Top Banner
Economic Redistribution and Ethnic Conflict: Why European Settler Colonies Fought Decolonization Wars Jack Paine * April 30, 2016 Abstract During the post-1945 decolonization era, 15 colonies fought major independence wars against their European colonizer. This paper argues that tensions over land redistribution in Eu- ropean settler colonies was an important cause of decolonization violence, focusing on Africa. In settler colonies, Europeans monopolized the best agricultural land and blocked political re- forms that could have alleviated grievances among Africans. Statistical evidence links larger European settler population shares to a robustly higher propensity for decolonization violence in Africa. This result holds whether controlling for a wide range of pre-colonial, colonial, or post-control covariates proposed in the literature, or instrumenting for climatic factors that impacted where Europeans could settle. * Post-doctoral Associate, Wallis Institute of Political Economy, University of Rochester, [email protected]. I thank Scott Abramson, Gretchen Helmke, Bethany Lacina, Alex Lee, Peter Lorentzen, Aila Matanock, Bing Powell, and Noam Yuchtman for helpful comments on earlier drafts. 1
41

Economic Redistribution and Ethnic Conflict: Why European ...

Nov 19, 2021

Download

Documents

dariahiddleston
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Economic Redistribution and Ethnic Conflict: Why European ...

Economic Redistribution and Ethnic Conflict:Why European Settler Colonies Fought Decolonization Wars

Jack Paine*

April 30, 2016

Abstract

During the post-1945 decolonization era, 15 colonies fought major independence warsagainst their European colonizer. This paper argues that tensions over land redistribution in Eu-ropean settler colonies was an important cause of decolonization violence, focusing on Africa.In settler colonies, Europeans monopolized the best agricultural land and blocked political re-forms that could have alleviated grievances among Africans. Statistical evidence links largerEuropean settler population shares to a robustly higher propensity for decolonization violencein Africa. This result holds whether controlling for a wide range of pre-colonial, colonial,or post-control covariates proposed in the literature, or instrumenting for climatic factors thatimpacted where Europeans could settle.

*Post-doctoral Associate, Wallis Institute of Political Economy, University of Rochester, [email protected] thank Scott Abramson, Gretchen Helmke, Bethany Lacina, Alex Lee, Peter Lorentzen, Aila Matanock, Bing Powell,and Noam Yuchtman for helpful comments on earlier drafts.

1

Page 2: Economic Redistribution and Ethnic Conflict: Why European ...

1 Introduction

We have little understanding of an important category of wars in the 20th century: conflict between

colonies and European imperialist powers during the process of decolonization. Although it is

perhaps remarkable that most colonies gained independence peacefully (Abernethy 2002, 325),

15 major decolonization wars occurred after 1945. These conflicts resulted in a total death toll of

nearly 1 million and exhibited a median casualty rate of 15,000. Among all decolonization wars and

post-independence civil wars since 1945, 18 percent of conflicts were fought against Europeans.1

Why did bargaining between Europeans and non-Europeans often result in costly violent conflict

rather than peaceful extrication?

Despite a voluminous literature on causes of civil conflict, most quantitative political science re-

search either omits decolonization wars entirely, or, more rarely, considers them to be a distinct

category of civil war (e.g., Fearon and Laitin 2003, 76-7). Scholars studying international war-

fare similarly have relatively little to say about decolonization wars. Nor have major theories of

violence or of political transitions been applied to studying decolonization conflicts. However,

the relative frequency and destructiveness of these wars—as well as their posited effects on post-

colonial outcomes such as durable authoritarian regimes (Huntington 1968; Levitsky and Way

2013) and internationalized civil wars (Roessler and Verhoeven 2015)—makes this an important

topic to study.2

Many historians have suggested that sizable European settler minority populations impeded re-

forms such as majority rule or full independence that could have prevented decolonization vio-

lence (e.g., Ajayi and Crowder 1985, section 64; Young 1994, 239-40). For example, Good (1976,1Figures calculated using Fearon and Laitin’s (2003) updated dataset, which codes conflicts that result in at least

1,000 battle deaths. The post-independence civil war tally only includes countries that have gained independence froma Western European power since 1945.

2Goldsmith and He’s (2008) article and Stein’s (2010) contemporaneous work-in-progress provide exceptions bystatistically assessing causes of decolonization wars.

2

Page 3: Economic Redistribution and Ethnic Conflict: Why European ...

713) states unequivocally: “Through its rigidity and wide-ranging intransigence, aided by the non-

intervention of the metropole, the [European] settler state itself ensures that the new African classes

have no way forward other than by a revolutionary path.” This paper argues and presents statistical

evidence that European settlers caused decolonization violence in Africa, the most relevant empiri-

cal setting for testing the theoretical implications. Whereas African colonies varied considerably in

their European settlement volume, almost no non-African post-World War II decolonization case

contained a significant European population.

Theoretically, tensions over economic redistribution between European settlers and Africans in-

creased incentives for violence. In settler colonies, Europeans monopolized the best agricultural

land and feared economic redistribution if the African majority gained control over policy. Fur-

thermore, because European settlers usually exerted strong political influence within the metropole

country, they successfully blocked concessions that could have alleviated Africans’ grievances.

This contrasted with non-settler colonies, where weaker vested economic interests did not impede

decolonization reforms. The theoretical framework situates the broader process of decolonization

negotiations within existing prominent theories of redistributive political transitions (Acemoglu

and Robinson 2006; Boix 2003) and of ethnic grievances and ethnic conflict (Cederman et al.

2013).

Empirical evidence from Africa strongly supports these theoretical implications. There is an ex-

tremely strong correlation between European percentage of the colonial population and a major de-

colonization war. Hypothetically increasing a colony’s European settler percentage from Ghana’s

one-tenth of one percent to Rhodesia’s level of roughly six percent increases the predicted probabil-

ity of violent decolonization from five percent to 78 percent. This correlation remains substantively

large and robustly statistically significant when adding sets of controls that capture pre-colonial de-

velopment, pre-colonial violence, geography, colonial factors, or standard covariates from the civil

war literature. The results are also robust to extensive sensitivity analysis: extreme bounds analy-

3

Page 4: Economic Redistribution and Ethnic Conflict: Why European ...

sis on the covariates, sample alterations, and using selection on observables to estimate bias from

unobservables.

A second set of statistical results reinforce these findings using instrumental variable analysis. I

used GIS data to construct a variable based on climate, rainfall, elevation, and tsetse fly conditions—

factors that strongly predict areas of Africa in which Europeans settled in large numbers. Regres-

sions using this variable as an instrument for European settlement generate a similarly substantively

large and robust correlation between European settlement and violent decolonization, and are ro-

bust to fairly large possible violations of the exclusion restriction.

2 Economic Redistribution and Ethnic Conflict: Existing Theories

This section reviews two major theoretical frameworks in political science for political transitions

or for violence—redistributive political transitions, and ethnic grievances and civil war—and pre-

views their potential implications for studying decolonization in Africa.

2.1 Redistributive Political Transition (RPT) Theories

Prominent political science theories study the relationship between economic redistribution and

political regime transitions (Acemoglu and Robinson 2006; Boix 2003), which I refer to as re-

distributive political transition (RPT) theories. Acemoglu and Robinson assume that class is an

important political cleavage and study an interaction between a small group of economic elites and

the majority group, called the masses. Elites are assumed to have higher income levels than the

masses. The governing actor can choose a tax rate that funds public goods. Because each group

is assumed to pay the same tax rate, higher taxes disproportionately affect the elites and, in effect,

redistribute income from the elite to the masses. Elites prefer a dictatorial regime—in which they

4

Page 5: Economic Redistribution and Ethnic Conflict: Why European ...

set the tax rate, and can therefore minimize redistribution—to a democracy in which policy is set

by the majority, who prefer a higher tax rate than do the rich. Democracy enables the masses to

redistribute at high rates.

Higher income inequality increases the difference in the amount of redistribution under a democ-

racy relative to a dictatorship. Because this raises the intensity of each actors’ preference for their

favored regime, an increase in income inequality heightens both actors’ incentives to violently in-

fluence a political transition. Specifically, assuming the present regime is a dictatorship, higher

income inequality creates stronger incentives for elites to repress the masses rather than to grant

democracy and acquiesce to high redistribution, and also raises incentives for the masses to launch

a revolution.

Not all types of economic activity cause fighting, however. One possible mitigating factor against

the economic inequality-violence relationship is high asset mobility (Boix 2003). Suppose inequal-

ity is high. If elites have highly mobile assets, then their threat to relocate their assets in reaction

to a high tax rate limits the amount of redistribution that would occur in a democracy. This reduces

elites’ incentives to use violence to prevent democratization. By contrast, immobile assets such as

land should trigger violence when distributed unequally.

Finally, actors may not be able to successfully bargain away their disagreements over the tax rate

because the player that controls the government—elites in dictatorship, masses in democracy—

cannot commit to implement a tax rate other than their most-preferred amount in most periods. In

the rare periods that the out-of-power actor can coerce the governing actor, higher inequality makes

it more difficult for the government to offer a palatable policy for the other actor. This is because

the out-of-power actor will frequently face an unfavorable tax rate in future periods. Overall, RPT

theories yield the key hypothesis:

5

Page 6: Economic Redistribution and Ethnic Conflict: Why European ...

Main RPT implication. If class is an important political cleavage, then more unequaldistribution of non-mobile economic assets increases tension over economic redis-tribution, which in turn raises incentives for political violence to influence politicaltransitions (additionally assuming that groups cannot commit to implement tax ratesother than their most-preferred amount).

2.2 Ethnic Grievances and Civil War (EGCW) Theories

A distinct literature argues that ethnic grievances trigger civil war (Cederman et al. 2010; Roessler

2011; Cederman et al. 2013; Wucherpfennig et al. 2015), which I refer to as ethnic grievance and

civil war (EGCW) theories. EGCW theories contend that the spread of nationalism to the colonial

world has caused ethnic cleavages to strongly influence access to political power at the center

in much of the post-colonial world. Choices regarding whom to include the governing coalition

have often followed ethnic lines, and excluded ethnic groups harbor grievances over unfavorable

political access. Excluded ethnic groups face heightened incentives to initiate a civil war against the

government. These authors have also compiled a novel dataset on politically relevant ethnic groups

and have empirically demonstrated that ethnic groups excluded from political power do indeed

fight civil wars at elevated rates. Therefore, EGCW research implies and empirically supports the

key hypothesis:

Main EGCW implication. Asymmetric access to political power at the center amongethnic groups increases incentives for political violence to control the state.

2.3 Applying the Theoretical Frameworks to Decolonization in Africa

Although disparate research traditions have produced these two theoretical frameworks, and neither

was constructed to study decolonization in Africa, they can be fruitfully combined to engender

implications for this substantive setting. RPT theories have mainly been substantively applied

to study transitions between dictatorship and democracy (and vice versa) in independent states.

6

Page 7: Economic Redistribution and Ethnic Conflict: Why European ...

However, the theoretical logic is potentially applicable to other types of political transitions such

as legislating majority rule under colonialism, or granting full independence. In Acemoglu and

Robinson (2006), the median voter in society (who belongs to the masses) chooses policy in a

democracy, but one could easily strip away the term “democracy” and instead simply call it a model

of transitions between majority and minority rule—as is appropriate for the present setting.

Finding new empirical domains for RPT theories is especially crucial when considering recent

empirical evidence demonstrating that these theories have minimal explanatory power for transi-

tions between dictatorships and democracy in the post-colonial world (among others, Haggard and

Kaufman 2012; Slater et al. 2014). A likely reason for this dissenting evidence is that few post-

colonial cases match a key scope condition from RPT models: the class-based rich-versus-poor

cleavage composes a key political cleavage. However, as argued below, African colonies—in par-

ticular settler colonies—do closely match this key scope condition. This fact makes these models

an appropriate framework for studying decolonization—intriguingly, perhaps a more substantively

relevant setting than the post-colonial cases that most existing empirical work on RPT theories has

examined.

Although EGCW theories focus primarily on grievances arising from asymmetric ethnic access to

political power at the center, their core ideas complement the RPT framework if ethnicity com-

poses a key political cleavage and if different ethnic groups exhibit stark economic disparities.

Rather than the spread of nationalism causing divisions across ethnic groups—EGCW theories’

focus—African colonies featured the even more concrete ethnic, or racial, distinction between na-

tive Africans and Europeans. This distinction was particularly sharp in settler colonies with sizable

European settlements, most of whom migrated to Africa in the late 19th or 20th century. And be-

cause Europeans dominated the political arena and excluded Africans from power prior to World

War II in all African colonies, the sharp economic inequality between Europeans and Africans in

settler colonies paralleled political inequalities.

7

Page 8: Economic Redistribution and Ethnic Conflict: Why European ...

EGCW theories become even more powerful when combined the logic of RPT theories because

the latter address an important strategic question. If asymmetric ethnic access to power causes

costly fighting, then why not expand political access at the center?3 The logic above explains why.

Suppose a small but dominant racial group controls both the political arena and a disproportionate

share of the country’s assets, and these assets are non-mobile. The government may prefer to

experience violence rather than to prevent fighting by granting majority rule—which would enable

the disprivileged racial group to redistribute their assets (assuming the out-of-power masses cannot

commit to limit redistribution if they take power).

A final relevant consideration is that one assumption in Acemoglu and Robinson’s (2006) core

model, that all elites have the same economic preferences, may be relaxed without qualitatively

altering the model’s implications. In the decolonization context, some elites (i.e., Europeans) had

stronger economic preferences for continued colonial rule than others. For example, politicians in

London or Paris frequently valued the colonies differently than did white farmers or white business

owners in Africa. Divisions among elites, however, can easily be incorporated into the RPT frame-

work if we additionally assume that different elite groups have varying levels of political influence,

and that elites’ aggregate choices are made by weighting the relative influence of different groups.

Therefore, if settlers are very politically influential in a particular colony, it is as if they are the ones

making the decision to expand the franchise or not even though other European actors may have

different preferences.

In sum, the core tenets of RPT and EGCW theories appear appropriate for studying decolonization

in Africa, and in particular for comparing prospects for violence in settler and non-settler colonies.

It remains to be demonstrated that settler colonies in Africa did indeed closely match the key scope

conditions from these theories in which violence occurs—as opposed to the relative absence of3Roessler (2011) provides a different modification of the EGCW framework to address this question in post-

independence Africa.

8

Page 9: Economic Redistribution and Ethnic Conflict: Why European ...

such conditions in non-settler colonies.

3 Divergent Reform Strategies in Settler and Non-Settler Colonies

Africans had very few rights or political representation in almost all colonies immediately after

World War II (Mamdani 1996; Collier 1982, 34-44).4 Changes during the war and in the following

decade, however, fundamentally altered the colonial calculus. Scholars have posited a wide range

of contributing factors. These include a shift in international opinion against colonialism driven

by the anti-colonial stances of the United States and Soviet Union (Young 1970), increased mobi-

lization ability of Africans after the war (Adu Boahen 1970), and decreased economic returns to

colonialism (Colgan 2015). European colonizers’ general response to these changes was to grant

reforms rather than to face rebellions. However, European settlers had strong economic interests

against majority rule and/or independence because land—an easily expropriated asset—was highly

unequally distributed and, under majority rule, Africans could not to commit to refrain from large-

scale redistribution away from a small minority group. Furthermore, European settlers successfully

aligned policy with their economic interests either by controlling the state themselves or by com-

posing a highly influential lobby in metropolitan politics. The logic of RPT and EGCW theories

implies that these conditions should cause decolonization violence.

The qualitative arguments for European settler colonies draw from the four major ones in Africa

(e.g., Mosley 1983, 1; Lutzelschwab 2013): South Africa, Southern Rhodesia/Zimbabwe, Algeria,

and Kenya. Notably, whereas questions of independence and majority rule were closely inter-

twined in Algeria and Kenya—as in almost every other African country—South Africa’s indepen-

dence in 1910 and Rhodesia’s quasi-independence in 1965 implied that their respective struggles4Partial exceptions include several colonies that had experienced limited elections (Collier 1982, 34-44) and a handful

of islands that enjoyed the same rights as metropolitan citizens (Owolabi 2015, Appendix I).

9

Page 10: Economic Redistribution and Ethnic Conflict: Why European ...

only concerned majority rule. However, I still conceptualize these conflicts (as well as South

West Africa/Namibia’s) as composing part of the decolonization period because the struggles pit-

ted Africans against direct colonial remnants.

An important caveat in the following discussion is that it does not, of course, suggest that land

inequality was the only cause of decolonization violence in Africa. Instead, it simply seeks to es-

tablish logic for why land inequality should have provided one important contributing factor.

3.1 Historical Background: Divergent Reform Patterns

Presented with changed circumstances, European powers faced a choice between granting reforms—

which eventually required allowing full independence—or potentially losing control over the pace

of reforms to armed guerrilla movements. This section briefly summarizes divergent reform pat-

terns in settler and non-settler colonies.

In almost all non-settler colonies, reform pre-empted rebellions. In France’s mainland Sub-Saharan

African colonies, reforms proceeded uniformly. Between 1945 and 1946, every colony began

electing delegates to the French Constituent Assembly, the French National Assembly, and the

new territorial assembly. The franchise expanded incrementally over the next decade until the Loi

Cadre in 1956, which granted universal suffrage. A constitutional referendum across the colonies

in 1958 for the first time introduced the possibility of secession, and all of mainland French Sub-

Saharan Africa became independent in 1960 after two years of de facto autonomous governance.

Britain pursued a decolonization strategy more tailored to the development of individual colonies,

which explains the differential timing of initial elections, first universal suffrage elections, and date

of independence across its African colonies. But it followed the same general thrust of reform:

“Power was devolved in an orderly series of steps, by which representative political institutions

were built around the bureaucratic core” (Young 1970, 488). Belgium hastily withdrew from its

10

Page 11: Economic Redistribution and Ethnic Conflict: Why European ...

central African colonies after a riot in Leopoldville in 1959 but also followed the general pattern

of granting reforms to prevent anti-colonial rebellions.

By contrast, settler colonies experienced minimal, or even negative, reforms. In South Africa, the

South Africa Act of 1909 initialized a series of racially exclusionary laws. These culminated with

the infamous apartheid state in 1948 that restricted where Africans could live and work and denied

any African political representation. Africans in South Rhodesia were offered a new constitution

in 1961 that offered a minimal compromise by granting 15 out of the 65 parliamentary seats to the

African majority (Encyclopaedia Britannica 2016c). However, after African leaders rejected the

new constitution, hardline policies by Europeans rejected any form of compromise. In Algeria, the

Blum-Violette Bill of 1936 offered French citizenship to a small percentage of the Muslim pop-

ulation without requiring them to renounce Islamic personal law (Lawrence 2013, 80). However,

this bill failed. Europeans also swept the rigged 1948 elections for the Algerian Assembly, which

denied representation to the nationalist movement (Toth 1994, “Polarization and Politicization”;

Spruyt 2005, 105).

3.2 Applying RPT and EGCW Theories to Explain Divergent Reform Patterns

Settler colonies possessed three attributes that match the scope conditions under which RPT and

EGCW theories imply that reforms should not occur. First, class/race was the main political cleav-

age. Second, settlers had strong economic preferences against majority rule. Third, settlers were

highly politically influential. The second and third conditions contrasted with the lack of politically

influential groups in non-settler colonies that rejected decolonization.

Europeans vs. Africans composed the main political cleavage. For RPT theories to be relevant,

class must compose an important political cleavage. For EGCW theories theories to be relevant,

ethnicity/race must compose an important political cleavage. The European-African race cleavage

11

Page 12: Economic Redistribution and Ethnic Conflict: Why European ...

was prominent in all African colonies, but was particularly pronounced—and coincided with the

class cleavage—when there were a large number of European settlers that dominated society, in

contrast to indirect colonial rule in most of the rest of Africa.

Supporting this claim, Gann and Duignan (1962, 142) offer the general view that in settler colonies,

“Class conflicts are overlaid and reinforced by racial differences.” Similarly, Good (1976, 611-2)

states, “the white community has a common and overriding interest in continuing the exploitation

of the black majority.” Providing examples, Southern Rhodesia featured a “dualistic society” in

which racial and economic cleavages coincided (Spiro 1963, 384). In Kenya, the “wealthy, ex-

patriate, white landowning elite . . . conspicuously dominat[ed] the African societies among whom

they dwelled.” This “spawned their antithesis: the conscious rural African masses aware of their

disadvantaged position in society” (Wasserman 1976, 2).

Although there were also differences within the white community, European settlers consciously

chose to emphasize racial over other differences. South Africa, for example, gained independence

in 1910 as a conglomeration of four previously separate white settler colonies, two of which were

English-speaking and two Afrikaans-speaking. Whites in South Africa therefore exhibited divisive

regional identities, in addition to their racial distinction from the African majority. Lieberman

(2003) documents how Europeans explicitly chose to enshrine the racial differences in the South

Africa Act of 1909—the founding constitution—and to minimize regional distinctions.

Settlers feared land redistribution. European settlers had strong economic interests against ma-

jority rule and/or independence for two main reasons: (1) highly unequal distribution of an easily

expropriated asset, land, and (2) under majority rule, Africans’ inability to commit to refrain from

large-scale land redistribution away from a small minority group.

First, European settlers tended to control a percentage of land far exceeding their population share.

The immobility of Europeans’ land and the ease of replacing Europeans with Africans to produce

12

Page 13: Economic Redistribution and Ethnic Conflict: Why European ...

agricultural output meant that a majority African government could easily redistribute away from

Europeans. This contrasts with other types of economic assets that can be moved in response to

redistributive pressures (Boix 2003) or that require greater bureaucratic capacity to redistribute

(Slater et al. 2014). For European settlers after World War II, Kahler (1981, 391) states: “The

probable political future of settler agriculture was not promising. Land and land hunger were the

driving forces behind nationalist movements from Rhodesia to Kenya to North Africa. The colonial

state had carefully prepared the way for European agriculture; any successor regime was likely to

threaten its property rights first. Few technological or other obstacles would prevent a successor

government from substituting African or Arab farmers for Europeans.” Table 1 summaries starkly

unequal land distribution patterns in the four main settler colonies.

Table 1. European Settler Economic Domination - Eve of World War II

Colony/country Eu. settler% of population

Eu. settler% alienated land

Eu. settler% cultivable land

South Africa 20.9 87 61Algeria 12.8 34 27Southern Rhodesia 4.6 50 58Kenya 0.6 7 25

Source: Lutzelschwab (2013, Tables 5.1 and 5.2). Land figures for Algeria exclude the Sahara.

Second, under majority rule, European settlers did not believe that Africans could commit to refrain

from large-scale redistribution away from a small minority group. If instead Africans could have

committed to low redistribution under majority rule, then this could have decreased incentives for

Europeans to adopt an intransigent stance toward reforms.

As one example, two historians (one of whom was Rhodesian) characterized the economic aspect

of Rhodesia’s power struggle. “Africans wish to gain control of the state; they are being resisted

by European groups who fear the perils of rapid deterioration of standards, or possible collapse,

and the dangers of minority status” (Gann and Duignan 1962, 142). Furthermore, “Should power

fall into African hands in Rhodesia, settlers fear the new rulers would insist on the expropriation of

13

Page 14: Economic Redistribution and Ethnic Conflict: Why European ...

white farms in the name of land reform” (Gann and Duignan 1970, 161).

Providing additional examples, Europeans in Algeria enjoyed relative prosperity that they would

not be able to duplicate if they returned to France, due to their low education levels, nor if they re-

mained in an independent Algeria (Spruyt 2005, 105). Wasserman (1976, 2) characterizes whites’

land domination in the Kenya Highlands as “the central economic issue” between Europeans and

Africans. “The fate of the ‘White Highlands,’ was the linchpin determining the future of the Eu-

ropean farming community and the colonial political economy. The European farmer holding

non-liquid assets in a threatening environment had to adapt in some way to his surroundings—if

only by leaving.” South Africa’s imposition of apartheid laws in 1948 also in part reflected a desire

for whites to maintain their economic dominance. According to Posel (2011, 321-322):

“As had been the case during the segregationism of the preceding decades, the apartheidproject was an attempt to sustain white political supremacy in ways that simultane-ously promoted the cause of white economic prosperity. Interventions to buttress thewhite racist regime, and the attendant political divisions and contestations, were there-fore inseparable from efforts to sustain and regulate practices of capital accumulationand economic growth with the class interests and conflicts associated with those.”

The stated redistributive goals of various African rebel organizations further support the plausibil-

ity of Europeans’ redistributive fears. In Rhodesia, “one of the strongest motivations for African

nationalists taking up arms was to win back the land that had been expropriated by the colonial set-

tlers” (Mlambo 2014, 220-221). Land reform negotiations composed a crucial part of the Lancaster

House Agreement of 1979 that yielded internationally recognized independence and majority rule

in Zimbabwe (191-3). In Algeria, the Soumman Declaration of 1956—the culmination of a foun-

dational meeting for the revolutionary group FLN—stated the need for agrarian reform and land

distribution as part of FLN’s broader independence goals (Kahler 1981, 391). In Kenya, histori-

ans and political scientists frequently cite Kikuyus’ belief that Europeans took their land as the

primary trigger of the Mau Mau rebellion (Wasserman 1976, 2; Decalo 1997, 188; Lutzelschwab

14

Page 15: Economic Redistribution and Ethnic Conflict: Why European ...

2013, 162).

Settlers were highly politically influential. European settlers’ control of the state—as in South

Africa and Southern Rhodesia—or heavy influence over metropole politics, as in Algeria and to

a lesser extent in Kenya—was crucial for blocking economic reforms that could have alleviated

Africans’ grievances. And although settlers were not the only contingent of the European elite that

decided policy in these colonies, they were highly politically influential. Therefore, their actions

can be directly linked to thwarted reforms.

South African whites governed a sovereign state and therefore had wide leeway to pursue policies

that enhanced their economic interests. Their choices directly engendered the discriminatory South

Africa Act of 1909 and apartheid policies. Although not governing a fully independent state, South-

ern Rhodesian whites had enjoyed self-governance since 1923 and the government pandered solely

to European interests (Spiro 1963, 366). After African leaders rejected the proposed constitution in

1961, the white electorate voted in a new right-wing party that was “committed to the maintenance

of white rule in the country” (Oliver and Atmore 2005, 272). The settler government unilaterally

declared independence in 1965 to preserve white rule, although Rhodesia’s independence was not

internationally recognized.

Although French settlers in Algeria did not control the colonial government, they did command

considerable influence over the colonial government and in Paris (Spruyt 2005). The colons “had

powerful allies in the [French] National Assembly, the bureaucracy, the armed forces, and the

business community, and were strengthened in their resistance by their almost total control of the

Algerian administration and police” (Toth 1994, “Political Movements”). Their influence under-

mined the Blum-Violette Bill of 1936 (Lawrence 2013, 80) and settlers rigged the 1948 elections

for the Algerian Assembly (Toth 1994, “Polarization and Politicization”; Spruyt 2005, 105).

Europeans in Kenya also exerted considerable influence. Good (1976, 613) quotes a historian who

15

Page 16: Economic Redistribution and Ethnic Conflict: Why European ...

claims that in Kenya, “the British administration did in fact follow a policy which almost invariably

allowed the interests of settlers to prevail at all the critical points where they conflicted with those

of the African population.” Unwilling to consider any degree of reforms before the Mau Mau

rebellion began in 1952, “when James Griffiths, Colonial Secretary in the reformist Attlee Labour

Government, visited Kenya in 1951, the settlers ‘indicated that any [political] changes imposed on

us against our wishes would be resisted, even to the extent of unconstitutional action”’ (611). Only

after the Mau Mau rebellion did the British government take more direct control of the colony—

despite considerable protest by whites—and allow it to become “part of Black Africa” (Gann and

Duignan 1962, 136).

Economic incentives for decolonization in non-settler colonies. By contrast, actors with strong

economic incentives to maintain colonial rule were not politically strong in non-settler colonies.

Most important, after World War II, colonies did not tend to economically benefit the metropolitan

country. Some argued that the colonies were necessary for economic recovery after the war, in

particular for providing a source of hard currency and natural resources. However, these economic

benefits were transient. Britain’s official historian of colonial development proclaimed with regard

to African decolonization:

“The economic considerations were fairly evenly matched [because, while Britainmight save on some types of expenditure, there might be costs resulting from reductionof special commercial advantages it enjoyed in the colonies]. Consequently, it was feltthat the economic interests in the United Kingdom were unlikely in themselves to bedecisive in determining whether or not a territory should become independent” (quotedin Fieldhouse 1986, 8).

Similarly for France, the huge subsidies it granted to its colonies undermined economic incen-

tives for continued colonial rule. By the 1950s, this economic reality had convinced many French

officials that autarkic assumptions about benefits of trading within the empire were flawed (Field-

house 1986, 14-7). These economic realities reflected the changed post-World War II international

16

Page 17: Economic Redistribution and Ethnic Conflict: Why European ...

economic system that made continued colonial rule unprofitable (Spruyt 2005, 65-86), especially

considering the alternative of facing anti-colonial rebellions.

Non-settler colonies also did not possess strong business lobby groups that rejected decolonization

reforms. This is important to consider because elites do not act as a unified bloc, and therefore

the best interest of the metropole as a whole does not necessarily prevail. Larger firms operating

in more modern industries relied less on colonial protectionism and therefore were more accepting

of African majority rule (Kahler 1981; Fieldhouse 1986, 9-12, 17-21; Spruyt 2005, 101-4, 124-

7). Many recognized the benefits of establishing a moderate nationalist elite to work with after

independence rather than potentially letting a guerrilla group take power following a prolonged

struggle (Spruyt 2005, 127). Fieldhouse (1986) explicitly contrasts the position of big businesses

operating in Africa—which believed they could adapt production to changed circumstances—with

smallholding white settlers in both British and French colonies (11-2, 18).

Although certain businesses were strongly pro-empire, the circumstances just described explain

why a business did not pose a unified front against decolonization. Some companies clamored

that decolonization would diminish Europeans’ external markets. Owners of fixed assets such

as railroads, utility companies, and mines all opposed reforms (Spruyt 2005, 125), as did pro-

tected French firms producing internationally uncompetitive products (Fieldhouse 1986, 17). This

split follows the economic logic proposed above: differences in prospective profits under Eu-

ropean versus African rule shaped a group’s stance toward decolonization. However, crucially,

pro-empire business lobbies were not politically powerful in either Britain or France (Fieldhouse

1986, 9-12, 17-21; Spruyt 2005, 101-4, 124-7).5 This may in part have resulted from the changed

post-World War II structure of European economies toward sectors involving motorization—which

requires less government protection that older industries to be profitable—as argued by Colgan5Spruyt (2005) titles his section on business’ reaction to decolonization in France, “Business Stands on the Sidelines”

(101) and in Britain, “The Heterogeneous Attitudes of the Business Community” (124).

17

Page 18: Economic Redistribution and Ethnic Conflict: Why European ...

(2015).

A final clarifying note for situating non-settler colonies into the RPT framework is that viewed from

a different perspective, they might seem ripe for violent political transitions. Economic inequality

was, of course, high even in non-settler colonies because European companies tended to control the

most valuable assets. However, as just discussed, colonies also entailed high governance costs that

mitigated the metropole’s benefits from European-controlled assets. Furthermore, many European

businesses operating in Africa could expect to profit in African-controlled regimes, at least relative

to the alternative of enduring armed rebellion. This highlights the importance of focusing on the

source of economic inequality and incorporating Boix’s (2003) consideration that it is crucial to

examine the extent of asset mobility or, more generally, the expected level of expropriation under

different political regimes.

3.3 Main Empirical Implication for Violent Decolonization

Combining the logic of RPT and of EGCW theories with empirical differences across colonies im-

plies that European settler colonies should have experienced violent decolonization more frequently

than non-settler colonies. Circumstances in non-settler colonies created incentives to grant politi-

cal reforms, which eventually culminated in universal suffrage elections and full independence. By

contrast, European settlers faced stronger economic incentives to thwart political reforms, which

triggered violence. This logic yields the main hypothesis to be tested:

Main hypothesis. Larger European settler populations should increase the propensityof decolonization violence.

18

Page 19: Economic Redistribution and Ethnic Conflict: Why European ...

4 Sample and Data

4.1 Sample

The sample includes every mainland African country (including North Africa), plus Madagascar,

that was colonized by a European country and gained independence after 1945. The only ex-

ceptions are former Italian colonies Libya and Somalia. Italy’s defeat in World War II placed

these colonies under temporary occupation by other European powers and eliminated any possible

linkage between colonial-era European settlers and violent secession from the colonizer. Indeed,

almost all of Libya’s fairly large Italian population had emigrated by the end of World War II

(Encyclopaedia Britannica 2016a). Additionally, because granting the franchise to the African ma-

jority was central to the process of decolonization, I consider the decolonization era to last until

1994 in South Africa despite its independence since 1910, until 1979 in Zimbabwe despite its in-

ternationally disputed independence since 1965, and until 1990 in Namibia because South Africa

effectively colonized this League of Nations Mandate territory after World War I.

4.2 Main Explanatory Variable: European Population Share

The European population share data uses three sources that estimate European settlers as a per-

centage of the population. Lawrence (2010) provides a data point for each French colony between

1946 and 1950, Mosley (1983) for southern British colonies and several others in 1960, and United

Nations (1965) for various colonies for up to three years ranging from 1946 to 1961.6 This yields

at least one data point for all but four countries in the sample. I consulted additional secondary

sources for these four countries that justified coding 0 settlers for each of them because none men-

tioned a European settler population. This coding rule follows Easterly and Levine (2012, 9), who6The latter two sources were identified using the replication data for Easterly and Levine (2012).

19

Page 20: Economic Redistribution and Ethnic Conflict: Why European ...

argue: “colonial histories (which are virtually all written by European historians) are extremely un-

likely to fail to mention significant European settlements.” For colonies in which multiple sources

provided a European settlers estimate, I average over the estimates. To prevent observations with

extreme values from driving the statistical results, the regressions use the natural log of European

settlers as a percentage of the population.

4.3 Dependent Variable: Violent Decolonization

A major decolonization war is conceptualized as a war fought for the purpose of gaining indepen-

dence from a colonial power that causes at least 1,000 deaths. The cases coded as major decolo-

nization wars for the statistical analysis below are Algeria, Angola, Cameroon, Kenya, Malagasy

(Madagascar), Morocco, Mozambique, Portuguese Guinea (Guinea-Bissau), South Africa, South-

ern Rhodesia (Zimbabwe), South-West Africa (Namibia), and Tunisia. This closely resembles

Fearon and Laitin’s (2003) list of wars fought over territories colonized by a Western European

power since World War II, with the following two differences. Namibia is coded as a decoloniza-

tion war because it meets Fearon and Laitin’s 1,000 death threshold (Lacina 2009, 397-8), whereas

Rwanda is not because the violence occurred between rival African ethnic groups prior to inde-

pendence rather than between Africans and Europeans to achieve independence (Encyclopaedia

Britannica 2016b). The list of decolonization conflicts used here is also nearly identical to cases

identified in the UCDP/PRIO armed conflict database (Gleditsch et al. 2002) that involved a Eu-

ropean country fighting in one of its colonies. They additionally include a minor decolonization

conflict fought across Mauritania, Morocco, and Western Sahara.

20

Page 21: Economic Redistribution and Ethnic Conflict: Why European ...

4.4 Alternative Explanations for Violent Decolonization

Although little statistical research has examined causes of decolonization wars specifically, large

literatures on causes of civil war and on historical legacies in Africa provide insight into possible

confounders for the European settlers-violent decolonization relationship.

Pre-colonial development. Although the most consistent finding in the civil war literature is that

richer countries are less likely to experience internal warfare, the relationship might go in the

opposite direction for decolonization wars. More developed societies might have been more able

to organize and to effectively challenge colonial rule after World War II. I measure pre-colonial

development with two different variables. The first is the percentage of each colony’s population

that belonged to “large paramount chiefdoms/small states” or “large states” prior to colonization.

Gennaioli and Rainer (2007) code this centralization variable using data from Murdock (1967).

The second is the years elapsed since a territory transitioned to agricultural production, which is

strongly correlated with modern development (Putterman 2008).

Pre-colonial violence. Nunn (2008) argues that the most significant interactions between Africans

and Europeans prior to colonization involved slave trading. Considering strong evidence linking a

country’s slave exports to negative contemporary economic and behavioral outcomes (Nunn 2008;

Nunn and Wantchekon 2011), colonies with more slave exports might have been more hostile to

European colonial rule. It is also possible that territories with more historical instances of warfare

would be more likely to experience conflict during decolonization. Besley and Reynol-Querol

(2014) demonstrate a strong positive correlation in Africa between a territory’s number of wars

from 1400 and 1700 and the probability of civil war in the post-colonial era.

Geography of rebellion. Another set of covariates captures geographic factors that existing research

argues affects rebels’ opportunities to combat and hide from the state. These include area and the

percentage of mountainous terrain (Fearon and Laitin 2003), the percentage of rugged territory

21

Page 22: Economic Redistribution and Ethnic Conflict: Why European ...

(Nunn and Puga 2011; Wantchekon and Ponce-Leon 2014), and a categorical variable that incor-

porates both the size of the territory and the number of population clusters to provide an aggregate

measure of the ability to broadcast power over the territory (Herbst 2000).

Colonial factors. The identity of the colonizer could influence prospects for violent decolonization.

Many scholars have argued that Britain had a relatively coherent plan for decolonization and more

flexible institutions than did the other colonizers, which diminished prospects for decolonization

violence (Young 1970, 488). This contrasted with Fourth Republic France’s weak parliamentary

system that enabled capture by special interest groups (Spruyt 2005). Decolonization in Portuguese

colonies may be linked to hardline policies associated with authoritarian regime survival (Wilson

1994, 181). I account for these possibilities by controlling for a full set of colonizer fixed effects:

British, French, Portuguese, Belgian, and Spanish. This leaves Namibia, the ex-German colony

ruled by South Africa from World War I onward, as the excluded basis category (and therefore, is

nearly equivalent to dropping Namibia from regressions that include colonizer fixed effects).

Related, Goldsmith and He (2008) argue that the strength of constraints on the executive in the

metropole impacts violent decolonization prospects. Because of collinearity with the colonizer

dummies, I account for this argument by also considering an alternative colonial-era specification

that controls for the colonizer’s xconst score in 1946 from the Polity dataset (Marshall and Gurr

2014) rather than the colonizer dummies. Regarding another potentially important colonial factor,

Lawrence (2013, 132-65) argues that in territories that were invaded or occupied during World

War II, nationalist forces faced an opening that enabled nationalist mobilization and increased the

likelihood of decolonization violence. I extended her coding of this variable for French colonies

only to all African colonies and also control for this factor.

Standard civil war correlates. Finally, it is also instructive to evaluate several commonly used co-

variates in the civil war literature: population,7 income per capita, and ethnic fractionalization. The7In one of the few published cross-national studies on European settlers and violent decolonization, Lawrence (2010)

22

Page 23: Economic Redistribution and Ethnic Conflict: Why European ...

former two are measured in 1950 using Maddison (2008), and the latter from Fearon (2003).

5 European Settlers and Decolonization Violence: OLS Results

Results from OLS regressions strongly support the main hypothesis by estimating a large effect for

European population share. This result is robust to considering numerous alternative hypotheses,

different combinations of covariates, altering the sample, and a fairly high extent of selection on

unobservables. Table 2 contains estimates of models of the following form:

Yi = β0 + βEEi +X ′iβX + εi, (1)

where Yi is violent decolonization, βE is the estimated effect of European settlement, Xi is a vector

of covariates that differs across the columns of Table 2, and εi is a random error term. Appendix

Table B1 shows that logistic regressions yield qualitatively identical results.

argues that this variable confounds their relationship.

23

Page 24: Economic Redistribution and Ethnic Conflict: Why European ...

Table 2. OLS ResultsDV: Violent decolonization

(1) (2) (3) (4) (5) (6) (7)ln(European pop. %) 0.212*** 0.219*** 0.236*** 0.218*** 0.203*** 0.163*** 0.238***

(0.0381) (0.0434) (0.0446) (0.0369) (0.0384) (0.0388) (0.0480)Ag. transition 8.35e-06

(6.48e-05)Political central. -0.0673

(0.194)ln(Slave exports/area) 0.0446

(0.0282)ln(Historical wars) 0.0541

(0.0534)ln(Mountains) 0.146***

(0.0499)Rugged terrain -0.151**

(0.0711)ln(Area) 0.0309

(0.0542)Herbst geography -0.0664

(0.0794)British colony -0.178

(0.343)French colony -0.0830

(0.343)Portuguese colony 0.480

(0.371)Belgian colony -0.135

(0.395)Spanish colony -0.669

(0.445)Metro. exec. constraints -0.0694**

(0.0296)WWII occupied 0.343**

(0.146)ln(Population) 0.0997**

(0.0411)ln(GDP/capita) -0.0600

(0.114)Ethnic frac. 0.271

(0.259)Observations 41 41 41 41 41 41 41R-squared 0.442 0.444 0.505 0.587 0.610 0.550 0.553

Notes: Table 2 summarizes a series of OLS regressions by presenting coefficient estimates, and standard error estimatesin parentheses. Estimates for the constant term are suppressed for expositional clarity. ∗∗∗p < 0.01, ∗∗p < 0.05,∗p < 0.1.

Column 1 presents a bivariate regression of violent decolonization on the log of European pop-

24

Page 25: Economic Redistribution and Ethnic Conflict: Why European ...

ulation share. The implied substantive magnitude is huge. Hypothetically increasing a colony’s

European settler percentage from Ghana’s one-tenth of one percent to Zimbabwe’s level of roughly

6% increases the predicted probability of violent decolonization from 5% to 78%. Furthermore,

the result is statistically significant despite a small sample size. Figure 1 depicts the corresponding

scatterplot.

Figure 1. Scatterplot of European Population Share and Violent Decolonization

GambiaSierra Leone

Congo Kinshasa

Burkina Faso

ChadTogo

NigerMauritania

RwandaSudanCentral African Republic

GhanaUgandaBenin

MaliGuineaTanzaniaBurundi

Cameroon

GabonMalawiLesotho

Ivory Coast

Congo Brazzaville

Guinea−Bissau

Botswana

KenyaMadagascar

Senegal

Mozambique

DjiboutiSwaziland

ZambiaEquatorial Guinea

MoroccoAngola

TunisiaZimbabwe

AlgeriaNamibia

Sout

−.5

0.5

11.

5Vi

olen

t dec

olon

izat

ion

−2 −1 0 1 2 3ln(European percent)

Columns 2 through 7 demonstrate that the coefficient estimate for European population share re-

mains robustly statistically significant across six groupings of covariates. These are sensible spec-

ifications because they distinguish different types of alternative explanations. But the consistent

finding is not an artifact of these specific covariate groupings. To take specification choices out of

the researcher’s hands, I ran an extreme bounds analysis (Leamer 1983) that regressed violent de-

colonization on European population share and all combinations of including up to four additional

covariates (the largest number of substantive covariates included in any Table 2 regression). The

coefficient estimate for European population share is statistically significant at 1% every one of

25

Page 26: Economic Redistribution and Ethnic Conflict: Why European ...

the 4,048 permutations. The result is also highly robust to sample modifications. For each of the

seven specifications in Table 2, I ran all combinations of dropping either one or two observations.

All but four of these 6,027 regressions are statistically significant at 1%, and all are significant at

5%.

Additionally, the magnitude of the coefficient estimate is quite stable across the various specifi-

cations in Table 2. This observation lends further credence to the robustness of the results by

suggesting that selection on unobservables would have to be quite strong relative to selection on

observables to explain away the positive effect estimate. More formally, a conventional heuristic is

to compare coefficient estimates of the main independent variable from regressions with covariates

(denoted as β̂cov) to the coefficient estimate from the bivariate regression (denoted as β̂biv) using

the metric:8

α̂ =β̂cov

β̂biv − β̂cov(2)

A higher value of α̂ corresponds with an effect more robust to hypothetically adding unobservable

covariates to the regression. Larger β̂cov in the numerator increases α̂ by creating a larger effect to

explain away. If β̂biv − β̂cov > 0, then a smaller difference between these two denominator terms

also raises α̂ because it implies that including the observed covariates reduces the effect estimate

by less—which in turn implies that additionally adding unobservables would need to diminish the

coefficient estimate by more to eliminate the positive effect estimate. If instead β̂biv−β̂cov < 0, then

the estimated effect is larger when controlling for observables, which implies the direction of the

bias from omitting the unobservables must go in the opposite direction as the bias from omitting the

observables in order to eliminate the positive effect estimate. This circumstance suggests a highly

robust result, at least absence a theory for why the bias from omitting observed versus omitting

unobserved covariates should go in opposite directions.8Altonji et al. (2005) provide an econometric derivation. Gonzalez and Miguel (2015, 31) list numerous recent

applied articles that use this technique.

26

Page 27: Economic Redistribution and Ethnic Conflict: Why European ...

Table 3 reports α̂ for European population share for each Table 2 specification with covariates,

thereby comparing the coefficient estimates from Columns 2 through 7—each of which include a

different set of observed covariates—with the coefficient estimate from Column 1. The results are

quite insensitive to hypothetically adding unobserved covariates to the regression. The European

population share coefficient estimate in each of Columns 2, 3, 4, and 7 is smaller than the European

population share coefficient estimate in the bivariate regression from Column 1, i.e., β̂biv − β̂cov <

0. The bias from omitting unobservables in the Column 5 regression must be a remarkable 25

times larger than the bias from not including the Column 5 covariates to explain away the European

settlers effect, whereas α̂ in Column 6 is lower but still quite high at 3.4. The latter figure is quite

similar to the corresponding α̂ for Altonji et al.’s (2005) main regression and justifies a similar

interpretation: “We find that selection on unobservables would need to be 3.55 times stronger

than selection on observables in the case of high school graduation, which seems highly unlikely”

(155).9

Table 3. Robustness of the European Settlers Estimate:Assessing Bias from Unobservables using Selection on Observables

Column in Table 2: (2) (3) (4) (5) (6) (7)α̂ for European population share -29.6 -9.7 -35.9 24.6 3.4 -9.1

Notes: Table 3 computes α̂ from Equation 2 for all six multiple regression specifications in Table 2.

Finally, it is useful to examine which existing explanations from the literature find support in Table

2. More mountainous terrain, weaker executive constraints in the metropole, occupation during

World War II, and higher population all covary significantly with a higher probability of a decol-

onization war. Intriguingly, none of the historical covariates used in previous research to explain

either economic development or post-independence civil wars have much explanatory power, nor9Oster (2015) demonstrates that the validity of this procedure relies upon assuming observed and unobserved con-

founders share an identical variance structure. Appendix B provides additional sensitivity analysis using her moregeneral approach.

27

Page 28: Economic Redistribution and Ethnic Conflict: Why European ...

does colonizer identity.

6 European Settlers and Decolonization Violence: IV Results

Instrumental variable regressions provide additional support for the main hypothesis. I used GIS

data to construct a variable based on climate, rainfall, elevation, and tsetse fly conditions—factors

that strongly predict areas of Africa in which Europeans settled in large numbers. Regressions using

this variable as an instrument for European settlement generate a similarly substantively large and

robust correlation between European settlement and violent decolonization. These findings are

robust to fairly large possible violations of the exclusion restriction.

6.1 Establishing a Valid Instrument: Possibilities for European Agricultural Settle-ments in Africa

Drawing from earlier influential research on European settlers, climatic factors that influenced

European settlement provide one possibility for a valid instrument. Rather than focusing on the

mortality rates of soldiers and settlers in the 19th century (Acemoglu et al. 2001) and associated

measurement problems (Albuoy 2012), however, this study draws from historians’ arguments about

required conditions for replicating large-scale European agricultural settlements in Africa (Mosley

1983, 5; Ajayi and Crowder, 1985; Bowman 1931, 200-40; Lutzelschwab 2013, 145). One con-

dition that enabled large-scale European-style farming settlements was the presence of temperate

climate, which is found at the northern and southern tips of the continent. The remainder of the

continent contains tropical climate, for which Europeans require moderate rainfall and high eleva-

tion to be able to replicate European farming practices. The presence of the tsetse fly in large parts

of Africa posed an additional impediment to European settlement.

To construct a variable that captures these conditions, I used GIS data for climate, rainfall, ele-

28

Page 29: Economic Redistribution and Ethnic Conflict: Why European ...

vation, and tsetse fly. Appendix A describes the coding procedure in detail. For each country, I

computed the percentage of territory10 that had (a) Mediterranean climate; or (b) moderate rainfall

(466mm-972mm in mean annual precipitation), high elevation (>800 meters above sea level), and

poor conditions for tsetse fly prevalence (the lowest quartile on Alsan’s 2015 tsetse fly suitability

index). Figure 2 depicts these conditions, and Appendix Figures A1 through A4 depict each factor

individually.

Figure 2. Area Suitable for Large-Scale European Settlement

Notes: The map depicts Mediterranean climate in green and moderate rainfall/high elevation/low tsetse fly areas inblue.

This is a reasonable instrument for studying the effect of European settlement on violent decol-

onization. All of the components of the instrument are exogenous in the sense that they are not

caused by political factors that could affect violent decolonization. Importantly, the tsetse fly data

comes from Alsan’s (2015) tsetse fly suitability index—which is derived from historical climate10I excluded desert and semi-desert area to eliminate territory where very few people, European or not, would settle.

29

Page 30: Economic Redistribution and Ethnic Conflict: Why European ...

data—rather than from colonial or post-colonial maps of tsetse fly prevalence, which may be af-

fected by climate change or by stronger states better able to control the fly (389). It is of course

possible that due to finite sample bias there is some degree of imbalance on confounders between

colonies with a large versus small percentage of its territory suitable for European settlement, but

(1) there is no a priori reason to believe this bias artificially supports the main hypothesis, and (2) I

use the same combination of covariates as above to demonstrate the robustness of the European set-

tlers coefficient estimate across various specifications. Additionally, I establish that the instrument

is strongly correlated with European population share, and only large violations to the exclusion

restriction would explain away the European settlers effect.

6.2 IV Results

Results from two-stage least square (2SLS) regressions strongly support the main hypothesis. Table

4 estimates simultaneous equation models composed of Equation 1 and:

Ei = β0,Z + βZZi +X ′iβX,Z + εZ,i, (3)

where Ei is the log of European population share, Xi is a vector of covariates that differs across

the columns of Table 4, Zi is the instrument, and εZ,i is a random error term. Each column of Table

4 contains 2SLS estimates of Equations 1 and 3 (Panel A), first-stage regressions of Equation 3

only (Panel B), and reduced form estimates consisting of a modified Equation 3 in which violent

decolonization replaces European population share as the dependent variable (Panel C). Each of the

seven columns contains the same of covariates as the corresponding Table 2 specification.

Column 1 presents a bivariate regression. Panel A demonstrates that, as in Table 2, the implied sub-

stantive magnitude is enormous. Hypothetically increasing a colony’s European population share

from Ghana’s one-tenth of one percent to Zimbabwe’s level of roughly 6% increases the predicted

30

Page 31: Economic Redistribution and Ethnic Conflict: Why European ...

probability of violent decolonization from 2% to 84%. This finding is qualitatively unaltered when

examining Columns 2 through 7 of Panel A.

Table 4. Instrumental Variable Results

(1) (2) (3) (4) (5) (6) (7)Panel A. 2SLS. DV: Violent decolonization

ln(European 0.239*** 0.290*** 0.326*** 0.197*** 0.260*** 0.214*** 0.235**pop %) (0.0634) (0.0923) (0.114) (0.0597) (0.0539) (0.0664) (0.0907)Observations 41 41 41 41 41 41 41R-squared 0.435 0.404 0.450 0.583 0.585 0.529 0.554

Panel B. First-stage. DV: ln(European pop %)ln(% area suitable 0.330*** 0.319*** 0.229*** 0.432*** 0.396*** 0.316*** 0.274***for Eu. agri.) (0.0696) (0.0940) (0.0834) (0.0919) (0.0627) (0.0695) (0.0734)Observations 41 41 41 41 41 41 41R-squared 0.366 0.385 0.433 0.462 0.636 0.476 0.605Partial F-test for IV 22.53 11.53 7.58 22.06 39.80 20.68 13.91

Panel C. Reduced form. DV: Violent decolonizationln(% area suitable 0.0789*** 0.0927*** 0.0748** 0.0849** 0.103*** 0.0678*** 0.0645**for Eu. agri.) (0.0248) (0.0337) (0.0305) (0.0332) (0.0216) (0.0223) (0.0305)Observations 41 41 41 41 41 41 41R-squared 0.206 0.222 0.250 0.307 0.573 0.468 0.326Additional covariatesin all regressions

None Pre-colonialdevelopment

Pre-colonialviolence

Geography ColonizerFE

Othercolonial

Standard

Notes: Table 4 summarizes a series of two-stage least square regressions in which the log of the percentage of a colony’sarea that is suitable for European agriculture instruments for the log of European population share. Panel A presentsthe coefficient estimate for European population share, and the standard error estimate in parentheses. Panel B presentsthe first-stage coefficient estimate for percentage of territory suitable for European agriculture, and the standard errorestimate in parentheses. Panel C presents the reduced form coefficient estimate for percentage of territory suitablefor European agriculture, and the standard error estimate in parentheses. This information is suppressed for all othercovariates in each panel for expositional clarity. *** p<0.01, ** p<0.05, * p<0.1.

Panel B presents the first-stage results. These findings verify that the instrument is, indeed, strongly

correlated with European population share. In all but one of the specifications, the partial F-test for

the instrument exceeds the conventional standard of 10 for a weak instrument (Staiger and Stock

1997). Figure 3 depicts the relationship from Column 1. The predominant trend is that relatively

large European settlements emerged where agricultural prospects were favorable but not where they

were unfavorable, albeit with notable exceptions such as Namibia (where South Africa promoted

31

Page 32: Economic Redistribution and Ethnic Conflict: Why European ...

European settlement even in economically unviable areas to ensure white dominance), Djibouti (a

tiny French military outpost with very few European settlers), and Lesotho (a remote mountainous

colony that Britain protected from incorporation into South Africa).

Figure 3. First-Stage Correlation Between Instrument and European Population Share

Benin

Botswana

Burkina FasoBurundi

Central African Republic

Chad

Congo Brazzaville

Congo Kinshasa

DjiboutiEquatorial Guinea

Gabon

Gambia

GhanaGuineaIvory Coast

LesothoMalawi

Mali

Mauritania

Niger Rwanda

Senegal

Sierra LeoneSudan

Swaziland

Tanzania

Togo

Uganda

Zambia

Algeria

Angola

CameroonGuinea−Bissau KenyaMadagascar

Morocco

Mozambique

Namibia South Africa

TunisiaZimbabwe

−20

24

ln(E

urop

ean

perc

ent)

−2 0 2 4 6ln(Suitable territory/area)

Because the exclusion restriction is unlikely to be perfectly satisfied, it is important to assess (1)

how badly the exclusion restriction would have to be violated for the results presented above to

be invalid and (2) the plausibility that exclusion restriction violations are indeed of a small enough

magnitude. Conley et al. (2012) provide a suitable method.11 They assume that instead of Equation

1, the dependent variable is generated by:

Yi = β0 + βEEi + γZi +X ′iβX + εi, (4)

If γ 6= 0, then the instrument directly affects the outcome, i.e., the exclusion restriction is not11The stated premise of their methods is: “Often the instrument exclusion restriction that underlies the validity of

the usual IV inference is suspect; that is, instruments are only plausibly exogenous. We present practical methods forperforming inference while relaxing the exclusion restriction” (260).

32

Page 33: Economic Redistribution and Ethnic Conflict: Why European ...

perfectly satisfied. Although it is likely that γ 6= 0 in any applied research situation, this is only

problematic for the 2SLS estimates if γ is positive and large. Table 5 states for each specification in

Table 4 the largest positive value of γ for which the 2SLS estimated effect of European population

share is exactly statistically significant at either the 5% and 10% significance levels (if instead

γ < 0, then the effect estimates in Table 4 would be negatively biased). If the true γ is larger

than the level stated in Table 5, then the 2SLS estimate for European settlement is not statistically

significant at the stated threshold. The number in parentheses states this γ threshold as a percentage

of the reduced form estimated effect of the instrument on decolonization wars when controlling for

the stated set of covariates (see Panel C of Table 4).

Table 5. Sensitivity of IV Results to Exclusion Restriction Violations

Column in Table 4: (1) (2) (3) (4) (5) (6) (7)Stat. sig. at 5% if γ ≤(% of reduced form estimate)

0.037(47%)

0.037(40%)

0.026(35%)

0.031(37%)

0.063(58%)

0.028(41%)

0.014(22%)

Stat. sig. at 10% if γ ≤(% of reduced form estimate)

0.043(55%)

0.045(49%)

0.034(46%)

0.038(45%)

0.069(67%)

0.034(50%)

0.022(34%)

Table 5 demonstrates that the 2SLS estimates are relatively insensitive to violations of the exclusion

restriction, which is somewhat remarkable considering the small sample size. For each specifica-

tion, at least one-third of the reduced form effect of the instrument on violent decolonization must

occur through channels other than European settlement for the European population share coeffi-

cient estimate to not be significant at least at the 10% level. For some specifications, this figure

is over 50%. At present we lack a coherent alternative hypothesis that would explain away the

European settler effect.

One possible exclusion restriction violation is that more favorable conditions for European settle-

ment also spurred greater pre-colonial economic and political development, and this in turn raised

violent decolonization prospects. Multiplying the coefficient estimates from two bivariate regres-

sions provides one estimate of this direct effect of the instrument: (1) political centralization on

33

Page 34: Economic Redistribution and Ethnic Conflict: Why European ...

the instrument (Appendix Table B3, Column 1) and (2) violent decolonization on political central-

ization (Column 2). This figure is 0.023, which is 38% smaller than the γ level needed to reduce

the European population share coefficient estimate from the Column 1 regression to the 5% signif-

icance level, and 46% smaller than the γ level for the corresponding 10% significance threshold.

Furthermore, 0.023 is likely a positively biased estimate of the extent to which the exclusion restric-

tion is violated because the instrument affects pre-colonial political development. The estimated

effect of political centralization on violent decolonization in a bivariate regression does not account

for the effect of European settlement. Regressing violent decolonization on both European settlers

and political centralization instead yields a negative coefficient estimate for political centralization

(Table B3, Column 3), which suggests that the bias from the exclusion restriction violation may go

in the opposite direction of the Table 4 findings.

7 Conclusion

During the post-1945 decolonization era, 15 colonies fought major independence wars against

their European colonizer. This paper argued that tensions over land redistribution in European

settler colonies was an important cause of decolonization violence, focusing on Africa. In set-

tler colonies, Europeans monopolized the best agricultural land and blocked political reforms that

could have alleviated grievances among Africans. Statistical evidence linked larger European set-

tler population shares to a robustly higher propensity for decolonization violence in Africa. This

result held whether controlling for a wide range of pre-colonial, colonial, or post-control covari-

ates proposed in the literature, or instrumenting for climatic factors that impacted where Europeans

could settle.

This paper provides three broader implications. First, it provides insight into an understudied type

of conflict, decolonization wars. Certainly European settlement was not the only cause of violent

34

Page 35: Economic Redistribution and Ethnic Conflict: Why European ...

decolonization, and we need additional statistical results to understand other contributors. This

paper also argues that bargaining over decolonization can be explained within existing theoretical

frameworks, redistributive political transitions (RPT) and ethnic grievances and civil war (EGCW).

Therefore, it may be fruitful to incorporate decolonization wars into general discussions of political

transitions and warfare rather than to treat them as a separate category or to ignore them.

Second, the long-term legacies of European colonial settlement have attracted considerable atten-

tion from political scientists and economists. Existing evidence suggests that European settlers

promoted both economic development (Acemoglu et al. 2001; Easterly and Levine 2012) and

democracy (Hariri 2012) because Europeans transplanted beneficial institutions from their home

countries. The present argument, however, highlights a negative consequence of European set-

tlement from increasing land inequality and causing tension with non-Europeans—regardless if

settlers recreated European institutions within their exclusive white communities in Africa. This

complements existing statistical research demonstrating that the beneficial legacies of European

settlement may not be particularly strong (Albuoy 2012; Paine 2016). At present, we lack a unified

framework for studying the multifaceted legacies of European settlement.

Third, the present findings also push back against recent empirical critiques of the RPT framework.

A considerable body of recent empirical evidence has demonstrated that RPT theories have minimal

explanatory power for transitions between dictatorships and democracy in the post-colonial world

(among others, Haggard and Kaufman 2012; Slater et al. 2014). A likely reason for this null

evidence is that class has only rarely composed an important political cleavage in these cases.

Although the scope of RPT theories may be more delimited than initially claimed in Acemoglu

and Robinson (2006) and Boix (2003), the present discussion highlights another substantive setting

(also see Boix 2013) in which RPT theories are highly empirically relevant.

35

Page 36: Economic Redistribution and Ethnic Conflict: Why European ...

References

Abernethy, David B. 2002. The Dynamics of Global Dominance: European Overseas Empires,1415-1980. New Haven, CT: Yale University Press.

Acemoglu, Daron and James A. Robinson. 2006. Economic Origins of Dictatorship and Democ-racy. Cambridge University Press: New York.

Acemoglu, Daron, Suresh Naidu, Pascual Restrepo, and James A. Robinson. 2013. “Democracy,Public Policy, and Inequality.” Newsletter for the Comparative Democratization Section of theAmerican Political Science Association 11(3): 2, 16-20.

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

Adu Boahen, A. 1970. “The Colonial Era: Conquest to Independence.” In Colonialism in Africa1870-1960, Volume 2, Eds. Peter Duignan and L.H. Gann. New York, NY: Cambridge UniversityPress, pp. 503–524.

Ajayi, A.J.F. and Michael Crowder. 1985. Historical Atlas of Africa. New York, NY: CambridgeUniversity Press.

Albouy, David Y. 2012. “The Colonial Origins of Comparative Development: An Empirical Inves-tigation: Comment.” American Economic Review, 102(6): 3059-76.

Alsan, Marcella. 2015. “The Effect of the TseTse Fly on African Development.” American Eco-nomic Review, 105(1): 382-410.

Altonji, Joseph G., Todd E. Elder, and Christopher R. Taber. 2005. “Selection on observed and un-observed variables: assessing the effectiveness of catholic schools.” Journal of Political Economy,113(1): 151-184.

Besley, Timothy and Marta Reynal-Querol. 2014. “The Legacy of Historical Conflict: Evidencefrom Africa.” American Political Science Review, 108(2): 319-336.

Boix, Carles. 2003. Democracy and Redistribution. Cambridge University Press: New York,NY.

Boix, Carles. 2013. “RMDs.” Newsletter for the Comparative Democratization Section of theAmerican Political Science Association, 11(3): 2, 12-15.

Bowman, Isaiah. 1931. The Pioneer Fringe. Worcester, MA: The Commonwealth Press.

Cederman, Lars-Erik, Andreas Wimmer, and Brian Min. 2010. “Why do ethnic groups rebel? Newdata and analysis.” World Politics 62(1): 87-119.

36

Page 37: Economic Redistribution and Ethnic Conflict: Why European ...

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

Colgan, Jeff. 2015. “The Political Economy of the End of Empires.” Providence, RI: Workingpaper, Brown University.

Collier, Ruth Berins. 1982. Regimes in Tropical Africa: Changing Forms of Supremacy, 1945-1975. Berkeley, CA: University of California Press.

Conley, Timothy G., Christian B. Hansen, and Peter E. Rossi. 2012. “Plausibly Exogenous.”Review of Economics and Statistics, 94(1): 260-272.

Decalo, Samuel. 1997. The Stable Minority: Civilian Rule in Africa. Gainsville, FL: FloridaAcademic Press.

Easterly, William and Ross Levine. 2012. “The European Origins of Economic Development.”NBER Working Paper Series.

Encyclopaedia Britannica Online Academic Edition. 2016a. “Libya.”http://academic.eb.com.ezp.lib.rochester.edu/EBchecked/topic/339574/Libya. Accessed 3/21/16.

Encyclopaedia Britannica Online Academic Edition. 2016b. “Rwanda genocide of 1994.”http://academic.eb.com.ezp.lib.rochester.edu/EBchecked/topic/1762747/Rwanda-genocide-of-1994?anchor=ref1111304. Accessed 4/23/16.

Encyclopaedia Britannica Online Academic Edition. 2016c. “Zimbabwe.”http://academic.eb.com.ezp.lib.rochester.edu/EBchecked/topic/657149/Zimbabwe. Accessed 3/21/16.

Fearon, James D. 2003. “Ethnic and Cultural Diversity by Country.” Journal of Economic Growth,8(2): 195-222.

Fearon, James D. and David D. Laitin. 2003. “Ethnicity, Insurgency, and Civil War.” AmericanPolitical Science Review, 97(1): 75-90.

Fieldhouse, D.K. 1986. Black Africa 1945-80: Economic Decolonization & Arrested Development.London, UK: Allen & Unwin.

Gann, Lewis D. and Peter Duignan. 1962. White Settlers in Tropical Africa. New York, NY:Penguin Press.

Gann, L.H. and Peter Duignan. 1970. “Changing Patterns of a White Elite: Rhodesian and OtherSettlers.” In Colonialism in Africa 1870-1960, Volume 2: The History and Politics of Colonialism1914-1960, Eds. L.H. Gann and Peter Duignan. Cambridge, pp. 92-170.

Gennaioli, Nicola and Ilia Rainer. 2007. “The Modern Impact of Precolonial Centralization inAfrica.” Journal of Economic Growth, 12: 185-234.

37

Page 38: Economic Redistribution and Ethnic Conflict: Why European ...

Gleditsch, Nils Petter, Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg, and HavardStrand. 2002. “Armed Conflict 1946-2001: A New Dataset.” Journal of Peace Research 39(5):615-637.

Goldsmith, Benjamin E. and Baogang He. 2008. “Letting Go Without a Fight: Decolonization,Democracy and War, 1900-94.” Journal of Peace Research, 45(5): 587-611.

Gonzalez, Felipe and Edward Miguel. 2015. “War and local collective action in Sierra Leone:A comment on the use of coefficient stability approaches.” Journal of Public Economics, 128:30-3.

Good, Kenneth. 1976. “Settler Colonialism: Economic Development and Class Formation.” TheJournal of Modern African Studies, 14(4): 597-620.

Haggard, Stephan, and Robert R. Kaufman. 2012. Inequality and Regime Change: DemocraticTransitions and the Stability of Democratic Rule. American Political Science Review 106(3): 1-22.

Hariri, Jacob Gerner. 2012. “The Autocratic Legacy of Early Statehood.” American PoliticalScience Review, 106(3): 471-494.

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

Huntington, Samuel P. 1968. Political Order in Changing Societies. New Haven: Yale UniversityPress.

Kahler, Miles. 1981. “Political Regime and Economic Actors: The Response of Firms to the Endof Colonial Rule.” World Politics, 33(3): 383-412.

Lacina, Bethany. 2009. “The PRIO Battle Deaths Dataset, 1946-2008, Version 3.0: Documentationof Coding Decisions.” Available at https://www.prio.org/Data/Armed-Conflict/Battle-Deaths/The-Battle-Deaths-Dataset-version-30.

Lawrence, Adria. 2010. “Driven to Arms? The Escalation to Violence in Nationalist Conflicts.”In Rethinking Violence: States and Non-State Actors in Conflict, Eds. Erica Chenoweth and AdriaLawrence. Cambridge, MA: MIT Press, pp. 143-171.

Lawrence, Adria. 2013. Imperial Rule and the Politics of Nationalism: Anti-Colonial Protest inthe French Empire. New York, NY: Cambridge University Press.

Leamer, Edward E. 1983. “Let’s Take the Con out of Econometrics.” American Economic Review73: 308-313.

Levitsky, Steven, and Lucan Way. 2013. “The Durability of Revolutionary Regimes.” Journal ofDemocracy, 24(3): 5-17.

38

Page 39: Economic Redistribution and Ethnic Conflict: Why European ...

Lieberman, Evan S. 2003. Race and Regionalism in the Politics of Taxation in Brazil and SouthAfrica. New York, NY: Cambridge University Press.

Lutzelschwab, Claude. 2013. “Settler Colonialism in Africa.” In Settler Economies in WorldHistory, Eds. Christopher Lloyd, Jacob Metzer, and Richard Sutch. Brill, pp. 141-167.

Maddison, Angus. 2008. “Statistics on World Population, GDP and Per Capita GDP, 1-2008 AD.”Available at http://www.ggdc.net/maddison/oriindex.htm. Accessed 4/10/16.

Marshall, Monty G. and Ted Robert Gurr. 2014. “Polity IV Project: Political Regime Characteris-tics and Transitions, 1800-2013.” Available athttp://www.systemicpeace.org/polity/polity4.htm. Accessed 4/10/16.

Mlambo, Alois S. 2014. A History of Zimbabwe. New York, NY: Cambridge University Press.

Mosley, Paul. 1983. The Settler Economies: Studies in the Economic History of Kenya and South-ern Rhodesia 1900-1963. New York, NY: Cambridge University Press.

Murdock, George P. 1967. Ethnographic Atlas. Pittsburgh, PA: University of Pittsburgh Press.

Nunn, Nathan. 2008. “The Long Term Effects of Africa’s Slave Trades.” The Quarterly Journal ofEconomics, February: 139-176.

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

Nunn, Nathan and Leonard Wantchekon. 2011. “The Slave Trade and the Origins of Mistrust inAfrica.” American Economic Review 101: 3221-3252.

Oliver, Roland and Anthony Atmore. 2005. Africa Since 1800. Fifth Edition. New York, NY:Cambridge University Press.

Oster, Emily, 2015. “Unobservable Selection and Coefficient Stability: Theory and Evidence.”Providence, RI: Working paper, Department of Economics, Brown University.

Owolabi, Olukunle P. 2015. “Literacy and Democracy Despite Slavery: Forced Settlement andPostcolonial Outcomes in the Developing World.” Comparative Politics, 48(1): 43-78.

Paine, Jack. 2016. “Improving Cross-National Regressions: Robustness Checks for RegionallyConcentrated Variables.” Rochester, NY: Working Paper.

Posel, Deborah. 2011. “The Apartheid Project, 1948-1970.” In Eds. Ross, Robert, Anne KelkMager, and Bill Nasson, The Cambridge History of South Africa, Volume 2: 1885-1994. NewYork, NY: Cambridge University Press, pp. 319-368.

Putterman, Louis. 2008. “Agriculture, Diffusion and Development: Ripple Effects of the NeolithicRevolution.” Economica, 75(300): 729-48.

39

Page 40: Economic Redistribution and Ethnic Conflict: Why European ...

Quandt, William B. 1969. Revolution and Political Leadership: Algeria, 1954-1968. Cambridge,MA: MIT Press.

Reno, William. 2011. Warfare in Independent Africa. New York, NY: Cambridge UniversityPress.

Roessler, Philip. 2011. “The Enemy Within: Personal Rule, Coups, and Civil War in Africa.”World Politics, 63(2): 300-346.

Roessler, Philip and Harry Verhoeven. 2015. “Why Comrades Go to War: Liberation Politics andthe Outbreak of Africa’s Deadliest Conflict.” Chapter from working book manuscript.

Slater, Dan, Benjamin Smith, and Gautam Nair. 2014. Economic Origins of Democratic Break-down? The Redistributive Model and the Postcolonial State. Perspectives on Politics, 12(2): 353-374.

Spiro, Herbert J. 1963. “The Rhodesias and Nyasaland.” In Five African States: Responses toDiversity, Ed. Gwendolen M. Carter. Ithaca, NY: Cornell University Press, pp. 361-470.

Spruyt, Hendrik. 2005. Ending Empire: Contested Sovereignty and Territorial Partition. Ithaca,NY: Cornell University Press.

Staiger, Douglas and James H. Stock. 1997. “Instrumental Variables Regression with Weak Instru-ments.” Econometrica, 65, 557-586.

Stein, Rachel M. 2010. “Decolonization and Conflict Over Territory in the Twentieth Century.”Stanford, CA: Working paper, Stanford University.

Toth, Anthony. 1994. “A Country Study: Algeria.” The Library of Congress Country Stud-ies.

United Nations. 1965. UN Economic Bulletin for Africa. New York: United Nations.

van de Walle, Nicolas. 2009. “The Institutional Origins of Inequality in Sub-Saharan Africa.”Annual Review of Political Science, 12: 307-27.

Wantchekon, Leonard and Omar Garcia-Ponce. 2014. “Critical Junctures: Independence Move-ments and Democracy in Africa.” Princeton, NJ: Working paper, Princeton University.

Wasserman, Gary. 1976. Politics of Decolonization: Kenya Europeans and the Land Issue 1960-1965. New York, NY: Cambridge University Press.

Wilson, H.S. 1994. African Decolonization. London, UK: Edward Arnold.

Wucherpfennig, Julian, Philipp Hunziker, and Lars-Erik Cederman. 2015. “Who Inherits theState? Colonial Rule and Post-Colonial Conflict.” American Journal of Political Science, forth-coming.

40

Page 41: Economic Redistribution and Ethnic Conflict: Why European ...

Young, Crawford. 1970. “Decolonization in Africa.” In Colonialism in Africa 1870-1960, Vol-ume 2: The History and Politics of Colonialism 1914-1960, Eds. L.H. Gann and Peter Duignan.Cambridge, pp. 450-502.

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

41