have a negative effect on the colonial franchise. The impact of inequality will be lower than that of labour market structure.
Data
A New Dataset on Representative Institutions
This subsection discusses the data used in the regressions to follow.Footnote 38
Suffrage
To examine trends in colonial political institutions, I have compiled a new dataset which codes, for each colony and annually from the first year of its establishment, all the different restrictions that governments placed on the suffrage. I have drawn on the extensive study of the suffrage in the thirteen colonies by McKinley, which I cross-checked against other more recent work, such as Rusk and Dinkin.Footnote 39 Even though the general reliability of the McKinley study was also confirmed by Stanley Engerman, it is very likely that the dataset measures representative institutions with error. This should be partly mitigated by the fact that historians agree that the suffrage restrictions on the books were binding, suggesting that the franchise rules that I coded should be highly correlated with those that were actually enforced.Footnote 40
I have created a suffrage index to use as the dependent variable in the regression analysis below. First, I extracted a list of all the franchise regulations in force in any particular colony from 1619 to 1775, such as those relating to income, property or religion.Footnote 41 For each colony–year observation and suffrage restriction, I created a variable that takes a value of 1 if voters needed to satisfy this particular regulation, and 0 otherwise. I have not focused on requirements for voting in local elections, as these differed very little from those for colony elections.
Suffrage restrictions were almost always introduced or removed in combination with others, so I aggregated these data into a suffrage index, and have not focused on restriction-by-restriction analysis instead. In order to do so, I needed to account for the fact that colonies that allowed voters to satisfy a choice of two or more voting requirements likely had a more liberal franchise than those which imposed one of those restrictions. For example, while from 1698 onwards, North Carolina required all voters to be freeholders, during the same period Maryland allowed electors who did not possess land to vote if they owned property worth at least 40 pounds. Therefore, I gave the freeholding restriction in North Carolina a weight of 1, while in Maryland the restrictions for (1) freeholding, and (2) income or property, each carry a weight of one quarter. As I assume that having a choice of two voting criteria is only half as restrictive as needing to satisfy one of those criteria, the sum of the two Maryland restrictions is thus one half. A similar approach is adopted when one suffrage requirement can be substituted with two or three others, yielding sums of one-third and a quarter, respectively. I added up all the different restrictions coded in this way and subtracted from this sum the maximum number of suffrage restrictions observed in the sample (6.5) to obtain the final suffrage variable used in the regressions. I gave periods without elections – for instance, when the representative assembly is appointed by the governor – a suffrage value of 0. As a result, a higher value of the suffrage index implies better political institutions.
To illustrate how my coding works, consider the case of South Carolina. From 1669 until 1691, the South Carolina suffrage index takes a value of 5.5, as the colony allowed all freemen to vote (obtained by subtracting a coding for the existence of a single restriction from the maximum number of restrictions in the sample (6.5)). Instead, in 1692, electors had to possess income or property worth ten pounds and to take an oath, so I coded the existence of restrictions for income or property, minimum income or property and oath taking, yielding a suffrage index of 3.5 (obtained by subtracting 3 from 6.5). In 1697, an additional requirement for three-month residence in the colony prior to election was introduced, which decreases the suffrage index to 2.5 (obtained by subtracting 4 from 6.5). In 1704, all previous requirements were kept, but voters were given a choice to satisfy either the ten-pound income or property requirement, or the possession of land of fifty acres. I gave a weight of one quarter for both of the latter requirements, which means that the suffrage index now takes a value of 4 (obtained by subtracting 2.5 from 6.5).
Of course, it must be acknowledged that this weighting scheme is crude. An accurate assessment of the relative importance of each restriction entails comparing, for each colony, the proportion of the population that was eligible to vote under each criterion.Footnote 42 In the absence of such evidence, I adopt three approaches to test the reliability of my coding. First, I experiment with several alternative operationalizations of the suffrage index, including (1) assigning equal weights to all component restrictions (which means that I do not account for any possible substitutions among restrictions); (2) assigning weights of 1/2 and 1/3 (instead of 1/4 and 1/9) to restrictions that can be substituted with one other or two other restrictions, respectively; and (3) using principal component analysis, which creates data-dependent weights.Footnote 43 The correlations with these new indices are quite high (above 0.8), and the empirical results are similar. Secondly, the effect of labour markets on the suffrage also survives when I replace my original suffrage index with a binary variable instead (which takes a value of 1 if any of the following restrictions are in place: income, freeholding, the existence of minimum freeholding or property, tax, residency or religion; and 0 otherwise).Footnote 44
Thirdly, I assemble some fragmentary data on the extent of suffrage in the colonies (covering mostly the eighteenth century) and investigate its correlation with the suffrage index. Figures A2 and A3, available online, show that the relationship is positive, as expected, and ranges from 0.237 if racial restrictions are accounted for to 0.356 if they are excluded. The correlation is also positive for the period 1775–1860 and takes a value of 0.4.Footnote 45 Therefore, I conclude that, while not perfect, the constructed suffrage index should be an informative measure of the quality of representative institutions in colonial British America.
Labour markets
I hypothesized that more liberal political institutions will be observed in those colonies that depend on white labourers, implying that colonies that find it harder to attract white labour should have more democratic regimes. To evaluate this argument empirically, one should ideally compile, for each colony and over time, data on the importance of indentured servants, free white labour, family labour, and slaves, as well as estimates of the demand for and supply of each labour group.
Since such detailed information is not available, I have made use of data on the percentage of each colony’s black population.Footnote 46 While this is a simple proxy based on fairly reliable data, it is an imperfect measure of the whites and blacks actually in the labour force.Footnote 47 Moreover, this measure does not capture differences in white labour scarcity across colonies. To account for the latter, I controlled for the density of the white population in each colony. I also collected colony-level data on life expectancy and the sex ratio of men over women, but I do not have enough observations to include these variables in the regressions,Footnote 48 so I have calculated simple correlations with the franchise index.Footnote 49 While population density is generally insignificant in the regressions – possibly because it proxies for other variables, such as differences in economic development – colonies with more men and a lower life expectancy had a more liberal suffrage, which is consistent with my theory.
I supplement this measure with two additional labour market proxies. First, to capture the arrival of slavery in the South in the late seventeenth and early eighteenth century, I have created a binary variable that is 1 for the South after 1700, and 0 for all other colonies and years. The idea is that such an indicator, albeit crude, would appropriately account for all three events which increased the supply of slaves in the South around 1700: the rising English share in the slave trade; the low profitability of West Indies sugar production; and the decrease in the supply of English indentured servants. Secondly, I also collected data on the prices of slaves in the Caribbean to construct an alternative proxy for the prevalence of black workers in the Southern colonies.Footnote 50 Since the Caribbean was the largest slave market throughout the colonial period, which set prices for the rest of the New World, high Caribbean slave prices should have decreased the demand for – and the number of – new slaves in the South.
Income inequality
According to the presented theory, income inequality, through its effect on redistribution, may be an important factor in determining the quality of colonial political institutions. Unfortunately, income distribution data for colonial America is not available, so I have utilized the percentage of a colony’s white landless population instead.Footnote 51 The argument is that as the percentage of the population who are landless increases, so does income inequality.
While this inequality proxy is clearly less precise than a more direct measure, such as a Gini coefficient, there are several reasons why it should capture at least some of the variation in the colonial income distribution. First, an extensive literature has documented a positive correlation between the unequal distribution of land and an unbalanced distribution of income.Footnote 52 This was particularly true for colonial America, where land – through its use for the production of export crops – was the major source of income and wealth, both in the North and the South. Moreover, historians agree that the main class divide in the colonies was between the few large landowners (colonial proprietors and their friends) and landless labourers (indentured servants in the South, or poor whites in the North).Footnote 53 The wealth of colonial elites, amassed from the very beginning through land grants and tax breaks granted by the Crown, changed little over time.
In addition, the share of the largest land-owning class (middle farmers owning between 100 and 500 acres) stayed relatively constant: for instance, in South Carolina, 60 per cent of landowners held between 100 and 500 acres.Footnote 54 As a result, taking into account those marginal farmers that shifted from landlessness to owning land (and vice versa) should track closely changes in the overall income distribution.
Indeed, the share of the white landless population correlates well with other (less extensive) measures of colonial economic inequality, such as the share of taxable wealth of the top 10 per cent of the population paying the largest wealth tax (the correlation is 0.389).Footnote 55 To dig deeper into the relationship between the white landless share and inequality, I looked beyond the colonial period and exploited land distribution data for the American states covering the period 1860–1920, the earliest period for which such data are available. While there is no information on the percentage of the population with no landholdings, there is data on the number of farms under 10 acres, which should approximate, at least to some extent, owning no land (farm size in the nineteenth-century United States ranged from 41 to 468 acres, with a mean of 175 acres). The correlation between the percentage of farms under ten acres and the land Gini coefficient is 0.589, which again suggests that percentage landless should account for a significant proportion of the variation in the colonial income distribution.
Additional control variables
Differences in economic development across colonies may have an independent effect on the colonial suffrage through at least two mechanisms. Colonies that are richer may not only be better able to afford the redistributive costs of a democratic regime, but higher income per capita may also make it more likely that citizens support the rule of law and democratic elections.Footnote 56
To capture economic development, I have relied on data on colonial urbanization and white population density.Footnote 57 The basic idea is that, in the absence of income per capita data, in the pre-industrial period large urban populations could thrive only in areas with high agricultural productivity and a developed transport network, and that only relatively prosperous areas could generally support dense populations.Footnote 58 Controlling for urbanization is also important for ruling out concerns that a rising percentage of landless persons may be driven by the rise of cities and the decline of agriculture, rather than by an increase in income inequality. While this issue is less relevant for the South, which remained mostly agricultural and relatively rural throughout the colonial period, it matters for the North, where urban centres gained economic importance, particularly in the eighteenth century.Footnote 59 As argued above, white population density may also proxy for the degree of labour scarcity of white workers. In the regressions to follow, I also control for the type of colonial settlement.Footnote 60 While charter colonies originated when the Crown granted patents and a land grant to a corporation or trustees, in other colonies a single or several proprietors were in charge of colonial affairs. In contrast, royal colonies were governed directly by England through an appointed governor. It is possible that proprietary and charter colonies, lacking the burden of royal supervision, may have had a stronger incentive to liberalize the suffrage in order to attract migrants.
Descriptive and Graphical Evidence
Table A1 in the online appendix presents summary statistics for the evidence I collected. For each colonial region, it breaks down the data into two panels: before and after 1700, the approximate arrival of slavery in the South. The third column in each of the two time panels reports the difference in the means across the two groups of colonies. This information is supplemented by the graphical evidence in Figures 1a and 1b. Figure 1a plots, for each Southern colony and using annual data, the suffrage index (left-hand-side panel), and percentage black (right-hand-side panel). Figure 1b does the same for the colonies in the North. For ease of reading, for each year, each graph presents three data points only: the maximum value of the suffrage or percentage black variable observed across all colonies in the region (circle), the minimum value (triangle), and the value obtained using Lowess smoothing (line).
Fig. 1 Evolution of suffrage and labour markets, by region Sources: See text. Notes: A higher value of suffrage indicates a more liberal political system. The black lines are obtained by locally weighted least squares smoothing (Lowess) over all colony-year observations. For each year, each graph presents three data points: the maximum value of the suffrage or percentage of population black variable observed across all colonies in the region (circle), the minimum value (triangle), and the value obtained using Lowess smoothing (dot).
A quick look at the data reveals several interesting patterns. First, while the black population of the South increased rapidly in the late seventeenth and early eighteenth centuries (reaching an average of 40 per cent by the 1750s), the average slave share in a typical Northern colony was less than 10 per cent during this period. Moreover, these differences in labour market structure tracked closely the evolution of political institutions in the two regions. In the seventeenth century, the franchise in the South was in fact very liberal (by around 1.3 restrictions as compared to the North). In contrast, an average Southern colony in the eighteenth century now required voters to satisfy 1.5 more requirements in order to vote, as compared to a colony in the North.Footnote 61 Nevertheless, the most dramatic change was observed in the South–North difference of the share of the black population: before 1700, this difference was only 5.9 percentage points, but after 1700 it increased nearly fivefold.
To study in more detail this reversal in Southern representative institutions, Figure 2 graphs, for each of the Southern colonies, the joint evolution of suffrage and labour markets. Figure 3 does the same for suffrage and inequality. Despite some existing difference in climate, crops and the timing of settlement across the Southern colonies, the regional-level relationships are consistent with this more disaggregated analysis. All four colonies started off with liberal suffrage institutions, but they deteriorated sharply in the late seventeenth and early eighteenth centuries. This political transformation was accompanied by an equally dramatic rise in the share of the slave population, which in all colonies exceeded 25 per cent of the population by the mid-1700s. Slavery was most prevalent in South Carolina, where, in 1700, nearly half of the population was black.
Fig. 2 Suffrage and labour markets within the South
Fig. 3 Suffrage and inequality within the South.
Figure 3 shows that the Southern political reversal is unlikely to have been driven by rises in inequality following the arrival of slavery. While inequality did increase to some extent in Maryland in the early eighteenth century and beyond, this was not the case in the other three colonies: in Virginia, the percentage of white landless persons stayed relatively constant, while in South and North Carolina, it actually went down.Footnote 62 In the North, both suffrage institutions and the composition of colonial labour markets changed little, while inequality shifts do not seem to correspond to movements in the suffrage.
Empirical Setup
Base Model
While the previous subsection has uncovered correlations that are consistent with the presented theory linking labour markets and the suffrage, it is important to examine this relationship in a more systematic way. In order to do so, I employ the following econometric model:
$$\eqalignno{ & {\rm Suffrage}_{{ct}} =\alpha {\plus}\beta _{1} {\rm Suffrage}_{{ct{\minus}1}} {\plus}\theta _{1} {\rm Labour}_{{ct{\minus}1}} {\plus}\theta _{2} {\rm Inequality}_{{ct{\minus}1}} \cr {\plus} & \beta _{2} {\rm Urbanization}_{{ct{\minus}1}} {\plus}\beta _{3} {\rm Density}_{{ct{\minus}1}} {\plus}{\bf X}_{{ct{\minus}1}} \beta _{4} {\plus}\gamma _{t} {\plus}\mu _{c} {\plus}{\varepsilon}_{{ct}} $$
where Suffrage ct is the suffrage index in colony c and time period t; Suffrage ct−1 is lagged suffrage; Labour ct−1 is percentage black; Inequality ct−1 is percentage white landless; Urbanization ct−1 is the share of the urban population; Density ct−1 is population density; X ct−1 is a vector of additional controls (dummies for the years in which each colony was proprietary or charter colony, with royal colony the omitted category); γ t is a year control; µ c is a colony fixed effect; and ε ct is the disturbance term. All independent variables are lagged by one period (three years, see more details below). Since a higher value of suffrage implies better political institutions, I expect θ 1 and θ 2 to be negative. As the share of whites, relative to blacks, increases, and percentage of white landless persons decreases (implying that inequality is lower), democratic institutions are more likely to emerge and survive.Footnote 63
The dataset uses an unbalanced time-series cross-section that covers the years 1619–1774. I have dropped observations for the Northern colonies during Dutch or Swedish rule in order to avoid any confounding effects arising from the differing identity of the colonizers. I also did not include observations for the years 1775–76, as the American Revolution was under way in this period. Furthermore, I excluded all the observations for Georgia, as Georgia was only established in 1732.Footnote 64 Since political institutions are ‘sticky’, I created three-year panels by taking three-year averages for all variables (results were similar when I used shorter or longer averages or an annual panel). I used robust standard errors as clustered standard errors are not accurate when the number of clusters is small, but also experimented with wild bootstrapped errors in Table 3 to account for the small number of clusters and minimize serial correlation, and obtained very similar results.Footnote 65 I captured the effect of permanent colony-level differences, such as climate or geography, by including colony fixed effects, while controls for urbanization, population density and the type of colonial settlement control for time-varying colonial characteristics.
Using a lagged dependent variable is important, since it not only controls for institutional persistence, but also removes serial correlation. Although there may be a bias when colony fixed effects are included together with the lagged dependent variable, this bias is small when the number of time periods per item is more than twenty or thirty (in my case, it is nearly forty periods per colony).Footnote 66 Even so, in the robustness section I have also obtained very similar results with the Arellano-Bond GMM estimator, suggesting that any fixed effects–lagged dependent variable bias is not a source of concern.
One issue about this econometric setup is that, by including colony fixed effects and a lagged dependent variable in the regressions, it may simply capture rapid responses to shifting circumstances very different from the historical events on which the paper focuses. I present results in Table A6 in the online appendix to address this concern: (1) without colony fixed effects; (2) without a lagged dependent variable; and (3) without both. While the coefficients on the labour variable are (unsurprisingly) several times stronger in the regressions without a lagged dependent variable, they are negative and highly significant in all specifications. Secondly, in the same table I have also constructed a ten-year panel, where all regression variables are averaged over ten years, rather than three, as in the baseline specification. Again, I obtained very similar results.Footnote 67
Instrumental Variables Approach
While the developed theoretical framework posits that the composition of colonial labour markets has a direct effect on the suffrage, it is plausible that this relationship may be confounded by reverse causality, with political institutions affecting migration, and thus the future composition of labour markets. In addition, the share of each colony’s black population may be endogenous to a variety of unobserved variables, such as differences in colonial leaders.
I addressed these concerns by instrumenting labour markets with a crop suitability index (crop index), calculated as follows. For each colony, I collected data on its (fixed) propensity to produce one of the two labour intensive colonial export crops (tobacco and rice). I compiled the same information for wheat, which can be grown in virtually any climate without significant amounts of labour. I interacted each colony’s suitability for rice and tobacco with their respective colonial prices, which change over time, and divided this sum by the suitability of each colony to produce wheat interacted with its price (I took the logarithm of the final quantity). Therefore, the instrument measures each colony’s revenue earning potential for tobacco and rice, relative to that of wheat.Footnote 68 Changes in the instrument within colonies are thus driven purely by price shocks to the three crops, while cross-colony differences could also arise because of differing crop suitabilities across colonies.Footnote 69
The logic behind the instrumental variable (IV) is as follows. Once the prices for tobacco and rice increase, landowners located in colonies which are suitable for producing the labour-intensive crops will seek to acquire additional labour. If the supply of English indentured servants is inelastic (as was the case in the late seventeenth and early eighteenth centuries), those farmers will turn to slaves, leading to an increase in the share of the black population. Although I do not have data on the export price of these three commodities during the time period in question, historians point out that the colonies were largely price takers,Footnote 70 so it is unlikely that the prices used in the construction of the IV were endogenous to colony-level factors. I discuss in more detail the exclusion restrictions of the IV in the next section.
I compiled the fixed crop suitability weights from the Food and Agriculture Organization of the United Nations Global Agro-Ecological Zones (GAEZ) 2010 database.Footnote 71 Although the latter data are not available historically, this should be a minor concern, as climate and soil conditions likely changed very little in the United States since colonial times. For each cell of approximately 56 by 56 kilometers, FAO combined information on the physical environment and the requirements for growing 154 crops, assuming different levels of input use and crop management. I made use of the published crop summary tables, which provided crop suitability and potential yield data for each of the American states. I defined land to be suitable for either tobacco or rice if it yields at least 40 per cent of the maximum possible yield for these crops.Footnote 72 I also assumed that cultivation occurs under rain-fed conditions (as mechanical irrigation did not become popular in the colonies and in the United States until after the Revolution), and low input intensity.Footnote 73
Results
Table 1 reports Ordinary Least Squares (OLS) and instrumental variable (IV) estimates for Equation 1, using four different specifications. First, the table shows that there is a negative relationship between the share of each colony’s black population and the suffrage index, both in column 1, which excludes colony fixed effects and inequality, and in column 2, which presents the full specification. In fact, the coefficient on labour markets in column 2 is higher than that in column 1, with inequality having, as hypothesized, a negative effect on the quality of political institutions that is significant but nearly two times weaker than that of labour markets. In addition, the impact of labour markets on the suffrage is quantitatively important. A 10 point increase in the percentage of the population that is black leads to a decrease in the suffrage index of approximately 0.159 restrictions, or close to 4.4 per cent, relative to the sample mean. Because I am also controlling for the lagged value of the suffrage index, it is more appropriate to consider the long-run effect of labour markets: a 10 percentage point increase in the percentage of the population that is black leads to a decrease in suffrage of roughly 0.773 restrictions, or 21.2 per cent.Footnote 74
Table 1 Suffrage in the Thirteen Colonies: Main Specification
Sources: See text.
Notes: The dataset is obtained by taking three-year averages for each colony. Column 4 only uses those colony-year observations with above median inequality. Independent variables are lagged by one period (three years). Additional controls include a dummy for when each colony was proprietary or charter (with royal as omitted category). A linear trend is used as ‘Year control’. Robust standard errors are in parentheses. ***p≤0.01, **p≤0.05, *p≤0.1.
Column 3 instruments labour market composition with the crop index variable. The coefficient on the percentage of population that is black is still negative and significant, and its magnitude is nearly 1.6 times stronger than that in the OLS specification. As argued earlier, since the percentage of the population that is black is calculated using population data, rather than labour force estimates, it is likely to measure the composition of colonial labour markets with error. If this measurement error is classical, its coefficient in the OLS regressions may be biased downward.
In the final column of Table 1, I restrict the sample to those observations that have a value of inequality (percentage of population landless) above the median value in the sample. The negative and significant relationship between the share of each colony’s black population and the suffrage index survives even when inequality is kept constant in this sample of year-colonies with unequal incomes, which makes it less likely that the percentage black is simply a proxy for economic inequality.
The results in Table 1 also suggest that variables such as urbanization, population density, and differences in colonial settlement patterns have limited impact on the quality of colonial political institutions. Unreported estimates in which urbanization and population density were entered separately, rather than together produced very similar results. In estimations which are available upon request, I also allowed inequality to vary non-linearly, which did not produce significant coefficients on the higher-order inequality terms and had little effect on the overall results.
Table 2 shows that the results in Table 1 are robust to using alternative measures of labour markets. Columns 1–3 substitute the share of each colony’s black population with a binary variable that is 1 for the South after 1700, and 0 for all other colonies and years.Footnote 75 Although crude, a significant relationship between this variable and the suffrage index should alleviate concerns that the labour markets–suffrage relationship is driven by something idiosyncratic in the percentage black measure. Not surprisingly, the coefficient estimates are significant, albeit much weaker than those in Table 1.
Table 2 Suffrage in the Thirteen Colonies: Alternative Measures of Labour Markets
Sources: See text.
Notes: Labour is measured as a binary indicator (1 for the South after 1700, and 0 for all other colony-year observations) in columns 1–3, and as Caribbean slave prices (with the sample restricted only to the South), in columns 4–5. The dataset is obtained by taking three-year averages for each colony. All independent variables are lagged by one period (three years). A linear trend is used as ‘Year control’. Additional controls include a dummy when each colony was proprietary or charter (with royal as omitted category). Robust standard errors are in parentheses. ***p≤0.01, **p≤0.05, *p≤0.1.
In columns 4–5, I use Caribbean slave prices as a measure of labour market structure. Since the Caribbean was the largest slave market throughout the colonial period which set prices for the rest of the New World, high Caribbean slave prices should decrease the demand for – and the number of – new slaves in the South. The expected relationship between this measure of labour markets and the suffrage index is, therefore, positive. I prefer to use Caribbean slave prices instead of prices in North America, as the latter are likely to be endogenous to domestic demand.
Because the number of slaves in the North was very small, it is likely that Northern slave-owners obtained black labourers from a different market, and at different prices, than did Southern slave-owners.Footnote 76 Unfortunately, I do not have data on Northern slave prices, so I need to restrict my analysis to colonies in the South. Moreover, the small sample size prevents me from conducting the IV analysis in this specification. Despite these caveats, the coefficient on Caribbean slave prices is – as expected – positive and significant. The estimates in column 5 imply that a one standard deviation increase in the price of a single slave improves the suffrage index in the South by around 0.420 restrictions, or approximately 13.4 per cent relative to the sample mean in the short run.Footnote 77
I test the robustness of these results in Table 3 below.Footnote 78 First, in Table 3 I include colony-specific trends, in addition to the colony fixed effects (columns 1–2). The labour coefficient is slightly smaller in magnitude but still significant, while inequality loses significance. Results are similar when I instead use regional trends (columns 3–4). In column 5, I use an annual panel to calculate the standard errors using the wild bootstrap method,Footnote 79 which is a more appropriate way for modelling within-colony correlation of the errors when the number of clusters is small, as compared to the conventional clustered errors method. Results are analogous to those presented earlier, and also survive when I experiment with other unreported specifications of the errors, such as a Prais–Winsten regression, clustered standard errors, Newey–West standard errors (with one, two or three lags), panel corrected standard errors (PCSEs),Footnote 80 and the inclusion of an additional lag of the dependent variable. The last column in Table 3 implements the Arellano–Bond GMM regression to account for the possible bias arising from the simultaneous inclusion of colony fixed effects and a lagged dependent variable. The coefficient on the percentage of the population that is black is again negative and significant.Footnote 81
Table 3 Suffrage in the Thirteen Colonies: Robustness Checks
Sources: See text.
Notes: The dataset used in columns 1–4 is obtained by taking three-year averages for each colony. Columns 5 and 6 use annual data. All independent variables are lagged by one period (three years for columns 1–4 and one year for columns 5 and 6). A linear trend is used as ‘Year control’. Additional controls include a dummy when each colony was proprietary or charter (with royal as omitted category). Robust standard errors are in parentheses. ***p≤0.01, **p≤0.05, *p≤0.1.
Instrumental Variable Validity and Exclusion Restrictions
The adopted IV strategy would be valid provided that three conditions are satisfied. First, the empirical analysis above illustrates that the relationship between the instrument and the share of each colony’s black population is indeed strong, with F-statistics significantly above 10. Moreover, the effect of crop index on the suffrage must work only through labour markets. More precisely, the IV should not affect the dependent variable directly, or through variables omitted from the model. I examine the validity of each of these assumptions below.
Could the crop index variable have a direct effect on political institutions? For example, a rise in the prices of labour intensive crops will make farmers expand production, and consequently increase their demands for labour. If slaves are not available, planters will respond by liberalizing the suffrage, and thus the crop index might affect the suffrage directly, even if the proportion of blacks in the population is constant. However, the evolution of tobacco, rice and wheat prices reveals that the only crop whose prices might justify a direct link between crop index and the suffrage is tobacco, as its prices rose rapidly during the period 1647–75.Footnote 82 Even so, tobacco planters would have been ambivalent about attracting additional labour, as this upward surge was volatile and relatively short lived. Indeed, when I drop from the regressions all years previous to 1675, the (unreported) results are nearly identical to those presented in Table 1.
A second concern is that the instrument may be correlated with other determinants of the dependent variable, either observed or unobserved. For instance, the crop index may simply capture differences in persistent factor endowments and economic inequality across colonies.Footnote 83 There are at least three reasons that mitigate this concern. On one hand, the inequality variable included in the regressions (percentage of the population landless) should capture at least some of the variation in the colonial income distribution. Moreover, the inclusion of colony fixed effects ensures that the effect of the instrument is not confounded by colony-specific fixed variables. By combining price and crop suitability data, the crop index provides information on each colony’s revenue earning potential for tobacco and rice, compared to that of wheat, without making a statement about the relative distribution of these crop revenues. This is also confirmed by the unconditional correlation between crop index and inequality in the sample (0.250), which is not overwhelming enough to raise concern.
On the other hand, the effect of the instrument on the suffrage could work through some omitted variable from the model. The very low correlation between the instrument and the regression residuals (less than 0.00001), the inclusion of colony fixed effects and time dummies, as well as the multiple robustness tests performed in the previous section and in the online appendix suggest that this is less likely to be the case.Footnote 84
Yet another possibility is that a rise in crop prices may also increase overall welfare, which in turn may directly affect the emergence and stability of political institutions. On one hand, as government revenues (largely based on crop export duties) grow and the costs of redistribution are partially offset, elites may be more likely to extend the suffrage. On the other hand, long-term rises in income per capita may lead to attitudinal changes in the population that make a democratic regime more likely to emerge and to be sustained.Footnote 85 Although some of these effects should be captured by the controls for urbanization and population density included in the regressions, these are admittedly less precise than a direct measure of income per capita, which unfortunately does not exist. An examination of one available income proxy for colonial Maryland (a welfare ratio),Footnote 86 provides some suggestive evidence against these hypotheses. In fact, the disfranchisement of landless whites in Maryland in the early eighteenth century was accompanied by a welfare ratio which was rising, rather than falling.Footnote 87
I conclude that, in line with Hypothesis 1, there is a consistent negative relationship between the proportion of each colony’s population that is black and the colonial suffrage. I also find support for Hypothesis 2. Inequality has a negative impact on the quality of colonial political institutions, but the magnitude of the effect is several times smaller than that of labour markets. In the next section, I discuss three potential validity concerns which may threaten the reliability of these estimates.Footnote 88
Alternative Explanations
The theoretical mechanism outlined above provides one reason why we may observe a connection between the composition of colonial labour markets and the quality of political institutions. Still, there may be several alternative explanations for this finding, which I review below.
The Relationship between Inequality and Labour Market Composition
An important concern about the validity of the empirical results is the potential relationship between labour market structure (proxied by percentage of the population that is black) and inequality (proxied by percentage of the population that is white landless). If percentage black simply captures the spread of plantation agriculture and thus increasing income inequality in the South, as compared to the North, the presented results will be spurious.
Although I attempt to address this concern by including a control for income inequality in all regressions, it is plausible that the only available inequality proxy that I have (the percentage of each colony’s white landless population) is less precise than a more direct measure (such as a Gini coefficient). Although detailed income inequality statistics for colonial America are not available, the existing evidence shows that the relationship between inequality and slavery during this time period is not clear cut. This is not surprising, as large Southern landowners were able to amass significant fortunes, as compared to indentured servants, even before slaves became widely available. In a companion paper,Footnote 89 I have traced the evolution of income inequality in Virginia from the early seventeenth century until the American Revolution, using a variety of sources such as land transfer records, wills, height data, and the distribution of slave holdings to show that slavery benefited equally both small and large farmers, and that income inequality in the colony did not increase substantially in the eighteenth century. What is more, on the eve of the American Revolution, the wealth distribution in New England was actually more unequal than that in the South.Footnote 90 In 1774, the richest 10 per cent of people who left probate records held 57 per cent of the net worth in New England, compared with 49 per cent in the South and 42 per cent in the Middle colonies.
Even so, it is instructive to investigate the relationship between the percentage white landless and percentage black in my dataset. If the arrival of slavery increased inequality because it gave the rich an opportunity to establish large plantations, then such a relationship should be particularly pronounced in the Southern colonies. Graphical evidence shows that there is a weak relationship between inequality and labour in the South.Footnote 91 Although the unconditional correlation between inequality and labour in the full sample is somewhat higher than that in the South (0.322), a regression analysis shows the lack of a relationship between the two variables.Footnote 92
The Threat of Revolution
A second explanation predicts that elites will extend the right to vote to the masses in order to commit themselves to economic concessions in the face of organized mass resistance.Footnote 93 However, several reasons weaken the conclusion that Southern elites implemented a restrictive political regime which favoured the landholding class only after revolutionary threats from the masses subsided in the early eighteenth century.
In Virginia, a series of rebellions in the period 1660–83 – such as the servant uprising of 1663, Bacon’s rebellion in 1676, and the tobacco cutting riots of 1683 – included poor whites and indentured servants who protested against their gradual loss of political rights.Footnote 94 However, subsequent gains in the franchise were short lived. For example, the 1676 decision to extend the suffrage to all freemen (rather than only freeholders) was rescinded in the same year, and only freeholders and householders could vote until 1683. Although after 1684 a new wording of the law did not prohibit landless freemen from voting and contained no punishment for irregular voting, essentially allowing all freemen to vote, this law was repealed in 1698, again granting the suffrage only to freeholders.Footnote 95
What is more, Virginia’s most liberal political institutions were adopted in the early seventeenth century, at a time when there were in fact few uprisings. More importantly, it was the switch from indentured servitude to slaves in the early eighteenth century that prompted whites to view blacks as the main threat to the colony’s tranquillity, thus underlying Virginia’s stability.Footnote 96 What is key, however, is that the shift to slavery also drove the tightening of the franchise in eighteenth-century Virginia. Slavery benefited not only large planters, but also smaller farmers who could afford even a few slaves. As the latter class expanded and prospered and the number of indentured servants and poor whites dwindled, the interests of big and small landholders became more closely aligned. This not only prompted the restriction of the suffrage to freeholders in 1699, but also made it increasingly costly for small farmers to participate in lower-class collective action, which improved the security situation in the colony.
Evidence from other colonies further questions the idea that colonial representative institutions were driven by revolutionary threats. In Maryland and the Carolinas, uprisings reflected disagreements within the elite as well as religious tensions, rather than franchise considerations. For instance, in the 1680s, opposers of Maryland’s proprietary government cited a list of grievances, such as Lord Baltimore’s appointment of his relatives for public offices; proprietary appointment of sheriffs; excessive fees and taxes; as well as anti-Catholic sentiments against the proprietors by the Protestant colonists. The property requirement for voting set in 1670 was also mentioned, but colonists complained about the fact that it was set by the proprietor, rather than the legislature. In North Carolina, the 1677 Culpeper Rebellion sought to stop payment to England of customs duties on exported tobacco, while Cary’s Rebellion (in 1704) resulted from a contest over the office of deputy governor. In South Carolina, political instability was rooted in disagreements between the proprietors and an anti-proprietary faction. In 1690, Seth Sothel, backed by nearly 500 people, staged a coup against the proprietor, while in 1719 there were clashes between the Assembly and the proprietor motivated by the desire of the colonists to overthrow proprietary rule.Footnote 97 In New England and New York, the Glorious Revolution in England prompted uprisings in Boston and New York City. Although in Massachusetts a property basis of the suffrage replaced the earlier religious qualifications, the franchise still remained rather restrictive.Footnote 98 There were no changes to political institutions in New York. In sum, a possible link between representative institutions and changing revolutionary threats is less than clear.
The Importance of Culture
In this subsection, I examine whether differences in migrants’ cultural backgrounds affected the quality of colonial political institutions, and, in particular, whether colonies with a wide franchise had a majority of inhabitants coming from more politically liberal European countries.
There is little data on the country of origin of each colony’s population until the first US census conducted in 1790. What we do know, however, is that by 1680, nearly 90 per cent of colonial Americans were English, and that by the early eighteenth century, more and more non-English migrants were arriving in the colonies, with the largest groups including the Scottish-Irish, Germans and Dutch.Footnote 99 An argument emphasizing the importance of culture, however, is less compelling, for at least three reasons.
First, no European settler experienced a suffrage as broad as the one granted in the Southern colonies. In England, the forty-shilling freehold requirement excluded nearly 98 per cent of adult males,Footnote 100 while in the Habsburg empire, no popular parliament existed, as Habsburg family members were routinely elected to the office of Holy Roman Emperor by seven hereditary electors. Similarly, although the Dutch republic had a relatively strong federal parliament (the States General) and a weak executive by the standards of late medieval and early modern period, popular political participation was limited. Members of the States General were selected by the provincial parliaments, selection for which was in turn limited to a list of regential families.Footnote 101
Second, any cross-colony cultural differences which were either fixed over time or followed linear time trends specific to each colony should be captured by the inclusion of colony fixed effects (present in most regressions) as well as colony time trends (see Table 3). Of course, it is possible that the rate of immigration of particular groups was non-linear. For instance, the arrival of large numbers of Scottish-Irish settlers in the South in 1717–75 may have been behind the decision of colonial elites to tighten the franchise. These migrants came from ‘lawless’ regions in Ulster, and were considered to be more violent and less hard-working than the Germans or the Dutch.Footnote 102
Conclusion
Why do elites choose to extend the suffrage to the masses? Using detailed data from the thirteen British American colonies, this article proposes one answer to this question: elites may wish to extend the right to vote to the lower classes in order to attract migrant workers. I develop a theory in which a representative assembly serves as a commitment device for any promises made to labourers by those in power, such as those related to compensation, labour contracts, or the allocation of land grants. While the theory also accommodates a negative relationship between income inequality and the extent of the franchise (as argued by earlier literature), an examination of both quantitative and historical evidence reveals a weaker and less robust impact of income inequality.
I have tested the theory by analysing a new time-series cross-section dataset from colonial British America covering more than 150 years. Colonial governments enforced a liberal franchise only in those colonies that relied on white workers, such as the seventeenth-century South and the North, and immediately rescinded political concessions once white labourers could be substituted with black workers, as in the South after the arrival of slavery. To rule out alternative explanations, I constructed a new instrument for the composition of colonial labour markets, experimented with several different measures of labour markets, controlled for a wide range of variables, including income inequality, and included colony fixed effects in all specifications.
My work helps identify the specific conditions under which institutions can be malleable and under which they persist, a fundamental research and policy question about which we still know relatively little. By emphasizing the link between the quality of labour and that of political institutions, I have suggested that democracy is less likely to emerge and survive in countries where economic production relies on easily replaceable, low-skill workers who have limited bargaining power vis-à-vis the ruling politicians. Shedding light on where such differences in labour markets come from, both across space and throughout history, may therefore be crucial for understanding how democracies originate and evolve.
