Exploration of the economic aspects of slavery has continued without interruption in the 30 years since the publication of Time on the Cross.¹

There is now recognition that plantations were profitable at the time of the abolition of both the slave trade and later slavery itself. It is also generally recognized that in the nineteenth century plantation output in the Americas increased massively, and that even though the price of what slaves in the Americas produced—cotton, coffee, and sugar—declined over several decades, the price of the slaves themselves increased.² On the broad interpretation of these patterns, there is, however, much less consensus. Many historians still see slavery as associated with low technology, at least in the field, and as insufficiently flexible to have survived long in an industrial environment.³

In the Caribbean the idea of the permanent decline of the British slave economy in the last quarter of the eighteenth century, that originated with nineteenth-century abolitionists, and was developed with racist undertones by Lowell Ragatz in the 1920s, continues to receive support.⁴

Indeed, recent work on the pre-1800 period has posed something of a paradox for slave societies. Ever larger estimates of wealth are juxtaposed in the literature with suggestions that sustained productivity gains in slave economies were small or nonexistent.⁵ The historical record before 1800 will never provide the rich material sustaining the detailed debates on the economics of slavery that have become the norm for the antebellum South; nevertheless, broad estimates of income and even of productivity trends are gradually emerging. More particularly, it is now possible to compare eighteenth-century trends in produce prices and slave prices—the basis for the striking findings of the last quarter century on productivity in the late slave Americas. In “Slave Prices, the African Slave Trade and Productivity in the Caribbean, 1674–1807,” we made use of data on the prices paid for slaves as they arrived in the Americas from Africa in conjunction with prices for the single product that most of these slaves were destined to produce—sugar—to draw inferences about productivity change and other aspects of the slave economy. We concluded that there were significant increases in both average slave productivity and total factor productivity, the latter as much as doubling over the course of the eighteenth century.

Two papers by Peter Mancall, Joshua Rosenbloom, and Thomas Weiss examining aspects of output and productivity growth in the Lower U.S. South, are along similar lines. In “Slave Prices and the South Carolina Economy, 1722–1809” (hereafter MRW), our focus here, they combine a slave price series, based largely on probate records, with a price series made up of the main export goods that slaves produced—indigo, cotton, and most importantly rice—to infer trends in average slave productivity. In their other paper, “Agriculture Productivity in the Lower South, 1720–1800,” they marshal an impressive volume of output, input and price data to draw broader implication about productivity growth. It is reassuring that, in certain aspects, the findings of Mancall, Rosenbloom, and Weiss, which are based on a greater variety of data than our estimates for the Caribbean, are similar to our own. They conclude, for example, that the market for slaves in the Americas was open in the long run, and average labor productivity and total factor productivity grew in some periods. Nevertheless, some of their central conclusions hearken back to a traditional view of slavery as inefficient and nonproductive. For example, they find significant differences in the pattern of slave prices between the Caribbean and South Carolina, and large variations even within the mainland colonies. Thus they conclude: “although the long-run supply curve was probably close to perfectly elastic, the short-run supply curve was relatively inelastic.”⁶

More importantly, MRW are generally pessimistic about productivity change, concluding in their paper on South Carolina that: “there was no apparent trend in the value of slave labor deflated by the price of its chief products,” a theme further developed in “Agricultural Productivity,” which analyses all of the Lower South.⁷

The data set, on which we drew to derive our results for the Caribbean, includes slave arrivals in South Carolina and other mainland reception points. During the eighteenth century South Carolina became a major producer of rice, and historians have recently come to focus on the skills that Africans acquired, before they were forced across the Atlantic, and the supposedly strong preferences that South Carolina planters, in particular, had for slaves from rice-growing regions of Africa.⁸

In fact, a comparative analysis of the provenance of slaves arriving in the tobacco-growing Chesapeake on the one hand and, on the other, Charleston, with its primarily rice-producing hinterland, reveals no such preference.⁹

The MRW sample is for slaves born in the Americas or Africans who had lived in the Americas long enough to enter the probate of their owner. MWR are able to identify the adult male slaves in their sample, but a concern in interpreting their series is the effect of age on slave value. Slaves who entered the probate records would have been older on average but more likely to have developed craft skills than the newly-arrived slaves; as well, one expects that the variation in age would have been much greater. One advantage of the slave trade data is the smaller differences in price due to age. Our series on the sales of newly arrived slaves is much less complete than MRW's, but given the larger sample sizes in those years for which data are available and the greater uniformity, the slave trade data provide a check on the MRW sample and allow us to draw other inferences about the nature of the slave market.

Our series of slave trade prices and the MRW series for years where there is overlap are shown in Table 1. Given the different types of sources, it is perhaps surprising that the slave trade records—records of actual sales, rather than valuations—are so similar to those taken from the probates. Excluding the years 1755, 1783, and 1784, which appear to be outliers even in terms of the MRW sample, the absolute differences average just 10 percent of the slave price. It thus appears that the lower productivity of newly arrived slaves was almost exactly offset by their younger age and, thus, longer productive lifetime. Although the number of years for which data are available is less than one would have liked, there was also little difference between the prices received for newly arrived slaves in South Carolina and the other mainland colonies. This finding is suggestive of a more open market than implied by the MRW comparison of South Carolina and Maryland. In fact, given that newly arrived slaves were more homogeneous, certainly in terms of age, the lower prices of slaves in Maryland after 1760 may indicate a difference in the composition of the slaves for whom prices were obtained.¹⁰

Some of the years for which MRW have data are based on few observations; as well, there are years missing from the MRW series for which slave trade data are available. Given the similarity in those years where there is overlap, we have merged the two data series and it is this combined series that is the basis of our subsequent analysis.¹¹

Within the Caribbean we found that, with the exception of short periods, the slave trade approximated an open market. Figure 1, which compares our Caribbean slave price series with the price series for South Carolina, indicates that the market for slaves in South Carolina was part of that broader American market. For the period 1722–1775 the average price of slaves in South Carolina was $160 as compared to $152 in the Caribbean. Over the period 1785–1806, South Carolina prices averaged $216 as compared to $227 in the Caribbean. In neither case do we reject the null hypothesis that the mean prices were the same.¹²

For the period 1785 to 1806 there is also no significant difference in the price trends, but from 1722 to 1775, the (constant dollar) price of slaves in South Carolina increased at an estimated 1.45 percent per year, which is significantly different from the annual 0.82 percent increase in the Caribbean.¹³

Based on a comparison of slave prices with the price of their main output, sugar, we argued in “Slave Prices,” for significant increases in average slave productivity throughout the Caribbean. MRW's approach to productivity was much the same, except in place of the sugar price they used combinations of rice, indigo, and cotton prices depending on the period. It is surprising then, given the similarity of the slave price series for the Caribbean and South Carolina, that MRW find little apparent trend in the ratio of slave prices to the prices of the export goods they produced. The lack of productivity growth is especially puzzling given that during the period 1722 to 1775 slave prices in South Carolina were increasing somewhat faster than in the Caribbean.¹⁴

Central to our calculation for the Caribbean was the price of sugar, the main slave output and almost the only cash crop. As noted, MRW use combinations of the price of rice, indigo, and cotton, although given the nature of the calculation it is the price of rice that dominates. For the period 1722 to 1731 the price of rice is from Peter Coclanis's work and after 1731 it is the Charleston wholesale price as reported by Arthur Cole.¹⁵

Basing prices on Charleston would seem appropriate, but there may be reason to believe that the series used by MRW greatly overstates the increase in rice prices, leading to a serious underestimate of productivity growth. Coclanis notes that his series for the period 1722 to 1731 should be regarded as a farm-gate price. Of course, shifting from a farm-gate measure to a wholesale market price would tend to increase the price even if there were no change in the underlying market. The extent of the adjustment is uncertain, but an indication of the magnitude is suggested by a comparison of the Coclanis-Cole series with the Philadelphia wholesale price, which is available for the entire the period. Adjusting for the currency exchange rates in South Carolina and Pennsylvania, the difference between the Philadelphia price and the Coclanis series averages 7.0 s/cwt (sterling) for the period 1722 to 1731.¹⁶ In the following decade, where the Charleston wholesale price is available the difference is 3.9 s/cwt. The gap of 3.1 s/cwt is about 30 percent of the Charleston wholesale price and, although other factors may have affected the price differential between the cities, the possibility remains that the smaller price differential after 1731 is due largely to a change in definition.¹⁷ If so this adjustment alone has a potentially significant effect on the productivity growth estimates.

Another concern is with possible variation in the quality of rice that was brought to the Charleston market. In the Cole data, the quality of rice at both Charleston and Philadelphia is listed as “unspecified.” Coclanis points out that lower quality rice was consumed locally. Some of this rice was “nonmerchantable” and hence would not have entered even the Charleston market; nevertheless, the incentive would have been to export the higher grades and sell more of the lower quality rice in Charleston.¹⁸

It is only beginning in 1803 that Cole reports the wholesale Charleston price of rice as “first quality.” Prior to that year variations in the Charleston price were potentially affected by changes in quality in addition to fluctuations in the market. The effect of quality on price appears to have been greater in the periods when exports were strong suggesting that less of the high-quality rice remained in the local market. To indicate the possible biases, we compare the wholesale price of rice at Philadelphia adjusted for transport costs to the MRW price series (see Figure 2).¹⁹ Overall the Charleston series trends much sharply higher than the Philadelphia series. There is further reason to believe that the Philadelphia series was more reflective of the price received by producers. An index of rice prices in England and Holland corresponds much more closely to the price in Philadelphia than in Charleston. For these reasons we use an index based on the Philadelphia price to derive implications for slave productivity.

We follow the MRW approach of using an export price index based on combinations of the prices of rice, indigo, and cotton.²⁰

The “real” price of slaves, that is the price deflated by the export price index, follows a more strongly upward trend than does the MRW series, largely because of the adjustment to the rice series (see Figure 3). As MRW found, the “real” slave price began to increase in the early 1740s, but we estimate price more than doubling by the time of the Revolution, a considerably greater increase than MRW estimate. There was a decline after the war, but by the end of the period, the “real” price of slaves was 70 percent above the 1722 level.

As noted by MRW and discussed at some length in Eltis et al., “Slave Prices,” the relation between the ratio of slave to output prices and slave productivity was affected by a number of factors. One of these was the discount rate. If we take the legal interest rate in South Carolina as appropriate, then the decline from 10 percent to 8 percent during the period before the war would imply an increase in productivity of about 75 percent rather than the doubling suggested by the price series alone.²¹

Sidney Homer and Richard Sylla, however, report interest rates for the American colonies, including the Lower South that imply little if any decline for loans that were secure. In fact, following the Revolution, interest rates were trending upward.²² On balance, then, it seems appropriate to take the price of slaves as reflective of their net marginal value product.

The “real” slave index that MRW report is based on export prices alone; yet as they carefully document in “Agricultural Labor Productivity,” a substantial part of slave output was consumed domestically. In order to provide a better measure of overall slave productivity rather than productivity based solely on exports, we have derived a slave price index that includes the major nonexported, slave-produced good, corn. We take the wholesale price of Indian corn in Charleston and assign it a weight of 0.5 in the output price index.²³

The result is another “real” slave price series (see Figure 3). Despite the emphasis of MRW on rising export prices, in fact the upward trend in the price of corn was greater, implying a somewhat smaller increase in the “real” slave price, although it is still considerable. For the period to the Revolution the increase in price is about 75 percent rather than the doubling based on exports alone. This result is hardly surprising given the emphasis that Coclanis and others have placed on productivity change in rice farming during the eighteenth century.

As was true of the Caribbean, increases in slave productivity were achieved despite rapid growth in the slave population. From 1722 to 1775, as MRW report, the slave population increased by a factor of six and over the entire period by a factor of 13.

²⁴

Thus, impressive gains in average slave productivity were achieved despite the great increase in labor supply. Drawing inferences from slave populations about total factor productivity requires data on other factor inputs, land of course being the most important. The lack of evidence on land that is of the quality of the slave and output price data makes our conclusion about total factor productivity more speculative. Nevertheless a variety of evidence points to a finding that total factor productivity was increasing, as it was in the Caribbean. Assuming a Cobb-Douglas specification with constant returns to scale, the relation between average labor productivity and total factor productivity is:

where π is average labor productivity, T is the input of land, L is the input of labor, α is labor's factor share, and superscripts refer to time. An alternate approach, appropriate if one has estimates of land productivity, is to compute total factor productivity as:

where τ is land productivity. Finally one could use a method similar to that used to derive the average productivity estimates by applying the dual of the production function and use output and factor prices. One such specification is:

where P is the price of slaves, R is the rental rate of land, and p is an index of output prices. In Table 2 we report indexes of total factor productivity for the period to the Revolution and for the entire period using each of the measures.

The results point to substantial increases in total factor productivity, with all the increase taking place by the time of the Revolution. Basing the calculation on export goods alone, we estimate using the three different measures that total factor productivity roughly doubled between 1722 and 1770–1774, implying an annual rate of increase over the period of 1.4 percent. According to measure A, which combines labor productivity with the land-labor ratio, total factor productivity increased 103 percent over the period 1722 to 1770–1774, whereas according to measure C, which is based on land prices, it increased by 81 percent. As MRW note and as our series also shows, much of the productivity increase took place after 1740. Important as well is that, despite the dislocation caused by the war, the effect on productivity appears to have been temporary. By 1806 the productivity of South Carolina agriculture was close to the prewar level. Our finding that the productivity of slave agriculture was similar in 1770–1774 and 1802–1806 is not very different from what MRW conclude, but the level that we estimate for each of these periods is much higher.²⁵

Measures of productivity based exclusively on prices are less firmly based than those which include data on outputs and inputs. Therefore it is reassuring that the estimates we derive are consistent with descriptions of eighteenth-century rice farming and the advances that were taking place in South Carolina. Irrigation technology, varieties of rice, and task organization improved in the decades after the successful experiments with growing rice in the 1690s.²⁶

Both Lewis Gray and Coclanis report estimates of land yields increasing by 50 percent, and it appears output per worker was going up as well: “a good working hand on a rice plantation [prior to 1748] produced about 2,250 pounds of clean rice per year, the figure for the average hand had apparently grown to about 3,000–3,600 pounds yearly during the second half of the eighteenth century, with good hands capable of even more.”²⁷ This anecdotal evidence implies increases in total factor productivity greater than 50 percent.

The evidence on South Carolina points to large advances in productivity associated mainly with rice farming, but for the major questions about slavery and its role in the development of the Americas, we need to observe the larger playing field. The South Carolina slave price data should be seen as just part of an assessment of trends in the plantation economy of the Americas as a whole. In 1790 South Carolina contained less than 3 percent of the Americas' slave population; and it produced rice and indigo, whereas the overwhelming reason slaves traveled across the Atlantic was sugar. Moreover, indigo, which accounted for as much as one-third of the region's export revenues, was heavily dependent on British bounties, and this revenue source largely disappeared after independence as the bounties were removed. But, despite the sudden variation in export crops, productivity remained high after the war; and over the entire period the increase in productivity paralleled that in the Caribbean. In the past two decades strong evidence of nineteenth-century technology gains has emerged for the export sectors of the U.S. South and other parts of the Americas. It now appears that such a pattern was well established before 1800.