THE CASH CROP REVOLUTION, COLONIALISM AND LEGACIES

OF SPATIAL INEQUALITY: EVIDENCE FROM AFRICA∗

CSAE Working Paper WPS/2020-12

Philip Roessler†
Yannick I. Pengl‡
Robert Marty§
Kyle Sorlie Titlow¶
Nicolas van de Walle‖

July 14, 2020

∗
We are grateful to the following colleagues who have provided valuable feedback on this paper: Leonardo Arriola,
Ariel BenYishay, Lars-Erik Cederman, Natalie Letsa, Carl Mueller-Crepon, Jan Pierskalla, Tom Pepinksy, Valeria
Rueda, Nils Weidmann, Martha Wilfahrt, and Steve Wilkinson. We also thank participants at the Centre for
Study of African Economies (CSAE) Seminar, University of Oxford; Working Group of African Political Economy
at University of California-Berkeley; Comparative-Historical Social Sciences (CHSS) Workshop at Northwestern
University; Comparative Politics of Colonialism Workshop, Ohio State University; Political Science Lecture Series,
Penn State University; International Development Seminar, William & Mary; Annual Meeting of the African Studies
Association; and Annual Meeting of the American Political Science Association. For excellent research assistance,
we thank Layla Abi-Falah, Brian Barnisin, Briana Cattelino, Ryan Corcoran, Daniel Sean Gurley, Rebecca Lamb,
Caroline Martin, Rosie Olwell, Márcia Pereia, Greg Robson, Damien Roux, Aaron Spitler, Kaylin Stigall, Henry
Young, and Eric Zybko. This research was supported by a U.S. National Science Foundation grant, award #1628498,
and Swiss National Science Foundation, award #P0EZP1_159076. The authors also acknowledge support from the
Global Research Institute and Center for Geospatial Analysis at William & Mary. Research for this paper occurred
while Philip Roessler was a fellow at the Blavatnik School of Government and affiliate at CSAE at Oxford University
in 2017-2018. He thanks them for their institutional support.
†
Department of Government, William & Mary, proessler@wm.edu
‡
International Conflict Research, ETH Zürich, yannick.pengl@icr.gess.ethz.ch
§
Development Impact Evaluation (DIME), World Bank, rmarty@worldbank.org
¶
Center for African Development, William & Mary, kyletitlow@gmail.com
‖
Department of Government, Cornell University, nv38@cornell.edu

t
Centre for the Study of African Economies
Department of Economics . University of Oxford . Manor Road Building . Oxford OX1 3UQ
T: +44 (0)1865 271084 . F: +44 (0)1865 281447 . E: csae.enquiries@economics.ox.ac.uk . W: www.csae.ox.ac.uk
 Abstract

We analyze the long-term effects of colonial cash crop extraction in Africa. Our concep-
tual framework focuses on the dynamic, interactive effects of geography, trade and colonial-
ism in the context of Africa’s structural change from the slave trades to export agriculture.
The adoption of cash crops shifted the loci of economic production to smallholder farmers
in areas suitable for cultivation. Concurrently, the cash crop revolution—tied to European
industrialization—led to the diffusion of economic imperialism beyond coastal Africa. Imperial
extractive economies fueled infrastructural development in highly-suitable zones but dislocated
production linkages to Europe and stymied the economic differentiation that otherwise might
have occurred. The result was economic agglomeration at the site of production but with
limited spillovers to nearby areas. Using agro-climatic suitability scores and historical data
on the source location of more than 95 percent of all exports across 38 African states, we
find that colonial cash crop production exhibited a large and positive long-run effect on local
development in terms of urbanization, road infrastructure, nighttime luminosity and house-
hold wealth. These effects rival or surpass other geographic and historical forces frequently
linked to subnational development in Africa. Exploring causal mechanisms, we show that
path dependence due to colonial infrastructure investments is the more important channel
than continued advantages in agricultural productivity. We also find that the positive local
effects of colonial cash crop extraction came at the expense of surrounding areas and thereby
entrenched deep spatial inequalities.

Keywords: agriculture, colonialism, spatial inequality, Africa

JEL Codes: F63, N57, O13, O18, Q17

 Colonialism almost never exploits the entire country. It is con-
tent with extracting natural resources and exporting them to
the metropolitan industries thereby enabling a specific sector
to grow relatively wealthy, while the rest of the colony contin-
ues, or rather sinks, into underdevelopment and poverty. In the
aftermath of independence the nationals who live in the pros-
perous regions realize their good fortune and their gut reaction
is to refuse to feed the rest of the nation. The regions rich in
groundnuts, cocoa and diamonds stand out against the empty
panorama offered by the rest of the country.

— Frantz Fanon, The Wretched of the Earth, 1961

1 Introduction

Few questions have attracted more attention—and provoked more debate—in economics
than the causes and consequences of variation in aggregate wealth (Smith, 1776; Diamond,
1997; Sachs, Mellinger and Gallup, 2001; Acemoglu, Johnson and Robinson, 2001). Tradi-
tionally studied at the country-level, there is a growing focus on spatial disparities in wealth
within countries (Nordhaus, 2006; Hodler and Raschky, 2014; Alesina, Michalopoulos and
Papaioannou, 2016; Henderson et al., 2018).1 In contrast to traditional theories of devel-
opmental economics that conceived of “unbalanced growth” as a positive lever for long-run
development (Hirschman, 1958), a new literature finds that severe disparities in subnational
wealth—especially when organized along ethnic lines—negatively affect economic perfor-
mance (Alesina, Michalopoulos and Papaioannou, 2016). Such ethno-regional disparities are
also found to elevate intra-state conflict risk as horizontal inequalities increase grievances,
worsen bargaining problems and undermine ethnic powersharing (Stewart, 2008; Ostby, 2008;
Cederman, Weidmann and Gleditsch, 2011). Spatial inequality thus may represent one of
the root causes of the vicious poverty-conflict trap that affects many low-income countries,
particularly in Africa—the region with the highest levels of spatial inequality in the world.2
1
For a useful scoping study, Michalopoulos and Papaioannou (2018).
2
See Table SM5.

1
 What accounts for Africa’s high levels of spatial inequality? Existing scholarship on
African political economy tends to fall along a classic divide on the endogenous versus ex-
ogenous effects of colonialism.
The former emphasizes continuity and persistence in African political topography and
economic geography from the pre-colonial period until the present (Herbst, 2000). On the
whole, this body of scholarship links development patterns across Africa to the recurring
and path dependent effects of the region’s unique biogeographical fundamentals,3 such as
the relative paucity of domesticable plants and animals, the vast continental interior with few
inlets or navigable rivers, and disease ecologies. In contrast, the second stream emphasizes
the exogenous impact of extractive institutions—the slave trades and colonialism—and their
disruptive effects on spatial equilibria of development.4
We build on this literature. In so doing, we bridge these two research streams. We ad-
vance an integrated framework to account for the economic geography of colonialism—that
is, why the colonial state invested where it did—but that also speaks to its potential distri-
butional consequences—whether if, as Fanon suggests in the epigraph, colonial extraction’s
development benefits for some regions came at the expense of others, entrenching spatial
inequality.
Our framework focuses on the impact of the continent’s structural transformation away
from the slave trades to export agriculture beginning from the end of the 18th century. We
conceive that this exogenous change in international trade reconfigured African economic
geography via the effects of three interlocking factors: a.) the shift in the value of geographic
fundamentals to favor agro-climatic suitability for cash crops (such as, oil palm, groundnuts,
cocoa, coffee, and cotton); b.) trade costs; and c.) imperial institutions.
3
Among other studies, see Diamond (1997), Gallup, Sachs and Mellinger (1999), Sachs, Mellinger and
Gallup (2001), Marshall and Hildebrand (2002), Olsson and Hibbs (2005), Alsan (2014), Fenske (2014),
Alesina, Michalopoulos and Papaioannou (2016), Mayshar et al. (2018.), Michalopoulos, Putterman and
Weil (2019).
4
Among other studies, see Nunn (2008), Nunn and Puga (2010), Nunn and Wantchekon (2011), Huillery
(2009), Huillery (2011), Michalopoulos and Papaioannou (2016), Jedwab and Moradi (2016), Jedwab, Kerby
and Moradi (2017).

2
 Crowded out by the slave trades for several centuries (Rodney, 1972; Inikori, 2014), com-
mercial export agriculture finally took off—starting in West Africa—after the slave trade’s
abolition in the early 1800s and in response to demand from industrializing states for agricul-
tural commodities (e.g., rubber, vegetable oils, cotton, cocoa and coffee) (Hogendorn, 1969;
Hopkins, 1973; Austin, 2009; Frankema, Williamson and Woltjer, 2018). Led in many places
by smallholder farmers, the cash crop revolution stimulated agricultural innovation; boosted
farmers’ incomes; induced migration to farming areas; and propelled investments in trans-
portation infrastructure (Hill, 1963; Hogendorn, 1969; Hopkins, 1973). Thus, in contrast to
the plantation economies of the Americas, West African cash crop agriculture tended to re-
semble the “yeoman farm system” of temperate staples (Auty, 1997; Hopkins, 1973).5 Areas
suitable for cash crop cultivation emerged as loci of localized economic agglomeration as it
spurred backward and forward linkages to, respectively, support cultivation and production,
and processing, marketing and transport (Hirschman, 1977).
However, as the cash crop trade lifted up smallholder farmers, it also pulled European
merchants deeper into the region. Occurring at the height of imperialism, European traders
leveraged mercantilist practices and institutions—such as gunboat commerce, trading mo-
nopolies, foreign-financed infrastructural investments, and, ultimately, colonialism—to dom-
inate Africa’s agricultural markets (Rodney, 1972). The spread of imperial institutions
profoundly shaped the developmental effects of Africa’s smallholder farming revolution.
European-financed transportation infrastructure, especially railways, fueled the cash crop
trade, leading to significant improvements in farmers’ living standards (Jedwab and Moradi,
2016; Moradi, Austin and Baten, 2013). But designed to vertically integrate agricultural
zones with European markets, this new transportation infrastructure—combined with pro-
tective trade policies (e.g., limiting imports to bulk, raw commodities)— dislocated forward
production outside of Africa, thwarting local manufacturing and stymieing the domestic eco-
5
On the classical economics literature on staples and development, see Innis (1930); Baldwin (1956);
North (1959); Hirschman (1977).

3
nomic differentiation that otherwise might have resulted.6 Beyond the prized cultivation
zones and trading hubs, other areas were either neglected or relegated as a labor reserve
and intentionally underdeveloped to create a cheap supply of labor to work in cash crop and
mining enclaves (Amin, 1972, 1974; Rodney, 1972; Cordell, Gregory and Piché, 1996).
Several important implications follow from our conceptual framework and this historical
context. One, we would expect agro-climatic suitability for cash crops and historical cash
crop production zones to be significant predictors of contemporary patterns of economic
development across the continent—with colonial infrastructure investments serving as one
of the primary channels. Two, while colonial extraction boosted economic agglomeration
in highly suitable areas, in the absence of strong domestic production linkages we would
expect limited positive spillover effects. And, if anything, given the exploitative nature of
colonialism, other peripheral regions may have been left worse off than they otherwise would
have been.
To test this argument, we build a comprehensive dataset of historical African economic
geography, including detailed geospatial information on sites of cash crop production and
mining. The latter data was extracted from a map published in 1961 depicting the source
location of more than 95 percent of all exports across 38 African states, standardized in
1957 U.S. dollars ($). Constructed by a team led by renowned geographer of Africa, William
Hance, the map draws on “hundreds of sources...including maps, articles, agricultural year-
books, reports of commodity boards, and product and regional studies” (Hance, Kotschar
and Peterec, 1961). As far as we know, the Hance dataset is the most exhaustive and gran-
ular representation of the spatial diffusion of the cash crop revolution across Africa, but has
never been systematically analyzed.
We validated Hance’s map by independently collecting administrative data from colonial
records and other historical sources on the volume and value of the most important agricul-
6
This is a central line of argument of dependency theory (Frank, 1966; dos Santos, 1970; Rodney, 1972;
Cardoso and Faletto, 1979). See also Hirschman (1977) who suggests that imperialism led to underdevelop-
ment via its negative effects on production and fiscal linkages—from which our argument draws inspiration.

4
tural commodities and minerals produced by each state (i.e., those that make up at least
10 percent of total exports at the end of colonialism) standardized in 1960 US$ and aggre-
gated at the level of subnational administrative units.7 Despite employing different units of
analysis and our less comprehensive coverage of agricultural and mineral commodities, over-
all and resource-specific correlations between our dataset and Hance’s are high (see Figure
SM1). Given the greater coverage and better granularity of Hance’s data, we employ his
spatial data in our analysis, measuring production at the extensive and intensive margins for
0.25×0.25 degree (longitude/latitude) grid cells (approximately 770 sq km at the equator).
To preview our results, we find that areas of historical cash crop production are signifi-
cantly better off today on a set of infrastructural and wealth measures than comparable cells
within the same African country. Colonial cash crop cells had a 16 percentage points higher
probability of having a quality road in 1998; close to a 20 percentage points higher likelihood
of emitting nighttime lights in 2015; a 19 percentage points higher likelihood of having a
city in 2015; and 14 percent of a standard deviation greater household wealth. Historical
cash crop production exhibits a comparable effect on contemporary roads, electrification and
cities as colonial mineral extraction, despite the more capital-intensive nature of mining.
To address endogeneity concerns, we use propensity score weighted regressions; employ
a randomization inference-type analysis that samples 1000 plausible spatial equilibria of
colonial resource extraction; and run spatial 2SLS-IV models instrumenting observed cash
crop production with the mean agro-climatic suitability score across the nine most important
African export crops (cocoa, coffee, cotton, groundnuts, oil palm, tea, sugarcane, tobacco,
bananas). This suitability measure predicts colonial production well and is unrelated to
pre-colonial development outcomes. Results from these additional analyses are in-line with
the baseline OLS models. In terms of the relative substantive significance of colonial cash
crop agriculture on long-run development, we show that, next to distance to the coast, cash
crop suitability surpasses or rivals other factors, such as caloric suitability or disease ecology.
7
For methods and sources, see Appendix SM2 “Sources of Historical Primary Commodity Production
by Country.”

5
We show that the persistence of the cash crop revolution’s agglomerating effects operates
more through the path dependency induced by colonial infrastructural investments in roads,
railways and power generation than continued agricultural production.
We further find that commercial export agriculture generally had moderate spillover
effects of up to 50 kilometers. Yet, in line with our argument that economic agglomeration
from colonial extraction arose only as it deepened spatial inequities, spillover effects on
electrification and urbanization are negative and significant after only 75 km. This suggests
that under colonial extractive institutions cash crop agriculture’s developmental benefits
came at the expense of other areas—which are worse off today than would be expected
based on their underlying characteristics.
Our paper contributes to existing scholarship on comparative economic development in
several important ways.
First, we contribute new geographic data on cash crop and mineral extraction in colonial
Africa as well as the economic organization of the colonial state. Despite the importance of
the cash crop revolution in shaping modern African economies, there has been a surprising
dearth of systematic research on the subject.8 Frankema, Williamson and Woltjer (2018)
offer a valuable long-run, macro-level perspective on the economic transition, drawing on
comprehensive commodity trade data they assembled. But the spatial implications of the
transition to cash crop agriculture have heretofore been understudied.9
The remarkable dataset assembled by Hance, Kotschar and Peterec (1961) allows us to
contribute new empirical evidence to a large and influential literature on the effects of natural
resources and primary commodity exports on economic development. (See, among others,
Sachs and Warner (2001); Auty (2001).) Whereas this literature has predominantly focused
8
The political and economic consequences of cash crops in Africa have generally been understudied.
Important exceptions include Bates (1981), Boone (2003), Kasara (2007) and the body of work by Austin
(e.g. 2008; 2009; 2010) as well as specific historical studies we cite throughout the paper.
9
This is further highlighted in a comprehensive survey of scholarship on precolonial and colonial lega-
cies on economic development in Africa (Michalopoulos and Papaioannou, 2020). Aside from Frankema,
Williamson and Woltjer (2018), Bates (1981), and Lowes and Montero’s study on rubber concessions in the
Congo Free State, the survey identifies few studies that address the transformative effects of the transition
to cash crop agriculture.

6
on general equilibrium effects via institutions, macroeconomic policy, and industrialization,10
in line with the broader subnational turn in the study of development growing attention is
paid to within-country effects (Dell, 2010; Michalopoulos and Papaioannou, 2018; Dell and
Olken, 2019; Mamo, Bhattacharyya and Moradi, 2019). In studying the spatial effects of
the cash crop revolution, we build from a classical economics literature on the role of staples
on regional (i.e., subnational) economic growth (Innis, 1930; Baldwin, 1956; North, 1959;
Hirschman, 1977; Engerman and Sokoloff, 1997; Sokoloff and Engerman, 2000). At the core
of this literature is an emphasis on the “technological nature of the production function”
of different agricultural crops based on their economies of scale—and thus whether planta-
tions or smallholder farms arise (North, 1959; Baldwin, 1956)—and their knock-on effects on
growth via “entrepreneurial initiative,” the density and horizontal-nature of production link-
ages, and human capital formation (North, 1959; Hirschman, 1977; Engerman and Sokoloff,
1997). Africa diverges from this model. Despite smallholder farming prevailing as the dom-
inant mode of cash crop production, the economic organization of the African state tended
to resemble Latin America’s enclave structures rather than North America’s economically
differentiated and dynamic regional economies. While a number of factors likely account for
this (e.g., legacies of slave trades, technological development), African historical economic
development in the 19th and 20th century points to the mediating effects of institutions on
factor endowments (Auty, 1997). In line with another prominent literature in economics
(Frank, 1966; dos Santos, 1970; Amin, 1972; Rodney, 1972; Cardoso and Faletto, 1979), the
political and economic institutions regulating the cash crop trade were significantly shaped
by international economic forces—namely the spread of imperialism as European trading
houses, backed by the metropole, sought to procure low-cost raw commodities to supply
manufacturers in Europe. While single export staple economies are expected to engender
weaker production domestic linkages relative to other trading systems (Baldwin, 1956; North,
1959), imperial economic policies likely weakened these economic structures even further by
10
See Van der Ploeg (2011) for a review.

7
suppressing local manufacturing and undercutting African traders (Hirschman, 1977).11
Third, in analyzing the long-run impact of the cash crop revolution and imperial con-
quest, our paper offers new evidence to speak to the debate on how much, if at all, colonial-
ism represented a break from existing spatial patterns of state-building and development.
Our analysis finds colonial economic geography marked a significant spatial shift in pat-
terns of development to areas highly suitable for cash crop cultivation. In West Africa,
the consequence was that colonialism disrupted the “regional complementarities” between
the forest and savanna zones that existed for centuries serving as a basis for long-distance
trade and shaping state formation (Amin, 1972). The Trans-Saharan trade continued to
flourish throughout the 19th century, despite technological change favoring sea trade over
caravan trade (Newbury, 1966; Moseley, 1992). Colonialism would be the death knell for the
Trans-Saharan trade. Even worse, in treating the densely populated regions on the southern
edge of the Sahel as labor reserves, colonialism would dislocate development around these
historical centers to the forest zones which were more suitable for cocoa and coffee (Amin,
1974; Ajayi and Crowder, 1985; Cordell, Gregory and Piché, 1996; Raynaut, 2002). Though
brief in duration, colonialism would bring about a significant and lasting divergence in West
African development trajectories.
Finally, our paper speaks to the debate on the long-run effects of colonialism on de-
velopment. At the country-level extractive colonialism is found to leave a legacy of weak
institutions, poor property rights and underdevelopment (Acemoglu, Johnson and Robinson,
2001, 2002; Lange, 2009). On the other hand, extractive processes may generate positive
local agglomerating effects via investments in infrastructure and manufacturing (Jedwab and
Moradi, 2016; Dell and Olken, 2019). One important question in terms of the local devel-
opmental effects of extractive systems, however, is whether these processes are generating
positive or negative spillovers. If the latter, then benefits for some are coming at the expense
11
The absence of a large settler population, which may have coalesced as an interest group to resist
mercantilist policies, likely compounded this effect (Engerman and Sokoloff, 1997; Acemoglu, Johnson and
Robinson, 2002).

8
of others. Our spillover analysis provides at least suggestive evidence that colonial extrac-
tion generated negative spillovers—making neighboring areas worse off than we would expect
based on underlying characteristics consistent with the mercantilist logic of colonialism laid
out above. This analysis helps to reconcile the seeming differences between national-level
and local-level impact of extractive colonialism. Rather than offsetting negative institutional
effects, the subnational extractive processes likely made them worse. Colonialism left much
of Africa trapped in a negative feedback loop of weak institutions and spatial inequities.

2 Relevant Literature

As described in the Introduction, existing scholarship on African economic development

revolves around the path dependent effects of underlying geographic factors or the legacies
of the region’s extractive institutions. In this section we briefly review the existing literature
before developing our conceptual framework.

2.1 The Long-run Effects of Africa’s Unique Biogeography

A large and influential body of scholarship attributes the prevalence of low-income states
across Africa to the region’s unique biogeography and the enduring consequences of its late
Neolithic transition to sedentary agriculture (Diamond, 1997; Olsson and Hibbs, 2005). A
combination of unfavorable locational fundamentals—erratic rainfall in the savanna; the rel-
ative paucity of domesticable plants and animals; endemic diseases such as trypanosomiasis
and malaria; tropical climate; productivity advantage of roots and tubers over cereals; vast
continental interior with few inlets or navigable rivers; and its north-south axis—not only
delayed the uptake of farming but, once adopted, trapped these societies in a low agriculture-
weak state equilibrium (Diamond, 1997; Gallup, Sachs and Mellinger, 1999; Sachs, Mellinger
and Gallup, 2001; Marshall and Hildebrand, 2002; Olsson and Hibbs, 2005; Alsan, 2014;

9
Scott, 2017; Mayshar et al., 2018.). Subsistence and shifting agriculture along with pastoral-
ism predominated, limiting population growth and contributing to the fractal nature of the
region’s state system (Herbst, 2000). Consequently, compared to Eurasia, the Neolithic Rev-
olution did not leave Africa on a path to deep economic and political stratification (Coquery-
Vidrovitch, 1975; Osafo-Kwaako and Robinson, 2013). In fact, unlike the rest of the world,
in Africa the relationship between agricultural potential, population density and historical
state formation is weak (Osafo-Kwaako and Robinson, 2013).12 Instead, the centralized
polities that emerged and endured tended to lie astride ecological boundaries and arise from
long-distance trade rather than intensive agricultural cultivation (Coquery-Vidrovitch, 1975;
Bates, 1983; Fenske, 2014).
Outside of Africa’s trade-based states, which appear to have some persistence as centers
of economic development (Fenske, 2014; Michalopoulos and Papaioannou, 2013), polities
tended to revolve around less hierarchical kinship-based groups shaped by local differences
in geographic and environmental conditions (Michalopoulos, 2012; Enke, 2019; Cervellati,
Chiovelli and Esposito, 2019). This social differentiation is posited to have led to varia-
tion in human capital formation and income disparities across ethnic lines (Alsan, 2014;
Michalopoulos, 2012).
The persistence of Africa’s biogeography on uneven patterns of development is proposed
to operate through various channels: 1.) path dependency—e.g., differential returns that
arise from variation in land endowments and human capital formation (Alesina, Michalopou-
los and Papaioannou, 2016); 2.) institutions—e.g., variation in political centralization (Gen-
naioli and Rainer, 2007; Alsan, 2014; Michalopoulos and Papaioannou, 2013; Fenske, 2014);
3.) culture—e.g., the adoption and transmission of different economic lifeways (Michalopou-
los, Putterman and Weil, 2019) or ethnogenesis (Michalopoulos, 2012; Cervellati, Chiovelli
and Esposito, 2019); and 4.) the recurring effects of locational fundamentals on subsequent
economic forces, such as globalization (Henderson et al., 2018) or the spread of Christian
12
Moreover, across the region, caloric suitability and cereal advantage are, if anything, negatively corre-
lated with pre-colonial political centralization.See Table SM6.

10
missionaries (Jedwab et al., 2019).

2.2 Extractive Economies

An alternative stream of research emphasizes the importance of institutions over geography,

especially the impact that extractive institutions have played. Two of the most pernicious
forms of economic extraction have dominated Africa over the last five hundred years: the
slave trades and colonialism. Both are shown to have persistent negative general equilibrium
effects on development via their institutional legacies. We discuss each in turn and analyze
their spatial implications.
The slave trades and the vicious gun-slave cycle it unleashed increased social fraction-
alization and mistrust that eroded legal and political institutions, increased rent-seeking
and undermined public goods provision (Nunn, 2008; Nunn and Wantchekon, 2011). One
paradoxical spatial implication of the slave trade is that in Africa—unlike the rest of the
world—areas with rugged terrain tend to be richer not poorer, despite significant barriers
to trade and other economic production, because of the protections it offered from slave
raiding (Nunn and Puga, 2010). However, as we show in Table SM19, in contrast to the
negative long-run effects on social institutions and country-level development, those areas
most affected by the slave trades did not permanently fall behind in terms of infrastructural
development within countries. The most-affected areas tend to have higher—not lower—
levels of contemporary economic agglomeration, due perhaps to their favorable geography
and proximity to the coast, the targeting of densely populated areas for slave raiding, or
the successful readjustment of markets and trade networks to other types of commerce.
Nonetheless, similar to the effects of the tsetse fly (Alsan, 2014), the slave trades likely en-
sured that areas with the most favorable underlying conditions—due to strong locational
fundamentals—have not realized their full economic potential.
Another important consequence of the slave trades is that it paved the way for the
European imperial conquest of Africa as it further weakened political centralization and

11
deepened racism (Rodney, 1972; Nunn, 2008). As is well-documented, European colonialism
created highly centralized, weak, extractive states (Rodney, 1972; Young, 1994; Herbst, 2000;
Acemoglu and Robinson, 2012). In contrast to the slave trades, colonialism appears to have
had more substantial effects on patterns of spatial development. Huillery (2009, 2011) finds
that, independent of underlying pre-colonial characteristics, such as political centralization,
trade and population density, the spatial distribution of colonial public investments in health
and education produced significant path dependent effects in French West Africa.13 .
Perhaps even more important in changing the contours of African economic geography
were colonial investments in transportation infrastructure (Jedwab and Moradi, 2016; Jed-
wab, Kerby and Moradi, 2017). Historically, high transportation costs have proven one
of the most significant barriers to trade in the region (Chaves, Engerman and Robinson,
2014). The colonial rail revolution in Africa—in which, in the span of a century 58,716
kilometers of railroad tracks were built—significantly reduced transportation costs (Jedwab
and Moradi, 2016). In doing so, it opened up previously underdeveloped areas to trade,14
spurring an increase in economic activities, agricultural production and mining, which in
turn increased labor demands. Strikingly, Jedwab and Moradi (2016) estimate that outside
of Africa’s largest cities, 55 percent of aggregate change in urbanization between 1900 and
1960 occurred within a 10-kilometer radius of colonial railroads.
What accounts for the spatial patterns of colonial investments that would prove so de-
cisive in reshaping African economic geography? Jedwab and Moradi (2016) emphasize the
role strategic considerations played in determining the siting of colonial railroads as the im-
perial powers competed for different territories and sought to consolidate internal control.
Yet according to their data, military domination was given as a motivation in only 35.5%
13
European settlements represented an important vehicle for local development as they drew greater levels
of colonial investments in public goods or economic production (Huillery, 2011)
14
The rail revolution was only made possible by the diffusion of new technologies that enabled construction
even across areas with unfavorable locational fundamentals. For example, the importation of the traction
engine, a self-propelled steam engine that was used to move heavy loads over rough terrain, was vital in
the construction of the Uganda Railway in areas which the tsetse fly was endemic and decimated transport
animals. “The Latest Emissaries of Civilisation: Traction Engines in Uganda,” The Railway News, Volume
70, December 10, 1898.

12
of the cases. The rest were linked to primary commodity extraction (mining and cash crop
agriculture)15 —and between minerals and agricultural commodities, most colonies were de-
pendent on the latter. (See SM2 “Sources of Historical Primary Commodity Production by
Country.”) As Crowder observed, reflecting on colonial infrastructure networks: “a map of
the railways and major roads in the 1930s represents a grid draining the exportable resources
of the interior towards the coastal ports (Crowder, 1978, page 246).”
In line with an influential stream of scholarship in economic history (McPhee, 1926;
Hance, 1964; Rodney, 1972; Hopkins, 1973; Lynn, 1997; Austin, 2009; Frankema, Williamson
and Woltjer, 2018), this underscores the importance of cash crop agriculture in structuring
colonial state-building and contemporary spatial inequality. Yet, there have been few sys-
tematic studies along this line of inquiry. In the next section, we develop our conceptual
framework that explains how the spread of the cash crop revolution and colonial economic
extraction shaped the modern African state before employing new, comprehensive data to
estimate their effects vis-à-vis other factors.

3 Conceptual Framework and Historical Context

In accounting for spatial patterns of economic development across Africa, we consider the
dynamic, interactive effects of geography, trade and institutions.

3.1 Geography

As discussed above, locational fundamentals have had a powerful effect on long-run devel-
opment across the continent. The value of different geographic characteristics is not fixed,
however; it can change as technology and market forces change (Hopkins, 1973; Henderson
et al., 2018). These economic shocks can interact with locational fundamentals to give rise
to new centers of agglomeration (Gallup, Sachs and Mellinger, 1999; Nunn and Qian, 2011;
15
See an earlier version of their paper (Jedwab and Moradi, 2012).

13
Bleakley and Lin, 2012; Henderson et al., 2018). This is an important dimension in ac-
counting for the spatial impact of the cash crop revolution in Africa from the 19th century
onward.
As is well-known, prior to the 19th century the global cash crop trade had differential
effects in Africa relative to other tropical regions, despite their similar ecological conditions.
Whereas the Columbian Exchange spurred sugar production in the Caribbean and Latin
America from the 16th century and opened European markets for other agricultural com-
modities, such as tobacco and cocoa, European imperial powers, led by Portugal, turned to
Atlantic Africa as a labor reserve for plantations in the New World (Inikori, 2007)—leading
to the enslavement of millions of people (Manning, 1990). (See previous discussion of effects
of slave trades in section 2.2.) The slave trade largely crowded out cash crop production for
three hundred years across the continent (Rodney, 1972; Inikori, 2014).16
By the end of the 18th century, however, the African cash crop trade began to increase,
starting with oil palm in West Africa, in response to rising European demand for the com-
modity for soap-making, candle-making and as an industrial lubricant (Lynn, 1997).17 The
abolition of the slave trade in the early 19th century accelerated West Africa’s economic
transformation (Hopkins, 1973; Austin, 2009). Beyond oil palm, European demand for other
oleaginous crops, such as groundnuts, as well as cotton, cocoa, coffee, tea, and tobacco fu-
elled the cash crop revolution that spread from West Africa to across the continent over the
next 150 years.
Outside of a few large-scale irrigated projects, such as the Gezira cotton scheme in Sudan,
agro-climatic suitability would condition the spread of cash crop agriculture. (See Figure
1 below). Groundnuts, cotton and tobacco, which require rainfall seasonality, clustered
16
An early exception was sugar production on the islands of São Tomé and Príncipe but, emblematically,
by the mid-17th century sugar exports sank and the economy revolved around servicing slave trade ships.
As the Liverpool General Advertiser characterized African trade in 1768, it consists in “slaves, teeth [ivory],
and gold ... [B]ut all this while, there is not the least use made of the land” (Cited in Eltis (2013)). This
gross generalization, however, overlooked increased agricultural production in food crops, from areas such as
Upper Guinea, to supply slave ships (Green, 2013).
17
On temporal trends in the cash crop trade in Africa, see Frankema, Williamson and Woltjer (2018).

14
in tropical and subtropical grasslands, savannas, and shrublands, whereas oil palm, cocoa
and robusta coffee, which need warmer temperatures, higher levels of rainfall and higher
humidity, thrived in tropical and subtropical moist broadleaf forests. These ‘new’ geographic
fundamentals would underlie the spatial distribution of the commercialization of agriculture
across the continent.

3.2 Trade

A second important dimension shaping the spatial distribution of the cash crop revolution
was the geography of markets and trade costs. Henderson et al. (2018) show that the
globalization of markets and technological advances in long-distance trade over the last 150
years led to the concentration of economic activity in coastal regions of late-developing
countries. With cash crop demand driven by industrializing states in Europe, commercial
agricultural zones were integrated with export markets. Thus, in West Africa, conditional
on suitability, cash crop cultivation tended to concentrate in areas closer to the coast first.
(These areas also tended to be the first exposed to the diffusion of cash crops and new
agricultural techniques.18 )
Trade costs are not fixed either, however. Infrastructural investments could significantly
lower them, opening up suitable areas in the hinterland for cash crop production (Jedwab
and Moradi, 2016).19 As noted above, in many cases demand for cash crops endogenously in-
creased investments in transportation infrastructure.20 One of the most prominent examples
was the British Cotton Growing Association’s (BCGA) lobbying of the British government
for a railway line to Northern Nigeria which the BCGA identified as the “salvation of Lan-
cashire” to reduce British cotton manufacturers’ dependence on American cotton (Hyam,
18
The classic example of this is the origin story of cocoa in Ghana in which Tetteh Quarshie, a Ghanaian
blacksmith and farmer, travelled to Fernando Po in the 1870s and brought cocoa seeds back with him, leading
to the successful introduction of the cash crop (Dickson, 1969).
19
As Jedwab and Moradi (2016) show in the Gold Coast, relative to headloading, the prevailing primary
means of transport, railways reduced transportation costs of cocoa by as much as 90 percent.
20
As one of the leading French traders in West Africa, Georges Borelli, put it in lobbying the French
government to build a railway in Dahomey, it would help “syphon [sic] off immense quantities of products,
up to the present time not used, towards the coast and towards Europe” (cited in Daumalin (2004)).

15
1968). Despite strong local pressure from colonial authorities and their agents, farmers in
Northern Nigeria spurned cotton cultivation for groundnuts, which gave a higher return, re-
quired less land and labor, and were consumed locally (Hogendorn, 1969, 1978). Nonetheless,
the railway fueled Northern Nigeria’s ascendancy as one of the leading groundnut production
zones in the world.21
This points to a corollary to Henderson et al’s (2018) elegant model of spatial develop-
ment in late developing countries. The rise of commercial export agriculture and connective
infrastructure potentially fuelled agglomeration in suitable areas beyond coastal regions.

3.3 Institutions

The emergence of new spatial equilibria and their developmental consequences are con-
currently shaped by a third dimension—the institutions in which markets and trade are
embedded. In states where regimes do not possess an encompassing interest in society and
power stems from monopolizing rents, elites are likely to structure the economy to enable
their capture of surplus arising from economic production—likely diminishing broader soci-
etal benefits and increasing inequality (Acemoglu, Johnson and Robinson, 2001, 2002; North
et al., 2009). In such systems, extractive, enclave economies tend to flourish; production is
organized around the harvesting and export of raw primary commodities (such as, minerals
and agriculture) from a few geographically-concentrated zones that possess weak forward or
backward linkages with other domestic sectors.22
The link between extractive enclave economies and natural resources points to an endoge-
nous relationship between institutions and factor endowments (Auty, 1997; Engerman and
Sokoloff, 1997; Hall and Jones, 1999; Sokoloff and Engerman, 2000; Easterly and Levine,
21
This vignette illustrates the top-down and bottom-up dynamics driving the cash crop revolution.
Externally-financed infrastructural investments were transformative but the agency of local producers also
were important (Hogendorn, 1969).
22
On extractive economies more generally see Auty (2001); Acemoglu and Robinson (2012). On the role
of forward and backward linkages on economic integration and growth, see Hirschman (1958, 1977). Classic
statements on enclave economies in Africa and Latin America include Rodney (1972); Cardoso and Faletto
(1979). For more recent analyses, including empirical tests see Leonard and Straus (2003); Conning and
Robinson (2009).

16
2003; Leonard and Straus, 2003). One influential study in this vein is by Engerman and
Sokoloff (1997). Building on a classic literature in economics on the differential effects of
food staples (e.g., wheat) versus export crops (e.g., cotton or tobacco) on domestic pro-
duction linkages (Baldwin, 1956; North, 1959),23 Engerman and Sokoloff develop a general
model on the long-run effects of factor endowments on institutions (Engerman and Sokoloff,
1997; Sokoloff and Engerman, 2000). They posit that, wheres Latin America’s agro-climatic
suitability for sugar gave rise to higher levels of political and economic inequality via sugar’s
economies of scale and the legacies of slave-dependent plantation production, North Amer-
ica’s suitability for grain cultivation and production in small family farms contributed to
greater economic equality and competitive markets.24
In Africa, however, the historical mode of cash crop production diverged from Latin
America. Given land abundance and labor scarcity, extensive agriculture proved more effi-
cient than more labor- and capital-intensive plantations, which generally failed (Austin, 1996;
Hopkins, 1973).25 Consequently, the locus of the economic revolution resided in smallhold-
ing farms—especially the farming that emerged organically in West Africa before colonialism
(Hogendorn, 1969).26
Yet, despite this dispersion of economic production in the hands of family farms, Africa
diverged from North America’s development trajectory. Two differences stand out. West
African smallholder agricultural production was concentrated in commodity exports destined
for Europe, rather than local grain markets, which likely engendered relatively thin domestic
production linkages akin to tobacco and cotton in North America (Baldwin, 1956; North,
1959; Price, 1974). Another important difference was the level of the European settler pop-
23
The nexus between agriculture (cash crops versus staple crops) and forward production linkages is also
argued to account for subnational differences in economic organization in North America (Baldwin, 1956;
North, 1959; Price, 1974).
24
See also Easterly (2007) who tests this relationship more broadly using sugar and grain soil suitability.
25
The most high profile failure was Cadbury’s model cocoa estate, which Austin (1996) attributes to its
exorbitant labor inputs vis-à-vis smallhold farms.
26
For example, a survey of cocoa farmers in western Nigeria in the 1950s found that more than half of
farms were less than 2.5 acres (Galletti, 1956); with cocoa farms in Gold Coast typically ranging from 2 to 6
acres (Hill, 1957). Similarly groundnuts in northern Nigeria were primarily cultivated by perhaps up to one
million peasant farmers, who devoted only one to two acres to the crop.

17
ulation. Engerman and Sokoloff (1997) as well as Acemoglu, Johnson and Robinson (2001,
2002) emphasize the effects of settlers on institutional development. In low settler impe-
rial outposts, extractive institutions prevailed—in which economic production was organized
around the extraction of raw commodities for the metropole. Whereas conquistadors were
the emissaries of extraction in 16th century Latin America (Acemoglu and Robinson, 2012),
in 19th century West Africa it was European trading houses (Young, 1994).
European traders leveraged the coercive and economic power of their home governments
to capture markets beyond the coastal ports where they were located during the slave trades
(dos Santos, 1970; Rodney, 1972).27 Instrumental in this regard was the use of, what can best
be described as, gunboat commerce28 to eliminate indigenous middlemen, control markets,
and drive down prices. By the Berlin Conference, French and British traders dominated
commercial trade on every major river in the region and nearly all ports from Saint-Louis in
Senegal to the Oil Rivers in Nigeria.
The spread of imperialism had a number of consequences. For one, as discussed, Euro-
pean capital financed new infrastructure, but railways and paved roads tended to vertically-
integrate agricultural zones with export markets (Rodney, 1972). Combined with protective
trade policies (e.g., limiting imports to bulk, raw commodities), this further weakened do-
mestic production linkages (as manufacturing was dislocated to Europe) and stymied the
domestic economic differentiation and horizontal integration that otherwise might have re-
sulted (Rodney, 1972; Hirschman, 1977). A case in point is the protective trade measures
France placed on groundnuts in the 1840s. France restricted groundnut imports to bulk
unshelled nuts on French vessels, which aided French merchants’ monopolization of the West
African commodity trade, while damaging local oil processing in the region (Brooks, 1975).
27
Aiding European traders’ penetration of western Africa was the the advent of the steamship and in-
creasing availability of quinine as a prophylaxis for malaria (Headrick, 2012).
28
With the abolition of the slave trade in 1807, the British Royal Navy started to patrol the coast of
West Africa to interdict slave traders. British merchants exploited the presence of the Royal Navy to call on
them to bombard villages that resisted their commercial activities on the Niger River and other waterways.
(Headrick, 2012) French traders did the same with the French Navy, which started to patrol West Africa
after 1820 (Daumalin, 2004).

18
 Moreover, buying oligopolies and monopolies, such as George Goldie’s United African
Company (later the Royal Niger Company), dominated export markets, squeezing out African
and non-European export firms.29 This not only hurt “the growth of a class of indigenous
capitalists,” (Meredith, 1988) but also drove down producer prices that, in some cases, such
as the uptake of groundnuts in northern Nyasaland, served as a barrier to the diversification
and spread of cash crop agriculture (Kerr, 2010).
The spread of imperial trade institutions paved the way for conquest and colonization.30
Colonialism did little to alter the enclave nature of primary commodity production that
was emerging before the Berlin Conference. If anything, it intensified its development as
colonial governments organized their territories around primary commodity production to
supply markets back home (Rodney, 1972; Havinden and Meredith, 1993).31 Toward this
end, colonial authorities sought to build additional infrastructure (roads, railways and power
stations) to reduce trade costs, while providing extension services to increase crop yields
(Hopkins, 1973). Just as importantly, the colonial state leveraged its administrative and
coercive power—for example, through taxation, forced labor, and indirect rule—to compel
farmers to cultivate cash crops (Crowder, 1978; Rodney, 1972; Boone, 1992).
However, the 19th century boom in cash crop production that in many ways propelled
the imperial conquest led to declining terms of trade in agricultural commodities by the
time of the Berlin Conference (Frankema, Williamson and Woltjer, 2018). Nonetheless,
colonial states remained steadfast in promoting adoption of cash crop agriculture. But once a
production area was opened to trade, there was a tendency to specialize in these zones. Severe
budget constraints imposed by the metropole further forced colonial governments to pick
and choose the agricultural schemes it could support (van de Walle, 2009; Gardner, 2012).
29
This eventually took the form of marketing boards before and during World War II.
30
European trading houses operating in Africa emerged as one of the leading proponents for formal
colonization (Young, 1994). For example, George Goldie of the United African Company attended the latter
part of the Berlin Conference to make the case that “on the lower Niger the British flag alone flew” (Wellesley,
1934).
31
Following the Mineral Revolution in South Africa in the 1870s and 1880s, imperial conquest and state
building in southern Africa revolved much more around mining—and were led by royally chartered companies,
such as the British South African Company and Union Minière du Haut-Katanga.

19
Beyond integrating these prized zones with export markets, investments in transportation
infrastructure were scarce. Few horizontal networks were constructed to facilitate intra-
African trade (Rodney, 1972).
Another potentially important spatial implication of extractive colonialism was the cre-
ation of labor reserves, in which colonial governments used the coercive arm of the state—
repression, high taxation rates and deprivation of local economic opportunities—to create
a cheap supply of labor to service cash crop and mining enclaves and work on colonial in-
frastructural projects (Amin, 1972, 1974; Cordell, Gregory and Piché, 1996). This policy is
most commonly associated with settler colonies—especially South Africa as it laid the foun-
dations for Apartheid (Wolpe, 1972)—where settlers dispossessed indigenous populations
from the most suitable land and forced them onto less productive areas that would become
reserves. But it was more universally adopted. In non-settler colonies, where the most
suitable land remained in indigenous hands, the colonial state targeted other communities—
often densely-populated groups residing in areas seen as less suitable for cash crops or too
costly to reach—as laborers to cheaply supply the ‘productive’ parts of their colonies (Amin,
1974).32 In effect this led to a policy of dividing their territories into productive and labor
extraction zones.

***

The cash crop revolution transformed Africa. How it did so was shaped by the interaction
of geography, trade and imperial institutions. Three implications follow from our framework:
1.) we expect agro-climatic suitability for cash crops and historical cash crop production
zones to be significant predictors of contemporary patterns of economic development across
the continent; 2.) colonial infrastructure investments to reduce trade costs served as a
key channel of path dependence; and 3.) while colonial extraction boosted agglomeration
32
In West Africa, the French identified the Mossi people as “exploitable” due to its high population density
in an “inhospitable land.” Quote by French doctor, Charles Crozat, who traveled to the Mossi Kingdom in
1890. Cited in Cordell, Gregory and Piché (1996). (The Mossi homeland falls in the bottom tercile of cash
crop suitability.) Subsequently, Upper Volta (Burkina Faso), which encompassed the Mossi people, was set
up as a labor reserve.

20
in highly suitable areas, in suppressing indigenous traders, dislocating production linkages
to Europe, and treating some areas as labor reserves, this stymied the broader domestic
economic gains that likely would have accrued and may have left some non-cash crop areas
worse off than expected based on their underlying characteristics.

Data and Methods

Measuring the spatial distribution of the cash crop revolution in

Africa

To test this argument, we draw on a novel and comprehensive dataset of historical African
economic geography, including detailed geospatial information on cash crop production at
the end of the 1950s across 38 countries. We rely on multiple sources of data. As explained
above, one remarkable dataset was built by a team led by William Hance, Kotschar and
Peterec (1961). In the late 1950s, Hance’s team mapped out the source location of more
than 95 percent of exports across 38 states in sub-Saharan Africa (excluding the Union of
South Africa and most island colonies) and conveyed them as points at the site of cultivation
or extraction. Each point represents a value of $289,270, standardized in 1957 U.S. dollars
($). The dataset covers 9 groups of cash crops, 20 minerals and metals, and forest, animal
and manufactured products. In total, it identifies 9,517 geocoded production points (See
the upper-right panel in Figure 1.) Published in the Geographical Review in 1961, the map
represented a landmark contribution to the study of African economic geography. Yet it was
generally overlooked. More attention was paid to the qualitative and descriptive analysis
in Hance’s (1964) magnum opus—The Geography of Modern Africa—that was published
shortly thereafter.

21
 Independently, we collected similar data as the Hance team.33 Drawing on colonial re-
ports, maps and other records documenting the location, volume and value of primary com-
modity production across 30 countries (28 colonies and the states of Liberia and Ethiopia),
the dataset provides administrative data on the subnational distribution of the most im-
portant commodity exports produced by each state (i.e., those that make up at least 10
percent of total exports at the end of colonialism) standardized in 1960 US$. We are thus
able to validate the Hance, Kotschar and Peterec (1961) dataset at the level of subnational
administrative units. Overall and resource-specific correlations are high (see Figure SM1).

Unit of analysis, outcome variables and alternative pathways

For our empirical analysis we aggregate the historical primary commodity data, the infras-
tructure and development outcomes as well as all control variables to 28,166 quarter-degree
grid cells (the mean land area of cells in our sample is 237 sq km). In the Appendix (see
Figures SM17 and SM18), we rerun many analyses at coarser geographic units aggregating
our data to grid cells of 0.5 and 1.0 degree resolution as well as ethnic group polygons from
Murdock’s “Tribal Map of Africa” (Murdock, 1959). For each spatial unit, we code binary
measures of whether cash crop or mineral production occurred above the threshold export
value of $289,270 in 1957, as measured by Hance, Kotschar and Peterec (1961). These colo-
nial cash crop and mineral dummies serve as the main predictors in our models and allow
a quantitative comparison of the effects of different commodity types. In alternative speci-
fications, we also use continuous operationalizations of our resource variables and take the
log (+1) of cell-level production value in $100 divided by land area in sq km (See Figures
SM14 and SM15).
Consistent with the expectation that the cash crop revolution transformed the long-
run spatial distribution of development within African countries, our main outcomes focus
on contemporary infrastructure and economic agglomeration: paved and improved roads;
33
For details, see SM2.

22
electrification; cities; and household wealth.34 To measure these variables, we use easy-to-
interpret dichotomous measures indicating if a cell had: (a) a quality road in 1998; (b)
satellite-detectable luminosity at night in 2015; (c) a city with at least 10,000 inhabitants;
and (d) the cell mean of asset-based household wealth from all available geocoded Demo-
graphic and Health Surveys (DHS) from the years 1990–2017. This measure is based on
asset ownership of 747,255 households surveyed in 26 African countries. The household
wealth analysis is based on the 24% subsample of our cells that contain DHS enumeration
areas. Replacing binary outcomes with logged continuous measures (quality road length per
100 sq km, urban population per sq km, average cell-level luminosity, night lights per 1000
inhabitants) does not substantively alter our results.
To account for the potential confounding effects of biogeographic fundamentals and the
legacy of the Neolithic Revolution, we control for calorie-weighted agricultural suitability,
elevation and terrain ruggedness, as well as disease prevalence (malaria and tsetse fly). Al-
ternative explanations of spatial inequality focus on: geographic variability and its effects
on social fractionalization and ethnic inequality (Alesina, Michalopoulos and Papaioannou,
2016); the effects of the slave trades (Nunn, 2008; Nunn and Puga, 2010); variation in
pre-colonial institutions (Boone, 2003; Gennaioli and Rainer, 2007; Michalopoulos and Pa-
paioannou, 2013; Wilfahrt, 2018); and technological changes that reduced shipping costs,
agglomerating economic activities around ports and natural harbors (Henderson et al., 2018;
Ricart-Huguet, 2017). To account for these factors and others, we also control for: distance
to the coast and navigable rivers, pre-colonial political institutions, slave affectedness, dis-
tance to historic trading and explorer routes, late 19th century urbanization, and proximity
to the founding colonial capital.
34
For sources of historical and contemporary development outcomes, all other variables and summary
statistics, see Tables SM1, SM2, SM3, and SM4.

23
Statistical analysis

We employ four different empirical strategies to estimate the long-term effects of colonial re-
source extraction on contemporary infrastructure and development outcomes: (i) simple OLS
regressions with country fixed effects and a host of geographic and historical pre-treatment
controls, (ii) inverse propensity score-weighted regressions, (iii) a spatial randomization in-
ference approach comparing outcomes in historical resource areas to those in 1000 similarly
plausible counterfactual spatial equilibria, and (iv) spatial instrumental variable models in-
strumenting colonial cash crop production with agro-climatic suitability scores while also
accounting for spatial patterns in the outcome variables. Each of these methods has its
advantages and limitations, yet consistent results across all of them increase the credibility
of our findings.

Baseline models. Our baseline models summarized in Figure 2 take the following form:

Yic = αc + β Resource Dummyic + λ Xic + ϵic

Yic is outcome Y for cell i nested in former colony c. The coefficient of interest is β identifying
the effect of colonial cash crop or mineral production. The fixed effects αc control for all
unobserved geographic and historical confounds at the level of colonies and enable an intuitive
interpretation of coefficients as measures of spatial inequality (i.e. cell-level deviation from
the country/colony mean). Xic contains the geographic and pre-colonial baseline controls
described above and ϵic is the error term. All models are estimated via OLS, which yields
linear probability models for the binary outcomes (a)-(c).35 We cluster standard errors at
the level of ethnic regions as defined in Murdock’s map (Murdock, 1959) representing the
closest known albeit imperfect approximation of pre-colonial political entities available for
35
We prefer the linear probability model over more traditional binary choice models due to its intuitive
interpretation of coefficients as marginal effects as well as its versatility in terms of accommodating fixed
effects, instrumental variables, and spatial parameters.

24
the whole continent.36 In Appendix Tables SM7 and SM8, we report detailed regression
tables with coefficient estimates and standard errors for all covariates.

Propensity score weights. As an alternative strategy to account for observable con-

founds, we use propensity score weighting. We first estimate a continent-wide propensity
score model of receiving the cash crop (mineral) treatment based on all geographic and
historic baseline covariates included in the OLS models as well as a cash crop-specific agro-
climatic suitability score from the FAO GAEZ database.37 We then run OLS regressions
with inverse propensity score weights, which under the (admittedly strong) assumption of
no unobserved confounding yield the average treatment effect on the treated (ATT) (Hi-
rano, Imbens and Ridder, 2003). Reweighting dramatically improves balance on observable
pre-treatment characteristics (see Figure SM2). For all 19 covariates across two treatments
(cash crops and minerals) and two samples (global vs. DHS subsample), the standardized
mean differences between treated and control observations remain below the conventional
threshold of 0.1.

Randomization inference: counterfactual spatial equilibria. We further proceed

with a randomization inference-type analysis inspired by Dell and Olken (2019). The basic
intuition is that the observed spatial distribution of colonial resource production is just one
realization of many possible spatial equilibria. According to our data, only 12% of grid
cells with above-median cash crop suitability saw actual production in 1957. In addition to
raw agro-climatic suitability, a host of other geographic and historical variables may have
determined the location of resource extraction (e.g. proximity to pre-colonial trade routes,
navigable rivers, or the coast). To account for this, we construct a set of 1000 similarly
plausible counterfactual distributions of both cash crop and mineral extraction. We regard
alternative spatial resource distributions as plausible if (i) counterfactual cells score similarly
36
In Figure SM16, we show that our results are robust to using Conley standard errors allowing for linearly
decaying spatial correlation within radii of 100, 200, 400, 800, 1600, and 3200 km (Conley, 1999).
37
We take the cell-specific mean across the nine most important African cash crops.

25
on geographic and historical baseline variables as real cash crop (mining) cells, and (ii) the
overall distribution of counterfactual cells exhibit comparable patterns of spatial clustering
as the actually observed resource data. The local determinants of colonial resource extraction
may vary across different crops and minerals as well as geographic and institutional contexts,
especially in the case of decentralized cash crop production by native farmers. However,
soil suitability, proximity to natural and pre-existing trade networks, terrain, the disease
environment, and pre-colonial development plausibly mattered in most contexts. In addition,
the counterfactual analysis enables us to gauge the latent developmental advantage implied
by the non-random assignment to resource extraction and, via the imposed spatial patterns,
safeguards against mistaking spatial clustering in predictors and outcomes for long-term
effects (Kelly, 2019).
We construct counterfactual resource distributions as follows: We treat the realized dis-
tributions of colonial cash crop and mineral production as clustered point processes and
estimate two separate models assuming a Thomas cluster process and fitting a spatial trend
that depends on the geographic and historical covariates described above. Using a clustered
point process model seems appropriate since the distribution of colonial cash crop and min-
eral extraction exhibits stronger spatial clustering than predicted by covariates alone. We
use the estimated parameters from the point process models to simulate 1000 counterfactual
spatial equilibria. We then aggregate the points of each simulation to our quarter-degree grid
and regress our outcome variables on a counterfactual cash crop (mineral) dummy without
adding any additional control variables or fixed effects. We run the same models with the
observed resource variables and compare the actual coefficients with the distribution of 1000
counterfactual estimates. We derive point estimates of the treatment effect by subtracting
the mean of all counterfactual coefficients from the actual estimate. Empirical p-values are
calculated by looking at the position of the actual coefficient within (or outside) the absolute
value distribution of counterfactual coefficients.38 The top panel of Figure 2 contains all point
38 r+1
More formally, the empirical p-values are calculated as n+1 , where r denotes the number of coun-
terfactual coefficients with greater absolute value than the real coefficient and n refers to the number of

26
estimates and empirical p-values, whereas the bottom two panels plot actual coefficients and
the distribution of their counterfactual counterparts.
To assess whether our simulation exercise achieves covariate balance, we also regress all
geographic and historical baseline covariates on both counterfactual and actual resource vari-
ables. Figures SM4 and SM5 show the coefficient distributions and indicate balance across
all observed covariates. Figure SM3 maps out the spatial distribution of our 1000 coun-
terfactual draws and visually compares them to the actual distribution of colonial resource
extraction.

Spatial instrumental variable analysis. All these strategies are based on observables
only and may still yield biased estimates if unobserved confounds varying within countries
make an area more likely to produce cash crops and simultaneously lead to more infras-
tructure investment and economic activity. We address this inferential threat by estimating
2SLS-IV models instrumenting observed cash crop production with the mean agro-climatic
suitability score across the nine most important African cash crops (cocoa, coffee, cotton,
groundnuts, oil palm, tea, sugarcane, tobacco, and bananas). Cash crop suitability is a
valid instrument if it predicts actual production and at the same time does not influence
infrastructure and subnational development through any other casual pathway. We argue
that this exclusion restriction plausibly holds since soil suitability for cash crops is not af-
fected by economic activity and, conditional on our baseline covariates—especially general,
non-cash-crop-related agricultural suitability and ethnic group-level agricultural dependence
prior to colonization—fertile soils for cash crops are unlikely to affect urbanization rates,
state building, infrastructure, and economic activity except through their impact on cash
crop production from the colonial era onwards.
In the Supplementary Materials (Figure SM8), we conduct placebo tests to probe the
plausibility of this assumption. We regress cell-level and ethnic group-level proxies of pre-
colonial development on cash crop suitability and the set of geographic baseline variables
simulations.

27
discussed above. Reassuringly, cash crop suitability is not significantly related to proximity
to pre-colonial trade routes and urban centers, town size, social class stratification, political
centralization, democratic selection of early elites, and the use of slavery (Murdock, 1967).39
However, similar spatial patterns in both our instrument and outcomes pose an additional
threat to the exclusion restriction (Betz, Cook and Hollenbach, 2019). We use a continuous,
spatially interpolated and therefore relatively smoothly varying instrument for a sharply
varying binary treatment constructed from point data, whereas our outcomes tend to vary
less sharply between cells than the colonial resource variables (see Figure 1). Under such
conditions, standard IV methods may yield estimates that are even more biased than simple
OLS (Betz, Cook and Hollenbach, 2019).
To account for spatial dependence in our data, we therefore estimate spatial 2SLS models
that include a spatial lag of the dependent variable instrumented with first- and second-order
spatial lags of the exogenous baseline controls. We do not have strong theoretical priors as to
whether the data generating process in fact follows a spatial lag model. Spatial correlation
in our outcome variables may also be due to spatially clustered unobservables or spillover
effects of right-hand side variables—a possibility we explore in more detail below. Instead,
we choose spatial lag 2SLS models as the most practical method to account for spatial
dependence while at the same time instrumenting for an endogenous treatment variable
(Betz, Cook and Hollenbach, 2019). In our main results, we report estimates from models
that use spatial lags based on an inverse-distance weighted nearest neighbor matrix including
the 24 most proximate cells. The robustness analysis in the Appendix shows that alternative
choices of the spatial weights matrix do not fundamentally change our results (See Figure
SM7). As for instrument strength, cash crop suitability strongly predicts actual production
in our first-stage regressions. The relevant first-stage F statistic is 34.2 in the road, city, and
night light models and 11.0 in the DHS household wealth analysis. We do not instrument
39
There is a statistically significant but substantively small association between cash crop suitability and
distance to European explorer routes indicating that, if anything, highly suitable areas are further away from
these routes.

28
colonial mineral production since we are not aware of a plausibly exogenous predictor of
colonial mining activities.

Results

The four maps in Figure 1 visually present the central argument and empirical approach:
cash crop suitability (upper-left panel), at least partially shaped the distribution of historical
cash crop production (upper-right panel), which in turn structured colonial infrastructure in-
vestments and the development of export-oriented cash crop enclaves (lower-left panel) that
have wrought severe and persistent subnational variation in economic development (lower-
right panel). Figure 2 reports coefficient plots derived from the four different statistical
methods that allow us to address concerns about endogeneity and more credibly estimate
the effect of historical cash crop cultivation on contemporary development outcomes. Our
baseline OLS models indicate that colonial cash crop cells had a 16 percentage points higher
probability of having a quality road in 1998; close to a 20 percentage points higher likelihood
of emitting nighttime lights in 2015; a 19 percentage points higher likelihood of having a
city in 2015; and 14 percent of a standard deviation greater household asset-based wealth.40
Historical cash crop production exhibits a comparable effect on contemporary roads, electri-
fication and cities as colonial mineral extraction, despite the more capital-intensive nature
of mining. The results from the propensity score-weighted regressions, randomization in-
ference and instrumental variable models align closely with the baseline OLS coefficients,
suggesting that our estimates are unlikely to be driven by omitted variables, reverse cau-
sation or measurement error. The S2SLS-IV results are robust to alternative definitions of
what constitutes the spatial neighborhood of a cell (Figure SM7).
40
See Tables SM7 and SM8 for OLS regression results with all controls.

29
Figure 1: Cash crop suitability, colonial cash crop production, infrastructure and subnational
development in Africa.
The blue/green shading in the upper panels shows agro-climatic suitability for cash crops. Each orange point
represents US $289,270 export value of cash crop production in 1957. Red crosses represent mining sites
producing varying export volumes. The lower-left panel illustrates road and railroad infrastructure around
1960. The lower-right panel overlays the colonial resource data with luminosity at night in 2015 as a proxy
for economic activity.

30
Comparing the effects of cash crop production with plausible coun-

terfactual sites

In addition to revealing alternative spatial equilibria that remained largely unrealized during
colonialism (see Figure SM3), the counterfactual analysis allows us to benchmark the latent
developmental effects of the non-random siting of cash crop production (e.g., due to under-
lying agricultural potential or proximity to historical urban centers or the coast) and then
more precisely estimate the impact of actual production. As illustrated in the lower panel in
Figure 2, locational fundamentals and historical processes appear to have predisposed coun-
terfactual cells to higher levels of development than the average cell, but colonial cash crop
production substantially amplified this effect by 136 to 196% depending on the outcome.

Robustness and heterogeneity

Overall the effect sizes are fairly consistent across the major five cash crops, with the excep-
tion of household wealth, in which coffee, cocoa and oil palm outperform cotton and ground-
nuts (See Figure SM9). Country-by-country regressions reported in Figures SM10 and SM11
show the results hold for most countries and are not driven by outliers; they also generally
persist across different imperial powers but are slightly higher in former British colonies (See
Figure SM12). Nor have path dependent effects been disrupted by different post-colonial
trajectories; results are stable across different levels of democracy, conflict-affectedness, and
resource dependence; as well as among landlocked and coastal countries (See Figure SM13).
The Online Appendix to this paper contains additional robustness checks: estimating cash
crop effects at the intensive margin (Figure SM14); using continuous instead of dichotomous
measures of colonial cash crops for the IV analysis (Figure SM15); employing Conley stan-
dard errors with varying cutoffs to account for spatially correlated residuals (Figure SM16);
using different grid cell resolutions and ethnic groups in Murdock as unit of analysis (Figures
SM17 and SM18); and comparing the effects of cash crops on long-run development to the

31
 Main Results
Colonial Resource Extraction & Contemporary Infrastructure and Development
Quality Road 1998 (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) DHS Household Wealth

Cash Crops (OLS) ● ● ● ●

Cash Crops (PS Weights) ● ● ● ●

Cash Crops (RI) ● p = 0.001 ● p = 0.001 ● p = 0.001 ● p = 0.006

Model

Cash Crops (S2SLS) ● ● ● ●

Minerals (OLS) ● ● ● ●

Minerals (PS Weights) ● ● ● ●

Minerals (RI) ● p = 0.004 ● p = 0.058 ● p = 0.01 ● p = 0.004

0.0 0.1 0.2 0.3 0.0 0.1 0.2 0.3 0.0 0.1 0.2 0.3 0.0 0.2 0.4
Coefficients and 95% Confidence Intervals

Treatment Effects: Actual vs. Counterfactual Cash Crop Distribution

1000 simulated cash crop counterfactuals based on clustered point process estimates
Quality Road 1998 (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) DHS Household Wealth
12.5
15
15 10.0
10
Density

10 7.5
10
5 5.0
5 5
2.5

0 0 0 0.0
0.1 0.2 0.3 0.1 0.2 0.3 0.4 0.1 0.2 0.3 0.0 0.1 0.2
Distribution of 1000 Counterfactual Coefficients
2.5 and 97.5 percentiles as well as mean of the counterfactual coefficient distribution in blue; actual treatment coefficient in red.

Treatment Effects: Actual vs. Counterfactual Mineral Distribution

1000 randomly sampled mineral counterfactuals based on clustered point process estimates
Quality Road 1998 (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) DHS Household Wealth
5 8 4

6 4
6 3
Density

3
4 4 2
2
2 2 1
1

0 0 0 0
−0.1 0.0 0.1 0.2 0.3 0.00 0.25 −0.1 0.0 0.1 0.2 0.3 −0.25 0.00 0.25 0.50
Distribution of 1000 Counterfactual Coefficients
2.5 and 97.5 percentiles as well as mean of the counterfactual coefficient distribution in blue; actual treatment coefficient in red.

Figure 2: The effects of colonial cash crop production on long-run development in Africa.
Upper Panel. Estimated effects of colonial cash crop and mineral production on four contemporary infras-
tructure and development outcomes. In each panel, the first and fifth estimates are based on OLS models
with country fixed effects; the second and sixth on propensity-score weighted regressions; the third and sev-
enth on our spatial randomization approach; and the fourth on spatial 2SLS models instrumenting colonial
cash crop production with agro-climatic suitability scores. Lower Panel. Actual versus counterfactual
coefficients from the randomization inference analysis.

32
slave trades and pre-colonial centralization (Figure SM19).

Mechanisms: serial correlation versus path dependence

The results raise the question of what accounts for the persistent effects of the cash crop
revolution on spatial inequality across Africa? We test for two potential pathways: se-
rial correlation and path dependence (Bleakley and Lin, 2012; Jedwab, Kerby and Moradi,
2017). Serial correlation links contemporary patterns of economic development to recurring
direct effects of locational fundamentals (i.e., the geographic or environmental conditions
that spurred the concentration of economic activities in the past exert a similar influence in
the present; Davis and Weinstein (2002)). Accordingly, high levels of contemporary devel-
opment in cash crop zones would be a function of the continuous agricultural production in
these highly suitable areas. In contrast, path dependence points to the increasing returns
that arise not necessarily from underlying drivers of initial economic activities but from the
clustering of capital and infrastructural investments in a given area (Krugman, 1991). Prior
investments not only lower costs going forward but also serve as a coordination mechanism
for subsequent distribution of factors of production (Jedwab, Kerby and Moradi, 2017). Con-
sistent with the pattern visualized in the lower-left panel in Figure 1, we report in Tables
SM9 and SM10 evidence that the historical transition to cash crop agriculture aligned with
colonial investments in infrastructure—not just in transportation networks, but also in power
generation (based on the digitization of a comprehensive map of power stations produced by
the United Nations Economic Commission for Africa from the late 1960s).
We perform a causal mediation analysis to investigate the impact of serial correlation
versus path dependence. We operationalize the serial correlation mechanism by using an
FAO estimate of the market value of total agricultural production across all crops in the
year 2000. To test for path dependence, we use data on early independence roads, railroads,
and electricity generation facilities. For all three of these infrastructure variables, we code
dummies indicating whether a cell is crossed by a colonial road, rail, or hosted a power plant

33
in the early post-independence period. As for the railroads, we only use those that were
constructed for other than mining or military purposes to make sure that the potential for
cash crop exports may have, at least partially, motivated rail construction.
We use sequential g-estimation as a method to test for the relevance of these two mech-
anisms (Acharya, Blackwell and Sen, 2016). This two-step estimation procedure allows to
estimate average controlled direct effects (ACDE) showing how much of the main effect
remains after taking a particular mechanism into account. Valid estimation of the ACDE
relies on the assumption of sequential unconfoundedness, i.e. no omitted variables biasing
the effect of the treatment and/or mediator on the outcome. To address potential violations
of this assumption, we use cash crop suitability instead of actual production in 1957 as the
reduced form treatment variable. In addition, we add all geographic covariates from our
baseline OLS models as potential pre-treatment confounders and all historical covariates as
potential intermediate confounders.
Figure 3 plots the results from our mediation analysis using cash crop suitability as
treatment variable. The suitability effects remain larger if we demediate the development
outcomes with contemporary crop production than with the infrastructure factor. The dif-
ference between the baseline coefficient and the ACDE can be interpreted as the part of the
total effect that is mediated by the respective causal pathways: 30-38% of the cash crop
suitability effect on contemporary development, depending on outcome, is due to early in-
frastructure, whereas only 16% – 22% of these effects is mediated by a continued advantage
in agricultural productivity. The infrastructure channel is even more important if we use
actual cash crop production in the mediation analysis (See Figure SM20). Here, we report
the more conservative demediated analysis with suitability due to potential confounding.
Since most highly suitable cells never saw actual production, the true effect probably lies
somewhere in between. Nonetheless, both results suggest that path dependence more than
serial correlation accounts for the long-run developmental effects of the cash crop revolution.

34
 Causal Mediation Analysis
Total and demediated direct effects (ACDE) of cash crop suitability
Quality Road 1998 (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) DHS Household Wealth
Total and Demediated Effects

Total Effect ● ● ● ●

ACDE (Crop Value) ● ● ● ●

ACDE (Infrastructure) ● ● ● ●

ACDE (Both) ● ● ● ●

0.000 0.025 0.050 0.075 0.100 0.000 0.025 0.050 0.075 0.100 0.000 0.025 0.050 0.075 0.100 0.000 0.025 0.050 0.075 0.100
Estimates and 95% Confidence Intervals (Cluster Bootstrap)

Figure 3: Causal Mediation Analysis.

Total effect of cash crop suitability as well as average controlled direct effects demediated with respect to
contemporary agricultural productivity and early independence infrastructure (rails, roads, power plants).
Throughout the infrastructure channel appears more important than agricultural productivity.

Benchmarking the effects of cash crops on long-run development

versus other geographic variables

To better grasp the importance of our findings, we compare the effects of agro-climatic cash
crop potential on patterns of long-run development to other geographic variables—conditions
favorable for food production to sustain dense populations; disease ecology; distance to coast;
distance to navigable rivers; elevation; and ruggedness. The analysis is at the cell-level—and
enables us to compare relative effects of cash crop suitability to other geographic factors
regularly cited in the literature as key determinants of development patterns in Africa.
All variables are standardized to mean 0 and sd 1 to compare effect sizes. Results,
reported in the lower panel of Figure 3, are at the continental-level (pooled) and within
modern international borders (using fixed effects). Overall the results confirm the impact
of “the daunting nature of Africa’s geography,” (Herbst, 2000) in particular the formidable
costs to trade for the region’s vast interior. Distance to the coast exhibits the largest effects
on contemporary development. Next to coastal distance, the effects of cash crop suitability
rival, and in many cases, surpass other important geographic factors, especially if we focus on
within-country variation. Results remain substantively the same if we use continuous instead
of binary outcome measures (See Figure SM21). One takeaway from this analysis is that

35
 Cash Crop Suitability vs. Other Geographic Fundamentals
Effect of a one−standard deviation increase in geographic covariate
Quality Road 1998 (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) DHS Household Wealth (Z)

Cash Crop Suitability (Pooled) ● ● ● ●

Cash Crop Suitability (Within) ● ● ● ●

Caloric Suitability (Pooled) ● ● ● ●

Caloric Suitability (Within) ● ● ● ●

TseTse Suitability (Pooled) ● ● ● ●

Predictor & Model

TseTse Suitability (Within) ● ● ● ●

Malaria Suitability (Pooled) ● ● ● ●

Malaria Suitability (Within) ● ● ● ●

Dist. Coast (Pooled) ● ● ● ●

Dist. Coast (Within) ● ● ● ●

Dist. Nav. River (Pooled) ● ● ● ●

Dist. Nav. River (Within) ● ● ● ●

Elevation (Pooled) ● ● ● ●

Elevation (Within) ● ● ● ●

Ruggedness (Pooled) ● ● ● ●

Ruggedness (Within) ● ● ● ●

−0.10 −0.05 0.00 0.05 0.10 −0.1 0.0 0.1 −0.05 0.00 0.05 −0.3 −0.2 −0.1 0.0 0.1 0.2
Coefficients and 95% Confidence Intervals

Figure 4: Benchmarking of coefficient sizes: Cash Crop Suitability vs. Other Locational
Fundamentals.
Compares the effects of a one-standard deviation increase in different geographic variables on contemporary
outcomes. Effects are estimated from OLS models with and without country fixed effects.

the prospective yield from the 19th century onward of agricultural crops of high commercial
(and limited nutritional) value was at least as consequential in shaping subnational wealth
disparities in Africa as the Neolithic potential to feed dense populations (Compare rows 1
and 2 in Figure 4 to rows 3 and 4).

Spillover effects, country-level inequality, and conflict legacies

In the final part of our analysis, we move beyond purely local effects of colonial cash crop
production. So far, our estimates show large effects on patterns of subnational development:
‘treated’ cells continue to be significantly better off than comparable cells within the same
country. The broader implications of these findings depend on whether and how colonial
resource extraction affected development outcomes beyond the immediate neighborhood.
Large-scale agglomeration patterns, as well as forward and backward linkages to other sec-
tors of the economy, may have spurred even more positive effects than our local estimates
suggest. As discussed, however, resource-exporting African countries are often characterized
as enclave economies with limited potential for positive spillovers. Worse yet, the positive
estimates we find may in part be due to adverse effects on resource-poor areas. Such negative

36
spillovers seem particularly likely under colonial institutions where extractive policies, such
as excessive taxation, labor coercion, restricted mobility, and the protection of European
economic interests, were common practice.
To gauge the effects of colonial resource extraction beyond the very local level, we run
models including variables for proximity to colonial cash crop and mineral production. More
specifically, we construct distance band dummy variables coding, for each grid cell, whether
it falls within 0-25, 25-50 …225-250 km of a colonial resource point. We then re-estimate
our baseline models replacing cell-level resource indicators with 20 distance band dummies
capturing proximity to both cash crop and mineral extraction. More formally, we run the
following specification:

∑
225−250 ∑
225−250
Yic = αc + βj Dist. Cropsjic + γj Dist. Mineralsjic + λ Xic + ϵic
j=0−25 j=0−25

The vectors βj and γj contain the coefficient estimates for all distance band dummies to
cash crops and minerals. Cells further away than 250 km from colonial resource extraction
serve as the baseline category. Xic and αc refer to the same outcomes, control variables and
country fixed effects used in our baseline analysis.
Positive coefficients on distance band indicators beyond 25 km would suggest positive
spillovers. Negative coefficients would be consistent with adverse spillover effects making
cells close to colonial resource extraction worse off than predicted by their baseline covariates.
Figure 5 presents coefficients for distance to cash crops. (See Figure SM22 for spillover
analysis with continuous outcomes). Across all outcomes, we find positive and significant
spillovers up to 50 km. For nighttime luminosity and urbanization, we find negative and
mostly significant spillovers between 75 and 250 km from colonial cash crop sites. These
results suggest that concentrated investments in colonial cash crop enclaves crowded out
development in other areas, which appear worse off today than predicted by geographic
conditions and precolonial factors. The effects on household wealth are neutral beyond 50
km, but note that only 24% of our grid cells contain DHS enumeration areas. The more

37
positive findings for road infrastructure are unsurprising, given that paved and improved road
segments typically cut through multiple cells and extensions of the road network mostly
connect pre-existing segments. Interestingly, the distance to mining coefficients exhibits
positive spillovers over a somewhat broader range and never turns negative and significant
(See Figure SM23).

Spillover Analysis: Cash Crops

Distance to Colonial Resources & Contemporary Outcomes
Quality Road 1998 (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) DHS Household Wealth

Cash Crops (0−25 km) ● ● ● ●

Cash Crops (25−50 km) ● ● ● ●

Cash Crops (50−75 km) ● ● ● ●

Distance Bins

Cash Crops (75−100 km) ● ● ● ●

Cash Crops (100−125 km) ● ● ● ●

Cash Crops (125−150 km) ● ● ● ●

Cash Crops (150−175 km) ● ● ● ●

Cash Crops (175−200 km) ● ● ● ●

Cash Crops (200−225 km) ● ● ● ●

Cash Crops (225−250 km) ● ● ● ●

0.0 0.1 0.2 −0.1 0.0 0.1 0.2 0.0 0.1 −0.1 0.0 0.1 0.2
Coefficients and 95% Confidence Intervals

Figure 5: Spillover effects of Colonial Cash Crop Production

Limited or even negative spillover effects of geographically concentrated colonial resource

enclaves suggest a legacy of severe spatial inequality. A descriptive analysis of country-level
aggregate inequality measures confirms this impression. We calculate Gini coefficients of
night light inequality in 2015 for each country in our sample (based on the distribution of
night light intensity in 2015 across quarter degree grid cells). Figure 6 plots this measure of
contemporary spatial inequality across an analogously constructed Gini coefficient of spatial
inequality in colonial resource extraction (based on the cell-level sum of production value in
1957 as coded by Hance, Kotschar and Peterec (1961)). Figure 6 shows a strong and posi-
tive correlation between the geographic concentration of historical resource extraction and
contemporary economic inequality. In combination with the spillover analysis, these associ-
ations suggest colonial extractive economies not only shaped local patterns of development
but produced macro-level consequences as well that continue to underlie African economic
geography.

38
 Colonial Resources & Contemporary Spatial Inequality Colonial Resources & Contemporary Spatial Inequality
Country−level Ginis based on 0.25° Cells Country−level Ginis based on 0.25° Cells (Outliers Removed)
1.00 COD TCD MRT
CAF
● ● ● MLI ● 1.00 COD CAF
●
TCD
MRT
● SOM ●● ● ● SSD ● ●
MLI ●
AGO ● AGO
NER SOM ● ● ●
● SDN ● ●
● NER ●
CMR COG CMR SSD
● SDN
● ●
MOZ GAB ZMB COG
● ● ● ●
GNB ●
ERI ZMB
● KEN
GIN ●● ● ● DJI GNB ● ●
GAB
●●
0.95 ●
NGA
TZA ETH ●
MOZ ERI
● ● KEN DJI
BFA ●
SEN GIN ● ●
●
● 0.95 NGA ETH BFA
Night Light Gini 2015

Night Light Gini 2015

GMB ● SEN ● TZA ●
UGA ZWE LBR
● ● ●
●
SLE UGA
●
BEN ●
ZWE
● ●
0.90 SLE
● LBR
BEN
●

GNQ
0.90
● ●
GHA

0.85 MWI
●

GNQ
TGO ● ●
● GHA
●
BDI CIV
●
0.85 MWI
●
RWA
●
0.80
TGO
●
●
CIV

0.4 0.6 0.8 1.0 0.8 0.9 1.0

Colonial Resource Gini (Cash Crops & Minerals) Colonial Resource Gini (Cash Crops & Minerals)
β = 0.236 R = 0.406
2
β = 0.431 R = 0.359
2

Figure 6: Colonial Resource and Contemporary Night Light Inequality at the Country Level

Legacies of Conflict?
Colonial Resources & Post−Independence Ethnic Conflict

Cash Crops (Y/N)

Minerals (Y/N)

Cash Crop Value (skqm, log)

Predictor

Mineral Value (sqkm, log)

Cash Crop Value (per capita, log)

Mineral Value (per capita, log)

−0.2 −0.1 0.0 0.1

Coefficients and 95% Confidence Intervals

Figure 7: Colonial Resource Endowments and Post-Independence Ethnic Rebellion

39
 In a final analysis, we investigate whether the spatial distribution of colonial resources
helps to account for patterns of post-colonial ethnic conflict. Unequal economic opportuni-
ties create potential for distributive conflict that may, in extreme cases, result in political
violence. Where economic inequalities align with politically salient ethnic group identities,
widespread grievances among relatively poor groups motivate armed mobilization against
the state (Stewart, 2008; Cederman, Gleditsch and Buhaug, 2013). In addition, economic
disadvantages and lower levels of historically inherited local state capacity provide armed ac-
tors with opportunities for recruitment and mobilization in ethnic settlement areas without
colonial resources (Collier and Hoeffler, 2004; Fearon and Laitin, 2003).
We use ethnic group-level data on settlement areas and rebellion from the Ethnic Power
Relations (EPR) dataset (Vogt et al., 2015) to test this notion.41 For all 36 countries in
our sample, we identify the ethnic groups that EPR codes as political relevant in the first
year of independence. We then aggregate our colonial resource data to the polygons of these
150 groups’ main settlement areas and code, for each group, whether it has been involved
in an ethnic civil war between independence and 2017. 35.7% of ethnic groups that did not
see any colonial cash crop production have fought at least one civil war post-independence,
whereas the number for groups with colonial cash crops is only 24.5%. The respective conflict
shares for groups with and without colonial mineral extraction are 24.1% and 28.7%. Figure
7 presents estimates from three linear probability models with country fixed effects, the
geographic and historical control variables from above, and Conley errors with a distance
cutoff of 400 km. Colonial cash crop production is associated with a significantly lower
likelihood of ethnic rebellion, regardless of whether we operationalize resource endowments
as binary indicators or logged 1960 production values per sqkm or capita. The estimates for
mineral resource endowments remain close to zero and statistically insignificant. This null
finding may have to do with less widespread economic benefits of mineral extraction or the
41
EPR provides lists of politically relevant ethnic groups for all independent countries since 1945 and codes
information about these groups’ political representation and involvement in intrastate conflict as defined by
the UCDP/PRIO dataset (Gleditsch et al., 2002). The GeoEPR companion data set provides spatial polygon
data on the main settlement areas of relvant ethnic groups (Wucherpfennig et al., 2011).

40
attractiveness of capital-intensive fixed point resources as a target for armed actors (Lujala,
2010; Dube and Vargas, 2013; Berman et al., 2017). Overall, however, our conflict analysis
suggests that the legacies of colonial resource extraction are not merely economic but, at
least in the case of cash crops, extend to political outcomes such as stability and conflict.

Discussion

We argue Africa’s contemporary economic geography was shaped by the cash crop revolution
that swept the continent from the late 1700s onward. The end of the slave trades and high
global demand for agricultural commodities increased the geographical advantages of areas
suitable for cash crop cultivation. The adoption and spread of commercial export agricul-
ture in the 19th and 20th centuries contributed to the emergence of new spatial equilibria
centering around highly-suitable regions. However, as the cash crop revolution also impelled
the imperial conquest of Africa, colonialism shaped these agglomerating processes. While
European capital fueled a revolution in transportation infrastructure that was a boon to
commercial export agriculture, extractive and mercantilist institutions structured the dis-
tribution of gains arising from this new mode of production and its broader impact on the
economy. Consequently, as the colonial cash crop revolution produced powerful and endur-
ing agglomerating effects, these came at the expense of surrounding areas, which ended up
significantly worse off than otherwise would be expected based on their underlying charac-
teristics.
An important scope condition of our analysis, however, is, given the extent of colonialism
in Africa, it is difficult to know the counterfactual of the long-run effects of commercial export
agriculture without colonial institutions and investments. As one of the few exceptions,
Ethiopia is potentially instructive. Coffee exports took off in Ethiopia in the first part of the
20th century, spurring the integration of the southern highlands into the Ethiopian Empire
through infrastructural and administrative investments. In turn, revenue from the coffee

41
trade contributed to state centralization and modernization (McClellan, 1980). Interestingly,
despite coffee’s economic importance, we do not see evidence of significant spatial disparities
between Ethiopia’s cash crop agricultural zones and the rest of the state (See Figure SM10).
Whether this is due to the absence of mercantilist institutions and stronger integrative
economic processes, a longer history of centralized statehood, or the (downward) leveling
effects of the Marxist revolution that brought an end to the Ethiopian Empire in the 1970s
requires future research.
Going beyond this single case, another fruitful avenue for research would be to scale up
to a global sample to leverage greater variation in the distribution of colonial institutions
and commercial export agriculture. Surprisingly, despite cash crop agriculture’s importance
to the emergence of early modern globalization from the 16th century onward, its impact on
spatial patterns of development around the world has largely been understudied.42 If Africa
is any indication, this represents a significant omission in our study of the subnational wealth
and inequality of nations.

42
For notable exceptions studying China, Indonesia, and Costa Rica, see Marden (2016), Edwards (2019),
and Méndez-Chacon and Patten (2019), respectively.

42
References
Acemoglu, Daron and James A. Robinson. 2012. Why Nations Fail: The Origins of Power,
Prosperity and Poverty. New York: Crown Publishers.
Acemoglu, Daron, Simon Johnson and James A. Robinson. 2001. “The Colonial Origins
of Comparative Development: An Empirical Investigation.” American Economic Review
91(5):1369–1401.
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.
Acharya, Avidit, Matthew Blackwell and Maya Sen. 2016. “Explaining Causal Findings
Without Bias: Detecting and Assessing Direct Effects.” American Political Science Review
110(3):512–529.
Ajayi, Ade and Michael Crowder. 1985. Historical Atlas of Africa. Cambridge, UK: Cam-
bridge University Press.
Alesina, Alberto, Stelios Michalopoulos and Elias Papaioannou. 2016. “Ethnic Inequality.”
Journal of Political Economy 124(2):428–488.
Alsan, Marcella. 2014. “The Effect of the Tsetse Fly on African Development.” The American
Economic Review 105(1):382–410.
Amin, Samir. 1972. “Underdevelopment and dependence in Black Africa—origins and con-
temporary forms.” The Journal of Modern African Studies 10(4):503–524.
Amin, Samir. 1974. Modern Migrations in Western Africa: Studies Presented and Discussed
at the Eleventh International African Seminar, Dakar, April 1972. London: Oxford
University Press.
Austin, Gareth. 1996. Mode of Production or Mode of Cultivation: Explaining the Failure
of European Cocoa Planters in Competition with African Farmers tn Colonial Ghana.
Springer pp. 154–175.
Austin, Gareth. 2008. “Resources, Techniques, and Strategies South of the Sahara: Revising
the Factor Endowments Perspective on African Economic Development, 1500–2000.” The
Economic History Review 61(3):587–624.
Austin, Gareth. 2009. “Cash Crops and Freedom: Export Agriculture and the Decline of
Slavery in Colonial West Africa.” International Review of Social History 54(01):1–37.
Austin, Gareth. 2010. “African Economic Development and Colonial Legacies.” International
Development Policy/ Revue Internationale de Politique Développment 1(1):11–32.
Auty, Richard M. 1997. “Natural resource endowment, the state and development strategy.”
Journal of International Development: The Journal of the Development Studies Associa-
tion 9(4):651–663.

43
Auty, Richard M., ed. 2001. Resource abundance and economic development. Oxford Uni-
versity Press.

Baldwin, Robert E. 1956. “Patterns of development in newly settled regions.” The Manch-
ester School 24(2):161–179.

Bates, Robert H. 1981. Markets and States in Tropical Africa. Berkeley, CA: University of
California Press.

Bates, Robert H. 1983. Essays on the Political Economy of Rural Africa. University of
California Press.

Berman, Nicolas, Mathieu Couttenier, Dominic Rohner and Mathias Thoenig. 2017. “This
Mine is Mine! How Minerals Fuel Conflicts in Africa.” American Economic Review
107(6):1564–1610.

Betz, Timm, Scott J Cook and Florian M Hollenbach. 2019. “Spatial interdependence and
instrumental variable models.” Political Science Research and Methods pp. 1–16.

Bleakley, Hoyt and Jeffrey Lin. 2012. “Portage and Path Dependence.” The Quarterly Journal
of Economics pp. 587–644.

Boone, Catherine. 1992. Merchant Capital and the Roots of State Power in Senegal, 1930-
1985. New York: Cambridge University Press.

Boone, Catherine. 2003. Political topographies of the African state: Territorial authority and
institutional choice. Cambridge University Press.

Brooks, George E. 1975. “Peanuts and Colonialism: Consequences of the Commercialization

of Peanuts in West Africa, 1830–70.” The Journal of African History 16(01):29–54.

Cardoso, Fernando Henrique and Enzo Faletto. 1979. Dependency and development in Latin
America (Dependencia y desarrollo en América Latina, engl.). Univ of California Press.

Cederman, Lars-Erik, Kristian Skrede Gleditsch and Halvard Buhaug. 2013. Inequality,
Grievances, and Civil War. New York: Cambridge University Press.

Cederman, Lars-Erik, Nils B. Weidmann and Kristian Skrede Gleditsch. 2011. “Horizontal
Inequalities and Ethnonationalist Civil War: A Global Comparison.” American Political
Science Review 105(03):478–495.

Cervellati, Matteo, Giorgio Chiovelli and Elena Esposito. 2019. Bite and Divide: Malaria
and Ethnolinguistic Diversity. Technical report CEPR Discussion Papers.

Chaves, Isaias, Stanley L. Engerman and James A. Robinson. 2014. Reinventing the Wheel:
The Economic Benefts of Wheeled Transportation in Early Colonial British West Africa.
New York: Cambridge University Press.

Collier, Paul and Anke Hoeffler. 2004. “Greed and Grievance in Civil War.” Oxford economic
papers 56(4):563–595.

44
Conley, Timothy G. 1999. “GMM estimation with cross sectional dependence.” Journal of
Econometrics 92(1):1–45.

Conning, Jonathan H and James A Robinson. 2009. “Enclaves and development: An empir-
ical assessment.” Studies in Comparative International Development 44(4):359–385.

Coquery-Vidrovitch, Catherine. 1975. “Research on an African Mode of Production.” Cri-

tique of Anthropology 2(4-5):38–71.

Cordell, Dennis D., Joel W. Gregory and Victor Piché. 1996. Hoe and Wage : A Social
History of a Circular Migration System in West Africa. Boulder, Colo.: Westview Press.

Crowder, Michael. 1978. Colonial West Africa: Collected Essays. London: F. Cass.

Daumalin, Xavier. 2004. Commercial presence, colonial penetration: Marseille traders in

west Africa in the nineteenth century. In From Slave Trade to Empire: European Coloni-
sation of Black Africa 1780s-1880s, ed. Olivier Pétré-Grenouilleau. Routledge pp. 225–246.

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

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

Dell, Melissa and Benjamin A Olken. 2019. “The Development Effects Of The Extractive
Colonial Economy: The Dutch Cultivation System In Java.” The Review of Economic
Studies .

Diamond, Jared M. 1997. Guns, Germs, and Steel: The Fates of Human Societies. New
York: W.W. Norton.

Dickson, Kwamina B. 1969. A Historical Geography of Ghana. New York: Cambridge

University Press.

dos Santos, Theotonio. 1970. “The structure of dependence.” The american economic review
60(2):231–236.

Dube, Oeindrila and Juan F Vargas. 2013. “Commodity price shocks and civil conflict:
Evidence from Colombia.” The Review of Economic Studies 80(4):1384–1421.

Easterly, William. 2007. “Inequality does cause underdevelopment: Insights from a new
instrument.” Journal of development economics 84(2):755–776.

Easterly, William and Ross Levine. 2003. “Tropics, Germs, and Crops: How Endowments
Influence Economic Development.” Journal of Monetary Economics 50(1):3–39.

Edwards, Ryan B. 2019. “Export Agriculture and Rural Poverty: Evidence from Indonesian
Palm Oil.” Unpublished Manuscript .

45
Eltis, David. 2013. The slave trade commercial agriculture in an African context. Boydell
Brewer Ltd p. 28–53.

Engerman, Stanley L. and Kenneth L. Sokoloff. 1997. Factor Endowments, Institutions,

and Differential Paths of Growth Among New World Economies. Stanford, CA: Stanford
Unviversity Press.

Enke, Benjamin. 2019. “Kinship, cooperation, and the evolution of moral systems.” The
Quarterly Journal of Economics 134(2):953–1019.

Fanon, Frantz. 1963. The Wretched of the Earth. London: Grove Press.

Fearon, James D and David D Laitin. 2003. “Ethnicity, Insurgency, and Civil War.” American
Political Science Review 97(01):75–90.

Fenske, James. 2014. “Ecology, Trade, and States In Pre-Colonial Africa.” Journal of the
European Economic Association 12(3):612–640.

Frank, Andre Gunder. 1966. “The development of underdevelopment.” Monthly Review

18:17–31.

Frankema, Ewout, Jeffrey Williamson and Pieter Woltjer. 2018. “An economic rationale for
the West African scramble? The commercial transition and the commodity price boom of
1835–1885.” The Journal of Economic History 78(1):231–267.

Galletti, Robert. 1956. Nigerian cocoa farmers: an economic survey of Yoruba cocoa farming
families. Oxford University Press.

Gallup, John Luke, Jeffrey D Sachs and Andrew D Mellinger. 1999. “Geography and Eco-
nomic Development.” International Regional Science Review 22(2):179–232.

Gardner, Leigh A. 2012. Taxing Colonial Africa The Political Economy of British Imperial-
ism. Oxford: Oxford University Press.

Gennaioli, Nicola and Ilia Rainer. 2007. “The Modern Impact of Precolonial Centralization
in Africa.” Journal of Economic Growth 12(3):185–234–234.

Gleditsch, Nils Petter, Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg and Håvard
Strand. 2002. “Armed conflict 1946-2001: A new dataset.” Journal of peace research
39(5):615–637.

Green, Toby. 2013. The export of rice millet from Upper Guinea into the sixteenth-century
Atlantic trade. Boydell Brewer Ltd p. 79–97.

Hall, Robert E and Charles I Jones. 1999. “Why do some countries produce so much more
output per worker than others?” The quarterly journal of economics 114(1):83–116.

Hance, William A, Vincent Kotschar and Richard J Peterec. 1961. “Source Areas of Export
Production in Tropical Africa.” Geographical Review 51(4):487–499.

46
Hance, William Adams. 1964. The Geography of Modern Africa. New York: Columbia
University Press.

Havinden, Michael Ashley and David Meredith. 1993. Colonialism and Development: Britain
and its Tropical Colonies, 1850-1960. New York: Routledge.

Headrick, Daniel R. 2012. Power over peoples: Technology, environments, and Western
imperialism, 1400 to the present. Princeton University Press.

Henderson, J Vernon, Tim Squires, Adam Storeygard and David Weil. 2018. “The Global
Distribution of Economic Activity: Nature, History, and the Role of Trade1.” The Quar-
terly Journal of Economics 133(1):357–406.

Herbst, Jeffrey. 2000. States and Power in Africa: Comparative Lessons in Authority and
Control. Princeton, NJ: Princeton University Press.

Hill, Polly. 1957. The Gold Coast Cocoa Farmer: a Preliminary Survey. Oxford University
Press.

Hill, Polly. 1963. The migrant cocoa-farmers of southern Ghana: A study in rural capitalism.
Cambrdige University Press.

Hirano, Keisuke, Guido W Imbens and Geert Ridder. 2003. “Efficient estimation of average
treatment effects using the estimated propensity score.” Econometrica 71(4):1161–1189.

Hirschman, Albert O. 1958. The Strategy of Economic Development. Yale University Press.

Hirschman, Albert O. 1977. “A generalized linkage approach to development, with special

reference to staples.” Economic development and cultural change 25:67.

Hodler, Roland and Paul A. Raschky. 2014. “Regional Favoritism.” The Quarterly Journal
of Economics 129(2):995–1033.

Hogendorn, Jan S. 1969. Economic Initiative and African Cash Farming: Pre-Colonial
Origins and Early Colonial Developments. In Colonialism in Africa, 1870-1960: The
Economics of Colonialism, ed. Lewis H. Gann and Duignan Peter. London: Cambridge
University Press pp. 283–328.

Hogendorn, Jan S. 1978. Nigerian Groundnut Exports: Origins and Early Development.
Zaria: Ahmadu Bello University Press.

Hopkins, Anthony G. 1973. An Economic History of West Africa. New York: Columbia
University Press.

Huillery, Elise. 2009. “History Matters: The Long-Term Impact of Colonial Public invest-
ments in French West Africa.” American Economic Journal: Applied Economics pp. 176–
215.

47
Huillery, Elise. 2011. “The Impact of European Settlement within French West Africa: Did
Pre-Colonial Prosperous Areas Fall Behind?” Journal of African Economies 20(2):263–
311.

Hyam, Ronald. 1968. Elgin and Churchill at the Colonial Office, 1905-08. Springer.

Inikori, Joseph E. 2007. “Africa and the globalization process: western Africa, 1450–1850.”
Journal of Global History 2(1):63–86.

Inikori, Joseph E. 2014. Reversal of Fortune and Socioeconomic Development in the Atlantic
World: A Comparative Examination of West Africa and the Americas, 1400–1850. In
Africa’s Development in Historical Perspective, ed. Emmanuel Akyeampong, Robert H
Bates, Nathan Nunn and James Robinson. New York: Cambridge University Press pp. 56–
88.

Innis, Harold Adams. 1930. The fur trade in Canada: An introduction to Canadian economic
history. Yale University Press.

Jedwab, Remi and Alexander Moradi. 2012. “Colonial investments and long-term devel-
opment in Africa: Evidence from Ghanaian railways.” Unpublished manuscript, George
Washington University and Sussex University .

Jedwab, Remi and Alexander Moradi. 2016. “The Permanent Effects of Transportation
Revolutions in Poor Countries: Evidence from Africa.” Review of Economics and Statistics
98(2):268–284.

Jedwab, Remi, Edward Kerby and Alexander Moradi. 2017. “History, Path Dependence
and Development: Evidence from Colonial Railways, Settlers and Cities in Kenya.” The
Economic Journal pp. 1467–1494.

Jedwab, Rémi, Felix Meier zu Selhausen, Alexander Moradi et al. 2019. The Economics of
Missionary Expansion: Evidence from Africa and Implications for Development. Technical
report African Economic History Network.

Kasara, Kimuli. 2007. “Tax Me If You Can: Ethnic Geography, Democracy, and the Taxation
of Agriculture in Africa.” American Political Science Review 101(1):159–172.

Kelly, Morgan. 2019. “The Standard Errors of Persistence.” CEPR Discussion Paper 13783
.

Kerr, Rachel Bezner. 2010. “Groundnuts as ‘economic crop’ or ‘wife of the home’ in Northern
Nyasaland.” Journal of Historical Geography 36(1):79–89.

Krugman, Paul. 1991. “Increasing Returns and Economic Geography.” Journal of political
economy 99(3):483–499.

Lange, Matthew. 2009. Lineages of Despotism and Development: British Colonialism and
State Power. Chicago: The University of Chicago Press.

48
Leonard, David K. and Scott Straus. 2003. Africa’s Stalled Development: International
Causes and Cures. Boulder, CO: Lynne Rienner Publishers.

Lowes, Sara and Eduardo Montero. 2017. “Blood Rubber: The Effects of Labor Coercion
on Institutions and Culture in the DRC.”.

Lujala, Päivi. 2010. “The Spoils of Nature: Armed civil conflict and rebel access to natural
resources.” Journal of peace research 47(1):15–28.

Lynn, Martin. 1997. Commerce and Economic Change in West Africa: The Palm Oil Trade
in the Nineteenth Century. New York: Cambridge University Press.

Mamo, Nemera, Sambit Bhattacharyya and Alexander Moradi. 2019. “Intensive and exten-
sive margins of mining and development: Evidence from Sub-Saharan Africa.” Journal of
Development Economics 139:28–49.

Manning, Patrick. 1990. Slavery and African life: Occidental, Oriental, and African slave
trades. Vol. 67 Cambridge University Press.

Marden, Samuel. 2016. “The Agricultural Roots of Industrial Development: ‘Forward Link-
ages’ in Reform-Era China.” Unpublished Manuscript .

Marshall, Fiona and Elisabeth Hildebrand. 2002. “Cattle before crops: the beginnings of
food production in Africa.” Journal of World Prehistory 16(2):99–143.

Mayshar, Joram, Omer Moav, Zvika Neeman and Luigi Pascali. 2018. “The Emergence of
Hierarchies and States: Productivity vs. Appropriability.” Working Paper.

McClellan, Charles W. 1980. “Land, Labor, and Coffee: The South’s Role in Ethiopian
Self-Reliance, 1889-1935.” African Economic History (9):69–83.

McPhee, Allan. 1926. The Economic Revolution in British West Africa. London: G. Rout-
ledge Sons.

Meredith, David. 1988. “The Colonial Office, British business interests and the reform of
cocoa marketing in West Africa, 1937–1945.” The Journal of African History 29(2):285–
300.

Michalopoulos, Stelios. 2012. “The Origins of Ethnolinguistic Diversity.” The American

Economic Review 102(4):1508.

Michalopoulos, Stelios and Elias Papaioannou. 2013. “Pre-Colonial Ethnic Institutions and
Contemporary African Development.” Econometrica 81(1):113–152.

Michalopoulos, Stelios and Elias Papaioannou. 2016. “The Long-run Effects of the Scramble
for Africa.” The American Economic Review 106(7):1802–1848.

Michalopoulos, Stelios and Elias Papaioannou. 2018. “Spatial patterns of development: A

meso approach.” Annual Review of Economics 10:383–410.

49
Michalopoulos, Stelios and Elias Papaioannou. 2020. “Historical legacies and African devel-
opment.” Journal of Economic Literature 58(1):53–128.

Michalopoulos, Stelios, Louis Putterman and David N Weil. 2019. “The influence of ancestral
lifeways on individual economic outcomes in Sub-Saharan Africa.” Journal of the European
Economic Association 17(4):1186–1231.

Moradi, Alexander, Gareth Austin and Jörg Baten. 2013. Heights and development in a cash-
crop colony: living standards in Ghana, 1870-1980. Technical report African Economic
History Network.

Moseley, Katherine P. 1992. “Caravel and Caravan: West Africa and the World-Economies,
ca. 900-1900 AD.” Review (Fernand Braudel Center) pp. 523–555.

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

Murdock, George Peter. 1967. “Ethnographic Atlas: A Summary.” Ethnology 6(2):109–236.

Méndez-Chacon, Esteban and Diana Van Patten. 2019. “Multinationals, Monopsony and
Local Development: Evidence from the United Fruit Company.” Unpublished Manuscript
.

Newbury, Colin W. 1966. “North African and Western Sudan Trade in the Nineteenth
Century: A Re-evaluation.” The Journal of African History 7(2):233–246.

Nordhaus, William D. 2006. “Geography and Macroeconomics: New Data and New Find-
ings.” Proceedings of the National Academy of Sciences of the United States of America
103(10):3510–3517.

North, Douglass C. 1959. “Agriculture in regional economic growth.” Journal of Farm Eco-
nomics 41(5):943–951.

North, Douglass C, John Joseph Wallis, Barry R Weingast et al. 2009. Violence and social
orders: A conceptual framework for interpreting recorded human history. Cambridge
University Press.

Nunn, Nathan. 2008. “The Long-Term Effects of Africa’s Slave Trades.” The Quarterly
Journal of Economics 123(1):139–176.

Nunn, Nathan and Diego Puga. 2010. “Ruggedness: The Blessing of Bad Geography in
Africa.” Review of Economics and Statistics 94(1):20–36.

Nunn, Nathan and Leonard Wantchekon. 2011. “The Slave Trade and the Origins of Mistrust
in Africa.” The American Economic Review 101(7):3221–3252.

Nunn, Nathan and Nancy Qian. 2011. “The Potato’s Contribution to Population and Ur-
banization: Evidence from a Historical Experiment.” The Quarterly Journal of Economics
126(2):593–650.

50
Olsson, Ola and Douglas A Hibbs. 2005. “Biogeography and Long-Run Economic Develop-
ment.” European Economic Review 49(4):909–938.

Osafo-Kwaako, Philip and James A. Robinson. 2013. “Political Centralization in Pre-Colonial

Africa.” Journal of Comparative Economics 41(1):6–21.

Ostby, Gudrun. 2008. “Polarization, Horizontal Inequalities and Violent Civil Conflict.”
Journal of Peace Research 45(2):143–162.

Price, Jacob M. 1974. “Economic Function and Growth of American Port Towns in the
Eighteenth Century.” Perspectives in American History 8:123–186.

Raynaut, Claude. 2002. Populations and Land: Multiple Dynamics and Contrasting Realities.
New York: Routledge.

Ricart-Huguet, Joan. 2017. “The origins of colonial investments in former British and French
Africa.” Princeton University .

Rodney, Walter. 1972. How Europe Underdeveloped Africa. London: Bogle-L’Ouverture

Publications.

Román, Miguel O, Zhuosen Wang, Qingsong Sun, Virginia Kalb, Steven D Miller, Andrew
Molthan, Lori Schultz, Jordan Bell, Eleanor C Stokes, Bhartendu Pandey et al. 2018.
“NASA’s Black Marble nighttime lights product suite.” Remote Sensing of Environment
210:113–143.

Sachs, Jeffrey D, Andrew D Mellinger and John L Gallup. 2001. “The Geography of Poverty
and Wealth.” Scientific American 284(3):70–75.

Sachs, Jeffrey D and Andrew M Warner. 2001. “The curse of natural resources.” European
economic review 45(4-6):827–838.

Scott, James C. 2017. Against the Grain: A Deep history of the Earliest States. New Haven,
CT: Yale University Press.

Shaver, Andrew, David B Carter and Tsering Wangyal Shawa. 2019. “Terrain ruggedness
and land cover: Improved data for most research designs.” Conflict Management and Peace
Science 36:191–218.

Smith, Adam. 1776. An Inquiry into the Nature and Causes of the Wealth of Nations.
London: W. Strahan and T. Cadell.

Sokoloff, Kenneth L and Stanley L Engerman. 2000. “Institutions, factor endowments, and
paths of development in the new world.” Journal of Economic perspectives 14(3):217–232.

Stewart, Frances. 2008. Horizontal Inequalities and Conflict: Understanding Group Violence
in Multiethnic Societies. New York: Palgrave Macmillan.

van de Walle, Nicolas. 2009. “The Institutional Origins of Inequality in Sub-Saharan Africa.”
Annual Review of Political Science 12:307–327.

51
Van der Ploeg, Frederick. 2011. “Natural resources: curse or blessing?” Journal of Economic
literature 49(2):366–420.

Vogt, Manuel, Nils-Christian Bormann, Seraina Rüegger, Lars-Erik Cederman, Philipp M

Hunziker and Luc Girardin. 2015. “Integrating Data on Ethnicity, Geography, and
Conflict: The Ethnic Power Relations Dataset Family.” Journal of Conflict Resolution
59(7):1327–1342.

Weiss, Daniel J, Samir Bhatt, Bonnie Mappin, Thomas P Van Boeckel, David L Smith,
Simon I Hay and Peter W Gething. 2014. “Air temperature suitability for Plasmodium
falciparum malaria transmission in Africa 2000-2012: a high-resolution spatiotemporal
prediction.” Malaria Journal 13(1):171.

Wellesley, Dorothy. 1934. Sir George Goldie, Founder of Nigeria: A Memoir. Macmillan.

Wilfahrt, Martha. 2018. “Precolonial legacies and institutional congruence in public goods
delivery: Evidence from decentralized West Africa.” World Politics 70(2):239–274.

Wolpe, Harold. 1972. “Capitalism and Cheap Labour-Power in South Africa: From Segre-
gation to Apartheid.” Economy and Society 1(4):425–456.

Wucherpfennig, Julian, Nils B Weidmann, Luc Girardin, Lars-Erik Cederman and Andreas
Wimmer. 2011. “Politically Relevant Ethnic Groups across Space and Time: Introducing
the GeoEPR Dataset.” Conflict Management and Peace Science 28(5):423–437.

Young, Crawford. 1994. The African Colonial State in Comparative Perspective. Yale Uni-
versity Press.

52
The Cash Crop Revolution, Colonialism and Legacies of
Spatial Inequality: Evidence from Africa

Supplementary Materials
SM1 Table of Contents

Table SM1. Description and Sources of Variables

Table SM2. Description and Sources of Variables (continued).

Table SM3. Summary Statistics (0.25 Degree Grid Cells)

Table SM4. Summary Statistics (Ethnic Homeland-Country Pairs)

Table SM5. Night-light Gini Coefficients by World Region

Table SM6. Agriculture, Grains and Early Statehood?

Table SM7. Baseline OLS Models (Cash Crops)

Table SM8. Baseline OLS Models (Minerals)

Table SM9. Cash Crop Suitability and Mediators

Table SM10. Colonial Cash Crop Production and Mediators

Figure SM1. Validation of Colonial Resource Data

Figure SM2. Covariate Balance Propensity Score Weights

Figure SM3 Distribution of Counterfactual Cells for RI Analysis

Figure SM4. Covariate Balance: Counterfactual vs. Actual Cash Crop Cells

Figure SM5. Covariate Balance: Counterfactual vs. Actual Mining Cells

Figure SM6. RI Treatment Effects with Continuous Outcomes

Figure SM7. Robustness of S2SLS-IV results

Figure SM8. Placebo Test: Cash Crop Suitability and Pre-Colonial Development

Figure SM9. Dissagregating Cash Crops

SM2
Figure SM10. Country-by-Country Analysis: Cash Crops

Figure SM11. Country-by-Country Analysis: Minerals

Figure SM12. Heterogeneity across Imperial Powers

Figure SM13. Heterogeneity Analysis

Figure SM14. Robustness: Continuous Measures & Intensive Margins

Figure SM15. Robustness: Continuous Measures & S2SLS Models

Figure SM16. Robustness: Conley Errors for Spatial Correlation

Figure SM17. Robustness: Changing the Spatial Unit of Analysis (Binary Predictors)

Figure SM18. Robustness: Changing the Spatial Unit of Analysis (Continuous Pred.)

Figure SM19. Cash Crops, Precolonial Statehood, and the Slave Trades.

Figure SM20. Causal Mediation Analysis (Actual Production 1957).

Figure SM21. Robustness: Suitability vs. Other Geographic Fundamentals (Continu-

ous Outcomes).

Figure SM22. Spillover Analysis (Cash Crops)

Figure SM23. Spillover Analysis (Minerals)

Appendix SM2. Sources of Historical Primary Commodity Production by Country

SM3
 Table SM1: Description and Sources of Variables

Variable Description Source

Quality Road 1998 Any paved or improved roads within the grid cell Michelin (1998) from JedwabTrans-
portation2016
Night Lights 2015 Any detectable night-time lights within the grid cell roman2018nasa
City 2015 Localities with 10,000 inhabitants or more in 2015 Africapolis.org
DHS Household Wealth Composite asset-based measure of a household’s liv- DHS (1990-2017)
ing standard from 93 surveys across 26 countries
Road Density 1998 (log) Natural log of the length of paved or improved roads Michelin (1998) from JedwabTrans-
per square kilometer portation2016
Night Lights 2015 (log) cell/polygon mean of light intensity roman2018nasa
Urban Population Density sum of population residing in cities with more than Africapolis.org
2015 (log) 10’000 inhabitants divided by land area in sqkm
Quality Road 1960 Paved and improved roads recorded by Michelin road Michelin1965NorthWest,Michelin1967CentralSouth,Mich
maps in the 1960s
Railway 1960 Railroads built by 1960 JedwabTransportation2016
Power Plant 1972 Location of hydropower and thermal plants UNECA1972energy
Crop Production Value 2000 grid-level estimate of the market value of all agricul- FAO GAEZ
(log) tural production
Colonial Cash Crops (Y/N) Production points greater than $289,270 (1957) hance1961source
across nine different cash crops
Colonial Minerals (Y/N) Mines producing greater than $289,270 (1957) across hance1961source
twenty different minerals
Colonial Cash Crop Value (log) Log of the total value of cash crops produced divided hance1961source
by land area in sqkm
Colonial Mineral Value (log) Log of the total value of cash crops produced divided hance1961source
by land area in sqkm
Cash Crop Suitability Mean agro-climatic suitability across nine cash crops FAO GAEZ
(cocoa, coffee, cotton, groundnuts, oil palm, tobacco,
tea, sugarcane, and bananas) in a given area
Caloric Suitability Potentially attainable caloric yield (post-1500) galor2016agricultural
across all crops suitable for cultivation in a given
area
TseTse Suitability Temperature-based tsetse fly suitability index cal- AlsanTsetse2014; harris2014updated
culated using the formula in Alsan (2015) and his-
torical temperature data (1901-2017) from the CRU
dataset
Malaria Suitability Temperature-based suitability index for malaria weiss2014air
transmission
Terrain Ruggedness Average terrain ruggedness within quarter-degree shaver2016terrain
cell or polygon
Elevation Mean elevation in m above sea level within cell or srtm.csi.cgiar.org
polygon
Dist. Navigable River (log) Minimum geodesic distance to a navigable river JedwabTransportation2016
Dist. Coast (log) Minimum geodesic distance to the coast ngdc.noaa.gov/mgg/shorelines/
Dist. European Explorer Minimum geodesic distance to European explorer Century Atlas Africa (1911) via Nun-
Route (log) routes nMistrust2011
Dist. 1900 City (log) Minimum geodesic distance to localities with more JedwabTransportation2016
than 10,000 inhabitants in 1900
Dist. Colonial Capital (log) Minimum geodesic distance to the colonial capital in Various
1900
Dist. Precolonial Trade Route Minimum geodesic distance to precolonial trade Ajayi1985
(log) routes

SM4
 Table SM2: Description and Sources of Variables (continued from Table SM1)

Variable Description Source

Slaves per Area (int.) Number of slaves taken from a Murdock ethnic group NunnMistrust2011
per sqkm (second tercile)
Slaves per Area (high) Number of slaves taken from a Murdock ethnic group NunnMistrust2011
per sqkm (third tercile)
Slaves per Area (missing) Indicates if slave trade data is missing NunnMistrust2011
Precolonial Agriculture (int.) Between 46 and 55% reliance on agriculture murdock1967ethnographic
Precolonial Agriculture (high) More than 55% reliance on agriculture murdock1967ethnographic
Precolonial Agriculture (miss- Indicates if precolonial agriculture variable is miss- murdock1967ethnographic
ing) ing
Precolonial Chiefdoms (Y/N) Indicates if ethnic group had some political central- murdock1967ethnographic
ization beyond the village level
Precolonial Statehood (Y/N) Indicates if ethnic group had state-like political cen- murdock1967ethnographic
tralization
Precolonial Centralization Indicates if political centralization variable is miss- murdock1967ethnographic
(missing) ing
Precolonial Town Size Average town/settlement size of Murdock ethnic murdock1967ethnographic
groups
Social Class Pre-colonial presence of class stratification murdock1967ethnographic
Headman Elected Democratic selection rule for local elites murdock1967ethnographic
Slavery Slavery as a common practice murdock1967ethnographic

SM5
 Table SM3: Summary Statistics (0.25 Degree Grid Cells)

Statistic N Mean St. Dev. Min Pctl(25) Pctl(75) Max

Quality Road 1998 (Y/N) 28,166 0.334 0.472 0 0 1 1
Night Lights 2015 (Y/N) 28,166 0.267 0.442 0 0 1 1
City 2015 (Y/N) 28,166 0.128 0.335 0 0 0 1
DHS Household Wealth 6,775 −0.391 0.549 −2.220 −0.721 −0.146 3.068
Road Density 1998 (log) 28,166 −2.676 2.767 −4.605 −4.605 0.940 7.461
Mean Night Lights 2015 (log) 28,166 −3.537 2.103 −4.605 −4.605 −3.698 7.742
Urban Population Density 2015 (log) 28,166 −4.096 1.386 −4.605 −4.605 −4.605 5.556
Night Lights per cap. 2015 (log) 28,166 −3.490 2.149 −4.605 −4.605 −3.461 12.398
Quality Road 1960 (Y/N) 28,166 0.107 0.308 0 0 0 1
Railway 1960 (Y/N) 28,166 0.018 0.134 0 0 0 1
Power Plant 1972 (Y/N) 28,166 0.027 0.163 0 0 0 1
Crop Production Value 2000 (log) 18,636 6.841 2.034 0.112 5.663 8.317 19.851
Colonial Cash Crop Dummy 28,166 0.081 0.273 0 0 0 1
Colonial Mineral Dummy 28,166 0.009 0.093 0 0 0 1
Colonial Cash Crop Value (log) 28,166 0.175 0.623 0 0 0 7
Colonial Mineral Value (log) 28,166 0.022 0.263 0 0 0 9
Cash Crop Suitability 28,166 0.000 1.000 −1.101 −0.871 0.588 4.774
Caloric Suitability 28,166 −0.000 1.000 −1.587 −0.880 0.868 1.767
TseTse Suitability 28,166 9.763 2.694 0.000 7.794 12.183 14.511
Malaria Suitability 28,166 0.562 0.202 0.000 0.395 0.732 0.958
Terrain Ruggedness 28,166 37.836 55.777 0.000 6.505 45.458 765.844
Elevation 28,166 608.049 440.281 −113.633 316.543 817.117 3,654.161
Dist. Navigable River (log) 28,166 5.016 1.214 0.026 4.337 5.913 7.186
Dist. Coast (log) 28,166 6.208 1.192 0.020 5.807 7.028 7.502
Dist. European Explorer Route (log) 28,166 4.608 1.240 0.018 3.917 5.516 6.906
Dist. 1900 City (log) 28,166 5.973 0.790 1.245 5.532 6.555 7.370
Dist. Colonial Capital (log) 28,166 6.222 0.733 0.327 5.825 6.776 7.562
Dist. Precolonial Trade Route (log) 28,166 4.803 1.290 0.000 4.086 5.752 7.284
Slaves per Area (int.) 28,166 0.204 0.403 0 0 0 1
Slaves per Area (high) 28,166 0.229 0.421 0 0 0 1
Slaves per Area (missing) 28,166 0.019 0.135 0 0 0 1
Precolonial Agriculture (int.) 28,166 0.462 0.499 0 0 1 1
Precolonial Agriculture(high) 28,166 0.265 0.441 0 0 1 1
Precolonial Agriculture (missing) 28,166 0.040 0.196 0 0 0 1
Precolonial Chiefdoms (Y/N) 28,166 0.619 0.486 0 0 1 1
Precolonial Statehood (Y/N) 28,166 0.126 0.332 0 0 0 1
Precolonial Centralization (missing) 28,166 0.095 0.293 0 0 0 1
Area on Land (sqkm) 28,166 732.293 96.842 0.001 736.064 764.317 769.316

SM6
 Table SM4: Summary Statistics (Ethnic Homeland-Country Pairs)

Statistic N Mean St. Dev. Min Pctl(25) Pctl(75) Max

Quality Road 1998 (Y/N) 1,074 0.827 0.379 0 1 1 1
Night Lights 2015 (Y/N) 1,074 0.778 0.416 0 1 1 1
City 2015 (Y/N) 1,074 0.709 0.455 0 0 1 1
DHS Household Wealth 711 −0.310 0.470 −1.772 −0.581 −0.086 1.824
Road Density 1998 (log) 1,074 −0.297 2.145 −4.605 −0.572 1.138 2.865
Mean Night Lights 2015 (log) 1,074 −1.603 2.496 −4.605 −4.379 0.284 6.564
Urban Population Density 2015 (log) 1,074 −2.656 1.714 −4.605 −4.605 −1.396 3.290
Mean Night Lights 2015 (log) 1,074 −1.422 2.374 −4.605 −4.007 0.377 5.755
Quality Road 1960 (Y/N) 1,074 0.443 0.497 0 0 1 1
Railway 1960 (Y/N) 1,074 0.098 0.297 0 0 0 1
Power Plant 1972 (Y/N) 1,074 0.303 0.460 0 0 1 1
Colonial Cash Crop Dummy 1,074 0.363 0.481 0 0 1 1
Colonial Mineral Dummy 1,074 0.067 0.250 0 0 0 1
Colonial Cash Crop Value (log) 1,074 0.389 0.744 0 0 0.4 5
Colonial Mineral Value (log) 1,074 0.067 0.362 0 0 0 4
Cash Crop Suitability 1,073 0.003 1.001 −1.588 −0.704 0.572 3.265
Caloric Suitability 1,073 −0.002 1.004 −2.957 −0.458 0.785 1.312
TseTse Suitability 1,074 10.245 2.600 0.000 8.588 12.387 14.464
Malaria Suitability 1,074 0.626 0.177 0.039 0.488 0.774 0.944
Terrain Ruggedness 1,074 47.581 52.247 0.603 13.940 60.510 413.404
Elevation 1,074 608.444 448.116 9.257 292.595 879.243 2,364.656
Dist. Navigable River (log) 1,074 4.705 1.138 0.196 4.107 5.460 7.128
Dist. Coast (log) 1,074 6.043 1.221 0.052 5.545 6.896 7.490
Dist. European Explorer Route (log) 1,074 4.634 1.203 0.119 4.006 5.466 6.865
Dist. 1900 City (log) 1,074 5.692 0.825 1.813 5.246 6.301 7.290
Dist. Colonial Capital (log) 1,074 6.078 0.808 2.668 5.628 6.703 7.492
Dist. Precolonial Trade Route (log) 1,073 4.472 1.274 0.000 3.771 5.398 7.237
Precolonial Chiefdoms (Y/N) 1,074 0.575 0.495 0 0 1 1
Precolonial Statehood (Y/N) 1,074 0.123 0.328 0 0 0 1
Slaves per Area (log) 1,070 0.186 0.555 0.000 0.000 0.032 3.774
Precolonial Agriculture 1,024 6.124 1.812 1.000 5.000 8.000 10.000
Precolonial Twon Size 402 4.206 2.026 1.000 3.000 7.000 8.000
Precolonial Class Stratification 950 0.575 0.495 0.000 0.000 1.000 1.000
Political Centralization 975 1.317 0.979 0.000 1.000 2.000 4.000
Headman Elected 1,074 0.068 0.252 0 0 0 1
Slavery (Murdock) 1,004 0.893 0.309 0.000 1.000 1.000 1.000
Area (sqkm) 1,074 19,709.700 37,242.160 108.765 2,567.194 20,928.760 482,598.900

SM7
Table SM5: Night-light Gini Coefficients across First and Second Administrative Units by
World Regions

ADM 1 ADM 2
2000 2010 2000 2010
Africa 0.693 0.673 0.867 0.847
Asia 0.604 0.572 0.763 0.747
South America 0.596 0.578 0.747 0.722
Oceania 0.468 0.482 0.744 0.739
North America 0.455 0.452 0.645 0.602
Europe 0.401 0.402 0.605 0.601

Table SM5 draws from AlesinaInequality2016. It reports Gini coefficients of nighttime

lights at first-level and second-level administrative divisions for all countries in the world.
Electrification inequality is highest in Africa compared to all world regions.

SM8
 Table SM6: Agriculture, Grains and Early Statehood?

Dependent Variable

Precolonial Centralization Precolonial State (Y/N)

Caloric Suitability −0.175∗∗∗ −0.025∗
(0.037) (0.013)

Cereal Advantage −0.182∗∗∗ −0.027∗∗

(0.037) (0.013)

Observations 986 978 986 978

Adjusted R2 0.180 0.181 0.115 0.112
Notes: OLS regression models with Murdock ethnic group polygons as the unit of analysis. Standard errors are clustered
at the country level. Significance codes: ∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01

Table SM6 reports OLS models regressing pre-colonial centralization as measured by

Murdock (1959, 1967) on two agricultural variables prominently discussed in the recent quan-
titative literature on early state formation. Some studies report an intricate link between
agricultural productivity and the emergence of early states DiamondGuns1997 whereas oth-
ers argue that only easy-to-tax cereal grains incentivize state formation mayshar2018emergence,
scott2017against. The first two columns in Table SM6 use Murdock’s ordinal variable as
coded, whereas the last two columns use a binary indicator of whether a particular ethnic
group had state-like political organization. Drawing from galor2016agricultural, the caloric
suitability measure is an estimate of a given territory’s potential caloric yield and its ability
to feed dense populations in premodern times. Cereal advantage measures the difference in
caloric yield between cereal grains on the one hand and roots and tubers on the other as de-
scribed by mayshar2018emergence. The correlation between these two agricultural variables
and Murdock’s early statehood indicators is negative suggesting that other mechanisms may
be more important in explaining the emergence of precolonial states in the African context.

SM9
 Table SM7: Baseline OLS Models (Cash Crops)

Baseline Analysis (OLS)

Quality Road (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) HH Wealth
Caloric Suitability 0.070∗∗∗ 0.049∗∗∗ 0.031∗∗∗ −0.048∗
(0.012) (0.009) (0.007) (0.026)
TseTse Suitability 0.001 −0.012∗∗∗ −0.002 0.0001
(0.004) (0.004) (0.002) (0.007)
Malaria Suitability −0.096 −0.128∗∗ −0.035 0.116
(0.071) (0.060) (0.039) (0.143)
Ruggedness −0.0003∗∗ −0.0001 −0.0002∗∗ −0.001∗∗∗
(0.0001) (0.0001) (0.0001) (0.0002)
Elevation 0.0001∗∗∗ 0.0002∗∗∗ 0.0001∗∗∗ 0.0004∗∗∗
(0.00003) (0.00002) (0.00002) (0.0001)
Dist. Navigable River (log) −0.031∗∗∗ −0.033∗∗∗ −0.024∗∗∗ −0.059∗∗∗
(0.007) (0.006) (0.004) (0.011)
Dist. Coast (log) −0.017∗ −0.086∗∗∗ −0.035∗∗∗ −0.127∗∗∗
(0.009) (0.008) (0.006) (0.016)
Dist. Trade Route 1900 (log) −0.020∗∗∗ −0.012∗∗∗ 0.0004 0.009
(0.006) (0.004) (0.003) (0.011)
Dist. Explorer Route (log) −0.066∗∗∗ −0.061∗∗∗ −0.039∗∗∗ −0.100∗∗∗
(0.011) (0.009) (0.007) (0.017)
Dist. City 1900 (log) −0.061∗∗∗ −0.048∗∗∗ −0.024∗∗∗ −0.132∗∗∗
(0.011) (0.011) (0.007) (0.019)
Dist. Colonial Capital (log) −0.009∗ −0.005 −0.003 −0.020∗∗
(0.006) (0.004) (0.003) (0.008)
Slavery (Med.) −0.016 −0.011 −0.016∗∗ −0.003
(0.019) (0.015) (0.008) (0.033)
Slavery (High) 0.008 0.006 −0.013 0.012
(0.019) (0.019) (0.013) (0.029)
Precolonial Agriculture (Med.) 0.047∗∗ 0.056∗∗∗ 0.018∗ 0.016
(0.020) (0.014) (0.009) (0.040)
Precolonial Agriculture (High) −0.004 0.044∗∗∗ 0.026∗∗ 0.013
(0.022) (0.015) (0.012) (0.044)
Precolonial Chiefdom −0.006 −0.013 0.006 0.002
(0.022) (0.013) (0.011) (0.044)
Precolonial State −0.010 −0.006 0.011 −0.054
(0.033) (0.023) (0.017) (0.051)
Precolonial Statehood Missing 0.062∗∗ −0.011 0.029∗ −0.069
(0.031) (0.021) (0.017) (0.052)
Colonial Cash Crops (Y/N) 0.164∗∗∗ 0.196∗∗∗ 0.187∗∗∗ 0.144∗∗∗
(0.017) (0.015) (0.015) (0.027)
Observations 28,166 28,166 28,166 6,775
Adjusted R2 0.193 0.366 0.199 0.236
Colony FE Yes Yes Yes Yes

Notes: OLS regression models. Column 4 restricts the sample to all grid cells with information on asset-based household
wealth from the DHS survey. All models include colony fixed effects and standard errors (in parentheses) are clustered
at the level of ethnic group polygons from George Peter Murdock’s “Tribal Map of Africa.” Significance codes: ∗ p<0.1;
∗∗ p<0.05; ∗∗∗ p<0.01

SM10
 Table SM8: Baseline OLS Models (Minerals)

Baseline Analysis (OLS)

Quality Road (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) HH Wealth
Caloric Suitability 0.070∗∗∗ 0.049∗∗∗ 0.031∗∗∗ −0.051∗
(0.012) (0.009) (0.007) (0.027)
TseTse Suitability 0.001 −0.012∗∗∗ −0.003 0.001
(0.004) (0.004) (0.002) (0.007)
Malaria Suitability −0.092 −0.124∗ −0.033 0.070
(0.072) (0.064) (0.043) (0.143)
Ruggedness −0.0003∗∗ −0.0001 −0.0002∗∗∗ −0.001∗∗∗
(0.0001) (0.0001) (0.0001) (0.0002)
Elevation 0.0001∗∗∗ 0.0002∗∗∗ 0.0001∗∗∗ 0.0004∗∗∗
(0.00003) (0.00002) (0.00002) (0.0001)
Dist. Navigable River (log) −0.031∗∗∗ −0.034∗∗∗ −0.025∗∗∗ −0.058∗∗∗
(0.008) (0.006) (0.004) (0.011)
Dist. Coast (log) −0.021∗∗ −0.091∗∗∗ −0.039∗∗∗ −0.133∗∗∗
(0.009) (0.009) (0.007) (0.015)
Dist. Trade Route 1900 (log) −0.020∗∗∗ −0.012∗∗∗ 0.0001 0.008
(0.006) (0.004) (0.003) (0.010)
Dist. Explorer Route (log) −0.070∗∗∗ −0.066∗∗∗ −0.044∗∗∗ −0.105∗∗∗
(0.011) (0.010) (0.009) (0.016)
Dist. City 1900 (log) −0.071∗∗∗ −0.059∗∗∗ −0.034∗∗∗ −0.143∗∗∗
(0.011) (0.011) (0.008) (0.018)
Dist. Colonial Capital (log) −0.009∗ −0.006 −0.004 −0.022∗∗∗
(0.006) (0.005) (0.003) (0.008)
Slavery (Med.) −0.018 −0.014 −0.018∗∗ 0.002
(0.019) (0.015) (0.008) (0.033)
Slavery (High) 0.011 0.010 −0.009 0.023
(0.019) (0.020) (0.014) (0.029)
Precolonial Agriculture (Med.) 0.048∗∗ 0.057∗∗∗ 0.019∗ 0.015
(0.020) (0.015) (0.010) (0.042)
Precolonial Agriculture (High) 0.002 0.051∗∗∗ 0.033∗∗ 0.017
(0.022) (0.016) (0.014) (0.047)
Precolonial Chiefdom −0.005 −0.012 0.007 −0.003
(0.022) (0.014) (0.012) (0.046)
Precolonial State −0.012 −0.009 0.008 −0.065
(0.033) (0.024) (0.018) (0.052)
Precolonial Statehood Missing 0.068∗∗ −0.004 0.036∗∗ −0.067
(0.032) (0.023) (0.018) (0.053)
Colonial Mining (Y/N) 0.154∗∗∗ 0.186∗∗∗ 0.223∗∗∗ 0.393∗∗∗
(0.035) (0.044) (0.039) (0.067)
Observations 28,166 28,166 28,166 6,775
Adjusted R2 0.187 0.355 0.183 0.236
Colony FE Yes Yes Yes Yes

Notes: OLS regression models. Column 4 restricts the sample to all grid cells with information on asset-based household
wealth from the DHS survey. All models include colony fixed effects and standard errors (in parentheses) are clustered
at the level of ethnic group polygons from George Peter Murdock’s “Tribal Map of Africa.” Significance codes: ∗ p<0.1;
∗∗ p<0.05; ∗∗∗ p<0.01

SM11
 Table SM9: Cash Crop Suitability, Early Infrastructure, Continued Agricultural Production

Dependent Variable

Crop Value 2000 (log) Quality Road ca. 1960 (Y/N) Rail 1960 (Y/N) Power Plant 1972 (Y/N)
Caloric Suitability 1.862∗∗∗ 0.031∗∗∗ 0.0001 0.004∗
(0.135) (0.008) (0.003) (0.002)

TseTse Suitability −0.229∗∗∗ −0.010∗∗∗ −0.001 −0.003∗∗∗

(0.038) (0.003) (0.001) (0.001)

Malaria Suitability 0.615 −0.046 −0.040∗ −0.020

(0.654) (0.044) (0.024) (0.014)

Ruggedness 0.002∗∗ −0.0002∗∗ −0.0001∗ 0.00002

(0.001) (0.0001) (0.00003) (0.00003)

Elevation 0.0005∗ 0.0001∗∗∗ 0.00000 0.00002∗∗∗

(0.0003) (0.00002) (0.00001) (0.00001)

Dist. Navigable River −0.419∗∗∗ −0.014∗∗∗ −0.004 −0.006∗∗∗

(0.069) (0.004) (0.003) (0.001)

Dist. Coast −0.325∗∗∗ −0.059∗∗∗ −0.012∗∗∗ −0.019∗∗∗

(0.067) (0.006) (0.003) (0.002)

Cash Crop Suitability 0.501∗∗∗ 0.049∗∗∗ 0.015∗∗∗ 0.010∗∗∗

(0.088) (0.007) (0.003) (0.003)

Observations 28,166 28,166 28,166 28,166

Adjusted R2 0.620 0.218 0.033 0.042
Colony FE Yes Yes Yes Yes
Notes: OLS regression models. All models include colony fixed effects and standard errors are clustered at the level
of ethnic group polygons from George Peter Murdock’s “Tribal Map of Africa.” Significance codes: ∗ p<0.1; ∗∗ p<0.05;
∗∗∗ p<0.01

SM12
 Table SM10: Colonial Cash Crop Production, Early Infrastructure, Continued Agricultural
Production

Dependent Variable

Crop Value 2000 (log) Quality Road ca. 1960 (Y/N) Rail 1960 (Y/N) Power Plant 1972 (Y/N)
Caloric Suitability 1.934∗∗∗ 0.038∗∗∗ 0.002 0.006∗∗∗
(0.138) (0.007) (0.003) (0.002)

TseTse Suitability −0.201∗∗∗ −0.006∗∗ 0.0002 −0.002∗∗

(0.038) (0.003) (0.001) (0.001)

Malaria Suitability 1.350∗∗ 0.015 −0.023 −0.009

(0.650) (0.041) (0.022) (0.014)

Ruggedness 0.002∗ −0.0002∗∗ −0.0001∗∗ 0.00001

(0.001) (0.0001) (0.00003) (0.00003)

Elevation 0.001∗∗ 0.0001∗∗∗ 0.00000 0.00002∗∗∗

(0.0003) (0.00002) (0.00001) (0.00001)

Dist. Navigable River −0.400∗∗∗ −0.011∗∗∗ −0.003 −0.005∗∗∗

(0.069) (0.004) (0.003) (0.001)

Dist. Coast −0.269∗∗∗ −0.050∗∗∗ −0.009∗∗∗ −0.016∗∗∗

(0.068) (0.006) (0.003) (0.002)

Colonial Cash Crops (Y/N) 1.168∗∗∗ 0.204∗∗∗ 0.071∗∗∗ 0.056∗∗∗

(0.129) (0.018) (0.011) (0.007)

Observations 28,166 28,166 28,166 28,166

Adjusted R2 0.620 0.218 0.033 0.042
Colony FE Yes Yes Yes Yes
Notes: OLS regression models. All models include colony fixed effects and standard errors are clustered at the level
of ethnic group polygons from George Peter Murdock’s “Tribal Map of Africa.” Significance codes: ∗ p<0.1; ∗∗ p<0.05;
∗∗∗ p<0.01

SM13
Figure SM1: Correlations between Hance’s primary commodity data (standardized in 1957
US$) with the primary commodity data data we collected from colonial reports, maps and
other records (standardized in 1960 US$; see additional appendix below). Data have been
aggregated to second-level (in some cases first-level) administrative districts (regions) at
independence (c. 1960).

SM14
 Covariate Balance Covariate Balance (DHS Sample)
Treatment Variable: Colonial Cash Crop Dummy Treatment Variable: Colonial Cash Crop Dummy

Propensity Score ● ● Propensity Score ● ●

Caloric Suitability ● ● Cash Crop Suitability ● ●

Cash Crop Suitability ● ● Dist. Explorer Route ● ●

Malaria Suitability ● ● Precolonial Agriculture (High) ● ●

Slavery (High) ● ● Slavery (High) ● ●

Ruggedness ● ● Dist. Navigable River ● ●

TseTse Suitability ● ● Malaria Suitability ● ●

Precolonial Agriculture (High) ● ● Precolonial State ●●

Dist. Explorer Route ● ● TseTse Suitability ● ●

Sample Sample
Elevation ●● Ruggedness ●●
● Unadjusted ● Unadjusted
Precolonial Agriculture (Med.) ●● Precolonial Chiefdom ●
●
● Adjusted ● Adjusted
Precolonial State ●● Precolonial Statehood Missing ●●

Precolonial Statehood Missing ●● Elevation ● ●

Precolonial Chiefdom ●● Slavery (Med.) ● ●

Slavery (Med.) ● ● Precolonial Agriculture (Med.) ● ●

Dist. Trade Route 1900 ● ● Caloric Suitability ● ●

Dist. Coast ● ● Dist. Trade Route 1900 ● ●

Dist. Colonial Capital ● ● Dist. Coast ● ●

Dist. City 1900 ● ● Dist. Colonial Capital ● ●

Dist. Navigable River ● ● Dist. City 1900 ● ●

−0.5 0.0 0.5 −0.50 −0.25 0.00 0.25 0.50

Standardized Mean Differences Standardized Mean Differences

Covariate Balance Covariate Balance (DHS Sample)

Treatment Variable: Colonial Mining Dummy Treatment Variable: Colonial Mining Dummy

Propensity Score ● ● Propensity Score ● ●

Caloric Suitability ● ● Cash Crop Suitability ● ●

Cash Crop Suitability ● ● Dist. Trade Route 1900 ● ●

Elevation ● ● Dist. Coast ● ●

Malaria Suitability ● ● Elevation ● ●

Ruggedness ● ● Precolonial State ● ●

TseTse Suitability ● ● Ruggedness ● ●

Precolonial Agriculture (Med.) ● ● Precolonial Agriculture (Med.) ● ●

Precolonial State ● ● Precolonial Agriculture (High) ● ●

Sample Sample
Slavery (Med.) ● ● Slavery (Med.) ● ●
● Unadjusted ● Unadjusted
Dist. Coast ●● Malaria Suitability ●
●
● Adjusted ● Adjusted
Slavery (High) ●● Precolonial Chiefdom ●●

Dist. Trade Route 1900 ●● Precolonial Statehood Missing ● ●

Precolonial Agriculture (High) ●● Slavery (High) ● ●

Dist. Colonial Capital ●● TseTse Suitability ● ●

Precolonial Statehood Missing ●● Dist. City 1900 ● ●

Precolonial Chiefdom ● ● Dist. Navigable River ● ●

Dist. Explorer Route ● ● Dist. Colonial Capital ● ●

Dist. City 1900 ● ● Dist. Explorer Route ● ●

Dist. Navigable River ● ● Caloric Suitability ● ●

−1.0 −0.5 0.0 0.5 1.0 −0.50 −0.25 0.00 0.25 0.50 0.75
Standardized Mean Differences Standardized Mean Differences

Figure SM2: Statistical balance of colonial cash crop and mineral production cells before
and after using propensity score reweighting hirano2003efficient.

SM15
Figure SM3: Figure S3a maps out counterfactual cash crop (upper panels) and mining
(lower panel) cells sampled from 1000 draws. See Methods section in the main text for
description of counterfactual sampling procedure. The right-hand panels also show actual
colonial production sites from hance1961source.

SM16
 Covariate Balance: Actual vs. Counterfactual Cash Crop Distribution
1000 simulated cash crop counterfactuals based on clustered point process estimates
Caloric Suitability TseTse Suitability Malaria Suitability Ruggedness
0.08
6 1.5 20
15 0.06
4 1.0
10 0.04
2 0.5 0.02
5
0 0.0 0 0.00
0.2 0.3 0.4 0.5 0.6 −0.5 0.0 0.5 1.0 1.5 0.00 0.05 0.10 0.15 −10 0 10 20 30

Elevation Dist. Navigable River Dist. Coast Dist. Trade Route 1900
0.03 0.012 0.020
0.006
0.015
0.02 0.008
0.004 0.010
0.002 0.01 0.004
0.005
0.000 0.00 0.000 0.000
−100 0 100 200 −125 −100 −75 −50 −300 −200 −100 0 −100 −50 0 50

Dist. Explorer Route Dist. City 1900 Dist. Colonial Capital Cash Crop Suitability
0.015 5
0.020 4
0.010
Density

0.015 0.010 3
0.010 0.005 2
0.005
0.005 1
0.000 0.000 0.000 0
−50 0 50 −200 −150 −100 −50 −250 −200 −150 −100 −50 0 0.3 0.5 0.7

Slavery (Med.) Slavery (High) Precolonial Agriculture (Med.) Precolonial Agriculture (High)
20 15
9
15 10
10
10 6
5 5
5 3
0 0 0 0
−0.10 −0.05 0.00 0.05 0.0 0.1 0.2 0.3 0.0 0.1 0.2 −0.05 0.00 0.05 0.10 0.15 0.20

Precolonial Chiefdom Precolonial State Precolonial Statehood Missing

10 15
10
10
5 5 5
0 0 0
−0.2 −0.1 0.0 0.1 −0.05 0.00 0.05 0.10 −0.05 0.00 0.05 0.10
Distribution of 1000 Counterfactual Coefficients
2.5 and 97.5 percentiles of the counterfactual coefficient distribution in blue; actual treatment coefficient in red.

Figure SM4: Statistical balance on key covariates among actual versus counterfactual cash
crop cells.

SM17
 Covariate Balance: Actual vs. Counterfactual Mineral Distribution
1000 randomly sampled mineral counterfactuals based on clustered point process estimates
Caloric Suitability TseTse Suitability Malaria Suitability Ruggedness
3 0.4 8
0.3 6 0.02
2
0.2 4
1 0.01
0.1 2
0 0.0 0 0.00
0.0 0.5 −4 −2 0 2 −0.2 −0.1 0.0 0.1 0.2 0 50 100 150

Elevation Dist. Navigable River Dist. Coast Dist. Trade Route 1900
0.0100
0.003 0.003
0.009 0.0075
0.002 0.006 0.002
0.0050
0.001 0.003 0.001 0.0025
0.000 0.000 0.000 0.0000
−300 0 300 600 900 −100 0 100 −500 −250 0 250 500 −100 0 100 200

Dist. Explorer Route Dist. City 1900 Dist. Colonial Capital Cash Crop Suitability

0.004 1.5
0.009
Density

0.004 0.003
0.006 1.0
0.002
0.002 0.5
0.003 0.001
0.000 0.000 0.000 0.0
−100 0 100 200 −300 −200 −100 0 100 200 300 −200 0 200 400 −0.5 0.0 0.5 1.0 1.5

Slavery (Med.) Slavery (High) Precolonial Agriculture (Med.) Precolonial Agriculture (High)
6 6 4 5
3 4
4 4 3
2
2 2 2
1 1
0 0 0 0
−0.2 0.0 0.2 0.4 −0.2 0.0 0.2 0.4 −0.2 0.0 0.2 0.4 −0.2 0.0 0.2

Precolonial Chiefdom Precolonial State Precolonial Statehood Missing

25
4
3 20
3 15
2
2 10
1 1 5
0 0 0
−0.4 −0.2 0.0 0.2 0.0 0.2 0.4 −0.10 −0.05 0.00 0.05 0.10 0.15
Distribution of 1000 Counterfactual Coefficients
2.5 and 97.5 percentiles of the counterfactual coefficient distribution in blue; actual treatment coefficient in red.

Figure SM5: Statistical balance on key covariates among actual versus counterfactual mining
site cells.

SM18
 Treatment Effects: Actual vs. Counterfactual Cash Crop Distribution
1000 simulated cash crop counterfactuals based on clustered point process estimates
Road Density 1998 (log) Night Lights 2015 (log) Urban Pop. Dens. 2015 (log) Night Lights 2015 p.c. (log)
3 3
3
2.0

2 1.5 2
Density

1.0
1 1 1
0.5

0 0.0 0 0
0.5 1.0 1.5 2.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 0.5 1.0 1.5
Distribution of 1000 Counterfactual Coefficients
2.5 and 97.5 percentiles as well as mean of the counterfactual coefficient distribution in blue; actual treatment coefficient in red.

Treatment Effects: Actual vs. Counterfactual Mineral Distribution

1000 randomly sampled mineral counterfactuals based on clustered point process estimates
Road Density 1998 (log) Night Lights 2015 (log) Urban Pop. Dens. 2015 (log) Night Lights 2015 p.c. (log)
2.0 1.25

0.75 1.00
1.0 1.5
Density

0.75
0.50 1.0
0.5 0.50
0.25 0.5
0.25

0.0 0.00 0.0 0.00

0 1 2 −1 0 1 2 −0.5 0.0 0.5 1.0 1.5 −1 0 1 2
Distribution of 1000 Counterfactual Coefficients
2.5 and 97.5 percentiles as well as mean of the counterfactual coefficient distribution in blue; actual treatment coefficient in red.

Figure SM6: Spatial randomization results replacing binary with continuous development
outcomes.

SM19
 Robustness of S2SLS Models: Definition of Spatial Neighbors
K−Nearest Neighbors (Inverse Distance Weights) vs. Simple Distance Cutoffs (No Weights)
Quality Road 1998 (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) DHS Household Wealth

OLS ● ● ● ●

KNN 8 ● ● ● ●

KNN 24 ● ● ● ●

KNN 50 ● ● ● ●
Model

KNN 100 ● ● ● ●

DIST 1 ● ● ● ●

DIST 2 ● ● ● ●

DIST 3 ● ● ● ●

DIST 4 ● ● ● ●

0.0 0.1 0.2 0.3 0.0 0.1 0.2 0.3 0.4 0.0 0.1 0.2 0.3 0.0 0.1 0.2 0.3 0.4 0.5
Coefficient of Instrumented Cash Crop Dummy and 95% Confidence Intervals

Instrument Strength in S2SLS Models: First Stage F

K−Nearest Neighbors (Inverse Distance Weights) & Simple Distance Cutoffs (No Weights)
Quality Road 1998 (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) DHS Household Wealth
11 40.7 40.7
KNN 8 19.2
40.7 29.1 8.8 38.3

10.8 34.2 45.6

KNN 24
17.5 34.2 17.9 34.2 11

20.4 12.5 7.5 69.4

Distance Matrix

KNN 50 25.9 26.3

26.3 26.3

21.8 29.4 27.9 173.1

KNN 100 19.3 21.8
21.8 6.3

DIST 1 42.7 42.7 42.7

21 22.8 8.9 20.6 55.8

36.9 19.6 13.8 125

DIST 2
27.8 36.9 10.8 36.9

36.9 20.1 27.8 9.4

DIST 3 26 27.8
27.8 133

21.7 21.7 32.4 9.2

DIST 4 186.8
32.6 21.7 34.2

0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 50 100 150
First Stage F Statistic

Endogenous Variable Cash Crops (Y/N) WxY

Figure SM7: Robustness of S2SLS-IV results

Figure SM7 reports results from spatial 2SLS-IV regression models but employs alter-
native definitions of what constitutes the spatial neighborhood of a cell. The first four IV
models use spatial lags on inverse-distance weighted nearest-neighbor matrices, whereas the
last four models use binary distance cutoffs at 0.35, 0.71, 1.06, and 1.41 lon/lat degrees. The
upper panel reports coefficients and confidence intervals for the instrumented colonial cash
dummy, whereas the lower panel reports F statistics for the two first stages in each model ((i)
cash crop dummy instrumented with suitability; (ii) spatial lag of the respective dependent
variable instrumented with first and second-order spatial lags of the baseline covariates).

SM20
 Placebo: Cash Crop Suitability & Precolonial Development
Conditional on colony fixed effects and geographic controls

Dist. Trade Route 1900 (log)

Dist. City 1900 (log)

Dist. Expl. Route 1911 (log)

Precolonial Outcome

Murdock: Town Size

Murdock: Social Class

Murdock: Jurisidctional Hierarchy

Murdock: Headman Elected

Murdock: Slavery

−0.1 0.0 0.1 0.2 0.3

Coefficients and 95% Confidence Intervals

Figure SM8: Placebo Test: Cash Crop Suitability and Pre-Colonial Development

Figure SM8 conducts placebo tests probing whether our suitability instrument predicts
pre-colonial development outcomes. Any developmental advantage of highly suitable areas
before the commercial transition to export agriculture could indicate a violation of the ex-
clusion restriction. All placebo regressions include distance to coast, navigable waterways,
caloric suitability, Tsetse fly and malaria suitability, ruggedness, and elevation. The first
three models are run at the grid cell-level and indicate that cells suitable for cash crops
are not significantly closer to pre-colonial trade routes or urban centers and, if anything,
further away from European explorer routes. The subsequent models are at the level of
Murdock’s ethnic homelands and show no significant associations with variables from Mur-
dock’s Ethnographic Atlas that can be seen as proxies for socioeconomic or institutional
development.

SM21
 Disaggregating Cash Crops
Results for the five most important agricultural exports
Quality Road 1998 (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) DHS Household Wealth

Cocoa ● ● ● ●
Cash Crop Dummy

Coffee ● ● ● ●

Cotton ● ● ● ●

Groundnut ● ● ● ●

Oil Palm ● ● ● ●

0.0 0.1 0.2 0.0 0.1 0.2 0.0 0.1 0.2 0.0 0.1 0.2 0.3
Coefficients and 95% Confidence Intervals

Figure SM9: Disaggregating effects by five main cash crops: cocoa, coffee, cotton, ground-
nuts, and oil palm.
Country−by−Country Analysis Country−by−Country Analysis
DV: Quality Road 1998 (Y/N); IV: Cash Crops (Y/N) DV: Night Lights 2015 (Y/N); IV: Cash Crops (Y/N)

LBR ● ZMB ●
GAB ● LBR ●
SOM ● TZA ●
ZMB ● GAB ●
SEN ● NER ●
NER ● SDN ●
TZA ● AGO ●
NGA ● RWA ●
SDN ● SEN ●
CMR ● CMR ●
TGO ● TCD ●
COD ● SLE ●
KEN ● MLI ●
GNB ● GIN ●
Country

Country

UGA ● GNQ ●
MWI ● KEN ●
TCD ● NGA ●
CIV ● SOM ●
GHA ● TGO ●
RWA ● COG ●
CAF ● CIV ●
AGO ● UGA ●
SLE ● GHA ●
GIN ● COD ●
COG ● BEN ●
BEN ● CAF ●
MOZ ● ZWE ●
ZWE ● ETH ●
MLI ● MWI ●
ETH ● MOZ ●
GNQ ● GNB ●

0.0 0.5 0.0 0.5 1.0

Coefficients and 95% Confidence Intervals Coefficients and 95% Confidence Intervals

Country−by−Country Analysis Country−by−Country Analysis

DV: City 2015 (Y/N); IV: Cash Crops (Y/N) DV: DHS Household Wealth; IV: Cash Crops (Y/N)

GAB ● GAB ●
ZMB ●
ZMB ●
LBR ●
KEN ● KEN ●

MWI ● AGO ●
SOM ●
ZWE ●
NER ●
GHA ● TZA ●
TZA ● MLI ●
TCD ●
LBR ●
SDN ●
NGA ● CMR ●
SEN ● GIN ●
ZWE ●
Country

Country

MWI ●
BEN ●
UGA ● NGA ●
CAF ●
GHA ●
GNQ ●
COG ● SLE ●

COD ● ETH ●
AGO ●
UGA ●
CMR ●
TGO ● COD ●
CIV ● RWA ●
SLE ●
SEN ●
GIN ●
ETH ● MOZ ●
MOZ ● TGO ●
RWA ●
CIV ●
MLI ●
GNB ● BEN ●

−0.4 0.0 0.4 −0.5 0.0 0.5 1.0

Coefficients and 95% Confidence Intervals Coefficients and 95% Confidence Intervals

Figure SM10: Country-by-country regressions of infrastructural and wealth outcomes on

colonial cash crop production.

SM22
 Country−by−Country Analysis Country−by−Country Analysis
DV: Quality Road 1998 (Y/N); IV: Minerals (Y/N) DV: Night Lights 2015 (Y/N); IV: Minerals (Y/N)

MOZ ● KEN ●

KEN ● DJI ●

UGA ● MOZ ●

COG ● AGO ●

ZMB ● ZMB ●

TZA ● COG ●

TGO ● SDN ●

GHA ● UGA ●

ZWE ● TZA ●
Country

Country
NGA ● SLE ●

GAB ● ZWE ●

DJI ● NGA ●

CAF ● COD ●

COD ● RWA ●

AGO ● GHA ●

RWA ● TGO ●

GIN ● LBR ●

SDN ● GAB ●

SLE ● CAF ●

SEN ● GIN ●

LBR ● SEN ●

−0.5 0.0 0.5 1.0 −0.5 0.0 0.5 1.0

Coefficients and 95% Confidence Intervals Coefficients and 95% Confidence Intervals

Country−by−Country Analysis Country−by−Country Analysis

DV: City 2015 (Y/N); IV: Minerals (Y/N) DV: DHS Household Wealth; IV: Minerals (Y/N)

MOZ ● UGA ●

TZA ●
MOZ ●
UGA ●
NGA ●
ZMB ●

GHA ● ZMB ●

NGA ●
TZA ●
DJI ●
AGO ●
ZWE ●

SLE ● GIN ●
Country

Country

AGO ●
GHA ●
RWA ●
ZWE ●
CAF ●

LBR ● COD ●

GIN ●
SLE ●
COD ●
RWA ●
TGO ●

KEN ● GAB ●

SDN ●
LBR ●
GAB ●
TGO ●
COG ●

SEN ● SEN ●

−0.5 0.0 0.5 1.0 −1 0 1

Coefficients and 95% Confidence Intervals Coefficients and 95% Confidence Intervals

Figure SM11: Country-by-country regressions of infrastructural and wealth outcomes on

colonial mining.

SM23
 Heterogeneity Across Imperial Powers
OLS Models with Colonial Cash Crop Dummy as Main Predictor
Quality Road 1998 (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) DHS Household Wealth

Britain ● ● ● ●

France ● ● ● ●
Colonizer

Portugal ● ● ● ●

Belgium ● ● ● ●

0.0 0.1 0.2 0.0 0.1 0.2 0.0 0.1 0.2 0.3 0.0 0.1 0.2
Coefficients and 95% Confidence Intervals

Heterogeneity Across Imperial Powers

OLS Models with Colonial Mineral Dummy as Main Predictor
Quality Road 1998 (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) DHS Household Wealth

Britain ● ● ● ●

France ● ● ● ●
Colonizer

Portugal ● ● ● ●

Belgium ● ● ● ●

−0.1 0.1 0.3 −0.1 0.1 0.3 0.5 0.7 0.1 0.3 −0.1 0.1 0.3 0.5 0.7 0.9 1.1
Coefficients and 95% Confidence Intervals

Figure SM12: Empire-by-empire regressions of main outcome variables on colonial cash crop
production (upper panel) and mining (lower panel).

SM24
 Heterogeneity Analysis
OLS Interaction Models with Colonial Cash Crop Dummy as Main Predictor
Quality Road 1998 (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) DHS Household Wealth

Polity IV (low) ● ● ● ●
Polity IV (med.) ● ● ● ●
Polity IV (high) ● ● ● ●
Moderator Variable & Value

VDEM (low)
VDEM (med.)
VDEM (high)

Conflict (low)
Conflict (med.)
Conflict (high)

Agr. Exports (low)

Agr. Exports (med.)
Agr. Exports (high)

Coastal Countries
Landlocked Countries

0.0 0.1 0.2 0.3 0.0 0.1 0.2 0.3 0.0 0.1 0.2 0.0 0.1 0.2 0.3 0.4
Marginal Effect and 95 % Confidence Intervals

Heterogeneity Analysis
OLS Interaction Models with Colonial Mineral Dummy as Main Predictor
Quality Road 1998 (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) DHS Household Wealth

Polity IV (low) ● ● ● ●
Polity IV (med.) ● ● ● ●
Polity IV (high) ● ● ● ●
Moderator Variable & Value

VDEM (low)
VDEM (med.)
VDEM (high)

Conflict (low)
Conflict (med.)
Conflict (high)

Mining Exports (low)

Mining Exports (med.)
Mining Exports (high)

Coastal Countries
Landlocked Countries

0.0 0.1 0.2 0.3 0.4 0.0 0.2 0.4 0.6 0.0 0.2 0.4 0.6 0.0 0.5 1.0
Marginal Effect and 95 % Confidence Intervals

Figure SM13: Heterogeneity by post-independence trajectories.

Figure SM13 reports the impact of colonial cash crops (upper panel) and mining (lower
panel) on contemporary development outcomes by various country sub-groups: different lev-
els of democracy; conflict-affectedness; agricultural/mineral export dependence; and land-
locked vs. coastal. Marginal effect estimates from multiplicative interaction models. Con-
tinuous moderators are made categorical based on sample terciles.

SM25
 Robustness: Continuous Measures & Intensive Margins
Colonial Resource Extraction & Contemporary Infrastructure and Development
Road Density 1998 (log) Night Lights 2015 (log) Urban Pop. Dens. 2015 (log) DHS Household Wealth

Cash Crops (Y/N) ● ● ● ●

Cash Crops (Y/N) & Y>0 ● ● ●

Cash Crop Value (log) ● ● ● ●

Cash Crop Value (log) & X>0 ● ● ● ●

Model

Minerals (Y/N) ● ● ● ●

Minerals (Y/N) & Y>0 ● ● ●

Mineral Value (log) ● ● ● ●

Mineral Value (log) & X>0 ● ● ● ●

0.0 0.5 1.0 0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 0.0 0.1 0.2 0.3 0.4 0.5
Coefficients and 95% Confidence Intervals

Figure SM14: Continuous outcomes and intensive margin effects

Figure SM14 shows estimates from models with continuous instead of dichotomous in-
frastructure and development outcomes. We also test for intensive-margin effects of our
resource variables by subsetting the sample to cells with non-zero production or at least one
quality road, city, or some night lights.
Figure SM15 shows robustness of our spatial IV results when using continuous versions
of the outcome and/or treatment variables.

Robustness: Continuous Measures & S2SLS Models

Colonial Resource Extraction & Contemporary Infrastructure and Development
Road Density 1998 (log) Night Lights 2015 (log) Urban Pop. Dens. 2015 (log) DHS Household Wealth

Cash Crops (Y/N, OLS) ● ● ● ●

Cash Crops (Y/N, S2SLS) ● ● ● ●

Model

Cash Crops (log, OLS) ● ● ● ●

Cash Crops (log, S2SLS) ● ● ● ●

0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0 0.0 0.1 0.2
Coefficients and 95% Confidence Intervals

Figure SM15: Spatial IV analysis using continuous outcome variables and/or continuous
versions of the cash crop treatment.

SM26
 Robustness: Conley Errors for Spatial Correlation
Point estimate and uncertainty for cash crop dummy
Quality Road 1998 (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) DHS Household Wealth

Baseline ● ● ● ●
Distance Cutoff for Conley Errors

100 km ● ● ● ●

200 km ● ● ● ●

400 km ● ● ● ●

800 km ● ● ● ●

1600 km ● ● ● ●

3200 km ● ● ● ●

0.00 0.05 0.10 0.15 0.20 0.00 0.05 0.10 0.15 0.20 0.250.00 0.05 0.10 0.15 0.20 0.00 0.05 0.10 0.15 0.20
Coefficients and 95% Confidence Intervals

Robustness: Conley Errors for Spatial Correlation

Point estimate and uncertainty for mineral dummy
Quality Road 1998 (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) DHS Household Wealth

Baseline ● ● ● ●
Distance Cutoff for Conley Errors

100 km ● ● ● ●

200 km ● ● ● ●

400 km ● ● ● ●

800 km ● ● ● ●

1600 km ● ● ● ●

3200 km ● ● ● ●

0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.1 0.2 0.3 0.0 0.1 0.2 0.3 0.0 0.2 0.4
Coefficients and 95% Confidence Intervals

Figure SM16: conley1999gmm standard errors with various distance cutoffs to address con-
cerns about spatial dependence and artificially small uncertainty estimates.

SM27
 Robustness: Changing the Spatial Unit of Analysis
Colonial Resources & Contemporary Outcomes
Quality Road 1998 (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) DHS Household Wealth

Cash Crops (Y/N; 0.25° Grid) ● ● ● ●

Cash Crops (Y/N; 0.5° Grid) ● ● ● ●

Cash Crops (Y/N; 1° Grid) ● ● ● ●

Cash Crops (Y/N; Ethnic Group) ● ● ● ●

Minerals (Y/N; 0.25° Grid) ● ● ● ●

Minerals (Y/N; 0.5° Grid) ● ● ● ●

● ● ● ●
Predictor & Spatial Unit

Minerals (Y/N; 1° Grid)

Minerals (Y/N; Ethnic Group) ● ● ● ●

0.0 0.1 0.2 0.0 0.1 0.2 0.3 0.4 0.0 0.1 0.2 0.3 0.0 0.2 0.4

Road Density 1998 (log) Night Lights 2015 (log) Urban Pop. Dens. 2015 (log) Night Lights 2015 p.c. (log)

Cash Crops (Y/N; 0.25° Grid) ● ● ● ●

Cash Crops (Y/N; 0.5° Grid) ● ● ● ●

Cash Crops (Y/N; 1° Grid) ● ● ● ●

Cash Crops (Y/N; Ethnic Group) ● ● ● ●

Minerals (Y/N; 0.25° Grid) ● ● ● ●

Minerals (Y/N; 0.5° Grid) ● ● ● ●

Minerals (Y/N; 1° Grid) ● ● ● ●

Minerals (Y/N; Ethnic Group) ● ● ● ●

0.0 0.5 1.0 1.5 0 1 2 0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5 2.0
Coefficients and 95% Confidence Intervals

Figure SM17: Binary resource variables across different spatial units of analysis.

In Figures SM17 and SM18, we check whether our results are stable to different definitions
of the spatial unit of analysis. For this purpose, we aggregate all variables to grids of 0.5 and
1.0 degree (long/lat) resolution (55.5 and 111 km, respectively, at the equator), resulting in
cells that are, on average, 4 and 16 times larger than in our baseline analysis. In addition, we
aggregate our data to ethnic homeland-country pairs as in (Michalopoulos and Papaioannou,
2013).
We use binary and continuous versions of our main outcome and predictor variables.
Figure SM17 presents results from models employing binary cash crop or mineral dummies
as main independent variable. With the exception of the urbanization outcomes, the binary
resource coefficients remain stable or decrease in size as the grid gets coarser. This seems
consistent with the limited or even negative spillover effects that we report in the main
paper as well as in Figures SM22 and SM23 below. The ethnic group-level coefficients are

SM28
 Robustness: Changing the Spatial Unit of Analysis
Colonial Resources & Contemporary Outcomes
Quality Road 1998 (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) DHS Household Wealth

Cash Crops (log; 0.25° Grid) ● ● ● ●

Cash Crops (log; 0.5° Grid) ● ● ● ●

Cash Crops (log; 1° Grid) ● ● ● ●

Cash Crops (log; Ethnic Group) ● ● ● ●

Minerals (log; 0.25° Grid) ● ● ● ●

Minerals (log; 0.5° Grid) ● ● ● ●

Predictor & Grid Cell Size

Minerals (log; 1° Grid) ● ● ● ●

Minerals (log; Ethnic Group) ● ● ● ●

0.00 0.05 0.10 0.00 0.05 0.10 0.15 0.00 0.05 0.10 0.15 0.200.0 0.1 0.2 0.3

Road Density 1998 (log) Night Lights 2015 (log) Urban Pop. Dens. 2015 (log) Night Lights 2015 p.c. (log)

Cash Crops (log; 0.25° Grid) ● ● ● ●

Cash Crops (log; 0.5° Grid) ● ● ● ●

Cash Crops (log; 1° Grid) ● ● ● ●

Cash Crops (log; Ethnic Group) ● ● ● ●

Minerals (log; 0.25° Grid) ● ● ● ●

Minerals (log; 0.5° Grid) ● ● ● ●

Minerals (log; 1° Grid) ● ● ● ●

Minerals (log; Ethnic Group) ● ● ● ●

0.0 0.2 0.4 0.6 0.0 0.5 1.0 0.00 0.25 0.50 0.75 1.00 0.0 0.5 1.0 1.5
Coefficients and 95% Confidence Intervals

Figure SM18: Continuous resource variables across different spatial units of analysis.

SM29
comparable to our baseline analysis using quarter-degree cells. A similar relationship between
grid cell and coefficient size emerges if we replace binary with continuous resource variables
(Figure SM18).

SM30
 Cash Crops, Precolonial Statehood, and the Slave Trades
Ethnic Homeland−Country−Pair Analysis
Road Density 1998 (log) Night Lights 2015 (log) Urban Population 2015 (log) DHS Household Wealth

Cash Crops (Y/N, Geo. Controls) ● ● ● ●

Cash Crops (Y/N, All Controls) ● ● ● ●

Variable & Model

Precolonial State (Y/N, Geo. Controls) ● ● ● ●

Precolonial State (Y/N, All Controls) ● ● ● ●

Slaves per area (log, Geo. Controls) ● ● ● ●

Slaves per area (log, All Controls) ● ● ● ●

−0.5 0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 −0.1 0.0 0.1 0.2
Coefficients and 95% Confidence Intervals

Figure SM19: Effect comparison: cash crops, pre-colonial centralization, slave trades

In Figure SM19 we compare the effects of colonial cash crop agriculture to two historical
factors often cited as shaping long-run development in Africa—the slave trades NunnSlave-
Trades2008, NunnRuggedness2010,NunnMistrust2011 and pre-colonial centralization Gen-
naioliPrecolonial2007, MichalopoulosPreColonial2013. Both of these variables are available
at the level of ethnic homelands as defined by murdock1959,murdock1967ethnographic. We
follow MichalopoulosPreColonial2013 and run models with ethnic homeland-country pairs
as the unit of analysis, include country fixed effects and cluster standard errors at the the
country level. We find that the historical effects of cash crop production on contemporary
infrastructural development at the subnational level are much more robust than those of
the other historical variables. The impact of pre-colonial centralization is modest and sen-
sitive to selection of control variables. And, compared to the slave trades’ negative effects
on social institutions and trust NunnMistrust2011, areas most affected by the slave trades
tend to have slightly higher levels of contemporary urbanization and wealth—due perhaps to
their favorable geography and proximity to the coast, the targeting of slave raids to initially
densely populated areas, or the persistence of markets organized around trade.

SM31
 Causal Mediation Analysis
Total and demediated direct effects (ACDE) of cash crop production
Quality Road 1998 (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) DHS Household Wealth
Total and Demediated Effects

Total Effect ● ● ● ●

ACDE (Crop Value) ● ● ● ●

ACDE (Infrastructure) ● ● ● ●

ACDE (Both) ● ● ● ●

0.0 0.1 0.2 0.0 0.1 0.2 0.0 0.1 0.2 0.0 0.1 0.2
Estimates and 95% Confidence Intervals (Cluster Bootstrap)

Figure SM20: Replication of the mechanism analysis from the main text with actual colonial
cash crop production instead of agro-climatic suitability as the main treatment variable.

Cash Crop Suitability vs. Other Geographic Fundamentals

Effect of a one−standard deviation increase in geographic covariate
Road Density 1998 (log, Z) Night Lights 2015 (log, Z) Urban Pop. Dens. 2015 (log, Z) Night Lights 2015 p.c. (log, Z)

Cash Crop Suitability (Pooled) ● ● ● ●

Cash Crop Suitability (Within) ● ● ● ●

Caloric Suitability (Pooled) ● ● ● ●

Caloric Suitability (Within) ● ● ● ●

TseTse Suitability (Pooled) ● ● ● ●

Predictor & Model

TseTse Suitability (Within) ● ● ● ●

Malaria Suitability (Pooled) ● ● ● ●

Malaria Suitability (Within) ● ● ● ●

Dist. Coast (Pooled) ● ● ● ●

Dist. Coast (Within) ● ● ● ●

Dist. Nav. River (Pooled) ● ● ● ●

Dist. Nav. River (Within) ● ● ● ●

Elevation (Pooled) ● ● ● ●

Elevation (Within) ● ● ● ●

Ruggedness (Pooled) ● ● ● ●

Ruggedness (Within) ● ● ● ●

−0.2 −0.1 0.0 0.1 0.2 −0.50 −0.25 0.00 0.25 −0.3 −0.2 −0.1 0.0 0.1 0.2 −0.4 −0.2 0.0 0.2 0.4
Coefficients and 95% Confidence Intervals

Figure SM21: Effect size comparisons from as in the main text but with continuous instead
of binary outcome variables.

SM32
 Spillover Analysis: Cash Crops
Distance to Colonial Resources & Contemporary Outcomes
Quality Road 1998 (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) DHS Household Wealth

Cash Crops (0−25 km) ● ● ● ●

Cash Crops (25−50 km) ● ● ● ●

Cash Crops (50−75 km) ● ● ● ●

Distance Bins

Cash Crops (75−100 km) ● ● ● ●

Cash Crops (100−125 km) ● ● ● ●

Cash Crops (125−150 km) ● ● ● ●

Cash Crops (150−175 km) ● ● ● ●

Cash Crops (175−200 km) ● ● ● ●

Cash Crops (200−225 km) ● ● ● ●

Cash Crops (225−250 km) ● ● ● ●

0.0 0.1 0.2 −0.1 0.0 0.1 0.2 0.0 0.1 −0.1 0.0 0.1 0.2
Coefficients and 95% Confidence Intervals

Spillover Analysis: Cash Crops

Distance to Colonial Resources & Contemporary Outcomes
Road Density 1998 (log) Night Lights 2015 (log) Urban Pop. Dens. 2015 (log) Night Lights 2015 p.c. (log)

Cash Crops (0−25 km) ● ● ● ●

Cash Crops (25−50 km) ● ● ● ●

Cash Crops (50−75 km) ● ● ● ●

Distance Bins

Cash Crops (75−100 km) ● ● ● ●

Cash Crops (100−125 km) ● ● ● ●

Cash Crops (125−150 km) ● ● ● ●

Cash Crops (150−175 km) ● ● ● ●

Cash Crops (175−200 km) ● ● ● ●

Cash Crops (200−225 km) ● ● ● ●

Cash Crops (225−250 km) ● ● ● ●

0.0 0.5 1.0 1.5 −0.5 0.0 0.5 1.0 −0.2 0.0 0.2 0.4 0.6 0.0 0.5
Coefficients and 95% Confidence Intervals

Figure SM22: Distance band regressions (cash crops) using continuous (lower panel) instead
of binary outcome variables (upper panel, for comparison).

SM33
 Spillover Analysis: Minerals
Distance to Colonial Resources & Contemporary Outcomes
Quality Road 1998 (Y/N) Night Lights 2015 (Y/N) City 2015 (Y/N) DHS Household Wealth

Minerals (0−25 km) ● ● ● ●

Minerals (25−50 km) ● ● ● ●

Minerals (50−75 km) ● ● ● ●

Distance Bins

Minerals (75−100 km) ● ● ● ●

Minerals (100−125 km) ● ● ● ●

Minerals (125−150 km) ● ● ● ●

Minerals (150−175 km) ● ● ● ●

Minerals (175−200 km) ● ● ● ●

Minerals (200−225 km) ● ● ● ●

Minerals (225−250 km) ● ● ● ●

−0.1 0.0 0.1 0.2 0.0 0.1 0.2 −0.05 0.00 0.05 0.10 0.15 0.20 0.0 0.2 0.4
Coefficients and 95% Confidence Intervals

Spillover Analysis: Minerals

Distance to Colonial Resources & Contemporary Outcomes
Road Density 1998 (log) Night Lights 2015 (log) Urban Pop. Dens. 2015 (log) Night Lights 2015 p.c. (log)

Minerals (0−25 km) ● ● ● ●

Minerals (25−50 km) ● ● ● ●

Minerals (50−75 km) ● ● ● ●

Distance Bins

Minerals (75−100 km) ● ● ● ●

Minerals (100−125 km) ● ● ● ●

Minerals (125−150 km) ● ● ● ●

Minerals (150−175 km) ● ● ● ●

Minerals (175−200 km) ● ● ● ●

Minerals (200−225 km) ● ● ● ●

Minerals (225−250 km) ● ● ● ●

−0.5 0.0 0.5 1.0 0.0 0.5 1.0 1.5 2.0 0.0 0.4 0.8 0.0 0.5 1.0 1.5
Coefficients and 95% Confidence Intervals

Figure SM23: Distance band regressions (minerals) using continuous (lower panel) and bi-
nary outcome variables.

SM34
References

Acemoglu, Daron and James A. Robinson. 2012. Why Nations Fail: The Origins of Power,
Prosperity and Poverty. New York: Crown Publishers.

Acemoglu, Daron, Simon Johnson and James A. Robinson. 2001. “The Colonial Origins
of Comparative Development: An Empirical Investigation.” American Economic Review
91(5):1369–1401.

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.

Acharya, Avidit, Matthew Blackwell and Maya Sen. 2016. “Explaining Causal Findings
Without Bias: Detecting and Assessing Direct Effects.” American Political Science Review
110(3):512–529.

Ajayi, Ade and Michael Crowder. 1985. Historical Atlas of Africa. Cambridge, UK: Cam-
bridge University Press.

Alesina, Alberto, Stelios Michalopoulos and Elias Papaioannou. 2016. “Ethnic Inequality.”
Journal of Political Economy 124(2):428–488.

Alsan, Marcella. 2014. “The Effect of the Tsetse Fly on African Development.” The American
Economic Review 105(1):382–410.

Amin, Samir. 1972. “Underdevelopment and dependence in Black Africa—origins and con-
temporary forms.” The Journal of Modern African Studies 10(4):503–524.

Amin, Samir. 1974. Modern Migrations in Western Africa: Studies Presented and Discussed
at the Eleventh International African Seminar, Dakar, April 1972. London: Oxford
University Press.

SM35
Austin, Gareth. 1996. Mode of Production or Mode of Cultivation: Explaining the Failure
of European Cocoa Planters in Competition with African Farmers tn Colonial Ghana.
Springer pp. 154–175.

Austin, Gareth. 2008. “Resources, Techniques, and Strategies South of the Sahara: Revising
the Factor Endowments Perspective on African Economic Development, 1500–2000.” The
Economic History Review 61(3):587–624.

Austin, Gareth. 2009. “Cash Crops and Freedom: Export Agriculture and the Decline of
Slavery in Colonial West Africa.” International Review of Social History 54(01):1–37.

Austin, Gareth. 2010. “African Economic Development and Colonial Legacies.” International
Development Policy/ Revue Internationale de Politique Développment 1(1):11–32.

Auty, Richard M. 1997. “Natural resource endowment, the state and development strategy.”
Journal of International Development: The Journal of the Development Studies Associa-
tion 9(4):651–663.

Auty, Richard M., ed. 2001. Resource abundance and economic development. Oxford Uni-
versity Press.

Baldwin, Robert E. 1956. “Patterns of development in newly settled regions.” The Manch-
ester School 24(2):161–179.

Bates, Robert H. 1981. Markets and States in Tropical Africa. Berkeley, CA: University of
California Press.

Bates, Robert H. 1983. Essays on the Political Economy of Rural Africa. University of
California Press.

Berman, Nicolas, Mathieu Couttenier, Dominic Rohner and Mathias Thoenig. 2017. “This
Mine is Mine! How Minerals Fuel Conflicts in Africa.” American Economic Review
107(6):1564–1610.

SM36
Betz, Timm, Scott J Cook and Florian M Hollenbach. 2019. “Spatial interdependence and
instrumental variable models.” Political Science Research and Methods pp. 1–16.

Bleakley, Hoyt and Jeffrey Lin. 2012. “Portage and Path Dependence.” The Quarterly Journal
of Economics pp. 587–644.

Boone, Catherine. 1992. Merchant Capital and the Roots of State Power in Senegal, 1930-
1985. New York: Cambridge University Press.

Boone, Catherine. 2003. Political topographies of the African state: Territorial authority and
institutional choice. Cambridge University Press.

Brooks, George E. 1975. “Peanuts and Colonialism: Consequences of the Commercialization

of Peanuts in West Africa, 1830–70.” The Journal of African History 16(01):29–54.

Cardoso, Fernando Henrique and Enzo Faletto. 1979. Dependency and development in Latin
America (Dependencia y desarrollo en América Latina, engl.). Univ of California Press.

Cederman, Lars-Erik, Kristian Skrede Gleditsch and Halvard Buhaug. 2013. Inequality,
Grievances, and Civil War. New York: Cambridge University Press.

Cederman, Lars-Erik, Nils B. Weidmann and Kristian Skrede Gleditsch. 2011. “Horizontal
Inequalities and Ethnonationalist Civil War: A Global Comparison.” American Political
Science Review 105(03):478–495.

Cervellati, Matteo, Giorgio Chiovelli and Elena Esposito. 2019. Bite and Divide: Malaria
and Ethnolinguistic Diversity. Technical report CEPR Discussion Papers.

Chaves, Isaias, Stanley L. Engerman and James A. Robinson. 2014. Reinventing the Wheel:
The Economic Benefts of Wheeled Transportation in Early Colonial British West Africa.
New York: Cambridge University Press.

Collier, Paul and Anke Hoeffler. 2004. “Greed and Grievance in Civil War.” Oxford economic
papers 56(4):563–595.

SM37
Conley, Timothy G. 1999. “GMM estimation with cross sectional dependence.” Journal of
Econometrics 92(1):1–45.

Conning, Jonathan H and James A Robinson. 2009. “Enclaves and development: An empir-
ical assessment.” Studies in Comparative International Development 44(4):359–385.

Coquery-Vidrovitch, Catherine. 1975. “Research on an African Mode of Production.” Cri-

tique of Anthropology 2(4-5):38–71.

Cordell, Dennis D., Joel W. Gregory and Victor Piché. 1996. Hoe and Wage : A Social
History of a Circular Migration System in West Africa. Boulder, Colo.: Westview Press.

Crowder, Michael. 1978. Colonial West Africa: Collected Essays. London: F. Cass.

Daumalin, Xavier. 2004. Commercial presence, colonial penetration: Marseille traders in

west Africa in the nineteenth century. In From Slave Trade to Empire: European Coloni-
sation of Black Africa 1780s-1880s, ed. Olivier Pétré-Grenouilleau. Routledge pp. 225–246.

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

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

Dell, Melissa and Benjamin A Olken. 2019. “The Development Effects Of The Extractive
Colonial Economy: The Dutch Cultivation System In Java.” The Review of Economic
Studies .

Diamond, Jared M. 1997. Guns, Germs, and Steel: The Fates of Human Societies. New
York: W.W. Norton.

Dickson, Kwamina B. 1969. A Historical Geography of Ghana. New York: Cambridge

University Press.

SM38
dos Santos, Theotonio. 1970. “The structure of dependence.” The american economic review
60(2):231–236.

Dube, Oeindrila and Juan F Vargas. 2013. “Commodity price shocks and civil conflict:
Evidence from Colombia.” The Review of Economic Studies 80(4):1384–1421.

Easterly, William. 2007. “Inequality does cause underdevelopment: Insights from a new
instrument.” Journal of development economics 84(2):755–776.

Easterly, William and Ross Levine. 2003. “Tropics, Germs, and Crops: How Endowments
Influence Economic Development.” Journal of Monetary Economics 50(1):3–39.

Edwards, Ryan B. 2019. “Export Agriculture and Rural Poverty: Evidence from Indonesian
Palm Oil.” Unpublished Manuscript .

Eltis, David. 2013. The slave trade commercial agriculture in an African context. Boydell
Brewer Ltd p. 28–53.

Engerman, Stanley L. and Kenneth L. Sokoloff. 1997. Factor Endowments, Institutions,

and Differential Paths of Growth Among New World Economies. Stanford, CA: Stanford
Unviversity Press.

Enke, Benjamin. 2019. “Kinship, cooperation, and the evolution of moral systems.” The
Quarterly Journal of Economics 134(2):953–1019.

Fanon, Frantz. 1963. The Wretched of the Earth. London: Grove Press.

Fearon, James D and David D Laitin. 2003. “Ethnicity, Insurgency, and Civil War.” American
Political Science Review 97(01):75–90.

Fenske, James. 2014. “Ecology, Trade, and States In Pre-Colonial Africa.” Journal of the
European Economic Association 12(3):612–640.

SM39
Frank, Andre Gunder. 1966. “The development of underdevelopment.” Monthly Review
18:17–31.

Frankema, Ewout, Jeffrey Williamson and Pieter Woltjer. 2018. “An economic rationale for
the West African scramble? The commercial transition and the commodity price boom of
1835–1885.” The Journal of Economic History 78(1):231–267.

Galletti, Robert. 1956. Nigerian cocoa farmers: an economic survey of Yoruba cocoa farming
families. Oxford University Press.

Gallup, John Luke, Jeffrey D Sachs and Andrew D Mellinger. 1999. “Geography and Eco-
nomic Development.” International Regional Science Review 22(2):179–232.

Gardner, Leigh A. 2012. Taxing Colonial Africa The Political Economy of British Imperial-
ism. Oxford: Oxford University Press.

Gennaioli, Nicola and Ilia Rainer. 2007. “The Modern Impact of Precolonial Centralization
in Africa.” Journal of Economic Growth 12(3):185–234–234.

Gleditsch, Nils Petter, Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg and Håvard
Strand. 2002. “Armed conflict 1946-2001: A new dataset.” Journal of peace research
39(5):615–637.

Green, Toby. 2013. The export of rice millet from Upper Guinea into the sixteenth-century
Atlantic trade. Boydell Brewer Ltd p. 79–97.

Hall, Robert E and Charles I Jones. 1999. “Why do some countries produce so much more
output per worker than others?” The quarterly journal of economics 114(1):83–116.

Hance, William A, Vincent Kotschar and Richard J Peterec. 1961. “Source Areas of Export
Production in Tropical Africa.” Geographical Review 51(4):487–499.

Hance, William Adams. 1964. The Geography of Modern Africa. New York: Columbia
University Press.

SM40
Havinden, Michael Ashley and David Meredith. 1993. Colonialism and Development: Britain
and its Tropical Colonies, 1850-1960. New York: Routledge.

Headrick, Daniel R. 2012. Power over peoples: Technology, environments, and Western
imperialism, 1400 to the present. Princeton University Press.

Henderson, J Vernon, Tim Squires, Adam Storeygard and David Weil. 2018. “The Global
Distribution of Economic Activity: Nature, History, and the Role of Trade1.” The Quar-
terly Journal of Economics 133(1):357–406.

Herbst, Jeffrey. 2000. States and Power in Africa: Comparative Lessons in Authority and
Control. Princeton, NJ: Princeton University Press.

Hill, Polly. 1957. The Gold Coast Cocoa Farmer: a Preliminary Survey. Oxford University
Press.

Hill, Polly. 1963. The migrant cocoa-farmers of southern Ghana: A study in rural capitalism.
Cambrdige University Press.

Hirano, Keisuke, Guido W Imbens and Geert Ridder. 2003. “Efficient estimation of average
treatment effects using the estimated propensity score.” Econometrica 71(4):1161–1189.

Hirschman, Albert O. 1958. The Strategy of Economic Development. Yale University Press.

Hirschman, Albert O. 1977. “A generalized linkage approach to development, with special

reference to staples.” Economic development and cultural change 25:67.

Hodler, Roland and Paul A. Raschky. 2014. “Regional Favoritism.” The Quarterly Journal
of Economics 129(2):995–1033.

Hogendorn, Jan S. 1969. Economic Initiative and African Cash Farming: Pre-Colonial
Origins and Early Colonial Developments. In Colonialism in Africa, 1870-1960: The
Economics of Colonialism, ed. Lewis H. Gann and Duignan Peter. London: Cambridge
University Press pp. 283–328.

SM41
Hogendorn, Jan S. 1978. Nigerian Groundnut Exports: Origins and Early Development.
Zaria: Ahmadu Bello University Press.

Hopkins, Anthony G. 1973. An Economic History of West Africa. New York: Columbia
University Press.

Huillery, Elise. 2009. “History Matters: The Long-Term Impact of Colonial Public invest-
ments in French West Africa.” American Economic Journal: Applied Economics pp. 176–
215.

Huillery, Elise. 2011. “The Impact of European Settlement within French West Africa: Did
Pre-Colonial Prosperous Areas Fall Behind?” Journal of African Economies 20(2):263–
311.

Hyam, Ronald. 1968. Elgin and Churchill at the Colonial Office, 1905-08. Springer.

Inikori, Joseph E. 2007. “Africa and the globalization process: western Africa, 1450–1850.”
Journal of Global History 2(1):63–86.

Inikori, Joseph E. 2014. Reversal of Fortune and Socioeconomic Development in the Atlantic
World: A Comparative Examination of West Africa and the Americas, 1400–1850. In
Africa’s Development in Historical Perspective, ed. Emmanuel Akyeampong, Robert H
Bates, Nathan Nunn and James Robinson. New York: Cambridge University Press pp. 56–
88.

Innis, Harold Adams. 1930. The fur trade in Canada: An introduction to Canadian economic
history. Yale University Press.

Jedwab, Remi and Alexander Moradi. 2012. “Colonial investments and long-term devel-
opment in Africa: Evidence from Ghanaian railways.” Unpublished manuscript, George
Washington University and Sussex University .

SM42
Jedwab, Remi and Alexander Moradi. 2016. “The Permanent Effects of Transportation
Revolutions in Poor Countries: Evidence from Africa.” Review of Economics and Statistics
98(2):268–284.

Jedwab, Remi, Edward Kerby and Alexander Moradi. 2017. “History, Path Dependence
and Development: Evidence from Colonial Railways, Settlers and Cities in Kenya.” The
Economic Journal pp. 1467–1494.

Jedwab, Rémi, Felix Meier zu Selhausen, Alexander Moradi et al. 2019. The Economics of
Missionary Expansion: Evidence from Africa and Implications for Development. Technical
report African Economic History Network.

Kasara, Kimuli. 2007. “Tax Me If You Can: Ethnic Geography, Democracy, and the Taxation
of Agriculture in Africa.” American Political Science Review 101(1):159–172.

Kelly, Morgan. 2019. “The Standard Errors of Persistence.” CEPR Discussion Paper 13783
.

Kerr, Rachel Bezner. 2010. “Groundnuts as ‘economic crop’ or ‘wife of the home’ in Northern
Nyasaland.” Journal of Historical Geography 36(1):79–89.

Krugman, Paul. 1991. “Increasing Returns and Economic Geography.” Journal of political
economy 99(3):483–499.

Lange, Matthew. 2009. Lineages of Despotism and Development: British Colonialism and
State Power. Chicago: The University of Chicago Press.

Leonard, David K. and Scott Straus. 2003. Africa’s Stalled Development: International
Causes and Cures. Boulder, CO: Lynne Rienner Publishers.

Lowes, Sara and Eduardo Montero. 2017. “Blood Rubber: The Effects of Labor Coercion
on Institutions and Culture in the DRC.”.

SM43
Lujala, Päivi. 2010. “The Spoils of Nature: Armed civil conflict and rebel access to natural
resources.” Journal of peace research 47(1):15–28.

Lynn, Martin. 1997. Commerce and Economic Change in West Africa: The Palm Oil Trade
in the Nineteenth Century. New York: Cambridge University Press.

Mamo, Nemera, Sambit Bhattacharyya and Alexander Moradi. 2019. “Intensive and exten-
sive margins of mining and development: Evidence from Sub-Saharan Africa.” Journal of
Development Economics 139:28–49.

Manning, Patrick. 1990. Slavery and African life: Occidental, Oriental, and African slave
trades. Vol. 67 Cambridge University Press.

Marden, Samuel. 2016. “The Agricultural Roots of Industrial Development: ‘Forward Link-
ages’ in Reform-Era China.” Unpublished Manuscript .

Marshall, Fiona and Elisabeth Hildebrand. 2002. “Cattle before crops: the beginnings of
food production in Africa.” Journal of World Prehistory 16(2):99–143.

Mayshar, Joram, Omer Moav, Zvika Neeman and Luigi Pascali. 2018. “The Emergence of
Hierarchies and States: Productivity vs. Appropriability.” Working Paper.

McClellan, Charles W. 1980. “Land, Labor, and Coffee: The South’s Role in Ethiopian
Self-Reliance, 1889-1935.” African Economic History (9):69–83.

McPhee, Allan. 1926. The Economic Revolution in British West Africa. London: G. Rout-
ledge Sons.

Meredith, David. 1988. “The Colonial Office, British business interests and the reform of
cocoa marketing in West Africa, 1937–1945.” The Journal of African History 29(2):285–
300.

Michalopoulos, Stelios. 2012. “The Origins of Ethnolinguistic Diversity.” The American

Economic Review 102(4):1508.

SM44
Michalopoulos, Stelios and Elias Papaioannou. 2013. “Pre-Colonial Ethnic Institutions and
Contemporary African Development.” Econometrica 81(1):113–152.

Michalopoulos, Stelios and Elias Papaioannou. 2016. “The Long-run Effects of the Scramble
for Africa.” The American Economic Review 106(7):1802–1848.

Michalopoulos, Stelios and Elias Papaioannou. 2018. “Spatial patterns of development: A

meso approach.” Annual Review of Economics 10:383–410.

Michalopoulos, Stelios and Elias Papaioannou. 2020. “Historical legacies and African devel-
opment.” Journal of Economic Literature 58(1):53–128.

Michalopoulos, Stelios, Louis Putterman and David N Weil. 2019. “The influence of ancestral
lifeways on individual economic outcomes in Sub-Saharan Africa.” Journal of the European
Economic Association 17(4):1186–1231.

Moradi, Alexander, Gareth Austin and Jörg Baten. 2013. Heights and development in a cash-
crop colony: living standards in Ghana, 1870-1980. Technical report African Economic
History Network.

Moseley, Katherine P. 1992. “Caravel and Caravan: West Africa and the World-Economies,
ca. 900-1900 AD.” Review (Fernand Braudel Center) pp. 523–555.

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

Murdock, George Peter. 1967. “Ethnographic Atlas: A Summary.” Ethnology 6(2):109–236.

Méndez-Chacon, Esteban and Diana Van Patten. 2019. “Multinationals, Monopsony and
Local Development: Evidence from the United Fruit Company.” Unpublished Manuscript
.

Newbury, Colin W. 1966. “North African and Western Sudan Trade in the Nineteenth
Century: A Re-evaluation.” The Journal of African History 7(2):233–246.

SM45
Nordhaus, William D. 2006. “Geography and Macroeconomics: New Data and New Find-
ings.” Proceedings of the National Academy of Sciences of the United States of America
103(10):3510–3517.

North, Douglass C. 1959. “Agriculture in regional economic growth.” Journal of Farm Eco-
nomics 41(5):943–951.

North, Douglass C, John Joseph Wallis, Barry R Weingast et al. 2009. Violence and social
orders: A conceptual framework for interpreting recorded human history. Cambridge
University Press.

Nunn, Nathan. 2008. “The Long-Term Effects of Africa’s Slave Trades.” The Quarterly
Journal of Economics 123(1):139–176.

Nunn, Nathan and Diego Puga. 2010. “Ruggedness: The Blessing of Bad Geography in
Africa.” Review of Economics and Statistics 94(1):20–36.

Nunn, Nathan and Leonard Wantchekon. 2011. “The Slave Trade and the Origins of Mistrust
in Africa.” The American Economic Review 101(7):3221–3252.

Nunn, Nathan and Nancy Qian. 2011. “The Potato’s Contribution to Population and Ur-
banization: Evidence from a Historical Experiment.” The Quarterly Journal of Economics
126(2):593–650.

Olsson, Ola and Douglas A Hibbs. 2005. “Biogeography and Long-Run Economic Develop-
ment.” European Economic Review 49(4):909–938.

Osafo-Kwaako, Philip and James A. Robinson. 2013. “Political Centralization in Pre-Colonial

Africa.” Journal of Comparative Economics 41(1):6–21.

Ostby, Gudrun. 2008. “Polarization, Horizontal Inequalities and Violent Civil Conflict.”
Journal of Peace Research 45(2):143–162.

SM46
Price, Jacob M. 1974. “Economic Function and Growth of American Port Towns in the
Eighteenth Century.” Perspectives in American History 8:123–186.

Raynaut, Claude. 2002. Populations and Land: Multiple Dynamics and Contrasting Realities.
New York: Routledge.

Ricart-Huguet, Joan. 2017. “The origins of colonial investments in former British and French
Africa.” Princeton University .

Rodney, Walter. 1972. How Europe Underdeveloped Africa. London: Bogle-L’Ouverture

Publications.

Román, Miguel O, Zhuosen Wang, Qingsong Sun, Virginia Kalb, Steven D Miller, Andrew
Molthan, Lori Schultz, Jordan Bell, Eleanor C Stokes, Bhartendu Pandey et al. 2018.
“NASA’s Black Marble nighttime lights product suite.” Remote Sensing of Environment
210:113–143.

Sachs, Jeffrey D, Andrew D Mellinger and John L Gallup. 2001. “The Geography of Poverty
and Wealth.” Scientific American 284(3):70–75.

Sachs, Jeffrey D and Andrew M Warner. 2001. “The curse of natural resources.” European
economic review 45(4-6):827–838.

Scott, James C. 2017. Against the Grain: A Deep history of the Earliest States. New Haven,
CT: Yale University Press.

Shaver, Andrew, David B Carter and Tsering Wangyal Shawa. 2019. “Terrain ruggedness
and land cover: Improved data for most research designs.” Conflict Management and Peace
Science 36:191–218.

Smith, Adam. 1776. An Inquiry into the Nature and Causes of the Wealth of Nations.
London: W. Strahan and T. Cadell.

SM47
Sokoloff, Kenneth L and Stanley L Engerman. 2000. “Institutions, factor endowments, and
paths of development in the new world.” Journal of Economic perspectives 14(3):217–232.

Stewart, Frances. 2008. Horizontal Inequalities and Conflict: Understanding Group Violence
in Multiethnic Societies. New York: Palgrave Macmillan.

van de Walle, Nicolas. 2009. “The Institutional Origins of Inequality in Sub-Saharan Africa.”
Annual Review of Political Science 12:307–327.

Van der Ploeg, Frederick. 2011. “Natural resources: curse or blessing?” Journal of Economic
literature 49(2):366–420.

Vogt, Manuel, Nils-Christian Bormann, Seraina Rüegger, Lars-Erik Cederman, Philipp M

Hunziker and Luc Girardin. 2015. “Integrating Data on Ethnicity, Geography, and
Conflict: The Ethnic Power Relations Dataset Family.” Journal of Conflict Resolution
59(7):1327–1342.

Weiss, Daniel J, Samir Bhatt, Bonnie Mappin, Thomas P Van Boeckel, David L Smith,
Simon I Hay and Peter W Gething. 2014. “Air temperature suitability for Plasmodium
falciparum malaria transmission in Africa 2000-2012: a high-resolution spatiotemporal
prediction.” Malaria Journal 13(1):171.

Wellesley, Dorothy. 1934. Sir George Goldie, Founder of Nigeria: A Memoir. Macmillan.

Wilfahrt, Martha. 2018. “Precolonial legacies and institutional congruence in public goods
delivery: Evidence from decentralized West Africa.” World Politics 70(2):239–274.

Wolpe, Harold. 1972. “Capitalism and Cheap Labour-Power in South Africa: From Segre-
gation to Apartheid.” Economy and Society 1(4):425–456.

Wucherpfennig, Julian, Nils B Weidmann, Luc Girardin, Lars-Erik Cederman and Andreas
Wimmer. 2011. “Politically Relevant Ethnic Groups across Space and Time: Introducing
the GeoEPR Dataset.” Conflict Management and Peace Science 28(5):423–437.

SM48
Young, Crawford. 1994. The African Colonial State in Comparative Perspective. Yale Uni-
versity Press.

SM49
SM2 Sources of Primary Commodity Production by

Colony

The dataset we compiled to validate hance1961source in Figure SM1 only includes primary
commodities that account for 10% or more of total exports as of 1960. Aggregate primary
commodity exports data come from the United Nations, Yearbook of International Statis-
tics for as close to 1960 as available. Country-specific subnational production data comes
from unique sets of sources for each colony or sets of colonies (French West Africa, French
Equatorial Africa), again as close as possible to the year 1960. We use these data sources to
validate the Hance map at the level of subnational administrative units (districts or regions).
See Figure S1 above.

Angola

• Primary commodity exports as of 1960

◦ Coffee (36%)

◦ Diamonds (17%)

• Sources of administrative and spatial primary commodity production data

◦ Angola. Repartição de Estatística Geral. 1959. Anuário Estatístico – Repartição

de Estatística Geral.

◦ Angola. Repartição de Estatística Geral. Censo geral da população, 1940. Lu-

anda: Imprensa Nacional, 1941.

◦ Simões, J. M. De Pitta, Ligia Marques, and J. Ferreira. 1966. Ocorrências

minerais. Luanda: Direção Provincial dos Servicos de Geologia e Minas.

◦ Van Dongen, Irene S. 1961. “Coffee Trade, Coffee Regions, and Coffee Ports in
Angola.” Economic Geography 37(4): 320-346.

SM50
Belgian Congo (DRC)

• Primary commodity exports as of 1960

◦ Copper (27%)

◦ Coffee (14%)

◦ Palm (13%)

• Sources of administrative and spatial primary commodity production data

◦ Atlas du Congo Belge et du Ruanda-Urundi. 1955. 11. Congo Belge et Ruanda-

Urundi: Productions Agricoles.

◦ Belgian Congo. Volume II. 1960. Brussels: Belgian Congo and Ruanda-Urundi
Information and Public Relations Office.

◦ Belgique. Ministère du Congo Belge et Ruanda-Urundi. 1959-1960. Bulletin

agricole du Congo Belge et Ruanda-Urundi = Landbouwkundig Tijdschrift voor
Belgisch-Congo en Ruanda-Urundi.

◦ United States. 1944. Belgian Congo Copper Mines (including cobalt, iron, lead,
zinc, radium, gold and limestone deposits). Washington, D.C.: Foreign Economic
Administration.

Cameroon

• Primary commodity exports as of 1960

◦ Cocoa (25%)

◦ Coffee (20%)

◦ Aluminium (20%)

• Sources of administrative and spatial primary commodity production data

SM51
 ◦ Service Colonial des Statistiques. Annuaire Statistique Du Cameroun 1947. Paris:
Ministère de la France d’Outre-Mer.

◦ Service Colonial des Statistiques. Annuaire Statistique Du Cameroun 1938-1945.

Paris: Ministère de la France d’Outre-Mer.

Central African Republic

• Primary commodity exports as of 1960

◦ Cotton (44.7%)

◦ Coffee (24.9%)

◦ Diamonds (12.1%)

• Sources of administrative and spatial primary commodity production data

◦ French Equatorial Africa. Haut-Commissariat. 1950-1955. Annuaire Statistique

de l’Afrique Equatoriale Française.

◦ France. Service de statistiques chargé des relations et de la coopération avec les

états d’Outre-Mer. Recensement Général De La Population (Décembre 1956).
Résultats Définitifs. Service de Statistique.

Chad

• Primary commodity exports as of 1960

◦ Cotton (80%)

• Sources of administrative and spatial primary commodity production data

◦ French Equatorial Africa. Haut-Commissariat. 1950-1955. Annuaire Statistique

de l’Afrique Equatoriale Française.

SM52
 ◦ Cabot, Jean. 1957. ‘La Culture Du Coton Au Tchad.” Annales de Géographie
66(358): 499-508.

◦ Cabot, Jean, and Christian Bouquet. 1972. Atlas Pratique Du Tchad. Paris:
Institut Geographique National.

Côte d’Ivoire

• Primary commodity exports as of 1960

◦ Coffee (45%)

◦ Cocoa (20%)

• Sources of administrative and spatial primary commodity production data

◦ Côte d’Ivoire. 1958. Inventaire économique de la Côte d’Ivoire (1947 à 1956).

Carte no. 6: Le Café en Côte d’Ivoire en 1957.

◦ Côte d’Ivoire. 1958. Inventaire économique de la Côte d’Ivoire (1947 à 1956).

Carte no. 7: Le Cacao en Côte d’Ivoire en 1957.

◦ Côte D’Ivoire. Service de L’Agriculture. Territoire de la Côte D’Ivoire. Rapport

Annuel 1957. VI Partie: Statistiques.

◦ United Africa Company, ltd. 1956. ”Cocoa in West Africa.” no. 18: 1-20.

Dahomey (Benin)

• Primary commodity exports as of 1960

◦ Palm products (66%)

• Sources of administrative and spatial primary commodity production data

◦ Rossignol, P, H Borius, Suzanne Truitard, and Jean Kerhor. 1944. Les Exporta-
tions Agricoles Des Cercles De L’Afrique Occidentale Et Du Togo Français. Paris:
Direction des Affaires Economiques.

SM53
Ethiopia

• Primary commodity exports as of 1960

◦ Coffee (56.5%)

• Sources of administrative and spatial primary commodity production data

◦ Ethiopian Economic Review. 1962. Addis Ababa.

Gambia

• Primary commodity exports as of 1960

◦ Groundnuts (88.2%)

• Sources of administrative and spatial primary commodity production data

◦ Second Annual Report of the Gambia Oilseeds Marketing Board and the Gambia
Oilseeds Marketing Company Ltd. Season 1950/1951. 1952. Bathurst.

Guinea

• Primary commodity exports as of 1960

◦ Bauxite/Aluminum (47.4%)

◦ Diamonds (12.6%)

◦ Bananas (10.7%)

◦ Coffee (10.5%)

• Sources of administrative and spatial primary commodity production data

◦ Lafosse, Georges. 1952. Guinée Française. Carte économique [1 : 2.000.000].

Paris: Institut Géographique National.

SM54
 ◦ Rossignol, P, H Borius, Suzanne Truitard, and Jean Kerhor. 1944. Les Exporta-
tions Agricoles Des Cercles De L’Afrique Occidentale Et Du Togo Français. Paris:
Direction des Affaires Economiques.

Gold Coast (Ghana)

• Primary commodity exports as of 1960

◦ Cocoa (60%)

• Sources of administrative and spatial primary commodity production data

◦ Gold Coast Survey Department. 1949. Atlas of the Gold Coast (Map 15. Agri-
cultural Products.).

◦ Ghana Cocoa Marketing Board. 1957-1961. Annual report and accounts for the
year ended.

◦ Gold Coast. Census Office. The Gold Coast Census of Population 1948: Report
and Tables.

◦ Great Britain Colonial Office. 1952. An Economic Survey of the Colonial Terri-
tories, 1951: The Gambia, the Gold Coast, Nigeria, Sierra Leone, and St. Helena.
Volume 3. London.

◦ Urquhart, D. H. 1955. Report on the Cocoa Industry in Sierra Leone and Notes
on the Cocoa Industry of the Gold Coast (‘Cocoa Production in the Gold Coast”.
p. 30.). Bournville: Cadbury Bros.

Kenya

• Primary commodity exports as of 1960

◦ Coffee (24%)

◦ Tea (24%)

SM55
 ◦ Sisal (11%)

• Sources of administrative and spatial primary commodity production data

◦ East African Statistical Department. 1955. Kenya Agricultural Census (Highland

and Asian Settled Area). Nairobi.

◦ Matheson, Alastair. 1959. Tea in Kenya: A Short Illustrated Account of an Ex-

panding and Progressive Industry - the Tea Industry of Kenya. Nairobi: Prepared
for the Tea Board of Kenya by the Dept. of Information.

◦ Colony and Protectorate of Kenya Government Printer. 1953. Report on the

Census of the Non-Native Population of Kenya Colony and Protectorate Taken
on the Night of the 25th February, 1948.

◦ Survey of Kenya. 1962. Atlas of Kenya: A Comprehensive Series of New and

Authentic Maps Prepared from the National Survey and Other Governmental
Sources, with Gazetteer and Notes on Pronunciation & Spelling. Nairobi.

Liberia

• Primary commodity exports as of 1960

◦ Iron Ore (47%)

◦ Rubber (42%)

• Sources of administrative and spatial primary commodity production data

◦ Church, RJ Harrison. 1969. “The Firestone Rubber Plantations in Liberia.”

Geography 54(4): 430–437.

◦ United States. Bureau of Mines. 1961. Geological Survey: Minerals Yearbook.

Washington DC.

◦ United States. Army. 1964. Area Handbook for Liberia. Washington, DC: US
Govt. Print. Off

SM56
 ◦ United States. 1964. Liberia: A Market for U.S. Products. Washington, DC: US
Govt. Print. Off.

French Soudan (Mali)

• Primary commodity exports as of 1960

◦ Coffee (56.5%)

• Sources of administrative and spatial primary commodity production data

◦ Rossignol, P, H Borius, Suzanne Truitard, and Jean Kerhor. 1944. Les Exporta-
tions Agricoles Des Cercles De L’Afrique Occidentale Et Du Togo Français. Paris:
Direction des Affaires Economiques.

Madagascar

• Primary commodity exports as of 1960

◦ Coffee (40%)

• Sources of administrative and spatial primary commodity production data

◦ Humbert, H., and G. Cours Darne. 1964. ”Carte Internationale du Tapis Végé-
tal et des Conditions Écologiques, République Malagache.” Pondichéry: Institut
Français de Pondichéry.

◦ Association des Géographes de Madagascar and Françoise Le Bourdiec. 1969.

Atlas de Madagascar (“Planche 33: Café, Cacao, Bananes.”). Tananarive: Bureau
Pour le Développement de la Production Agricole.

Mozambique

• Primary commodity exports as of 1960

SM57
 ◦ Cotton (32%)

◦ Sugar (13%)

◦ Cashew (13%)

• Sources of administrative and spatial primary commodity production data

◦ Bravo, N. S. 1963. A Cultura Algodoeira na Economia do Norte de Moçambique.

Lisboa.

◦ Mozambique. 1962. Atlas de Moçambique (30. Agricultura). Lourenço Marques:

Empresa Moderna.

◦ Mozambique. 1951. Recenseamento Agricola: Repartição Técnica de Estatística:

259–283. Lourenco Marques: Imprensa Nacional.

◦ Mozambique. 1951. Recenseamento Agricola: Repartição Técnica de Estatística:

414–425. Lourenco Marques: Imprensa Nacional.

Nyasaland (Malawi)

• Primary commodity exports as of 1960

◦ Tobacco (45%)

◦ Tea (40%)

• Sources of administrative and spatial primary commodity production data

◦ Great Britain Colonial Office. 1952. An Economic Survey of the Colonial Ter-
ritories, 1951:The Central African and High Commission Territories, Northern
Rhodesia, Nyasaland, Basutoland, Bechuanaland, and Swaziland. Volume 1.
Map of Northern Rhodesia and Nyasaland, Main Cash Products. London.

◦ Federation Of Rhodesia And Nyasaland. 1959. Report on an Economic Survey

of Nyasaland, 1958–1959. Salisbury: The Government Printer.

SM58
 ◦ Federation Of Rhodesia And Nyasaland. 1958. Report on the Agricultural and
Pastoral Production of Southern Rhodesia, Northern Rhodesia and Nyasaland,
1956–1957. Salisbury: The Government Printer.

◦ Federation Of Rhodesia And Nyasaland. 1961. Tobacco Production Map of

the Federation of Rhodesia and Nyasaland. Salisbury: Federal Department of
Trigonometrical Topographical Surveys.

Niger

• Primary commodity exports as of 1960

◦ Groundnuts (70–80%)

• Sources of administrative and spatial primary commodity production data

◦ République du Niger. Cultures Vivrières; Cultures Industrielles (unsourced map

from Library of Congress received in 1962).

◦ République du Niger. Carte Administrative: Production Agricole, Arachide

Coque (unsourced map at Library of Congress, 1964).

Nigeria

• Primary commodity exports as of 1960

◦ Palm products (26%)

◦ Cocoa (24%)

◦ Groundnuts (15%)

• Sources of administrative and spatial primary commodity production data

◦ Buchanan, Keith M., and John Charles Pugh. 1955. Land and People in Nigeria:
The Human Geography of Nigeria and Its Environmental Background. London:
University of London Press.

SM59
 ◦ Dorward, D. C. 1978. Annual Departmental Reports Relating to Nigeria and
British Cameroons. East Ardsley, UK: EP Microform.

◦ Great Britain Colonial Office. 1952. An Economic Survey of the Colonial Terri-
tories, 1951: The Gambia, the Gold Coast, Nigeria, Sierra Leone, and St. Helena.
Volume 3. London.

◦ Nigeria Cocoa Marketing Board and R. Galletti. 1956. Nigerian Cocoa Farmers:
An Economic Survey of Yoruba Cocoa Farming Families (“Cocoa Producing Areas
of the Western Region of Nigeria.”). London: Oxford University Press.

◦ Nigeria Groundnut Marketing Board. 1954. Fifth Annual Report of the Nigeria
Groundnut Marketing Board. Season 1953-54 (pp. 14–15). Lagos.

◦ Nigeria Oil Palm Produce Marketing Board. 1950. First Annual Report 1949
(pp. 18–25). Brighton, UK: King, Thorne, and Stace, Ltd.

◦ Nigeria Department of Statistics. 1955. Population census of Nigeria, 1952–53.

Lagos.

Nothern Rhodesia (Zambia)

• Primary commodity exports as of 1960

◦ Copper (90%)

• Sources of administrative and spatial primary commodity production data

◦ Northern Rhodesia Copper Industry Service Bureau. 1961. Northern Rhodesia

Chamber of Mines Yearbook. Kitwe.

◦ Northern Rhodesia Department of Geological Survey. 1959. Mineral Map of

Northern Rhodesia. Lusaka.

◦ Northern Rhodesia Census. 1951. Lusaka: Government Printer.

SM60
 ◦ Rhodesia and Nyasaland. 1959. Federation of Rhodesia and Nyasaland: Designed
and printed by the Government Printer (Scale [ca. 1:4,000,000]). Salisbury.

Ruanda-Urundi (Rwanda, Burundi)

• Primary commodity exports as of 1960

◦ Coffee (75%)

• Sources of administrative and spatial primary commodity production data

◦ Belgique. Ministère du Congo Belge et Ruanda-Urundi. 1959-1960. Bulletin

Agricole du Congo Belge et Ruanda-Urundi = Landbouwkundig Tijdschrift voor
Belgisch-Congo en Ruanda-Urundi.

◦ Belgique. Ministère du Congo Belge et du Ruanda-Urundi. 1959. La situation

économique du Congo Belge et du Ruanda-Urundi.

◦ Belgique. Office de l’information et des relations publiques pour le Congo Belge

et le Ruanda-Urundi. 1960. Ruanda-Urundi: L’Economie I-II.

◦ Union des producteurs de café du Congo Belge et du Ruanda-Urundi. 1958. Le

café au Congo Belge et au Ruanda-Urundi.

Senegal

• Primary commodity exports as of 1960

◦ Groundnuts (80%)

• Sources of administrative and spatial primary commodity production data

◦ Rossignol, P, H Borius, Suzanne Truitard, and Jean Kerhor. 1944. Les Exporta-
tions Agricoles Des Cercles De L’Afrique Occidentale Et Du Togo Français. Paris:
Direction des Affaires Economiques.

SM61
 ◦ Senegal. 1965. Cartes pour servir a l’aménagement du territoire. Ouvrage réalisé
par l’equipe de l’aménagement du territoire: Pierre Metge et al.

Sierra Leone

• Primary commodity exports as of 1960

◦ Diamonds (55%)

◦ Iron ore (18%)

• Sources of administrative and spatial primary commodity production data

◦ Clarke, John Innes. 1966. Sierra Leone in Maps. London: University of London
Press.

◦ Hall, P K. 1969. The Diamond Fields of Sierra Leone. Freetown, Sierra Leone:
Geological Survey Division.

◦ Sierra Leone. 1931. Sierra Leone report of census for the year 1931. Freetown:
Government Printer.

Somalia

• Primary commodity exports as of 1960

◦ Bananas (70%)

• Sources of administrative and spatial primary commodity production data

◦ Committee for the World Atlas of Agriculture. 1976. World Atlas of Agriculture,
Vol. 4: Africa: Commercial Farming Areas (Somalia, 1965).

◦ World Bank. 1957. The Economy of the Trust Territory of Somaliland. Wash-
ington D.C: International Bank for Reconstruction and Development.

SM62
Southern Rhodesia (Zimbabwe)

• Primary commodity exports as of 1960

◦ Tobacco (30%)

• Sources of administrative and spatial primary commodity production data

◦ Federation of Rhodesia and Nyasaland. 1961. Tobacco Production Map of the

Federation of Rhodesia and Nyasaland. Salisbury: Director of Federal Surveys.

◦ Federation Of Rhodesia And Nyasaland. 1958. Report on the Agricultural and

Pastoral Production of Southern Rhodesia. Salisbury: The Government Printer

Sudan

• Primary commodity exports as of 1960

◦ Cotton (55%)

• Sources of administrative and spatial primary commodity production data

◦ Republic of the Sudan Ministry of Economics and Finance. 1958. Sudan Cotton
Review. Khartoum.

◦ Republic of the Sudan Sudan Survey Department. 1957. Sudan Cotton. Khar-
toum.

Tanganyika (Tanzania)

• Primary commodity exports as of 1960

◦ Sisal (27%)

◦ Coffee (17%)

◦ Cotton (16%)

SM63
 • Sources of administrative and spatial primary commodity production data

◦ Guillebaud, Claude William. 1958. An Economic Survey of the Sisal Industry of

Tanganyika. Welwyn, UK: Tanganyika Sisal Growers Association.

◦ East African Statistical Department. 1953. Report on the Analysis of the Sample
Census of African Agriculture, 1950. Nairobi.

◦ Tanganyika Department of Lands & Mines. 1938. Tanganyika Territory: Sisal

Estates. Dar-es-Salaam.

◦ Tanganyika Department of Lands & Survey. 1956. Atlas of Tanganyika. Dar-es-

Salaam.

Togo

• Primary commodity exports as of 1960

◦ Cocoa (28%)

◦ Coffee (25%)

• Sources of administrative and spatial primary commodity production data

◦ Rossignol, P, H Borius, Suzanne Truitard, and Jean Kerhor. 1944. Les Exporta-
tions Agricoles Des Cercles De L’Afrique Occidentale Et Du Togo Français. Paris:
Direction des Affaires Economiques.

◦ Republic Togolaise. 1962. L’Enquete Agricole 1961/62, Resultats Provisoires.

Uganda

• Primary commodity exports as of 1960

◦ Coffee (40%)

◦ Cotton (35%)

SM64
 • Sources of administrative and spatial primary commodity production data

◦ Uganda, Department of Lands and Surveys. 1962. Atlas of Uganda (49. Coffee,
Tea, Sugar, Tobacco). Entebbe.

◦ Uganda, Department of Lands and Surveys. 1962. Atlas of Uganda (47. Cotton).
Entebbe.

◦ Uganda. 1953. Report on the census of the non-native population of Uganda

protectorate, taken on the night of the 25th February, 1948. Nairobi: W. Boyd.

◦ Uganda Protectorate. 1961. Annual Report of the Department of Agriculture,

For the Year Ended 31st December, 1960 (Appendix IV: Crop Production).

Zanzibar

• Primary commodity exports as of 1960

◦ Cloves (75%)

• Sources of administrative and spatial primary commodity production data

◦ Zanzibar Govt. Printer. 1947. Annual Report of the Agricultural Department.

◦ Sheriff, A. 1991. Zanzibar under Colonial Rule. Athens, OH: Ohio University
Press.

◦ Zanzibar Blue Book. 1946.

65