The Origins of Colonial Investments

Joan Ricart-Huguet∗ September 14, 2016

Abstract

Recent literature has documented the economic and political consequences of colonialism, but we know less about the origins of colonial investments. I present evidence that they were very unequally distributed within 16 French and British African colonies, even when adjusted for population. How did colonial powers allocate their investments? Eco- nomic history emphasizes the role of natural resources while recent literature in political economy argues that settlers explain the origin of colonial institutions. I argue that observable geographic features—locational fundamentals—led some locations to become centers of pre-colonial trade, which in turn increased colonial settlement and investments not only in infrastructure but also in health and education. Disparities did not diminish during the colonial period because those locations became centers of colonial economic activity and beneﬁted from complementarities between investments, consistent with a logic of increasing returns. JEL: F63, H50, N37, N57

∗Department of Politics, Princeton University, Princeton, NJ 08540. Email: [email protected]. I thank Carles Boix, Evan Lieberman, Marc Ratkovic and Leonard Wantchekon, Jennifer Widner, Tsering Wangyal Shawa, Bill Guthe, Torben Behmer, Brandon Miller de la Cuesta, Costantino Pischedda and seminar participants at the Contemporary African Political Economy Research Seminar (CAPERS), the Politics and History Network and Princeton University for useful comments. I am grateful to Elise Huillery and Bob Woodberry for sharing the relevant data. Jeremy Darrington and Elizabeth Bennett helped me locate multiple data sources. Seth Merkin Morokoﬀ, Luise Zhong and Bruno Schaﬀa provided excellent research assistance.

1 1 Introduction

The last two decades have seen a surge in the literature examining the economic and political consequences of colonialism, notably in the Americas, Africa and South Asia.1 Existing claims on the effects of colonial institutions, investments and practices are wide-ranging.2 Some of this literature uses arbitrary or “as-if” random variation in a particular aspect of colonial rule to identify those effects, such as spatial discontinuities in agricultural revenue collection systems in India (Banerjee and Iyer, 2005), in extractive forced labor in Peru (Dell, 2010), or in missionary school location (Wantchekon and Garc´ıa-Ponce, 2015). However, colonial investments are generally not random and they are important (i) to better understand the impact of colonialism on subsequent political and economic development, (ii) because they may be mechanisms linking deep pre-colonial explanations to current outcomes, and (iii) because they fundamentally affected welfare and development during colonial rule.

I present novel evidence showing that colonial investments were very unequally distributed across and within 16 British and French colonies in East and West Africa.3 This paper focuses on the large within-colony variation in investments among the over 300 districts in those 16 colonies. One important advantage of the within-colony focus is that it allows me to hold institutions constant, since they hardly varied across districts within the same colony. Investments in some districts were orders of magnitude larger than in others within the same colony, even taking population into account, as Figure 1 shows. This highly unequal distribution applies to each of the three main types of investments in both empires: education, infrastructure, and health (Figures 13-18 in the Appendix). Further, these inequalities persisted during the colonial period under study: districts receiving higher investments in the early 1900s continued to receive more decades later.

Why did Europeans invest so much more in some districts than in others? It is diﬃcult to answer this question beyond particular cases by examining colonial documents because these lack a systematic investment strategy (Section 3). Scholars have long argued and provided evidence for the role of natural resources. Curtin et al. (1995, p. 447) argue that “European capital was invested where exploitable resources promised the most extractive returns” (Huillery, 2010, p. 271), notably the natural resources of settler colonies such as South Africa and in extractive East and West African colonies such as Guinea and Ghana. It is no coincidence that Ghanaian gold and

1See Nunn(2009) for a concise review of the economic literature on the topic. 2For example, see Acemoglu et al. (2001) and Mahoney (2010) on the effects of colonial institutions; Huillery (2009), Ricart-Huguet (2015), and Cagéand Rueda (2016) on investments; and Guardado (2014) and Wilkinson (2015) on colonial practices. 3The eight French West African (AOF) colonies are Benin (formerly Dahomey), Burkina Faso (Upper Volta), Cote d’Ivoire, Guinea, Mali (French Soudan), Mauritania, Niger and Senegal. The eight British colonies are the main territories under the British Colonial Office (CO): Ghana (Gold Coast), Kenya, Malawi (Nyasaland), Nigeria, Tanzania (Tanganyika), Sierra Leone, Uganda and Zambia (Northern Rhodesia).

2 Figure 1: Public health and education by district (1910-1939)

Cote d'Ivoire Public health staff Public health staff per 100,000 people 40 50 40 30 30 20 20 10 10 0 0

Man Kong Man Daloa KongLahou DaloaGuigloLahou Tabou Guiglo Indenie Assinie Tabou Gouros Baoule Gouros Baoule AgnebyIndenie Assinie SeguelaOdienne Agneby Bassam Lagunes OdienneSeguela BassamLagunes Nzicomoe Nzicomoe Tagouanas Bondoukou Sassandra BondoukouTagouanas Sassandra Nigeria Students Students per 100,000 people 15,000 2,000 1,500 10,000 1,000 5,000 500 0 0

Oyo Oyo Niger Ilorin ZariaKano Warri IjebuOndo Kano Ilorin NigerZaria Warri Ijebu Benue KabbaOgojaBornu LagosBenin Owerri Benue Bornu Ogoja Kabba Owerri OndoBenin Lagos Plateau Katsina Bauchi Sokoto Onitsha Calabar SokotoBauchi Plateau Katsina Onitsha Calabar Adamawa Abeokuta Adamawa Abeokuta Cameroons Cameroons

Sierra Leonean diamonds ended up in British hands. “With a great sense of the practical, the British had for a long time been snapping up the best coastal sectors. [...] Britain took possession of the territories with the richest resources and best future, although without any geographical ties: the Niger delta, basis of future Nigeria; the Gambia estuary, Sierra Leone and the Gold Coast” (Chi-Bonnardel, 1973, p. 50). More recently, Wantchekon and Stanig (2015, p.5) show that “colonial infrastructure can be predicted [...] by the presence of extractive resources (mines and quarries) but not by soil quality. This is in line with conventional wisdom in economic history: colonial powers were not after farmland but minerals.”

Natural resources are a plausible explanation for colonial investments in infrastructure. How- ever, their role in non-extractive investments such as health and education is less clear. In fact, why did the colonial state invest in health and education in these extractive colonies at all? There are several reasons. First, the 1885 Berlin Conference declared the principle of Eﬀective Occupa- tion, by which claims to the territory depended on actual presence (Young, 1994, p. 100). These claims gain veracity if the conqueror builds facilities such as schools or inﬁrmaries. Second, public health investments have positive externalities that reduce the risk of contagious disease among

3 Africans and Europeans. Third, the French created a class of “´evolu´es”(evolved) just like the British created a class of “Europeanised”—two racist terms with similar meanings—who would become colonial civil servants thanks to their European education. Some attribute these expenditures partly to the “white man’s burden”—the presumed European responsibility to educate at least some among the colonized (Lugard (1922), Suret-Canale (1971)). A more prosaic reason was the need to reduce the colonial payroll and maximize revenue by having Africans rather than Europeans occupy junior civil service positions.

The most influential argument to explain variation in non-extractive colonial investments such as public health and education focuses on the size of settler communities: more settlers led to more inclusive colonial institutions (Acemoglu et al., 2001) and higher levels of district colonial investment (Huillery, 2010), both of which rely on partly exogenous variation in settlement size. The former article argues that European settler mortality rate was lower where disease environment was more favorable, while the latter argues that the number of settlers was higher in areas where local chiefs offered less resistance or were less hostile. Indeed, settlers not only had a voice in the colonial government but sometimes even acted as a lobby, affecting the decisions of the colonial administration (Gardner, 2012).

The settlers and natural resources explanations are not without problems, however. Settlers can hardly be a fundamental explanation for investments since most arrived in East and West Africa in the late 19th and 20th century, centuries after European exploration and conquest had begun. There were very few settlers pre-colonially, unlike in neo-Europes and in many Latin American colonies, and settlement was very low even in the 20th century, with all 16 colonies except Kenya populated by less than 10,000 settlers prior to 1940. Their late arrival means that colonial settlers inﬂuenced but also responded to levels of investments, with causality running both ways. Data show that a couple of districts in each colony attracted the lion’s share of Europeans; an important question is why Europeans settled where they did within each colony.

The natural resources explanation also has limitations. Natural resources sometimes drove initial interest in a colony, but that is different from showing that they drove district-level investments in infrastructure. The answer to that more precise question remains unclear because much of the prominent work on colonialism is cross-national (e.g. Engerman and Sokoloff (2012), Ma- honey (2010)) or uses current rather than historical data on natural resources (e.g. Acemoglu et al. (2001), Wantchekon and Stanig (2015)). Using disaggregated historical data, this paper shows that natural resources have a modest influence on within-colony investments, even in infrastructure.

Instead, I show that investments in infrastructure, education and health have a common origin. Key geographic features led some places to be centers of pre-colonial trade, mostly focused

4 on slavery, which in turn explains colonial settlements and investments.4 Early explorers and traders had very limited information about the territory, so locational fundamentals (Davis and Weinstein, 2002)—notably natural harbors and capes—inﬂuenced where they landed and therefore the location of initial trading posts between the 16th century and the 19th century, the period of slavery and the Triangle trade.5 This is unlike in South America and South Asia, where primary commodities dominated international trade since the 1500s. And yet, although the slave trade was illegal and residual in British and French African colonies by the late 19th century, those centers of pre-colonial trade had already started developing before colonialism by selling Africans, usually from places further inland, to European traders.

Pre-colonial trading centers were usually located on the coast in a natural harbor (e.g. Mombasa in Kenya) or in a cape (e.g. Dakar in Senegal), or were aided by the presence of a navigable river or a river basin (e.g. Bamako in Mali).6 Hence, I estimate the long-term causal effect of pre-colonial trade on investments, using locational fundamentals as an instrumental variable and find that trade is a primary cause of investments, both because it is a root cause as well as a quantitatively important one. Further, I show that the relevance of pre-colonial commerce is not restricted to these early “colonial development hubs”: distance from early trading centers helps explain the limited diffusion of investments we observe within each colony. The diffusion is limited because colonialism in most of East and West Africa was shorter-lived and shallower than in other regions such as South Africa and certainly South Asia, where trading companies were more important and long-lasting and colonial infrastructure more developed (Jha (2012), Gaikwad (2014)).7 Hence, investments in East and West African colonies remain highly concentrated in a few districts throughout the colonial period (1890s-1950s).

The effect of pre-colonial trade is larger than that of natural resources or post-treatment colonial socioeconomic variables. The advantage of pre-colonial trading centers persisted into the colonial period because they became centers of economic activity and benefited from complementarities between investments in infrastructure, education and health. The pattern is consistent with a logic of increasing returns (Krugman, 1991b) and in particular with agglomeration economies in urban areas (Krugman (1991a), Rosenthal and Strange (2004)), although in a very different context: instead of competitive agents—firms—in a private market, these are monopolistic public agents—colonial state administrators—allocating public finances. In this light, settlers and colonial administrators become an important mechanism rather than a root cause to explain why districts

4In most of East and West Africa, “pre-colonial” refers to the period starting in the 1500s up to the 1885 Berlin Conference, and “colonial” refers to the period between 1885 and 1960. 5In the Triangular trade, Europeans sold manufactured goods to Africa in exchange for slaves that were sold in the Americas. The Americas, in turn, provided cotton, sugar and other primary commodities to Europe. 6Jha(2012) ﬁrst made this case for natural harbors in South Asia. 7The famous British East India Company lasted almost three centuries while the British Royal Africa Company only lasted one. French chartered companies in Africa such as the French West India Company or the Senegal Company were not prominent either.

5 close to early trading centers received much higher investments during the colonial period: they were attractive to Europeans because they already had basic infrastructure in place. I show that the size of the settler community in each district was an important short-term factor aﬀecting within-colony investment allocations that reinforced this feedback loop between early trade and colonial investments.

In sum, this paper makes three contributions. One concerns the data collection effort de- scribed in more detail in Section 4. Second, I show the fundamental role of pre-colonial trade, as opposed to natural resources or settlers, in explaining the loci of colonial state activity. It is fundamental because pre-colonial trade is a primary cause as well as a quantitatively important one, increasing average district colonial investments by an order of magnitude, from roughly 7,000FRA to 70,000FRA (1910 constant francs). Instrumenting for natural harbors and capes increases confidence in the causality of the finding, while colony fixed effects control for institutions and other-colony level unobserved factors. This result adds to two existing papers that instrument pre-colonial trade with natural harbors in a different context—India—to explain different outcomes—tolerance (Jha, 2012) and economic development (Gaikwad, 2014).8 Further, I show that colonial investment diffusion is partly a result of early trade, thus showing that its importance extends beyond pre-colonial development hubs. Finally, I apply the logic of increasing returns to a new context—the public finances of the colonial state—and suggest that settlers are an important mechanism for the observed historical persistence.

The rest of the paper proceeds as follows. Section 2 develops the argument and section 3 provides the relevant historical background and evidence. Sections 4 and 5, respectively, detail the data sources and present descriptive results. Sections 6 and 7 present the econometric results. Section 8 concludes.

2 Locational fundamentals and pre-colonial trade

“It is not an exaggeration that between 1550 and 1800 Europeans learned virtually nothing new about the lands beyond the African coastline. [...] By 1875, in fact, European possessions in Africa still only comprised the coastal forts and trading stations and a few tiny colonies.” Foster (1967, p. 45, 51)

Locational fundamentals are “crucial [geographic] characteristics” of a territory that “change little over time—even if their economic meaning may have evolved. For example, there are advantages of being near a river [or] on the coast, on a plain instead of a mountain or desert, etc.” (Davis

8In another paper, Jia (2014) shows that historical treaty port cities in China are more developed today.

6 and Weinstein, 2002, p. 1270). “The central question in [economic geography] is how to explain the distribution of economic activity across space—across countries of the world, across regions within a country, and across cities,” as Davis and Weinstein (2002, p. 1269) explain.9 Locational fundamentals are useful in that endeavor especially in a pre-industrial context where geography affects and restricts mobility and economic activity (Diamond, 2005) more than in the industrial era. Geographic characteristics that change little over time are also useful because of their exogeneity in two senses. First, they are physical factors originating independently of the social world. In the last century, engineering has greatly affected geography, and yet we can still distinguish natural lakes, such as Lakes Baikal and Titicaca, from man-made ones, such as Lakes Volta, Nasser and reservoirs more generally. Second, locational fundamentals are exogenous because they can affect colonial investments without being affected by them—at least without modern engineering. For example, Koromojo in Kenya and Félouin Mali were the only dams in East and West Africa until the 1950s.

I argue that the main determinant for early European trade was the presence of observable geographic features—natural harbors, capes and rivers—in a few locations along the East and West African coast. The early exploration of the Americas, Africa and Asia was an imperial enterprise with a clear economic motivation. As early as in the 1600s, Africa became a key component of the Triangle trade and of the Atlantic slave trade in particular.10 How did Europeans decide where to dock in order to trade in the absence of man-made docks? Inductively, we observe that the Portuguese and later the French ﬁrst landed in three places as they descended the Northwest African coast. One was Ras Nouadhibou (Cap Blanc), currently divided between Western Sahara and Mauritania. They also established trade in what would become the cities of Saint Louis and Dakar in Senegal because the former is a natural harbor formed by the Senegal river mouth and the latter is located in a cape called Cap-Vert.

In other words, Europeans landed where coastal geography was favorable. For one, naviga- tion technology during the Age of Sail depended much more on environmental factors and wind patterns than later in the Age of Steam (Feyrer and Sacerdote, 2009).11 Europeans observed variation mostly in coastal characteristics because knowledge about the socioeconomic and geological characteristics of the territory was very limited until the 19th century. While natural harbors, capes and coastal rivers were not the only factors aﬀecting sail or location of pre-colonial trade, their presence did make setting sail and hence pre-colonial trade easier. Prior to colonization, only

9The authors consider the three leading theories in economic geography: locational fundamentals, increasing returns, and random growth theory, and test the extent to which each of them is consistent with the distribution of population in Japanese cities from the Stone Age to the present. See paper and sources therein for a detailed explanation. 10Young(1994, p. 103) provides a brief discussion of chartered companies in Africa and their relation to the colonial state. 11Choice was restricted even along that geographic dimension, since African coastal geography is more even and terrain less rugged than in other regions such as the South America or South Asia.

7 areas that had long extracted minerals from known deposits inﬂuenced trading locations. The most famous cases in the colonies under study are perhaps the diamond mines of Sierra Leone and the Gold Coast (current Ghana), where the British disrupted a Trans-Saharan gold trade that had existed for centuries (Young, 1994, p. 134). Even then we can partly see a geographic logic: two key locations where the British landed were the natural harbor of Tagrin Bay in Freetown and Cape Coast in Ghana, a country without natural harbors. East Africa has a longer history of nav- igation than West Africa largely because of Arab explorations and the Arab slave trade (Hourani and Carswell, 1995, p. 83). However, an equivalent logic applies. Mombasa, a natural harbor, and Zanzibar, an island, became centers of Arab-African trade centuries before the Europeans arrived in the 1500s. Locational fundamentals such as natural harbors and capes along the coast and inland rivers matter, especially in a pre-industrial context, because they help us understand early trade patterns.

2.1 Early trade and colonial investments

Did the initial diﬀerences between trading and non-trading locations persist into and during the colonial period? The Age of Steam in the 19th century drastically reduced the sailboat’s dependence on the topograhy of a territory, and railroads provided an alternative to rivers for hinterland penetration. Inland areas in a colony developed in a few instances, notably in Kenya’s Rift Valley. The particular locations where pre-colonial trade was established might have been no more relevant than other similar locations had they not inﬂuenced later colonial settlement and investments. In fact, that is precisely what happened. Proximity to trading centers became even more advanta- geous during the colonial period, starting in the late 19th century. They allowed stronger linkages to the international economy, as Laitin (1982) explains in the case of Western Nigeria.

Some of the early trading centers, such as Abidjan, Freetown and Lagos became small colonies and then colonial capitals, further establishing their preeminence within the colony. As Horowitz (1985, p. 151) argues, “groups located near the colonial capital, near a rail line or port, or near some center of colonial commerce—the sitting of which was usually determined by capricious factors, such as a harbor or a natural resource to be exploited—were well situated to take up opportunities as they arose.” Capitals were rarely chosen for their central geographic location within the colony, unlike in today’s Nigeria (Abuja) or Cote d’Ivoire (Yamoussoukro). Colonial state borders were not even defined by the time the capital was chosen, which restricted the set of available localities. The persistent concentration of economic activity in a few places in each colony had spillover effects on other colonial investments, notably in hospitals and schools such as the Ecole´ William Ponty in Gorée(Dakar) and Fourah Bay College in Freetown. In other words, investments in education and health partly followed early investments in basic infrastructure. It

8 Figure 2: Timeline of major historical periods in East and West Africa 1450 1600 1850 1890 1960

Early exploration Pre-colonial Triangle trade Settlement Colonial period

1885 Berlin Conference

Note: Dates are approximate. Pre-colonial trade and colonization periods vary between colonies.

was cheaper for settlers and colonial oﬃcials to establish schools and hospitals in places where basic infrastructure was already in place and hence with lower initial costs.

Similar patterns of concentration and persistence have been theorized elsewhere, originally in Krugman’s application of increasing returns theory to economic geography to explain why countries develop an “industrialized core” while the rest remains an “agricultural periphery.” Krugman’s (1991b, p. 483) key insight is that, “in order to realize scale economies while minimizing transport costs, manufacturing firms tend to locate in the region with larger demand, but the location of demand itself depends on the distribution of manufacturing.” This logic can help explain patterns of public investments in a very different context: the early stages of the pre-industrial colonial state. Instead of competitive agents—firms—in a private market, we have monopolistic public agents—colonial state administrators—deciding on the allocation of public finances. And yet, as Pierson (2000, p. 254) argues, some characteristics of increasing returns are particularly intense in the political decision-making process, notably (i) reduced competition in the public sector, (ii) an actor’s shorter time horizon and (iii) institutional status quo biases. The first and third reason apply to colonial officials. Colonial administrators did not face competition in allocating public finances and their limited spatial choice and knowledge likely reinforced their biases towards investing in “known territory,” especially in the early stages in the colonial state—from 1821 for Ghana to 1922 for Niger, depending on the timing of colonization. Each particular colony had very few locations with basic infrastructure—locations analogous to Krugman’s industrial core, with lower fixed costs—while all other districts were part of a periphery that was agricultural, pastoral or nomadic. Transportation costs were also very high unless basic transportation infrastructure was in place.

In sum, many colonies in East and West Africa were very far from Weberian states. They were ruled on the cheap and with thin administration on the ground, so European officials faced severe informational and budget constraints (Herbst, 2000). Given that large nominally colonized areas in the hinterland remained mostly unexplored for some time even after colonial borders were defined, initial fixed costs of investment in a new location were compounded by uncertainty about the future profitability of investments. In these circumstances, administrators, notably the colonial

9 governor, usually preferred low-risk investments in the core of the colony to high-risk investments in the periphery.12

3 Historical context

The ‘command and control’ of this [British] empire was always ramshackle and quite often chaotic. To suppose that an order uttered in London was obeyed round the world by zealous proconsuls is an historical fantasy (although a popular one). Darwin(2012, p. xii)

The allocation of investments across districts did not follow a simple decision rule. If it did, investment strategies contained in the colonial documents and records would be perhaps sufficient to explain variation in colonial investments. Investments were not proportional to district population as the needs of the native population were not a major concern: colonial officials on the ground responded to their superiors (upward accountability), not to the colonized (downward accountability).13 Investments were somewhat more responsive to European settlers. The interests of the settlers, however, did not necessarily coincide with those of the British Colonial Office or the French Ministry of the Colonies, or with those of other pressure groups such as missionaries and traders (Darwin, 2012). Again, officials were accountable to their superiors and not to the European minority.14 In fact, settlers in Kenya went as far as to create the European Taxpayer’s Protection League to shift the tax burden away from Europeans to the native population (Gard- ner, 2012, p. 98); they wanted to receive public investments but not pay for their cost. In sum, investments responded to, but were not simply a function of, the settler population. And even if investments partly responded to settlers, this paper investigates why a district was more heavily settled or chosen as the colonial capital in the first place.

3.1 Administration and budgets

While there was no simple decision rule to allocate investments, examining British and French colonial institutional structures provides us with some interesting insights on the public ﬁnances of the colonial state and on the similarities between empires. These similarities are not emphasized in the literature because much historical work on colonial policy focuses on how French ideas of assimilation and direct rule diﬀered from British ideas of association and indirect rule (Crowder

12Economic models on the temporal persistence of inequality address a similar type of question, with individuals deciding between consumption and investment of assets (Boix, 2010, p. 492). 13In that case, the per capita plots in Figure 1 would be mostly ﬂat. French West Africa did establish district quotas based on population for military conscription, but these were systematically violated (Echenberg, 1991). 14That started to change after World War II with the introduction of Legislative Councils in multiple colonies.

10 (1964), Sharkey(2013), Strang(1994)) and because studies of colonial ﬁnance and administration focus on one empire or the other (Delavignette(1968), Suret-Canale(1971), Gardner(2012), Constantine(1984)). 15 At the top of the colonial hierarchy there were the Ministry of the Colonies in Paris and the Colonial Oﬃce in London. These ministries sent a Governor (Lieutenant Governor in French West Africa) to the colony that acted as the main link between the metropole and the civil servants residing in the colony, which included administrators, teachers, judges, engineers, doctors and nurses. Each district (cercle in French West Africa) was led by a district head formally called District Commissioner (Commandant de Cercle).

District heads stationed far from the colonial core had much latitude in local policy and implementation because of “physical distances and no means of communication. [...] The administrative organization [in French West Africa] was officially centralized but effectively decentralized,” making disrict heads “the real chiefs of the French empire” (Huillery (2009, p. 181), Delavignette (1968)). District heads in both empires were in charge of administration, taxation, justice and other public services (Suret-Canale, 1971, p. 72). They were also in charge of relations with village chiefs, and district heads partly relied on them for policy implementation and revenue collection, often using translators.16 Local chiefs in French West African villages were a useful but certainly subordinate figure whose “influence [was limited] to small areas” (Huillery, 2009, p. 181). Similarly, “local chiefs [in British colonies] had a guise of autonomy in their local jurisdictions, but they were actu- ally guided and supervised by the British colonial administrators” (Strang, 1994, p. 149). Lugard’s (1922) promotion of indirect rule is well-known for some British colonies, but its practice was not unique to them. Sharkey (2013) considers the difference in that regard between the two empires a matter of degree, consistent with the move by Gerring et al. (2011, p. 378) “to understand systems of [direct and indirect] rule along a continuum that reflects the degree of central control.”17

The most interesting comparison for the purpose of this article concerns the sources of revenue and budget allocation procedures. London and Paris paid for military expenditures, and even then soldiers were African to minimize costs.18 The main difference is that the British Colonial 15There are historical studies spanning both empires on other issues, notably economic history (Hopkins, 1973) and labor policy (Cooper, 1996). 16For a collection of essays on the importance of translators and interpreters in French and British colonial Africa, see Lawrance et al. (2006). 17The administrative organization of the French and British colonies also presented some differences. There were two layers of hierarchy between the district heads and the metropolitan ministry, but these were different between empires. French West Africa was a political federation under the Governor General, based in Senegal, that commanded and coordinated the eight aforementioned Lieutenant Governors in each colony. The eight British colonies were the main eight territories in the continent controlled by the Colonial Office. However, they did not have the equivalent figure of the Governor General. Instead, Governors were directly under the Secretary of State for the Colonies. Another difference is that most British colonies were divided in provinces, while French West Africa presented no administrative layers between the colony and the districts. The British Provincial Commissioner was the layer between the Governor and the District Commissioner. 18In West Africa, colonial army soldiers were called Senegalese tirailleurs (Echenberg, 1991), a misnomer since they were recruited throughout the French West African territory. The equivalent British army was called the Royal

11 Oﬃce did not fully pay for defense costs, and hence some revenue had to be raised locally, whereas the French Ministry of the Colonies covered defense expenditures with metropolitan taxes. One reason Britain and its colonies divided military expenditures is that the Colonial Oﬃce was often pressured by the Treasury to contribute more and spend less (Stammer, 1967, p. 194).

The most important fact is that revenue for public investments in both empires was raised domestically from export tariffs and a variety of other taxes. Resource constraints led both empires to tax African subjects as well as settlers: “the cost of colonization was endured by local populations rather than French taxpayers and, more precisely, mostly by households rather than firms” (Huillery, 2009, p. 182). The British poll and hut taxes on Africans and income tax on Asian and European settlers served a similar purpose (Gardner, 2012, p. 8). These sources of internal revenue were limited, however, and they help explain the difficulties many colonies endured building adequate infrastructure beyond the colonial core. Given that revenue was raised domestically, differences in public investments between colonies resulted partly from the fact that some colonies raised much more revenue than others for a variety of reasons, including revenue collection capacity and volume of trade in commodities. British and French colonies had to apply for grants in aid to the respective ministries to complement colonial tax and tariff revenues, especially if they wanted to conduct large infrastructure expenditures. In any case, colonial budgets had to pay for public investments in both empires. Similar to the British, only “in the aftermath of World War II France abandoned the requirement that colonies should pay for themselves and began assuming the cost of some public programs” (Lawrence, 2013, p. 118).19

Before then, the existence of a federal budget in French West Africa was intended partly to reduce those limited sources of internal revenue in poorer colonies. The federal budget was mostly raised from custom duties and used to pay for the cost of the federal government and fund large public works (Huillery (2009), Newbury (1960)).20 British colonies also charged tariﬀs of course, but tariﬀ revenues remained in the colony given the lack of federal structures. This partly explains why public investments were more unequal in British colonies (Figure 4 on page 17): higher inequality was in part a result of institutional design.

The lack of a systematic investment strategy in British colonies is perhaps the most surprising aspect of colonial investment policy, especially if we use current Weberian states as our reference categories. It is less surprising if we consider that communications and knowledge of the territory

West African Frontier Force (Haywood, 1964); and in East Africa they were called King’s African Rifles (Parsons, 1999). 19It is important to distinguish between the period up to 1939 and the post-World War II period. “One of the striking aspects of the economy of French West Africa (FWA) is the extent of its development since the end of World War II. Between the turn of the century, when settled administration began in most parts of the area, and 1939, changes in the economic and social structure of the country were few.” (Berg, 1960, p. 392) 20As Newbury explains, it was mainly a redistributive institution. The socialization of tariffs from more economically dynamic colonies like Cote d’Ivoire allowed poorer colonies like Niger to undertake some public works, and it also avoided the creation of a financial burden on France.

12 were also limited, which made policy coordination difficult between the core and the periphery: “the ‘command and control’ of this [British] empire was always ramshackle and quite often chaotic” (Darwin(2012, p. xii), see also Delavignette (1968, p. 63)). Further, cost minimization was a first- order objective. “The concerns of British parliamentarians that the Empire would become a drain on the British Treasury” led “the imperial government to delegate the costs and financial risks of imperial expansion whenever possible” (Gardner, 2012, p. 18). As already mentioned, the metropole constantly strove to reduce the quantity of funds—especially grants-in-aid—available to colonial governments (Constantine (1984, p. 14, 84), Gardner (2012, p. 9)). Given meager budgets and the mostly military role of metropolitan taxes, and especially given the tentative and improvised nature of colonial rule (Darwin, 2012), there was little room for long-term and detailed investment strategies.

While expenditures presumably responded to particular needs, “no explicit investment strategy can be found in [French colonial] local budgets. Motivations reported at the beginning of each local budget explain the general level of annual resources but do not motivate the spatial distribution of public goods provision” (Huillery, 2009, p. 181). British local budgets present a remarkably similar focus on detailed descriptions and administration rather than policy: “colonial tax and spending patterns did not follow a similar logic throughout British Africa” (Frankema, 2011, p. 147) because “[Britain] did not strive to apply a common financial policy to the various dependencies” beyond “general instructions [...] from the Secretary of State for the Colonies” (Stammer, 1967, p. 194). In sum, while colonies strived to increase sources of revenue from cash crops, mineral resources, local taxes and trade tariffs, investments did not follow simple decision rules because their rationales were ill-defined.

4 Data sources

I combine various historical data sources to uncover the determinants of investment levels. To the best of my knowledge, this paper presents the most extensive data on colonial investments at the colonial district level for French West Africa—collected by Huillery (2009)—and for the main eight British colonies under the Colonial Oﬃce—original data collection: Benin (formerly Da- homey), Burkina Faso (Upper Volta), Cote d’Ivoire, Ghana (Gold Coast), Guinea, Kenya, Malawi (Nyasaland), Mali (French Soudan), Mauritania, Niger, Nigeria, Senegal, Sierra Leone, Tanzania (Tanganyika), Uganda, Zambia (Northern Rhodesia). One important advantage of focusing on these colonies is their rather homogeneous colonial institutional structure within each empire, as explained in Section 3. French West Africa was a federation, and these eight British colonies had a much more similar colonial structure compared to other territories such as Egypt and South Africa. Unlike French Algeria or British Southern Africa, which had been colonized before 1900,

13 most of East and West Africa was not even adequately mapped before 1850 and was not integrated into the French and British empires until late in the 19th century.21 Perhaps due in part to their similar colonial experience, all of these 16 colonies became independent around 1960.

Another important advantage is that record-keeping procedures were very similar within each empire, owing again to the fact that French West Africa was a federation and that British colonies all reported directly to the Colonial Office. The yearly Comptes Définitifs (Final Budgets) and the Blue Books, as the Colonial Office called its yearly reports, were standardized across colonies for the purposes of accountability.22 This comparability does not directly extend to Sudan, for instance, because it was under the control of the Foreign Office. The British National Archives and the Archives Nationales d’Outre-Mer keep most of the original records. Huillery (2009) collected the original French records for selected years in the 1910-1939 period.23 I collected British colonial records from 1915, 1920, 1927, 1928 and 1938 as a function of availability in various libraries.24 British and French records are often organized or can be grouped at the level of districts. While the colonial records have many gaps, they often contain detailed information on demographics, education, infrastructure investments and other activities.25

The colonial records are complemented with a set of sources that provide information on physical and geographic characteristics of the territory. Many of these sources contain data I georeference and analyze using ArcGIS, such as colonial maps with district boundaries published by the Colo- nial Oﬃce (Figure 20 in Appendix).26 Huillery (2010) collected several of those physical and geographic attributes for French West Africa, such as distance between the district capital and the coast, altitude, the presence of navigable rivers—which I complemented with a map by C.S. Ham- mond (1921)—and pre-colonial trading posts. I extend those variables to include British colonies and collect others not included by Huillery. The sources for pre-colonial trading posts in British colonies come from Curtin et al. (1995) and Slave Voyages (2013). For natural harbors and capes most data come from Ramsar (2016) and Ports.com (2016).27 Altitude is a rough proxy for disease environment, notably for malaria (World Health Organization, 2016), but to better capture disease

21There are exceptions. Part of the hinterland in Senegal and Ghana had been colonized much earlier. Also, Zambia was incorporated as Northern Rhodesia into the Colonial Oﬃce only in 1924. Tanzania, then Tanganyika, was under German occupation prior to World War I and incorporated to the British Empire as a League of Nations Mandate only in 1922. 22They were not the only empires to keep such records. The Portuguese and the Belgian records I examined are similarly detailed. 23Technically the panel extends to 1956, but data are mostly missing post-1939. 24The main diﬃculty is to make the data usable for statistical purposes, which took close to two years for each author. Figure 19 is a page of a British Blue Book. 25While not the topic of this paper, in a related paper I am combining the British and French colonial data with recent sources such as the Demographic and Health Surveys (DHS) to explain subnational variation in post-colonial levels of public goods provision. 26Interestingly, colonial district boundaries are relevant today. With some exceptions, many have changed little over time and around 80% of them remain in place in the 21st century. Districts today are often partitions of a larger colonial district. 27Tables 21 and 22 in the Appendix provide tests of covariate balance for both indicator variables.

14 environment I use a geocoded map of malaria prevalence around 1900 (Lysenko and Semashko, 1968) and FAO tse-tse ﬂy data (Alsan, 2015). Those data are important in tropical Africa, “often referred to as ‘the white man’s grave”’ (Darwin, 2012, p. 138). I also georeference several historical maps of natural resources. One is the map by Hubert (1922) presented in Figure 21, likely the most comprehensive on the issue for pre-1940 West Africa. The other is a detailed worldwide map by Kuhne (1927). I complement these two main sources with an early publication by the United States Geological Services (USGS, 1921) that also has world coverage but is much less detailed.

Large plantations like cotton or crops were important in some colonies, and investments could partly be a function of crop development. Because they are often an endogenous choice by the colonizer, I use geological soil characteristics such as nutrient and oxygen availability (FAO/IIASA, 2012). These are current measures due to lack of detailed historical data and hence far from ideal, but they are useful to the extent soil quality does not change dramatically over time (Wantchekon and Stanig, 2015, p. 27). Otherwise, the paper purposely avoids contemporary databases on natural resources and diseases that are commonly used in some well-known studies (e.g. see Parker (1997) in Acemoglu et al. (2002)) to avoid obvious reverse causality.

Finally, the paper includes data on relevant colonial socioeconomic characteristics, such as a historical map on the presence of Islam across the continent (Bartholomew, 1913). For some other pre-colonial characteristics, notably whether the district was located in a pre-colonial kingdom or in an acephalous society, I extend Huillery’s variables to include British colonies. I also draw from the Murdock (1959) dataset on pre-colonial ethnic group characteristics because it provides useful proxies of pre-colonial economic and political development, such as intensity of agriculture, settlement patterns, the size of local communities and level of political centralization.

5 Descriptive results

Figure 3 presents levels of infrastructure investments by colony. When we adjust investments per capita, Ghana’s levels of investments are similar to Guinea’s and much higher than those of any other British colony. The colonies received little help from the metropole, so this diﬀerence is largely due to Ghana’s colonial government raising more revenue locally and hence having much larger budgets than the other British colonies (Hopkins, 1973, p. 190). Figure 4 quantiﬁes inequality in infrastructure investments by computing Gini indices by colony. The indices are calculated using district expenditures, and they are above 0.7 in the average British colony and around 0.6 in the average French colony.28 Figures 5 and 6 show that Gini indices are usually above 0.4 for educational and health investments, and inequality is especially high in British colonies. Figures

28Typically, Gini indices or coeﬃcients use individual income data to measure economic inequality, where 0 means perfect equality and 1 perfect inequality.

15 13-18 in the Appendix present the district expenditures from which Gini indices are calculated. The ﬁgures show that inequality was pervasive across colonies even if we examine per capita investments.29 British colonies, at least measured by public investments, were even more unequal than French ones. This could partly be the result of the redistributive federal budgets in French West Africa, but between-colony variation could be the result of many colony and empire-speciﬁc factors that are not the focus of this paper. Figure 7 presents the variation in a map along with the presence of the most valued natural resources by district. The correlation pattern varies by colony: it appears to be quite strong in Ghana but nonexistent in Nigeria, and overall it might be weaker than expected. Figure 3: Infrastructure expenditures in British and French colonies (1910-1939, in 1910 FRA)

Total Per 1 million people 2.0e+06 4.0e+06 1.0e+06 2.0e+06 0 0

Kenya Kenya Malawi Zambia Uganda Nigeria Ghana Malawi Zambia Nigeria Uganda Ghana Tanzania Tanzania Sierra Leone Sierra Leone

Total Per 1 million people 4.0e+06 2.0e+06 2.0e+06 1.0e+06 0 0

Mali Mali Niger Benin Niger Benin Senegal Guinea Senegal Guinea Mauritania Mauritania Burkina Faso Cote d'Ivoire Burkina Faso Cote d'Ivoire

Tables4,5 and 6 in the Appendix present summary statistics of investments in infrastructure. Table 4 shows that infrastructure expenditures in a given district-year were around 50,000FRA, in 1910 real French francs, in each empire. They are a bit more evenly spread in French West Africa. More importantly, French per capita expenditures doubles the British. Table 5 shows that the diﬀerence is driven by low investments in British East African colonies—the French had no colonies in East Africa. Within West Africa, due to Ghana and Nigeria, we observe higher British expenditures but similar per capita levels between empires and with very similar spread. Europeans settled earlier in West Africa and raised higher taxes, on average, hence the higher expenditures. Finally, Table 6 breaks down the four main types of infrastructure. The average British and French district receives close to no investments in sewage/water sanitation or in electricity/lighting. There

29The denominator includes the local and settler population.

16 Figure 4: Infrastructure Gini indices by colony

Infrastructure expenditures Infrastructure expenditures per capita 1 1 .8 .8 .6 .6 .4 .4 .2 .2 0 0

Mali Mali Niger Niger Kenya Ghana Benin Kenya Benin Nigeria MalawiUganda Zambia Guinea Uganda GhanaMalawi ZambiaNigeria Guinea Tanzania Senegal Tanzania Senegal Mauritania Mauritania Sierra Leone BurkinaCote Faso d'Ivoire Sierra Leone Burkina Faso Cote d'Ivoire British colonies French colonies British colonies French colonies

Figure 5: Education Gini indices by colony

Students Students per capita 1 1 .8 .8 .6 .6 .4 .4 .2 .2 0 0

Mali Mali Kenya BeninNiger Kenya NigerBenin UgandaNigeriaGhana Zambia Guinea UgandaGhana Nigeria Zambia Guinea Tanzania Tanzania Sierra Leone Burkina FasoCote d'Ivoire Sierra Leone Burkina Faso Cote d'Ivoire British colonies French colonies British colonies French colonies

Figure 6: Health Gini indices by colony

Health staff Health staff per capita 1 1 .8 .8 .6 .6 .4 .4 .2 .2 0 0

Mali Mali Kenya Niger Benin Kenya Niger Benin Uganda Nigeria GhanaZambiaMalawi Guinea Uganda Ghana MalawiZambia Nigeria Guinea Tanzania Senegal Tanzania Senegal Mauritania Mauritania Sierra Leone Cote d'Ivoire Burkina Faso Sierra LeoneBurkinaCote Faso d'Ivoire British colonies French colonies British colonies French colonies

17 Figure 7: Public infrastructure investments by district (1910-1939 average) and location of base and precious metals (1922)

was no such system in place in most districts because setup was costly. Colonial records show that most expenditures were concentrated in buildings and premises of various sorts—port authorities, residences of district oﬃcers in core and remote areas—and in transportation—mostly railroads, roads, bridges and harbors.

6 Results

This section begins by examining what locational fundamentals aﬀected the placement of early pre- colonial trading posts. Next, I show that the presence of pre-colonial trading posts, instrumented by coastal geography, increase colonial investments not only in infrastructure but also in education and health. Settlers help explain this long-term eﬀect. I also present some further results splitting the sample by empire given the large literature that compares British and French colonialism.30

Models in Table 1 predict whether a colonial district had a pre-colonial trading post as a function of locational fundamentals. I use both probit (equations 1 and 3) and linear probability models (equations 2 and 4) because the former adequately models binary variables but the second

30For a recent quantitative comparison using the colonial split of French and British Cameroons, see Lee(2012).

18 is akin to the first stage of the Two-Stage Least Squares (2SLS) models used below.31 Natural harbors and capes increase the probability of pre-colonial trading centers in coastal districts, as expected (models 1 and 2), although the relationship is not deterministic—not all trading posts were in natural harbors and the other way around (see Table 20 in the Appendix). In the case of inland colonies (models 3 and 4), pre-colonial trade refers to the first trading post established in what would later become a colony (e.g. in Niger, Niamey equals 1 and the other districts 0). As expected, coastal distance negatively affects the probability that a particular district was the first trading post in the colony, whereas the partial correlation with a navigable river is positive. Given the importance of terrain ruggedness (Nunn and Puga, 2012) and disease environment (Alsan, 2015) for long-term African development, it is interesting that neither affects the location of early trading posts even in inland colonies—except for the tsetse fly marginally in one model. This is further evidence that early trading posts were established with very limited knowledge of the environment.

Table 1: Pre-colonial trade and locational fundamentals

(1) (2) (3) (4) Coastal, probit Coastal, LPM Inland, probit Inland, LPM Natural harbor or cape indicator 1.40∗∗ 0.40∗∗ (0.50) (0.15) Coastal distance, in 100km -0.47† -0.03∗ (0.24) (0.01) Navigable river indicator -0.25 -0.04 0.94 0.16∗ (0.55) (0.17) (0.71) (0.07) Terrain ruggedness 1.08 0.27 -1.60 -0.07 (0.87) (0.24) (1.48) (0.09) Malaria prevalence index 0.61 0.19 0.27 0.01 (0.58) (0.18) (0.68) (0.04) Tsetsefly index 0.43 0.12 -1.60 -0.09† (0.57) (0.16) (1.09) (0.05) Constant -2.95 -0.45 2.21 0.39† (2.21) (0.59) (2.86) (0.21) Observations 54 56 101 101 Adjusted R2 0.08 0.04 Pseudo R2 0.26 0.31 Notes: p < 0.10, * p < 0.05, ** p < 0.01. Models include colony fixed effects.

The two-stage least squares (2SLS) models in Tables 8–12 in the Appendix identify the causal eﬀect of pre-colonial trade and of some competing explanations, notably natural resources, on colonial investments. The exclusion restriction claim is that natural harbors and capes aﬀect

31Using a logit or probit for the ﬁrst stage of 2SLS is Wooldridge’s “forbidden regression” because the standard errors of the second stage are not estimated consistently. The formal ﬁrst stage regression is in Table7 in the Appendix.

19 colonial investments only because they enabled early pre-colonial trade, as argued by Jha (2012), Gaikwad (2014), and this paper (Section 2). The 2SLS models take the following form:

T T T Tij = β0 + β1Gij + L β2k + N β3k + S β4k + ηj + ij (1) T T T log(Yij) = β0 + βIV Tcij + L β2k + N β3k + S β4k + ηj + ij, (2)

where Yij is the colonial investment of interest in district i in colony j, G stands for natural harbors and capes, L stands for other locational fundamentals, N for natural resources and soil quality,

S for colonial district population, area and pre-colonial socioeconomic characteristics, and ηj are colony fixed effects. Given the highly unequal investment distributions, ordinary least squares (OLS) models are logged to reduce dependence on extreme observations.32 While all models include country fixed effects, model 1 in each table includes only L, model 2 adds N and model 4 adds S. Models 1, 2 and 4 purposely omit variables that would suffer from simultaneity, such as the number of settlers in the district or an indicator for colonial capitals. African population and district area are two exceptions included in models 4, 5 and 6 to account for explanations simply based on demographic and spatial size of the district. Models 3 and 5 include the number of settlers in the district as a mechanism. Model 6 is equivalent to model 4 but subsets the data to colonies with natural resources. The Wald F-statistic is between 13 and 17 in all models, above the Stock and Yogo convention of 10 showing that the instrument is not weak. The first stage is reported in Table 7 in the Appendix.

Figure 8 plots some key coefficients from tables 8-12 to visualize the overall pattern of signifi- cance. Pre-colonial trade increases all three types of investments but not the likelihood of having a railroad.33 Navigable rivers were important for exploration and resource extraction in the absence of a railroad, suggesting they were substitute modes of transportation. It is more surprising that rivers do not affect health or educational investments. They only increase public investments other than railroads in some models when the sample is restricted to inland colonies.34 The indicator for noble metals and diamonds increases infrastructure investments, especially railroads, but not human capital investments in teachers, students or public health staff. This is an important result: while the effect of early trade spills over to health and education, the effect of natural resources does not, consistent with the extractive nature of the latter. Finally, higher malaria prevalence

32Estimates from seemingly unrelated regressions (SUR) would be identical because, although outcome variables are indeed correlated, the right-hand side of the equation is the same in all models. Also, a linear probability model and a logit model yield the same results in Table 9, where presence of a railroad is a binary outcome. 33Two thirds of districts with pre-colonial trading posts had a railroad, but the importance of the railroad is precisely that it facilitated extractive exports even in coastal places without pre-colonial trade. 34This is partly due to the large diﬀerence in investment levels between coastal and non-coastal colonies, with the former dominating the latter.

20 is negatively correlated with all investments and signiﬁcantly so in railroad and settler models, suggesting Europeans avoided moving into areas where malaria was meso, hyper or holoendemic.35

Figure 8: Eﬀects of instrumented pre-colonial trade and locational fundamentals on colonial investments

Infrastructure Railroad

Pre-colonial trading post

Navigable river

Malaria prevalance

Gold, silver or diamonds

Health staff Students

Pre-colonial trading post

Navigable river

Malaria prevalance

Gold, silver or diamonds

-2 -1 0 1 2 3 4 -2 -1 0 1 2 3 4 Regression coefficients

Note: These are coeﬃcients from model 3 in Tables 8, 9, 10 and 11. Conﬁdence intervals shown at the 95% and 90% level.

Figure 9 presents quantitative estimates of the predicted levels of investments in the models above by switching the variable of interest—pre-colonial trade and natural resources—from 0 to 1. A coastal district with a pre-colonial trading post receives infrastructure expenditures an order of magnitude larger than another coastal district without it, from roughly 7,000FRA to 70,000FRA (1910 prices). The effect is not only statistically significant and causally identified; it is also a very large marginal effect. For comparison, consider the marginal effect of natural resources. While they were a motivation for colonial expansion and raised a colony’s overall revenue, the increase in public investments for districts possessing them were relatively modest. A district with gold, silver or diamonds almost quadruples the expenditures of one without any of these three resources, but the difference is between roughly 8,000FRA and 2,000FRA. The effect of pre-colonial trade is absolutely and relatively larger. Some colonies like Benin and Kenya did not have natural resources, however, so using the full sample may be underestimating their importance. We can provide a harder test for pre-colonial trade by restricting the sample to colonies with natural

35Tables8-12 show that other fundamentals such as terrain ruggedness and soil quality are not signiﬁcant predictors of investments. Other models including individual measures of soil quality such as nutrient availability, oxygen availability, rooting conditions also show null eﬀects.

21 Figure 9: Marginal eﬀects of pre-colonial trade and natural resources on colonial investments 12 9.5

67,556FRA 9 11 7,745FRA 10 8.5 9 8 (1910-1939), in 1910 FRA 6,983FRA (1910-1939), in 1910 FRA Logged infrastructure expenditures 2,149FRA 8 7.5 Logged infrastructure expenditures in coastal districts 0 1 0 1 Pre-colonial trading post Gold, silver, diamonds (1920)

Note: These are the marginal effects for model 3 in Table 8. The left plot restricts the sample to coastal districts (n = 57) to increase comparability. Confidence intervals shown at the 95% level. resources such as Ghana and Guinea.36 The results of this exercise are presented in Model 6 of Tables 8-12. Results show that trade is just as significant, statistically and substantively, in this subset of colonies. Most investments were located in pre-colonial trading posts that were usually infamous for a past closely connected to the slave trade.

6.1 Settlers and diﬀerences by empire

The aggregate or total effect of early trading posts goes through multiple historical channels such as settler presence and colonial capitals, some of which were already trading posts centuries earlier. As expected, pre-colonial trading posts correlate with settlers (ρ = 0.22) and colonial capital indicators (0.35). In turn, settlers correlate with investments in infrastructure (0.41), health (0.48) and education (0.46)—the respective correlations for colonial capital indicators are 0.21, 0.34 and 0.22. Models 3 and 5 in tables 8-12 of the Appendix show that colonial settlers account for all the effect of pre-colonial trade on infrastructure investments, about half of the effect on education and health, and around one third of the effect on missions. In other words, settlers in the 19th and 20th century naturally chose locations that were centers of economic activity, usually of pre-colonial

36Colonies excluded due to lacking gold, silver and diamonds as of 1920: Benin, Kenya, Malawi, Niger and Uganda. The other 11 colonies had at least one of the three resources.

22 trade, which compounded the advantages of these particular locations and their surroundings (e.g. Lagos Colony and the surrounding Southwestern Nigeria) even further.

The second exercise before concluding this section compares the importance of some covariates in the British and French empires. While this is not the focus of this paper, I provide a brief discussion in light of the large literature comparing empires. Figure 10 presents coefficients resulting from the same model (equation2) as in Figure 8, but now split by empire. The importance of pre-colonial trade is very similar in magnitude between empires. First, the figure shows that all investments are positively correlated with population size. Second, the British invested less in areas with greater Muslim presence, a factor that only marginally affected French public educational investments.37 The well-known case of low education provision in Muslim Northern Nigeria, ruled indirectly, vs. higher investments in the non-Muslim South, ruled directly, may apply more broadly within the British empire in Africa. Colonialists in both empires held many racial prejudices;38 however, British investments seem to discriminate more than the French based on the number of ethnic groups in the district, pre-colonial intensity of agriculture and pre-colonial political centralization. The patterns in Figure 10 are interesting because, albeit correlational, they suggest the British might have discriminated more based on observable ethnic and socioeconomic traits than the French.

The estimates on pre-colonial political centralization are particularly interesting because they are either null or negative whether I use the ordinal measure from Murdock (reported in Figure 10) or dichotomous measures for kingdoms and acephalous societies, and even bivariate correlations are around 0. If pre-colonial kingdoms provided better public services today (Gennaioli et al., 2013), it is likely not because of higher levels of colonial investments. In conclusion, this section estimated the causal effect of pre-colonial trade on colonial investments and examined correlationally other possible determinants of public investments. Pre-colonial trade trumps the importance of social factors (e.g. pre-colonial kingdoms) and of natural resources insofar as public investments are concerned. 37Islam does not affect mission location in French West Africa but there were few missions to begin with apart from Senegal and Benin. 38Examples of racism by British colonial officials abound. Mamdani (1996), among others, highlights that racist stereotyping was pervasive: “The Baganda proper [in developed central Uganda] are eager to become educated [...] with a zest which is almost pathetic” (Herbertson and Howarth, 1914, p. 297). But French colonial officials were far from race blind: “the Wolof [in developed Western Senegal] was spoiled and had become a terrible snob,” no longer fit to be a tirailleur [soldier] (Echenberg, 1991, chapter 2), while the Bambara [from Southern Mali] was not very intelligent but possessed “all the strong warrior’s virtues.”

23 Figure 10: Eﬀects of socioeconomic variables on colonial investments

Infrastructure Students Colonial district population, logged

Prevalence of Islam (1910)

Pre-col. # of ethnic groups

Pre-col. intensity of agriculture

Pre-col. political centralization

Health staff Missions Colonial district population, logged

Prevalence of Islam (1910)

Pre-col. # of ethnic groups

Pre-col. intensity of agriculture

Pre-col. political centralization -1 0 1 -1 0 1 Regression coefficients French colonies British colonies

Note: These are the coeﬃcients for model 3 in Tables 8, 10, 11 and 12 splitting the sample by empire. Conﬁdence intervals shown at the 95% and 90% level.

7 Diﬀusion and persistence of colonial investments

In this section, I provide some further evidence for the importance of trade into the colonial period and for the role of settlers as a investment diﬀusion mechanism. I also conduct a horse race between early missions and early trade and show that the latter better explains colonial investments. In the second half of the section, empirical evidence suggests that inequality between district investments remained constant and sometimes even increased during colonial times, consistent with a logic of increasing returns.

7.1 Colonial investment diﬀusion

Investments are concentrated in a few districts in each colony (Figures 13-18 in Appendix), and development hubs—Freetown in Sierra Leone, Saint Louis in Senegal—received much higher investments than other nearby territories. These districts are often colonial capitals and/or pre-colonial trading posts. However, other districts usually receive a positive, even if small, amount of investment. Is this limited diﬀusion of investments throughout the colony partly a function of early trade? Or is the relevance of early trade circumscribed to these usually coastal “colonial development hubs”?

24 There are historical and economic reasons why investment diffusion may not be a function of pre-colonial trade in the African continent. Historically, the Principle of Effective Occupation adopted at the Berlin Conference meant that empires needed to develop some presence throughout the territory. In some colonies like Niger, French occupation was not a commercial but a military affair. Budget-constrained colonial states may have invested small amounts somewhat uniformly in inland and remote border districts in order to strengthen their territorial claims. Even without the Principle, it was usually necessary to build the district officer headquarters and basic infrastructure such as secondary roads to facilitate travel and revenue collection. Economically, motives to invest inland were sometimes orthogonal to early trade. Some natural resources were located far from the coast, as in Guinea and Mauritania, while territory between the coast and those inland resources was less profitable. The fertile lands of Kenya’s Rift Valley were developed even though they were 500km from the main trading center, Mombasa, and much of the territory in between was underdeveloped.

However, we also know that “agglomeration economies attenuate with distance” (Rosenthal and Strange, 2004, p. 2120). Colonial state expansion was progressive and limited because of European ﬁnancial and manpower constraints. In this gradual logic of expansion, pre-colonial enclaves could have been important departing points since colonial states did not extend randomly to recently conquered non-coastal lands. Neighboring districts may beneﬁt from proximity to pre-colonial trading centers if there are spatial investment spillovers, especially after World War I when economic activity and hence investments increased. I test these competing hypotheses econometrically with OLS models:

log(Yik) = β0 + β1T + β2DT + β3DC + β4P + β5E + ηk + ik, (3) where Y is the outcome of interest, T is an indicator for pre-colonial trade post, DT is the distance between the district capital and the nearest pre-colonial trading post, DC is the distance between the district capital and the coast, P is logged population, η are country ﬁxed eﬀects and E is the logged number of Europeans. The pre-colonial trading post indicator and coastal distance control means that we are examining the within-coastal colony variation among districts without a trading post that is not already explained by coastal distance. I use geodesic distances between district capitals advisedly because it eliminates the endogeneity that would be introduced by taking into account local geography such as hills and rivers and especially man-made infrastructure such as roads.

Table2 shows that distance from a pre-colonial trading post reduces all types of investments. The results suggest that the limited diffusion of colonial investments we observe is partly a function of early trade, consistent with findings in India (Gaikwad, 2014). The negative effect of DT likely goes through many channels. Table 3 shows the results of the model in equation 3 that includes

25 European settlers (E) as a mechanism. Interestingly, distance from a pre-colonial post is no longer signiﬁcant—albeit it remains positive—for non-extractive public investments (health and education). However, it remains signiﬁcant and very similar in size for investments in infrastructure and missions. Infrastructure investments were largely not driven by settlers because of their extractive nature which required mostly forced labor. Results are also consistent with missions operating under colonial state consent but trying to reach areas in which settlers—or even the colonial state—were not very present.

Table 2: Diﬀusion of investments (1910-1939) across districts within coastal colonies

(1) (2) (3) (4) (5) Infrastructure Railroads Education Health Missions Pre-colonial trading post indicator 1.41† 0.03 1.84∗∗ 1.22∗∗ 0.40∗ (0.64) (0.11) (0.31) (0.23) (0.17) Distance from post, in 100km -0.27∗∗ -0.05∗∗ -0.14† -0.06∗∗ -0.05∗ (0.07) (0.02) (0.06) (0.02) (0.02) Constant 13.16∗∗ 2.07† -2.24 -0.55 0.37 (2.59) (1.05) (3.18) (1.16) (0.91) Observations 211 211 202 210 211 Adjusted R2 0.33 0.15 0.36 0.26 0.42 Notes: † p < 0.10, * p < 0.05, ** p < 0.01. Clustered standard errors by colony in parentheses. All models include colony ﬁxed eﬀects and control for district population and distance from the coast. All outcome variables are logged for normality.

Table 3: Settlers as a colonial investment diﬀusion mechanism

(1) (2) (3) (4) (5) Infrastructure Railroads Education Health Missions Pre-colonial trading post indicator 0.35 -0.13 1.39∗∗ 0.76∗∗ 0.26 (0.68) (0.12) (0.29) (0.14) (0.19) Distance from post, in 100km -0.21∗∗ -0.04∗∗ -0.10 -0.04 -0.04∗∗ (0.05) (0.01) (0.06) (0.04) (0.01) European population, logged 1.00∗∗ 0.13∗∗ 0.51∗ 0.43∗∗ 0.11† (0.21) (0.01) (0.17) (0.05) (0.06) Constant 1.82 -0.12 -2.76 -4.67∗∗ -1.92† (2.18) (0.47) (3.96) (1.03) (0.99) Observations 200 200 191 199 200 Adjusted R2 0.50 0.26 0.44 0.46 0.44 Notes: † p < 0.10, * p < 0.05, ** p < 0.01. Clustered standard errors by colony in parentheses. All models include colony ﬁxed eﬀects and control for district population and distance from the coast. All outcome variables are logged for normality.

26 7.2 Early trade and early missions

Given the recent importance in the literature on missionaries (Woodberry (2012), Cagéand Rueda (2016)), I exploit distance from the initial locations of economic and missionary activity in each of the 16 colonies, not just coastal ones, to show that the first pre-colonial trading posts are better predictors of colonial investments than the first pre-colonial missions. In particular, I code the location of the first trading post and the first mission in each colony as well as the distance between that location and each district capital in the colony. The models in Tables 13, 14, 15 and 16 in the Appendix follow the specification

log(Yij) = β0 + β1Tij + β2kDTij + β4Mij + β5DMij + ηj + ij, (4) where Y is the colonial investment of interest in district i in colony j, T (M) is an indicator for the first trading post (missionary school) in each colony and DT (DM) is the distance between each district capital and the trading post (missionary school). η are country fixed effects and are clustered standard errors. The results show that an indicator for the initial trading post in a colony and its distance from other district capitals in the colony are good predictors of 20th century investments in infrastructure, education and health. The difference is not simply driven by first trading posts becoming the colonial capital—which occurs in 5 of 16 colonies.39 The models show a primacy of early trade over early missions—even in public education. Only the number of missions in a district (Table 16 in the Appendix) is better predicted by distance from the first school in the colony, which is not surprising since it was invariably a mission.

7.3 Persistence during colonial times

I now provide evidence for the persistence of investments during the colonial period.40 The collected data from British records are a cross-section for health and education in some colonies, and although we observe infrastructure investments twice or thrice it is too limited for panel data analysis. The French records collected by Huillery (2009) are also an unbalanced panel with many gaps, but many observations are repeated over a few years. The gaps arise in part because colonial record keeping, though extremely extensive and intensive for its time, was not always systematic or consistent. Within those limitations, I present correlations over time for the three key investments

39As of 1900, these were Mombasa (Kenya), Lagos (Nigeria), Saint Louis (Senegal), Freetown (Sierra Leone) and Entebbe (Uganda). The capitals of Kenya, Senegal and Uganda have changed since then. 40Pre-20th century diﬀerences in investments are diﬃcult to measure for several reasons. For one, the timing of initial pre-colonial presence varied widely by colony, ranging from the 1500s in some West African coastal locations such as the Island of Goree and Lagos to the end of the 19th century in the Niger or Malawi. That helps explain why there are no similarly systematic pre-colonial records. Further, initial European presence came invariably in the form of trading posts or missionary stations, not in the form of public investments since there was no colonial state.

27 and then I test whether disparities in investments change over time for the case of French West Africa.

Across years with suﬃcient data in the 1910-1939 time frame, the number of students and teachers in French West African districts correlates at around 0.8, as shown in Figure 11. The persistence in health staﬀ is almost just as high, although it varies somewhat from corr(health1928, health1939) =

0.54 to corr(health1915, health1928) = 0.83. The persistence in infrastructure investments is over 0.95 for any of the three pairwise correlations in Figure 11. Districts that received no infrastructure in 1915 continue to receive none in 1928, and those high receivers in 1915 continue to be at the top.41 In the case of British colonies, the number of schools, students and health staﬀ all also correlate at 0.8 or above. For infrastructure investments, the correlation is 0.57 (Figure 12). The pattern is overall similar to the French data, and it shows that disparities persist during the interwar period in both empires, even if shocks such as the Great War and the Great Depression reduced overall levels of revenue collection and therefore investment (Gardner, 2012, p. 6).

Figure 11: Persistence of public investments in French West Africa

Teachers, logged Students, logged

1915 1913

4 8

2 1928 6 1933

0 4 4 8

2 1939 6 1936

0 4

Infrastructure investments, logged Health personnel, logged

1915 1915

12 6 8 1916 4 1928 4 2 0 0 12 6 8 1928 4 1939 4 2 0 0

Note: The correlation matrices show continuity in logged levels of public investments over time. Both X and Y axis use the same logged scale.

Since the French data are a less incomplete panel, albeit unbalanced, I can examine whether investment patterns over time converge or diverge. I use autoregressive models with one lag (AR1)

41Data on infrastructure for the 1930s was too incomplete.

28 to model inequality over time for the three types of investment—infrastructure, education and health (equation5). One advantage of AR1 models, as opposed to a simple serial correlation, is that the constant controls for deterministic trends.42 I also examine whether initial levels of colonial investments (Iit0 ) predict subsequent diﬀerences (equation 6):

Iit = α + γIit−1 + βIit0 + it (5)

4Iit = Iit − Iit−k = α + βIit0 + it (6)

Investments (I) are indexed by district i and year t. Models 1 and 2 in Tables 17, 18 and 19 in the Appendix present levels of investments, as in the serial correlations of Figure 11. Models 3 and 4 are logged to reduce dependence on outliers. Equation 5 is an AR1 process that includes the initial value (t0) of I in the time series to adjust for baseline levels of I (models 1 and 3 of Tables 17, 18 and 19 in the Appendix). β > 0 indicates increasing divergence and β < 0 indicates convergence when compared to the distribution at t0. This approach is more rigorous but the cost is a reduced sample size. Hence, equation 6 also considers changes in investments over time (models 2 and 4 of Tables 17, 18 and 19). Because the panel is unbalanced and incomplete, the main diﬀerence is that the previous value It−k is not necessarily the previous year but the nearest previous year for which there is data available, which in some cases may be several years. These models are more ﬂexible in determining the previous observation and hence the regressions include larger sample sizes. I report both for completeness because there can be biases in either modeling strategy given the gaps in the data.

All models except for one indicate either increasing or constant disparities in educational, health and infrastructure investments. Table 17 in the Appendix shows that teachers per district in 1915 is a good predictor of increase in district teachers later on. Table 18 shows that districts with high health staff may have more health staff later on (models 1 and 3) or in any case not fewer (models 2 and 4). The results on infrastructure investments (Table 19) vary depending on the specification. These mixed findings on infrastructure suggest that increasing returns may apply more to public services that clearly benefit from complementarities and economies of scale, such as schools and hospitals, than to basic infrastructure. In sum, models in this section do not show or disentangle the mechanisms of persistence—future research should. They show that distance from early trading posts matters for colonial investments; that disparities remain stable or increase, consistent with increasing returns to investments in the colonial core; and that settlers are likely an important mechanism in this feedback loop of weakly increasing disparities.

42In other words, the constant would capture a constant increase across districts due to inﬂation (already ac- counted for by using real 1910FRA), a larger budget, or other factors.

29 8 Conclusions

This paper establishes and quantifies the striking inequality in colonial investments in infrastructure, education and health in British and French colonies. By using disaggregated data, I can move beyond the traditional cross-country analysis to find that coastal districts that were pre-colonial trading centers—instrumented by natural harbors and capes, adapting Jha (2012)—receive around ten times more investments than coastal districts that were not. Instrumenting for natural harbors and capes increases confidence in the causality of the finding, while colony fixed effects control for institutions and other-colony level unobserved factors. Further, and although trade in Africa was mostly coastal and colonialism short-lived compared to other continents, I show that the relevance of early trade for all three types of investments is not restricted to these usually coastal colonial development hubs but that it extends to other district of the colony as an inverse function of geodesic distance from the trading center, consistent with earlier findings showing that “agglomeration economies attenuate with distance” (Rosenthal and Strange, 2004, p. 2120). The persistent concentration of economic activity in a few places in each colony had spillover effects on other colonial investments, notably in hospitals and schools because investments in education and health were easier in locations where the initial infrastructure was already in place. In particular, it was cheaper for settlers and colonial officials to establish schools and hospitals in places with lower initial costs.

Natural resources were a motivation for the colonial enterprise, but they are poor predictors of within-colony investment allocations or, more simply, of where the money goes. They increase investments in infrastructure and in railroads in particular, but the eﬀect does not spill over into education or health. Even in the case of infrastructure expenditures, the increase in district investments due to natural resources is smaller than that of pre-colonial trading posts relatively and absolutely (Figure 9). Possible alternative explanations such as disease environment and pre- colonial political centralization have marginal or null explanatory power, showing that inferences from particularly well-known cases (e.g. Buganda kingdom in Uganda) may be unwarranted.

There are at least three areas for future research. First, descriptive results show that investments in infrastructure, education and health in British colonies were even more unequal than in French. Between-colony differences are due in part to the more redistributive and federal institutional design of French West Africa. However, British investments are also more unequal within-colony. Future research could explore those differences. In this paper I only present some correlational evidence that the British were more discriminatory in their investments insofar as, unlike the French, they invested less in districts with (i) higher pre-colonial political centralization, (ii) higher Muslim presence, and (iii) higher ethnic diversity. Those correlational findings are consistent with the notion that British practiced divide and rule to a greater extent than the French (Wucherpfennig et al., 2015).

30 Second, in a companion paper (Ricart-Huguet, 2016) I study the consequences of colonial investments for political elite formation by using novel biographical data on over 900 post-colonial ministers in those former colonies. Current explanations of cabinet formation emphasize short-term bargaining and Francois et al.(2012) argue that cabinet shares are largely proportional to population. Instead, my data reveal that some districts were represented in post-colonial cabinets much more than others, even adjusting for population. I show that district political returns—proxied by minister shares—do not derive from colonial investments or levels of development in general but from educational investments in particular. While the logic applies to both empires, district returns are type-speciﬁc: they derive only from missionary education in British colonies and only from public education in French colonies. I argue that these political leaders are a byproduct of the revenue-maximizing strategy of colonial administrators, who recruited civil servants dis- proportionately from some districts based on numeracy and literacy rather than on some ethnic group characteristic. Uneven distributions of power after the end of foreign rule have a structural long-term component—colonial human capital resulting from formal education—and they might mediate the relationship between colonialism and current political and economic development. This is precisely the third line of research, in which I examine the long-term impact of colonial investments on current outcomes such as welfare and literacy, following Huillery (2009), as well as on political participation. Examining these downstream consequences is a relevant area of research to understand current disparities in economic and political development in former colonies.

References

Acemoglu, Daron, Simon Johnson and James A. Robinson. 2002. “Reversal of fortune: Geography and institutions in the making of the modern world income distribution.” The Quarterly Journal of Economics 117(4):1231–1294. Acemoglu, Daron, Simon Johnson and James Robinson. 2001. “The colonial origins of comparative development: An empirical investigation.” The American Economic Review 91(5):1369–1401. Alsan, Marcella. 2015. “The Eﬀect of the TseTse Fly on African Development.” The American Economic Review 105(1):382–410. Banerjee, Abhijit and Lakshmi Iyer. 2005. “History, institutions, and economic performance: The legacy of colonial land tenure systems in India.” American Economic Review 95(4):1190–1213. Bartholomew, John George. 1913. A Literary and Historical Atlas of Africa and Australasia. Edinburgh: John Bartholomew and Co. Berg, Elliot J. 1960. “The Economic Basis of Political Choice in French West Africa.” American Political Science Review 54(2):391–405. Boix, Carles. 2010. “Origins and Persistence of Economic Inequality.” Annual Review of Political Science 13(1):489–516.

31 Cag´e,Julia and Valeria Rueda. 2016. “The Long-Term Eﬀects of the Printing Press in Sub-Saharan Africa.” American Economic Journal: Applied Economics 8(3):69–99.

Chi-Bonnardel, Regine Van. 1973. The Atlas of Africa. New York: The Free Press.

Constantine, Stephen. 1984. The Making of British Colonial Development Policy 1914-1940. Lon- don: Frank Cass & Co. Ltd.

Cooper, Frederick. 1996. Decolonization and African Society: The Labor Question in French and British Africa. Cambridge and New York: Cambridge University Press.

Crowder, Michael. 1964. “Indirect Rule: French and British Style.” Africa: Journal of the Inter- national African Institute 34(3):197–205.

C.S. Hammond. 1921. Hammond’s Business Atlas of Economic Geography. New York: C. S. Hammond & company, inc.

Curtin, Philip, Steven Feierman, Leonard Thompson and Jan Vansina. 1995. African History: From Earliest Times to Independence. 2nd ed. Pearson.

Darwin, John. 2012. Unﬁnished Empire. New York: Bloomsbury Press.

Davis, Donald R. and David E. Weinstein. 2002. “Bones, Bombs, and Breakpoints: The Geography of Economic Activity.” American Economic Review 92(5):1269–1289.

Delavignette, Robert. 1968. Freedom and Authority in French West Africa. London: Frank Cass & Co. Ltd.

Dell, Melissa. 2010. “The Persistent Eﬀects of Peru’s Mining Mita.” Econometrica 78(6):1863–1903.

Diamond, Jared. 2005. Guns, Germs, and Steel. New York: W. W. Norton & Company.

Echenberg, Marion. 1991. Colonial Conscripts. The Tirailleurs S´en´egalaisin French West Africa, 1857-1960. Portsmouth, NH: Heinemann.

Engerman, Stanley L. and Kenneth L. Sokoloﬀ. 2012. Economic Development in the Americas since 1500. New York: Cambridge University Press.

FAO/IIASA. 2012. “Harmonized World Soil Database (version 1.2).”. URL: http://webarchive.iiasa.ac.at/Research/LUC/External-World-soil-database/HTML/

Feyrer, James and Bruce Sacerdote. 2009. “Colonialism and Modern Income: Islands as Natural Experiments.” Review of Economics and Statistics 91(2):245–262.

Foster, Phillip. 1967. Africa South of the Sahara. London: MacMillan.

Francois, Patrick, Ilia Rainer and Francesco Trebbi. 2012. “How Is Power Shared In Africa?”.

Frankema, Ewout. 2011. “Colonial taxation and government spending in British Africa, 1880-1940: Maximizing revenue or minimizing eﬀort?” Explorations in Economic History 48(1):136–149.

Gaikwad, Nikhar. 2014. “East India Companies and Long-Term Economic Change in India.”.

32 Gardner, Leigh A. 2012. Taxing Colonial Africa: The Political Economy of British Imperialism. Oxford and New York: Oxford University Press.

Gennaioli, Nicola, Andrei Shleifer, Rafael La Porta, Florencio Lopez-de Silanes and Andrei Shleifer. 2013. “Human Capital and Regional Development.” The Quarterly Journal of Economics pp. 105–164.

Gerring, John, Daniel Ziblatt, Johan Van Gorp and Juli´anAr´evalo. 2011. “An Institutional Theory of Direct and Indirect Rule.” World Politics 63(3):377–433.

Guardado, Jenny R. 2014. “Oﬃce-Selling, Corruption and Long-term Development in Peru.”.

Haywood, Austin Hubert Wightwick. 1964. The History of the Royal West African Frontier Force. Gale & Polden, Aldershot.

Herbertson, A. J. and O. J. R. Howarth, eds. 1914. The Oxford Survey of the British Empire. Volume III. Oxford University Press.

Herbst, Jeﬀrey. 2000. States and power in africa. Princeton, New Jersey: Princeton University Press.

Hopkins, Antony Gerald. 1973. An Economic History of West Africa. New York: Routledge.

Horowitz, Donald L. 1985. Ethnic Groups in Conﬂict. Berkeley and Los Angeles: University of California Press.

Hourani, George Fadlo and John Carswell. 1995. Arab Seafaring in the Indian Ocean in Ancient and Early Medieval Times. Princeton: Princeton University Press.

Hubert, Henry. 1922. “Substances Min´eralesUtiles de l’Afrique Occidentale.”.

Huillery, Elise. 2009. “History Matters: The Long-Term Impact of Colonial Public Investments in French West Africa.” American Economic Journal: Applied Economics 1(2):176–215.

Huillery, Elise. 2010. “The impact of European settlement within French West Africa: Did pre- colonial prosperous areas fall behind?” Journal of African Economies 20(2):263–311.

Jha, Saumitra. 2012. “Trade, Institutions and Ethnic Tolerance: Evidence from South Asia.” American Political Science Review 107(4):806–832.

Jia, Ruixue. 2014. “The Legacies of Forced Freedom: China’s Treaty Ports.” The Review of Economics and Statistics 96(4):596–608.

Krugman, Paul. 1991a. “History and Industry Location : The Case of the Manufacturing Belt.” The American Economic Review 81(2):80–83.

Krugman, Paul. 1991b. “Increasing Returns and Economic Geography.” Journal of Political Econ- omy 99(3):483–499.

Kuhne, F. 1927. “Map of the mineral deposits of the world.”.

Laitin, David D. 1982. “Capitalism and hegemony: Yorubaland and the international economy.” International Organization 36(04):687–713.

33 Lawrance, Benjamin N., Emily Lynn Osborn and Richard L. Roberts. 2006. Intermediaries, In- terpreters, and Clerks: African Employees in the Making of Colonial Africa. University of Wisconsin Press.

Lawrence, Adria K. 2013. Imperial Rule and the Politics of Nationalism. New York: Cambridge University Press.

Lee, Alexander. 2012. “Comparing British and French Colonial Legacies: A Discontinuity Analysis of Cameroon.” Quarterly Journal of Political Science 7(4):365–410.

Lugard, Frederick. 1922. The Dual Mandate in British Tropical Africa. Edinburgh and London: William Blackwood and Sons.

Lysenko, A.Y. and I.N. Semashko. 1968. “Geography of malaria: A medico-geographical study of an ancient disease.”.

Mahoney, James. 2010. Colonialism and Postcolonial Development. Spanish America in Compar- ative Perspective. Cambridge University Press.

Mamdani, Mahmood. 1996. Citizen and Subject. Princeton: Princeton University Press.

Murdock, George Peter. 1959. Africa: Its Peoples and Their Culture History. New York: McGraw- Hill Book Company.

Newbury, CW. 1960. “The Formation of the Government General of French West Africa.” Journal of African history 1(1):111–128.

Nunn, Nathan. 2009. “The importance of history for economic development.” Annual Review of Economics 1(1):65–92.

Nunn, Nathan and Diego Puga. 2012. “Ruggedness: The Blessing of Bad Geography in Africa.” Review of Economics and Statistics 94(February):20–36.

Parsons, Timothy H. 1999. The African Rank-and-File. Oxford: James Currey.

Pierson, Paul. 2000. “Increasing Returns, Path Dependence, and the Study of Politics.” The American Political Science Review 94(2):251–267.

Ports.com. 2016. “Ports.com.”. URL: http://ports.com/

Ramsar. 2016. “Ramsar Sites Information Service.”. URL: https://rsis.ramsar.org/

Ricart-Huguet, Joan. 2015. “Books, Goods and Guns. The Colonial Origins of Political Elite Formation.”.

Ricart-Huguet, Joan. 2016. “Who Governs? Education as Colonial Revenue Maximization.”.

Rosenthal, Stuart S. and William C. Strange. 2004. Evidence on the Nature and Sources of Agglomeration Economies. In Handbook of Regional and Urban Economics. Vol. 4 Elsevier Inc. pp. 2119–2171.

34 Sharkey, Heather J. 2013. African Colonial States. In The Oxford Handbook of Modern African History, ed. John Parker and Reid Richard. Oxford: Oxford University Press chapter 8, p. 536.

Slave Voyages. 2013. “The Trans-Atlantic Slave Trade Database.”. URL: http://slavevoyages.org/

Stammer, D. W. 1967. “British Colonial Public Finance.” Social and Economic Studies 16(2):191– 205.

Strang, David. 1994. British and French political institutions and the patterning of decolonization. In The Comparative Political Economy of the Welfare State, ed. Thomas Janoski and Alexan- der M. Hicks. Cambridge and New York: Cambridge University Press p. 395.

Suret-Canale, Jean. 1971. French Colonialism in Tropical Africa, 1901-1945. London: Hurst Publishers.

USGS. 1921. World atlas of commercial geology. Part I. Distribution of mineral production. Wash- ington, D. C.: U.S. Geological Survey.

Wantchekon, Leonard and Omar Garc´ıa-Ponce. 2015. “Critical Junctures: Independence Move- ments and Democracy in Africa.”.

Wantchekon, Leonard and Piero Stanig. 2015. “The Curse of Good Soil? Land Fertility, Roads, and Rural Poverty in Africa.”.

Wilkinson, Steven I. 2015. Army and Nation: The Military and Indian Democracy since Indepen- dence. Harvard University Press.

Woodberry, Robert D. 2012. “The Missionary Roots of Liberal Democracy.” American Political Science Review 106(02):244–274.

World Health Organization. 2016. “Malaria.”. URL: http://www.who.int/ith/diseases/malaria/en/

Wucherpfennig, Julian, Philipp M Hunziker and Lars-erik Cederman. 2015. “Who Inherits the State? Colonial Rule and Post-Colonial Conﬂict.” American Journal of Political Science 00(0):1– 17.

Young, Crawford. 1994. The African Colonial State in Comparative Perspective. New Haven and London: Yale University Press.

35 Appendix

A Tables and ﬁgures

Table 4: Infrastructure expenditures (1910-1939 average) in East and West African districts (in 1910 FRA)

N mean sd min median max British colonies Infrastructure expenditures 200 44,086 153,461 0 2,834 1,551,032 Infrastructure exp. per 100,000 people 200 31,633 110,059 0 3,108 1,134,658 French colonies Infrastructure expenditures 112 51,240 130,562 0 12,112 1,150,341 Infrastructure exp. per 100,000 people 112 96,397 276,465 0 14,451 1,940,058

Table 5: Infrastructure expenditures (1910-1939 average) in West African districts (in 1910 FRA)

N mean sd min median max British colonies Infrastructure expenditures 66 107,974 249,909 0 20,117 1,551,032 Infrastructure exp. per 100,000 people 66 70,282 180,251 0 9,706 1,134,658 French colonies Infrastructure expenditures 112 51,240 130,562 0 12,112 1,150,341 Infrastructure exp. per 100,000 people 112 96,397 276,465 0 14,451 1,940,058

Table 6: Infrastructure expenditures by category (1910-1939 average) in East and West African districts (in 1910 FRA)

N mean sd min median max British colonies Buildings 200 18,364 79,394 0 1,175 974,900 Transportation 200 10,159 25,484 0 219 168,122 Sewage/water 200 8,784 45,625 0 0 487,848 Electricity/lighting 200 6,779 41,014 0 0 451,596 French colonies Buildings 112 22,688 87,264 0 5,854 907,381 Transportation 112 20,535 49,458 0 3,423 276,964 Sewage/water 112 7,238 30,153 0 409 220,133 Electricity/lighting 112 780 3,735 0 0 32,367

36 Table 7: First stage of 2SLS (equation 1)

(1) (2) (3) (4) (5) (6)

Natural harbor or cape indicator 0.49∗∗ 0.49∗∗ 0.48∗∗ 0.46∗∗ 0.46∗∗ 0.50∗∗ (0.10) (0.10) (0.10) (0.10) (0.09) (0.09) Locational fundamentals Distance from the coast, in 100km -0.02∗ -0.02∗ -0.02† -0.01 -0.01 -0.00 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Navigable river indicator (1910) -0.00 -0.01 0.00 -0.01 -0.00 -0.00 (0.03) (0.03) (0.03) (0.03) (0.03) (0.04) Terrain ruggedness -0.01 -0.01 -0.02 -0.03 -0.02 -0.01 (0.04) (0.04) (0.04) (0.05) (0.05) (0.09) Malaria prevalence index (1900) 0.02 0.03 0.03 0.02 0.03† 0.01 (0.02) (0.02) (0.01) (0.02) (0.01) (0.02) Tsetse fly prevalence index (1970) -0.01 -0.01 -0.01 0.01 0.00 -0.01 (0.03) (0.03) (0.03) (0.03) (0.03) (0.04) Natural resources and soil quality Gold, silver or diamonds indicator (1920) -0.03† -0.03† -0.02 -0.02 -0.02 (0.02) (0.02) (0.02) (0.02) (0.02) Base metals indicator (1920) -0.03 -0.01 0.00 0.01 -0.00 (0.03) (0.02) (0.02) (0.02) (0.02) Soil quality index (2000) -0.02 -0.02 -0.02 -0.02 -0.03 (0.03) (0.03) (0.02) (0.02) (0.02) (Pre-)Colonial socioeconomic characteristics European population, logged 0.01 0.01† (0.01) (0.00) African population, logged 0.01 0.01 0.02 (0.02) (0.02) (0.03) Area in km2, logged -0.07∗ -0.05∗ -0.08∗ (0.03) (0.02) (0.03) Prevalence of Islam (1910) 0.02 0.02 -0.01 (0.03) (0.03) (0.03) Ethnic Fractionalization Index 0.06 0.04 0.05 (0.07) (0.06) (0.08) Agriculture (none to irrigation) -0.03∗ -0.03∗ -0.03∗ (0.01) (0.01) (0.01) Settlements (nomadic to complex) -0.02† -0.01 -0.01 (0.01) (0.01) (0.01) Pre-colonial political centralization 0.02 0.02 -0.01 (0.02) (0.02) (0.02) Slavery (absence to prevalent) 0.01 0.01 0.02 (0.02) (0.02) (0.02) Constant 0.14 0.28 0.21 0.89∗ 0.63 0.88 (0.09) (0.21) (0.23) (0.41) (0.41) (0.51) Observations 312 312 301 312 301 228 Adjusted R2 0.31 0.31 0.33 0.36 0.36 0.38 Notes: † p < 0.10, * p < 0.05, ** p < 0.01. Clustered standard errors by colony in parentheses. All models include colony fixed effects The first stage is identical for all second stages, hence only one first stage is printed in the paper.

37 Table 8: Logged district infrastructure expenditure (1919-1939, in 1910 FRA)

(1) (2) (3) (4) (5) (6) Pre-colonial trading post indicator 2.97∗∗ 3.03∗∗ -0.70 3.66∗∗ -0.09 3.35∗ (1.02) (1.05) (0.86) (1.34) (1.24) (1.33) European population, logged 1.19∗∗ 1.17∗∗ (0.15) (0.12) Locational fundamentals Distance from the coast, in 100km -0.11 -0.09 -0.05 -0.24∗ -0.18∗ -0.22∗∗ (0.13) (0.11) (0.08) (0.10) (0.08) (0.05) Navigable river indicator (1910) -0.28 -0.20 -0.03 -0.27 -0.06 -0.34 (0.47) (0.42) (0.35) (0.43) (0.40) (0.45) Terrain ruggedness 0.84∗∗ 0.88∗∗ 0.40 0.81∗∗ 0.48 0.50† (0.27) (0.28) (0.30) (0.23) (0.52) (0.26) Malaria prevalence index (1900) -0.32 -0.41 0.32 -0.32 0.47 -0.46 (0.35) (0.33) (0.35) (0.30) (0.32) (0.43) Tsetse fly prevalence index (1970) -0.21 -0.20 -0.06 -0.49 -0.20 -0.13 (0.32) (0.32) (0.27) (0.32) (0.26) (0.29) Natural resources and soil quality Gold, silver or diamonds indicator (1920) 0.73∗ 0.63† 0.48† 0.51† 0.44† (0.35) (0.37) (0.25) (0.28) (0.24) Base metals indicator (1920) 0.37 -0.06 0.14 -0.20 0.20 (0.33) (0.25) (0.35) (0.33) (0.35) Soil quality index (2000) 0.32 0.14 0.22 0.08 0.21 (0.22) (0.29) (0.24) (0.30) (0.29) (Pre-)Colonial socioeconomic characteristics African population, logged 0.57∗∗ 0.09 0.52∗ (0.16) (0.24) (0.26) Area in km2, logged 0.67∗∗ 0.62∗ 0.51∗ (0.18) (0.29) (0.25) Prevalence of Islam (1910) -0.63∗ -0.37 -0.38 (0.30) (0.27) (0.25) Ethnic Fractionalization Index -1.33 -0.63 -0.97 (1.29) (0.83) (1.66) Agriculture (none to irrigation) 0.33 0.23 0.68∗∗ (0.29) (0.20) (0.23) Settlements (nomadic to complex) 0.07 -0.03 0.21 (0.19) (0.17) (0.18) Pre-colonial political centralization 0.23 0.24 -0.06 (0.27) (0.25) (0.32) Slavery (absence to prevalent) -0.49 -0.24 -0.90† (0.41) (0.34) (0.48) Constant 5.44∗∗ 3.46∗ -1.09 -7.83∗ -7.48 -5.94 (1.70) (1.47) (1.91) (3.96) (5.12) (5.63) Observations 312 312 301 312 301 228 Adjusted R2 0.35 0.35 0.53 0.39 0.55 0.41 Wald F Statistic 24.88 24.41 25.38 22.93 25.93 28.34 Notes: † p < 0.10, * p < 0.05, ** p < 0.01. Clustered standard errors by colony in parentheses. All models include colony fixed effects

38 Table 9: Presence of a colonial railroad

(1) (2) (3) (4) (5) (6) Pre-colonial trading post indicator -0.01 0.03 -0.38 -0.05 -0.45 0.18 (0.23) (0.24) (0.28) (0.25) (0.28) (0.18) European population, logged 0.14∗∗ 0.14∗∗ (0.01) (0.02) Locational fundamentals Distance from the coast, in 100km -0.04† -0.04† -0.02† -0.04∗ -0.03∗ -0.04∗∗ (0.02) (0.02) (0.01) (0.02) (0.01) (0.02) Navigable river indicator (1910) -0.21∗∗ -0.20∗∗ -0.17∗∗ -0.21∗∗ -0.18∗∗ -0.21∗∗ (0.07) (0.07) (0.05) (0.06) (0.05) (0.07) Terrain ruggedness 0.03 0.05 -0.01 0.02 -0.02 0.05 (0.08) (0.09) (0.09) (0.08) (0.06) (0.15) Malaria prevalence index (1900) -0.08∗ -0.09∗∗ -0.01 -0.10∗∗ -0.02 -0.16∗∗ (0.03) (0.03) (0.03) (0.04) (0.04) (0.05) Tsetse fly prevalence index (1970) -0.03 -0.04 -0.02 -0.06 -0.02 -0.01 (0.05) (0.05) (0.04) (0.05) (0.05) (0.05) Natural resources and soil quality Gold, silver or diamonds indicator (1920) 0.22∗ 0.21∗ 0.20∗ 0.20∗∗ 0.21∗ (0.09) (0.08) (0.08) (0.08) (0.08) Base metals indicator (1920) -0.01 -0.02 0.00 -0.00 -0.01 (0.06) (0.06) (0.07) (0.06) (0.07) Soil quality index (2000) 0.03 0.01 0.02 0.01 0.00 (0.03) (0.03) (0.03) (0.03) (0.03) (Pre-)Colonial socioeconomic characteristics African population, logged 0.08∗ 0.02 0.11∗∗ (0.04) (0.03) (0.04) Area in km2, logged -0.04 -0.04 0.03 (0.05) (0.03) (0.03) Prevalence of Islam (1910) -0.02 0.03 0.11† (0.08) (0.07) (0.06) Ethnic Fractionalization Index -0.26∗ -0.19† -0.40∗∗ (0.12) (0.10) (0.12) Agriculture (none to irrigation) 0.03 0.03 0.03 (0.03) (0.03) (0.04) Settlements (nomadic to complex) 0.01 -0.00 0.04∗ (0.02) (0.02) (0.01) Pre-colonial political centralization 0.04 0.03 0.02 (0.04) (0.03) (0.05) Slavery (absence to prevalent) -0.03 -0.02 -0.03 (0.04) (0.04) (0.04) Constant 0.77∗∗ 0.60∗∗ -0.01 0.19 0.17 -0.64 (0.21) (0.20) (0.21) (0.65) (0.30) (0.55) Observations 312 312 301 312 301 228 Adjusted R2 0.19 0.20 0.33 0.23 0.32 0.23 Wald F Statistic 24.88 24.41 25.38 22.93 25.93 28.34 Notes: † p < 0.10, * p < 0.05, ** p < 0.01. Clustered standard errors by colony in parentheses. All models include colony fixed effects

39 Table 10: Logged district students (1919-1939)

(1) (2) (3) (4) (5) (6) Pre-colonial trading post indicator 4.46∗∗ 4.38∗∗ 2.25∗ 4.19∗ 2.39† 3.97∗ (1.31) (1.35) (1.02) (1.64) (1.36) (1.65) European population, logged 0.67∗∗ 0.55∗∗ (0.11) (0.14) Locational fundamentals Distance from the coast, in 100km -0.02 -0.01 -0.00 -0.05 -0.03 -0.14∗∗ (0.10) (0.10) (0.08) (0.07) (0.05) (0.05) Navigable river indicator (1910) -0.27 -0.19 -0.01 -0.21 -0.07 -0.09 (0.23) (0.24) (0.13) (0.22) (0.18) (0.21) Terrain ruggedness 0.51 0.49 0.28 0.38 0.28 0.94† (0.41) (0.40) (0.20) (0.43) (0.25) (0.48) Malaria prevalence index (1900) 0.02 -0.02 0.40∗∗ -0.16 0.20 -0.22 (0.19) (0.15) (0.14) (0.10) (0.14) (0.16) Tsetse fly prevalence index (1970) 0.16 0.19 0.25 -0.11 0.04 -0.33 (0.31) (0.29) (0.31) (0.23) (0.23) (0.26) Natural resources and soil quality Gold, silver or diamonds indicator (1920) 0.11 0.07 -0.12 -0.09 0.01 (0.30) (0.23) (0.23) (0.19) (0.30) Base metals indicator (1920) 0.46† 0.16 0.30 0.06 0.13 (0.25) (0.18) (0.21) (0.17) (0.19) Soil quality index (2000) 0.27∗ 0.16 0.13 0.06 0.11 (0.13) (0.13) (0.14) (0.12) (0.13) (Pre-)Colonial socioeconomic characteristics African population, logged 0.83∗∗ 0.61∗ 0.70∗ (0.21) (0.24) (0.30) Area in km2, logged 0.10 0.01 0.20 (0.18) (0.09) (0.15) Prevalence of Islam (1910) -0.77∗∗ -0.59∗∗ -0.54∗ (0.20) (0.20) (0.27) Ethnic Fractionalization Index -1.00∗ -0.65∗ -1.04∗ (0.39) (0.25) (0.45) Agriculture (none to irrigation) 0.24∗∗ 0.22† 0.11 (0.08) (0.13) (0.10) Settlements (nomadic to complex) 0.16 0.07 0.09 (0.10) (0.12) (0.07) Pre-colonial political centralization -0.16 -0.17 -0.16 (0.16) (0.13) (0.21) Slavery (absence to prevalent) -0.02 0.12 -0.12 (0.13) (0.12) (0.13) Constant 0.97 -0.78 -3.16∗ -9.61∗∗ -8.57∗ -7.16† (1.38) (1.74) (1.59) (2.95) (3.51) (4.20) Observations 288 288 277 288 277 219 Adjusted R2 0.39 0.40 0.57 0.50 0.61 0.54 Wald F Statistic 28.04 26.47 27.22 22.94 26.09 29.25 Notes: † p < 0.10, * p < 0.05, ** p < 0.01. Clustered standard errors by colony in parentheses. All models include colony fixed effects

40 Table 11: Logged district health staﬀ (1919-1939)

(1) (2) (3) (4) (5) (6) Pre-colonial trading post indicator 2.66∗∗ 2.64∗∗ 1.34∗ 2.81∗∗ 1.55∗ 3.10∗∗ (0.93) (0.94) (0.63) (0.99) (0.65) (1.07) European population, logged 0.42∗∗ 0.38∗∗ (0.06) (0.06) Locational fundamentals Distance from the coast, in 100km 0.01 0.00 0.01 -0.04 -0.02 -0.03 (0.06) (0.05) (0.03) (0.04) (0.04) (0.04) Navigable river indicator (1910) 0.08 0.10 0.15 0.05 0.12 0.01 (0.17) (0.19) (0.13) (0.16) (0.13) (0.17) Terrain ruggedness 0.06 0.01 -0.16 -0.02 -0.14 0.12 (0.16) (0.15) (0.11) (0.17) (0.13) (0.27) Malaria prevalence index (1900) -0.10 -0.10 0.16∗ -0.18∗∗ 0.07 -0.24∗∗ (0.07) (0.07) (0.07) (0.07) (0.05) (0.08) Tsetse fly prevalence index (1970) 0.08 0.08 0.13 -0.03 0.07 0.18 (0.18) (0.18) (0.13) (0.20) (0.15) (0.19) Natural resources and soil quality Gold, silver or diamonds indicator (1920) 0.02 -0.01 -0.07 -0.06 -0.07 (0.15) (0.10) (0.16) (0.12) (0.17) Base metals indicator (1920) 0.30∗ 0.16 0.21 0.12 0.17 (0.12) (0.12) (0.13) (0.13) (0.16) Soil quality index (2000) 0.02 -0.05 -0.01 -0.05 0.03 (0.06) (0.05) (0.06) (0.04) (0.06) (Pre-)Colonial socioeconomic characteristics African population, logged 0.36∗∗ 0.20∗∗ 0.44∗∗ (0.10) (0.07) (0.13) Area in km2, logged 0.10 0.07 0.12 (0.10) (0.06) (0.11) Prevalence of Islam (1910) -0.17 -0.06 0.01 (0.12) (0.07) (0.12) Ethnic Fractionalization Index -0.39† -0.13 -0.26 (0.23) (0.21) (0.26) Agriculture (none to irrigation) 0.23∗∗ 0.20† 0.14 (0.08) (0.11) (0.09) Settlements (nomadic to complex) 0.05 0.02 0.01 (0.05) (0.05) (0.05) Pre-colonial political centralization -0.09 -0.09 0.02 (0.09) (0.10) (0.06) Slavery (absence to prevalent) -0.13 -0.05 -0.12 (0.11) (0.10) (0.15) Constant 0.46 0.32 -1.30∗∗ -4.06∗ -3.70∗∗ -5.27∗ (0.58) (0.70) (0.47) (2.06) (0.87) (2.11) Observations 311 311 300 311 300 227 Adjusted R2 0.30 0.31 0.56 0.37 0.58 0.26 Wald F Statistic 18.05 17.67 19.03 15.96 18.48 19.12 Notes: † p < 0.10, * p < 0.05, ** p < 0.01. Clustered standard errors by colony in parentheses. All models include colony fixed effects

41 Table 12: Logged number of missions per district (1919-1939)

(1) (2) (3) (4) (5) (6) Pre-colonial trading post indicator 0.95∗∗ 0.91∗∗ 0.55† 0.90∗∗ 0.64∗ 1.07∗∗ (0.28) (0.29) (0.29) (0.27) (0.28) (0.20) European population, logged 0.12∗∗ 0.08∗∗ (0.03) (0.02) Locational fundamentals Distance from the coast, in 100km -0.02 -0.02 -0.01 -0.05† -0.04 -0.03 (0.03) (0.03) (0.03) (0.03) (0.03) (0.04) Navigable river indicator (1910) 0.05 0.06 0.07 0.02 0.03 -0.02 (0.06) (0.06) (0.05) (0.05) (0.05) (0.05) Terrain ruggedness 0.21 0.17 0.13 0.14 0.11 0.53∗∗ (0.25) (0.24) (0.19) (0.20) (0.17) (0.19) Malaria prevalence index (1900) -0.07 -0.07 0.00 -0.14∗ -0.09† -0.16∗∗ (0.07) (0.07) (0.07) (0.06) (0.05) (0.04) Tsetse fly prevalence index (1970) 0.15∗ 0.15∗ 0.17∗ 0.07 0.10† 0.08 (0.06) (0.06) (0.07) (0.05) (0.05) (0.07) Natural resources and soil quality Gold, silver or diamonds indicator (1920) -0.08 -0.09 -0.15 -0.15 -0.11 (0.08) (0.08) (0.11) (0.11) (0.10) Base metals indicator (1920) 0.21∗∗ 0.17∗ 0.15† 0.14 0.05 (0.08) (0.08) (0.08) (0.09) (0.08) Soil quality index (2000) 0.02 0.00 -0.01 -0.02 -0.00 (0.04) (0.04) (0.04) (0.04) (0.04) (Pre-)Colonial socioeconomic characteristics African population, logged 0.19∗∗ 0.15∗∗ 0.19∗∗ (0.05) (0.05) (0.05) Area in km2, logged 0.04 0.03 -0.00 (0.04) (0.04) (0.03) Prevalence of Islam (1910) -0.15∗∗ -0.13∗ -0.13† (0.06) (0.06) (0.07) Ethnic Fractionalization Index -0.09 -0.04 -0.07 (0.17) (0.15) (0.17) Agriculture (none to irrigation) 0.17∗∗ 0.16∗∗ 0.15† (0.06) (0.06) (0.09) Settlements (nomadic to complex) 0.07∗∗ 0.06∗ 0.03 (0.02) (0.02) (0.03) Pre-colonial political centralization -0.01 -0.01 -0.07 (0.06) (0.07) (0.05) Slavery (absence to prevalent) -0.05 -0.04 -0.11 (0.07) (0.07) (0.10) Constant 1.19∗∗ 1.06∗∗ 0.60 -1.45∗ -1.33∗ -0.79 (0.26) (0.36) (0.44) (0.65) (0.67) (0.94) Observations 312 312 301 312 301 228 Adjusted R2 0.38 0.39 0.45 0.49 0.51 0.48 Wald F Statistic 24.88 24.41 25.38 22.93 25.93 28.34 Notes: † p < 0.10, * p < 0.05, ** p < 0.01. Clustered standard errors by colony in parentheses. All models include colony fixed effects

42 Figure 12: Persistence of public investments in British East and West Africa

Schools, logged Students, logged

pre-1930 pre-1930

5 10

1938 5 1938

0 0

Infrastructure investments, logged Health staff, logged

pre-1930 pre-1930

15 6

10 4 1938 1938 5 2

0 0

Table 13: Logged district infrastructure expenditure (1919-1939, in 1910 FRA)

(1) (2) (3) First British/French pre-colonial trading post 1.83∗ 1.76† (0.69) (0.96) Distance from first British/French trading post, in 100km -0.18† -0.15∗ (0.09) (0.07) First British/French pre-colonial school 1.30 0.10 (0.87) (1.05) Distance from first British/French school, in 100km -0.13 -0.05 (0.11) (0.05) Constant 10.08∗∗ 9.98∗∗ 10.14∗∗ (0.31) (0.36) (0.37) Observations 312 312 312 Adjusted R2 0.37 0.35 0.36 Notes: † p < 0.10, * p < 0.05, ** p < 0.01. Clustered standard errors by colony in parentheses. All models include colony fixed effects

43 Table 14: Logged district students (1919-1939)

(1) (2) (3) First British/French pre-colonial trading post 1.03† 0.81† (0.51) (0.42) Distance from first British/French trading post, in 100km -0.18∗ -0.17∗∗ (0.06) (0.05) First British/French pre-colonial school 0.94† 0.35 (0.52) (0.39) Distance from first British/French school, in 100km -0.11 -0.02 (0.08) (0.03) Constant 5.97∗∗ 5.76∗∗ 5.99∗∗ (0.20) (0.26) (0.25) Observations 288 288 288 Adjusted R2 0.43 0.41 0.43 Notes: † p < 0.10, * p < 0.05, ** p < 0.01. Clustered standard errors by colony in parentheses. All models include colony fixed effects

Table 15: Logged district health staﬀ (1919-1939)

(1) (2) (3) First British/French pre-colonial trading post 0.86∗ 0.71† (0.35) (0.35) Distance from first British/French trading post, in 100km -0.08∗∗ -0.06∗∗ (0.02) (0.02) First British/French pre-colonial school 0.71† 0.23 (0.35) (0.29) Distance from first British/French school, in 100km -0.07∗ -0.03∗ (0.03) (0.01) Constant 2.22∗∗ 2.18∗∗ 2.25∗∗ (0.08) (0.10) (0.07) Observations 311 311 311 Adjusted R2 0.37 0.35 0.37 Notes: † p < 0.10, * p < 0.05, ** p < 0.01. Clustered standard errors by colony in parentheses. All models include colony fixed effects

44 Table 16: Logged number of missions per district (1919-1939)

(1) (2) (3) First British/French pre-colonial trading post 0.08 -0.07 (0.18) (0.15) Distance from first British/French trading post, in 100km -0.04† -0.02† (0.02) (0.01) First British/French pre-colonial school 0.18 0.22 (0.16) (0.15) Distance from first British/French school, in 100km -0.05∗∗ -0.04∗∗ (0.02) (0.01) Constant 0.26∗∗ 0.28∗∗ 0.31∗∗ (0.06) (0.05) (0.06) Observations 312 312 312 Adjusted R2 0.37 0.38 0.38 Notes: † p < 0.10, * p < 0.05, ** p < 0.01. Clustered standard errors by colony in parentheses. All models include colony fixed effects

Table 17: Disparities in educational investments per district (1915-1939)

(1) (2) (3) (4) levels (eq. 5) FD (eq. 6) levels (eq. 5) FD (eq. 6) Teachers, lagged 0.86∗∗ (0.03) Teachers in 1915 0.17∗∗ 0.19∗∗ (0.04) (0.07) Teachers, logged ﬁrst lag 0.64∗∗ (0.04) Teachers in 1915, logged 0.31∗∗ -0.01 (0.04) (0.03) Constant -0.04 0.01 0.08∗ 0.10∗ (0.09) (0.31) (0.03) (0.05) Observations 374 749 374 749 Standard errors in parentheses † p < 0.10, ∗ p < 0.05, ∗∗ p < 0.01

45 Table 18: Disparities in health investments per district (1915-1939)

(1) (2) (3) (4) levels (eq. 5) FD (eq. 6) levels (eq. 5) FD (eq. 6) Health staff, lagged 0.40∗∗ (0.08) Health staff in 1915 0.82∗∗ 0.04 (0.09) (0.10) Health staff, logged first lag 0.61∗∗ (0.06) Health staff in 1915, logged 0.35∗∗ 0.01 (0.06) (0.04) Constant -0.70 0.08 0.05 -0.01 (0.44) (0.90) (0.06) (0.07) Observations 178 305 178 305 Standard errors in parentheses † p < 0.10, ∗ p < 0.05, ∗∗ p < 0.01

Table 19: Disparities in infrastructure investments per district (1915-1939, in 1910 FRA)

(1) (2) (3) (4) levels (eq. 5) FD (eq. 6) levels (eq. 5) FD (eq. 6) Infrastructure investments, lagged 0.32∗∗ (0.07) Infrastructure investments in 1915 0.27∗ -0.45∗∗ (0.13) (0.12) Infrastructure investments, logged ﬁrst lag 0.67∗∗ (0.07) Infrastructure investments in 1915, logged 0.11 -0.01 (0.07) (0.07) Constant 9770.81∗ 1795.77 1.71∗∗ -0.23 (3991.53) (4288.49) (0.32) (0.38) Observations 181 223 181 223 Standard errors in parentheses † p < 0.10, ∗ p < 0.05, ∗∗ p < 0.01

Table 20: Average colonial district infrastructure expenditures (in 1910 FRA)

Pre-colonial trading post Yes No 357,723FRA 96,441FRA Yes Natural harbor n = 12 n = 9 or cape 189,235FRA 38,416FRA No n = 10 n = 25 Some districts (e.g. Accra, Ouidah) without natural harbors or capes nonetheless were centers of pre-colonial trade. That became easier with 19th century technological advances in the Age of Steam. While the number of districts in each cell is small, colonial investments were relatively low in places that did were not centers of pre-colonial trade even if they possessed a natural harbor or cape (e.g. Boke).

46 Table 21: Covariate balance between coastal districts with pre-colonial trading posts (Yes) and those without (No)

(1) (2) (3) No Yes p-value Natural harbor or cape indicator 0.265 0.545 0.034 Number of ethnic groups in the district 4.412 2.909 0.069 Gathering 0.202 0.061 0.090 Hunting 0.706 0.949 0.102 Fishing 1.298 1.267 0.884 Intensity of agriculture (none to irrigation) 3.168 2.906 0.261 Settlement patterns (nomadic to complex) 6.275 6.565 0.515 Political centralization (acephalous to kingdoms) 2.282 2.465 0.476 Slavery (absence to prevalent) 2.251 2.287 0.836 Prevalence of Islam (1910) 0.618 0.545 0.753 Malaria prevalence index (1900) 2.911 3.233 0.134 Tsetse ﬂy prevalence index (1970) 1.663 1.858 0.472 N=56. 22 districts had a pre-colonial trading posts; 34 did not. The table shows balance along a set of mostly pre-colonial covariates except in the instrument, as expected, and marginally (p < 0.1) in the number of ethnic groups and in the importance of gathering as an economic activity as deﬁned in Murdock (1959).

Table 22: Covariate balance between coastal districts with natural harbor or capes (Yes) and those without (No)

(1) (2) (3) No Yes p-value Number of ethnic groups in the district 4.257 3.095 0.167 Gathering 0.104 0.218 0.180 Hunting 0.869 0.690 0.234 Fishing 1.106 1.585 0.022 Intensity of agriculture (none to irrigation) 2.936 3.279 0.143 Settlement patterns (nomadic to complex) 6.248 6.626 0.400 Political centralization (acephalous to kingdoms) 2.343 2.372 0.909 Slavery (absence to prevalent) 2.268 2.262 0.972 Prevalence of Islam (1910) 0.543 0.667 0.592 Malaria prevalence index (1900) 3.104 2.927 0.419 Tsetse fly prevalence index (1970) 1.822 1.603 0.420 N=56. 21 districts have a natural harbor or cape; 35 do not. The table shows balance along a set of mostly pre-colonial covariates except in fishing, measured as percentage of the population of the ethnic group engaged in fishing as defined in Murdock (1959).

47 Figure 13: Public expenditures by district in British colonies (1920-1938, in 1910 FRA)

Ghana Kenya 15,000 1.5e+06 10,000 1.0e+06 5,000 500000 0 0 Digo Teita Meru Kilifi Kitui Ho EmbuLamuNandi Uasin Wa Ada Elgeyo Ravine BaringoForthallLaikipia Kisumu KiambuNairobiNakuru SefwiKetaBirim WestsukTurkana KrachiLawra Gonja Accra TanariverNaivasha Northnyeri MachakosMasaikenMombasa Wenchi ObuasiSunyaniBekwai Ashanti WasawKumasi Ahanta Southnyeri Transnzoia Voltariver Saltpond WinnebaAkwapim Navrongo MampongMamprusi Dagomba Capecoast NorthkavirondoSouthkavirondo Northernfrontier Westernakim Centralkavirondo Malawi Nigeria 30,000 800000 600000 20,000 400000 10,000 200000 0 0

Kota Oyo Dowa Ncheu Dedza Cholo Warri Ondo Niger Ijebu Ilorin Kano Zaria Likoma Mlanje Zomba Kabba BornuBauchi Benue Benin Ogoja Owerri Lagos Karonga Mzimba Blantyre Lilongwe KatsinaSokoto Calabar PlateauOnitsha Chinteche Lowerriver Adamawa Abeokuta Fortjohnston Cameroons

Sierra Leone Tanzania 150000 250000 200000 100000 150000 100000 50,000 50,000 0

0 Pare Ufipa Rufiji Lindi Kilwa Iringa Mbulu MoshiTanga Kilosa MasasiNewala KondoaSingidaNjombeMaswaUlanga ArushaSongeaKigoma TaboraBukobaMbeya Bo MasaitzaRungweTunduruHandeniMusoma KwimbaKahama Mwanza DodomaManyoni Mikindani Mpwapwa Morogoro Kono Biharamulo BagamoyoUsambaraShinyanga KareneTonkolili Sherbro Bonthe Bombali Daressalaam Kenema Pujehun Kailahun Portloko Freetown Koinadugu Moyamba Uganda Zambia 400000 80,000 60,000 300000 40,000 200000 20,000 100000 0 0 Mpika Isoka Ndola ChinsaliLundazi Mkushi SerenjeKalabo SolweziMongu KalomoLusaka Toro Balovale LuwinguMankoya MumbwaPetaukeSenangaSesheke Kasama Abercorn Teso Kigezi Acholi Namwala Kasempa BrokenhillMazabuka LangoAnkole Central Mengo Kawambwa Mporokoso Livingstone Masaka BudamaBunyoro Busoga Fortrosebery Fortjameson WestnileKaramoja Mubende

48 Figure 14: Public expenditures by district in French colonies (1910-1939, in 1910 FRA)

Benin Burkina Faso 20,000 150000 15,000 100000 10,000 50,000 5,000 0 0 Say Kaya Fada Dori Gaoua Mono Allada Ouidah Borgou Dedougou Djougou Atacora Savalou Cotonou Abomey KoudougouTenkodogo Ouahigouya Portonovo Moyenniger Bobodioulasso Ouagadougou Cote d'Ivoire Guinea 150000 1.5e+06 100000 1.0e+06 500000 50,000 0 0

Pita Man Labe Boffa Boke Daloa Kong Siguiri Beyla Kindia GuigloGourosLahou Indenie Tabou Assinie Baoule Dabola MamouKankan Odienne Seguela Agneby BassamLagunes Koumbia Macenta Conakry Nzicomoe Nzerekore KouroussaForecariah BondoukouTagouanas Sassandra GueckedouKissidougou

Mali Mauritania 30,000 250000 200000 20,000 150000 100000 10,000 50,000 0 0

Gao San Kita Nara Adrar Nema Nioro Mopti Gorgol Tagant Trarza Macina Segou Kayes Assaba Brakna Gourma Niafunke Koutiala Sikasso Bamako GoundamBafoulabe Bougouni Guidimaka Satadougou Bandiagara Tombouctou Baiedulevrier Niger Senegal 8,000 400000 300000 6,000 200000 4,000 100000 2,000 0 0 Baol Bakel Podor Louga Thies Dakar Dagana Matam Bilma Tivaouane Saintlouis Dosso Konny Zinder Goure Casamance Sinesaloum Nguigmi Tahoua Agadez Niamey Hautegambie Tessaoua Tambacounda

49 Figure 15: Students by district in British colonies (1920-1938))

Ghana Kenya 4,000 15,000 3,000 10,000 2,000 5,000 1,000 0 0

Digo Kilifi Kitui Ho Wa Lamu EmbuUasinNandiTeita Meru Ada Keta Elgeyo NairobiRavine Nakuru Gonja Sefwi Birim Accra Laikipia Turkana Baringo Kiambu Forthall Kisumu KrachiLawra Obuasi Ahanta Masaiken TanariverNaivashaWestsuk Bekwai Wenchi SunyaniWasawAshanti Kumasi Northnyeri MachakosMombasaSouthnyeri VoltariverSaltpondWinneba Akwapim Transnzoia Navrongo Mamprusi Dagomba Mampong Capecoast Westernakim Northernfrontier SouthkavirondoNorthkavirondo Centralkavirondo Malawi Nigeria 1 15,000 10,000 5,000 -1 0

Kota Oyo Cholo Dedza Dowa Ncheu Niger Ilorin ZariaKano Warri IjebuOndo Benin Likoma Mlanje Zomba Benue KabbaOgojaBornuBauchi Lagos Owerri Blantyre Karonga Lilongwe Mzimba Plateau Katsina Sokoto Onitsha Calabar Chinteche Lowerriver Adamawa Abeokuta Fortjohnston Cameroons

Sierra Leone Tanzania 8,000 6,000 6,000 4,000 4,000 2,000 2,000 0 0 Pare Lindi Bo UfipaRufiji Kilwa MoshiIringa Maswa UlangaMbulu Kilosa MasasiMbeya Arusha TaboraTanga Kono HandeniKondoaKwimbaManyoniNewalaNjombeTunduru Singida KahamaRungwe SongeaKigomaBukoba Mwanza Masaitza Mikindani Musoma Morogoro Dodoma Tonkolili Karene PortlokoBombali Bonthe Sherbro Shinyanga UsambaraBagamoyo Mpwapwa Pujehun Kenema Kailahun Freetown Biharamulo Koinadugu Moyamba Daressalaam Uganda Zambia 15,000 600 400 10,000 200 5,000 0 0

Isoka Mpika Ndola Toro Chinsali Kalabo Lundazi MkushiMongu SerenjeSolwezi Kalomo Lusaka Teso AbercornBalovale Kasama LuwinguMankoya MumbwaPetaukeSenangaSesheke Kigezi Acholi Lango AnkoleCentral Mengo Kasempa Namwala Brokenhill Budama Bunyoro BusogaMasaka Mporokoso MazabukaLivingstone Karamoja Mubende Westnile Fortrosebery Kawambwa Fortjameson

Note: British colonial records do not provide disaggregated education data for Malawi.

50 Figure 16: Students in French colonies (1910-1939)

Benin Burkina Faso 250 1,500 200 1,000 150 100 500 50 0 0 Say Kaya Fada Dori Gaoua Mono Borgou Allada Ouidah Dedougou Djougou Atacora Savalou Cotonou Abomey TenkodogoKoudougou Ouahigouya Portonovo Moyenniger BobodioulassoOuagadougou Cote d'Ivoire Guinea 1,500 1,000 1,000 500 500 0 0

Pita Man Beyla Boffa Boke Labe Daloa Lahou Kong Dabola SiguiriKindia TabouGouros Guiglo Indenie Assinie Baoule KankanMamou Conakry Seguela Odienne BassamAgneby Lagunes MacentaKoumbia Nzicomoe KouroussaForecariah Nzerekore Tagouanas Bondoukou Sassandra Gueckedou Kissidougou Mali Mauritania 1 1,000 500 -1 0

Gao San Nara Kita Adrar Nema Nioro Mopti Gorgol Trarza Macina Segou Kayes Assaba Brakna Tagant Gourma Koutiala Niafunke Sikasso Bamako GoundamBougouniBafoulabe Guidimaka Satadougou BandiagaraTombouctou Baiedulevrier Niger Senegal 150 4,000 3,000 100 2,000 50 1,000 0 0 Baol Bakel Louga Podor Thies Dakar MatamDagana Bilma Konny Dosso Goure Zinder Tivaouane Saintlouis Agadez Tahoua Nguigmi Niamey SinesaloumCasamance Tessaoua HautegambieTambacounda

Note: Student data for Mauritania is missing in Huillery (2009)

51 Figure 17: Public health staﬀ by district in British colonies (1920-1938))

Ghana Kenya 150 150 100 100 50 50 0 0 Digo TeitaMeru KilifiKitui Ho Embu Lamu Nandi Uasin Ada Wa ElgeyoLaikipia Ravine Baringo Kiambu NakuruForthall KisumuNairobi Sefwi Keta Birim WestsukTurkana Krachi Lawra Gonja Accra MasaikenNaivashaNorthnyeri Tanariver MachakosMombasa Ashanti Obuasi Bekwai Wenchi SunyaniWasaw AhantaKumasi Transnzoia Southnyeri Saltpond Voltariver AkwapimWinneba Mampong Navrongo Mamprusi CapecoastDagomba NorthkavirondoSouthkavirondo Northernfrontier Westernakim Centralkavirondo Malawi Nigeria 10 80 8 60 6 40 4 20 2 0 0

Kota Oyo Cholo Dedza Dowa Ncheu Ijebu Ondo IlorinNiger Warri ZariaKano Likoma Mlanje Zomba BauchiBenueBornu Ogoja BeninKabba Owerri Lagos Karonga Lilongwe Mzimba Blantyre Katsina Sokoto OnitshaPlateau Calabar Chinteche Lowerriver Adamawa Abeokuta Fortjohnston Cameroons

Sierra Leone Tanzania 50 80 40 60 30 40 20 20 10 0 0 Pare Lindi Bo RufijiIringa UfipaKilwa Moshi MasasiMaswa Mbulu Ulanga Arusha Kilosa MbeyaTangaTabora Kono HandeniKwimbaManyoni NewalaNjombeTunduruKigoma SingidaKondoaSongea RungweKahamaMwanzaBukoba Masaitza Mikindani MorogoroMusomaDodoma Karene Portloko SherbroTonkoliliBombali Bonthe Mpwapwa Bagamoyo ShinyangaUsambara Kailahun Kenema Pujehun Freetown Biharamulo Koinadugu Moyamba Daressalaam Uganda Zambia 6 250 200 4 150 2 100 50 0 0

Isoka Mpika Ndola Toro Chinsali KalaboKalomo Lundazi Mkushi SerenjeSolweziMongu Lusaka Teso AbercornBalovale LuwinguMankoya MumbwaPetaukeSenangaSeshekeKasama Lango AnkoleCentral Acholi Kigezi Mengo Kasempa Namwala Brokenhill Budama Masaka Busoga Bunyoro MazabukaMporokoso Livingstone KaramojaMubende Westnile FortjamesonFortrosebery Kawambwa

52 Figure 18: Public health staﬀ in French colonies (1910-1939)

Benin Burkina Faso 40 50 40 30 30 20 20 10 10 0 0 Say Fada Dori Kaya Gaoua Mono Allada Borgou Ouidah Dedougou Atacora Djougou Savalou Abomey Cotonou TenkodogoOuahigouya Koudougou Portonovo Moyenniger BobodioulassoOuagadougou Cote d'Ivoire Guinea 40 15 30 10 20 5 10 0 0

Pita Man Beyla Boffa Labe Boke Daloa KongLahou SiguiriDabola Kindia Guiglo Indenie Assinie Tabou Gouros Baoule Kankan Mamou Conakry SeguelaOdienne Agneby Bassam Lagunes Macenta Koumbia Nzicomoe Nzerekore Forecariah Kouroussa Tagouanas Bondoukou Sassandra Gueckedou Kissidougou Mali Mauritania 6 80 60 4 40 2 20 0 0

Kita Gao San Nara Adrar Nema Nioro Mopti Assaba Gorgol Brakna Tagant Trarza SegouMacina Kayes Koutiala Niafunke Gourma Sikasso Bamako Bafoulabe BougouniGoundam Guidimaka Bandiagara TombouctouSatadougou Baiedulevrier Niger Senegal 15 60 10 40 20 5 0 0 Baol Bakel Podor Louga Dakar Thies Dagana Matam Bilma Goure Dosso Konny Zinder Tivaouane Saintlouis Tahoua Agadez Nguigmi Niamey CasamanceSinesaloum Tessaoua Hautegambie Tambacounda

53 B Historical materials

Figure 19: Pages of a Blue Book page for Uganda, 1945 (left) and of a Compte D´eﬁniffor Benin, 1928 (right)

COMPTE DÉFINITIF 1928 xxxm

3 W S CRÉDITS S J O S u DEPENSES «J ' par Pu H S NATURE DES DEPENSES

Cercle de Cotonou.

II 17 Entretien des routes et ponts 10.700 » 10.565 70 Entretien des routes. 8 Entretien des marchés et caravansérails ; 3.000 » 2.717 82 Réparation ie la toiture du mar- 9 Entretien des : immeubles, gîtes d'étapes et ché de cotonou et uadigconnage. puits 1.250 » 350 50 Réparation do l'école de fiodomey, , . crépissage des murs et badigeonnage. Totaux. 14.950 » 13 634 02

17 2 2 Autres — dépenses imprévues. Aménagement d'un champ d'aviation entre Kadjèhoun et Godomey 3 084 » 2 084 » Aménagement d'un champ d'aviation entre Kadjèhoun et (iodomey. Totaux 3.084 » 2.084 »

20 2 1 Construction de ponts et de routes dans les cercles 25.200 » 24.773 50 Grosses réparations a la route de , Cotonou. Totaux 25.200 » 24.773 50

Cercle de Djougou

III 7 Entretien des routes et ponts 37.000 » 19.608 » Travaux courants d'entretien des routes Djougou-Sèméré-Djougou- Onklou sur tout leur parcourt). 8 Entretien des mai chés et caravansérails. 1.000 » 020 » Entretien courant des bâtiments du marché et

2 3 Construction de bâtiments pour logement de fonctionnaires 4.000 » 610 » Construction d'un pavillon de 3 pièces avec vôrandah, toiture chaume. Totaux 4 000 » 010 »

20 2 l Construction de et pnnts routes nouvelles.' 32.000 » 16.148 » Construction de la route de Djou- , gou a N'Dali. Totaux 32.000 » 16.148 »

6 3 Travaux divers aux rpostes médicaux... 5 000 » 4.180 » construction d'un dispensaire a , Djougou. Totaux 5.000 » 4.180 »

Cercle du Borgou.

11 1 7/ PntrptiPii HP« rontP< entretien aes roiue» etPt pontsnnntt; -il.UUUk\ 000 » 49i».o/u 876 » Entretien et amélioration desroules ,iu cercle. Remise en état de la route Nikki. Kallé, route Guessou, Senendé. Construction d'un pont sur rivière Kala, réfection des ponts du cercle. 8 Entretien des marchés et caravansérails 1.000». 1-009» decgKcu^^en8e^ffr "^ tion des toits des caravansérails des 9 Entretien des immeubles, gîtes d'étapes et subdivisions. puits .. . . * 4' 750 » 4.697 » lîntretiendes immeubles du cercle et creusement des puits du poste. Construction d'un gîte d'étapes â Bori. 10 1U entretienFntfpripn aesHPS Cimctieiespimptipr'PS ~ou250 » 235~o<; » Crépissage et blanchiement des tombes, réfection du mur d'entou- ,, „ , . . . . t m • 11 Réparation et transformation des Tribunauxi. raSe~ â Bem béréké. .f.. ,„ . , ,. . m .. , . indigènes 3oOocn »„. dou-JKO » Transformation du Tribunal de Parakou. Réfection toiture, aména- salle d'audiences. Crépissage 1V, „ fementes murs.

54 Figure 20: Colonial map of Nigeria (1948)

Figure 21: Natural resources in West Africa