WP 2015: 16

Colonial legacy, state-building

and the salience of
ethnicity in Sub-Saharan
Africa

Merima Ali, Odd-Helge Fjeldstad,

Boqian Jiang, Abdulaziz B. Shifaz
Chr. Michelsen Institute (CMI)
is an independent, non-profit
research institution and a
major international centre in
policy-oriented and applied
development research. Focus
is on development and human
rights issues and on international
conditions that affect such
issues. The geographical focus is
Sub-Saharan Africa, Southern and
Central Asia, the Middle East and
Latin America.

CMI combines applied and

theoretical research. CMI
research intends to assist policy
formulation, improve the basis
for decision-making and promote
public debate on international
development issues.
 Colonial legacy, state-building and the salience of
ethnicity in Sub-Saharan Africa∗

Merima Ali†, Odd-Helge Fjeldstad‡, Boqian Jiang§, Abdulaziz B. Shifa‡

December 2015

Abstract

Ethnicity has received increased attention in studies of Africa’s economic

and institutional development. We present evidence on the long-term effects of
Britain’s “divide-and-rule” colonial strategy that deliberately fostered ethnic
rivalries to weaken and control locals. Using micro data from Sub-Saharan
Africa, we find that citizens of Anglophone (as compared to Francophone)
countries are more likely to: (1) attach greater importance to ethnic identity
(vis-á-vis national identity); (2) have weaker norms against tax evasion; and
(3) face extortion by non-state actors. We address endogeneity concerns using
IV regression and regression-discontinuity. These results suggest that Britain’s
divide-and-rule strategy may have undermined state-building.

JEL Classification: N47, N97, O17, O55, H10

Keywords: Colonial legacy; development; ethnicity; state capacity; Sub-Saharan
Africa.
∗
This article was prepared with financial support from the International Center for Tax and
Development (www.ICTD.ac). Useful comments on earlier drafts were received from Ragnar Torvik
and participants at the 2014 ICTD Annual Conference in Arusha.
†
Chr. Michelsen Institute and Syracuse University.
‡
Chr. Michelsen Institute, International Centre for Tax and Development, and African Tax
Institute.
§
Syracuse University.

1
1 Introduction
The role of ethnic rivalries in the institutional and economic development of states
has received significant attention. Higher ethnic fractionalization is found to be
correlated with slower economic growth, higher incidence of civil conflicts, weaker
state capacity and under-provision of public good (see, e.g., Mauro, 1995; Easterly
and Levine, 1997; Alesina et al., 2003; Miguel and Gugerty, 2005). The implication of
these empirical patterns appears to be most severe in Africa due to the higher level
of ethnic fractionalization among African countries. Despite a lot of emphasis on
the detrimental impact of ethnic rivalry on the construction of national identity and
state-building (Acemoglu et al., 2014a), the legacies of colonialization in fomenting
these rivalries and undermining state-building remains poorly understood. In this
study, using nationally representative micro data from Sub-Saharan Africa (SSA), we
examine whether differences in occupation strategies during the colonial era adopted
by the two dominant powers, France and Britain, led to different outcomes with
regard to the construction of national identity and state capacity in present-day
Africa.
Our study is motivated by the differences in colonial strategies used by the two
major colonial powers in Africa to control local populations. As outlined by Frederick
Lugard, the most prominent British colonial practitioner in Africa, Britain adopted
the strategy of divide-and-rule where ethnic cleavages were fostered and exploited to
weaken and control the local populations (Khapoya, 2010). Existing ethnic divisions
were institutionalized into the colonial power structure by subduing local chiefs. In
many instances where there were no clear ethnic divisions, the colonial masters went
as far as inventing new “ethnic” groups and installing chiefs to govern the locals
(Khapoya, 2010). In contrast, systematic exploitation of ethnic rivalries played a
less prominent role in France’s control strategy (Miles, 1994). We will return to a
more detailed discussion of the historical background in Section 2 below.
In this paper, we present empirical evidences on the impact of this distinction
in colonial strategy on state-building in post-colonial Africa. Our data are drawn
from round 5 of the Afrobarometer surveys, which are nationally representative value
surveys. Our sample covers 21 countries, consisting of twelve Anglophone and nine
Francophone states. We compare Anglophone and Francophone countries with re-
spect to three outcomes that are related to ethnic rivalry and state-building.
One may expect that rivalries among ethnic groups fostered by the divide-and-
rule colonial strategies would undermine the construction of shared national identity.
Thus, the first outcome variable captures the salience of ethnic identity as compared

2
to national identity. The other two outcome variables capture two key aspects of
state capacity – the “fiscal” and “public good” aspects.On the fiscal side, we look at
citizens’ attitude toward tax obligations. One way in which inter-ethnic rivalry and
weak sense of nationhood hinder state-building could be by lowering the willingness
to contribute to the provision of public goods (Miguel and Gugerty, 2005). Another
important aspect of state capacity is the ability of the state to provide public good.
The Afrobarometer data have useful information on one of the most important public
goods that a state is supposed to provide - the protection of citizens against extortion
by non-state actors (e.g., organized gangs). The widespread prevalence of extortion
by non-state actors could be considered as an indicator of the state’s inability to
“monopolize violence”.
Results from the OLS regressions are consistent with the view that state-building
in Anglophone countries is challenged by the legacy of British rule. On average,
citizens in Anglophone countries tend to have a weaker sense of nationhood, are less
likely to view tax compliance as an important obligation and are more likely to face
extortion by non-state actors.
Even though the patterns revealed by the OLS results are themselves interesting,
causal interpretation requires a more robust analysis. To mitigate endogeneity con-
cerns, we deploy two identification strategies. First, we construct an instrumental
variable to instrument the likelihood that an African territory becomes a British or
French colony. Construction of the instrument is guided by historical patterns in
the expansion of colonial territories. During the “scramble for Africa” (from the late
19th to early 20th century), European colonizers moved from a few coastal areas
of Africa and expanded their control inland. Their strategy was to gradually ex-
pand their control by incorporating lands adjacent to their existing colonies. Due
to transportation constraints, the control of adjacent lands required less logistics.
Thus, a key determining factor for whether a territory would be colonized by France
or Britain was its distance from an existing French or British colony. We construct
an instrument based on this idea, and find that it has a strong predictive power for
the probability of a territory becoming a French or British colony. We also show
that the instrument is uncorrelated with observable factors (e.g., pre-colonial level
of state centralization) that may affect the outcome variables.
In the second identification strategy, we exploit the fact that the Afrobarometer
data have a wide geographic coverage in each country. We use regression discontinuity
(RD) focusing on a subset of observations that reside in areas around the borders
between Anglophone and Francophone countries. In our benchmark regression, we
consider observations that reside within 100 km from the borders. Given the arbitrary

3
nature of colonial borders, results from this RD analysis are expected to capture a
“border effect”. The findings from both the instrumental variable regression and RD
analysis confirm the OLS results.
Our paper intersects several previous studies on ethnicity, state capacity and
colonial legacy. There has been increasing interest in the role of ethnicity. The first
set of studies, some of which are cited above, examine the effect of ethnic structure on
economic and institutional outcomes. More recently, Hjort (2014) has show that the
negative effect of ethnic diversity can also hinder productivity within private firms.
Furthermore, this negative effect worsens during periods of inter-ethnic tensions.
Hjort’s findings speak to a seminal study by Leeson (2005), who argues that ethnic
diversity in itself does not lead to detrimental outcomes. Leeson discusses several
historical examples where pre-colonial Africans managed to establish inter-ethnic
cooperation through a wave of “traditional” institutions, which were latter disrupted
by colonial rules.
Our paper is also related to studies on the legacy of conquests and colonialism.
These studies find that the legacies could be multi-faceted, both in terms of outcomes
and mechanisms. Acemoglu et al. (2001; 2002) show that the legacy of institutional
changes left behind by colonial powers has affected current institutional and economic
outcomes. The effect depends largely on the strategies adopted by colonizers to
maximize their benefit. In a more recent study, Acemoglu et al. (2014b) show the
legacies of indirect rule practiced by Britain in Sierra Leone. Feyrer and Sacerdote
(2009) find that longer colonial occupation is associated with higher current per
capita GDP, and the extent of this effect varies by colonizer (e.g., larger for Dutch
and British colonies than Portuguese). La Porta et al. (2008) emphasize the legal
system that the colonizers put in place (i.e., civil versus common laws) and its impact
on later institutional outcomes. Dell (2010) focuses on the long-term effect of forced
labor. Iyer (2010) studies the legacy of direct versus indirect rule in India, and finds
that the indirect rule is associated with higher provision. Using data from Cameroon,
Lee and Schultz (2012) also find similar results. Examples of other papers studying
the long-tem effects of colonialism include Banerjee and Iyer (2005), Michalopoulos
and Papaioannou (2013a; 2013b) and Huillery (2009). Both the focus and the findings
of existing studies vary, demonstrating the complexity of colonial legacy. Our paper
provides one further channel - the salience of ethnic tensions- through which colonial
legacy may affect current institutional and economic outcomes.
The remainder of the paper is organized as follows. The next section will pro-
vide brief historical background about the different colonial rulings and their conse-
quences. This is followed by discussion of the data and results in Sections 3 and 4.

4
The paper ends with some concluding remarks in Section 5.

2 Historical Background
Britain’s divide-and-rule colonial strategy to control local populations in Africa, first
formulated by the renowned British colonial master Lord Frederick, is characterized
by two key components.
The first component involves identifying locally powerful men from the pre-
colonial power structure, such as tribal chiefs, and making them part of the colonial
administration (Crowder, 1964). In areas where identifiable tribes or tribal chiefs
did not exist, the British colonizers went as far as creating them (Khapoya, 2010).
The main task of the chiefs was to control the local population and extract taxes on
behalf of their colonial masters (Mamdani, 1996). In return, the chiefs would pocket
a share of the tax revenues. The chiefs would also receive British support to suppress
resistance from the locals, empowering themselves over the local population (Migdal,
1988; Mann and Roberts, 1991; Mamdani, 1996).
Several examples from the colonial periods illustrate the British strategy. Lugard
himself implemented divide-and-rule first in Nigeria by using local chiefs from the
Islamic emirates in Northern Nigeria (Mamdani, 1996). Existing indigenous chiefs
were also used in colonial administration in Buganda Kingdom (Lwanga-Lunyiigo,
1987). There are also a number of cases where British colonizers created “chiefs” and
“tribes”. In Tanganyika, entirely new “chiefs” and “tribes” were created by British
colonizers following the takeover from Germany (Khapoya, 2010). The “warrant
chiefs” among the Igbos in South-Eastern Nigeria are also another examples of those
who were appointed as local representatives of the British colonial administration
(Afigbo, 1972). Before British colonial rule, the Igbos were not politically organized
as a single ethnic group; rather, they belonged to several groups that were politically
fragmented.
A second key component of the divide-and-rule strategy was fostering and exploit-
ing rivalries among ethnic groups to weaken and control the local populations. This
strategy primarily aimed at undermining cooperation among various ethnic groups
that could have led to a more unified and stronger resistance against the colonial
power. Instances where colonial soldiers from selected ethnic groups were utilized to
put down resistance from other ethnic groups have been widely documented across
British colonies. For example, the British used the Nubians to control Akoholi from
the early 1870s in what is now Uganda (Lwanga-Lunyiigo, 1987). The Nubians, who

5
comprised the core of the colonial army in Uganda, also helped defeat kingdoms of
other ethnic groups such as the Bunyoro and Buganda kingdoms in the 1890s, and
helped put down violence throughout Uganda (ibid). After colonizing Buganda, the
British also used Bugandan chiefs to help govern and administer other territories that
had less centralized systems of political organization and thus were harder for the
British to deal with directly. Britain’s colonial strategy also promoted segregation
of the local population along tribal lines that undermined inter-ethnic cooperation
and integration (Blanton et al., 2001).
Several historians agree that exploitation and fostering of ethnic rivalries were not
the major features of France’s colonial strategy (see, e.g., Crowder, 1964; Mamdani,
1996; Miles, 1994). First, local administrative boundaries (called ‘cantons’) did not
necessarily represent specific ethnic groups, and often cut across preexisting political
boundaries (Crowder, 1964). Thus, the local administrative boundaries in French
colonies did not hinder various ethnic groups from belonging to the same political
unit (ibid). This stands in marked contrast with the way in which Britain divided and
segregated its colonies along ethnic lines. Second, France deliberately suppressed the
local chiefs and significantly undermined their power (Whittlesey, 1937). The chiefs
in British colonies, as compared to their counterparts in French colonies, tended to
have more power in local governments.
Why the French did not follow Britain’s style of control is a matter of speculation.
According to the renowned political geographer Whittlesey (1937), the deliberate
weakening of traditional chiefs and the nominal power of the canton chiefs in French
colonies were the result of France’s policy of centralized administration and assimila-
tion of the Africans into French culture: “France is in Africa to make Frenchmen out
of Africans.” Although a full-scale policy of assimilation was perhaps undesirable
and impractical, the integration policy was noted in numerous incidences, such as
education, language, and in few cases, rights of citizenship (Crowder, 1964). Educa-
tion in the colonies was modeled on the French system, and the use of French was
encouraged throughout the colonies as the language of the government and commerce
(Clapham, 1985). Educated natives were also allowed to assume administrative po-
sitions. In a few instances, citizenship rights were granted to local educated elites,
as in four cantons in Senegal (Crowder, 1964).
Miles (1994) provides an illustrative example of this distinction in British and
French colonial rulings using a case study from Hausaland. During the pre-colonial
era, the Hausa people lived under an amalgamation of loose political unions without
any particular distinctions between the Nigerian and Nigérien sides of the border.
As is the case with the arbitrary nature of many colonial borders with respect to pre-

6
colonial ethnic and political structures, the Hausa were divided between the French
and British colonies (present-day Niger and Nigeria, respectively) and the two regions
experienced divergent colonial experiences. Miles (1994) notes,

Politically and administratively, the driving ethos behind French colonial

policy in Africa was centralization: a single structure, a single hierarchy,
a single set of rules. Structurally, this ethos was reflected in their pol-
icy of federation: unlike the British, who treated their various colonies
throughout West Africa as discrete administrative entities, the French
merged theirs into a single Afrique Occidentale Francaise (p. 95 ).

The legacy of this distinction in colonial strategy and its impact on state-building
has been a subject of controversy among scholars. On the one hand, Herbst (2014)
argues that colonizers lacked effective control in most part of their official territory
except in ports and capital cities. As a result, differences in style of colonial con-
trol can only play a minor role, if any, in explaining the nature of modern African
States. Herbst (2014) emphasizes that the role of colonization is limited mostly to
border demarcations (following the Berlin conference) that accelerated the forma-
tion of states with clearly defined borders. The other line of literature suggests that
colonization, and particularly the indirect rule of the British colonial administration,
plays an important role in state-building in present-day Africa (see, e.g., Mamdani,
1996, 2007; Blanton et al., 2001; Lange, 2004; Acemoglu et al., 2014a).
Broadly speaking, there are two possible reasons in the literature as to why the
divide-and-rule strategy may pose a lasting challenge to state-building in Anglophone
Africa. First, an important part of state-building is the construction of national iden-
tity that inculcates a sense of solidarity among fellow citizens, and historical rivalries
among ethnic groups within a state could make it difficult to construct a strong
sense of nationhood. Animosities against ethnic groups could be highly persistent
(Voigtländer and Voth, 2012). Holmén (1990) notes that divide-and-rule strategies
may have prevented the spread of national identity even after independence and
instead created “poor and unintegrated nations”. In Uganda, for example, Lwanga-
Lunyiigo (1987) argues that the use of Buganda to conquer other territories “estab-
lished a lasting hatred between the Baganda and a sizable chunk of the rest of Uganda
– a weak premise on which to build a nation.” Badru (2010) attributes the massacre
of Igbos in Nigeria’s civil war following independence to the animosity towards the
Igbos that had resulted from the colonial strategy of using members of the Igbo to
control other populations. Mamdani (2007) argues that the deep ethnic divisions

7
created in Nigeria during colonization allowed for fragmentation of Nigeria’s politics
along ethnic lines.
Second, the chiefs that were empowered in British colonies during the colonial
era still tend to retain significant power in many Anglophone countries, which can
undermine the process of building a strong central state should such a state be viewed
as a threat to the chiefs’ power. If the emergence of a strong central state would
pose a threat to locally powerful chiefs, the chiefs would have an incentive to use
their current power to keep the state weak. For example, Badru (2010) notes that in
Nigeria, the national elites “found it convenient, like the British colonialists did, to
maintain the divisiveness as a means to sustaining their hold on power”. Acemoglu et
al. (2014b) also describes how the unconstrained power of local chiefs in Sierra Leone
allows them to control civil society and influence both local and national governance.
The power and influence of the local chiefs minimizes the incentive of the state to
broadcast its power, especially in peripheries, as this might trigger resistance from
local chiefs who would like to retain control.

3 Data Source and Description

Our primary data source is the most recent round of the Afrobarometer survey
(Round 5), a nationally representative survey of several countries in Africa under-
taken during 2011 and 2012. The survey collects data on adult citizens about their
attitudes towards democracy and governance, markets, and civil society, among other
topics. In addition, the survey also gathers data on individuals’ socio-economic char-
acteristics, such as education, age, asset ownership, political participation and access
to services.
The questionnaires vary slightly across countries that some survey questions are
not asked in all countries.1 As a result, even though the Round 5 survey covers 31
countries, only 21 countries have data that can serve the purposes of our analysis.
We report the list of countries in our analysis along with year of colonization and
name of colonizer in the appendix (Table A.4). The table also reports the number of
respondents in the Afrobarometer survey and the survey year. We have twelve An-
glophone and nine Francophone countries. Cameroon is excluded from the analysis
due to the ambiguity of its colonial status. Since both Britain and France colonized
parts of Cameroon, it is not straightforward to define Cameroon’s colonial status as
1
This could partly be due to the political sensitivity of some survey questions.

8
either Anglophone or Francophone.2
The next section describes the construction of the main outcome variables using
data in the survey (in Section 3.1), followed by a description of moments of the
variables in the analysis.

3.1 Main outcome variables

As discussed in Section 1, the empirical analysis focuses on the comparison of Anglo-
phone and Francophone countries with regard to three outcomes: (1) the importance
of national identity relative to ethnic identity, (2) citizens’ view towards tax compli-
ance, and (3) the prevalence of extortion by non-state actors. To this end, we focus
on three variables in the Afrobarometer dataset that are considered to capture these
three outcomes.
The first variable includes responses regarding the importance of national versus
ethnic identity. In the questionnaire, respondents were asked to choose among five
statements ranked with respect to the importance of national versus ethnic identity.
Taking the Kenyan generic questionnaire as an example, the survey question reads:

• Let us suppose that you had to choose between being a Kenyan and being a
[member of RESPONDENT’S ethnic group]. Which of the following statements
best expresses your feelings?

(a) I feel only Kenyan.

(b) I feel more Kenyan than a [member of RESPONDENT’S ethnic group].
(c) I feel equally Kenyan and a [member of RESPONDENT’S ethnic group].
(d) I feel more a [member of RESPONDENT’S ethnic group] than Kenyan.
(e) I feel only a [member of RESPONDENT’S ethnic group].

Using answers to this question, we construct two alternative indices to measure

one’s sense of national identity. The first index, labeled National Identity Rank, as-
sumes values ranging from 0 to 4, where higher values are assigned to statements
corresponding to a higher salience of national identity. The second index, labeled Na-
tional Identity Binary, is a binary variable that assumes 1 if the respondent chooses
2
Note, however, that British Cameroon is relatively small in terms of size comprising only 9%
of the whole area of Cameroon.

9
either statement (a) or (b), i.e., if the respondent puts a higher importance on na-
tional than on ethnic identity. Otherwise, National Identity Binary equals 0.
The second outcome variable focuses on the extent to which respondents view tax
compliance as an important obligation. Respondents are asked to choose among three
ranked statements that reflect varying levels of compliance attitude. The generic
question in the survey reads:

• Please tell me whether you think that not paying the taxes people owe on their
income is:

(a) not wrong at all.

(b) wrong but understandable.
(c) wrong and punishable.

We constructed two indices to measure attitudes towards tax compliance. The

first variable, labeled Compliance Rank, equals 0, 1 or 2 if the respondent chooses
statement (a), (b) or (c), respectively. Thus, a higher value of Compliance Rank
captures a stronger attitude in favor of tax compliance. We also constructed an
alternative binary index of compliance attitude, labeled Compliance Binary, that
equals 1 if the respondent chooses statement (c) (i.e., “wrong and punishable”).
Otherwise, Compliance Binary equals 0.
Finally, the third outcome relates to the extent to which the state protects its
citizens from extortion by non-state actors. The survey asks respondents to assess the
prevalence of extortion by non-state actors in their community. The survey question
reads:

• In the last year, how often have powerful people or groups other than govern-
ment, such as criminals or gangs, made people in your community or neigh-
bourhood pay them money in return for protecting them, their property or their
businesses?

(a) Never.
(b) Only once.
(c) A few times.
(d) Often.

10
 Using these responses, we constructed two indices to measure the level of extortion
by non-state actors. The first index, labeled Extortion Rank, equals 0, 1, 2 or 3 if
the respondent chooses option (a), (b), (c) or (d), respectively. A higher value of
“Extortion Rank” indicates a higher prevalence of extortion by non-state actors.
We also constructed a binary index, labeled Extortion Binary, that equals 0 if the
respondent replies “never” and 1 otherwise.

3.2 Summary statistics

Table 1 presents means and standard deviations of the variables in our analysis,
both for the whole sample as well as for the Anglophone and Francophone sub-
samples. The top panel includes variables sourced from the Afrobarometer survey,
where respondents are the unit of analysis. To account for country-level differences,
we also include some macro variables in our analysis. These are presented in the
bottom panel.
The mean of the national identity index, National Identity Rank is larger for Fran-
cophone countries by 0.25 standard deviations (2.92 vs 2.63). Similarly, the share
of respondents who prioritize national identity over ethnic identity, as captured by
National Identity Binary, is 59% in Francophone countries whereas it is only 43%
in Anglophone countries. We also see that the tax compliance index Compliance
Rank is higher in Francophone countries (1.46 vs 1.31). The share of respondents
who view tax evasion as “wrong and punishable” is higher by 9 percentage points in
Francophone countries (55% vs 46%). The more striking difference is the prevalence
of extortion by non-state actors. Looking at Extortion Binary, Anglophone respon-
dents are 3.25 times more likely to report payments to non-state actors than are
Francophones. The other measure of extortion by non-state actors, Extortion Rank,
also shows a large difference between Francophones and Anglophones (0.08 vs 0.21,
a difference of 0.23 standard deviations).
We include a number of socio-economic characteristics of respondents in our anal-
ysis, such as age, gender, levels of education, wealth, employment status and area of
residence. Francophone respondents are more likely to be from urban areas and less
likely to be employed. The share of employed respondents is very low, partly a result
of a very low labor force participation rate (68% and 57.4% for males and females,
respectively). In both Anglophone and Francophone subsamples, male and female
respondents are equally represented. The education measure ranges from 0 = “no
formal schooling” to 9 = “postgraduate qualifications”. Respondents in Anglophone
countries are younger and tend to have a higher level of education.

11
 Table 1: Descriptive Statistics

Panel A: Individual level

Francophone Anglophone Total
mean sd mean sd mean sd
(1) (2) (3) (4) (5) (6)
Nation identity rank 2.92 (1.20) 2.63 (1.12) 2.72 (1.15)
Nation identity binary 0.59 (0.49) 0.43 (0.49) 0.48 (0.50)
Compliance rank 1.46 (0.65) 1.31 (0.71) 1.36 (0.69)
Compliance binary 0.55 (0.50) 0.46 (0.50) 0.49 (0.50)
Extortion payment rank 0.08 (0.40) 0.21 (0.61) 0.17 (0.56)
Extortion payment binary 0.04 (0.20) 0.13 (0.33) 0.10 (0.30)
Urban 0.40 (0.49) 0.35 (0.48) 0.37 (0.48)
Age 37.77 (14.40) 36.46 (14.45) 36.88 (14.45)
Employment 0.20 (0.40) 0.39 (0.49) 0.33 (0.47)
Education 2.37 (2.22) 3.53 (1.87) 3.16 (2.06)
Male 0.50 (0.50) 0.50 (0.50) 0.50 (0.50)
Wealth 0.51 (0.29) 0.47 (0.26) 0.48 (0.27)
N 9317 - 20136 - 29453 -
Panel B: Country level
West Africa 0.89 (0.33) 0.25 (0.45) 0.52 (0.51)
East Africa 0.11 (0.33) 0.50 (0.52) 0.33 (0.48)
Landlocked country 0.33 (0.50) 0.42 (0.51) 0.38 (0.50)
Ethnic fractionalization 0.75 (0.07) 0.67 (0.21) 0.70 (0.17)
Pre-colonial state centralization 0.45 (0.23) 0.64 (0.30) 0.56 (0.29)
Former German colony 0.11 (0.33) 0.17 (0.39) 0.14 (0.36)
N 9 - 12 - 21 -
This table reports means and standard deviations of the variables. The top/bottom panel reports
individual-/country-level variables. Standard deviations are in parentheses. The means and standard
deviations are reported for the whole sample as well as for Francophone and Anglophone samples.

12
 We construct an individual wealth index following the standard factor analysis
method.3 The Afrobarometer data provide information on the ownership of radio,
TV, car, water, latrine and type of roof material. We first implement the principal-
component factoring method to find the main factors among the five items that
capture individuals’ level of wealth. Then, we generate a wealth index as an average
of the main factors. To account for heterogeneity among countries, we do the factor
analysis for each country separately. For instance, TV, car, water and latrine are the
four main factors for Ghanaian respondents according to the principal-component
factoring result. The individual wealth index for a Ghanaian respondent is then
calculated as the number of items owned divided by four.
The lower panel presents the country-level variables. These variables are meant
to control for factors that may affect the outcome variables independently of colonial
status though they may still be correlated with colonial status. These may include
factors like geography, ethnic composition and pre-colonial institutions.
In order to account for differences in precolonial institutions, we consider the
pre-colonial centralization measure provided by Gennaioli and Rainer (2007).4 The
hierarchy data describe ethnic group centralization level in Africa before European
colonization. Gennaioli and Rainer (2007) use countries’ ethnic composition from
the Soviet Atlas to calculate the share of each country’s non-European population
belonging to centralized groups identified from Murdock data. The centralization
index varies from 0 (non-centralized/fragmented) to 1 (centralized). We see that
Anglophone countries on average have a higher level of pre-colonial centralization.
In our regressions, we also account for the possible effect of German occupation
in some African countries. One Francophone and two Anglophone countries in our
sample were former German colonies (Tanzania, Namibia and Togo) prior to World
War I5 ; however, after the defeat of Germany, these countries were later transferred
to Britain and France.
To account for differences in ethnic composition, we use the ethnic fractionaliza-
tion index from Alesina et al. (2003). The index measures the probability that two
3
Lawrence Hamilton, Statistics with STATA, 2013, Cengage: Chapter 12 (p. 337)
4
The data are created by matching ethnic group jurisdictional hierarchy data collected by an-
thropologist George P. Murdock and ethnic group composition data published in Atlas Narodov
Mira by Miklukho-Maklai Ethnological Institute in the Soviet Union in 1964.
5
In addition, two Anglophone countries – Ghana and Nigeria – both had a small part of land
transferred from German ownership (British Togo joined Ghana, and British Cameroon joined
Nigeria); however, the majority of the country was under British rule. Thus, these two countries
are not treated as former Germany colonies.

13
randomly selected individuals from a country belong to different ethnic groups.6 We
see that Francophone countries have a higher level of fractionalization.

4 Empirical Results
The regression equation under consideration is:

Outcomei,e,c = β × Anglophonec + X′c Γ + X′i,e,c Ω + εi,e,c (1)

where i indexes individuals, e ethnic groups, and c countries. Outcomei,e,c is one of

the three individual-level outcomes: (1) the salience of national identity, (2) attitude
towards tax compliance, and (3) payment to non-state actors. These outcomes are
measured using the variables constructed in Section 3.1. Anglophonec is a dummy
for Anglophone countries – it equals 1 for Anglophone respondents and 0 for Fran-
cophones. β is our coefficient of interest – it captures the effect from Anglophone
countries. Xc and Xi,e,c are vectors of country- and individual-level controls, respec-
tively. In all of the regressions, we cluster standard errors both by ethnic group and
country.7
We will first report estimation results from OLS regressions. OLS coefficients
are interesting as they depict the correlations between the outcome variables and
colonial status; however, one can have reasonable endogeneity concerns that the
OLS estimates may be biased. That is, the differences in the three main outcomes
between Francophone and Anglophone countries, rather than being results of colonial
legacies, could be driven by some unobservable differences between the two sets of
countries. This may, for example, be the case if Britain, with the aim of using the
divide-and-rule strategy, targeted regions that have more ethnic divisions to begin
with. France, on the other hand, with the objective of implementing a more direct
rule, may have focused on regions that consist of relatively homogeneous ethnic
groups. This colonial strategy would then imply that the current differences in the
level of inter-ethnic friction between Anglophone and Francophone countries, rather
resulting from colonial legacy, might already have existed before colonization.8
6
The construction of the ethnic fractionalization measure used data ranging from 1983 for Ghana
and Uganda to 2001 for Kenya; however, as also argued by the authors, group shares in ethnic
fractionalization measures are generally considered to be sufficiently stable over a relatively long
period of time.
7
The ethnicity clustering is based on Murdock’s ethnographic map.
8
However, H. L. Hesseling documented in his book “Divide and Rule. The Partition of Africa,

14
 We attempt to address endogeneity concerns using two identification strate-
gies. First, using geographic variations during the “scramble for Africa,” we con-
struct a novel instrument for the likelihood that an African country becomes a
British (or French) colony. In our second identification strategy, we apply regression-
discontinuity focusing on observations in areas near shared borders between Anglo-
phone and Francophone countries. The results from the instrumental variable and
regression discontinuity will be presented in Sections 4.2 and 4.3, respectively.

4.1 Benchmark regression results

We start with the OLS results on the salience of national versus ethnic identities. In
Table 2, the dependent variable is National Identity Rank – an index ranging from 0
to 4, with larger values indicating a higher sense of national identity. The first column
in the Table presents results where we only control for a set of geographic indicators.
These include two region dummies, one for East and one for West Africa. The bench-
mark group is Southern Africa. These controls are meant to account for differences
in location that may be correlated with the type of colonizers. For example, West
Africa was predominantly colonized by France and East Africa by Britain. Thus, the
geographic controls help address the concern that the regional differences may affect
state-building in other ways than the colonial legacy (e.g., through differential expe-
rience of slave trade). We also include a dummy for landlocked countries. Access to
the sea may affect state-building through a complex web of several possible channels.
For example, sea access may help state-building by facilitating international trade
and economic integration, which in turn may encourage inter-ethnic cooperation.
We see that all of the geographic indicators in column (1) are significant. The
coefficient on Anglophone is negative and statistically significant. As measured by
National Identity Rank, the average level of national identity in Anglophone countries
is lower by 0.53 than in Francophone countries. For comparison, this effect is about
20% of the mean of National Identity Rank, which is equal to 2.72 (see Table 1).
Column (2) controls for the pre-colonial level of state centralization and ethnic
fractionalization. The former variable is meant to capture the legacy of state cen-
tralization during the pre-colonial era, whereas the latter controls for the effect of
ethnic fragmentation on one’s sense of nationhood.9 A negative coefficient on Ethnic
fractionalization is consistent with the notion that constructing national identity can
1880-1914 ” (p. 177-8) that the massive French occupation in West Africa was driven more by the
navy’s desire to redeem itself from past humiliation than by any strategic concern.
9
See Section 3 for the construction of these variables

15
Table 2: OLS results for national identity. Dependent variable is National Identity
Rank

(1) (2) (3) (4)

∗∗ ∗∗∗ ∗∗∗ ∗∗∗
Anglophone −0.53 −0.52 −0.49 −0.50
(0.24) (0.17) (0.16) (0.16)
West Africa −0.29 0.11 0.18 0.18
(0.26) (0.21) (0.18) (0.18)
∗∗∗ ∗∗∗ ∗∗∗
East Africa 0.18 0.66 0.67 0.67
(0.14) (0.20) (0.15) (0.15)
∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗
Landlocked −0.37 −0.47 −0.38 −0.38
(0.13) (0.10) (0.08) (0.07)
∗∗∗ ∗∗∗ ∗∗∗
Ethnic fractionalization −1.67 −1.72 −1.72
(0.46) (0.43) (0.43)
∗ ∗∗ ∗∗
Pre-colonial state centralization −0.40 −0.47 −0.47
(0.21) (0.20) (0.20)
∗∗ ∗∗
Former German colony 0.27 0.27
(0.12) (0.12)
Urban 0.01
(0.04)
Age 0.00
(0.00)
Employment 0.01
(0.03)
Education 0.01
(0.01)
∗∗∗
Male 0.06
(0.02)
Wealth −0.06
(0.09)
R-squared 0.04 0.07 0.07 0.07
Observations 28967 28967 28967 28967
Number of country clusters 21 21 21 21
Number of ethnic clusters 316 316 316 316
Robust standard errors clustered by country and ethnic group are in parentheses. *** Significant
at 1%, ** significant at 5%, * significant at 10%.
16
be a challenging task in ethnically diverse societies. Notice that the sign on West
Africa reverses as we move from column (1) to (2). This is because West Africa
on average has a higher level of fractionalization and lower level of national identity;
however, even though the average national identity level is lower in West Africa, once
we control for differences in ethnic fractionalization, this is no longer the case. The
coefficient on Pre-colonial state centralization is negative. On the one hand, a higher
level of centralization may enhance the sense of national identity as people share a
long history of cooperation within the same state. On the other hand, centralization
may foster ethnic animosity if the central state is dominated by a powerful ethnic
group. From the results in column (2), the latter effect seems to dominate. The
coefficient on Anglophone remains fairly stable as we include these controls.
Some countries in our sample were former German colonies up until World War
I, and transferred to Britain and France thereafter. To account for the possibility
that this may have affected the style of British and French rule in those countries,
column (3) includes a dummy for former German colonies. We see that sense of
national identity appears to be higher in former German colonies. The coefficient on
Anglophone stays more or less the same.
Finally, column (4) includes individual socio-economic characteristics. Except for
gender, no other variables show a significant coefficient, suggesting that individual
characteristics have a less important effect on the salience of ethnic versus national
identity. Thus, unsurprisingly, the coefficient on Anglophone does not change signif-
icantly in response to controlling for the individual characteristics.
Table 3 reports results for the second outcome – citizens’ attitude towards tax
compliance. The dependent variable is Compliance Rank. We include the same set of
controls that we discuss above. Unlike the results from national identity, individual
socio-economic characteristics seem to matter significantly for attitudes towards tax
compliance. Our coefficient of main interest, the coefficient on Compliance Rank, is
largely insignificant; however, the signs are consistently negative across all columns.
Thus, even though the lack of statistical significance means that the evidence is not
conclusive, the negative signs are in line with the idea that attitudes towards tax
compliance are weaker among Anglophones. In fact, as we will see in Section 4.3,
the coefficient becomes significant in alternative specifications where we attempt to
address endogeniety concerns.
We now turn to results on the third outcome – payment to non-state actors.
In Table 4, the dependent variable is Extortion Payment Rank, whose higher value
indicates a more frequent payment to non-state actors. We see that the coefficient on
Anglophone is significantly positive. That is, respondents in Anglophone countries

17
 Table 3: OLS results for tax compliance. Dependent variable is Compliance Rank

(1) (2) (3) (4)

Anglophone −0.03 −0.01 −0.01 −0.02
(0.07) (0.07) (0.07) (0.07)
West Africa 0.12 0.12 0.13 0.14
(0.13) (0.14) (0.14) (0.13)
East Africa −0.09 −0.04 −0.04 −0.03
(0.13) (0.11) (0.11) (0.10)
Landlocked −0.04 −0.03 −0.02 −0.00
(0.07) (0.05) (0.06) (0.06)
∗∗ ∗∗ ∗∗
Ethnic fractionalization −0.36 −0.36 −0.31
(0.16) (0.16) (0.15)
∗∗∗ ∗∗∗ ∗∗∗
Pre-colonial state centralization −0.29 −0.30 −0.29
(0.11) (0.11) (0.11)
Former German colony 0.02 0.02
(0.06) (0.05)
∗∗∗
Urban 0.08
(0.02)
∗∗∗
Age 0.00
(0.00)
Employment −0.03
(0.02)
∗∗∗
Education 0.03
(0.01)
∗∗
Male 0.02
(0.01)
Wealth −0.02
(0.05)
R-squared 0.03 0.03 0.03 0.04
Observations 28079 28079 28079 28079
Number of country clusters 21 21 21 21
Number of ethinic clusters 316 316 316 316
Robust standard errors clustered by country and ethnic group are in parentheses. *** Significant
at 1%, ** significant at 5%, * significant at 10%.

18
report payments to non-state actors more frequently than in Francophone countries.
The magnitude of this effect, at about 0.18, is also large compared to the mean of
Extortion Payment Rank in the whole sample, which stands at 0.17 (see Table 1).
As a robustness check, we run the regressions where we used the binary outcome
measures (National Identity Binary, Compliance Binary and Extortion Payment Bi-
nary) as our dependent variables. We also run alternative regression models (i.e.,
OLS, probit and ordered logit). We found similar results from the robustness checks.
The results from the robustness regressions are reported in the appendix (Table A.1).
To summarize, the broad picture from the above empirical patterns suggests
that Anglophone countries appear to face a relatively difficult challenge in state-
building, to the extent of the dimensions of state outcomes captured by the dependent
variables. The findings show that, in Anglophone countries, the construction of
national identity is significantly mitigated and social norms toward tax compliance
are weaker. Moreover, the state is less likely to protect its citizens from extortion by
non-state actors.

4.2 Results from instrumental variable estimates

This section presents results from the instrumental variable regressions as a first
attempt to address the endogeneity concerns that may arise in the above estimates.
One needs an instrumental variable that is correlated with the likelihood that an
African country becomes a French or British colony, but that is not directly correlated
with the outcome variables. Before turning to the IV regression results, we first
describe the construction of our instrument.

A. Construction of the instrument

To construct the instrument, we exploit the patterns of colonial expansion during the
massive scramble for Africa by European powers that started in the late 19th century.
Before the 1880s, European settlers mainly stayed along the African coastline, and
very few had explored inland. Then, in the early 1880s, tension among European
powers increased. To avoid the looming threat of military conflict in Africa over
territorial claims, European powers held the Berlin Conference (1884-85) where they
reached a broad consensus on how to stake claims in Africa. The main outcome
of the conference was that actual control on African soil would essentially imply
territorial ownership. This outcome ushered in an era of rapid colonial expansion to
inland Africa. Since controlling nearby areas was logistically less costly than distant

19
Table 4: OLS results for extortion payments. Dependent variable is Extortion Pay-
ment Rank

(1) (2) (3) (4)

∗∗ ∗∗∗ ∗∗∗ ∗∗∗
Anglophone 0.17 0.18 0.18 0.18
(0.07) (0.06) (0.05) (0.05)
∗∗
West Africa 0.16 0.11 0.10 0.10
(0.07) (0.08) (0.08) (0.07)
∗∗∗ ∗ ∗ ∗∗
East Africa 0.11 0.08 0.08 0.08
(0.02) (0.04) (0.04) (0.04)
Landlocked −0.02 0.00 −0.01 −0.00
(0.02) (0.03) (0.03) (0.03)
Ethnic fractionalization 0.03 0.03 0.04
(0.11) (0.11) (0.10)
Pre-colonial state centralization −0.11 −0.11 −0.10
(0.08) (0.08) (0.08)
Former German colony −0.02 −0.02
(0.03) (0.03)
Urban 0.02
(0.03)
∗∗
Age −0.00
(0.00)
Employment −0.00
(0.02)
Education −0.00
(0.00)
Male 0.00
(0.01)
∗∗
Wealth 0.10
(0.04)
R-squared 0.02 0.02 0.02 0.02
Observations 28384 28384 28384 28384
Number of country clusters 21 21 21 21
Number of ethnic clusters 316 316 316 316
Robust standard errors clustered by country and ethnic group are in parentheses. *** Significant
at 1%, ** significant at 5%, * significant at 10%.
20
areas, colonial forces expanded their territory primarily by annexing areas adjacent
to existing colonies.
This expansion pattern implies that whether a given African country ends up
being a French or British colony is strongly influenced by the country’s distance from
the nearest existing British and French colonies – a pattern that we exploit in order
to construct the instrumental variable. For each African country, we measure its
distance from the nearest British and French colonies in the year before the country
was colonized. We then use the ratio of these distances to predict the colonizer.
For example, consider Malawi – an Anglophone country. Its colonization by Britain
began in 1889. The nearest British colony prior to 1889 was Zimbabwe (830 kms),
and the nearest French colony was Antsiranana (1675 kms). The idea is that the
relative proximity of Malawi to Zimbabwe was a significant factor in Malawi ending
up as a British (rather than a French) colony. For each of the countries in our sample,
the distances to to nearest existing French and British colonies in the year prior to
colonization are provided in the appendix (see Table A.4).10 In all cases, we consider
the straight-line distance between the centroids of each geographic unit.
The instrumental variable, Relative Distance, is defined as:
Distance to the nearest existing British colony
Relative Distance =
Distance to the nearest existing French colony

In order for the IV regression to deliver a consistent estimate, it is not enough that
the instrument provides a significant correlation in the first stage; we also need that
the instrument should not affect the outcome variables other than through its effect
on the endogenous variable. That is, the instrument should not only be relevant, but
it also needs to be exogenous.
Admittedly, as is the case with any instrument, one cannot conclusively rule
out all possible channels by which the instrument may affect the outcome variables.
While aware of this caveat, we find it quite hard to imagine plausible scenarios where
distance to the nearest existing French and British colonies is not orthogonal to fac-
tors that may potentially affect the outcome variables in our analysis. Our outcome
variables, like inter-ethnic rivalries, seem to be primarily driven by domestic situa-
tions within the territory such as the level of ethnic fractionalization and historical
inter-ethnic conflicts. On the other hand, the exact distances of the nearest British
and French colonies from a territory at the time of the territory’s occupation (which
10
The years of colonization are determined according to various historical sources listed in the
appendix.

21
are the source of variation for our instrument) are likely to be determined by histor-
ical situations that the colonizers faced in the lead-up to the conquest of the nearest
colonies. These historical situations may include, among other factors, the colonizers’
military resources, the level of resistance that the colonizers had faced and difficulty
of the terrain during previous conquests. For example, in 1888 (the year of Ugan-
dan occupation), Britain possessed colonies much closer to Uganda than did France.
Our identifying assumption is that these differences of distance are not primarily the
results of domestic situations within Uganda. Instead, they were outcomes of histor-
ical factors that led Britain to occupy the southern African territories (and France to
occupy the western African territories) in the earlier years – factors that do not seem
to be results of the domestic situations at the time within the Ugandan territory.
Thus, it seems plausible to assume that these distance differences are orthogonal to
domestic factors within Uganda and hence our outcome variables.
Moreover, in an attempt to diagnose possible violations of the exogeneity as-
sumption more systematically, we regress the instrument on each of the country-level
control variables to check whether there is a sign of correlation between the instru-
ment and observable factors that can potentially affect the outcome variables. Table
5 presents the results. We see that none of the control variables show significant
correlation with the instrument, indicating no violation of the exclusion restriction.

B. Results

Turning to the IV regressions, Table 6 reports the results for the salience of national
identity. The dependent variable is National Identity Rank. We include control
variables following the patterns in our OLS specifications in the previous tables. In
the bottom panel, we report the first-stage coefficient on the instrument Relative
Distance and F-stat on the excluded instrument. The coefficient on the instrument
has the expected sign and is significant. The F-stat is also fairly large.
The first column controls only for geographic variables. We see that the co-
efficient on Anglophone is negative and significant, implying that the salience of
national identity is relatively low in Anglophones and reaffirming the results from
the OLS estimations. The second column controls for ethnic fractionalization and
pre-colonial state centralization. Both of the controls enter significantly. Column (3)
includes an indicator variable for whether the country was a former German colony.
This control is also significant. Finally, we control for individual-level respondent
characteristics. Notice that even though many of the controls enter significantly
(particularly national-level variables), the coefficient on Anglophone remains stable

22
Table 5: Coefficient from regressions of the instrument on country characterstics.

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

West Africa 0.39
(0.33)
East Africa -0.06
(0.36)
Landlocked country -0.29
(0.34)
Ethnic fractionalization 1.38
(0.98)
Pre-colonial state centralization -0.54
(0.60)
Former German colony -0.01
(0.48)
R-squared 0.07 0.00 0.04 0.09 0.04 0.00
Observations 21 21 21 21 21 21
Robust standard errors are in parentheses. *** Significant at 1%, ** significant at
5%, * significant at 10%.

23
Table 6: Results from IV regressions. The dependent variable is National Identity
Rank

(1) (2) (3) (4)

∗ ∗∗ ∗∗∗ ∗∗∗
Anglophone −0.42 −0.43 −0.42 −0.42
(0.24) (0.18) (0.16) (0.16)
West Africa −0.22 0.16 0.22 0.22
(0.27) (0.20) (0.18) (0.18)
∗∗∗ ∗∗∗ ∗∗∗
East Africa 0.18 0.66 0.67 0.67
(0.14) (0.20) (0.15) (0.15)
∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗
Landlocked −0.36 −0.46 −0.37 −0.37
(0.13) (0.10) (0.08) (0.08)
∗∗∗ ∗∗∗ ∗∗∗
Ethnic fractionalization −1.68 −1.73 −1.72
(0.48) (0.44) (0.44)
∗∗ ∗∗ ∗∗
Pre-colonial state centralization −0.42 −0.48 −0.49
(0.19) (0.19) (0.19)
∗∗ ∗∗
Former German colony 0.28 0.28
(0.11) (0.11)
Urban 0.01
(0.04)
Age 0.00
(0.00)
Employment −0.00
(0.03)
Education 0.00
(0.01)
∗∗∗
Male 0.06
(0.02)
Wealth −0.05
(0.09)
First-stage coefficient on the excluded instrument −0.51 −0.51 −0.51 −0.49
F-stat on the excluded instrument 59.59 64.78 87.63 97.47
R-squared 0.04 0.07 0.07 0.07
Observations 28967 28967 28967 28967
Robust standard errors clustered by country and ethnic group are in parentheses. *** Significant
at 1%, ** significant at 5%, * significant at 10%.
24
across columns. This is reassuring since the stability of the coefficient on Anglophone,
despite inclusion of controls that enter significantly, suggests that the effect in the
IV regression appears to be insulated from confounding factors.
Table 7 presents results for tax compliance. The dependent variable is Compliance
Rank. As is the case with the OLS estimates, even though the signs are consistently
negative, the tax compliance results lack statistical significance.
Table 8 reports the results for the prevalence of extortion by non-state actors.
We see that the IV results reaffirm the findings from OLS regressions, namely, that
Anglophones are more likely to face extortion by non-state actors. Once again, the
coefficient on Anglophone remains stable as we include more controls.
As a robustness check, we run the IV regressions with the binary outcome mea-
sures (National Identity Binary, Tax Compliance Binary and Extortion Binary) as
our dependent variables, and found similar results (see Table A.2 in the appendix).

4.3 Regression Discontinuity

Turning now to the second approach to address the endogeneity, we implement re-
gression discontinuity (RD) on a limited set of observations in our sample that reside
in areas near borders shared by Anglophone and Francophone countries. There is
little disagreement among African scholars that African borders were for the most
part arbitrarily drawn by colonizers. As the British prime minister at the time
Lord Salisbury said, “we have been engaged in drawing lines upon maps where no
white man’s feet have ever trod; we have been giving away mountains and rivers and
lakes to each other, only hindered by the small impediment that we never knew ex-
actly where the mountains and rivers and lakes were.” 11 Thus, the borders typically
divided communities that belonged to relatively homogeneous groups that shared
similar ethnicity, political organization and agro-economic zones (see Michalopoulos
and Papaioannou (2011) for more systematic evidence). This means that disconti-
nuities at borders can be reasonably interpreted as “border effects,” i.e., effects of
colonial border demarcations as opposed to, say, pre-colonial differences.
Of the twenty-one countries in our sample, nine have borders shared by an Anglo-
phone and a Francophone country, all in West Africa. Three of them are Anglophone
countries (Ghana, Nigeria and Sierra Leone), and the rest are Francophone (Guinea,
Cote d’Ivoire, Burkina Faso, Togo, Benin and Niger). As shown in Figure 1, there are
11
Geographical Journal, vol. xxciii, Proceedings, Mar. 9, 1914.

25
Table 7: Results from IV regressions. The dependent variable is Tax Compliance
Rank

(1) (2) (3) (4)

Anglophone −0.04 −0.04 −0.04 −0.05
(0.07) (0.05) (0.05) (0.06)
West Africa 0.11 0.11 0.11 0.12
(0.13) (0.13) (0.13) (0.12)
East Africa −0.09 −0.04 −0.04 −0.03
(0.12) (0.11) (0.11) (0.10)
Landlocked −0.04 −0.03 −0.03 −0.01
(0.06) (0.05) (0.06) (0.06)
∗∗ ∗∗ ∗∗
Ethnic fractionalization −0.35 −0.36 −0.31
(0.15) (0.15) (0.14)
∗∗∗ ∗∗∗ ∗∗∗
Pre-colonial state centralization −0.29 −0.29 −0.28
(0.11) (0.11) (0.10)
Former German colony 0.02 0.02
(0.06) (0.06)
∗∗∗
Urban 0.08
(0.02)
∗∗∗
Age 0.00
(0.00)
Employment −0.03
(0.03)
∗∗∗
Education 0.03
(0.01)
∗∗
Male 0.02
(0.01)
Wealth −0.03
(0.06)
First-stage coefficient on the excluded instrument −0.51 −0.51 −0.51 −0.49
F-stat on the excluded instrument 59.24 63.92 86.46 95.69
R-squared 0.03 0.03 0.03 0.04
Observations 28079 28079 28079 28079
Robust standard errors clustered by country and ethnic group are in parentheses. *** Significant
at 1%, ** significant at 5%, * significant at 10%.
26
 Table 8: Results from IV regressions. The dependent variable is Extortion Rank

(1) (2) (3) (4)

∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗
Anglophone 0.19 0.19 0.19 0.20
(0.05) (0.04) (0.04) (0.04)
∗∗∗ ∗ ∗ ∗∗
West Africa 0.17 0.11 0.11 0.11
(0.05) (0.07) (0.06) (0.06)
∗∗∗ ∗ ∗ ∗∗
East Africa 0.12 0.08 0.08 0.08
(0.02) (0.04) (0.04) (0.04)
Landlocked −0.02 0.00 −0.00 0.00
(0.02) (0.03) (0.03) (0.03)
Ethnic fractionalization 0.03 0.03 0.04
(0.10) (0.10) (0.09)
Pre-colonial state centralization −0.12 −0.11 −0.10
(0.08) (0.08) (0.08)
Former German colony −0.02 −0.02
(0.04) (0.03)
Urban 0.02
(0.03)
∗∗∗
Age −0.00
(0.00)
Employment −0.00
(0.02)
Education −0.00
(0.00)
Male 0.00
(0.01)
∗∗
Wealth 0.10
(0.04)
First-stage coefficient on the excluded instrument −0.51 −0.51 −0.51 −0.49
F-stat on the excluded instrument 59.95 65.34 87.11 96.49
R-squared 0.02 0.02 0.02 0.02
Observations 28384 28384 28384 28384
Robust standard errors clustered by country and ethnic group are in parentheses. *** Significant
at 1%, ** significant at 5%, * significant at 10%.

27
 Table 9: Results from regression-discontinuity

National Identity Rank Compliance Rank Extortion Payment Rank

(1) (2) (3) (4) (5) (6)
∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗
Angolophone −0.32 −0.35 −0.11 −0.11 0.14 0.15
(0.06) (0.06) (0.04) (0.04) (0.03) (0.03)
∗
Age −0.00 0.00 −0.00
(0.00) (0.00) (0.00)
∗∗
Male 0.03 0.04 0.00
(0.03) (0.02) (0.01)
∗∗
Employment 0.03 −0.05 0.00
(0.04) (0.02) (0.02)
∗∗∗
Urban −0.03 0.06 0.01
(0.04) (0.02) (0.02)
∗∗∗ ∗∗
Education 0.03 0.01 0.00
(0.01) (0.00) (0.00)
∗ ∗∗∗
Wealth 0.11 0.01 0.15
(0.07) (0.04) (0.03)
∗∗ ∗∗∗
Distance 1.30 −0.23 0.74
(0.52) (0.29) (0.28)
∗∗∗
Distance square −1.96 1.06 −1.74
(1.27) (0.71) (0.63)
∗ ∗∗∗
Distance cube 1.02 −0.78 1.10
(0.87) (0.47) (0.40)
R-squared 0.15 0.16 0.08 0.09 0.08 0.08
Observations 5749 5749 5600 5600 5517 5517
This table reports results from regression-discontinuity. The observations are drawn from respon-
dents residing within 100 km from shared borders between Anglophone and Francophone countries.
Fixed-effects for ethnic groups and border segments are included in the regressions. Robust stan-
dard errors clustered by ethnic group are in parentheses.
*** Significant at 1%, ** significant at 5%, * significant at 10%.

28
 Figure 1: Border

six segments of shared borders among these countries.12 Using geodata, we located
the village of each respondent.13 In the baseline regression, we selected respondents
that were located within 100 kilometers of the borders. We also did robustness checks
with 75 and 125 km and found similar results. Thanks to the wide geographic cov-
erage of the Afrobarometer surveys within each country, we have more than 5,500
observations for our analysis (in the baseline regression).
Table 9 reports results from the discontinuity regressions for national identity,
tax compliance and prevalence of extortion by non-state actors. The dependent
variables are National Identity Rank, Compliance Rank and Extortion Payment Rank.
In all of the regressions, we include fixed effects for each of the border segments
to account for potential heterogeneity among different segments arising from, for
example, differences in how the borders are drawn.14 We also include fixed effects
12
The shared border between Nigeria and Cameroon is excluded from our six border group. In
1961, southern part of British Cameroons joined Cameroon while the northern part joined Nigeria;
thus, the current border between Nigeria and Cameroon is not the original dividing line between
French and British rule.
13
The geodata on Afrobarometer respondents are from Kotsadam et al. (2015). We would like to
thank Eivind Hammersmark Olsen for collecting the geodata and sharing it with us.
14
Some borders, like the one between Ghana and Burkina Faso, may be more reflective of natural
differences (due to the Volta River between the two borders) than others like the Nigeria-Niger

29
for each ethnic territory. This helps account for differences in ethnic compositions
across borders. In order to account for the possible effect of distance from the borders,
we control for distance from borders. The polynomial distance terms are meant to
capture potential non-linearities in the effect of distance.
Using the binary outcome measures – National Identity Binary, Compliance Bi-
nary and Extortion Payment Binary – as the dependent variables delivers similar
results (see Table A.3 in the appendix).
The results reiterate the empirical patterns that we have uncovered in the previous
tables. Anglophones report a lower salience of national identity. They also report
weaker attitudes towards tax compliance, and face a higher frequency of extortion
by non-state actors.15 One important pattern is that the coefficient for attitudes
towards tax compliance, which was not significant in the OLS and IV regressions, is
now significant.

5 Concluding remarks
The role of ethnicity in economic and institutional development of states has received
increased attention in recent years. Despite the emphasis on the detrimental impact
of ethnic rivalry on the construction of national identity and state-building, the role of
historical factors in fomenting these rivalries and undermining state-building remains
poorly understood. This paper examines the legacies of different styles of colonial rule
in Africa. Our study is motivated by debates surrounding the distinction between
the strategies adopted by Britain and France, as well as their implications for long-
term state-building in Africa. While Britain adopted the “divide-and-rule” strategy,
which is argued to have fostered (and sometimes created) local ethnic rivalries, the
exploitation of inter-ethnic rivalries played a less prominent role in control strategy
adopted by France.
Two key features characterize the divide-and-rule strategy. The first involves
identifying (or creating) locally powerful men from the pre-colonial power structure,
such as tribal chiefs, and making them part of the colonial administration. The
second involves fostering and exploiting rivalries among ethnic groups to weaken
and control local populations. This strategy is primarily aimed at undermining
cooperation among various ethnic groups that might otherwise have led to a more
border.
15
There are slight variations in the number of observations across columns due to missing values
for some variables in the Afrobarometer data.

30
unified and stronger resistance against the colonial power.
Broadly speaking, there are two possible reasons in the literature as to why the
divide-and-rule strategy may pose a lasting challenge to state-building in Anglophone
Africa. First, historical rivalries among ethnic groups within a state could make it
difficult to construct a strong sense of nationhood. Second, the chiefs that were
empowered in British colonies during the colonial era still tend to retain significant
power in many Anglophone countries, which can undermine the process of building
a strong central state if such a state may be viewed as a threat to the chiefs’ power.
These arguments contrast with the view that the distinction in colonial occupation
strategy should matter very little for post-colonial state development in Africa since
the colonizers’ control, which tended to be limited to capitals and coastal towns, did
not appear to be strong enough to affect post-colonial institutions.
The empirical findings lend support to the hypothesis that Britain’s divide-and-
rule strategy has had a lasting negative impact on state-building. We find that
citizens in Anglophone countries tend to report a lower salience of national identity.
Anglophone respondents are also less likely to view paying taxes as an important
obligation, which is a sign of weaker norms towards tax compliance. They are also
more likely to face extortion by non-state actors, indicating the weakness of Anglo-
phone states to monopolize violence. Furthermore, these results hold both in the
instrumental variable and regression discontinuity identification strategies we deploy
in our analysis.
In addition to highlighting the role of history in fostering inter-ethnic rivalries, the
results also suggest that policies adopted by those who control the central government
can either exacerbate or mitigate frictions among ethnic groups. Colonial policies
are likely to be one of the many factors that affect rivalries among ethnic groups.
Thus, future research could provide valuable insights in order to identify what kind
of government policies influence inter-ethnic cooperation and state-building.

References
Acemoglu, Daron, Isaias N. Chaves, Philip Osafo-Kwaako, and James A.
Robinson, “Indirect Rule and State Weakness in Africa: Sierra Leone in Compar-
ative Perspective,” in “African Successes: Sustainable Growth” NBER Chapters,
National Bureau of Economic Research, Inc, June 2014.

, Simon Johnson, and James A. Robinson, “The Colonial Origins of Compar-

31
 ative Development: An Empirical Investigation,” The American Economic Review,
2001, 91 (5), 1369–1401.

, , and , “Reversal of Fortune: Geography and Institutions in the Making

of the Modern World Income Distribution,” The Quarterly Journal of Economics,
2002, 117 (4), 1231–1294.

, Tristan Reed, and James A. Robinson, “Chiefs: Economic Development

and Elite Control of Civil Society in Sierra Leone,” Journal of Political Economy,
2014, 122 (2), 319 – 368.

Afigbo, A.E., The Warrant Chiefs: indirect rule in southeastern Nigeria, 1891-1929
Ibadan history series, Longman, 1972.

Alesina, Alberto, Arnaud Devleeschauwer, William Easterly, Sergio

Kurlat, and Romain Wacziarg, “Fractionalization,” Journal of Economic
Growth, 2003, 8 (2), 155–194.

Badru, Pade, “Ethnic conflict and state formation in post-colonial Africa: a com-
parative study of ethnic genocide in the Congo, Liberia, Nigeria, and Rwanda-
Burundi,” Journal of Third World Studies, 2010, 27 (2), 149–172.

Banerjee, Abhijit and Lakshmi Iyer, “History, Institutions, and Economic Per-
formance: The Legacy of Colonial Land Tenure Systems in India,” American Eco-
nomic Review, 2005, 95 (4), 1190–1213.

Blanton, Robert, T. David Mason, and Brian Athow, “Colonial Style and
Post-Colonial Ethnic Conflict in Africa,” Journal of Peace Research, 2001, 38 (4),
473–491.

Clapham, Christopher ., Third World Politics: An Introduction Madison, WI:

University of Wisconsin Press, 1985.

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

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

Easterly, William and Ross Levine, “Africa’s Growth Tragedy: Policies and
Ethnic Divisions,” The Quarterly Journal of Economics, 1997, 112 (4), 1203–1250.

32
Feyrer, James and Bruce Sacerdote, “Colonialism and Modern Income: Islands
as Natural Experiments,” The Review of Economics and Statistics, 2009, 91 (2),
245–262.

Gennaioli, Nicola and Ilia Rainer, “The modern impact of precolonial central-
ization in Africa,” Journal of Economic Growth, September 2007, 12 (3), 185–234.

Herbst, J., States and Power in Africa: Comparative Lessons in Authority and
Control: Comparative Lessons in Authority and Control Princeton Studies in In-
ternational History and Politics, Princeton University Press, 2014.

Hjort, Jonas, “Ethnic Divisions and Production in Firms,” The Quarterly Journal
of Economics, 2014, 129 (4), 1899–1946.

Holmén, Hans, State, cooperatives and development in Africa, Vol. 86, Nordic
Africa Institute, 1990.

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

Iyer, Lakshmi, “Direct versus Indirect Colonial Rule in India: Long-Term Conse-
quences,” The Review of Economics and Statistics, November 2010, 92 (4), 693–
713.

Khapoya, V.B., The African Experience: An Introduction, Longman, 2010.

Kotsadam, Andreas, Eivind Hammersmark Olsen, Carl Henrik Knutsen,

and Tore Wig, “Mining and local corruption in Africa,” 2015.

Lange, Matthew K., “British Colonial Legacies and Political Development,” World
Development, 2004, 32 (6), 905 – 922.

Lee, Alexander and Kenneth A. Schultz, “Comparing British and French Colo-
nial Legacies: A Discontinuity Analysis of Cameroon,” Quarterly Journal of Po-
litical Science, 2012, 7 (4), 365–410.

Leeson, Peter T., “Endogenizing fractionalization,” Journal of Institutional Eco-

nomics, June 2005, 1 (01), 75–98.

Lwanga-Lunyiigo, S., The Colonial Roots of Internal Conflict in Uganda, Makerere

University, Makerere Institute of Social Research, 1987.

33
Mamdani, Mahmood, Citizen and Subject: Contemporary Africa and the Legacy
of Late Colonialism, Princeton, NewJersey: Princeton University Press, 1996.

, Political violence and state formation in post-colonial Africa, International De-

velopment Centre, Open University, 2007.

Mann, K. and R. Roberts, Law in colonial Africa Social history of Africa, Heine-
mann Educational Books, 1991.

Mauro, Paolo, “Corruption and Growth,” The Quarterly Journal of Economics,

1995, 110 (3), 681–712.

Michalopoulos, Stelios and Elias Papaioannou, “The Long-Run Effects of the

Scramble for Africa,” Working Paper 17620, National Bureau of Economic Re-
search November 2011.

and , “National Institutions and Subnational Development in Africa,” The

Quarterly Journal of Economics, 2013, 129 (1), 151–213.

and , “Pre-Colonial Ethnic Institutions and Contemporary African Develop-

ment,” Econometrica, 2013, 81 (1), 113–152.

Migdal, Joel S, Strong societies and weak states: state-society relations and state
capabilities in the Third World, Princeton University Press, 1988.

Miguel, Edward and Mary Kay Gugerty, “Ethnic diversity, social sanctions,
and public goods in Kenya,” Journal of Public Economics, December 2005, 89
(11-12), 2325–2368.

Miles, W. F., Hausaland Divided: Colonialism and Independence in Nigeria and

Niger, Ithaca and London: Cornell University Press, 1994.

Porta, Rafael La, Florencio Lopez de Silanes, and Andrei Shleifer, “The
Economic Consequences of Legal Origins,” Journal of Economic Literature, 2008,
46 (2), 285–332.

Voigtländer, Nico and Hans-Joachim Voth, “Persecution Perpetuated: The

Medieval Origins of Anti-Semitic Violence in Nazi Germany,” The Quarterly Jour-
nal of Economics, 2012, 127 (3), 1339–1392.

Whittlesey, Derwent, “British and French colonial technique in west Africa,”

Foreign Affairs, 1937, 15 (2), 362–373.

34
Appendix A Country details and robustness re-
gressions

Table A.4: Country details

Country (colonizer) Sample size Colony established* Nearest British/ French colony Dbri /Df ra (RD)∗∗
Benin (F) 1186 1892- 1900 Lagos (B port) / Cotonou (F port) 373 / 368 (1.01)
Botswana (B) 675 1880 - 1885 South Africa / Antsiranana (F port) 863 / 3069 (0.28)
Burkina Faso (F) 1000 1896 - 1919 Ghana / Mali 480 / 622 (0.77)
Cote d’Ivoire (F) 2169 1886 - 1893 Ghana / Grand-Bassam (F port) 470 / 310 (1.52)
Ghana (B) 2269 1821 - 1867 Cape Coast (B port) / Dakar (F port) 320 / 1908 (0.16)
Guinea (F) 1004 1890 - 1898 Freetown (B port) / Conakry (F port) 336 / 319 (1.05)
Kenya (B) 833 1888 - 1895 Mombasa (B port) / Antsiranana (F port) 514 / 1923 (0.27)
Lesotho (B) 2069 1867 -1868 South Africa/ Antsiranana (F port) 360 / 3124 (0.12)
Madagascar (F) 1126 1885 - 1896 South Africa / Antsiranana (F port) 2684 / 884 (3.04)
Malawi(B) 732 1889 - 1891 Zimbabwe / Antsiranana (F port) 830 / 1675 (0.50)
Mali(F) 2317 1890 - 1892 Ghana / Mauritania 1113 / 833 (1.33)
Namibia(B) 1133 1915 -1915 Botswana / Congo 731 / 2441 (0.30)
Niger (F) 2357 1895 - 1900 Lagos (B port) / Mali 1422 / 1436 (0.99)
Nigeria (B) 2165 1879 - 1900 Lagos (B port)/ Dakar (F port) 641 / 2901 (0.22)
Senegal (F) 1166 1850 - 1865 Freetown (B port) / Dakar (F port) 680 / 333 (2.04)
Sierra Leone (B) 2142 1800- 1895 Freetown (B port) / Dakar (F port) 160 / 938 (0.17)
Tanzania (B) 1071 1924 - 1924 Malawi / Madagascar 787 / 2004 (0.39)
Togo (F) 918 1920 - 1920 Ghana / Benin 340 / 197 (1.73)
Uganda (B) 871 1888 - 1894 Mombasa (B port) / Antsiranana (F port) 992 / 2420 (0.41)
Zambia (B) 1132 1886 - 1890 Botswana / Antsiranana (F port) 1108 / 2403 (0.46)
Zimbabwe (B) 1118 1885 - 1888 Botswana / Antsiranana (F port) 774 / 2302 (0.34)
* This column lists the periods during which control of each colony’s territory took place.
** RD ≡ D Db
f
, where Db (Df ), denotes distance to nearest British (French) colony.

35
 Table A.1: Results from OLS, probit and ordered logit models

Dependent variable Model Specification (1) (2) (3) (4)

∗∗ ∗∗∗ ∗∗∗ ∗∗∗

National Identity Rank Ordered Logit -0.93 -0.92 -0.86 -0.88
(0.40) (0.29) (0.27) (0.27)
∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗
National Identity Binary OLS -0.26 -0.25 -0.24 -0.24
(0.09) (0.06) (0.06) (0.06)
∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗
National Identity Binary Probit -0.67 -0.66 -0.62 -0.63
(0.26) (0.18) (0.17) (0.17)
Observations: 28967

Compliance Rank Ordered Logit -0.09 -0.04 -0.03 -0.08

(0.22) (0.20) (0.20) (0.20)

Compliance Binary OLS -0.03 -0.02 -0.02 -0.03

(0.06) (0.05) (0.05) (0.05)

Compliance Binary Probit -0.08 -0.05 -0.04 -0.07

(0.15) (0.14) (0.13) (0.13)
Observations: 28079
∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗
Extortion Rank Ordered Logit 1.35 1.43 1.46 1.49
(0.43) (0.35) (0.35) (0.33)
∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗
Extortion Binary OLS 0.68 0.71 0.73 0.74
(0.21) (0.17) (0.17) (0.16)
∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗
Extortion Payment Binary Probit -0.67 -0.66 -0.62 -0.63
(0.26) (0.18) (0.17) (0.17)
Observations: 28384

Location control Yes Yes Yes Yes

Ethnic fraction and centralization control No Yes Yes Yes
Former German colony control No No Yes Yes
Individual control No No No Yes
Number of country/ethnicity clusters: 21/316

This table reports estimated coefficients on Anglophone from alternative specifications. The first
(second) column describes the dependent variables (regression models). The last four columns include
alternative sets of controls in the regressions, which are listed in the bottom rows. Standard errors
∗ ∗∗
clustered by country and ethnic group are in parentheses. significant at 10%, significant at 5%,
∗∗∗
significant at 1%.

36
Table A.2: Results from IV regressions. The dependent variables are National Iden-
tity Binary (Panel A), Compliance Binary (Panel B) and Extortion Payment Binary
(Panel C).

Panel A: National Identity Binary

∗ ∗∗∗ ∗∗∗ ∗∗∗
Anglophone -0.19 -0.19 -0.19 -0.19
(0.10) (0.07) (0.07) (0.06)
R-squared 0.05 0.08 0.09 0.09
Observations 28967 - - -

Panel B: Tax Complaince Binary

Anglophone -0.04 -0.04 -0.04 -0.04
(0.05) (0.04) (0.04) (0.04)
R-squared 0.02 0.03 0.03 0.03
Observations 28079 - - -

Panel C: Extortion Payment Binary

∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗
Anglophone 0.12 0.12 0.12 0.12
(0.03) (0.02) (0.02) (0.02)
R-squared 0.02 0.03 0.03 0.03
Observations 28384 - - -

Location control Yes Yes Yes Yes

Ethnic fraction and centralization control No Yes Yes Yes
Former German colony control No No Yes Yes
Individual control No No No Yes

This table reports the IV regression results for three dependent variables. The list
of control variables included in the regressions are provided below the coefficients.
Standard errors clustered by countries and ethnic groups are in parenthesis.
∗ ∗∗ ∗∗∗
significant at 10%, significant at 5%, significant at 1%.

37
 Table A.3: Results from regression discontinuity

National Identity Binary Compliance Binary Extortion Payment Binary

(1) (2) (3) (4) (5) (6)
∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗
Angolophone −0.18 −0.19 −0.08 −0.08 0.09 0.09
(0.03) (0.03) (0.03) (0.03) (0.02) (0.02)
∗
Age −0.00 0.00 −0.00
(0.00) (0.00) (0.00)
∗∗∗
Male 0.00 0.04 0.00
(0.01) (0.01) (0.01)
Employment 0.01 −0.02 0.00
(0.02) (0.02) (0.01)
∗
Urban −0.02 0.03 0.00
(0.02) (0.02) (0.01)
∗∗∗
Education 0.01 0.01 0.00
(0.00) (0.00) (0.00)
∗∗ ∗∗∗
Wealth 0.07 0.03 0.08
(0.03) (0.03) (0.02)
∗∗ ∗∗∗
Distance 0.49 −0.16 0.54
(0.23) (0.24) (0.15)
∗∗∗
Distance square −0.70 0.87 −1.27
(0.54) (0.57) (0.33)
∗ ∗∗∗
Distance cube 0.39 −0.68 0.80
(0.36) (0.38) (0.21)
R-squared 0.19 0.19 0.08 0.09 0.10 0.11
Observations 5749 5749 5600 5600 5517 5517
This table reports results from regression-discontinuity. The observations are drawn from respon-
dents residing within 100 km from shared borders between Anglophone and Francophone countries.
Fixed-effects for ethnic groups and border segments are included in the regressions. Robust stan-
dard errors clustered by ethnic group are in parentheses.
*** Significant at 1%, ** significant at 5%, * significant at 10%.

38
Appendix B Sources for determining the years of
colony establishment
Benin: Benin. Worldmark Encyclopedia of Nations. 2007. Encyclopedia.com
Botswana: Thomas Tlou & Alec Campbell, History of Botswana, 2nd edn. Gaborone:
Macmillan. (1997)
Burkina Faso: [1] Klein, Martin A. Slavery and Colonial Rule in French West
Africa. (1998) London: Cambridge University Press. [2] Lea, David; Rowe, Anna-
marie.A Political Chronology of Africa (2001). Taylor & Francis.
Cote d’Ivore: Cote d’Ivore. Worldmark Encyclopedia of Nations. 2007. Encyclo-
pedia.com
Ghana: James L. McLaughlin and David Owusu-Ansah. A Country Study: Ghana,
Chapter 1: Historic Setting. Library of Congress Federal Research Division (Novem-
ber 1994)
Guinea: Guinea. Worldmark Encyclopedia of Nations. 2007. Encyclopedia.com
Kenya: [1] The London Gazette: no. 25854. pp. 50175020. 7 September 1888. [2]
Oxford Dictionary of National Biography September 2004. Oxford University Press.
Lesotho: Eldredge, EA A South African Kingdom: The pursuit of security in nineteenth-
century Lesotho,1993, Cambridge University Press.
Madagascar: Helen Chapin. Library of Congress Country Studies: Madagascar.1994.
Archived from the original on 1 February 2011.
Malawi: De Robeck, A Pictorial Essay of the 1898 Provisions of British Central
Africa Nyasaland. Privately Published, N.P., 1982.
Mali: [1] Klein, Martin A. Slavery and Colonial Rule in French West Africa. (1998)
London: Cambridge University Press. [2] Lea, David; Rowe, Annamarie. A Political
Chronology of Africa (2001). Taylor & Francis.
Namibia: Namibia. Worldmark Encyclopedia of Nations. 2007. Encyclopedia.com
Niger: Niger. Worldmark Encyclopedia of Nations. 2007. Encyclopedia.com
Nigeria: [1] Helen Chapin Metz, ed. Nigeria: A Country Study. Washington: GPO
for the Library of Congress, 1991. [2] Temple, Charles Lindsay.Northern Nigeria.The
Geographical Journal(1912)40(2): 149163.
Senegal: [1] Chisholm, Hugh, ed. Faidherbe, Louis Lon Csar. Encyclopedia Britan-
nica (11th ed.).(1911) Cambridge University Press. [2] Klein, M. (2007). France’s

39
African Colonies. In T. Benjamin (Ed.)Encyclopedia of Western Colonialism since
1450, (Vol. 2). (pp. 493) Detroit: Macmillan Reference USA.
Sierra Leone: Christopher Fyfe, A History of Sierra Leone. London, 1962 :p 486
Tanzania: A Modern History of Tanganyika.Cambridge University Press, May 10,
1979.
Togo: Togo. Worldmark Encyclopedia of Nations. 2007. Encyclopedia.com
Uganda: [1] The London Gazette: no. 25854. pp. 50175020. 7 September 1888.
[2] Galbraith, John S, 1970,Italy, the British East Africa Company, and the Benadir
Coast, 18881893, The Journal of Modern History 42. 4, pages 549-563
Zambia, Zimbabwe: E A Walter, (1963).The Cambridge History of the British
Empire: South Africa, Rhodesia and the High Commission Territories, Cambridge
Universiry Press.

40
CMI WORKING PAPERS
This series can be ordered from:

CMI (Chr. Michelsen Institute)

Phone: +47 47 93 80 00
Fax: +47 47 93 80 01
E-mail: [email protected]

P.O.Box 6033 Bedriftssenteret,

N-5892 Bergen, Norway
Visiting address:
Jekteviksbakken 31, Bergen

Web: www.cmi.no

Printed version: ISSN 0804-3639

Electronic version: ISSN 1890-5048
Printed version: ISBN 978-82-8062-573-0
Electronic version: ISBN 978-82-8062-574-8

This paper is also available at:

www.cmi.no/publications

I
Ethnicity has received increased attention in studies of Africa’s economic and
institutional development. We present evidence on the long-term effects of
Britain’s “divide-and-rule” colonial strategy that deliberately fostered ethnic
rivalries to weaken and control locals. Using micro data from Sub-Saharan
Africa, we find that citizens of Anglophone (as compared to Francophone)
countries are more likely to: (1) attach greater importance to ethnic identity
(vis-a-vis national identity); (2) have weaker norms against tax evasion; and
(3) face extortion by non-state actors. We address endogeneity concerns using
IV regression and regression-discontinuity. These results suggest that Britain’s
divide-and-rule strategy may have undermined state-building.