The business of barter on the pre-colonial Gold Coast

ABSTRACT Trade on the Gold Coast in the eighteenth century was dominated by non-monetized barter trade. In this paper, a large dataset of barter transactions are used to study the social embeddedness of the trade. The data shows that prestige goods such as alcohol to a disproportionate degree were exchanged for other prestige goods such as gold. Guns – but also cheaper types of textiles – were to a disproportionate degree exchanged for slaves in particular. The evidence thus helps to shed light on the social valuation of various imported commodities on the Gold Coast at this time.


Introduction
Karl Polanyi famously argued that we must take the embeddedness of economic relationships into consideration, not least when studying pre-modern societies (see, for example, Polanyi 1957, 250). Following in this tradition, it has been argued that socalled separate spheres of exchange were common in pre-colonial West Africa. The basic idea behind this concept is that it, for some reason, was considered impossible or improper to exchange certain goods for just any other type of goods. Any exchange therefore could occur only, or at least primarily, within the same sphere of commensurable commodities (Bohannan 1955(Bohannan , 1959. Economic historians have certainly criticized this characterization of pre-colonial West African trade (see, for example, Dorward 1976). Nonetheless, the theory has been recurring in the literature on economic anthropology ever since, and it has been argued that it is applicable both to other societies in Africa and to societies elsewhere in the world (see, for example, Dalton 1961;Dalton 1977;Barth 1981;Gregory 1982;Kopytoff 1986;Piot 1991;Gregory 1996;Gregory 1997;Sillitoe 2006). There has also been a discussion as to why there might exist separate spheres of exchange (see, for example, Kopytoff 1986;Espeland & Stevens 1998;Maurer 2006;Sillitoe 2006).
In monetized trade, the very fungibility of money allows it to be used as a general medium of exchange. The seller of a commodity assumes that it will be possible to purchase anything (i.e. anything exchangeable) using the money gained from selling a good (ignoring for the moment any transaction costs). In barter trade, on the other hand, the personal relations of the barter relationships become crucial. Seller A has to make sure not only that he or she can find a buyer B, and ascertain an acceptable price for the commodity sold, but alsosince there exists no general medium of exchangethat B in turn has goods to sell that A is interested in buying in return. One example of this phenomenon can be found in the historical and economic-historical scholarship on pre-colonial African external trade. An important hypothesis that has been put forth in this literature is that there existed a special guns-for-slaves cycle in the external trade. The most basic evidence for this cycle is quite simple: slaves were for many years the main export from Western Africa, and guns were one important import. It is by now uncontroversial that slaves were the main exports from Western Africa at least during the eighteenth century, and likewise that guns were just one among many goods imported (see, for example, Gemery & Hogendorn 1990;van den Boogaart 1992;Eltis 1994). But the hypothesis of a guns-for-slaves cycle has been argued by some scholars to exhibit a systematic relationship, whereby guns more or less were always required in exchange for slaves (for some classic contributions see Kea 1971;White 1971;Davidson 1971:69-70;Inikori 1977:350-1;Richards 1980:47; see also a critique in Curtin 1975:325). David Northrup has more recently claimed that even though there might exist a superficial correlation between firearms imported to, and slaves exported from, some parts of Western Africa, this is not sufficient evidence to say that there was an actual, systematic relationship between the two. Northrup has instead suggested that the majority of the arms 'must have gone into the hands of common people, who used them for hunting, for selfdefense, and for firing at funerals and other ceremonial occasions, rather than for war' (Northrup 2009, 99-105, quote on page 104). It is unfortunately rather unclear what evidence Northrup was basing this particular claim on. Ann Ruderman has, even more recently, argued that there are no echoes of any guns-for-slaves thesis in the correspondence between Europe and Africa that she consulted (Ruderman 2016, 37). The discussion has recently been revived by Warren Whatley, who has employed advanced statistical methods on aggregate trade data, to show that the relationship between the two variables is not simply spurious, but that there actually existed a cycle whereby imports of guns into West Africa increased the exports of slaves from the region, and vice versa (Whatley 2018; see also Whatley & Gillezeau 2011).
Whatley's valuable contribution focused upon the particular pattern of guns exchanged for slaves, and ignored the trade in other goods and how that trade might have been embedded in various social institutions. It has therefore not been well studied if a trade pattern whereby guns to a large extent were exchanged primarily for slaves was unique to these particular goods at this time, or if similar patternswhere a particular type of good was exchanged only or primarily for another particular type of goodcould be found for other goods as well. There have been some suggestions in the previous literature that this might have been the case. As mentioned earlier, several anthropologists following in the footsteps of Polanyi have thus suggested that there indeed existed certain spheres of exchange in various parts of Africa (and elsewhere in the world), but have then generally not undertaken an historical analysis. Eugenia Herbert has, on the other hand, suggested that more than half of the gold exported from West Africa in the early years of contact with European traders was bartered for copper in particular (Herbert 1984, 126-7), hinting at the existence of a special copper-for-gold trade at that time.

Research design
This article contributes to these fields of research by assembling quantitative micro-level data on the business of barter in all goods traded in pre-colonial West Africa, focusing on the trade on the Gold Coast during the early eighteenth century in particular. The region and period are chosen due to the very rich sources available from this particular region at this time. It is most plausible to assume that trade was embedded in different social institutions in different locations, and that it might have changed over time. We can for that reason not assume that the Gold Coast was representative of West Africa in general, nor that the specific time period under study would be representative of the Gold Coast's long pre-colonial history.
The research question of the present paper is: were there any disproportionately common barter exchanges on the eighteenth-century Gold Coast, and, if so, how can these be explained? The terminology of 'disproportionately common barter exchanges' is here used to describe specific barter exchanges (commodity A exchanged for commodity B) that were more/less common than what would be expected from the aggregate trade data. Table 1 shows an attempt to illustrate this hypothetically. Panel A shows what here will be called proportionate barter exchanges. Assume for simplicity that the figures represent individual barter exchanges for standard units of the commodities in question estimated by the trading partners to be of equal value. In the example in panel A, slaves are more popular than gold in total, and textiles are in aggregate likewise more popular than both arms and alcohol. The individual barter exchanges involved are, however, proportionate to these aggregate figures. If the historical data looked like this, there would be no suggestion that the individuals that barter away slaves were doing so primarily in exchange for, say, arms (and vice versa), but that they were as likely to barter them in exchange for alcohol or textiles as an individual bartering away gold were. Panel B, in contrast, illustrates what here will be called disproportionate barter exchanges. In this example, the aggregate amount of goods bartered is the same, but the individual barter exchanges are not. Individuals that barter away gold in this example only do so in exchange for two of the other goodseither alcohol or textiles and furthermore disproportionately much in exchange for alcohol in particular. Individuals that barter away slaves in this hypothetical example do so primarily in exchange for guns, and disproportionately little in exchange for alcohol. Vice versa: all guns are in this panel's example exchanged for slaves, and none for gold.
The example in Table 1 is thus an illustration of what is here meant by disproportionately common barter exchanges. This terminology is used here since the concept of 'spheres of exchange' is so intimately connected to a theoretical framework attempting to explain these same 'spheres', as described earlier. The concept of 'cycles' of trade, for its part, would on the other hand imply that there is a feedback mechanism involved, so that trade in good A increases trade in good B (such as the guns-for-slaves cycle whereby imports of guns enable greater enslavement, and thereby a greater number of slaves exported). If disproportionately common barter exchanges can be found, this could certainly be due to a specific 'sphere of exchange' or a specific 'cycle' of trade. Importantly, however, the mere existence of such patterns of the barter trade does not per se support either of these two previous theories, and it says nothing about the intentions of the agents involved. Such patterns can, however, suggest that certain barter exchanges might have been embedded in special social relations. Shedding light on any such patterns might therefore increase our understanding of the economic and social history of the region, for example regarding the social classes to which the people trading particular goods with the British Royal African Company (RAC) might have belonged. The paper will study the barter transactions at the micro-level between a European Company, the RAC, and African counterparts on the Gold Coast during the first half of the eighteenth century. It thereby extends previous research by Klas Rönnbäck (2014Rönnbäck ( , 2016 into the trade in this region by analysing this trade in greater detail. In the accounts of the RAC from this period of time, it is possible to find empirical data on individual barter transactions, or transactions summarized over shorter periods of time (e.g. monthly or bimonthly). When data for a number of different transactions have been summarized in the source, the summary is done by commodity bartered for. It thereby becomes possible to study exactly what goods the individual African exports were bartered for. Some illustrations of what the source reveals might be in order. On 6 February 1725, for example, the accounts of the RAC record a transaction where nine chests of corn were 'purchased' (the term used in the accounts) with three large perpetuanas (a durable woollen textile) and three sheets (of unspecified quality). The perpetuanas were valued by the Company at five ackies each, thus amounting to 15 ackies in total (an ackie was a weight of gold, equivalent to 1/16 of an ounce, or approximately 1.8 grams). The sheets were valued at one ackie each, adding up to three ackies. The total value of the barter transaction is therefore recorded as one ounce and two ackies of gold (the equivalent of 18 ackies). The same day, four male slaves were 'purchased in barter' at the value of 2 marks, 5 ounces, 9 ackies and 6 takas of gold for all four of them. The company paid for the slaves using a sorting of trade goods including a number of different commodities: 16 guns, 27 perpetuanas, 15 kegs of tallow, three anchors of brandy, seven gallons of rum, 40 dozen knives, one dozen padlocks and one carpet. The purported value of each individual commodity bartered for is recorded in the accounts (BNA T/70/389, p. 100). The 'assortments' or 'sortings' of goods (Polanyi 1964) involved in the barters thus varied considerably between every transaction. In general, however, a barter transaction as recorded in the accounts of the RAC included a single African commodity (of varying quantity) being bartered for a number of different European commodities. Only the very rare entry records the sale of two African commodities at the same time, and in all of these cases it is goods from the same category of commodities (e.g. various African provisions, such as corn, palm oil or yams) summarized together in one entry in the accounts. By assembling data on a large number of such transactions, it thereby becomes possible to study if there is any pattern to which goods were bartered for which. The sources only exceptionally reveal anything about the company's African trading partners, for example, who they were. Most often the source remains silent about exactly who the Europeans were bartering with. It is therefore necessary to remain cautious when interpreting the data.
The barter transactions were of very different magnitude, ranging from the barter of very small amounts of, for example, provisions, to very large numbers of slaves or gold reported under the same heading in the RAC accounts. There was, furthermore, a great variety of commodities involved: in total there were around 25 different commodities that the European company acquired from African agents. As previous research has shown, slaves and gold were the two main exports from the Gold Coast. Other goods sold on the coast include provisions (for the people living in the RAC forts, as well as to provision the slave ships) and ivory, as well as a number of other goods (e.g. beeswax, redwood trees, gum or canoes). In exchange, the RAC bartered away more than 180 different commodities. Several of these commodities were, furthermore, accounted for in different units, which are not always easy to convert into comparable units (e.g. textiles traded both by piece and by yard). In this paper, the European commodities sold to the African buyers have been classified into a smaller number of categories of commodities, broadly following the categorization of Alpern (1995): textiles of various sorts (here including clothing); metal and metalware; firearms (including gunpowder); beads, coral and cowries; alcohol; tobacco; European provisions; and other goods. As is known from previous research, some of these goods were not of European origin, but re-exports. For simplicity, I will use the term 'European goods' here as shorthand for the goods that the European company sold, regardless of whether the goods really were produced in Europe originally.
The multitude of types of commodities, the different magnitudes of the barter transactions, and the different units of measure makes it hard to aggregate the barter transactions in terms of the quantities involved. In order to allow for aggregating the trade data from the various barter transactions, this paper will therefore make use of the values reported for the goods involved reported in the RAC's accounts. The values seem to refer to what the good in question was estimated to be worth in monetary terms on the African coast (the occasional entry also reports the London price for the same good). Local goods are generally accounted for in local gold weightsmarks, ounces, ackeys, takus and dambas of gold (Rönnbäck 2016, 68). In many cases, the data is, however, reported in pounds sterling. An official exchange rate (£4 per ounce of gold) is then seemingly used as conversion factor between the currencies throughout the period under study. Exactly how the prices for individual commodities traded were established is unfortunately not revealed in the source. As gold apparently was a unit of account, at least for the RAC (but potentially also for the parties that the RAC was exchanging goods with), it might be discussed whether these transactions indeed should be considered to be true barter transactions, or if gold were a form of implicit money underlying the transactions. As a rule, no money seems to have been exchanged during the transactions (possibly except, that is, in the case of when gold itself was transacted). For that reason, it seems proper to classify the trade as effectively being based on a form of barter.
Previous research has shown that African prices for foodstuffs seem to have fluctuated in response to, for example, climate and demand conditions (Rönnbäck 2014). Certain price differences can also be seen in the prices for European goods assembled for this study. To take just two examples, the reported prices for broadcloth were mostly in the range of 15-25 shillings per yard, whereas knives were sold for the equivalent of 4-7 shillings per dozen. The data would thus seemingly indicate that the prices reported in the sources would reflect some form of market valuation of the goods. Similar to Rönnbäck's findings (2014), the prices in the source do not suggest that there was any inflation or deflation in the region at the time, but prices reported shifted in very different ways for the different commodities involved.
All barter transactions recorded in the accounts of the RAC (the so-called T70-series at the British National Archives), for five benchmark years (1725, 1730, 1735, 1740 and 1744/ 5), have been assembled in a database by the author of this paper. In total, the dataset is made up of 1,268 barter transactions of African goods. Other transactions (gifts and dashes, wage or rent payments and so on) have not been included in this database. There seem to be no large individual transactions that skew the results. For simplicity, I will henceforth make use of the terminology of the sale and purchase of goods (terms which also were used in the accounts of the RAC), in order to make the direction of the flow of commodities clearer, but it must be kept in mind that these were barter transactions that involved no monetary exchange (unless we count the European purchasing of gold as a transaction where the Europeans sell goods for money).
It is finally important to note one further important limitation of the research design: the data used comes from a European company. In the literature on spheres of exchange, it has, however, been argued that the supposedly traditional system of separate spheres of exchange was undermined by and eventually disappeared as a consequence of the contact with a capitalist, monetary economy (Bohannan 1959;Sillitoe 2006;see, however, Parry andBloch 1989 or Piot 1991 for a different point of view). The Gold Coast had been in contact with European traders since the sixteenth century. If Bohannan and others are correct in that contact with a capitalist economy would undermine the existence of separate spheres of exchange, any such patterns would perhaps already have disappeared by the time this study is concerned with. The source used is thus from the start biased against finding any pattern of separate spheres of exchange, at least. It is therefore crucial, as John Lie has put it, that the 'embeddedness approach must itself be embedded in larger, historically transient, social structures' (Lie 1997, 351).

Aggregate exports and imports to the Gold Coast by the RAC, 1725-45
A first issue is to determine the aggregate patterns of the trade by the RAC: what were the main commodities bought and sold on the Gold Coast in the years under study in this paper? In total, the trade during the benchmark years was reported to have a value of approximately £105,000 sterling for the five benchmark years, or around £21,000 sterling per year on average. The distribution of this trade between different types of goods is shown in Figures 1-2. Figure 1 shows data on the goods sold by the RAC trading on the West African coast in the 1720s to 1740s. In terms of value, textiles were clearly the most important category of commodities in the trade, accounting for half of the total value of the trade in this period. Alcohol and arms (most importantly, guns and gunpowder) were also very important import goods, accounting for approximately 10% of the total value of the trade each. Metals and metalwareincluding various products such as jugs, pots and pans of brass  or pewteraccounted for trade of approximately the same magnitude. Other goods, including cowries and beads, European provisions, tobacco and other items all accounted for smaller shares of the total trade. Overall, the data fit well with the aggregate estimates from previous research on the constitution of the Western African imports. Compared, for example, to the two benchmarks estimated by Eltis and Jennings (1988 , Table 2) for the 1680s and 1780s, respectively, the data presented here for the 1720-1740s in virtually every case fall almost perfectly in between the values estimated for the respective benchmarks. Figure 2 shows the value of the goods purchased by the RAC during the period under study. As can be seen, slaves constituted the clear majority of the purchases (51%) in terms of value, followed by gold (36%). Other goods bought by the RAC include ivory (5%) and African provisions (6%). The data here indicates that the composition of RAC purchases during these benchmark years was considerably less biased towards slave purchases than what some previous researchers have claimed was the case for West African exports in general (see, for example, Gemery and Hogendorn 1990: Appendix 1). This seeming discrepancy might be a reflection of a bias because of differences in the regional composition of the external trade. As the name implied, gold was certainly an important commodity from the Gold Coast in particular, to a much larger extent than from other parts of the West African coast. It might also reflect the trade patterns of the RAC in particular, after they had lost the monopoly they previously had been granted on the British trade in slaves (Davies 1975: ch. 3.III). Further studies, disaggregating the data on West African exports geographically, might reveal if this is solely due to geographical differences, or if the importance of goods other than slavesnot the least the exports of gold from the Gold Coasthave been underestimated in previous research. Provisions played only a minor role in the trade, even thoughas recent research has shownthe RAC seem to have purchased comparatively much provisions on the African coast (relative to ships from other European nations, which purchased much more of their provisions in Europe) (Dalrymple-Smith & Frankema 2017, Table 2).
Disaggregating the trade chronologically and geographically between different ports along the Gold Coast gives a somewhat more detailed picture ( Figure 3). As can be seen in Figure 3, the values of the transaction shifted quite substantially between the five benchmark years. In 1725, the total amount of goods bartered was accounted for as having a value of almost £50,000 sterling. The three following benchmark years exhibit trade at a value of around £15,000-20,000 sterling per year. In 1744-5, in contrast, the value of the trade was only £2,500. The decline is the strongest for the RAC's trade in slaves, but there was a marked decline in the trade in all other goods as well. Figure 4 shows the trade from a couple of the most important ports along the Gold Coast, as well as from a number of other locations where the RAC was trading along the coast. The single most important port was at the Cape Coast, followed by Accra. Four other ports -Commenda, Sekoni, Tantumkweri and Winnebahwere also of quite considerable importance for the RAC during this period. Overall, there were no enormous differences in the types of goods bartered for in these different ports, but it seems to have been possible to acquire most of the major traded commodities in all of these ports. The main exception was ivory: during these benchmark years, no ivory at all was purchased in some of the ports on the Gold Coast, such as Tantumkweri or Winnebah. A  disproportionately large share of all ivory purchased was, in contrast, purchased outside of the actual Gold Coast, either to the 'windward' or to the 'leeward' of the Gold Coast, i.e. potentiallysince the primary source reveals no more detailed informationin current-day Côte d'Ivoire and Benin/Nigeria, respectively.

The business of barter
We now turn to the key issue of this paper, which is to study if there existed any disproportionately common barter exchanges. Figure 5 shows the business of barter, reporting the transactions of goods according to the type of African commodity purchased. Each bar in the graph thus represent a single category of European commodities sold by the European company, and the division of each bar shows what sorts of commodities the Europeans acquired in exchange for these commodities in the barter process (as a share of the value of all transactions of that category of European commodity). If the sortings of European goods exchanged for African goods looked similar for all types of goods traded, the bars would also look similar. As can be seen in Figure 5, this is apparently not the case. Alcohol was thus, for example, seemingly often used when acquiring gold, whereas textiles or arms often used when bartering for slaves.
The data underlying the figure is summarized in Table 2. Table 2 furthermore includes data on the expected distribution (i.e. what the values would be if there were no relationship between the variables, but that all trade assortments of European goods looked similar regardless of what African goods were purchased). The expected value is calculated for each cell in the table by multiplying the row total with the column total, and dividing it by the sum total of the whole table. The expected value of the cell in the top left corner (alcohol bartered for African provisions) is thus the row total of the Alcohol row (9.9) multiplied by the column total for the African provisions column (6.3), divided by the sum total of the whole table (100.0), i.e. 0.6.
The absolute figures (measured in shillings) are reported in a table in the appendix to the paper (see appendix Table A1). A simple Chi-square test of independence for the whole of the distribution shows a statistically significant result, but that might not be surprising given the sample size. More importantly, therefore, several of the results for the individual observations in each cell of Table 2 also differ from the expected value at statistically significant levels. In the table, the cells differing from the expected values (at statistically significant levels) have all been marked in grey: dark grey for the ones substantially higher than expected, and light grey for the ones substantially lower than expected. This shows that some barter exchanges were much more common than ought to be expected from the aggregate trade data (and vice versa, that some other barter exchanges were much less common than expected). Alcohol thus does on an aggregate level indeed seem to have been more commonly bartered for gold than it was for other commodities; European provisions were likewise more commonly bartered for gold; arms (including gunpowder) were more commonly bartered for slaves; metals and metalware were more commonly bartered for ivory; and textiles were seemingly more commonly bartered for either slaves or African provisions than would be expected from the aggregate trade data. Less common barter exchanges than expected were, on the other hand, alcohol in exchange for slaves; European provisions in exchange for slaves; arms in exchange for gold; and textiles in exchange for gold. Table 3 breaks down the data chronologically, focusing on these key barter patterns noted here, in order to analyse whether these patterns remained stable, or whether any of them changed over time.    Note: Cells with observed values different than expected, at statistically significant levels, have been highlighted in the table: dark grey for the observations higher than expected, and light grey for the observations lower than expected. Source: Database assembled from BNA T/70/389-421.
The data thus suggests that there were several disproportionately common barter exchanges. The perhaps most disproportionately common exchange was the barter of alcohol in exchange for gold (and disproportionately less common in exchange for slaves). Gold accounted for roughly one third of the value of what the RAC purchased from African agents on the Gold Coast during this period. Much of this was paid for with alcohol in particular, so that more than two thirds (69%, as measured by the value of the transactions) of all alcohol sold by the Europeans was used for the purchase of gold only. The pattern is remarkably robust both chronologically, as can be seen in Table 3, and for all locations along the Gold Coast represented in the sample (see appendix Table A2). As the source reveals little to nothing about who the African agents acquiring this alcohol were, we cannot from this source tell who the final consumers of this alcohol were. Previous scholarship has, however, shown how the consumption of alcohol on the Gold Coast was associated with traditional religion to a large extent. Alcohol was, for example, important during funeral rituals and other rites de passage. It was also of the utmost importance for the odwira, the yam harvest festival held in honour of deceased kings in September every year. Imported spirits were also important when receiving guests, and the ability to serve the guests with such alcohol could enhance the status of the host (Dumett 1974, 81-5;Akyeampong 1995, 266-8;Akyeampong 1996, ch. 1-2;Nugent 2014). In Asante, for example, as Emmanuel Akyeampong has put it: 'alcohol symbolized royal power, patronage, generosity' (Akyeampong 1996, 6). Because of the intimate connection between alcohol and power, the male elders tried to monopolize the use of alcohol, for example by fostering a temperance ethic except for when alcohol was used for religious or ritual purposes under the control of the elders (Akyeampong 1996, 15, 21-3). One hypothesis is, thus, that the social traditions that the consumption of alcohol was embedded in made imported alcohol a particular prestige good for the elite. Elites were furthermore in control of much of the gold exports. Future research might shed further light on whether this imported alcohol was consumed only by the members of the elite, or whether it also was used, for example, to control gold producers in particular, or even society as a whole. If the latter was the case, it could be relevant to talk of a cycle of trade in this case, where the social prestige gained from the control over imported alcohol had an effect upon the production (and exports) of gold.
In a similar fashion, both tobacco and European provisions of various sorts (e.g. beef or butter) were minor trade goods on the African coast, but disproportionately commonly used when bartering for gold in particular. The difference between the observed and expected observations are in these cases, too, statistically significant for virtually all benchmark periods, and for most geographical locations in the sample. This might be an indication that both tobacco and European provisions were also prestige goods primarily demanded by local African elites.
Yet another disproportionately common barter exchange was textiles in exchange for slaves (and disproportionately less commonly used when bartering for gold). Textiles were the single largest type of commodity used in the barter trade on the coast, accounting for between 40% and 70% of the value of the aggregated purchases, respectively, depending on the type of commodity purchased. Regardless of what sort of commodity the Europeans purchased (staple foods such as corn or yams, luxuries such as gold, or human slaves), textiles could thus always be, and were in the vast majority of each individual case, used in such transactions. Indeed, it is quite rare to find any transaction in the accounts of the RAC where no textiles at all were included, but odd examples do exist. Textiles were, however, disproportionately often used when bartering for slaves in particular. Textiles were, on the other hand, used relatively less than statistically expected in transactions when Europeans were purchasing gold. These relationships hold true not only for the aggregate dataset, but also when disaggregating the data chronologically, as well as for most places on the Gold Coast when disaggregating the data geographically (see Table 3 and Appendix Table A2). Some of the other relationships involving the trade in textiles that the data in Table 2 would hint at (textiles bartered for African provisions and textiles bartered for ivory) are not robust when disaggregating the data. One hypothesis explaining the pattern of textiles bartered primarily for slaves, and less commonly for gold, might be that the trade in slaves was less controlled by African elites than the trade in gold was, and therefore less associated with luxurious prestige goods than the trade in gold. This is not to say that all the textiles exported to Africa were of low quality. There were undoubtedly many highly luxurious types of textiles exported to the region. There was, however, a great diversity among the textiles, catering to different consumers. Figure 6 shows the barter exchanges when disaggregating the types of textiles crudely into two different groups, based on the average coast price per piece of each type of textile (all types of textiles with an average coast price of 30 shillings or more per piece classified as 'expensive' textiles, and all other textiles classified as 'cheaper' textiles).
As can be seen in Figure 6, the expensive types of textiles are to a much greater extent bartered for gold (and correspondingly a lesser extent for slaves) than the cheaper types of textiles were. Disaggregating the data in Table 2 by quality of the textiles confirms the pattern (see Appendix Table A3): that the barter of textiles for slaves is disproportionately common (and vice versa, that the barter of textiles for gold is disproportionately uncommon) is true only for the cheaper types of textiles, and not for the more expensive types of textiles. The pattern for the expensive textiles does thus seem more similar to that of the trade in other prestige goods, such as gold, described earlier.
Another disproportionately common barter exchange was arms (here including not only pistols, guns and muskets, but also gunflints and gunpowder) in exchange for slaves. In total, slaves made up roughly half of the value of all European purchases on the Gold Coast during this period, but 64% of the arms sold (again as measured by the value of the transactions) were bartered solely for slaves. This evidence fits well with the general guns-for-slaves cycle of trade shown previously by Whatley (2018). The relationship is, however, not entirely consistent in the sample studied hereneither geographically nor over timeso there is a statistically significant difference between the observed and expected values for this type of barter exchange only during some of the benchmark years, and for some of the geographical locations (see Table 3 and Appendix Table A2). The relationship was thus strong during the 1720s and 1740s, respectively, but there was no statistically significant relationship during the 1730s; moreover, the relationship was particularly strong in only two locationsat the Cape Coast and at Commenda. One hypothesis that could explain such a pattern is that this is correlated with the intensity of armed conflicts in the region. In previous research, scholars have identified at least eight conflicts in the region between 1720 to 1728, several of them including the Fante nation (the polity controlling the region around the RAC's main trading stations on the Gold Coast). The last years of the 1720s and the first half of the 1730s seem to have been much more peaceful in the region, with only three conflicts identified by scholars from 1729 to 1736 (and none of them involving the Fante nation). Only by the late 1730s and early 1740s did conflicts again intensify in the region (nine conflicts identified by scholars between 1737 and 1745). The most protracted conflict of this last period involved the Fante nation, and at this time both the Danish and the Dutch trading stations also got involved in some of the conflicts (Freeman-Grenville 1973, 124-6;Boahen 1973, 37-9: Sanders 1980Thornton 1999, 70-2;Shumway 2011, xi). That the relationship between guns and slaves is strong during the 1720s and 1740s, but not in the 1730s, might thus perhaps be a reflection of the level of conflict in the region around these benchmark periods, something which also would correlate well with the development of the external slave trade from the region at this time (Whatley 2018, 101).
A final disproportionately common barter exchange was metals or metalwareoften, for example, knives, but also various types of pewterwarein exchange for ivory. We can at this stage only hypothesize about the reasons for this relationship. One hypothesis is that it is a reflection of the demand from local hunters, who supplied much of the ivory bought by the RAC; they might thus have demanded everyday tools or metals that would be worked into tools, rather than prestige goods such as alcohol or luxurious textiles demanded by African elites.

Concluding discussion
The trade patterns shown in this paper might at first glance seem to exhibit some likeness to what the theory of separate spheres of exchange would predict, where prestige goods (such as alcohol) to a disproportionate extent are exchanged for other prestige goods (such as gold), and subsistence goods are exchanged primarily for other subsistence goods. The evidence presented in this paper can however at least clearly reject a strong interpretation of this theory, according to which there existed strictly separated spheres of exchange between which no exchange at all could take place. In the trade studied in this paper, this was clearly not the case: both subsistence goods (such as corn, palm oil or yams) and prestige goods (such as slaves or gold) could apparently be sold in exchange for any type of European commodity bought. Judging from the barter transactions studied, there was no specific composition of barter which seems to have been completely unacceptable for the African traders.
There were, however, some barter transactions that seem to have been preferred more than others. One disproportionately common barter exchange on the Gold Coast was an exchange of guns for slaves. This evidence fits well with much previous research on a guns-for-slaves cycle of trade, wherein the product purchased by the African agents (guns) was used for the gathering (supply) of the product that the same agents sold (slaves), and which therefore would be associated with the fluctuations in the external slave trade from the region. Other disproportionately common barter exchanges revealed by the empirical evidence, which in contrast seem to remain comparatively stable over the period under study, include that various types of alcohol that were bartered for gold to a disproportionate extent. Textiles, furthermore, remained the single most important type of commodity sold throughout the period under study in this paper, and virtually always were included at least in some small measure in the assortments of goods bartered for African goods. Disaggregating the category of textiles into two sub-groups by how expensive the textiles were shows, however, that more expensive textiles were bartered for gold more commonly than the less expensive textiles were. Tobacco was likewise bartered for gold to a disproportionate extent. The patterns to the barter trade in these goods might thus be a reflection of the social institutions that the trade was embedded in, where the social elite controlled much of the trade in gold and thereby was able to enjoy the consumption of various imported luxury goods.

Disclosure statement
No potential conflict of interest was reported by the author. Table A2. Actual and expected distribution of barter trade relationships as share of all barter relationships at that location (actual distribution in %, expected distribution in % in parentheses).