Revisiting the Profitability of Slavery

Slave Hiring Rates and the Return on Investments in Slaves in the Antebellum US South

In: Journal of Global Slavery
Klas Rönnbäck University of Gothenburg Sweden Gothenburg

Search for other papers by Klas Rönnbäck in
Current site
Google Scholar
View More View Less
Open Access


This article revisits the scholarly debate on the profitability of historical slavery. The article examines the case of the antebellum US South, using slave hire rates as a proxy for the net rent on investments in slavery. It employs empirical data and a more advanced methodological approach to the issue than in previous research. The results suggest that the profitability of slavery was much higher than what most previous research has shown, around 14–15 per cent per year on average after adjusting for mortality risk, but that the return also fluctuated over time. It was on average more profitable for Southern capital owners to invest in slaves than investing in many alternatives such as financial instruments or manufacturing activities in the US South, as long as slavery remained a legal institution.


This article revisits the scholarly debate on the profitability of historical slavery. The article examines the case of the antebellum US South, using slave hire rates as a proxy for the net rent on investments in slavery. It employs empirical data and a more advanced methodological approach to the issue than in previous research. The results suggest that the profitability of slavery was much higher than what most previous research has shown, around 14–15 per cent per year on average after adjusting for mortality risk, but that the return also fluctuated over time. It was on average more profitable for Southern capital owners to invest in slaves than investing in many alternatives such as financial instruments or manufacturing activities in the US South, as long as slavery remained a legal institution.

1 Introduction

It was long held that slavery was a rather irrational institution in economic terms; slaves were thought to be expensive to purchase, and by many believed to be less productive than non-enslaved workers, and therefore could be quite unprofitable for many masters to own. This idea that slavery was unprofitable, and hence economically irrational, has generated a long debate on the profitability of businesses linked to slavery, ranging from companies and merchants involved in the transatlantic slave trade to the profitability of sugar plantations in the Caribbean or cotton plantations in the United States. This research has clearly refuted the idea that the exploitation of slaves was unprofitable in general. At issue in modern-day research is, instead, just how profitable slavery was in comparison with other relevant investment opportunities. Simply put, would an owner of slaves have gained more from selling the slaves and instead investing his capital in something other than the slaves, such as in corporate or government bonds, or in manufacturing industries?

This article contributes to this literature by studying the profitability of slavery in the case of the antebellum United States. This is arguably a critical case for our understanding of how profitable slavery could be, as this was one of the most important economic sectors in the nineteenth-century Americas. Methodologically, the present article extends upon previous research by Robert Evans, who used the cost of hiring a slave as a proxy for the net rent that a slave-owner could make from his ownership of the slave. While Evans’ method generally was well received by other scholars, his empirical research was highly criticized for possible biases in the sample of data collected on slave hire rates and has for that reason largely been dismissed in the historiography on the economics of slavery. This article makes four novel contributions to this field of research. Firstly, Evans’ empirical data on slave hires is complemented with data from three other datasets on slave hire rates to analyze whether critics of Evans’ estimates were right. The datasets have all been collected independently of each other, allowing for an analysis of the possible biases in them. Secondly, the gains and losses on the capital invested in the slaves—due to generally rising prices of slaves during the period under study, but also changes in the value of the individual slaves—is included in the estimates. Thirdly, the datasets employed allow for constructing a time-series (instead of point estimates from particular years as in much previous research), and thereby allow us to analyze how the profitability changed over time, from the early nineteenth century to the outbreak of the Civil War in 1860. It can thereby help settle the controversy over whether particular benchmark years studied in previous research were atypical or not. Fourthly, the estimates incorporate an analysis of the cost of slave life insurance in order to control for the mortality risk of slaves, when comparing the return on investments in slaves with other alternative economic opportunities.

2 Previous Research on the Profitability of Slavery in the Americas

In recent years, there has been an intensive interest in the role slavery played in the development of modern capitalism, in Europe as well as in the Americas. Some scholars have argued that slavery played a key role for the development of capitalism, most importantly in Southern United States during the nineteenth century.1 Other scholars, who in many cases have employed a broader geographical approach to the issue, have instead argued that slavery at most played a marginal, or perhaps even a negative role, for the development of modern capitalist societies.2 The recent controversies bring back several issues concerning slavery that have preoccupied scholars of historical slavery for decades. One such key issue, and the one that will be in focus in the present article, is how profitable slavery really was for the traders and masters involved. A classic idea held that slavery was (or at least could be) unprofitable. Slave masters did not necessarily have a capitalist mindset, or the intention of maximizing their profits, but owned slaves for other reasons such as embracing tradition or a hierarchical social order as important values.3 But if slavery was an unprofitable relic of pre-capitalist societies, other scholars asked, why did it manage to survive for so long? While slavery was multifaceted, with social, cultural, religious factors all playing a part in the development of the institution, economic factors were certainly at the heart of the institution. This has spawned a wave of research into the profitability of the transatlantic slave trade, of slavery in general in the Americas, and in particular of slavery in the United States. I will here discuss this research in that order.

2.1 Transatlantic Slave Trade

Many scholars have attempted to study the profitability of the transatlantic slave trade. Most of these studies have been based on the returns from particular samples of ships involved in the slave trade during the eighteenth century. The estimated return on slave voyages varies greatly depending on the sample studied, ranging from –1 per cent to +30 per cent on average per voyage.4 Other scholars have instead attempted to undertake a cost-revenue analysis on a macro-level. The variation between estimates is somewhat smaller using this method—the estimates are all in the range of 7 to 32 per cent per year—but the range is still large depending on the data employed and the assumptions made in these calculations.5

2.2 Slave Labor Exploitation in the Americas Outside of the United States

William Sharp studied the profitability of exploiting slaves in mines in Colombia during the eighteenth century, finding that the profits varied greatly, but exceeded 10 per cent per year for most of the mines studied.6 Most research has, however, focused upon the profitability of slave labor on plantations. Some of this research has then focused upon plantations cultivating sugar in the Caribbean and in Brazil, particularly during the eighteenth century. The results suggest that average rates of return around 5–10 per cent per year might have been quite normal.7 Taking the changing market value of the plantations themselves into account, the total rate of return on investment in these plantations might have been in the range of 15–20 per cent per year during the eighteenth century.8 Nicholas Radburn and Justin Roberts have also studied the practice of hiring out slaves in the form of “jobbing gangs” in the British Caribbean in the eighteenth century. Anecdotal evidence suggests that the profitability of owning and hiring out a slave could be very high, around 15 per cent or even higher per year.9

2.3 Slave Labor Exploitation in the United States

Much research has finally been dedicated to studying the profitability of slave plantations in the antebellum United States in particular. The research was pioneered by Alfred Conrad and John Meyer.10 Other scholars have later returned to this issue.11 A common thread in many of the studies in the field is that they attempt to estimate the profitability from the value of the output of cotton plantations. The results of the key estimates are all in the range of 4.5 to 10 per cent return per year.12 Estimates of the profitability of sugar plantations in antebellum Louisiana suggest an average profit of around 9–10 per cent per year.13 If the estimates from many of the cotton plantations would be correct, the profits from exploiting slaves would many times have been lower than what an owner of capital could have gained from investing for example in financial instruments, such as corporate or government bonds, in the US South.14 This begs the question of whether the slave masters were economically irrational, or whether some of these previous estimates of the profitability of slavery are misleadingly low. The estimates are in all cases based on a number of assumptions, for example about the productivity of the slaves, or the average price of the output. Much of the discussion on this previous research has thus been concerned with how realistic some of these assumptions really were.15 Another problem with this previous research is that most of the studies only calculate data for one particular year (or at best for a couple of years). The cotton yields, and hence the profitability of the cotton plantations, for these particular years might be (and have by some critics indeed been claimed to be) very unrepresentative of average levels.

A completely different approach to studying the profitability of slavery in the antebellum US South was instead suggested by Robert Evans: to use data on slave hires as a proxy for the net rent that a slave owner could get from owning a slave.16 The owner of a slave would not have hired out his/her slave if it would have been more profitable to exploit the slave’s labor on the owner’s plantation. If it, on the other hand, would have been highly profitable for owners to hire out their slaves, many other slave-owners would undoubtedly also have opted for this, so that an increased competition on the hiring market would have reduced the slave hire rates. Slave hire rates might thus be used as a proxy for the rent that an owner could expect from his/her ownership of a slave. Put relative to the capital investment in the slave (proxied for example by the purchasing price), it is possible to estimate the relative profitability of the investment in slavery. For this purpose, Evans collected a large sample of data on the cost of hiring a slave from the records of railroad companies and other similar institutions. Evans’ data suggests that the return on investments in male slaves was considerably higher than previous estimates based on plantation data, in the range of 9 to 17 per cent per year.17 Sarah Hughes study, of a much smaller sample of slave hires, suggest an average annual return of 17 per cent of the appraised value of the individual slave for male slaves, and 9 per cent for female slaves.18 Several scholars have agreed that Evans’ method essentially was sound, while others have criticized his empirical data.19 Butlin for instance pointed out that most of Evans’ data date from the period 1852 to 1860, and that the sample is very limited for other years.20 More important is that there might be some systematic biases in Evans’ dataset. Firstly, the dataset might include some skilled slave laborers, so that the hire rates for average slaves might have been lower than what Evans estimated.21 Secondly, the hire rates in Evans’ dataset most probably include a risk premium, as many owners might have considered it more risky to hire slaves out to railroad companies and similar (often urban) ventures than it would have been to have them work on plantations.22 Criticism of Evans’ data was thus raised by several scholars, but was never substantiated by empirical evidence showing that the sample in reality was as biased as the critics claimed. Even though the criticism was unsubstantiated, Evans’ study has received very little attention ever since.23

3 Historical Context: Slave Hiring in the Americas

The practice of slave hiring was by no means novel in the nineteenth-century United States, nor was it unique to this country. Nicholas Radburn and Justin Roberts have traced the origin of slave hiring in the British Caribbean to the early eighteenth century, finding the first piece of evidence from Barbados in 1708.24 The practice seems to have developed in the United States around the same period of time. Jonathan Martin has traced the origin of the practice within the United States to the early eighteenth century. In 1712, the institution was prevalent enough in South Carolina to merit legislation. Evidence of the practice can also be found in Virginia from 1718. In Georgia, slave hiring can at least be dated to 1733.25

Slave hiring was initially a response to a demand for temporary labor, particularly for the harder or more dangerous tasks on a plantation. Slave hiring did not, however, become very important in many parts of the United States, such as the Chesapeake, until the 1750s. This shift has been attributed the growing importance of slave hiring in this particular region to a structural change in the Chesapeake economy at the time when many farmers switched from growing tobacco to growing wheat. As wheat cultivation has greater seasonal fluctuations in labor demand than tobacco cultivation, it became increasingly common for planters to hire slaves during the labor-intensive harvest.26 Farmers in the Chesapeake might also have pioneered the emergence of annual slave hires.27 This practice then spread across the South, and in many counties became the most common type of slave hiring contract.28 Slaves were then often hired starting on 1 January, New Year’s Day, in auction-like settings. Other contracts were entered into privately with family, friends, neighbors or business partners at various times of the year. Newspapers carried ads both from people wanting to hire, as well as from owners wanting to hire out, slaves.29 By the late eighteenth century, hiring agents start to appear in the primary sources, brokering contracts between hirers and owners. The agents also helped spread the practice of slave hiring geographically and socially, so that owners no longer needed to hire out slaves only through their own private networks.30 The slaves could be hired to perform a wide range of tasks, varying from domestic work, or working on plantations, to woodcutting, mining or railroad construction.31 During the nineteenth century, slave hires became increasingly common in Southern United States. Estimates differ, but scholars have suggested that from 5 to 15 per cent of the total US slave population might have been hired out every year.32 Estimates from the British Caribbean suggest that a similar share of the slaves might have been hired out annually there.33

While the practice might have developed in response to temporary (often seasonal) demands for labor, over time it developed into a much more complex institution. One of the key advantages for the hirers was that it remained a more flexible way of acquiring labor than purchasing a slave.34 At the same time, hiring a slave was in practice much cheaper than hiring a free laborer in the antebellum South, even when the costs for goods that a master was to provide to the hired slave (such as clothes or food) is taken into account.35 The practice could also be beneficial for financially strained small farmers or entrepreneurs, as they then would not have had to put up all the capital necessary to purchase a slave all at once.36 Clement Eaton has also found examples of persons who refused to own a slave because of “antislavery convictions”, but at the same time had no qualms hiring a slave from another master.37 The practice of slave hiring could even be found in ostensibly free states, such as Illinois.38

From the slave-owners’ perspective, flexibility was also a key advantage. Slave hiring was a means of reducing expenses for or earning money from slaves that—for shorter or longer periods of time—were superfluous to the owner’s own labor needs.39 Several scholars have also stressed how investing in slaves to be hired out could be a part of a lifecycle investment for non-planters, perhaps most importantly for overseers; investing first in slaves so that they, by hiring out the slaves, could accumulate enough capital to eventually also invest in a plantation of their own.40 Slaves could also be hired out by guardians administering estates on behalf of minors who had inherited slaves.41 Other agents simply found that investing in a slave was a way of supplementing their own income from other types of work, or to fund their retirement, without having any intention of ever becoming planters themselves.42 For some owners, hiring out a slave could also be a means of getting rid of slaves that they had troubles managing themselves, while still making a profit from the slave’s labor.43 A predominant motivation for many owners to hire out slaves throughout the antebellum South simply was, however, as Martin for example has put it, “to increase returns on their slave capital. […] Hiring out slaves made good economic sense, by providing returns that could rival those on other business ventures in the South.”44

4 Method

Estimating the profitability of any economic or business venture is fraught with problems. In focus here is the private profitability for the slave-masters, not including what some have called the “external effects” of slavery in the form of social costs (such as institutions for the surveillance and repression of slaves) nor the huge economic, social and psychological costs for the slaves themselves.45 In this article, the private profitability of slavery to the slave masters will be estimated via the total rate of return on investments in slavery. This rate of return will be estimated using data on slave hires as a proxy for the net rent earned from the capital investment in slaves. The method was pioneered long ago by Robert Evans. The method thus uses these net rents to estimate the annual return on investments by calculating the ratio of the net rent to the capital investment in the slave (for example proxied by the purchase price of slaves).

The main advantage of the method, Evans himself argued, is that the estimates are based directly on market data rather than being residuals, and that it requires observations of comparatively few variables, compared to alternative methods employed in the literature.46 An additional factor worth emphasizing is that the method allows for constructing annual time-series, in contrast to previous research, which only has been based on one or a few benchmark years. This allows for estimating changes in the profitability of slavery over business cycles (taking inflation into account).

Methodologically, the current article will make some key contributions to this literature. First, the total rate of return will be calculated, including not only the net rent on the capital invested (proxied by the slave hire rates), but also the capital gains or losses made by the slave-owner. And second, four different datasets on slave hire rates will be employed as proxies of the net rent from owning a slave. The four different datasets are employed in order to try to control for the possible biases in the individual datasets, suggested in previous critique of Evans’ estimates. Including both capital gains/losses and net rent in the calculation is in line with how the return on investments generally is calculated in modern research on financial history.47 Put in the form of an equation, the return on investments—or shorter, the profit—will be estimated in the following way:


4.1 Capital Investment in Slaves

In order to know the return on investment in slaves for the slave-owners, we first need to know how high the capital investments in slaves were. The purchase price of slaves in the antebellum South will here be used as a proxy for this. This has been studied empirically by several scholars previously. A key series of data, summarized by Richard Sutch, refer to the price of “prime male field hands” in New Orleans.48 Attempts have also been made to estimate the price of the average slave population throughout the South, taking the gender and age-profile of slaves into consideration.49 In line with this research, the price of an average male slave is here assumed to have been 90 per cent of the price of a “prime field hand,” based on previous research on the valuation of slaves, and the price of female slaves is assumed to have been 75 per cent of the price of a male slave.50 There were also geographical differences between the Upper and Lower South. Here, the price ratio between the Upper and Lower South has been calculated for benchmark years reported in previous research, and the ratio has been interpolated for the years between these benchmarks.51

4.2 Capital Gains/Losses

Secondly, the capital gains or losses in one year is calculated as the product of the change in slave prices because of fluctuations on the market for slaves, the change in capital value due to slaves born during the year, and the change in the capital value due to ageing of the individual slaves. The yearly changes in the price of slaves on the market is gathered from the same sources as discussed above. The change in the capital invested in slavery due to the birth of slave children has been included in some previous research.52 For simplicity, an assumption will be made that the capital invested in female slaves (only) increased equivalent to the annual growth of the of the slave population.53 The owner of a slave would, on the other hand, experience that the value of capital invested in slaves decreased due to ageing (and/or eventual death) of the individual slaves. In general, mortality rates were not particularly high among adult slaves.54 The mortality rates among slave children were certainly much higher than among non-slave children, but the average mortality rates among adult slaves was on par with the mortality rates of non-slaves of the same age.55 The mortality rates did, however, depend upon location and type of work, so that mortality rates among slaves forced to work on sugar plantations in the antebellum South for example was much higher than among the slave population in general.56 The estimates here will not take such differences into account, but only work with more general estimates of mortality rates. Previous research on the age-profile of slave prices allows for calculating some rough proxies for this change in the valuation of slaves.57 Judging from this data, a slave was in general considered to be in the prime of his life from the age of 20 to the age of 40. As their labor could be exploited for many years, it is presumed that the category of “prime” slaves was valued 1 per cent less per year of age on average relative the prime valuation. If we include all (male) slaves, the prices of older slaves (particularly 50 years and older) quickly dropped towards zero, so it is assumed that the valuation of a slave decreased 3 per cent per year of age when all (male) slaves are studied. Female slaves seem to have been valued to have a much shorter prime period of their life, from the perspective of the masters, so it is here expected that the value of female slaves dropped 5 per cent per year of age.

4.3 Net Rent

The hiring rates of slaves will be considered as proxy for the net rent that an owner of a slave could make from his ownership. As was noted above, critics have argued (but never substantiated with evidence) that Evans’ own dataset might suffer from biases in the form of skill and/or risk premia. Four different datasets of slave hire rates will for that purpose be employed in this study, in order to try to control for such biases. The data has been collected independently of each other and from different sources by Robert Evans, Robert Fogel and Stanley Engerman, Claudia Goldin and Robert Margo, respectively.

The sample of data collected by Robert Evans includes 5,768 observations of year-long contracts. Virtually all observations (94 per cent) come from the Upper South.58 Evans also collected data on monthly slave hires (in total 883 observations, 61 per cent of which are from the Upper South). There are only male slaves in Evans’ dataset. Most of the observations are from institutions such as railroad companies hiring slaves from the owners. The dataset does not report the occupation of the slaves hired, so Evans eventually included data on slave hires when “the source indicated that it probably represented a healthy adult male performing relatively unskilled labor.”59

The dataset collected by Robert Fogel and Stanley Engerman is made up of 20,253 observations, collected from probate records of slave-owners. The observations come primarily from rural areas of the South; 39 per cent from the Lower South and 61 per cent from the Upper South.60 62 per cent of the slaves hired out were male, and 36 per cent female (the remainder of unknown gender). Most of the observations (89 per cent) are for yearly hires, a tiny fraction of the contracts covers a period longer than one year, and the remainder are for shorter terms (from one to eleven months). The dataset does include variables for the occupation and age of the slaves hired, but data is missing on these variables for the overwhelming majority of the observations (99.8 per cent and 97.2, respectively). The dataset also includes a variable with information on whether the slaves had a handicap in some way (e.g. aged, disabled, or sick). The vast majority (87 per cent) were not reported to have any handicaps, and were thus presumably able-bodied adults. Chad Dacus has noted that the dataset includes some observations where the reported hire cost is negative, speculating over whether these in reality might have been gifts rather than actual slave hires.61 Outliers such as these have been excluded from the sample used in this article.

Claudia Goldin’s dataset is based on data from urban agents in cities in Virginia hiring slaves from the owners, for five benchmark periods from 1820 to 1860. The dataset is much smaller than the three other datasets (in total 381 observations, 65 per cent male and 35 per cent female).

Robert Margo’s dataset, finally, is based on contracts from US government forts hiring slaves from the owners for civilian work at the fort. All slaves hired by these forts were male. The sample is made up of 5,006 observations of slave-hires, out of which 66 per cent were for monthly slave hires and 34 per cent for daily slave hires. The vast majority of the observations (86 per cent of the sample) come from the Lower South. This is particularly the case for monthly slave hires, where 98 per cent of the observations are from the Lower South. In contrast to the other datasets, Margo’s data contain information on the occupational categories of all the slaves hired. The largest group (48 per cent of the sample) were simply categorized as “laborer.” Other important occupations among the slaves hired by the forts included carpenters and teamsters.

The annual data is the most direct measure to use in order to estimate the return that an owner could realize from owning a slave, for two reasons. First, the annual data shows what an owner actually received for hiring out a slave for a whole year, making it easy to calculate the annual net rent from the investment. Hiring rates for shorter periods as a rule seem to have been higher than for annual contracts, so a slave owner could temporarily earn even higher rates of return than the ones estimated in this study. These temporarily high returns could, however, come at the price of not being able to hire out a slave all the time. There might in addition have been seasonal fluctuations in the slave hiring rates.62 Secondly, the annual slave hire contracts also seem to be the most standardized, so that someone hiring a slave for a full year as a rule would have to pay for the subsistence cost of the slave, including food, housing and medical bills. For shorter hires—daily or weekly—these costs would instead often (but not necessarily always) fall upon the owner of the slave.63

The data from the four datasets are all unbalanced time-series. 97 per cent of all observations in Evans’ dataset are dated from the period 1852–1860.64 92 per cent of the data in Margo’s dataset on the other hand comes from the period 1835–1842. Fogel and Engerman’s dataset is the least concentrated to one particular sub-period, and as a rule has fair samples of data (more than 20 observations per year) from the 1810s onwards, and large samples of 100 observations or more from the 1830s onwards. Geographically, the data is reported by state in the antebellum South in three of the datasets, and by broad region—the Upper and the Lower South—in the fourth (by Evans). Due to the comparatively limited sample sizes, the data has in this article been aggregated by broad region in all datasets. Alabama, Florida, Georgia, Louisiana, Mississippi, North Carolina, and South Carolina are then included in the Lower South, whereas Arkansas, Kentucky, Maryland, Oklahoma, Tennessee, and Virginia are included in the Upper South.

Finally, a very important issue is whether the slave hire rates in the samples were representative of average slave hire rates, and whether the slaves hired were representative of the total slave population. This includes several aspects, including risk and skill premia, possible discounts, and selection effects among slaves hired.

  • Risk premia. Claudia Goldin have pointed out that slave-owners might have considered hiring slaves out to railroad companies and similar ventures (constituting Evans’ dataset) to be very risky, and therefore demanded a risk premium for such hires.65 The risk of slaves running away might, however, have been much higher in urban areas, so Goldin’s dataset might also suffer from such a risk premium, compared to the profit a slave-owner could expect to make in rural parts of the antebellum South. Having slaves work in government forts might likewise have been perceived as riskier for the owner than exploiting their labor on plantations, so it is not possible to rule out the possibility that Margo’s dataset on slave hire rates, too, includes a certain risk premium.

  • Skill premia. Previous scholars have noted that Evans’ dataset might contain some more highly skilled individuals, as Evans was unable to always control for this, so that the slave hire rates also include a certain skill premium.66 Several of the urban slaves in Goldin’s study were also skilled, so it is also possible that these hire rates also include a certain skill premium.67 It is not possible to rule out that this might be an issue in Fogel and Engerman’s dataset, too, as data on the occupation of the slaves hired out in this dataset is missing for the overwhelming majority of the observations in the sample. The only dataset where a skill premium can be ruled out with certainty is the one collected by Margo, as the occupations are clearly recorded for all entries in this dataset, allowing for selecting to use only (unskilled) laborers.

  • Discounts. While many of the hiring contracts were entered into by parties with little or no social relations outside of the market, for example via hiring agents, some slaves were hired by relatives, friends, or neighbors of the owner. In such cases, the person hiring the slaves might have received a discount on the hiring rate, thus exhibiting a downward bias compared to going market hiring rates. This is potentially the case primarily in the dataset of rural hiring rates collected by Fogel and Engerman. Indeed, as was noted above, the dataset contains some outliers with extremely low (in some cases thus even negative) hiring rates, supporting the hypothesis that this might have been the case at least for some observations. Primarily, the institutional hirers underlying the datasets collected by Margo and Evans, respectively, in contrast probably suffer little from such bias.

  • Selection effects. The slaves hired out were not necessarily representative of the total slave population. Selection effects might have worked in opposite directions for the different datasets. Agents hiring slaves would have wanted to hire as productive slaves as possible, so the samples of data based on such sources (the datasets by Evans, Goldin, and Margo) might all suffer from such a positive selection effect. The selection effect might have been the strongest in the case of the sample assembled by Margo, as the government most probably only would have wanted to hire less rebellious slaves to work in the military forts. Fogel and Engerman’s dataset, based on probate records of the slave-owners, would on the other hand potentially show a negative selection effect: previous research has for example shown that some owners preferred to hire out slaves they had problems managing themselves, and keep the best and most productive slaves to work on the own plantation.

Comparing the slave hire rates in the four different samples of data shows little support for Butlin’s critique of Evans’ sample of data: Evans’ monthly hire rates exhibit comparatively similar hire rates to the hire rates for unskilled slave laborers in Margo’s dataset (see appendix figure A2). Since we can be quite certain that there is no skill premium involved in the observations used from Margo’s dataset (as the occupation of the slaves are known for all these observations, and the selected observations in the figure only include unskilled laborers), it seems reasonable to conclude that there is little or no skill premia involved in Evans’ dataset, either. It is, on the other hand, not possible to rule out Goldin’s critique of Evans’ data, that there might be a risk-premium included in these series of hire rates.

Another issue is whether the slaves hired were representative of the broader slave population, or if they were positively or negatively selected in either sample? It is not possible to tell this for certain in any case, but if that would have been the case for the sample of data collected by Fogel and Engerman, the cost of hiring unskilled free laborers in the antebellum South on average would have been around 200 per cent higher than the cost of hiring a slave.68 That the labor markets for free and enslaved labor would have been so badly integrated seems improbable. It seems more plausible to assume that the slave in this particular sample might have been negatively selected to some extent, or that there might have been certain discounts involved when hiring out slaves to relatives or neighbors. The slave hire rates in this dataset could then suffer from a quite substantial downward bias compared to what a slave-owner could expect to profit from exploiting an average slave. The cost of hiring (and what a slave-owner thus could expect to profit from hiring out) an average slave would then plausibly have fallen somewhere between the estimates by Fogel and Engerman on the one hand, and by the three other scholars on the other hand.

5 Results

Figures 1a–b show the results for the total return on investments in slaves in the Lower and Upper South, respectively, using the data from Fogel & Engerman as a lower boundary, and the data from Evans as an upper boundary for the estimates.


Figure 1a

Total real return on investment in slaves in the United States, 1806–1860, by year

Citation: Journal of Global Slavery 8, 1 (2023) ; 10.1163/2405836X-00801003


Figure 1b

Total real return on investment in slaves in the United States, 1806–1860, by year

Citation: Journal of Global Slavery 8, 1 (2023) ; 10.1163/2405836X-00801003

Sources: slave hires rates; Evans = Evans, “The Economics of American Negro Slavery, 1830–1860”; F&E = Fogel and Engerman, “Slave Hires, 1775–1865”. Capital gains from price changes based on Sutch, “Slave Prices, Value of the Slave Stock, and Annual Estimates of the Slave Population: 1800–1862”, tables Bb209–211. Capital gains for female slaves due to population growth based on Sutch, table Bb214. Capital losses due to ageing based on Fogel and Engerman, Time on the Cross, fig. 18; adjusted for inflation based on consumer price index from Peter Lindert and Robert Margo, “Consumer Price Indexes” in Historical Statistics of the United States (Cambridge University Press, 2006), table Cc2

As can be seen in the figures, the return on investment fluctuated substantially between the years. This is primarily due to changes in the purchase price of the slaves. These changes were clearly related to business cycles in the antebellum Southern economy. Cotton had, for example, increased in price for several years until 1818, due to increasing demand from Britain in particular. In 1819, however, cotton prices dropped drastically, leading to the “Panic of 1819,” with repercussions for several years after.69 This would also have repercussions on the market for slaves, so that the prices for slaves tumbled in the following years—having peaked at more than $ 1,000 for a prime male field hand in 1818, the price of such slaves dropped to less than $ 600 some five years later. The same pattern would repeat itself some two decades later, with the “Panic of 1837.” This panic was not solely driven by the cotton market, but has been attributed to many different factors. The effects nonetheless lasted for several years.70 These effects again had major repercussions for the market for slaves: the price of a prime male field hand peaked at around $ 1,300 in 1837, but was some five years later again below $ 600. In terms of the return on investment, these drastic price-drops would mean a substantial capital loss for any slave-owner who during these depressed times, for one reason or another, had to realize the capital that they had invested in the slaves. The rent that a slave-owner could gain from exploiting the labor of a slave—here proxied by the slave hire rates—were in contrast much more stable over time. The slave hire rates in general moved in tandem with the development of the purchase price of the slaves—possibly with a slight lag. The effect was that the net rent from slave hiring normally was in the range of 8–10 per cent per year on the capital invested in the slave.

It is also noteworthy that the gender differences are minor: male and female slaves seem to have provided the owner with quite similar total return on investment. The minor differences that do occur on an aggregate level (see table 1, panel A) mainly occur because of missing data either for male or female slaves some particular years—for years of comparable data, there are only very small gender-differences in the estimated figures. This lends credibility to the different assumptions and proxies employed when estimating the total return on investment by gender (discussed further in the methods section above).

Table 1

Total real return on investments in slaves in the antebellum South, total for the period 1806–1860 (per cent per year)





Average return







(per cent per year,



geo. mean)






Panel A: plain hiring rates