Paper for seminars at York University and University of Western Ontario, October 2003
THE ECONOMICS OF SLAVERY AND THE TRANSITION TO FREE LABOUR IN PRECOLONIAL WEST AFRICA: AN ASANTE (GHANA) CASE-STUDY, 1807-C.19501 Gareth Austin
Department of Economic History, London School of Economics
This paper offers a West African test of the standard economic explanation for the use of slave labour, and goes on, in the same geographical context, to evaluate the most influential economic model of institutional change as a tool for explaining the eventual end of slavery.
We have long known that the nineteenth and early twentieth centuries saw the growth of agricultural export economies in West Africa, uneven as it was over the decades and between different parts of the region. The process began on the coast during the decline of the Atlantic slave trade (conventionally dated from British abolition, 1808), and was much extended, in output and in geographical scope, during the early decades of colonial rule.2 More recently, we have learned that the nineteenth-century, pre-colonial, expansion of commodity production was by no means based solely on overseas markets.3
Until the 1970s at least, the general assumption was that the labour involved in all this was supplied overwhelmingly by family members, increasingly supplemented in the twentieth century by migrant wage labourers. The latter moved seasonally between the zones of cash-crop cultivation and their own homes, which were in areas where soil quality and/or transport costs were such as to render export production on their own account unprofitable. Scholars recognised that ‘domestic’ slavery existed in various societies in pre-colonial West Africa, and that colonial governments had generally been slow to act on their international commitments to abolish it. Over the last thirty years, however, this picture has been altered by an accumulation of evidence that the nineteenth-century episodes of growth in goods output stimulated, and were reinforced by, the purchase of captives generated within the region (mostly by wars and raids).4 This took advantage of the fact that the closure of the Atlantic slave market led to lower prices for slaves within West Africa. The fact that the sharp fall (by about half) in slave prices on the coast between 1807 and 1820 was largely reversed from the 1820s reflected, at least in part, the higher level of internal demand for slaves to fuel the growing output of marketed goods.5
African institutions of slavery retained the distinctive feature – by contrast with slavery in Europe and the New World – that slave status tended to be alleviated for the descendants of slaves. Thus there was a tendency towards assimilation rather than to the perpetuation of a slave caste (the timing and completeness of this tendency varied greatly between slave-holding societies within Africa). This feature has been the foundation of an influential argument that slavery in Africa was a social mechanism for the incorporation of outsiders more than an instrument of labour exploitation.6 The nineteenth-century expansion of slave-holding, however, was strongly associated with the application of slave labour to production, for both sale and subsistence.7
This paper tackles two issues which arise in this setting. First, if there was a market demand for labour, why did it take the form of a market in labourers themselves rather than in their services? Second, what was the relationship between the expansion and subsequent demise of slavery and the growth of extra-subsistence production? Specifically, how far did slavery contribute to the ‘cash crop revolution’ in West Africa; and, conversely, did the growth of export agriculture itself abolish the conditions which had made slave labour profitable, thereby facilitating, even ‘inducing’ the transition to hired labour?
The sections below examine these matters for the case of Asante (or Ashanti), by which I mean the forest heartland of the pre-colonial kingdom of Asante (c.1701-1896). This corresponds to the Ashanti and Brong-Ahafo regions of the Republic of Ghana of today. The Asante kingdom was one of the major states in nineteenth-century Africa, dominating much of what is now Ghana by the imposition of tributary relationships. While most of the food-crop output did not pass through markets, the nineteenth-century economy had a substantial extra-subsistence dimension.8 Kola nuts were the main export to northern (savanna) markets, gold (or in the 1880s and 1890s, rubber) was the main export to the south and thereby to Europe. There was also lively internal trade, notably in handicrafts (such as cotton cloths) and in forest products (such as edible snails). There was a general-purpose currency in the form of gold dust, which was measured in an elaborate system of units, and whose use was compulsory for transactions within Asante. Following the British occupation in 1896, Asante producers rapidly – and largely at their own initiative – adopted an exotic crop, cocoa.9 In 1911 ‘Ghana’ (to use the modern name) became the world’s largest producer of cocoa beans. Though most of the output came initially from the Gold Coast Colony, to the south of Asante, the latter rapidly became a major supplier, ultimately surpassing the Colony’s contribution to the Ghanaian total.
The main discussion has three parts. Section 1 provides a brief description of slavery in Asante, and of its decline. Sections 2 and 3 analyse the economics and political economy of, respectively, the ‘choice’ of slave labour in nineteenth-century Asante, and the colonial-era transition to hired labour.
1. Asante slavery and its decline10 In Asante, as elsewhere in West Africa, there is no suggestion that the institutions of slavery were new in the nineteenth century. As elsewhere in the region and in the world, first-generation slaves were mostly foreigners. Some were Akan-speakers, like the Asantes themselves, but mostly captured in war from the southern Akan states. Some were Ewes, from what is now the Volta Region of Ghana; again generally captured by Asante forces. The great majority, however, seem to have been what the Asantes called nnonkofoo, people from the northern savanna. Though some were obtained directly through capture, and others were supplied in tribute from subordinate polities, in the nineteenth century northern slaves were obtained mostly by purchase, in the market of Salaga (or later, Kintampo) to the immediate north of Asante. The geographer Boaten commented in 1970 that ‘The many domestic slaves (sic) found in practically all the Asante homes attest to the extent and popularity of the slave trade with the north’.11 As the quotation implies, slaves were commonly incorporated within commoner households. Alternatively, they were added to chiefs’ establishments, often living either at the chief’s palace or in villages populated specifically by chief’s servants. Whether belonging to a chief or to a commoner, slaves – male and female – were put to work, contributing to the genesis of virtually all the products of the Asante economy. Not all slaves were closely supervised: some seem to have enjoyed a degree of autonomy, farming on sharecrop terms. Female – very rarely male – slaves were usually married by free Asantes. Because Asante society was matrilineal, strictly speaking their descendants remained outside the Asante kinship system. In practice, though, they were included, as members of cadet lines. The children of slaves – and in principle their own descendants – had an obligation to work for senior members of the household (or for the chief, as the case may be); but it seems agreed that more of their labour time was likely to be their own than in the case of first-generation slaves.
Two observations are particularly important for the discussion in Section 2. One is that the matrilineal character of Asante kinship – relatively unusual in West Africa – created a specific non-economic incentive for the acquisition of slaves which did not apply in patrilineal societies. A man’s children by a free wife belonged to her lineage, not to his: thus his own heir was not his own child but, typically, his sister’s child. By marrying a slave woman, however, he could obtain children who belonged to him, and could inherit from him.
The other observation is that there is no unambiguous evidence that hired labour existed in pre-colonial Asante. Thus individuals wishing to recruit workers beyond those available from their households (which of course probably already included first-generation slaves and/or their descendants) obtained coerced labour in some form. Besides slavery, there were two other categories of such labour: pawning12 and corve. Both were important, but there is no dispute that slaves were the prime source of non-free labour; and the slave trade was the main means for acquiring them.
Britain occupied Asante in January 1896, and promptly declared the slave trade abolished, making serious attempts at enforcement which seem to have largely successful by the end of the century. In contrast, but as in many other European colonies in West Africa, slave-holding remained legal for an extended period. In this case, it was in June1908 that slavery (together with debt bondage) was prohibited. Some prosecutions followed, but the decline of slavery on the ‘social ground’ had begun already (with some slaves running away) and, on the other hand, was protracted in the sense that chiefs, especially, continued to expect ‘slaves’ or their descendants to continue to perform at least token services for them. Again, as will be seen below, female slavery seems to have outlived male slavery. For men at least, slave labour seems to have largely disappeared by c.1918. Some slaves (male and female) left their masters’ villages and headed home. This option was open mainly to the minority of Akan slaves, who did not have all that far to travel. Most of the northerners, and especially their descendants, seem to have stayed but refused to labour for their former masters.
By no coincidence, as will be argued later, regular wage contacts – for a season or year – began to appear in Asante agriculture, specifically in cocoa production, from about 1918, i.e. from when – or shortly after - slave labour ceased to be widely available. The hiring of migrant labourers, recruited mainly from the northern savanna, had replaced the purchase of northern people. The growth of the free labour market continued throughout the colonial period, though slowed by the Depression and the Second World War. In 1956-7 the first relatively comprehensive survey of Asante cocoa farming found a ratio of hired labourers to farmers of 1.9: 1. By then – since the 1930s – hired labour contracts increasingly took the form of sharecrop rather than wage terms. But that is a different, though related, story.
2. The Nieboer hypothesis and slavery in late pre-colonial Asante, 1807-96: Hobson’s choice The classic economic interpretation of slavery, which we may call for generic convenience the ‘Nieboer hypothesis’, explains the existence of widespread slavery as a profit-maximizing response to a scarcity of labour in relation to land, in economies in which capital, however desirable, was not essential for subsistence (as it might be in pastoral societies).13 The most influential restatements of this tradition, both succinct, were published over a quarter of a century ago. Domar’s 1970 paper14 was widely discussed by students of serfdom and slavery in Europe and the Americas. For the West African context the Nieboer hypothesis received its classic formulation in five pages by Hopkins in 1973.15 Hopkins argued that land was abundant in relation to labour, and that in response, though ‘there was a factor market in land . . . it was very limited’ because ‘land was not scarce enough to acquire a market value’.16 Because of the easy access to land, potential labourers were physically and occupationally mobile and so enjoyed a strong bargaining position in relation to would-be employers. Hence the latter chose slavery in preference to wage labour, which, Hopkins insisted, was a genuine alternative: ‘the use of slave rather than wage labour was a matter of deliberate choice on the part of African employers’.17
Since the early 1970s there has been a vast amount of research on the history of slavery within Africa;18 yet systematic tests of the Nieboer hypothesis in relation to particular cases of slavery in the continent remain very rare.19 So far as I am aware, this study is the first to attempt a quantification - admittedly crude - of the costs of slave compared to free labour in an African case. Maybe the critics have been satisfied with the various a priori grounds which have been offered for dismissing the hypothesis, contentions which will be discussed below. Perhaps more sympathetic scholars have considered the issue sufficiently explored. A particular barrier (but should it not be a stimulus?) to further exploration is the model’s notorious indeterminancy. Nieboer, Domar and Hopkins all agreed, in effect, that an economy in which expansion of output is constrained only by the cost of labour is not sufficient for the existence of slavery ‘as an industrial system’.20
The nineteenth century in Asante, as in West Africa generally, is a particularly apposite period to consider in relation to the hypothesis, for Nieboer further posited that commerce was a supplementary stimulus to the growth of slavery: because extra-subsistence production increased the demand for labour, notably beyond the level where the ‘drudgery’ could be performed solely by [‘free’] women.21 To underline a theme introduced at the start of the paper, in Asante, as in West Africa generally, the era from the beginning of the end of the Atlantic slave trade to the colonial partition was marked by uneven but major growth in extra-subsistence production.22 Asante shared in the expansion of both the intra-regional trade, and the Atlantic commodity trades. That the period also saw a great increase in slave-ownership within West Africa, and that slaves were widely put to work producing for the market, as the literature has increasingly established, is as the model would predict.
The argument developed in this section is that the hypothesis that slavery was a profit-maximizing response to land abundance and labour scarcity is justified in this case; but that the sense in which it was a ‘choice’ needs to be clarified, for in economic terms it was Hobson’s choice; while the concept of economic rent based on monopoly of natural resources and of locational advantage may provide one of the tools needed in the general search for a more determinate account of the political economy of slavery in African and even world history. The discussion is organized as follows. We need first to ask whether the prevailing factor ratio and production function fit the Nieboer prediction. This question is addressed in Section A, which goes on to highlight an important political economy aspect of resource allocation between Asante and the areas from which it imported captives. Section B considers, in relation to this case, the two standard objections to the hypothesis in the West Africanist literature. Section C compares the cost of free and slave labour, and explores the implications of this for the economic significance of coercion in post-1807 Asante.
2A. Natural-resource abundance and nineteenth-century Asante This short section is intended to summarise some points documented in my forthcoming book. I would be happy to enlarge on them in discussion. My argument there is that the land abundance/labour scarcity model is overly aggregative as a description of nineteenth-century Asante. First, it overlooks the fact that the supply of capital, in its various forms, was an important determinant of short-term output. Second, the homogeneity of land as a factor of production disappears when the supply of this resource is considered not as a stock but as a flow. But I also argue that, while both these points matter greatly in the context of choice of technique, they are not crucial to the issue of labour scarcity over comparatively long timespans: those relevant to a decision on whether to acquire a slave. Over periods longer than a few months, in agriculture, or perhaps a couple of years in craft and mining production, output was a function very largely of inputs of labour. It was the level of labour inputs that determined the area of land cleared for food crops, the provision of capital goods in weaving and mining, and whether enough seed was planted for the following year. In this perspective even for activities undertaken during the agricultural off-season, when the opportunity cost of labour was low, it was still the level of labour inputs that principally decided the volume of (in this case non-agricultural) output. The crucial point here is that the forest heartland of Asante was very lightly populated relative to the carrying capacity of the land. We may conclude that, certainly in general and probably virtually everywhere, land for cultivation was abundant - unless the supply was constrained artificially, by political intervention - in the strict sense that its marginal product must have been zero. This in turn implies that the marginal product of labour must have been very close to its average product, as in Domar’s formulation.23
In the absence of economies of scale in production potential employers were constrained by a lack of means by which they could profitably afford to pay workers more than the latter could make by working for themselves.24 But as free subjects of the Asante kingdom, they shared a collective natural-resource advantage over one set of potential workers: those from the savanna societies to the north. Their enjoyment of the natural-resource rent25 was secured by the might of the Asante state, and reinforced by the exploitation of locational advantage: possession of the strategic position in the transit trade, through the forest, between the coast and the savanna, which the government used not only to give Asante traders a national monopoly but also to obstruct savanna polities from acquiring European weaponry. This did not mean that Asante did most of its own slave-catching: as we have seen, it could buy (or demand as tribute) slaves from the more powerful savanna states, which they in turn had obtained often from acephalous societies. The economic rent which Asante obtained by institutionalizing its natural-resource advantage was particularly significant for slavery: as will become apparent once we have considered the options available to Asante’s rulers.
2B. Objections to the Nieboer-Hopkins approach: politics and culture left out? Hopkins’s analysis has provoked criticism over the years from marxist, substantivist and other directions: notably on the grounds that, in effect, it is either too determinist or not determinist enough. The former objection is that the hypothesis takes relative scarcities as given, and thus overlooks the possibility that the more powerful historical actors had a choice between slavery and free labour: specifically, it fails to allow the possibility that private appropriation of land could turn natural abundance into artificial scarcity. As Hymer had already remarked of the notion of land surplus: ‘land, even when it is plentiful, can be made scarce if certain groups gain the power of excluding others from its use’.26 It has also been observed of the Nieboer hypothesis generally that a ruling class in a labour-scarce, land-abundant economy might have further options for achieving income redistribution in its favor besides slavery: a landowners’ or employers’ cartel, or regressive taxation imposed on the free peasantry.27
The other objection is that the actors had no choice; for cultural reasons. Goody, for example, maintained - contrary to Hopkins - that the question of a ‘choice between wage labour and slavery, as practical alternatives’ did not arise. In Goody’s view Hopkins’s ‘argument from economic principles appears to neglect the location of slavery in a social system’.28 Thornton has put forward a related view: that the reason why slaves were the characteristic ‘form of private, revenue-producing property’ in Atlantic Africa, whereas land fulfilled the same function in Europe, lay in ‘divergence’ between the two regions’ legal systems.29 Thornton’s implicit assumption is that cultural difference is a sufficient explanation: that an alternative property system, such as one in which labour-time rather than persons were sold, was simply not conceivable whatever its economic implications.
Let us consider both sets of objections in relation to nineteenth-century Asante. A cartel of labour-users would seem impracticable because of the very large numbers of such potential principals dispersed over a wide area in which communications were restricted by the high forest and the disease-dictated reliance on head loading.30 It is slightly easier to conceive of access to land being restricted: of the physical abundance of land being turned into institutional scarcity, turning most of the population into landless labourers. Indeed the Asante courts manipulated the ownership of rights of sovereignty and political control over land: by imposing fines on individual chiefs obliging some of them to mortgage these rights.31 Engerman has suggested, however, that ‘often the success of political interventions will itself depend upon the land-labor ratio, influencing the ability to control the population’.32 In nineteenth-century Asante, to create an artificial scarcity of land as a factor of production would have required far more detailed administration and enforcement than the state could afford. Moreover a relatively generous exercise of the stools’ ultimate control over land was fundamental to the chiefs’ enjoyment of support or at least acquiescence from their subjects. More prosaically, we noted earlier that the high-forest environment was a source of heavy transactions costs when it came to territorial demarcation. It is therefore no surprise that, while the ownership of the land itself was vested in chiefdoms, cultivation rights - access to land as a factor of production - were, institutionally as well as physically, virtually free in most of the country at least. Strangers, and sometimes subjects, had to hand over a share of gold or other winnings from the non-renewable endowment of the lands: but such resources were themselves limited.33
Arguably, the one really practicable alternative to slavery would have been the tribute option: making a free peasantry pay tax. But such a tax or internal tributary system without slavery would have been economically sub-optimal for both the ruling class and the peasantry. The Asante state did indeed get much of its revenue from the mass of free households, for example by taking a share of gold output.34 However, as noted earlier, free commoner households and individuals themselves became major importers of slaves. Consequently the size of those households’ extra-subsistence output, and therefore their fiscal potential, was to a large extent a function of their resources of slaves (and pawns).35
In broad rational-choice terms it makes sense to suggest that the most efficient system of exploitation from the viewpoint of the rulers of Asante was based on the importation of slaves. This is because the gains could be shared - unequally - across the ‘host’ society. Government and free subjects would have a joint interest in maintaining in servitude a group who would remain a minority of the population. In principle this was much less difficult in resource terms than, in particular, it would be for the governing elite to try to dispossess the mass of the population. Ultimately, the chiefs had nothing to gain from seeking to overturn the natural factor ratios that they could not achieve more easily by using their power of coercion, not to make the abundant factor scarce, but to make the scarce factor - selectively - cheaper. In terms of self-interest it benefitted chiefs and free subjects to join together in the exploitation of outsiders, paid for to a great extent by the economic rent derived from monopoly control of particularly valuable natural resources.
It seems likely, though at this stage of knowledge it can only be a very plausible assumption, that Asante’s contemporary reputation as an intensive importer and user of slaves was justified. ‘Slavery prevails here [in Asante], as it does in most parts of Western Africa, but on a larger scale.’36 If indeed the scale of slaveholding, and the importance of slaves’ contribution to output, were relatively high in Asante compared to most other polities in the region, it is reasonable to suggest that this was a function of Asante’s advantage in economic rent.
Hopkins’s insistence on the notion of choice was, at least in part, intended to highlight his rejection of any assumption that wage-labour was culturally unthinkable in precolonial West Africa. In the Asante case it is very plausible that the idea was not much ‘thought’ before the twentieth century, yet it would be wrong to infer from this absence the presence of a solid cultural template preventing any practice that did not fit into familiar slots. We have an 1881 instance of expatriate Asante princes readily becoming employers of wage labour, and in Asante youths showed themselves similarly willing to accept casual wage employment when it became an economic option early in the colonial period.37 Thus Hopkins’s emphasis on cultural flexibility is vindicated in this case. But, as we will see in the next section, there was a decisive barrier to wage labour in nineteenth-century Asante. This barrier was not cultural but material.
Applying the Nieboer hypothesis entails a further pair of assumptions about Asante politics and culture respectively: that the distribution of coercive power was sufficiently unequal to make slavery physically possible, and that this was ideologically acceptable to those holding the coercive power. It is evident that this was the case: slaves were held, slavery was a legally-enforceable status, and there is no evidence of pressure for abolition from free members of the society. As in most other slave-holding societies in world history, as a form of exploitation slavery was directed mostly outwards, in that (as we have seen) most of the slaves were foreigners, especially nnonkofoo (singular, odonko) from the northern savanna. Furthermore, again as in other slave-holding communities, the specific category of foreigners from whom most slaves were obtained tended to be seen as alien and inferior. McCaskie comments that is ‘abundantly clear from reported speech and traditions, [that] the term [odonko] suggested a brute animality to the Asante, and connoted a non-human status.’38 This needs to be balanced by the recollection that Asante men were happy to marry northern women and to count any offspring as their own. It is necessary to avoid the anachronistic assumption that the act of enslavement was in all eras felt by its perpetrators as an enormity. But it is reasonable to suppose that in nineteenth-century Asante society it was culturally, and therefore psychically, harder to envisage and treat fellow Akan, especially Asante, as slaves than it was with northerners, ‘othered’ as nnonkofoo. Meanwhile, this time in common with most other African slave-holding societies but in sharp contrast to New World slavery, a process of partial assimilation of the descendants of slaves reduced the risk of creating a hereditary slave caste.39 These characteristics can be said to have reduced the psychic, political and military costs and risks in holding slaves.