The Long-Term Effects of Africa's Slave Trades (Article begins on next page) The Harvard community has made this article openly available. Please share how this access benefits you. Your story matters. Citation Nunn, Nathan. 2008. The long-term effects of Africa's slave trades. Quarterly Journal of Economics 123(1): 139-176. Published Version doi:10.1162/qjec.2008.123.1.139 Accessed September 18, 2017 8:27:39 AM EDT Citable Link http://nrs.harvard.edu/urn-3:HUL.InstRepos:3710252 Terms of Use This article was downloaded from Harvard University's DASH repository, and is made available under the terms and conditions applicable to Other Posted Material, as set forth at http://nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of- use#LAA THE LONG-TERM EFFECTS OF AFRICA’S SLAVE TRADES* NATHAN NUNN Can part of Africa’s current underdevelopment be explained by its slave trades? To explore this question, I use data from shipping records and historical documents reporting slave ethnicities to construct estimates of the number of slaves exported from each country during Africa’s slave trades. I find a robust negative relationship between the number of slaves exported from a country and current economic performance. To better understand if the relationship is causal, I examine the historical evidence on selection into the slave trades and use instrumental variables. Together the evidence suggests that the slave trades had an adverse effect on economic development. I. INTRODUCTION Africa’s economic performance in the second half of the twentieth century has been poor. One, often informal, explanation for Africa’s underdevelopment is its history of extraction, characterized by two events: the slave trades and colonialism. Bairoch (1993, p. 8) writes that “there is no doubt that a large number of negative structural features of the process of economic underdevelopment have historical roots going back to European colonization.” Manning (1990, p. 124) echoes Bairoch but focuses on the slave trades, writing, “Slavery was corruption: it involved theft, bribery, and exercise of brute force as well as ruses. Slavery thus may be seen as one source of precolonial origins for modern corruption.” Recent empirical studies suggest that Africa’s history can explain part of its current underdevelopment. These studies focus on the link between countries’ colonial experience and current economic development (Grier 1999; Englebert 2000a, 2000b; * A previous version of this paper was circulated under the title “Slavery, Institutional Development, and Long-Run Growth in Africa.” I am grateful to the editor, Edward Glaeser, and three anonymous referees for comments that substantially improved this paper. I also thank Daron Acemoglu, Robert Bates, Albert Berry, Loren Brandt, Jon Cohen, Bill Easterly, Stanley Engerman, Azim Essaji, Joseph Inikori, Martin Klein, Pat Manning, Ted Miguel, Jim Robinson, Aloysius Siow, Ken Sokoloff, Dan Trefler, Chris Udry, Jeffrey Williamson, and seminar participants at the University of British Columbia, the University of California Los Angeles, the University of California San Diego, Harvard University, the University of Michigan, New York University, Pennsylvania State University, the University of Rochester, the University of Southern California, the University of Toronto, York University, the CIFAR, the SED Conference, the CEA Meetings, the SSHA Meetings, the ITAM Summer Camp in Macroeconomics, the IEHC, the NBER, and the WGAPE meetings for valuable comments and suggestions. I thank Maira Avila and Ken Jackson for excellent research assistance. C 2008 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology. The Quarterly Journal of Economics, February 2008 139 140 QUARTERLY JOURNAL OF ECONOMICS Acemoglu, Johnson, and Robinson 2001, 2002; Bertocchi and Canova 2002; Lange 2004). However, the other important event in Africa’s history, its slave trades, has yet to be examined empirically. There are reasons to expect that the slave trades may have been at least as important as official colonial rule for Africa’s development. For a period of nearly 500 years, from 1400 to 1900, the African continent simultaneously experienced four slave trades. By comparison, official colonial rule lasted from 1885 to about 1960, a total of approximately 75 years. This paper provides the first empirical examination of the importance of Africa’s slave trades in shaping subsequent economic development. In doing this, I construct measures of the number of slaves exported from each country in Africa in each century between 1400 and 1900. The estimates are constructed by combining data from ship records on the number of slaves shipped from each African port or region with data from a variety of historical documents that report the ethnic identities of slaves that were shipped from Africa. I find a robust negative relationship between the number of slaves exported from each country and subsequent economic performance. The African countries that are the poorest today are the ones from which the most slaves were taken. This finding cannot be taken as conclusive evidence that the slave trades caused differences in subsequent economic development. An alternative explanation that is just as plausible is that countries that were initially the most economically and socially underdeveloped selected into the slave trades, and these countries continue to be the most underdeveloped today. In other words, the slave trades may be correlated with unobserved country characteristics, resulting in biased estimates of the effect of the slave trades on economic development. I pursue a number of strategies to better understand the reason behind the relationship between slave exports and current economic performance. First, I review the evidence from African historians on the nature of selection into the slave trades. I also use historic data on pre–slave trade population densities to examine whether it was the less developed parts of Africa that selected into the slave trades. Both sources of evidence show that it was actually the most developed areas of Africa that tended to select into the slave trades. I discuss the reason behind this seemingly paradoxical relationship in detail. Second, I use instruments to estimate the causal effect of the slave trades on subsequent economic development. The instruments are the sailing distances THE LONG-TERM EFFECTS OF AFRICA’S SLAVE TRADES 141 from each country to the nearest locations of demand for slave labor in each of the four slave trades. Like the OLS coefficients, the IV coefficients are negative and significant, suggesting that increased extraction during the slave trades caused worse subsequent economic performance. I then explore the precise channel of causality underlying the relationship between slave exports and economic development. Using historical evidence as a guide, I examine whether the procurement of slaves through internal warfare, raiding, and kidnapping resulted in subsequent state collapse and ethnic fractionalization. I find that the data are consistent with these channels. These findings complement the research of Engerman and Sokoloff (1997, 2002), which shows that slavery in the New World resulted in the evolution of institutions that were not conducive to economic growth.1 My results show that not only was the use of slaves detrimental for a society, but the production of slaves, which occurred through domestic warfare, raiding, and kidnapping, also had negative impacts on subsequent development. The paper is structured as follows. In the following section, I provide a description of Africa’s slave trades, providing a detailed historical overview of the manner in which slaves were procured and the resulting adverse effects. In Section III, I describe the construction of the slave export figures. Section IV documents the correlations that exist in the data, and Section V turns to the issue of causality. In Section VI, guided by the historical evidence, I examine the potential channels of causality. Section VII concludes. II. HISTORICAL BACKGROUND Between 1400 and 1900, the African continent experienced four simultaneous slave trades. The largest and most well-known is the trans-Atlantic slave trade where, beginning in the fifteenth century, slaves were shipped from West Africa, West-Central Africa, and Eastern Africa to the European colonies in the New World. The three other slave trades—the trans-Saharan, Red Sea, and Indian Ocean slave trades—were much older and pre-dated the trans-Atlantic slave trade. During the trans-Saharan slave trade, slaves were taken from south of the Saharan desert to 1. Also see Lagerl¨of (2005) and Mitchener and McLean (2003) for related evidence. 142 QUARTERLY JOURNAL OF ECONOMICS Northern Africa. In the Red Sea slave trade, slaves were taken from inland of the Red Sea and shipped to the Middle East and India. In the Indian Ocean slave trade, slaves were taken from Eastern Africa and shipped either to the Middle East and India or to plantation islands in the Indian Ocean. A number of characteristics of Africa’s slave trades make them distinct from previous slave trades. First, the total volume of slaves traded was unprecedented. During the trans-Atlantic slave trade alone, approximately 12 million slaves were exported from Africa. Another 6 million were exported in the other three slave trades. These figures do not include those who were killed during the raids or those who died on their journey to the coast. The total effect of the slave trades, according to calculations by Patrick Manning (1990, p. 171), was that by 1850 Africa’s population was only half of what it would have been had the slave trades not taken place. Africa’s slave trades were also unique because, unlike previous slave trades, individuals of the same or similar ethnicities enslaved one another. This had particularly detrimental consequences, including social and ethnic fragmentation, political instability and a weakening of states, and the corruption of judicial institutions. The most common manner in which slaves were taken was through villages or states raiding one another (Northrup 1978; Lovejoy 1994). Where groups of villages had previously developed into larger-scale village federations, relations between the villages tended to turn hostile (e.g., Azevedo 1982; Inikori 2000; Hubbell 2001). As a result, ties between villages were weakened, which in turn impeded the formation of larger communities and broader ethnic identities. Kusimba (2004, p. 66) writes that “insecurity confined people within ethnic boundaries constructing spheres of interaction.” Because of this process, the slave trades may be an important factor explaining Africa’s high level of ethnic fractionalization today. This is significant for economic development given the established relationship between ethnic fractionalization and long-term economic growth (Easterly and Levine 1997). Because of the environment of uncertainty and insecurity at the time, individuals required weapons, such as iron knives, spears, swords or firearms, to defend themselves. These weapons could be obtained from Europeans in exchange for slaves, who were often obtained through local kidnappings. This further perpetuated the slave trade and the insecurity that it caused, THE LONG-TERM EFFECTS OF AFRICA’S SLAVE TRADES 143 which in turn further increased the need to enslave others to protect oneself (Mahadi 1992; Hawthorne 1999, pp. 108–109). Historians have named this vicious cycle the “gun–slave cycle” (e.g., Lovejoy 2000) or the “iron–slave cycle” (e.g., Hawthorne 2003). The result of this vicious cycle was not only that communities raided other communities for slaves, but also that members of a community raided and kidnapped others within the community. Well-documented examples come from the Balanta of modern day Guinea-Bissau, the Minyanka of modern day Mali (Klein 2001), and the Makua, Chikunda, and Yao of East Central Africa (Alpers 1969, pp. 413–414, 1975, p. 225; Isaacman 1989, pp. 191–192, 196). Generally, the consequence of internal conflict was increased political instability and in many cases the collapse of preexisting forms of government (Lovejoy 2000, pp. 68–70). In sixteenthcentury northern Senegambia, the Portuguese slave trade was a key factor leading to the eventual disintegration of the Joloff Confederation, which was replaced by the much smaller kingdoms of Waalo, Kajoor, Baol, Siin, and Saalum. Further south, in southern Senegambia, the same pattern is observed. Prior to the slave trades, complex state systems were in the process of evolving. However, this evolution stagnated soon after the arrival of the Portuguese in the 15th century (Barry 1998, pp. 36–59). Similar patterns of instability have also been documented in Eastern Africa (e.g., Isaacman [1989]; Mbajedwe [2000]). In the late 19th century, the slave trades resulted in the disintegration of the Shambaa kingdom, the Gweno kingdom, and the Pare states in East Africa’s Pangani valley (Kimambo 1989, p. 247; Mbajedwe 2000, pp. 341–342). The most dramatic example may be the Kongo kingdom of West-Central Africa. As early as 1514, the kidnapping of local Kongo citizens for sale to the Portuguese had become rampant, threatening social order and the King’s authority. In 1526, Affonso, king of Kongo, wrote to Portugal complaining that “there are many traders in all corners of the country. They bring ruin to the country. Every day people are enslaved and kidnapped, even nobles, even members of the king’s own family” (Vansina 1966, p. 52). This break-down of law and order was partly responsible for the weakening and eventual fall of the once powerful state (Inikori 2003). For many of the other Bantu-speaking ethnicities, stable states also existed in earlier periods, but by the time the slave trades were brought to an end, few ancient states remained (Colson 1969, pp. 36–37). 144 QUARTERLY JOURNAL OF ECONOMICS Preexisting governance structures were generally replaced by small bands of slave raiders, controlled by an established ruler or warlord. However, these bands were generally unable to develop into large, stable states. Colson (1969, p. 35) writes that “both the bands and the new states they created retained an air of improvisation. Few band leaders were able to hand power to a legitimate successor. Even where a band leader had become the ruler of a state, succession remained a problem. Leadership was a personal role, rather than an established office.” The slave trades also contributed to political instability by causing the corruption of previously established legal structures. In many cases, it became common to obtain slaves by falsely accusing others of witchcraft or other crimes (Koelle 1854; Northrup 1978; Lovejoy 2000). Klein (2001, p. 59) writes that “communities began enslaving their own. Judicial penalties that formerly had taken the form of beatings, payment of compensation or exile, for example, were now converted to enslavement.” Often, leaders themselves supported or even instigated this abuse of the judicial system (Mahadi 1992; Hawthorne 1999, 2003; Klein 2001). To protect themselves and their communities from being raided, leaders often chose to pay slaves as tribute, which were often obtained through the judicial system. Hawthorne (1999, 2003) provides detailed studies of this process among the Cassanga of modern day Guinea Bissau. The chief of the Cassanga used the “red water ordeal” to procure slaves and their possessions. Those accused of a crime were forced to drink a poisonous red liquid. If they vomited, then they were judged to be guilty. If they did not vomit, they were deemed not guilty. However, for those that did not vomit this usually brought death by poisoning. Their possessions were then seized and their family members were sold into slavery. Evidence from research showing a relationship between a country’s history of state development and subsequent economic performance suggests that these effects of the slave trades may be important for current economic development (Bockstette, Chanda, and Putterman 2002; Chanda and Putterman 2005). Others have argued that Africa’s underdevelopment is a direct result of state failure, which stems from Africa’s weak and unstable precolonial political structures (Herbst 1997, 2000). Because Africa’s slave trades were an important factor affecting political underdevelopment, they may be a central reason behind Africa’s weak states today. THE LONG-TERM EFFECTS OF AFRICA’S SLAVE TRADES 145 III. SLAVE EXPORT DATA Because I am interested in examining the effects of the slave trades that resulted because of the procurement of slaves, my measure of interest is the total number of slaves taken from each country during the four slave trades between 1400 and 1900.2 I use two types of data to construct the slave export estimates. The first are data that report the total number of slaves exported from each port or region in Africa. I refer to these as shipping data. For the trans-Atlantic slave trade, the data are from the updated version of the Trans-Atlantic Slave Trade Database constructed by Eltis et al. (1999). The database records information for 34,584 voyages from 1514 to 1866. The shipping data are originally from various documents and records located around the world. Because, in most European ports, merchants were required to register their ships and declare the volume and value of goods transported for each ship and voyage, typically, there exists a number of different registers and documents. In the database, 77% of the trans-Atlantic slave voyages after 1700 have shipping information from more than one source; the average number of sources for each voyage is six. It is estimated that the database contains 82% of all trans-Atlantic slaving voyages ever attempted (Eltis and Richardson 2006). Data for the early period of the Atlantic slave trade not covered by the Trans-Atlantic Slave Trade Database are from Elbl (1997). For the Indian Ocean, Red Sea, and trans-Saharan slave trades, data are from Austen (1979, 1988, 1992). The data are based on estimates from all available documents, records, and accounts by observers and government officials on the location and volume of slave exports. With the shipping data one can calculate the number of slaves that were shipped from each coastal country. However, this does not give an accurate indication of where slaves were originally captured. Slaves shipped from the ports of a coastal country may have come from a country located further inland. To estimate the number of slaves shipped from the coast that came from inland countries, I also use a second source of data that reports the ethnic identity of slaves shipped from Africa. This information comes from a variety of sources, such as records of sale, slave registers, 2. Ideally, I would also like to include people that entered into local domestic slavery. However, the necessary data to construct these estimates do not exist. 146 QUARTERLY JOURNAL OF ECONOMICS slave runaway notices, court records, church records, and notarial documents. There were a number of ways of identifying the ethnicity or “nation” of a slave. The easiest was often by a slave’s name. Slaves were often given a Christian first name and a surname that identified their ethnicity (e.g., Tardieu [2001]). As well, a slave’s ethnicity could often be determined from ethnic markings, such as cuts, scars, hairstyles, or the filing of teeth (Karasch 1987, pp. 4–9). Oldendorp (1777, p. 169) writes that “the people of all Negro nations are marked with certain cuts on the skin. As far as I have been able to learn from the Negroes themselves, these serve to distinguish one nation from another.” Because slaves were legally defined as property, those engaged in the buying and selling of slaves had a strong incentive to correctly identify the birthplace or “nation” of slaves (Wax 1973). Moreno Fraginals (1977, p. 190) writes that “the slave trade was the business that involved the greatest amount of capital investment in the world during the eighteenth and nineteenth centuries. And a business of this size would never have kept up a classificatory scheme had it not been meaningful (in overall general terms, in keeping with reality) in designating in a very precise way the merchandise that was being traded.” Information on the ethnicities of slaves shipped during the trans-Atlantic slave trade come from 54 different samples, totalling 80,656 slaves, with 229 distinct ethnic designations reported. Table I summarizes information about the samples used in the trans-Atlantic slave trade. The table reports the location, the years covered, the number of slaves, and the number of ethnicities that could be identified for each sample. Similar tables for the other three slave trades are reported in Nunn (2007). The ethnicity data for the Indian Ocean slave trade come from six samples, with a total of 21,048 slaves and 80 different ethnicities reported. The data for the Red Sea slave trade are from two samples: one from Jedda, Saudi Arabia, and the other from Bombay, India. The samples provide information for 67 slaves, with 32 different reported ethnicities. For the trans-Saharan slave trade two samples are available: one from central Sudan and the other from western Sudan. The samples provide information on the origins of 5,385 slaves, with 23 different ethnicities recorded. The shipping data from Austen (1992) also provide additional information on which caravan slaves were shipped on, the city or town that the caravan originated in, the destination of the THE LONG-TERM EFFECTS OF AFRICA’S SLAVE TRADES 147 TABLE I SLAVE ETHNICITY DATA FOR THE TRANS-ATLANTIC SLAVE TRADE Num. Num. Location Years ethnic. obs. Record type Valencia, Spain 1482–1516 77 2,675 Crown records Puebla, Mexico 1540–1556 14 115 Notarial records Dominican Republic 1547–1591 26 22 Records of sale Peru 1548–1560 16 202 Records of sale Mexico 1549 12 80 Plantation accounts Peru 1560–1650 30 6,754 Notarial records Lima, Peru 1583–1589 15 288 Baptism records Colombia 1589–1607 9 19 Various records Mexico 1600–1699 28 102 Records of sale Dominican Republic 1610–1696 33 55 Government records Chile 1615 6 141 Sales records Lima, Peru 1630–1702 33 409 Parish records Peru (Rural) 1632 25 307 Parish records Lima, Peru 1640–1680 33 936 Marriage records Colombia 1635–1695 6 17 Slave inventories Guyane (French Guiana) 1690 12 69 Plantation records Colombia 1716–1725 33 59 Government records French Louisiana 1717–1769 23 223 Notarial records Dominican Republic 1717–1827 11 15 Government records South Carolina 1732–1775 35 681 Runaway notices Colombia 1738–1778 11 100 Various records Spanish Louisiana 1770–1803 79 6,615 Notarial records St. Dominique (Haiti) 1771–1791 25 5,413 Sugar plantations Bahia, Brazil 1775–1815 14 581 Slave lists St. Dominique (Haiti) 1778–1791 36 1,280 Coffee plantations Guadeloupe 1788 8 45 Newspaper reports St. Dominique (Haiti) 1788–1790 21 1,297 Fugitive slave lists Cuba 1791–1840 59 3,093 Slave registers St. Dominique (Haiti) 1796–1797 56 5,632 Plantation inventories American Louisiana 1804–1820 62 223 Notarial records Salvador, Brazil 1808–1842 6 456 Records of manumission Trinidad 1813 100 12,460 Slave registers St. Lucia 1815 62 2,333 Slave registers Bahia, Brazil 1816–1850 27 2,666 Slave lists St. Kitts 1817 48 2,887 Slave registers Senegal 1818 17 80 Captured slave ship Berbice (Guyana) 1819 66 1,127 Slave registers Salvador, Brazil 1819–1836 12 871 Manumission certificates Salvador, Brazil 1820–1835 11 1,106 Probate records Sierra Leone 1821–1824 68 605 Child registers Rio de Janeiro, Brazil 1826–1837 31 772 Prison records Anguilla 1827 7 51 Slave registers Rio de Janeiro, Brazil 1830–1852 190 2,921 Free africans’ records Rio de Janeiro, Brazil 1833–1849 35 476 Death certificates 148 QUARTERLY JOURNAL OF ECONOMICS TABLE I (CONTINUED) Num. Num. Location Years ethnic. obs. Record type Salvador, Brazil 1835 13 275 Court records Salvador, Brazil 1838–1848 7 202 Slave registers St. Louis/Goree, 1843–1848 21 189 Emancipated slaves Senegal Bakel, Senegal 1846 16 73 Sales records d’Agou´e, Benin 1846–1885 11 70 Church records Sierra Leone 1848 132 12,425 Linguistic and British census Salvador, Brazil 1851–1884 8 363 Records of manumission Salvador, Brazil 1852–1888 7 269 Slave registers Cape Verde 1856 32 314 Slave census Kikoneh Island, 1896–1897 11 185 Fugitive slave records Sierra Leone caravan, and in some cases the ethnic identity of the slaves being shipped. To illustrate how I combine the ethnicity data with the shipping data to construct my estimates I use an example, which is shown in Figure I. The figure is a hypothetical map of the western coast of Africa, with each square representing a country. From the shipping data, I first calculate the number of slaves shipped from each coastal country in Africa. In this example 100,000 slaves were shipped from Country A and 250,000 were shipped from Country C. The problem with relying on the shipping data alone is that many of slaves shipped from Country A may have come from Country B, which lies landlocked behind Country A. Then, using the ethnicity data, I calculate the ratio of slaves from each coastal country relative to any landlocked countries located inland of the coastal country. This requires that I map ethnicities to countries and aggregate up to the country level. In practice, this step relied on a great amount of past research by African historians, linguists, and ethnographers. The sources most heavily used are Koelle (1854), Murdock (1959), Curtin (1969), Higman (1984), and Hall (2005). Assume that the ratio of slaves from Country A relative to Country B is 4 to 1. This ratio suggests that 20% of the slaves shipped from Country A were actually from Country B. Therefore, the estimated number of slaves from Country B is 20,000 and from Country A is 80,000. Assume that the ratio of slaves THE LONG-TERM EFFECTS OF AFRICA’S SLAVE TRADES 149 FIGURE I An Artificial Map of the West Coast of Africa FIGURE II Ethnic Boundaries Defined by Murdock (1959) and Modern Political Boundaries 150 QUARTERLY JOURNAL OF ECONOMICS from Country C to D to E is 3 to 1 to 1. The same procedure then yields an estimate of 150,000 slaves from Country C and 50,000 each from Countries D and E. In practice, these calculations are performed separately for each slave trade. As well, because slaves were increasingly taken from further inland as each slave trade progressed, the calculations are also performed separately for each of the following time periods: 1400–1599, 1600– 1699, 1700–1799, 1800–1900.3 Because ethnicities tended to be much smaller than countries, the mapping of ethnicities into countries generally is not problematic. This is illustrated in Figure II, which shows African ethnicities based on Murdock’s (1959) classification, as well as modern political boundaries. From the figure it is apparent that ethnicities are much smaller than modern boundaries, and therefore ethnicities generally map cleanly into one country.4 An important assumption in this procedure is that slaves shipped from a port within a country are either from that country or from countries directly to the interior. However, in reality, some slaves shipped from a country’s coast may have originated from a neighboring coastal country. In Nunn (2007), using three samples of slaves for which we know both the ethnicity of the slaves and the port that they were shipped from, I test the validity of this assumption and the overall accuracy of the estimation procedure. I find that for each of the three samples my procedure correctly identifies the origins of between 83% and 98% of the slaves in the samples. A second source of measurement error arises because slaves from the interior will tend to be underrepresented in the ethnicity samples. This is because only slaves who survived the voyage outside of Africa are in the ethnicity samples. All else equal, the further inland a slave originated, the longer the journey was, and the more likely it was that he or she died along the way. Because the rates of mortality during the slave trades were extremely high, this form of measurement error may be significant.5 However, as I show formally in the Appendix, the undersampling of slaves from 3. See Nunn (2007) for all of the finer details of the construction procedure. 4. In instances where an ethnicity is located in more than one country, I map the ethnicity into the multiple countries using land area as weights. This is explained in detail in Nunn (2007). 5. Estimates of cross-Atlantic mortality rates ranged from 7% to 20% depending on the time period and the length of the voyage (Curtin 1969, pp. 275–286; Lovejoy 2000, p. 63). Death rates during the trek to the coast are known with less certainty, but estimates range from 10% to 50% (Vansina 1990, p. 218; Lovejoy 2000, pp. 63–64). THE LONG-TERM EFFECTS OF AFRICA’S SLAVE TRADES 151 the interior results in OLS estimates that are biased toward zero. As well, one can use instruments that are uncorrelated with the measurement error to derive consistent estimates. I do this in Section V.B. After the data have been constructed, I have estimates of the number of slaves shipped from each country in Africa during each of the four slave trades during four different time periods: 1400–1599, 1600–1699, 1700–1799, 1800–1900. Table II reports the estimated total number of slaves exported from each country, as well as the total disaggregated by slave trade. Overall, the estimates are consistent with the general view among African historians of where the primary slaving areas were. During the trans-Atlantic slave trade, slaves were taken in greatest numbers from the “Slave Coast” (Benin and Nigeria), West-Central Africa (Zaire, Congo, and Angola), and the “Gold Coast” (Ghana). All of these countries appear among the top exporting countries on the list. Ethiopia and Sudan are also among the top exporting countries because they were the primary suppliers of slaves shipped during the Red Sea and Saharan slave trades. The low number of slave exports from South Africa and Namibia confirms the view of African historians that these areas exported “virtually no slaves” (Manning 1983, p. 839). The relative magnitudes of exports from geographically close countries are also consistent with the qualitative evidence from the African history literature. Manning (1983, p. 839) writes that “some adjoining regions were quite dissimilar: Togo exported few slaves and the Gold Coast many; Gabon exported few slaves, and the Congo exported many.” The estimates are consistent with Manning’s observation. Exports from Togo are far less than from Ghana, and exports from Gabon are less than from the Republic of Congo. IV. BASIC CORRELATIONS: OLS ESTIMATES I begin by examining the relationship between past slave exports and current economic performance. I normalize the total number of slaves taken from each country by its size, measured by land area.6 Figure III shows the relationship between the natural log of the number of slaves exported in all four slave trades 6. Alternative measures of country size are also possible. As I show in Table A.2 in the Appendix, using the average population of a country between 1400 and 1900 yields nearly identical results. Similarly, using arable land area rather than land area also yields essentially identical results. 152 QUARTERLY JOURNAL OF ECONOMICS TABLE II ESTIMATED TOTAL SLAVE EXPORTS BETWEEN 1400 AND 1900 BY COUNTRY Trans- Indian Trans- Red All slave Isocode Country name Atlantic Ocean Saharan Sea trades AGO Angola 3,607,020 0 0 0 3,607,020 NGA Nigeria 1,406,728 0 555,796 59,337 2,021,859 GHA Ghana 1,614,793 0 0 0 1,614,793 ETH Ethiopia 0 200 813,899 633,357 1,447,455 SDN Sudan 615 174 408,261 454,913 863,962 MLI Mali 331,748 0 509,950 0 841,697 ZAR Democratic 759,468 7,047 0 0 766,515 Republic of Congo MOZ Mozambique 382,378 243,484 0 0 625,862 TZA Tanzania 10,834 523,992 0 0 534,826 TCD Chad 823 0 409,368 118,673 528,862 BEN Benin 456,583 0 0 0 456,583 SEN Senegal 278,195 0 98,731 0 376,926 GIN Guinea 350,149 0 0 0 350,149 TGO Togo 289,634 0 0 0 289,634 GNB Guinea-Bissau 180,752 0 0 0 180,752 BFA Burkina Faso 167,201 0 0 0 167,201 MRT Mauritania 417 0 164,017 0 164,434 MWI Malawi 88,061 37,370 0 0 125,431 MDG Madagascar 36,349 88,927 0 0 125,275 COG Congo 94,663 0 0 0 94,663 KEN Kenya 303 12,306 60,351 13,490 86,448 SLE Sierra Leone 69,607 0 0 0 69,607 CMR Cameroon 66,719 0 0 0 66,719 DZA Algeria 0 0 61,835 0 61,835 CIV Ivory Coast 52,646 0 0 0 52,646 SOM Somalia 0 229 26,194 5,855 32,277 ZMB Zambia 6,552 21,406 0 0 27,958 GAB Gabon 27,403 0 0 0 27,403 GMB Gambia 16,039 0 5,693 0 21,731 NER Niger 133 0 0 19,779 19,912 LBY Libya 0 0 8,848 0 8,848 LBR Liberia 6,790 0 0 0 6,790 UGA Uganda 900 3,654 0 0 4,554 ZAF South Africa 1,944 87 0 0 2,031 CAF Central African 2,010 0 0 0 2,010 Republic EGY Egypt 0 0 1,492 0 1,492 ZWE Zimbabwe 554 536 0 0 1,089 NAM Namibia 191 0 0 0 191 BDI Burundi 0 87 0 0 87 GNQ Equatorial Guinea 11 0 0 0 11 DJI Djibouti 0 5 0 0 5 BWA Botswana 0 0 0 0 0 THE LONG-TERM EFFECTS OF AFRICA’S SLAVE TRADES 153 TABLE II (CONTINUED) Trans- Indian Trans- Red All slave Isocode Country name Atlantic Ocean Saharan Sea trades CPV Cape Verde Islands 0 0 0 0 0 COM Comoros 0 0 0 0 0 LSO Lesotho 0 0 0 0 0 MUS Mauritius 0 0 0 0 0 MAR Morocco 0 0 0 0 0 RWA Rwanda 0 0 0 0 0 STP S˜ao Tom´e & Principe 0 0 0 0 0 SWZ Swaziland 0 0 0 0 0 SYC Seychelles 0 0 0 0 0 TUN Tunisia 0 0 0 0 0 FIGURE III Relationship between Log Slave Exports Normalized by Land Area, ln(exports/area), and Log Real Per Capita GDP in 2000, ln y between 1400 and 1900 normalized by land area and the natural log of per capita GDP in 2000.7 As shown in the figure, a negative 7. Because the natural log of zero is undefined, I take the natural log of 0.1. As I show in the Appendix, the results are robust to the omission of these zero-export countries. 154 QUARTERLY JOURNAL OF ECONOMICS relationship between income and slave exports is apparent in the raw data. I further examine this relationship by controlling for other country characteristics that are also potentially important for current income. My baseline estimating equation is (1) ln yi = β0 + β1 ln(exportsi/areai) + Ciδ + Xiγ + εi, where ln yi is the natural log of real per capita GDP in country i in 2000, and ln(exportsi/areai) is the natural log of the total number of slaves exported between 1400 and 1900 normalized by land area. Per capita GDP data are for the year 2000 and are from Maddison (2003). Ci is a vector of dummy variables that indicate the origin of the colonizer prior to independence. These are included to control for the other significant event in Africa’s past, colonial rule. Xi is a vector of control variables that are meant to capture differences in countries’ geography and climate. OLS estimates of (1) are reported in Table III. The first column reports estimates of (1) with colonizer fixed effects only. In the second column, I also include controls that capture the potential importance of geography for long-term economic development: distance from the equator, longitude, minimum monthly rainfall, average maximum humidity, average minimum temperature, and proximity to the ocean measured by the natural log of coastline divided by land area. All factors, except longitude, influence whether a country has a tropical climate, which affects the prevalence of infectious disease and agricultural productivity (Kamarck 1976; Sachs et al. 2001). Longitude is also included to capture differences between the eastern and western parts of the African continent.8 In both specifications, the estimated relationship between slave exports and per capita income are negative and statistically significant. One concern with the estimates in columns (1) and (2) is that they may be biased because of the inclusion of small islands and North African countries in the sample. In column (3), I omit island and North African countries, dropping Morocco, Algeria, Tunisia, Libya, Egypt, Seychelles, Mauritius, Comoros, S˜ao Tom´e and Principe, and the Cape Verde Islands. As shown, dropping these countries makes little difference. The estimated coefficient 8. A related concern is that the slave exports variable may simply capture a West African fixed effect. I show in the Appendix that the results are robust to also controlling for region fixed effects. THE LONG-TERM EFFECTS OF AFRICA’S SLAVE TRADES 155 TABLE III RELATIONSHIP BETWEEN SLAVE EXPORTS AND INCOME Dependent variable is log real per capita GDP in 2000, ln y (1) (2) (3) (4) (5) (6) ln(exports/area) −0.112∗∗∗ −0.076∗∗∗ −0.108∗∗∗ −0.085∗∗ −0.103∗∗∗ −0.128∗∗∗ (0.024) (0.029) (0.037) (0.035) (0.034) (0.034) Distance from 0.016 −0.005 0.019 0.023 0.006 equator (0.017) (0.020) (0.018) (0.017) (0.017) Longitude 0.001 −0.007 −0.004 −0.004 −0.009 (0.005) (0.006) (0.006) (0.005) (0.006) Lowest monthly −0.001 0.008 0.0001 −0.001 −0.002 rainfall (0.007) (0.008) (0.007) (0.006) (0.008) Avg max humidity 0.009 0.008 0.009 0.015 0.013 (0.012) (0.012) (0.012) (0.011) (0.010) Avg min −0.019 −0.039 −0.005 −0.015 −0.037 temperature (0.028) (0.028) (0.027) (0.026) (0.025) ln(coastline/area) 0.085∗∗ 0.092∗∗ 0.095∗∗ 0.082∗∗ 0.083∗∗ (0.039) (0.042) (0.042) (0.040) (0.037) Island indicator −0.398 −0.150 (0.529) (0.516) Percent Islamic −0.008∗∗∗ −0.006∗ −0.003 (0.003) (0.003) (0.003) French legal origin 0.755 0.643 −0.141 (0.503) (0.470) (0.734) North Africa 0.382 −0.304 indicator (0.484) (0.517) ln(gold prod/pop) 0.011 0.014 (0.017) (0.015) ln(oil prod/pop) 0.078∗∗∗ 0.088∗∗∗ (0.027) (0.025) ln(diamond −0.039 −0.048 prod/pop) (0.043) (0.041) Colonizer fixed Yes Yes Yes Yes Yes Yes effects Number obs. 52 52 42 52 52 42 R2 .51 .60 .63 .71 .77 .80 Notes. OLS estimates of (1) are reported. The dependent variable is the natural log of real per capita GDP in 2000, ln y. The slave export variable ln(exports/area) is the natural log of the total number of slaves exported from each country between 1400 and 1900 in the four slave trades normalized by land area. The colonizer fixed effects are indicator variables for the identity of the colonizer at the time of independence. Coefficients are reported with standard errors in brackets. ∗∗∗, ∗∗, and ∗ indicate significance at the 1%, 5%, and 10% levels. for slave exports remains negative and significant, and the magnitude of the estimated coefficient actually increases.9 9. One may also be concerned that the inclusion of the countries in southern Africa—namely South Africa, Swaziland, and Lesotho—may also be biasing the results. As I report in the Appendix, the results are robust to also omitting this 156 QUARTERLY JOURNAL OF ECONOMICS In column (4), I include additional control variables to account for potential differences between islands or North African countries and the rest of Africa. Two core differences between North Africa and the rest of Africa is that North African countries are predominantly Islamic and that they all have legal systems based on French civil law. To capture these differences, I also include a measure of the percent of the population that is Islamic and a French legal origin indicator variable. I also include a North Africa fixed effect and an island fixed effect. As shown, including these additional control variables does not affect the slave exports coefficient, which remains negative and statistically significant. The final factor that I control for is differences in countries’ endowments of natural resources. In column (5), I also control for the natural log of the annual average per capita production between 1970 and 2000 of gold, oil, and diamonds. As shown, controlling for countries’ production of these natural resources does not alter the results.10 In column (6), I simultaneously include all control variables and drop islands and North African countries from the sample. Again the results remain robust.11 The estimated magnitudes of the relationship between slave exports and income are not only statistically significant but also economically meaningful. Calculating the standardized beta coefficients of the estimates, one finds that a one-standard-deviation increase in ln(exports/area) is associated with between 0.36 to 0.62 standard deviation decrease in log income. If for purely illustrative purposes one interprets the OLS estimates as causal, then according to the estimate from column (5), for a country initially with the mean level of income of $1,249, a one-standard-deviation decrease in the slave export variable will raise income to $1,864, which is a 50% increase in income. V. ECONOMETRIC ISSUES: CAUSALITY AND MEASUREMENT ERROR Although the OLS estimates show that there is a relationship between slave exports and current economic performance, it group of countries. The table also shows that the estimates are robust to the omission of potentially influential observations. 10. Looking at Figure III, one can see that Equatorial Guinea (GNQ), which has a low value of slave exports and has recently witnessed extremely rapid economic growth because of the discovery of large offshore oil reserves, is a potential outlier. As I show in the Appendix, the results are also robust to the omission of Equatorial Guinea from the sample. 11. The island and North Africa indicator variables drop out of the equation when island and North African countries are omitted from the sample. THE LONG-TERM EFFECTS OF AFRICA’S SLAVE TRADES 157 remains unclear whether the slave trades have a causal impact on current income. An alternative explanation for the relationship is that societies that were initially underdeveloped selected into the slave trades, and these societies continue to be underdeveloped today. Therefore, we observe a negative relationship between slave exports and current income, even though the slave trades did not have any effect on subsequent economic development. In this section, I pursue two strategies to evaluate whether there is a causal effect of the slave trades on income. First, using historic data and qualitative evidence from African historians, I evaluate the importance and characteristics of selection into the slave trades. As I will show, the evidence suggests that selection was important, but it was usually the societies that were the most prosperous, not the most underdeveloped, that selected into the slave trades. Given this evidence, it is unlikely that the strong relationship between slave exports and current income is driven by selection. Instead, selection will tend to bias the OLS estimates towards zero. Second, I use the distance from each country to the location of the demand for slaves as instruments for slave exports. The IV (instrumental variables) results confirm the OLS estimates. V.A. Historical Evidence on Selection during the Slave Trades A large proportion of the early trade between Africans and Europeans was in commodities other than slaves. During this time, only societies with institutions that were sufficiently developed were able to facilitate trade with the Europeans. Between 1472 and 1483, the Portuguese sailed south along the west coast of West-Central Africa, testing various points of entry, looking for trading partners. They were unable to find any societies north of the Zaire river that could support trade. Vansina (1990, p. 200) writes that “the local coastal societies were just too small in terms of people and territory; their economic and social institutions were too undifferentiated to facilitate foreign trade.” Sustained trade did not occur until the Portuguese found the Kongo kingdom, located just south of the Zaire river. Because the Kongo kingdom had a centralized government, a national currency, and well-developed markets and trading networks, it was able to support trade with the Europeans. When European demand turned almost exclusively to slaves, the preference to trade with the most developed parts of Africa continued. Because the more prosperous areas were also the most densely populated, large numbers of slaves could be efficiently 158 QUARTERLY JOURNAL OF ECONOMICS FIGURE IV Relationship between Initial Population Density and Slave Exports obtained if civil wars or conflicts could be instigated (Barry 1992; Inikori 2003). As well, societies that were the most violent and hostile, and therefore the least developed, were often best able to resist European efforts to purchase slaves. For example, the slave trade in Gabon was limited because of the defiance and violence of its inhabitants toward the Portuguese. This resistance continued for centuries, and as a result the Portuguese were forced to concentrate their efforts along the coast further south (Hall 2005, pp. 60–64). Using data on initial population densities, I check whether it was the more prosperous or less prosperous areas that selected into the slave trades. Acemoglu, Johnson, and Robinson (2002) have shown that population density is a reasonable indicator of economic prosperity. Figure IV shows the relationship between the natural log of population density in 1400 and ln(exports/area). The data confirm the historical evidence on selection during the slave trades.12 The figure shows that the parts of Africa that were 12. The relationship is similar if one excludes island and North African countries, or if one normalizes slave exports by population rather than land area. THE LONG-TERM EFFECTS OF AFRICA’S SLAVE TRADES 159 the most prosperous in 1400, measured by population density, tend also to be the areas that were most impacted by the slave trades. A second potential source of selection may be that societies that initially had domestic slavery may have selected into the slave trades. If this is the case, then the estimates may be biased by a negative relationship between domestic slavery and subsequent economic development, similar to that documented by Engerman and Sokoloff (1997, 2002) in the Americas. The historic evidence indicates that in the areas of Africa that were part of the older Islamic slave trades there was domestic slavery, but it is unclear whether domestic slavery was a cause or a consequence of the external slave trades. Whether the parts of Africa that were untouched by the Islamic trades had chattel slavery prior to European contact has been the subject of an old debate among African historians (e.g., Fage [1962]; Rodney [1970]). Since this debate, evidence has been brought forth suggesting that domestic slavery may not have existed prior to the trans-Atlantic slave trade. Hilton (1985) provides evidence showing that in the sixteenth century, words that originally meant “servant” or “prisoner” were altered to take on the meaning of a “traded slave.” Historian and Anthropologist Jan Vansina (1989), using more detailed linguistic data, confirms Hilton’s finding, showing that in West Central Africa there was no word for slave. Vansina maps the origin of the word “pika,” which originally meant servant, but took on the meaning of “traded slave.” The word originated at the coastal ports engaged in the slave trade and spread to the inland communities that were also involved in the trade (Vansina 1989, 1990). Recent studies of other regions also suggest that prior to the external slave trade domestic slavery did not exist (e.g., Harms [1981]; Inikori [2000]; Hall [2005, p. 16]). V.B. Instrumental Variables The second strategy that I pursue is to use instruments that are correlated with slave exports, but are uncorrelated with other country characteristics. This strategy has the added benefit of yielding potentially consistent estimates even though slave exports are measured with error. If the instruments are also uncorrelated with the measurement error in slave exports that arises from the undersampling of slaves from the interior, then unlike OLS, IV yields consistent estimates. 160 QUARTERLY JOURNAL OF ECONOMICS As instruments for slave exports, I use the distances from each African country to the locations where slaves were demanded. The validity of the instruments relies on the presumption that although the location of demand influenced the location of supply, the location of supply did not influence the location of demand. If sugar plantations were established in the West Indies because the West Indies were close to the western coast of Africa, then the instruments are not valid. However, if instead many slaves were taken from western Africa because it was relatively close to the plantation economies in the West Indies, then the instruments are potentially valid. According to the known history of the slave trades, it was the location of demand that influenced the location of supply and not vice versa. The location of the demand for African slaves was determined by a number of factors, all unrelated to the supply of slaves. In the West Indies and the southern United States, slaves were imported because of climates suitable for growing highly valued, globally traded commodities such as sugar and tobacco. The existence of gold and silver mines was a determinant of the demand for slaves in Brazil. In the northern Sahara, Arabia, and Persia, slaves were needed to work in salt mines, and in the Red Sea area slaves were used as pearl divers. The instruments measure the distance from each country to the most important destinations in each of the slave trades. The four instruments are 1. The sailing distance from the point on the coast that is closest to the country’s centroid to the closest major market of the Atlantic slave trade. I use the nine largest importers of slaves, which are Virginia, USA; Havana, Cuba; Haiti; Kingston, Jamaica; Dominica; Martinique; Guyana; Salvador, Brazil; and Rio de Janeiro, Brazil.13 2. The sailing distance from the point on the coast that is closest to the country’s centroid to the closest of the two major slave destinations of the Indian Ocean slave trade: Mauritius and Muscat, Oman. 3. The overland distance from a country’s centroid to the closest port of export for the trans-Saharan slave trade. The markets are Algiers, Tunis, Tripoli, Benghazi, and Cairo. 13. Data on slave imports are from Eltis and Richardson (2006). There is a significant drop in the volume of slave imports between the ninth and tenth largest markets. Because of this natural break, I use the top nine markets. THE LONG-TERM EFFECTS OF AFRICA’S SLAVE TRADES 161 FIGURE V Example Showing the Distance Instruments for Burkina Faso 4. The overland distance from a country’s centroid to the closest port of export for the Red Sea slave trade. The ports are Massawa, Suakin, and Djibouti.14 The instruments are illustrated in Figure V, which shows the four distances for Burkina Faso. The ports in each of the four slave trades are represented by different colored symbols, and the shortest distances by colored lines. Details of the construction of the instruments are given in the Appendix.15 The IV estimates are reported in Table IV. The first column reports estimates without control variables, the second column includes colonizer fixed effects, and the third and fourth columns include colonizer fixed effects and geography controls. In column (4), the sample excludes islands and North African countries. 14. For island countries, one cannot reach the ports of the Saharan or Red Sea slave trades by traveling overland. For these countries I use the sum of the sailing distance and overland distance. 15. An alternative strategy is to also include the distance from the centroid to the coast (which is also shown in Figure V) as an additional instrument, since this distance is part of the total distance to the markets in the Indian Ocean and trans-Atlantic slave trades. The results are essentially identical if this distance is also included as an additional instrument. 162 QUARTERLY JOURNAL OF ECONOMICS TABLE IV ESTIMATES OF THE RELATIONSHIP BETWEEN SLAVE EXPORTS AND INCOME (1) (2) (3) (4) Second Stage. Dependent variable is log income in 2000, ln y ln(exports/area) −0.208∗∗∗ −0.201∗∗∗ −0.286∗ −0.248∗∗∗ (0.053) (0.047) (0.153) (0.071) [−0.51, −0.14] [−0.42, −0.13] [−∞, +∞] [−0.62, −0.12] Colonizer fixed No Yes Yes Yes effects Geography controls No No Yes Yes Restricted sample No No No Yes F-stat 15.4 4.32 1.73 2.17 Number of obs. 52 52 52 42 First Stage. Dependent variable is slave exports, ln(exports/area) Atlantic distance −1.31∗∗∗ −1.74∗∗∗ −1.32∗ −1.69∗∗ (0.357) (0.425) (0.761) (0.680) Indian distance −1.10∗∗∗ −1.43∗∗∗ −1.08 −1.57∗ (0.380) (0.531) (0.697) (0.801) Saharan distance −2.43∗∗∗ −3.00∗∗∗ −1.14 −4.08∗∗ (0.823) (1.05) (1.59) (1.55) Red Sea distance −0.002 −0.152 −1.22 2.13 (0.710) (0.813) (1.82) (2.40) F-stat 4.55 2.38 1.82 4.01 Colonizer fixed No Yes Yes Yes effects Geography controls No No Yes Yes Restricted sample No No No Yes Hausman test .02 .01 .02 .04 (p-value) Sargan test (p-value) .18 .30 .65 .51 Notes. IV estimates of (1) are reported. Slave exports ln(exports/area) is the natural log of the total number of slaves exported from each country between 1400 and 1900 in the four slave trades normalized by land area. The colonizer fixed effects are indicator variables for the identity of the colonizer at the time of independence. Coefficients are reported, with standard errors in brackets. For the endogenous variable ln(exports/area), I also report 95% confidence regions based on Moreira’s (2003) conditional likelihood ratio (CLR) approach. These are reported in square brackets. The p-value of the Hausman test is for the Wu–Hausman chi-squared test. ∗∗∗, ∗∗, and ∗ indicate significance at the 1%, 5%, and 10% levels. The “restricted sample” excludes island and North African countries. The “geography controls” are distance from equator, longitude, lowest monthly rainfall, avg max humidity, avg min temperature, and ln(coastline/area). The first-stage estimates are reported in the bottom panel of the table. The coefficients for the instruments are generally negative, suggesting that the further a country was from slave markets, the fewer slaves it exported.16 The exception is the distance 16. The specifications assume a linear first-stage relationship. The estimates are similar if one also allows for a nonlinear relationship between slave exports THE LONG-TERM EFFECTS OF AFRICA’S SLAVE TRADES 163 from the Red Sea ports, which is never significant, and is even positive in one specification. The second-stage estimates are reported in the top panel. Because the first stage F-statistics are low, I also report conditional likelihood ratio (CLR) confidence intervals. The estimates for ln(exports/area) are all negative and statistically significant. In column (3), the confidence interval is unbounded, which is a consequence of the low first-stage F-statistic. The point estimates range from −0.20 to −0.29. These magnitude are significantly larger than the magnitudes of the OLS estimates. This is not surprising, because the measurement error in the slave export estimates and the selection of the initially most prosperous societies into the slave trades are both expected to bias the OLS estimates towards zero. A potential concern with the Table IV estimates is that the distances to the slave markets may be correlated with distances to other locations that are important for economic development. The likelihood of this can be assessed by estimating the reduced form relationship between the distance instruments and income within Africa and outside of Africa.17 I find that within Africa, the four distance instruments are positively correlated with income, and all coefficients, except for the Red Sea coefficient, are highly significant. Being further from slave markets was good for growth. However, outside of Africa, there is no clear relationship between the distance instruments and income. If the results of Table IV are driven by the relationship between the distances to slave markets and distances to other locations, then one would expect to also observe a positive relationship between the distance measures and income outside of Africa. However, this is not the case. Overall, the IV results confirm the negative relationship between slave exports and income estimated by OLS. They also suggest that the OLS estimates may even be a lower bound estimate of the strength of the slave trade–income relationship. VI. POSSIBLE CHANNELS OF CAUSALITY I now turn to the channels through which the slave trades may have affected economic development. I view this analysis as and the distance instruments. The results are also similar if one uses the average or median distance to the ports in each trade, rather than the minimum distance. 17. I thank Ted Miguel for suggesting this check. 164 QUARTERLY JOURNAL OF ECONOMICS FIGURE VI Relationship between Slave Exports and Current Ethnic Fractionalization preliminary and exploratory. With only 52 observations it is not possible to pin down the precise channels and mechanism underlying the relationships with any reasonable degree of certainty. My strategy here is to simply investigate whether the data are consistent with the historic events described in Section II. An important consequence of the slave trades was that they tended to weaken ties between villages, thus discouraging the formation of larger communities and broader ethnic identities. I explore whether the data are consistent with this channel by examining the relationship between slave exports and a measure of current ethnic fractionalization from Alesina et al. (2003). As shown in Figure VI, there is a strong positive relationship between the two variables.18 This is consistent with the historic accounts of the slave trades impeding the formation of broader ethnic identities. This consequence of the slave trades is important because of the increasing evidence showing that ethnic fractionalization is an 18. The results are also similar if other measures of ethnic fractionalization are used. THE LONG-TERM EFFECTS OF AFRICA’S SLAVE TRADES 165 important determinant of a variety of factors necessary for economic development. Since the seminal article documenting the link between ethnic diversity and economic growth by Easterly and Levine (1997), subsequent research by La Porta et al. (1999), Alesina et al. (2003), Aghion, Alesina, and Trebbi (2004), and Easterly, Ritzan, and Woolcock (2006) looks more deeply into why ethnic fractionalization is important for development. These studies find that ethnic diversity is important for social cohesion, domestic institutions, domestic polices, and the quality of government. As well, Alesina, Baquir, and Easterly (1999), Miguel and Gugerty (2005), and Banerjee and Somanathan (2006) find that ethnic fractionalization reduces the provision of public goods, such as education, health facilities, access to water, and transportation infrastructure, all of which are important for economic development. A second, and closely related, consequence of the slave trades was the weakening and underdevelopment of states. To examine whether the data are consistent with this channel, I consider the relationship between slave exports and the level of state development following the slave trades. To do this I use a measure of precolonial state development from Gennaioli and Rainer (2006). The measure is constructed using ethnographic data from Murdock (1967) on the indigenous political complexity of ethnic groups, measured by the number of jurisdictional hierarchies beyond the local community. The original measure ranges from 0 to 4, with 0 indicating “stateless” societies and 4 indicating societies with “large states” (Murdock 1967, p. 52). Using this data, Gennaioli and Rainer (2006) construct a measure of the proportion of a country’s indigenous population that belongs to an ethnic group that falls into category 2, 3, or 4. The relationship between slave exports and nineteenthcentury state development is shown in Figure VII. The negative relationship between slave exports and state centralization shown in the figure is consistent with the historic accounts of the slave trades causing long-term political instability, which resulted in weakened and fragmented states. Recent empirical research shows that a country’s history of state development is an important determinant of current economic performance. Bockstette, Chanda, and Putterman (2002) and Chanda and Putterman (2005) find that “state antiquity,” measured using an index of the depth of experience with statelevel institutions, is positively correlated with real per capita GDP 166 QUARTERLY JOURNAL OF ECONOMICS FIGURE VII Relationship between Slave Exports and Nineteenth-Century State Development growth between 1960 and 1995. Looking within Africa, Gennaioli and Rainer (2006) find that countries with ethnicities that had centralized precolonial state institutions today provide more public goods, such as education, health, and infrastructure. Herbst (1997, 2000) also focuses on the importance of state development for economic success, arguing that Africa’s poor economic performance is a result of postcolonial state failure, the roots of which lie in the underdevelopment and instability of precolonial polities. Herbst (2000, chaps. 2–4) argues that because of a lack of significant political development during colonial rule, the limited precolonial political structures continued to exist after independence.19 As a result, Africa’s postindependence leaders inherited nation states that did not have the infrastructure necessary to extend authority and control over the whole country. Many states were, and still are, unable to collect taxes from their citizens, and as a result they are also unable to provide a minimum level of public goods and services. 19. On the continuity between Africa’s precolonial and postcolonial political systems also see Hargreaves (1969, p. 200). THE LONG-TERM EFFECTS OF AFRICA’S SLAVE TRADES 167 FIGURE VIII Paths of Economic Development Since 1950 A corollary of Herbst’s argument is that the impact of the slave trades may have been felt most strongly after colonial independence. This is because this is when precolonial political structures suddenly increased in importance, as they became central determinants of the success of the newly formed state. Using Figure VIII, I examine whether the evolution of incomes since 1950 is consistent with this hypothesis. The figure shows average per capita GDP between 1950 and 2000 for two groups of African countries.20 One group consists of the 26 countries with the lowest measures of ln(exports/area), and the other is the 26 countries with the highest measures of ln(exports/area). As shown in the figure, throughout the period low-slave-export countries are richer on average than high slave export countries. Also interesting, however, is the difference in the evolution of income between the two groups of countries. Although the low-slave-export countries were richer in the early 1950s when most countries were still under colonial rule, the income gap between the two groups increased 20. The averages are weighted by each country’s population in 2000. 168 QUARTERLY JOURNAL OF ECONOMICS significantly over time and became most pronounced after the late 1960s and early 1970s, when most countries had gained indepen- dence.21 This pattern is consistent with the slave trades affecting early state development, which may have mattered during colonial rule but mattered much more after independence. Because those parts of Africa that were most severely impacted by the slave trades tended to have the least developed political systems, after independence these countries continued to have weak and unstable states, as well as slower economic growth. VII. CONCLUSIONS Combining data from shipping records and data from historical documents reporting slave ethnicities, I have constructed estimates of the number of slaves exported from each country in Africa during Africa’s four slave trades. I found a robust negative relationship between the number of slaves taken from a country and its subsequent economic development. I pursued a number of strategies to better understand if the relationship is causal or spurious. If countries that were initially underdeveloped selected into the slave trades, and if these countries continue to be underdeveloped today, then this may explain the observed relationship between slave exports and current income. I first reviewed the historical evidence on the characteristics of African societies that were most affected by the slave trades. The qualitative and quantitative evidence show that it was actually the most developed parts of Africa, not the least developed, that tended to select into the slave trades. I also used the distances from each country to the locations of the demand for slaves as instruments to estimate the causal effect of the slave trades on economic development. The IV estimates confirmed the OLS results, suggesting that increased extraction during the slave trades resulted in worse economic performance. I then examined the channels of causality underlying the relationship between slave exports and economic development. I showed that the data are consistent with historic accounts suggesting that the slave trades impeded the formation of broader ethnic groups, leading to ethnic fractionalization, and that the 21. In 1950 only four African countries were independent. Ethiopia had never been colonized, and Liberia, South Africa, and Egypt had previously gained independence. By 1969, 42 of Africa’s 52 countries had gained independence, and by 1976, 48 countries were independent. THE LONG-TERM EFFECTS OF AFRICA’S SLAVE TRADES 169 slave trades resulted in a weakening and underdevelopment of political structures. APPENDIX A. Deriving the Bias from the Undersampling of Slaves from the Interior In this section I show that the undersampling of slaves from the interior of Africa will result in OLS estimates of the effect of slave exports on income that are biased toward zero. To see this, denote the true number of slaves taken from country i by s∗ i , the observed number of slaves by si, distance to the coast by di, and economic development by yi. All variables are expressed as deviations from means. Assume the true relationship between the number of slaves exported and distance to the coast is given by (2) s∗ i = −αdi + εi, where α > 0 and εi is i.i.d. drawn from a normal distribution. The relationship between the observed number of slaves exported, si, and the distance to the coast, di, is given by (3) si = s∗ i − γ di + νi, where γ > 0 and νi is uncorrelated with εi. The true relationship between slave exports and development is given by (4) yi = −βs∗ i + ωi, where β > 0 and ωi is uncorrelated with all other variables. If one estimates yi = bsi + ξi by OLS, then the estimated relationship between si and yi is (5) ˆb = i si yi i s2 i . Substituting (2) into (3) gives (6) si = −(α + γ )di + εi + νi. Similarly, (2) and (4) give (7) yi = βαdi − βεi + ωi. 170 QUARTERLY JOURNAL OF ECONOMICS Substituting (6) and (7) into (5) and taking the plim of ˆb gives (8) plim ˆb = −β σ2 s∗ + γ ασ2 d σ2 s∗ + 2γ (α + γ )σ2 d + σ2 ν , where σ2 s∗ = α2 σ2 d + σ2 ε . First, consider the case when the only source of measurement error is classical measurement error. Then γ = 0 and (8) reduces to the standard formula for attenuation bias: plim ˆb = −β[σ2 s∗ / (σ2 s∗ + σ2 ν )]. Next, consider the measurement error introduced by the undersampling of slaves from the interior. The result of this is that the underestimation of slave exports is increasing in a country’s distance from the coast: γ > 0. Looking at (8), it is apparent that 2γ (α + γ ) > γ α, and therefore the presence of nonclassical measurement error also biases the estimated coefficient toward zero, reinforcing the attenuation bias resulting from classical errors in variables. B. Data Real per capita GDP data are from Maddison (2003). Land area, which is used to calculate ln(export/area), ln(coastline/area), and population density in 1400, is measured in millions of square kilometers, and is from Parker (1997). Historic population figures, measured in thousands of people, are from McEvedy and Jones (1978). For some groups of smaller countries, population data are only disaggregated to a regional level. In these cases the data are disaggregated to the country level using the distribution of population in 1950 from the United Nations. Data on the identity of the colonizer before independence are from the Political Regimes and Regime Transitions in Africa, 1910–1994 data set, which is described in Bratton and van de Walle (1997). Distance from the equator is the absolute value of the latitude of each country’s centroid, measured in degrees. Longitude is the longitude of each country’s centroid, also measured in degrees. The centroid of each country is calculated using the Centroid Utility in ArcGIS. For five countries where the centroid falls outside the land borders of the country (Gambia, Somalia, Cape Verde, Mauritius, and Seychelles) the point within the country closest to the centroid is used. The location on the coast that is closest to each country’s centroid is identified using the Proximity Utility in ArcGIS. THE LONG-TERM EFFECTS OF AFRICA’S SLAVE TRADES 171 TABLE A.1 SUMMARY STATISTICS Variable Mean Std. dev. Min Max N ln real per capita GDP in 2000 7.13 0.83 5.38 9.27 52 ln(exports/area) 3.26 3.89 −2.30 8.82 52 ln(exports/pop) 9.26 3.68 3.91 14.4 52 Distance from equator 13.6 9.86 0.2 36 52 Longitude 16.7 20.2 −24.0 57.8 52 Lowest monthly rainfall 8.87 16.1 0 69 52 Avg max humidity 71.7 11.9 35 95 52 Avg min temperature 8.75 7.49 −9.0 19 52 ln(coastline/area) −0.24 3.24 −4.61 6.98 52 Island indicator 0.10 0.30 0 1 52 Percent Islamic 35.3 39.1 0 100 52 French legal origin 0.65 0.48 0 1 52 North Africa indicator 0.10 0.30 0 1 52 ln(gold prod/pop) −7.48 5.66 −13.8 3.08 52 ln(oil prod/pop) −6.71 4.03 −9.21 3.24 52 ln(diamond prod/pop) −5.49 2.40 −6.91 2.19 52 Atlantic distance 7.38 3.28 3.64 16.40 52 Indian distance 6.93 4.24 0.03 16.78 52 Saharan distance 3.51 1.57 0.31 6.64 52 Red Sea distance 3.44 1.47 0.06 6.47 52 TABLE A.2 ROBUSTNESS AND SENSITIVITY CHECKS Specification Coef. Std. Err. N R2 Normalizing slave exports by average −0.103∗∗∗ (0.035) 52 0.77 population from 1400 to 1900 Omitting zero slave export countries −0.104∗∗ (0.041) 41 0.84 Omitting N. Africa, islands, −0.140∗∗∗ (0.040) 38 0.70 GNQ, LSO, SWZ, and ZAF Including five region fixed effects −0.099∗∗ (0.036) 52 0.80 Omitting influential observations −0.091∗∗∗ (0.031) 42 0.90 Notes. The table reports OLS estimates of (1), with the full set of control variables from Table III included. The dependent variable is the natural log of real per capita GDP in 2000, ln y. Each row of the table reports estimates from one regression. ∗∗∗, ∗∗, and ∗ indicate significance at the 1%, 5%, and 10% levels. The “five region fixed effects” in row 4 are North, West, Central, East, and South Africa. In row 5, influential observations were omitted if Cook’s distance was greater than 4/N, where N is the number of observations. Lowest monthly rainfall is the average total rainfall, measured in millimeters, in the driest month of the year. Average maximum humidity is the average of the maximum afternoon humidity, measured in percent, during the hottest month of the year. Average minimum temperature is the lowest average monthly 172 QUARTERLY JOURNAL OF ECONOMICS temperature measured in degrees Celsius. The data are from meteorological data taken over a 30-year period and reported in Parker (1997). Countries’ total coastline, which is used along with land area to calculate ln(coastline/area), is measured in thousands of kilometers and is from Parker (1997). The percent Islamic variable is the percentage of a country’s population that is Islamic. The data are from Parker (1997). Data on countries’ legal origins are from La Porta et al. (1999). All countries in the sample are coded as being either British common law or French civil law countries. Data on the production of diamonds, crude petroleum, and mined gold are from the British Geological Survey’s World Mineral Statistics and World Mineral Production. All three variables are measured as the natural log of the average annual production per thousand inhabitants from 1970 to 2000. Diamonds include both gemstones and industrial diamonds and are measured in thousands of carats. Crude petroleum is measured in thousands of tonnes, and mined gold is measured in kilograms. Ethnic fractionalization is from Alesina et al. (2003). The measure of nineteenth-century state development is from Gennaioli and Rainer (2006). When taking the natural log of variables that may take on the value of zero, I replace the zero observations with 1 × 10n , where n is the largest integer value possible subject to 1 × 10n being less than the smallest nonzero observation in the data. The distance instruments measure the shortest sailing distances to the locations of demand in the trans-Atlantic and Indian Ocean slave trades and the shortest overland distances to the locations of demand in the Red Sea and trans-Saharan slave trades. The distances are calculated using the great circle distance between two locations. The formula for this is dij = (arccos{sin(Lai) sin(Laj) + cos(Lai) cos(Laj) cos(Loi − Loj)}× 111.12)/1000, where dij is the distance in thousands of kilometers between location i and j, Lai is the latitude of location i in degrees, and Loi is the longitude of location i in degrees. When calculating the sailing distances, I do not allow ships to sail across land or through the Suez Canal, which was not completed until 1869. For voyages from northern Africa in the trans-Atlantic slave trade, I calculate the sailing distance through the Strait of Gibraltar to the closest market in the Atlantic Ocean. For the distance from these countries in the Indian Ocean slave trade, I calculate the sailing distance through the Mediterranean Sea, and then south THE LONG-TERM EFFECTS OF AFRICA’S SLAVE TRADES 173 around the Cape of Good Hope. When calculating distances from East African countries during the trans-Atlantic slave trade, and from West African countries during the Indian Ocean slave trade, I calculate the sailing distance around the Cape of Good Hope. HARVARD UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH REFERENCES Acemoglu, Daron, Simon Johnson, and James A. Robinson, “The Colonial Origins of Comparative Development: An Empirical Investigation,” American Economic Review, 91 (2001), 1369–1401. ——, “Reversal of Fortune: Geography and Institutions in the Making of the Modern World Income Distribution,” Quarterly Journal of Economics, 117 (2002), 1231–1294. Aghion, Philippe, Alberto Alesina, and Francesco Trebbi, “Endogenous Political Institutions,” Quarterly Journal of Economics, 119 (2004), 565–613. Alesina, Alberto, Reza Baquir, and William Easterly, “Public Goods and Ethnic Divisions,” Quarterly Journal of Economics, 114 (1999), 1243–1284. Alesina, Alberto, Arnaud Devleeschauwer, William Easterly, Sergio Kurlat, and Romain Wacziarg, “Fractionalization,” Journal of Economic Growth, 8 (2003), 155–194. Alpers, Edward A., “Trade, State, and Society among the Yao in the Nineteenth Century,” The Journal of African History, 10 (1969), 405–420. ——, Ivory and Slaves in East Central Africa (London: Heinemann International, 1975). Austen, Ralph A., “The Trans-Saharan Slave Trade: A Tentative Census,” in The Uncommon Market: Essays in the Economic History of the Atlantic Slave Trade, Henry A. Gemery and Jan S. Hogendorn, eds. (New York: Academic Press, 1979). ——, “The 19th Century Islamic Slave Trade from East Africa (Swahili and Red Sea Coasts): A Tentative Census,” Slavery & Abolition, 9 (1988), 21–44. ——, “The Mediterranean Islamic Slave Trade out of Africa: A Tentative Census,” Slavery & Abolition, 13 (1992), 214–248. Azevedo, Mario, “Power and Slavery in Central Africa: Chad (1890–1925),” Journal of Negro History, 67 (1982), 198–211. Bairoch, Paul, Economics and World History: Myths and Paradoxes (Chicago: University of Chicago Press, 1993). Banerjee, Abhijit and Rohini Somanathan, “The Political Economy of Public Goods: Some Evidence from India,” mimeo, 2006. Barry, Boubacar, “Senegambia from the Sixteenth to the Eighteenth Century: Evolution of the Wolof, Sereer, and ‘Tukuloor,”’ in General History of Africa: Volume 5, Africa from the Sixteenth to the Eighteenth Century, B.A. Ogot, ed. (Berkeley: University of California Press, 1992). ——, Senegambia and the Atlantic Slave Trade (Cambridge, UK: Cambridge University Press, 1998). Bertocchi, Graziella, and Fabio Canova, “Did Colonization Matter for Growth? An Empirical Exploration into the Historical Causes of Africa’s Underdevelopment,” European Economic Review, 46 (2002), 1851–1871. Bockstette, Valeri, Areendam Chanda, and Louis Putterman, “States and Markets: The Advantage of an Early Start,” Journal of Economic Growth, 7 (2002), 347– 369. Bratton, Michael, and Nicolas van de Walle, “Political Regimes and Regime Transitions in Africa, 1910–1994,” Technical Report 6996, ICPSR, 1997. Chanda, Areendam, and Louis Putterman, “State Effectiveness, Economic Growth, and the Age of States,” in States and Development: Historical Antecedents of Stagnation and Advance, Matthew Lange and Dietrich Rueschemeyer, eds. (New York: Palgrave MacMillan, 2005). 174 QUARTERLY JOURNAL OF ECONOMICS Colson, Elizabeth, “African Society at the Time of the Scramble,” in Colonialism in Africa, 1870–1960. Volume 1: The History and Politics of Colonialism, 1870–1914, L.H. Gann and Peter Duignan, eds. (Cambridge UK: Cambridge University Press, 1969). Curtin, Philip D., The Atlantic Slave Trade: A Census (Madison: The University of Wisconsin Press, 1969). Easterly, William, and Ross Levine, “Africa’s Growth Tragedy: Policies and Ethnic Divisions,” Quarterly Journal of Economics, 112 (1997), 1203–1250. Easterly, William, Jozef Ritzan, and Michael Woolcock, “Social Cohesion, Institutions, and Growth,” Working Paper 94, Center for Global Development, 2006. Elbl, Ivana, “Volume of the Early Atlantic Slave Trade, 1450–1521,” Journal of African History, 38 (1997), 31–75. Eltis, David, Stephen D. Behrendt, David Richardson, and Herbert S. Klein, The Trans-Atlantic Slave Trade: A Database on CD-Rom (New York: Cambridge University Press, 1999). Eltis, David, and David Richardson, “Missing Pieces and the Larger Picture: Some Implications of the New Database,” mimeo, Emory University, 2006 Engerman, Stanley L., and Kenneth L. Sokoloff, “Factor Endowments, Institutions, and Differential Paths of Growth Among New World Economies: A View from Economic Historians of the United States,” in How Latin America Fell Behind, Stephen Harber, ed. (Stanford, CA: Stanford University Press, 1997). ——, “Factor Endowments, Inequality, and Paths of Development Among New World Economies,” Working Paper 9259, National Bureau of Economic Research, 2002. Englebert, Pierre, “Pre-Colonial Institutions, Post-Colonial States, and Economic Development in Tropical Africa,” Political Research Quarterly, 53 (2000a), 7– 36. ——, “Solving the Mystery of the Africa Dummy,” World Development, 28 (2000b), 1821–1835. Fage, John D., An Introduction to the History of West Africa, 3rd ed. (Cambridge, UK: Cambridge University Press, 1962). Gennaioli, Nicola, and Ilia Rainer, “The Modern Impact of Precolonial Centralization in Africa,” mimeo, Stockholm University, 2006. Grier, Robin M., “Colonial Legacies and Economic Growth,” Public Choice, 98 (1999), 317–335. Hall, Gwendolyn Midlo, Slavery and African Ethnicities in the Americas: Restoring the Links (Chapel Hill, NC: University of North Carolina Press, 2005). Hargreaves, John D., “West African States and the European Conquest,” in Colonialism in Africa, 1870–1960, L.H. Gann and Peter Duignan, eds. (Cambridge, UK: Cambridge University Press, 1969). Harms, Robert, River of Wealth, River of Sorrow: The Central Zaire Basin in the Era of the Slave and Ivory Trade, 1500–1891 (New Haven, CT: Yale University Press, 1981). Hawthorne, Walter, “The Production of Slaves Where There Was No State: The Guinea-Bissau Region, 1450–1815,” Slavery & Abolition, 20 (1999), 97–124. ——, Planting Rice and Harvesting Slaves: Transformations along the GuineaBissau Coast, 1400–1900 (Portsmouth, NH: Heinemann, 2003). Herbst, Jeffrey, “Responding to State Failure in Africa,” International Security, 21 (1997), 120–144. ——, States and Power in Africa (Princeton, NJ: Princeton University Press, 2000). Higman, Barry W., Slave Populations of the British Caribbean, 1807–1834 (Baltimore, MD: Johns Hopkins University Press, 1984). Hilton, Anne, The Kingdom of Kongo (Oxford: Clarendon, 1985). Hubbell, Andrew, “A View of the Slave Trade from the Margin: Souroudougou in the Late Nineteenth-Century Slave Trade of the Niger Bend,” Journal of African History, 42 (2001), 25–47. Inikori, Joseph E., “Africa and the Trans-Atlantic Slave Trade,” in Africa Volume I: African History Before 1885, Toyin Falola, ed. (Durham, NC: Carolina Academic Press, 2000). ——, “The Struggle against the Trans-Atlantic Slave Trade,” in Fighting the Slave Trade: West African Strategies, A. Diouf, ed. (Athens, OH: Ohio University Press, 2003). THE LONG-TERM EFFECTS OF AFRICA’S SLAVE TRADES 175 Isaacman, Allen F., “The Countries of the Zambezi Basin,” in General History of Africa, VI, J. F. A. Ajayi, ed. (Paris: Heinemann International, 1989). Kamarck, Andrew M., The Tropics and Economic Development (Johns Hopkins University Press, Baltimore, MD, 1976). Karasch, Mary C., Slave Life in Rio de Janeiro (Princeton University Press, Princeton, NJ, 1987). Kimambo, I. N., “The East African Coast and Hinterland, 1845–1880,” in General History of Africa, VI, J.F.A. Ajayi, ed. (Heinneman International, Paris, 1989), 234–269. Klein, Martin, “The Slave Trade and Decentralized Societies,” Journal of African History, 42 (2001), 49–65. Koelle, Sigismund Wilhelm, Polyglotta Africana; or A Comparative Vocabulary of Nearly Three Hundred Words and Phrases, in More than One Hundred Distinct African Languages (Church Missionary House, London, 1854). Kusimba, Chapurukha M., “Archaeology of Slavery in East Africa,” African Archaeological Review, 21 (2004), 59–88. La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert W. Vishny, “The Quality of Government,” Journal of Law, Economics, and Organization, 15 (1999), 222–279. Lagerl¨of, Nils-Petter, “Geography, Institutions and Growth: The United States as a Microcosm,” (2005), mimeo, York University. Lange, Matthew K., “British Colonial Legacies and Political Development,” World Development, 32 (2004), 905–922. Lovejoy, Paul E., “Background to Rebellion: The Origins of Muslim Slaves in Bahia,” Slavery & Abolition, 15 (1994), 151–180. ——, Transformations in Slavery: A History of Slavery in Africa, Second Edition (Cambridge University Press, Cambridge, UK, 2000). Maddison, Angus, The World Economy: Historical Statistics (Organisation for Economic Co-operation and Development, Paris, 2003). Mahadi, Abdullahi, “The Aftermath of the Jih¯ad in the Central Sudan as a Major Factor in the Volume of the Trans-Saharan Slave Trade in the Nineteenth Century,” in The Uncommon Market: Essays in the Economic History of the Atlantic Slave Trade, Elizabeth Savage, ed. (Frank Cass, London, 1992), 111– 128. Manning, Patrick, “Contours of Slavery and Social Change in Africa,” American Historical Review, 88 (1983), 835–857. ——, Slavery and African Life (Cambridge University Press, Cambridge, UK, 1990). Mbajedwe, Patrick U., “Africa and the Trans-Atlantic Slave Trade,” in Africa Volume I: African History Before 1885, Toyin Falola, ed. (Durham, NC: Carolina Academic Press, 2000), 335–358. McEvedy, Colin, and Richard Jones, Atlas of World Population History (Harmondsworth: Penguin Books, 1978). Miguel, Edward, and Mary Kay Gugerty, “Ethnic Diversity, Social Sanctions, and Public Goods in Kenya,” Journal of Public Economics, 89 (2005), 2325– 2368. Mitchener, Kris James, and Ian W. McLean, “The Productivity of U.S. States since 1880,” Journal of Economic Growth, 8 (2003), 73–114. Moreira, Marcelo J., “A Conditional Likelihood Ratio Test for Structural Models,” Econometrica, 71 (2003), 1027–1048. Moreno Fraginals, Manuel, “Africa in Cuba: A Quantitative Analysis of the African Population in the Island of Cuba,” in Comparative Perspectives on Slavery in New World Plantation Societies, Vera Rubin and Arthur Truden, eds. (New York: New York Academy of Sciences, 1977). Murdock, George Peter, Africa: Its Peoples and Their Cultural History (New York: McGraw-Hill Book Company, 1959). ——, Ethnographic Atlas (Pittsburgh: University of Pittsburgh Press, 1967). Northrup, David, Trade without Rulers: Pre-colonial Economic Development in South-Eastern Nigeria (Oxford: Clarendon Press, 1978). Nunn, Nathan, Data Appendix for the Long-Term Effects of Africa’s Slave Trades,” mimeo, University of British Columbia, 2007. 176 QUARTERLY JOURNAL OF ECONOMICS Oldendorp, Christian Georg Andreas, C.G.A. Oldendorp’s History of the Mission of the Evangelical Brethren on the Caribbean Islands of St. Thomas, St. Croix, and St. John (1777) [Reprinted by Karoma Publishers Inc., in 1987, Ann Arbor]. Parker, Philip M., National Cultures of the World: A Statistical Reference (London: Greenwood Press, 1997). Rodney, Walter, A History of the Upper Guinea Coast (Oxford: Clarendon Press, 1970). Sachs, Jeffrey D., Andrew D. Mellinger, and John L. Gallup, “The Geography of Poverty and Wealth,” Scientific American, 284 (2001), 70–76. Tardieu, Jean-Pierre, “Origins of the Slaves in the Lima Region in Peru (Sixteenth and Seventeenth Centuries)” in From Chains to Bonds: The Slave Trade Revisited, Doudou Diene, ed. (New York: UNESCO Publishing, 2001), 43–45. Vansina, Jan, Kingdoms of the Savanna (Madison: University of Wisconsin Press, 1966). ——, “Deep-Down Time: Political Tradition in Central Africa,” History in Africa, 16 (1989), 341–362. ——, Paths in the Rainforests (Madison, WI: The University of Wisconsin Press, 1990). Wax, Darold D., “Preferences for Slaves in Colonial America,” Journal of Negro History, 58 (1973), 371–401. This article has been cited by: 1. S. Bhattacharyya. 2009. Root Causes of African Underdevelopment. Journal of African Economies 18:5, 745-780. [CrossRef] 2. William Easterly. 2009. Can the West Save Africa?. Journal of Economic Literature 47:2, 373-447. [CrossRef] 3. Laura N Beny, Lisa D Cook. 2009. Metals or Management? Explaining Africa's Recent Economic Growth Performance. American Economic Review 99:2, 268-274. [CrossRef] 4. James D Fearon, Macartan Humphreys, Jeremy M Weinstein. 2009. Can Development Aid Contribute to Social Cohesion after Civil War? Evidence from a Field Experiment in Post-Conflict Liberia. American Economic Review 99:2, 287-291. [CrossRef] 5. Gareth Austin. 2008. The ‘reversal of fortune’ thesis and the compression of history: Perspectives from African and comparative economic history. Journal of International Development 20:8, 996-1027. [CrossRef]