On the theory of Ethnic Conflict

1 Introduction
Ethnic conáict is frequently in the news, and pervasive throughout history. In many countries
ethnic groups are or have been visiting violence on each other, sometimes on a horriÖc scale
(the word ìgenocide,î by deÖnition, refers to a type of ethnic conáict). Fearon and Laitin
(2003) identify no less than 58 ethnic civil wars between 1945 and 1999, constituting 51% of
the total number of civil wars.
1
Less visible and newsworthy, but quite possibly much more
pervasive, is non-violent ethnic conáict. Non-violent ethnic conáict can take multiple forms.
In some countries ethnic groups compete through overtly ethnic parties, vying for power.
In others a dominant group discriminates against and exploits the others. As Esman (1994)
succinctly puts it ìwhen an ethnic group gains control of the state, important economic assets
are soon transferred to the members of that communityî (p. 229).
Yet, and crucially, ethnic conáict is by no means universal in ethnically heterogenous
societies: in many countries ethnic groups coexist peacefully. Nor is it constant over time:
many ethnically heterogenous societies experience long (sometimes very long) periods of
fairly harmonious ethnic relations before or after periods of conáict. Why do some countries
experience ethnic conáict and others donít? Why does ethnic conáict wax and wane over
time in the same country? The goal of this paper is to contribute to the e§ort to answer
these questions.
We view each society as being endowed with a set of wealth-creating assets, such as
land and mineral resources. There is therefore an incentive for a subset of agents to form
a group to wrest control of these assets from the rest of the population. Once a group has
won control over the countryís riches, however, it faces the task of enforcing the exclusion
of non-members. In particular, agents not belonging to the winning group will attempt to
inÖltrate it, so as to participate in the distribution of the spoils. For example, they will apply
for land titles, or for government jobs. This inÖltration defeats the winning groupís purpose,
as it dilutes the ìdividendî each original member receives. In large communities of millions
of citizens it can be quite costly to keep track of the genuine members so as to successfully
discriminate against the non-members. Our key idea is that, when groups can be formed
along ethnic lines, ethnic identity can be used as a marker to recognize potential inÖltrators.
By lowering the cost of enforcing membership in the winning group, ethnic diversity makes
it less susceptible to ex-post inÖltration by members of the losing one. Hence, for group
that expects to prevail in a conáict, a bid for a countryís resources is an ex-ante more
proÖtable proposition if this bid occurs along ethnic lines than if it occurs along non-ethnic
(and therefore more porous) lines.

1
Of the remaining 56 civil wars an additional 20 is classiÖed as ìambiguous,îin the sense that Fearon and
Laitin are not sure whether it was faught along ethnic lines or not (their deÖnition of ìethnicî war).


An important implication of this idea is that not all ethnic distinctions are equally
e§ective ways of enforcing group membership. In particular, it is possible that some types
of ethnic identities are harder to shed than others. One reason for this is that some ethnic
identities are more easily observed by members of other groups. The clearest case of this
is the case of skin color, or other physical characteristics that di§er markedly among ethnic
groups. Ceteris paribus ethnic boundaries based on physical di§erences should be easier to
police than boundaries based on non-visible di§erences. Another reason why not all ethnic
cleavages are equally resistant to passing is that the psychic costs of giving up oneís ethnic
identity may vary with the nature of that identity. For example, in some cases passing from
one group to the other may require religious conversion, while in others both origin and
destination groups have the same religion. Abandoning oneís religious identity may be more
costly psychologically than abandoning other traits of oneís cultural identity. Furthermore
some religions create physical markers, such as circumcision or scarring, that further increase
the cost of passing.
2
To capture this heterogeneity, we build on the notion of ethnic distance.
3
In our
model ethnic distance is the cost to be born by a member of one group to successfully
pass himself as a member of the other group. In general, we would expect ethnic distance
to be maximal when there are di§erences in skin color and other physical characteristics
that make passing all but impossible. Distance may be fairly high in the case of religious
di§erences among groups. Language barriers could plausibly be argued to be a somewhat
weaker source of distance. Potential inÖltrators can assimilate through learning the language,
or more realistically through having oneís children do so. Finally, ethnic cleavages that are
only marked by a shared sense of identity or history, unsupported by additional di§erences
of color, religion, language, or other observable characteristics, should be associated with the
lowest levels of ethnic distance.
4
Having established the concept of ethnic distance we can give a preview of how our
2Maimonides in the late 12th century explains the practice of circumcision as a way of preventing
ìstrangersî from saying they are members of the faith. ìFor sometimes people say so for the purpose of
obtaining some advantage.î Needless to say the practice has backÖred when circumcision has been used to
identify Jews for the purposes of persecution, as depicted most memorably in Luis Malleís Au Revoir les
Enfants (1987).
3
A notion of group distance is also important in Esteban and Rayís (1994, 1999) measures of polarization
and their subsequent work on ethnic conáict (reviewed below). However in their context distance is best
interpreted as distance in preference or income space, not in terms of ease of migration among groups.
4
Needless to say, ethnic bounaries can be and often are multi-dimensional, involving various combination of
physical, religious, linguistic, and other cultural di§erences. Ethnic distance is the cumulative e§ect of these
di§erences. Also, ethnic distance is not country speciÖc but is speciÖc to each pair of ethnic groups within a
country. Group A could be ìcloseî to group B but distant from C, etc. While in the paper we mostly focus
on the case of just two groups, we also discuss a multiple-group extension.



model answers the questions that motivate the paper. In particular, the probability of observing ethnic conáict is:
1. Increasing in ethnic distance. The further the distance, the most limited the passing from the losing/exploited group into the winning/dominant group, and hence the
greater the reward from conáict behavior for the latter.
2. Inverted-U shaped in expropriable assets ñ resources that can be captured through
conáict behavior ñ as a share of overall income. In our model an increase in the
share of expropriable assets has two opposing e§ects on the intensity of conáict. It
increases the ìprizeî to be gained by the winning group, and hence its incentive to
seek conáict. But it also increases the incentive for the losers to pass into the winning
group, enhancing the dilution e§ect from inÖltration, and thus reducing the incentive
for conáict by the prospective dominant group. In our model the net e§ect turns out to
be inverted-U shaped, with conáict intensity being maximized for intermediate levels
of the expropriable-resource share in total wealth.
3. Increasing in the wealth of the losing/exploited group. This is because passing often
involves having to give up part of oneís assets. For example, starting a life under a new
ethnic identity may require moving to a di§erent region where one is not previously
known, which in turn may require abandoning assets such as land, a shop, or oneís own
residence. Oneís human capital may also be location speciÖc. Hence, the cost of passing
is increasing in oneís wealth, so the richer the ìvictimî group the less concerned with
inÖltration and dilution the dominant group will be.

4. The e§ect of pre-conáict groups sizes on the likelihood of conáict is complex. In a
baseline model where the only possible form of ethnic conáict is exploitation of a weak
group by a strong group, conáict is monotonically decreasing in the pre-conáict size of
the strong group. This is because the larger the strong group the smaller the per-capita
gain from exploitation. However on an extended model where the weaker group can
also ìÖght backî to protect some of its own access to the countryís resources, or even
exploit the stronger groupís passivity to appropriate them to itself, the probability of
conáict may become U shaped in the stronger groupís size. This is because as the
weaker group becomes extremely small it has extremely large per-capita beneÖts from
5
Clearly there are other possible reasons why conáict is more likely when the side which is expected to lose
is rich. For example, the winner may expect to appropriate some of the losing sideís private assets. But the
mechanism we highlight, the e§ect of the losersíwealth on the incentive to pass, is unique to our framework
and, more importantly, it applies even to the component of private wealth that cannot be expropriated (for
example human capital).


exploiting the fact that group A has correspondingly low incentives for conáict, and it
becomes possible to observe forms of exploitation of the majority by the minority.
5. Decreasing in the destructiveness of conáict, and decreasing in the wealth of the prevailing group (because conáict destroys some of this wealth).
The mechanisms in points (1)-(3) are consequences of our notion of ethnic distance
and therefore unique to our framework. The U shape in relative size result in point (4) is
not particularly dependent on our notion of distance but as far as we know is novel in the
literature. The results in (5) would operate in a very large class of models of civil conáict.
6
Given these results, cross-country di§erences in proneness to civil conáict would result
from di§erences in all the determinants of conáict just listed, and transitions from conáict
to peace would equally be driven by changes over time in these determinants. Particularly
likely seem changes in the share of expropriable assets in total wealth, and we discuss below
a number of historical examples where we conjecture such changes may have led to long-run
changes in ethnic relations. Changes in ethnic distance due to changes in the perceived psychic
costs of passing seem also possible, and so are of course changes in the relative wealth of the
groups, changes in relative group size (for example due to migration or di§erential population
growth), and of course changes in conáict technology that may make conáict more or less
destructive.
Another source of distance is of course geography. Our model applies equally well to
groups that form based on the geographical base of their membership. When one groupís army
enters a city in enemy territory, its soldiers can be pretty conÖdent that the overwhelming
majority of the civilians they encounter belong to the enemy group. Hence, our theory of
conáict among geographically separated groups is isomorphic to our theory of ethnically
distant groups, and one may therefore be able to use our model, together with other relevant
state variables identiÖed in this paper, to explain changes over time in the intensity of interregional (and perhaps even international) conáict.
The rest of the paper is organized as follows. In the next section we discuss related
literature. This discussion is structured along the following lines. First, we discuss literature
that provides empirical and historical underpinnings to our key notion of ethnic distance,
and how it relates to ìpassingîbetween ethnic groups. Since passing is the central strategic
6
Obviously we are not saying that conáict will only arise in societies with deep ethnic divides. If the
beneÖts of conáict are large enough, a group aiming to exclude the rest of the population may arise even in
relatively homogenous societies: this group will tolerate a certain amount of leakage and/or will be willing
to pay relatively large costs to set up artiÖcial methods to enforce membership (e.g. party a¢ liation). We
are merely saying that, ceteris paribus, distance increases the likelihood for conáict, particularly if the other
conditions listed in the text are satisÖed.


concern of the dominant group in our model, this discussion serves to highlight the relevance
of the ideas developed in the paper. Other parts of the literature review section deal with
alternative theories of ethnic conáict, existing empirical evidence on ethnic conáict, and
(brieáy) the various ways in which the notion of ìethnicityî has been conceptualized in the
literature.

Our theory is developed formally in Sections 3-4. Section 3 sets out a benchmark
model of exploitation with two ethnic groups that captures the basic idea of ethnic distance
and highlights the non-monotonic role of expropriable resources. Section 4 develops a number
of extensions. In one extension we distinguish between two types of conáict: exploitation of
one group at the hands of the other, and open conáict between the two groups. We show that
our insights are robust to this richer description of ethnic relations, and we make predictions
as to when conáict will take the form of exploitation and when it will be bilateral. In a
second extension we allow for a distinction between ìleadersî and ìfollowersî within each
group. Once again our main insights go through. A third extension considers the possibility of
multi-dimensional ethnic cleavages, and hence multiple groups. Finally we look at a simple
dynamic extension of our baseline static model. The results from the dynamic analysis
are broadly in line with the baseline. However one new insight is that when factoring the
possibility of passing the dominant group may optimally manipulate the timing at which it
begins and cease to exploit the other. As a result, changes in ethnic relations can occur well
before or long after changes in the exogenous state variables, and ethnic relations can change
multiple times between peace and conáict even if the state of economy changes only once.
Section 5 discusses how our model can contribute to the understanding a number
of historical examples that we think our theory sheds some light on, including Black-White
relations in the United States and South Africa; Hutu-Tutsi relations in Rwanda and Burundi;
Muslims and Hindus in India; and others.

Click here to view the entire document