The Journal of African History http://journals.cambridge.org/AFH

Additional services for The Journal of African History:

Email alerts: Click here Subscriptions: Click here Commercial reprints: Click here Terms of use : Click here

The Impact of the Atlantic Slave Trade on Africa: A Review of the Literature

Paul E. Lovejoy

The Journal of African History / Volume 30 / Issue 03 / November 1989, pp 365 ­ 394 DOI: 10.1017/S0021853700024439, Published online: 22 January 2009

Link to this article: http://journals.cambridge.org/abstract_S0021853700024439

How to cite this article: Paul E. Lovejoy (1989). The Impact of the Atlantic Slave Trade on Africa: A Review of the Literature. The Journal of African History, 30, pp 365­394 doi:10.1017/ S0021853700024439

Request Permissions : Click here

Downloaded from http://journals.cambridge.org/AFH, IP address: 131.111.164.128 on 28 Nov 2012 Journal of African History, 30 (1989), pp. 365-394 365 Printed in Great Britain

THE IMPACT OF THE ATLANTIC SLAVE TRADE ON AFRICA: A REVIEW OF THE LITERATURE1

BY PAUL E. LOVEJOY

African History and the Atlantic Slave Trade THE significance of the Atlantic slave trade for African history has been the subject of considerable discussion among historians and merits attempts from time to time to review the literature. The present such attempt addresses several, but not all, the key issues that have emerged in recent years. These are, in order of discussion here: What was the volume of the Atlantic slave trade ? More specifically, what were the demographic trends of the trade with respect to regional origins, ethnicity, gender and age ? Finally, what was the impact of the slave trade on Africa ? In brief, what is the state of the debate over the slave trade ? My own position in the debate is clear: the European slave trade across the Atlantic marked a radical break in the history of Africa, most especially because it was a major influence in transforming African society.

The history of slavery involved the interaction between enslavement, the slave trade, and the domestic use of slaves within Africa. An examination of this interaction demonstrates the emergence of a system of slavery that was basic to the political economy of many parts of the continent. This system expanded until the last decades of the nineteenth century. The process of enslavement increased; the trade grew in response to new and larger markets, and the use of slaves in Africa became more common. Related to the articulation of this system, with its structural links to other parts of the world, was the consolidation within Africa of a political and social structure that relied extensively on slavery.2

The transformation thesis identifies slavery as a central feature of African history over the past millennium. The Atlantic trade was only one, although a major, influence on the transformation of society. The Muslim slave trade was also important, and other internal African developments strongly influenced social change as well. According to this interpretation, the task of the historian is to weigh the relative importance of the various factors that incorporated Africa into an 'international system of slavery' that included Africa, the Americas, western Europe and the Islamic world.3 David Eltis has challenged this interpretation. On the basis of his study of the nineteenth-century Atlantic trade and an analysis of the value of the Atlantic trade between the 1680s and 1860s, Eltis has concluded that neither the scale nor the value of the Atlantic trade was sufficiently large to have had

1 I would like to thank Patrick Manning, J. S. Hogendorn, Joseph C. Miller, Jose Curto, Charles Becker, Allen Isaacman, and Stuart Schwartz for their comments on this article. 2 Paul E. Lovejoy, Transformations in Slavery. A History of Slavery in Africa (Cambridge, 1983), 22. 3 Ibid. 275-

14 AFH30 366 PAUL E. LOVEJOY more than a marginal influence on the course of African history. According to Eltis, the slave trade for most regions and most periods was not a critically important influence over the course of African history. At the very least, those who would place the slave trade as central to West African and west-central African history should be able to point to stronger common threads, if not themes, across African regions than have so far come to light.4 With respect to slavery, he claims that 'whatever the origins or nature of structural changes in African slavery, it is unlikely that external influences could have been very great'.5 The contribution which Eltis makes is two-fold, it seems to me. First, he brings more precision to an analysis of the volume and direction of the Atlantic trade in the nineteenth century. His study of nutritional trends, age and gender, and mortality are particularly significant.6 Secondly, he has articulated a model of economic development for the pre-colonial era that must be taken seriously, although I disagree with his conclusions. Did exports determine the extent of economic change, as measured by standard economic indicators ? He concludes that an export-led model of economic development has little to offer in interpreting African history. Climate and human genius, according to Eltis, were far more important than the export sector.7 Eltis bases his startling conclusions on an analysis of the relative importance of the Atlantic trade on Africa, as determined on a per capita basis. In a study undertaken jointly with Lawrence C. Jennings, it is claimed that the annual average per capita trade of those parts of Africa involved in the Atlantic slave trade was significantly less than in other parts of the Atlantic basin and that the African share in world trade declined in relative terms from the 1680s to the 1860s.8 They conclude that neither the absolute nor the relative value of Atlantic trade was very great; in general, foreign trade had only a weak influence on African economies. According to Eltis, ... on the assumption that the improbably low figure of 15 million people lived in West Africa [in the 1780s] at subsistence levels, then imports from Atlantic trade may be taken at about 9 per cent of West African incomes in the 1780s. With assumptions that are more in accord with reality (i.e. a population of 25 million or more and domestic production in excess of subsistence), then imports decline in importance to well below 5 per cent. For other decades in the century when both slave prices and exports were lower, imports would have been much less significant. For west-central Africa, population densities were much lower but import/income ratios could not have been much greater.9 4 David Eltis, Economic Growth and the Ending of the Transatlantic Slave Trade (New York, 1987), 77. 5 Ibid. 225. 6 Also see his ' Fluctuations in the age and sex ratios of slaves in the nineteenth-century transatlantic slave traffic', Slavery and Abolition, vn, 3 (1986), 257-72; and 'Nutritional trends in Africa and the Americas: heights of Africans, 1819-1839', Journal of Inter- disciplinary History, xn (1982). 7 According to Eltis (Economic Growth, 15), 'There can...be no doubt that the slave trade was of critical economic importance to the nineteenth-century Atlantic basin as long as it lasted. The only part of the basin where this was not the case was Africa...'. 8 David Eltis and Lawrence C. Jennings, 'Trade between Western Africa and the Atlantic world in the pre-Colonial era', American Hist. Review, XLIII, 4 (1988), 936-59. 9 Eltis, Economic Growth, 72. THE ATLANTIC SLAVE TRADE 367 Indeed, 'The majority of Africans... would have been about as well off, and would have been performing the same tasks in the same socioeconomic environment, if there had been no trading contact' with Europe.10 Eltis even advances the astonishing conclusion for Asante that'... the ratio of the level of exports, either before or after 1807, to any plausible population estimates of Asante suggests that the slave trade can never have been important'.11 The rise of commodity exports in the nineteenth century had virtually no impact on Africa either. According to Eltis, ... the slave and commodity trades together formed such a small percentage of total African economic activity that either could expand without there being any impact on the growth path of the other [I]n the mid-nineteenth century neither the slave nor the commodity traders were large enough to have to face the problem of inelastic supplies of the factors of production.12 In short, neither the Atlantic slave trade nor its suppression had much influence on African history. The following review of the recent literature on the Atlantic slave trade provides a context in which to assess the revisionist interpretation of Eltis (and Jennings). I begin with the new studies on the volume of the slave trade, in which a consensus seems to have emerged. I then consider the analytical refinements in the regional and ethnic origins of the exported slave population. The demographic data allow a closer examination of the gender and age profile of the trade, which is the subject of the next section of this article. Finally, I return to an assessment of the arguments of Eltis, particularly with regard to the demographic impact of the trade on Africa. A number of important themes are not considered, including the economic significance of slavery in Africa and the importance of imported commodities on African society and economy. Nonetheless, I believe that I can demonstrate that Eltis' provocative con- clusions are seriously flawed.

The Volume of the Atlantic Slave Trade Twenty years ago, Philip Curtin insisted that a scientific scrutiny of slavery required the statistical study of the Atlantic slave trade. Through an analysis of published material, he estimated that European slavers imported 9,566,100 Africans into the Americas, Europe and the Atlantic islands for the entire period of European involvement in the slave trade.13 He allowed a considerable margin of error because many of his sources were not based on archival research. As any good historian should, he restrained his analysis within limits. His initial discovery was startling enough. Earlier figures—15 million, 50 million and more — were nothing but guesses. A proper application of statistical theory could establish the probable range of demographic change. For those interested in African history, Curtin's research has led to con- siderable controversy. His effort at quantification has been challenged,14 and 10 Eltis and Jennings, 'Western Africa and the Atlantic world', 958. 11 Eltis, Economic Growth, 74. 12 Ibid. 183. 13 Philip D. Curtin, The Atlantic Slave Trade : A Census (Madison, Wisconsin, 1969), 268. 14 For a summary of the debate, see Paul E. Lovejoy, ' The volume of the Atlantic slave trade: a synthesis \J. Afr. Hist., xxin, 4 (1982), 473-501; and Transformations in Slavery. 368 PAUL E. LOVEJOY indeed the attempt to derive a global estimate for the trade has been criticized.15 Some of those who have believed that the slave trade had a devastating effect on African societies have not welcomed the hard look at facts and statistical probability. A reduction of the figure for slave exports might indicate that the trade had only a marginal impact on Africa. According to this interpretation, which was never articulated in such bold terms before the revisionist conclusions of Eltis and Jennings, fewer slaves might mean less oppression. Any assessment of the impact of the slave trade on Africa has to estimate the scale of exports over time and for specific exporting regions, and hence has to deal with the projections based on existing knowledge, always allowing for gaps in the data. Many studies of this type have been completed, and more are likely to be done.16 Analysis depends upon a continuing reassessment of the volume and direction of the export trade as new material becomes available. The most important new material pertains to the French trade, assembled in Jean Mettas, Repertoire des expeditions ne'grieresfranfaises au XVIIIe siecle.17 Mettas's data are analyzed by Charles Becker and David Richardson.18 Richardson has also compiled new material on the British trade in the eighteenth century and has reinterpreted the available data on the North American trade.19 Finally, Jose Curto has uncovered new material on the Portuguese trade.20 The revisions of Richardson, Becker and Curto tend to confirm the broad parameters of Curtin's 1969 study, which are revised upward but within acceptable limits. The known scale of the slave trade was on the order of 11,863,000 slaves shipped across the Atlantic, with a death rate during the Middle Passage reducing this total by 10-20 per cent, which means that 9-6-10-8 million slaves were imported into the Americas. Curtin expected such refinement. The cautious historian expects that upward revision is more probable than downward.

16 David Henige,' Measuring the immeasurable: the Atlantic slave trade, West African population and the Pyrrhonian critic', .7. Afr. Hist., xxvn, 2 (1986), 295-313. 16 Curtin undertook one such regional analysis; see his Economic Change in Precolonial Africa: Senegambia in the Era of the Slave Trade (Madison, Wisconsin, 1975). His interpretation has been challenged; see especially Charles Becker, ' La Senegambie a l'epoque de la traite des esclaves. A propos d'un ouvrage recent de Philip D. Curtin', Revue francaise d'histoire d'outre-tner, LXIV (1977), 203-44. F°r a general survey of the regional studies, see Lovejoy, Transformations in Slavery. 17 Vol. 1, Nantes (Paris, 1978,8. Daget, ed.); Vol. 11, Ports autres que Nantes (Paris, 1984, S. and M. Daget, eds.). Serge Daget has also completed an inventory of the illegal trade of the nineteenth century; see Repertoire des Expeditions ne'grieres franfaises a la traite ille'gale (1814-1850) (Nantes, 1988), but I have not attempted to analyse the data therein. Instead, I am relying on Eltis's research, which is based on different data. 18 David Richardson, 'Slave exports from west and west-central Africa, 1700-1810: new estimates of volume and distribution', J. Afr. Hist., xxx, 1 (1989), 1-22; Charles Becker, 'Note sur les chiffres de la traite atlantique francaise au XVIIIC siecle', Cahiers a"etudes africaines, xxvi (1986), 633-79. 19 David Richardson, 'The eighteenth-century British slave trade: new estimates of its volume and distribution', Research in Economic History, xn (1988); and Richardson, 'Slave exports'. 20 Jose C. Curto,' Recounting the numbers: the legal Angolan slave trade, 1710-1830', unpublished. I wish to thank Jose Curto for showing me his unpublished work. THE ATLANTIC SLAVE TRADE 369 Richardson has undertaken the most complete attempt at a new synthesis, but only for the eighteenth century. He concentrates on the period 1700—1810, although he neither used Becker's work nor had access to Curto's research. Since Richardson made the same calculations as Becker and Curto's different data base confirms what is already known, it might be expected that some level of confirmation is possible. It is rewarding to those of us who synthesize that Richardson concludes: ' Lovejoy's re-estimate of the overall volume of the trade in this period is perhaps reasonably accurate but... his assessment of both the temporal and coastal distributions of the trade needs to be modified in several important ways'.21 One expects adjustments; one is repeatedly sur- prised at the equilibrium that seems to form around proper statistical analysis. Curtin understood statistics in 1969. Others have had to learn slowly. Errors seem to cancel themselves, but the acceptable figures for the volume of the trade seem to inch upwards nonetheless. In order to understand the state of the debate, it is necessary to look at the different sectors of the trade in the eighteenth and early nineteenth centuries and adjust for the various modifications that seem to be required. I begin with the French trade, and then consider the British, North American and Portuguese sectors in turn. Finally I end with a reconsideration of the nineteenth-century trade. While it can be expected that modifications in the estimated volume before 1700 are also in order, the level of the trade was much lower then, and consequently revisions are likely to have little impact on our assessment of the total trade.22 The new analysis of the French sector rests on Mettas' compilation of shipping data that include the names of ships, often including the number of slaves on board, that operated from Africa between 1707 and 1793. The material on Nantes was published in 1978, but only with the publication of information on the other French ports in 1984 has it been possible to substitute the Mettas material for earlier analyses of French sources.23 Becker first completed the substitution. His tabulation of the Mettas data totals i,oi7,oio,24 which includes 12,845 slaves delivered to the in the Indian Ocean. Becker's total for the Americas is 1,004,165, but that figure does not always include slaves who died in the Middle Passage. Furthermore, Becker provides examples of other gaps in the Mettas data and suggests that the real level of French exports from Africa approached 1.5 million. In recalculating Becker's figures, I reach a preferred estimate of 1,150,000.

21 'Slave exports', 2. 22 See, for example, Ivana C. Elbl, 'The Portuguese trade with West Africa, 1440- 1521' (Ph.D. thesis, University of Toronto, 1986). According to Elbl's interpretation, some upward revision is called for. 23 This is why I did not use the Nantes volume in my 1982 synthesis. 24 'La traite atlantique francaise', 633-79, ar>d personal communication. 26 The gaps include at least 3,848 slaves on twenty-two ships not reported in the Mettas inventory, and another twenty-six ships whose cargoes are unknown. For Senegambia alone, Becker has information on other vessels that are also not included in the Mettas data. In addition, Becker refers to documents that conflict with the Mettas materials for certain periods. In a registry of slaves introduced into the Americas between 1728 and 1740, for example, 203,522 slaves arrived on 723 ships. The Mettas material records only 132,851 slaves on 418 vessels. These discrepancies alone account for an additional 82,000 37° PAUL E. LOVEJOY Richardson also tabulated the Mettas data. His total for 1710-93 is 1,017,700, and he estimates the French trade from 1700—1809 at 1,052,000.26 Unlike Becker, Richardson makes no allowance for gaps in the Mettas data, other than for the period before 1707 and after 1793. My 1982 estimate is virtually the same as Richardson's figure. Based primarily on Robert Stein's calculations of Admiralty records, I concluded that the French trade totalled 1,180,300, but I erroneously included French imports into the Mascarene islands in the eighteenth century, which probably totalled 160,000 slaves.27 The British trade, according to Richardson's new data, involved the export of 3,120,000 slaves between 1700 and 1810, which is some 342,700, or 12-3 per cent, more than my estimate (following Anstey, Inikori, Curtin, Drescher, et al.).2S My synthesis appears to have under-estimated the portion of the British trade in the first three-quarters of the century and hereby stands corrected.29 Richardson has used previously unknown shipping data for the period before 1750 and has reanalysed the available data on shipping for the period after

slaves. Other documents reveal similar discrepancies. All figures are lower than the Mettas data; see Becker, 'Traite atlantique francaise', 665-8. A figure of 1,082,000 is certainly the lower limit for the period of the Mettas data (1707-93). When estimates for the period before 1707 and after 1793 are included, an adjusted estimate for Becker's figure would be at least 1,126,000 and easily 1,150,000. Becker's projection of rs million for the French trade as a whole, including 160,000 slaves shipped to the Mascarene islands, represents a further increase of 15 per cent. Whether or not he is correct in this projection remains to be proven. 26 Richardson, 'Slave exports', io, 14. The similarity between Richardson's and Becker's calculations of the Mettas data cannot be explained. Becker included the Mascarene trade; Richardson states that he did not. Nonetheless, both reached virtually identical totals based on the same number of ships. A more thorough comparison of their methodologies might resolve the difference. 27 Richardson ('Slave exports') correctly observes that my earlier estimate of the French trade to the Americas wrongly included the Mascarene trade. I relied on Robert Stein, The French Slave Trade in the Eighteenth Century : An Old Regime Business (Madison, Wisconsin, 1979); and Stein, 'Measuring the French slave trade, 1713- 1792/3', J. Afr. Hist., xix, 4 (1978), 515-21. Richardson incorrectly assesses my error at 125,000, not 160,000; see J.-M. Filliot, La Traite des esclaves vers les Mascareignes au XVIII' siecle (Paris, 1974), 45-51. 28 Richardson, 'Slave exports', 3-4. For the earlier research, see Lovejoy, 'Volume of the Atlantic slave trade', 473-501 and the citations therein, especially Roger Anstey, The Atlantic Slave Trade and British Abolition, 1760-1810 (London, 1975); Anstey, 'The volume and profitability of the British slave trade, 1761-1807', in Stanley Engerman and Eugene Genovese (eds.), Race and Slavery in the Western Hemisphere: Quantitative Studies (Princeton, 1975); J. E. Inikori, 'Measuring the Atlantic slave trade: an assess- ment of Curtin and Anstey', J. Afr. Hist., xvn, 2 (1976), 197-223; Philip D. Curtin, 'Measuring the Atlantic slave trade once again', J. Afr. Hist., xvn, 4 (1976), 595-605; Anstey, 'The British slave trade 1751-1807; a comment', J. Afr. Hist., xvn, 4 (1976), 606-7 > and Seymour Drescher, Econocide: British Slavery in the Era of Abolition (Pittsburgh, 1977), 205-13. 29 Richardson criticizes my synthesis for failing ' to examine critically the basic sources of information used by the various protagonists in the debate over the volume of the British slave trade' ('British slave trade"). Richardson is correct, but after reviewing the various calculations, I accepted the arguments of Anstey and Drescher; Lovejoy, 'Volume of the Atlantic slave trade', 486-7. THE ATLANTIC SLAVE TRADE 371 1750.30 He calculated the average number of slaves per ship for each of Britain's main ports (Liverpool, Bristol and London), and then adjusted for ship-board mortality, again based on decadal averages.31 Meticulous research enabled him to construct annual figures for slave shipments. One serious problem with Richardson's calculations limits their reliability, however meticulous the research. Richardson did not properly adjust the clearance records for those British ships bound for Africa that did not carry slaves to the Americas. Various estimates put the level of non-slave shipping and foreign ships bound for Africa at 5 per cent of the total, while the proportion of ships lost at sea, seized or otherwise failing to complete their voyages represented another 5 per cent (this figure occasionally rising much higher during war years).32 Richardson eliminated any obvious examples from his analysis but otherwise decided not to make allowances. By his own admission, his estimates ' represent probably the maximum levels of the British slave trade'.33 If a raw figure of 5 per cent were used to account for non- delivery of slaves, then Richardson's estimate is reduced by 156,000 slaves, thereby suggesting a total figure of 2,964,000. This allowance for non-slave deliveries would suggest that Richardson's calculations indicate an upward revision of 186,700 or 6-7 per cent (not 342,700 or 123 per cent, as he argues). Estimates of the North American trade are still problematic, but Richardson has reinterpreted the work of other scholars, most notably Jay Coughtry and Herbert Klein.34 Richardson uses Coughtry's study for Rhode Island but incorporates data from Klein on North American trade to Cuba. He supple- ments his analysis with scattered data from elsewhere. As a result, Richardson estimates that the North American trade totalled some 208,000 slaves, considerably less than Anstey's estimate of 294,000, a figure that I adopted. According to Richardson, North Americans traded 145,000 slaves for the whole of the eighteenth century, a figure much lower than Anstey's estimate (194,200) for 1763-1799.35 Richardson reckoned that North American shippers bought 34,000 slaves before 1760, a period not covered in Anstey's estimate

30 Richardson, 'British slave trade'. It should be noted that Richardson has located new data on the annual volume of shipping, the number of slaves delivered per vessel in the Americas, and the mortality of slaves in the Atlantic crossing, but no new data on the average number of slaves loaded on British ships in Africa. 31 Richardson's estimates for the British trade vary. In 'British slave trade', he estimates the trade from 1700 to 1807 at 3,039,050 (I have subtracted his estimate for 1699 from his total, with an allowance for mortality of 20 per cent). In 'Slave exports', he rounds his figures to the nearest thousand and adds in additional figures for slaves shipped via , which affect his estimates for the period 1710-29, an upward revision of 81,000. 32 The few foreign ships that somehow crept into British statistics and those ships that did not carry slaves affect Richardson's estimate and should be eliminated from the analysis, while most ships that failed to complete their voyages still carried slaves and hence should be included in an assessment of volume. 33 Richardson, 'British slave trade'. 34 Richardson, 'Slave exports', 7-9, and full citations therein, but see Jay Coughtry, The Notorious Triangle: Rhode Island and the African Slave Trade (Philadelphia, 1981) and Herbert S. Klein, 'The Cuban slave trade in a period of transition, 1790-1843', Revue franfaise d'histoire d'outre-mer, LXII (1975), 72-5. 35 Roger Anstey, 'The volume of the North American slave-carrying trade from Africa, 1761-1810', Revue franfaise d'histoire d'outre-mer, LXII (1975), 65. 372 PAUL E. LOVEJOY because Anstey (and I) assumed North American imports were included in general British figures.36 Richardson also estimated that the North American trade was 63,000 in the first decade of the nineteenth century, thereby reaching his total of 208,000. His logic seems clear, and until further analysis proves otherwise, the downward revision that he proposes seems acceptable. Curto's revisions of the Portuguese trade are interesting for what they do not show. Despite laborious work in the archives, Curto uncovered little that was new. His research suggests some modest revisions in the export figures for Luanda and Benguela for the eighteenth and nineteenth centuries, but these revisions hardly affect the total or the regional and temporal distributions.37 Curto's new shipping data confirm existing figures, although a few annual returns are higher. He argues quite plausibly that the higher figures are more acceptable. He concludes that a more accurate total is approximately seven per cent more than previous .estimates. For the period of this study (1710-1830), he has accounted for an additional 12,736 slaves, 2,349 exported from Luanda and 10,393 from Benguela.38 This is a modest upward revision indeed. As noted, Richardson apparently did not know of Curto's work, but it hardly affects his analysis. Richardson concludes that the level of slave exports grew substantially over the eighteenth century, rising from 36,000 per year in 1700-09 to a peak of almost 80,000 per year in the 1780s. Such a view suggests that historians have tended to under-estimate the level of slave exports for the first three-quarters of the century and to over-estimate the level for the last quarter. Growth was more evenly distributed than previously thought. For the nineteenth century, David Eltis has provided yet a further revision of the volume of the slave trade. According to his latest reckoning, the nineteenth-century trade was 3 per cent higher than my 1982 calculations, which were based on his data. His new total for 1811-60 exceeds his old one by 81,000 slaves.39 Eltis does not explain the discrepancy between his earlier calculations and his latest revisions. Having reviewed the recent contributions, it is now possible to attempt a new synthesis. The combination of the various studies is as follows: Richardson has claimed to have revised my estimate of the British trade upward by 342,700; but if allowance is made for non-slave deliveries at 5 per cent, a figure of 186,700 is suggested. Curto has identified an additional 12,736 slaves in the Portuguese trade. Eltis is now claiming that an additional 81,000 slaves were exported between 1811 and 1867. These substitutions (rounded to the nearest thousand) represent increases of 281,000.40 An estimate for the French trade, based primarily on a conservative assessment of Becker's analysis, is 30,000 less than my earlier estimate (which incorrectly included the Mascarene trade), while Richardson's downward revision for the North American trade is 86,000. The net increase is 165,000 slaves. The new total is 11,863,000, which is higher than Curtin's 1969 estimate but still within an acceptable margin of error.41 36 As indeed may have been the case. " 'Angolan slave trade'. 38 In a personal communication, Curto has indicated that his data may warrant a revision that might be a few thousand higher still. 39 Eltis, Economic Growth, 241-54. 40 Richardson's analysis suggests an upward revision of 437,000. 41 If Richardson's estimates are used, then the new total would be in the order of 11,911,000. THE ATLANTIC SLAVE TRADE 373 In order to compare the new estimate with Curtin's original one, it is necessary to convert Curtin's figure for imports into the Americas (9,566,100) into an export figure from Africa. The difference includes the number of slaves who died during the Middle Passage, which suggests that anywhere from 10-20 per cent more slaves actually left Africa than reached the Americas. Curtin's estimate of imports would have required an additional 1-2—2-4 million exports. In addition, Curtin recognized a considerable margin of error. Any figure as high as 11 million slaves imported into the Americas would be readily acceptable within Curtin's framework. An estimate of 11,863,000 slaves exported from Africa, allowing for losses at sea of 10-20 per cent, would mean that 9-6-10-8 million slaves would have been imported into the Americas, which is well within Curtin's limits.42 The major problem in calculating the scale of the Atlantic trade, a problem that is unlikely to be solved, is an evaluation of the extent of missing data. The figure of 11,863,000 is the best estimate of slave exports to the Americas on the basis of current information, but there is no question that some slaves went unrecorded and that the available data contain errors that are difficult, if not impossible, to detect. How much allowance should be made for these factors is difficult to assess.43

Regional and Ethnic Origins of Slaves If the new data only modestly affect an assessment of the total volume of the Atlantic slave trade, the same cannot be said about our understanding of the regional origins of slaves, particularly in the eighteenth century. Richardson's analysis represents a major revision of my 1982 synthesis.44 His principal conclusions can be summarized as follows: First, the Bight of Biafra was more important and its involvement began earlier than previously thought. Sec- ondly, the French and British were more active on the Loango coast than earlier statistical analysis revealed. Thirdly, Richardson's analysis of the destinations of North American ships fills an important gap in our under- standing of regional origins. The implication of the first two points is that the southward shift in search of slaves started earlier and was more pronounced than I allowed for. The implication of the third helps one to re-assess the regional distribution of the West African market. Perhaps Richardson's most important discovery concerns the trade of the Bight of Biafra, which he shows to have been substantial from at least the beginning of the eighteenth century.45 This discovery arises from new information on British shipping. It confirms references to trade in the Niger

42 Compare with Lovejoy, 'Volume of the Atlantic slave trade', 496. 43 Joseph Miller's analysis of the export trade from west-central Africa suggests that there may be less missing data than some scholars might like to think, at least for the crucial eighteenth and nineteenth centuries. Portuguese records are particularly rich; the other countries were involved in an ' illegal' trade from the Portuguese perspective, but their shipments show up in their national records, especially as reflected in import series and shipping records. The regional distribution of slave exports within west-central Africa requires adjustment, but not the overall level of exports; see 'The origins and destinations of slaves in the eighteenth-century Angolan slave trade' (unpublished). I wish to thank Miller for letting me read this important contribution. 44 Richardson, 'Slave exports'. 46 Richardson, 'British slave trade'; 'Slave exports', Tables 6 and 7. 374 PAUL E. LOVEJOY Delta, which Curtin's (and hence my) analysis did not account for, and it makes sense in terms of backward projections of political history in Igbo country.46 Until now, it seemed that the Bight of Biafra did not begin to export slaves in large numbers until the 1730s; thereafter the trade appeared to have expanded rapidly. Richardson's data show a more gradual but nonetheless significant expansion that began before 1700. Richardson's other major revision concerns the trade of west-central Africa. The British and the French (according to Mettas) were heavily involved in buying slaves there from at least the 1740s, particularly in the region north of the Zaire River.47 The British obtained 20 per cent of their slaves from the Loango coast in the eighteenth century. The French, particularly from the 1740s, developed their trade on the Loango coast to a level that came to rival that of the Bight of Benin. The implication of the French and British presence is that west-central Africa was even more prominent in the trade than my regional projections of 1982 indicated, and those projections were startling enough. I concluded that west-central Africa exported more slaves than any other region, was prominent early in the trade, and continued as a major exporter to the very end of the trade. Not only does Richardson confirm these observations; it is possible to conclude that perhaps as many as 40 per cent of all slaves came from the interior of Angola and the Zaire River basin. Mettas' data, as analysed by Richardson, confirm that the French took most of their slaves from the Bight of Benin, with west-central Africa becoming a secondary region of concentration from the 1740s. The French purchased relatively few slaves in the Bight of Biafra. By contrast, North American shippers obtained many of their slaves in Sierra Leone, with secondary concentrations in Senegambia and the Gold Coast. While the North American share of the trade was relatively small, the implications of Richardson's analysis on the study of slavery in the United States are readily apparent. Slaves imported on North American ships tended to come from the Upper Guinea Coast. The relative importance of various regions can now be reassessed. It is clear that I over-estimated the suddenness, although not the intensity, of the southward shift of the trade in the eighteenth century; there was a shift, but it was more gradual. The proportion of slaves from the Bight of Biafra also appears to have been greater than I allowed for, with most of the increase destined for British colonies. By comparison, the Bight of Benin —the Slave Coast - retreats in significance slightly. Previously, it was thought that the Bight of Benin exported 50 per cent more slaves than the Bight of Biafra. Richardson estimates that both areas contributed about the same number of slaves.48 This analysis accepts Manning's assignment of those Dutch slaves

46 See David Northrup, Trade Without Rulers : Pre-Colonial Economic Development in South-Eastern Nigeria (Oxford, 1978); F. I. Ekejiuba, 'The Aro trade system in the nineteenth century', Ikenga, 1, 1 (1972), 11-26; and Lovejoy, Transformations in Slavery, 82-3, 99-100. 47 Richardson, 'Slave exports'. Joseph C. Miller demonstrates that slaves from west- central Africa, whether exported from Loango, Luanda or Benguela, came from interior regions that overlapped considerably (Way of Death : Merchant Capitalism and the Angolan Slave Trade, 1730-1830 (Madison, Wisconsin, 1988), 140-244); see also Miller, 'Origins and destinations of slaves'. 48 Richardson, 'Slave exports', 17-18. THE ATLANTIC SLAVE TRADE 375

Table i. Regional origins of the French slave trade (eighteenth century)

Plantation Total trade inventories

Region No. % No. 0/ /o Senegambia 83,860 80 1,380 10-3 Sierra Leone 48,730 47 206 '•5 Windward Coast — — 253 19 Gold Coast 65,150 62 633 47 Bight of Benin 351,240 33'5 4,552 34-1 Bight of Biafra 57,230 5'5 i,245 93 West-Central 441,410 42-1 4,928 37-0 Africa South-East — — 137 i-o Africa Total 1,047,620 ioo-o 13,334 998

Source: Richardson, 'Slave exports'; Geggus, 'French shipping and plantation records'. designated as ' Mina' to the Bight of Benin.49 Sierra Leone also appears to have been more important than previously thought, primarily because of the inclusion of the North American trade. Estimates of the regional origins of slave exports have obvious implications for our understanding of the ethnic backgrounds of slaves. More than 40 per cent (perhaps close to 50 per cent) of all slaves shipped to the Americas came from Bantu-speaking peoples, and most came from matrilineal societies. The overwhelming majority of slaves from the Bight of Biafra were Igbo, with a secondary concentration of Ibibio. Those from the Bight of Benin were more diverse in ethnic origin, but most were Gbe-speaking (Ewe-Fon) or Yoruba, with a heavy sprinkling of several interior populations, especially the Gurma cluster. There were also recognizable numbers of Hausa and Nupe by the early nineteenth century. Akan, from the Gold Coast, were in strong evidence, while exports from the stretch of coast from Sierra Leone to Senegambia display a more complicated pattern, with slaves divided between those who came from close to the coast and those who came from the interior (particularly Bambara). A number of scholars have examined the ethnic backgrounds of slaves in the

49 Patrick Manning, 'The slave trade in the Bight of Benin, 1640-1890', in H. A. Gemery and J. S. Hogendorn (eds.), The Uncommon Market: Essays in the Economic History of the Atlantic Slave Trade (New York, 1979), 141; Johannes Postma did not break down the Dutch figures sufficiently to resolve this question; see 'The origin of African slaves: the Dutch activities on the Guinea Coast, 1675-1795', in Engerman and Genovese, Race and Slavery, 49; and ' The Dutch slave trade: a quantitative assessment', Revue francaise d'histoire d'outre-mer, LXII (1975), 237. Manning believes that the Dutch tended to buy gold at their forts on the Gold Coast and slaves further east. 376 PAUL E. LOVEJOY

Americas. Following in the path of Gabriel Debien and his associates,50 David Geggus has used plantation inventories to examine the origins of slaves on French plantations.51 B. W. Higman has conducted similar research in the British Caribbean on the basis of census material, and Mary Karasch has studied burial records, the Court of Mixed Commission's papers on captured slave ships, and custom house registries for nineteenth-century Rio de Janeiro.52 Together with the data collected by S. W. Koelle on freed slaves in Sierra Leone in the nineteenth century, there is now a considerable amount of information on slave ethnicity.53 Geggus compiled a sample of 13,334 slaves for whom ethnic origins are provided in French plantation inventories. The regional origins of these slaves is roughly compatible with the regional origins for the French trade as a whole, as analysed by Richardson from Mettas' data (see Table 1). The plantation inventories appear to over-represent slaves from Senegambia and the Bight of Biafra and under-represent those from the Sierra Leone, Windward and Gold Coasts and also from west-central Africa. Geggus's Windward Coast can be combined with Sierra Leone for purposes of analysis here. If the coast from Senegambia to the Windward Coast is treated as a unit, the plantation inventories reveal that 13-7 per cent of all slaves came from this area, while the shipping records reveal that 12-7 per cent had their origins there. Since there was considerable overlap in the Sierra Leone and Senegambia areas, the difference between the various percentages is probably not significant. Geggus demonstrates that almost four-fifths (78-5 per cent) of the slaves whose ethnic origins were reported can be identified with one of six ethnic categories - Congo, Gbe (Ewe-Fon), Yoruba, Igbo, Bambara, and Akan (Table 2). 'Congo' indicates little more than a variety of Bantu-speaking peoples in west-central Africa, and 'Ewe-Fon' may well include people from further inland but who had become identified with coastal society, either because they were slave imports from elsewhere or because they reached the coastal ports through Ewe-Fon hands. The same can be said of Yoruba and Akan.54

50 See, especially, Gabriel Debien, Les Esclaves aux Antilles francaises (XVIl'-XVIII* siecles) (Basse-Terre and Fort-de-France, 1974). 51 David Geggus, 'Sex ratio, age and ethnicity in the Atlantic slave trade: data from French shipping and plantation records', J. Afr. Hist., xxx, 1 (1989), 23-44. 52 B. W. Higman, Slave Populations of the British Caribbean, 1807-1834 (Baltimore, 1984), 126-133, 442-458; and Mary C. Karasch, Slave Life in Rio de Janeiro, 1808-1850 (Princeton, 1987), 8-28, 371-83. 53 Philip D. Curtin and Jan Vansina, ' Sources of the nineteenth century Atlantic slave trade', J. Afr. Hist., v, 2 (1964), 185-208; E. M. Chilver, P. M. Kaberry and R. Cornevin, 'Sources of the nineteenth century slave trade: two comments', J. Afr. Hist., vi, 2 (1965), 117-120; Curtin, Atlantic Slave Trade, 244-57; 289-98; Northrup, Trade Without Rulers, 60-1; 231-3. Also see S. W. Koelle, Polyglotta Africa (London, 1854; reprinted with new introduction by P. E. H. Hair, Graz, Austria, 1963). 54 Geggus' sample can be compared with the smaller sample collected by Arlette Gautier, who has tabulated the ethnic origins of 1,812 slaves in southern St Domingue between 1721 and 1770. Her data are from the Archives de Nippes, which Geggus appears not to have used. Gautier also confirms the six principal ethnic categories, which comprised 78-3 per cent of her sample; see Arlette Gautier, 'Les origines ethniques des esclaves de Saint-Domingue d'apres les sources notariales', Canadian J. Afr. Studies, XXIII, i (1989). If Gautier's and Geggus' samples are combined, some of my conclusions would require adjustment, but the relative importance of the various ethnic categories THE ATLANTIC SLAVE TRADE 377

Table 2. Principal ethnic identifications on French plantations

Number of Ethnic category slaves Percentage

Congo 4,56i 34-2 Gbe (Ewe-Fon) 1,962 14-7 Yoruba 1,580 11-8 Igbo 1,129 8-5 Bambara 718 5-4 Akan 520 3'9 Other 2,864 21-5 Total I OO'O

Source: Geggus, 'French shipping and plantation records'. I have designated ' Mina' and ' Caramenty' as Akan. It should be noted that the distinction between 'Mandingue' and 'Bambara' may be artificial. For 'Ewe-Fon' I have included 'Arada', 'Adia', 'Foeda' ('Hweda'), and 'Fond'.

The relatively small number of ethnic groupings on French plantations is even more striking if the Gurma cluster ('Chamba'),55 Hausa, Mondonga, Malinke, Kotokoli, Nupe and Susu are included. Together, these groupings comprised an additional 11*4 per cent of the sample (1,514 out of 13,334). The thirteen largest ethnic categories comprised almost 90 per cent of the total sample. The data seem to indicate that masters perceived relatively homo- geneous groupings of slaves. would not change. The data compiled for Guadeloupe would alter these percentages further; see Nicole Frisch, 'Les esclaves de la Guadeloupe a la fin de l'ancien regime d'apres les sources notariales (1779-1789)', Bulletin de la Societe d'histoire de la Guadeloupe, LXIII/LXIV (1985). Ethnic origins of slaves in southern St Domingue, 1721-jo

Adjusted Ethnicity Per cent* percentage!

Ewe-Fon 190 152 Congo 188 324 Igbo 152 9-3 Yoruba 90 115 Akan 8-4 4'4 Bambara 7'9 S'7 Sub-total 78-3 78-5 Other 21-7 2I'S Total 1000 1000

* Based on Gautier's analysis. | Combination of Gautier's sample of 1,812 and Geggus' sample of 13,334. 55 In many of the sources that identify the ethnicity of exported slaves, there is a category 'Chamba', 'Thiamba', etc. The correct identification is with the related Gurma languages, not with Chamba, an important ethnic group in the Benue River Valley. 378 PAUL E. LOVEJOY

Table 3. Ethnic identifications of slaves from the Bight of Benin, 1721-97

Ethnicity Number Percentage

Gbe (Ewe-Fon) 1,962 43'2 Yoruba 1,580 34-8 Gurma ('Chamba') 297 6-5 Hausa 287 63 Kotokoli 166 3'7 Nupe 161 3-5 Bariba 84 19 Total 4,537 99.9

Source: Geggus, 'French shipping and plantation records'.

Mary Karasch's study of the ethnic origins of slaves in Rio de Janeiro reveals a similar pattern.86 In nineteenth-century Rio, there were seven principal ethnic groups, and several minor ones. The seven — Mina, Cabinda, Congo, Angola (or Loanda), Cassange, Benguela and - were associated with an African region of origin. Mina came from the Bight of Benin and were largely Muslim, which suggests that the slaves so identified came from the interior. Mozambique combined the slaves from south-east Africa. The 'nations' from west-central Africa actually represented four ethnic clusters: Kongo, Mbundu, Lunda-Cokwe, and Ovimbundu-Ngangela. The smaller ethnic groups (Gabao, Anjico, Monjola, Moange, Rebola or Libolo, Cajenge, Cabunda, Quilimane, Inhambane, Mucena and Mombaca) further confirm this clustering, since most of these are of west-central or south-eastern African origin.57 Karasch's data, which cover the period 1820-52, demonstrate that a degree of ethnic amalgamation was occurring under slavery in the Americas. Other ethnic origins were subsumed. Ethnicity under slavery tended to be identified with the commercial system through which slaves passed in Africa; that is the region and/or port of export. What slaves perceived is another matter, but it is likely that ethnic affiliations assumed new meanings under slavery and that the 'common' ethnic labels reflected historical amalgamation. The extent of such change should become the subject of research.58

56 Karasch, Slave Life in Rio, 8-28, 371-83. 57 Ibid., 10-20. 88 The data on ethnicity in early nineteenth-century Bahia reveal a very different pattern, which reflects the close ties between Bahia and the Bight of Benin. According to Joao Jose Reis, a very large proportion of identifiable ethnic groups between 1819 and 1836 were from West Africa. In one sample of 1,161 manumissions (1819-36), 76-2 per cent of individuals came from the Bight of Benin, and among the urban slave population of Salvador (sample: 1,480 slaves) between 1820 and 1835, 64^6 per cent of slaves came from the Bight of Benin. In contrast to Rio de Janeiro, far fewer slaves came from west- central Africa (i3'4 per cent in the first sample and 24'1 per cent in the second sample). 'Mina' might refer to Akan and/or Ewe-Fon in this sample. It should be noted that the number of Hausa, Borno and Nupe slaves was a significant proportion of the total (15-5 per cent in the first sample and 13-9 per cent of the second sample), which reflects the importance of the Sokoto jihad as a supplier of slaves, especially since some Yoruba slaves would also have been a product of the jihad. Reis provides an excellent analysis of revolts THE ATLANTIC SLAVE TRADE 379 Geggus's material makes possible a more detailed analysis of the ethnic identifications for Senegambia and the Slave Coast, where it is clear that many slaves came from the distant interior. Ethnic groups identified for Senegambia include 718 Bambara in a sample of 1,380 (52-0 per cent). Other categories (Senegal, Malinke, Fulbe) could have come from the coastal region or the interior, but probably were mostly of coastal origin or slaves who had come from further inland but whose identification was coastal. These data confirm what we know about the Senegambia trade. The ethnic configuration for the Slave Coast is also revealing, although again it is important to recognize that slaves who were identified as Gbe (Ewe-Fon), Yoruba, or Hausa, in particular, might have had some other ethnic origin. At least 21-9 per cent of the slaves whose ethnicity was identified came from the distant interior and were not otherwise identified as Yoruba or Gbe (Table 3). It should be noted that these findings are significantly different from Manning's analysis of the ethnic origins of slaves from the Bight of Benin. According to Manning, 72 per cent of the slaves exported between 1720 and 1800 (a comparable period to that analysed by Geggus) were Gbe, 15-5 per cent were Yoruba, and only 125 per cent came from the interior (the Gurma in Bahia in this period, particularly with reference to Islam and Yoruba orisha; see Joao Jose Reis, 'Slave rebellion in Brazil: the African Muslim uprising in Bahia, 1835', (Ph.D. thesis, University of Minnesota, 1983).

Ethnicity among slaves in Bahia, i8igs6

Urban slave Manumissions population Ethnicity (1819-36) (1820-35)

Yoruba 237 28'6 Ewe 24-8 16-2 West-Central Africa 134 24-1 ' Mina' 12-1 60 Hausa 96 9-5 Nupe 39 2-9 ' Borno' 2-I 1 "4 Igbo/Ibibio i'3 18 Other g-2 94 Total 1 oo-1 999 Sample I,l6l 1,480

Source: Reis, 'Slave rebellion in Brazil', 198.

Stuart Schwartz has established that a similar pattern in ethnicity prevailed in Bahia since at least the 1780s. In an inventory from 1803, two-thirds of imported slaves were from the Bight of Benin and only one-third were from west-central Africa (sample: 6,992). Yoruba, Nupe and Hausa together constituted about one-third of the slave population in the early nineteenth century; see S. Schwartz, Sugar Plantations in the Formation of Brazilian Society: Bahia, 1550-1835 (Cambridge, 1985), 341, 437. 380 PAUL E. LOVEJOY cluster, Nupe and Hausa).59 Manning's data appear to over-estimate Gbe and under-estimate other ethnic groups. Higman's data for the British Caribbean reveal a similar pattern of identification of slaves with several, major ethnic groupings. The most common were Congo, Malinke, Igbo, Akan and ' Moco,' with lesser concen- trations of Gurma ('Chamba'), Bambara, Temne, Susu, Hausa and Fulbe.60 There is a strong overlap in the ethnic identifications of the British and French samples. The most striking difference in regional origins for the two trades relates to the Bights of Benin and Biafra. The British trade reveals relatively few slaves from the Bight of Benin and a large percentage from the Bight of Biafra. The French trade shows the reverse, which is to be expected. The information on ethnic origins confirms that both the British and French were also heavily involved in the trade of west-central Africa. These data can be compared with other materials on the ethnic origins of slaves. Koelle's data, as analyzed by Curtin,61 show a much greater number of ethnic groups, although it should be recognized that Koelle deliberately tried to enlarge his sample to incorporate as many languages as possible. His data were obtained from free slaves in Sierra Leone, in contrast to the data used by Geggus (and Debien previously), which derive from plantation records in the Americas. Koelle was conducting research on the origins of slaves in order to collect linguistic materials, while those who compiled plantation records were gathering incidental information that they felt would be useful in controlling the slave population. The purpose of enquiry corresponds to the sophistication of data collection. Koelle's material is biased towards ethnic diversity; plantation records are biased towards ethnic homogeneity. Higman's material on the ethnic origins of slaves in the British Caribbean, which is derived from census data, falls midway between the diversity of the Koelle inventory and the relative ethnic concentration of the Geggus analysis.62 There are several ways to explain the discrepancy in the extent of ethnic diversity other than the relative sophistication of techniques in data collection. First, the patterns for the eighteenth and nineteenth centuries may have been different, with a greater concentration of ethnic groups in the eighteenth century than in the nineteenth century, but this seems unlikely. Secondly, data based on plantation inventories and census records may reveal a degree of ethnic amalgamation on plantations that was not as pronounced on the African coast, where ethnic distinctions were more clearly recognized. Only future research and analysis will determine which of these factors is most important and whether or not there are other factors. The time has probably come for such analysis. The material on ethnicity, including that analyzed by Curtin, Debien, Higman and Geggus, is sufficiently large that a full study is called for.

59 Patrick Manning, Slavery, Colonialism and Economic Growth in Dahomey, 1640-IQ60 (Cambridge, 1982), 335-9. According to Manning, total slave exports from 1720 to 1800 were 799,300 slaves, of which 576,000 were Gbe (Aja), 124,300 Yoruba, 81,600 Voltaic, 7,900 Nupe, and 10,500 Hausa. These estimates were based on a combination of several New World slave inventories, including some from Brazil, which may explain the differences in the percentages. 60 Higman, Slave Populations, 129, 131-3. 61 Curtin, Atlantic Slave Trade, 244-57, 289-98. Also see Northrup, Trade without Rulers, 60-2, 231. 62 Higman, Slave Populations, 126-33; 442-57. THE ATLANTIC SLAVE TRADE 381

Age and Gender Profile of the Exported Slave Population Recent publications have contributed a great deal to our knowledge about the age and gender profile of the exported slave population. Geggus has compiled the available data on the late seventeenth and eighteenth centuries and also includes some material from the nineteenth century.63 Eltis has performed a similar task for the nineteenth-century trade, while Joseph Miller has. undertaken a useful analysis of age and gender for west-central Africa in the eighteenth and nineteenth centuries.64 In a series of tables, Geggus compares the Mettas data with earlier studies.65 His analysis includes the age and gender profile of 721,949 slaves, approxi- mately 6 per cent of the whole trade, who were shipped to the Americas between 1636 and 1867. The various samples cover most national carriers and cover most of the important American colonies, although North American ships are not included, and the Portuguese are under-represented. While data are relatively sparse before the last quarter of the seventeenth century, there is considerable information on the period after 1675. Based on Geggus' study, I calculate that the ratio of males to females was about 181:100 for the trade as a whole from the seventeenth century until the end of the trade in the nineteenth century; that is, 64-4 per cent of the slaves were male and 35-6 per cent were female. The eighteenth-century French trade displayed an overall sex ratio of 179 males per 100 females (64-2 per cent male), virtually the same as the overall pattern. Geggus concludes that 'the slave- traders ' oft-stated target of two males for every female appears to have been only rarely attained'.66 Technically, Geggus is correct for the seventeenth and eighteenth centuries and is entirely wrong for the nineteenth century. The figures that he uses indicate that the slave traders took 63-0 per cent males before 1800 (170:100) and 68-5 per cent males after 1800 (217:100). Eltis' analysis confirms the high ratio of males in the nineteenth-century trade. On average slavers were able to purchase more than two males for every female, although it should be noted that it was more likely a result of African supply patterns than European demand.67 Indeed, ratios of males to females rose to 70-8 per cent on the coast north of the Zaire River between 1811 and 1867 and as high as 80 per cent in south-eastern Africa in the same period.68 A more accurate assessment would be as follows: the slave traders seldom achieved the desired goal of two males for every female in the seventeenth century; they continued to have difficulties in West Africa, but not in west- central Africa, in the eighteenth century; and they easily attained that mark in west-central Africa, and often in West Africa as well, in the nineteenth century. Whether or not they did depended upon the section of the coast and the period. 63 Geggus, 'French shipping and plantation records'. 64 David Eltis, ' Fluctuations in the age and sex ratios of slaves in the nineteenth- century transatlantic slave traffic', Slavery and Abolition, vn, 3 (1986), 257-72; Eltis, Economic Growth, 69-70, 255-9; Miller, Way of Death, 346-9, 387-8. 65 Geggus, 'French shipping and plantation records', 24, Table I, but strangely not including Eltis, 'Age and sex ratios'. 66 Geggus, 'French shipping and plantation records', 26. 67 Eltis, 'Age and sex ratios', 267-9. 68 Ibid., 259. It should be noted that Geggus ('French shipping and plantation records', 26) overstates his case in claiming that 'in no trade from Africa did the known proportion of males exceed, or even equal, 70 per cent'. 382 PAUL E. LOVEJOY The lowest reported male/female ratios were in the seventeenth century. In two samples (2,064 and 2,269 slaves), the ratios were nearly equal (105 and 108 males for every 100 females or 51-2 per cent and 51-9 per cent male, respectively), while in a third sample (3,086) the ratio was 138:100 or 58-0 per cent male. These samples are small but suggestive. There seems to have been no particular preference for males or females on the African coast at the time, in contrast to later periods. African demand for women appears not yet to have developed enough to influence the ratio of males to females in the export trade. Such an influence appears to have increased during the course of the trade, although regional variations were usually strong. While it is known that the volume of the trade increased substantially in the seventeenth century, it has not hitherto been known when volume began to affect local politics and society. The available data may provide a clue to timing, at least for those portions of the coast that were first drawn into the trade in a serious way. The proportions of males to females in the eighteenth-century trade reveal clear developments. First, the proportion of males appears to have increased substantially, assuming that the few samples from the seventeenth century are an accurate indication of sex ratios in that period. Secondly, the increase was fairly gradual for West Africa, with male/female ratios still well below the desired ratio of two males for every female. Thirdly, the increase in the percentage of males from west-central Africa appears to have been dramatic, even reaching the preferred level of European slavers.69 For West Africa as a whole, the sex ratio for French shipping was 158:100, or 61-2 per cent male. Sierra Leone was low (57"3 per cent male), and even the Bight of Benin, which supplied the majority of slaves shipped by the French from West Africa, had a male/female ratio not much above the levels of the seventeenth century (61-7 per cent male). Ratios were lowest in the Bight of Biafra (539 per cent male). Data based on British and Dutch shipping for the Bight of Biafra also reveal very low proportions of males in the seventeenth century which continued through the eighteenth century. When French, British and Dutch shipping are taken together, the proportion of males to females purchased in the Bight of Biafra was only 129:100, or 56-4 per cent male (sample: 20,492).70 In west-central and south-eastern Africa, by contrast, the French obtained a ratio of two males for every female (679 per cent male) in the eighteenth century.71 A remarkable shift occurred, probably in the 1740s, when the French began to concentrate on the Loango coast. The ability to buy males may explain why west-central Africa became a major destination of slavers from the Netherlands and Britain as well. The goal of purchasing twice as many males as females appears to have been first realized there. Attempts to replicate this experience in West Africa seem to have failed in the eighteenth century. The Dutch stand out as an anomaly in the trade because of their success in obtaining male slaves, and it may be that they were a factor in the growing

69 Geggus, 'French shipping and plantation records'. These conclusions are based on Geggus' analysis of French shipping records for all regions and French, British and Dutch records for the Bight of Biafra. 70 Ibid. Geggus includes British, Dutch and French samples for his calculation. The French sample alone was even lower: 53-9 per cent male. 71 Ibid. THE ATLANTIC SLAVE TRADE 38.3 demand for males. Between 1675 and 1740, the Dutch maintained a ratio of 228:100 (sample: 36,121), or 60/5 per cent males, which is very high for any period and extremely high for the late seventeenth and early eighteenth centuries.72 The Dutch were the dynamic force in the export trade at the time, which may indicate a connection between expanded demand for slaves and a shift in sex ratios towards more males. Furthermore, the regional origins of the Dutch sample are unclear and may indicate a move in Dutch slaving from West Africa to west-central Africa, where males seem to have been easier to buy. An analogy with St Domingue is suggestive. In the eighteenth century, St Domingue was the richest and fastest-growing plantation colony and conse- quently was able to buy 'better' (i.e. male) slaves than its competitors. The Dutch may have been able to out-bid the British, French and Portuguese in the search for males.73 The shift towards higher ratios of males to females that appears to have begun with the Dutch may have been a result of an effort to establish and/or maintain a competitive advantage. High male ratios characterized the nineteenth-century trade, when the trade was under pressure from abolitionists. Buyers may have wanted strong workers in order to maximize returns on their investment, precisely when African sellers wanted to retain women. Eltis' data, based on a sample of 114,225 slaves between 1811 and 1867, reveal that 68-4 per cent were males.74 The Cuban trade, for example, consisted of 60/6 per cent males (229:100) between 1811 and 1867 (sample: 5i,577).75 The male/female ratio tended to rise for all regions. The Bight of Biafra, which exported almost as many females as males from the late seventeenth century and continuing through the eighteenth century, experienced a con- siderable shift. In the nineteenth century, the Bight still had the lowest male/female ratio of any major exporting region, but the proportion of males rose from 53-9 per cent in the eighteenth century to 66-1 per cent in the nineteenth century.76 A reconsideration of Geggus' calculations of male/female ratios for French slave merchants reveals that African middlemen were selling two males for every female during most of the eighteenth century in west-central and south- eastern Africa but not in the Bights or elsewhere in West Africa. According to Mettas' data, as summarized by Geggus, 67-9 per cent of slaves purchased in 72 Johannes Postma, 'Mortality in the Dutch slave trade, 1675-1795', in Gemery and Hogendorn, Uncommon Market, 257. Earlier statistics on the Dutch trade reveal only a slight difference in sex ratios between West Africa and west-central Africa. In Dutch shipments between 1637 and 1645, West Africa exported 1,791 males in a total trade of 3,086, or 58-0 per cent male, while west-central Africa exported 1,286 males in a total trade of 2,064, or 63'3 per cent. It is worth noting, moreover, that the West African ratio was lower than the pattern for the eighteenth century, when the ratio was 61-2 per cent, at least in the French trade. The west-central Africa ratio may have been on the rise and was already above the eighteenth-century ratio for West Africa. Statistics from the Portuguese trade, if available, should clarify the situation. For the 1637-45 period, see Ernst van den Boogaart and Pieter C. Emmer, 'The Dutch participation in the Atlantic slave trade, 1596-1650', in Gemery and Hogendorn, Uncommon Market, 366. 73 Geggus, 'French shipping and plantation records'. 74 Eltis, 'Age and sex ratios', 257, 259. 75 Geggus, 'French shipping and plantation records'. 76 In the nineteenth century, the ratio was 195:100, based on a sample of 24,502; see Geggus, 'French shipping and plantation records'. Eltis ('Age and sex ratios', 259) reports a similar ratio for the period 1811-67: 65-7 per cent. 384 PAUL E. LOVEJOY west-central Africa were males. Considering that the sample was large (46-1 per cent of 137,400 slaves for whom gender and regional origin were identified), this fact is particularly significant and runs counter to the conclusion that Geggus reached. For west-central and south-eastern Africa combine'd, 67-9 per cent of the exported slave population was male (32-1 per cent female), a ratio of 212:100. Whether or not the same pattern prevailed for British and Portuguese shipping remains to be discovered, but in the nineteenth century the proportion of males rose even further in west-central Africa, reaching levels close to 70 per cent.77 The age profile of the exported population indicates that children were under-represented early in the trade but that they became significantly more numerous as time passed. Regional variation also marked the trade in children. Admittedly there are difficulties in determining which slaves were 'children,' but there is no question that the trade increasingly concentrated on ever younger populations, even as there was a shift towards more males. Patrick Manning has estimated that children aged 14 or younger probably constituted 34 per cent of a stable population in Africa during the slave trade era, given what is known about patterns of fertility and mortality.78 Before the nineteenth century, only 19-5 per cent of exported slaves were 'children', however defined.79 According to Geggus, the French trade had a higher percentage of children (26-5 per cent) than the non-French trade (16-2 per cent), but slavers in general appear not to have purchased a proportionate number of children. It is not known whether this was a result of African or European preferences. Considering the fact that children, especially males, would come to dominate the trade in the nineteenth century, the fact that they were under-represented in the eighteenth century is worth noting. Geggus correctly observes that the French bought more children than their competitors before 1800, but he inadvertently distorts the picture. Slave merchants, apparently without regard to nationality, bought fewer children before 1800 than after. The regional variations in the age structure of the eighteenth-century trade were considerable. More children were purchased in Sierra Leone (35-0 per cent), the Bight of Biafra (30^9 per cent) and the region of west-central Africa (30-4 per cent) than in the other exporting regions, although again it should be noted that, except from Sierra Leone, the proportion of children was still lower than the number of children in the African societies from which the children came. In British, French, and Dutch samples for the Bight of Biafra in the seventeenth and eighteenth centuries, children comprised 28-2 per cent of the trade (sample: 16,427), which is remarkably low in a society in which much of the enslaved population had been kidnapped and probably tended to be young.80 The percentage of children taken from Africa rose dramatically. Again, west- central Africa led the way, and the shift was most dramatic there. Already by

" Eltis, 'Age and sex ratios', 259. 78 Patrick Manning, 'The impact of slave trade exports on the population of the western coast of Africa, 1700-1850', in S. Daget (ed.), Actes du colloque sur la traite des noirs (Paris/Nantes, 1988). See also Manning, Slavery and African Life (Cambridge, forthcoming). I wish to thank Manning for showing me this important study. 79 Geggus, 'French shipping and plantation records'. I have not included the Cuban sample, 1790-1829, in this calculation. 80 Ibid. THE ATLANTIC SLAVE TRADE 385 1784-95, 'the majority of slaves... were children, mostly boys'.81 In samples derived from the Portuguese and Spanish trade to Brazil and Cuba between 1790 and 1867, 41-7 per cent of all slaves were children (sample: i8o,586).82 Eltis has calculated an average of 41-4 per cent for all regions of Africa between 1811 and 1867.83 In some regions the increase was very large. The area immediately north of the Zaire River exported 52-5 per cent children; the proportion of children from Angola was 59-0 per cent, and cargoes from south- eastern Africa contained an alarming 6i-o per cent children. The Bight of Benin exported the lowest proportion of children of any region in the nineteenth century, 33-2 per cent, approximately the same ratio as in society as a whole.84 The proportion of children from the Bight of Biafra rose to 38-9 per cent, just slightly more than the upper Guinea coast (38-2 per cent).85 The Atlantic trade had become a trade in children, and the skewing of the export population followed the pattern with respect to males — the matrilineal areas of west-central and south-eastern Africa dispensed with males and children in disproportionate numbers. West Africa followed suit, but more slowly and to a lesser extent. More sophisticated analysis may be able to discern when, where and how quickly the shift toward more males and younger people occurred. The difficulty in identifying children may exaggerate the pattern, and hence the present analysis should be accepted as a challenge rather than as a definitive answer to the question: Were children as important in the trade as the recent literature seems to suggest ? Until that question is addressed, it is difficult to suggest correlations between change in age and gender profiles and internal African developments. There appears to have been a relationship between a high proportion of males and distance from the coast. Manning initially suggested that internal African demand, disease and other factors lowered the number of women and children from the far interior who were available for export. Manning appears to be correct with respect to males but his observations on children need to be qualified, if the nineteenth-century trade is an indication of the willingness to move children great distances.86 In the Bight of Biafra, where the sex ratio of the exported population was relatively balanced and slaves came from near the coast, the number of children was high in proportion to their numbers in society. For west-central Africa, however, many slaves came from the distant interior, particularly in the nineteenth century.87 There were more males, and 81 According to Miller (Way of Death, 159, 346-8, 387-9), one of the reasons west- central Africa led the way in the shift toward more children related to attempts at 'tight packing' as a result of Portuguese regulations on shipping slaves. Also see Eltis, ' Sex and age ratios', 262, for a graphic portrayal of the nineteenth-century trend. 82 Geggus, 'French shipping and plantation records', Table 1. 83 Eltis, 'Age and sex ratios', 259. 84 Ibid. 259. 86 Ibid. 259. Geggus, 'French shipping and plantation records', has virtually the same ratio for the Bight of Biafra as Eltis. 88 Manning argues that ' Children had no special premium price in the interior, but they tended to be kept there because a long march to the coast would cost more in mortality and maintenance than what the child would bring on arrival'; see ' Impact of slave trade exports'. His argument appears to be correct for small children but not for older boys; see Miller, Way of Death, 159, 347-8, 387-9. 87 Miller (Way of Death, 148, 149, 153-4, J 59-64, 387-9) emphasizes distance from the coast as a factor that influenced the ratio of males to females in the export trade and notes that there was a tendency to bring increasing numbers of boys from the interior. 386 PAUL E. LOVEJOY they tended to be ever younger. Clearly there was an age below which boys could not be moved without great losses, and young children and babies (boys and girls) usually travelled with their mothers, but not at any distance with ease. The age and sex profile reveals that boys of an age when they could withstand forced marches tended to be a valuable commodity.88 That age depended upon individual boys, but was probably around ten and above. The* greatest proportion of males came from south-eastern Africa, which is an ironic gloss on Manning's rule that there was a correlation between distance that slaves travelled and gender. Mozambique was the furthest distance from the Americas, and proportionately more males came from there than any other exporting region. It is not clear why. Variations based on different regions and different periods require careful scrutiny by historians to see whether or not they can be correlated with historical changes in the exporting regions. No attempt is made here to revise my earlier analysis,89 although Manning is in the process of doing so.90

Impact of the Atlantic Slave Trade on Africa The discussion of the volume of the trade, the regional and ethnic origins of the exported population, and the sex and age profiles of slaves should indicate that much of Eltis' revisionist interpretation cannot be accepted. Otherwise, these factors would not matter to African history, although they are crucial to the history of slavery in the Americas. While the slight modifications in the volume and direction of the Atlantic slave trade do not affect the argument of Eltis (and Jennings), the issues remain: did the slave trade have a dramatic impact on exporting regions ? Did the suppression of the Atlantic trade in the nineteenth century have a significant effect on the course of slavery ? My informed opinion is that both the trade and its suppression were major factors in African history, and to show this I will examine, in order, the following issues: (1) the economic impact of the trade; (2) its demographic implications; and (3) the incidence of slavery in Africa. There are certainly other issues, but these will have to suffice. One of the principal conclusions of Eltis and Jennings seems likely but only modifies my analysis: Africa's share of world trade from the late seventeenth until the mid-nineteenth century was relatively small in comparison with other parts of the Atlantic world, and Africa's share of that trade declined in relative terms during the period of the slave trade.91 Eltis and Jennings use an estimate of jCo-8-j£n for per capita incomes in western Africa for the 1780s. They calculate that the export trade amounted to £o-i per person each year.92 The

88 Patrick Manning, 'Contours of slavery and social change in Africa', American Hist. Review, LXXXVIII, 4 (1983), 847; Manning, 'Impact of slave trade exports'; Geggus, 'French shipping and plantation records'; Miller, Way of Death, 159-69. 89 See Transformations in Slavery. 90 See his forthcoming monograph, Slavery and African Life. 91 Eltis and Jennings, 'Western Africa and the Atlantic world', 957. Eltis and Jennings define ' western Africa' as the region that supplied slaves to the Atlantic trade, including West Africa and west-central Africa. 92 Eltis relies on Henry A. Gemery and Jan S. Hogendorn, 'The economic costs of West African participation in the Atlantic slave trade: a preliminary sampling for the eighteenth century', in Gemery and Hogendorn, Uncommon Market, 153, for the estimate THE ATLANTIC SLAVE TRADE 387 proportion of exports to total income certainly appears to be very low by comparison with other parts of the Atlantic basin, although the room for error in measuring per capita income and the value of the export trade are enormous. This observation on the value of the export trade can be accepted, but the implications that Eltis and Jennings draw from it cannot. The ratio of the value of the external trade to per capita income is not an accurate indicator of the impact of the slave trade on Africa. Per capita income in western Africa was certainly very low by the standards of other parts of the Atlantic basin. Africa was very poor. Almost any incremental increase over subsistence would have had a disproportionate impact on the economy. Eltis and Jennings quantify the relative poverty of Africa, but they are wrong to conclude that the lack of prosperity was an accurate gauge of the degree of isolation from the impact of the slave trade. The simulation model developed by Patrick Manning provides one way to establish that the slave trade had a significant, indeed devastating, impact on Africa.93 Manning's model is a statistical means of measuring demographic change under conditions of enslavement, slave trade and slave exports. His analysis is based on the demography of the Atlantic slave trade and certain broad assumptions about demographic change that establish the parameters of the historical possibilities. Manning estimates that the population of those areas of West and west-central Africa that provided slaves for export was in the order of 22-25 million in the early eighteenth century. He projects a growth rate for that population during the eighteenth and early nineteenth centuries of 0-5 per cent, which he considers the maximum possibility. His simulation model 'suggests that no growth rate of less than one per cent could have counterbalanced the loss of slaves in the late eighteenth-century'.94 He uses a counterfactual argument to make the same point: 'With a growth rate of 0-5 %, the 1700 populations... [of] 22 to 25 million would have led to 1850 populations of from 46 million to 53 million, more than double the actual 1850 popu- lation'.95 Manning concludes that 'the simulation of demographic impact of the Atlantic slave trade provides support for the hypothesis of African population decline through the agency of the slave trade'.96 Furthermore, the simulation model suggests that the incidence of slavery increased in Africa. According to Manning, 'As an accompaniment to the estimated nine million slaves landed in the New World [between 1700 and 1850]..., some twenty-one million persons were captured in Africa, seven million of whom were brought into domestic slavery, and five million of whom suffered death within a year of capture'.97 As the discussion of the sex profile of the export trade makes clear, more women were retained in Africa than men. Not only did the slave population increase, therefore, but the incidence of polygyny increased as well. Indeed, the two phenomena were closely related. By 1770 the Atlantic trade resulted in a slave population in the Americas of approximately 2,340,000.98 Manning's simulation suggests that the slave on per capita income, although he inexplicably lowers Gemery's and Hogendorn's upper limit; see Eltis, Economic Growth, 72. For the estimated value of the export trade, see Eltis and Jennings, 'Western Africa and the Atlantic world', 956. 93 Manning, ' Impact of slave trade exports'; and Slavery and African Life. 94 Manning, 'Impact of slave trade exports'. 95 Ibid. 9° Ibid. «" Ibid. 98 Robin Blackburn, The Overthrow of Colonial Slavery, 1776-1848 (London, 1988), 5. 388 PAUL E. LOVEJOY population in West and west-central Africa could not have been much different. It is safe to say that the slave population was at least 10 per cent of the total population of 22—25 million and that the proportion of slaves was rising. Manning concludes that there were 3 million slaves in those parts of Africa that serviced the Atlantic trade at the turn of the nineteenth century, virtually the same number as in the Americas." The dramatic growth in the African slave population is the transformation that I highlighted initially in 1979 and more fully in 1983.10° The trans- formation was the result of a dialectical relationship between slavery in the Americas and the enslavement, trade and use of slaves in Africa. The production of slaves for the Americas also produced slaves for Africa. It is difficult to prove that the Atlantic slave trade caused the transformation of slavery in Africa, but it is likely. The simulation model, as well as the thesis that I advanced in Transformations in Slavery, consider that enslavement, trade and the use of slaves were interrelated, across the Atlantic and across the Sahara. African, European, and Muslim merchants wanted slaves, and African, European and Muslim slave-owners used slaves. The low value of exports and imports apparently confused Eltis and Jennings so that they did not perceive the importance of this interrelationship, but an examination of any part of Africa that was a supplier of the export trade reveals the dialectic. Miller's study of Angola provides the most dramatic example of impact of the slave trade on a region in Africa. As the export data make clear, approximately 40 per cent of all slaves in the Americas came from west-central Africa, and Miller estimates that deaths in Africa related to capture and enslavement roughly equalled the number of slaves exported, that is 50,000- 60,000 per year in the last half of the eighteenth century. In addition, 'fully as many more people [were] seized as slaves but left to reside in other parts of western central Africa'.101 Total population displacement would have been in the order of 100,000-120,000 per year. Admittedly, Miller paints this as a worst-possible scenario, but even if demographic change was less severe it must have been dramatic. No one has argued as much, but it may be that matrilineality and the export trade were interrelated. They certainly reinforced each other. Miller's analysis confirms the gender and age structure of the trade. In pursuing a discovery made earlier by John Thornton, Miller shows that the sex ratio of the population in those areas most heavily involved in the export trade was strongly skewed towards women and girls.102 Polygyny was a central institution of wealth and political power, and slaves (females) were most

99 Manning, Slavery and African Life, ch 4. 100 'Indigenous African slavery', in Michael Craton (ed.), Roots and Branches : Current Directions in Slave Studies (Toronto, 1979), 19-61; and Transformations in Slavery. As I have acknowledged, Walter Rodney, Philip Curtin, John Fage and other scholars provided-the inspiration for this insight. 101 Miller, Way of Death, 153. Miller is able to postulate which areas were most severely hit and when (see map, 148). 102 Miller, Way of Death, 160-5; see also John Thornton, 'The slave trade in eighteenth-century Angola: effects on demographic structures', Canadian J. Afr. Studies, xiv, 3 (1980), 417-27; Thornton, 'An eighteenth century baptismal register and the demographic history of Manguenzo', in African Historical Demography (Edinburgh, '977)> 1. 4°5~'5> Thornton, 'The demographic effect of the slave trade on Western Africa, 1500-1850', in African Historical Demography (Edinburgh, 1981), n, 691-720. THE ATLANTIC SLAVE TRADE 389 heavily concentrated around the principal courts of the region.103 Miller refers to the centralization that was associated with the slave trade as the 'great transformation', which is solace to the theorist.104 On the basis of Miller's analysis, it is impossible to conclude that the slave trade had a marginal impact on west-central Africa.105 Eltis is on shaky ground in suggesting that the suppression of the slave trade was not a decisive event for Africa. According to his interpretation, the increased incidence of slavery in the nineteenth century was unconnected with the collapse of the Atlantic trade. Instead, increased demand for slaves arose from ' rejuvenated Islam' and late in the century from European demand for primary products from Africa.106 Indeed there was a dramatic increase in slavery as a result of the jihads and Muslim commerical expansion in East Africa.107 Although Eltis does not explicitly argue the point, his proposed revision is one of timing and concentration, not of substance, and he presents no new data. I remain unconvinced. Certainly there was an expansion in production, based on slave labour, late in the nineteenth century, but that development is also part of the transformation thesis. Eltis disputes the thesis that the slaving frontier continued to move inland during the early nineteenth century and that the number of slaves in Africa increased dramatically in these decades. He bases this conclusion on an analysis of slave prices: ...because the price of all slaves declined, it seems clear that although domestic [African] demand increased, it did not increase sufficiently to offset the decline in trans-atlantic demand. As a consequence, the number of slaves traded as well as the price of those slaves declined during the century...; accordingly, suppression must have meant some reduction in the enslavement of Africans.108 Eltis is correct that the price of slaves dropped between the 1790s and the i8aos and continued at a depressed level for most of the continent for the rest of the century. Slaves were cheap, and in real terms prices may have continued to fall in many of the major slave markets, but this does not mean that enslavement decreased in intensity.109 The cheapness of slaves reflects two factors, the generally low prices of basic commodities in Africa and the glut of the slave market. Before the suppression of the trans-Atlantic trade, according to Eltis, the demand for slaves was divided into two sectors: one within Africa and the second in the Americas. (In fact, there was a third sector - the external Muslim market.) The collapse of American demand inevitably depressed prices in

103 Miller, Way of Death, 135. 104 Ibid. 135-9. 105 Curiously, Miller concludes that the slave trade had a marginal impact on the level of population west-central Africa, despite his convincing and overwhelming array of data to the contrary; see Way of Death, 165-9. It should be noted that Miller attempts to distinguish between sheer demographic impact and all other kinds of influence, mostly institutional, and his demographic analysis tries to identify population relocations, shifts in sex and age ratios, among other factors. 106 Eltis, Economic Growth, 226. 107 Ibid. 225. Eltis claims that this interpretation is a modification of my ' transformation thesis', but in fact an appreciation of the Islamic factor, virtually unrelated to the trans- Atlantic trade, is an integral part of my argument; see Transformations in Slavery, 184-219. 108 Eltis, Economic Growth, 227. 109 Ibid. 41, 260-4; see also Manning, Slavery and African Life. 39.O PAUL E. LOVEJOY Africa. He concludes that the African market did not increase sufficiently to offset the loss of American sales; according to Eltis demand and supply declined, although he only provides slave prices as evidence. Prices did not rebound after the decline of the 1790—1820 period, it is true, but the reason was a combination of factors. Indeed Eltis shows that American demand did not decline, in the aggregate, in this period but regained its former heights. Decline only began in the 1850s, well after the period that is crucial to his argument. Demand is an expression of price, so if prices were generally low, then demand might appear to have been depressed. In fact, the contrary could have been true, if more were known about the price structure of African economies. African demand for slaves was determined by the value of what slaves could produce, and this marginal revenue product was primarily a function of the value of food consumption, housing costs, social requirements, taxation and similar expenditures that were not much affected by international markets. Any qualitative assessment of the eighteenth and nineteenth centuries would judge Africa to have been comparatively poor, as the ratio of exports to per capita income reveals. Prices in general were low, and the price structure would have influenced the cost of slaves accordingly. Eltis and Jennings, in concentrating on export-led economic development, have raised an important consideration: more needs to be known about internal African price structures. Until such information is available, however, it is hard to justify their conclusions, and there is sufficient evidence available to argue the contrary. Eltis' conclusion that enslavement declined in the early nineteenth century contradicts the research of most Africanist historians who have written on the period. In almost every part of Africa for most of the century, enslavement was rampant. Slaves were generated on a scale previously unknown, as can be attested by the following examples: the wars of the mfecane in southern and central Africa, the activities of Arabs, Swahili, Yao, Nyamwezi, Chikunda and others in eastern and south-eastern Africa, and the raiding of Muslim slavers in the southern Sudan and north-central Africa. None of these cases have much, if anything, to do with the suppression of the Atlantic slave trade, and hence could be dismissed by those who favour the Eltis thesis. But what about the phenomenal levels of enslavement during the jihads, often in areas that once did or could have fed the Atlantic trade ? How are the collapse of the Lunda states and the havoc of enslavement instigated by the Cokwe to be explained ? Can the insecurity of Igbo country in the nineteenth century and the enslavement resulting during the Yoruba wars be easily dismissed ? The combined impact of these phenomena was to maintain a glutted market and hence a depressed price for slaves almost everywhere. The 'transformation thesis' holds that the external slave trade, particularly the trans-Atlantic sector but also the Islamic market, shaped slavery and society in Africa, and that internal factors intensified slavery as the external trade contracted. The enslavement of people and the growth of the slave population in Africa continued apace for the whole of the nineteenth century, despite local variations. As yet there are few estimates of the scale of the African slave population, but some insights can be gained by a comparison of certain parts of West Africa in c. 1900 with the Americas on the eve of emancipation there.110 110 It could be argued that a more accurate comparison would include slaves and the descendants of slaves. The purpose of the present comparison is intended to suggest the THE ATLANTIC SLAVE TRADE 39I

Table 4. Estimated slave populations in the Americas (late eighteenth century)

Region/country Slave population

French West Indies 575.000 British West Indies 467,000 Spanish West Indies 80,000 (Caribbean 1,122,000) Brazil 1,000,000 U.S.A. 575.000 Mainland Spanish America 271,000 Total 2,968,000

Source: Herbert Klein, African Slavery in Latin America and the Caribbean (New York, 1986), 295.

The slave population of the Americas rose from 2,340,000 in c. 1770 and peaked at 2,968,000 by the end of the century (Table 4).111 The revolt of St Domingue reduced this total considerably; St Domingue had a slave pop- ulation of 480,000 in 1791.112 The independence of mainland Spanish America after c. 1820, with its slave population of a couple hundred thousand, and the emancipation of 674,000 British slaves in 1834 reduced the total further, but the number of slaves continued to expand in the Spanish Caribbean, the U.S.A. and Brazil, reaching a peak just before the emancipation of U.S.A. slaves in the early 1860s. In i860, there were almost 4 million slaves in the U.S.A. and another 1-5—i-8 million slaves in Brazil and the Spanish Caribbean, for an estimated total of 5-5—5-8 million slaves.113 With the freeing of slaves in the U.S.A., the slave population declined considerably to a level well below two million. Puerto Rico had 47,000 slaves in 1867; Cuba 288,000 slaves in 1871, and Brazil 1,511,000 slaves in 1872. With the final emancipation of slaves in Cuba in 1880 and Brazil in 1888, slavery came to an end in the Americas. We may compare the American figures with those of the Western Sudan that have been assembled by Martin Klein (Table 5).114 Various estimates between 1905 and 1913 put the slave population of Haut-Senegal-Niger at about 702,000, or 18 per cent of the total population of 3,942,000. The slave population of French Guinea was 490,000, or 34-6 per cent of the total magnitude of the issue, not its complexity. To include the freed black population of the Americas in the analysis would require an inclusion of the descendants of slaves in Africa, which would greatly complicate the discussion, particularly since it would mean that the scale of ' slavery' in Africa would increase correspondingly. The usual argument is that slaves and their descendants tended to be ' assimilated' , and hence the numbers of people affected by the ' slavery' legacy was geometrically greater in Africa than the Americas. 111 Blackburn, Overthrow of Colonial Slavery, 5; Herbert Klein, African Slavery in Latin America and the Caribbean (New York, 1986), 295. 112 Curtin, Atlantic Slave Trade, 78. 113 Blackburn estimates the total in i860 at six million; see Overthrow of Colonial Slavery, 544. 114 Martin A. Klein, 'The demography of slavery in Western Soudan: the late nineteenth century', in Dennis D. Cordell and Joel W. Gregory (eds.), African Population and Capitalism : Historical Perspectives (Boulder, Colorado, 1987), 52, 54. 392 PAUL E. LOVEJOY

Table 5. Estimated slave populations, Africa and the Americas (nineteenth century)

Region/country Date Slave population Americas British West Indies 1817 753,ooo British West Indies 1832 674,000 Cuba 1817 199,000 Puerto Rico 1820 22,000 Spanish West Indies 1860-1 412,000 U.S.A. i860 3,954,ooo Puerto Rico 1867 47,000 Cuba 1871 288,000 Brazil 1872 1,511,000 Africa Haut-Senegal-Niger 1905-13 702,000 French Guinea I9°5-!3 490,000 (French Western Sudan) 1905-13 1,192,000 Sokoto Caliphate c. 1900 2,500,000

Source: Eltis, Economic Growth, 297 fn; H. Klein, African Slavery, 297; Franklin Knight, Slave Society in Cuba during the Nineteenth Century (Madison, Wisconsin, 1970), 22, 63; M. Klein, 'Demography of Slavery', 52, 54; Hogendorn and Lovejoy, Legal Status Abolition. population (1,418,000). For the French Sudan as a whole, there were approxi- mately 1,192,000 slaves in a total population of 5,134,000, but these estimates were made after the slave exodus that occurred during and immediately after the French conquest. That exodus reached a climax in 1905-06, by which time hundreds of thousands of slaves had fled. Before the exodus, the slave population was probably in the order of i-5-2 million. The extent of slavery in the Sokoto Caliphate was comparable. J. S. Hogendorn and I have calculated that the number of slaves in the Caliphate probably represented one-quarter of the total population of 10 million in 1900.115 Both the percentage of slaves and the scale of the population are intended as conservative estimates. While there is a slight overlap between Klein's figures and our own, these are not significant. Eight of the thirty emirates in the Caliphate came under French rule, but only two are included in Klein's sample, and they were both small emirates. Whether Hogendorn's and my figures are accepted or not, there can be little doubt that the Sokoto Caliphate may well have been the second or third largest slave society in modern history. Only the United States in i860 (and maybe Brazil as well) had more slaves than the Caliphate did in 1900. About a decade after the final emancipation of slaves in the Americas, there were at least twice as many slaves in Islamic West Africa as there had been in Brazil and Cuba in 1870 and at least as many as in the U.S.A. at the start of its Civil War. These comparisons are striking evidence that slavery in Africa 115 J. S. Hogendorn and Paul E. Lovejoy, Legal Slavery Abolition. The Decline of Slavery in Northern Nigeria (Cambridge, forthcoming). THE ATLANTIC SLAVE TRADE 393 has to be taken seriously by historians of both Africa and the Americas.116 As should be obvious, no attempt is made here to estimate the slave populations of other parts of Africa, particularly areas that fed the Atlantic slave trade, but it is known that the percentage of slaves in Asante, the Yoruba states, the Igbo country and elsewhere was high.

Conclusion The economic costs of the slave trade in African economies and societies were severe, despite Eltis' interpretation to the contrary. First, the low per capita income from the trade indicates that the economic advantages of exporting slaves were nowhere near large enough to offset the social and political costs of participation. Secondly, the size of the trade, including enslavement, related deaths, social dislocation and exports, was sufficient to have had a disastrous demographic impact. Thirdly, because the rise of produce exports started at such a low base and at a time when slave exports were becoming less important, western Africa suffered a relative decline in its position in world trade. There were other heavy costs which Eltis has failed to appreciate. It is possible to calculate the gross barter terms of trade and per capita income from the slave trade and compare western Africa with other parts of the world. But it is difficult to assess the full costs of ' producing' slaves because of the nature of enslavement. In an economic sense, as Robert Paul Thomas and Richard Bean have demonstrated, slaves were a 'free good', like fish, as far as those doing the enslaving were concerned.117 There were costs associated with 'production', but the real cost in human terms included the loss of life from enslavement, subsequent famines and disease. Furthermore, the destruction of property during wars and raids also represented a loss. Manning's simulation model has attempted to account for some of these costs, although it will never be possible to do the kind of analysis that is possible in measuring the volume and direction of the trans-Atlantic trade itself. Miller has come closest to demonstrating the effects of this impact on a particular region, but his analysis, too, is based on a considerable degree of conjecture. When the indirect costs of enslavement and trading are taken into con- sideration, the insights of Eltis and Jennings take on a new meaning. Rather than demonstrating that the Atlantic slave trade had virtually no impact on western Africa, it can be concluded that the impact was in fact strongly negative, although profound.

116 Unfortunately, all too few Americanist historians take this point seriously. Black- burn, for example, analyses the end of 'colonial slavery' only in the colonies of the Americas (Overthrow of Colonial Slavery). For an example of an historian who does make the attempt to consider slavery in a trans-Atlantic framework; see Klein, African Slavery. Both Eltis and Miller try to bridge the Atlantic gap as well by examining change throughout the Atlantic basin, including the interior of Africa. They reach remarkably different conclusions, nonetheless. In my opinion, Miller's approach is truly global and represents a dramatic break with past scholarship. He has succeeded in following slavery from deep within west-central Africa to Brazil and Portugal. He has set an example that will be difficult to follow. Eltis has made a similar attempt, but less successfully from an African perspective. 117 Robert Paul Thomas and Richard Bean,' The fishers of men: the profits of the slave trade', J. Econ. Hist., xxxiv, 4 (1974), 885-914. 394 PAUL E. LOVEJOY

SUMMARY Recent revisions of estimates for the volume of the trans-Atlantic slave trade suggest that approximately 11,863,000 slaves were exported from Africa during the whole period of the Atlantic slave trade, which is a small upward revision of my 1982 synthesis and still well within the range projected by Curtin in 1969. More accurate studies of the French and British sectors indicate that some revision in the temporal and regional distribution of slave exports is required, especially for the eighteenth century. First, the Bight of Biafra was more important and its involvement in the trade began several decades earlier than previously thought. Secondly, the French and British were more active on the Loango coast than earlier statistics revealed. The southward shift of the trade now appears to have been more gradual and to have begun earlier than I argued in 1982. The greater precision in the regional breakdown of slave shipments is confirmed by new data on the ethnic origins of slaves. The analysis also allows a new assessment of the gender and age profile of the exported population. There was a trend toward greater proportions of males and children. In the seventeenth century, slavers purchased relatively balanced proportions of males and females, and children were under-represented. By the eighteenth century, west-central Africa was exporting twice as many males as females, while West Africa was far from attaining such ratios. In the nineteenth century, by contrast, slavers could achieve those ratios almost anywhere slaves were available for export, and in parts of west-central and south-eastern Africa the percentage of males reached unprecedented levels of 70 per cent or more. Furthermore, increasing numbers of slaves were children, and again west-central Africa led the way in this shift while West Africa lagged behind considerably. This review of the literature on the demography of the slave trade provides a context to assess the revisionist interpretation of David Eltis, who has argued recently that the slave trade and its suppression were of minor importance in African history. It is shown that Eltis' economic arguments, based on an assessment of per capita income and the value of the export trade, are flawed. The demography of the trade involved an absolute loss of population and a large increase in the enslaved population that was retained in Africa. A rough comparison of slave populations in West Africa and the Americas indicates that the scale of slavery in Africa was extremely large.