# Oxford University Press 2004 Oxford Economic Papers 56 (2004), 563–595 563
All rights reserved doi:10.1093/oep/gpf064
Greed and grievance in civil war
By Paul Collier* and Anke Hoefﬂery
*Centre for the Study of African Economies, University of Oxford
yCentre for the Study of African Economies, University of Oxford, 21 Winchester
Road, Oxford OX2 6NA; e-mail: anke.hoefﬂer@economics.ox.ac.uk
We investigate the causes of civil war, using a new data set of wars during 1960–99.
Rebellion may be explained by atypically severe grievances, such as high inequality,
a lack of political rights, or ethnic and religious divisions in society. Alternatively,
it might be explained by atypical opportunities for building a rebel organization.
While it is difﬁcult to ﬁnd proxies for grievances and opportunities, we ﬁnd that
political and social variables that are most obviously related to grievances have little
explanatory power. By contrast, economic variables, which could proxy some grievances
but are perhaps more obviously related to the viability of rebellion, provide
considerably more explanatory power.
1. Introduction
Civil war is now far more common than international conﬂict: all of the 15 major
armed conﬂicts listed by the Stockholm International Peace Research Institute for
2001 were internal (SIPRI, 2002).
In this paper we develop an econometric model which predicts the outbreak
of civil conﬂict. Analogous to the classic principles of murder detection, rebellion
needs both motive and opportunity. The political science literature explains
conﬂict in terms of motive: the circumstances in which people want to rebel are
viewed as sufﬁciently rare to constitute the explanation. In Section 2 we contrast
this with economic accounts which explain rebellion in terms of opportunity:
it is the circumstances in which people are able to rebel that are rare. We discuss
measurable variables which might enable us to test between the two accounts
and present descriptive data on the 79 large civil conﬂicts that occurred
between 1960 and 1999. In Section 3 we use econometric tests to discriminate
between rival explanations and develop an integrated model which provides a
synthesis. Section 4 presents a range of robustness checks and Section 5 discusses
the results.
This analysis considerably extends and revises our earlier work (Collier and
Hoefﬂer, 1998). In our previous theory, we assumed that rebel movements incurred
net costs during conﬂict, so that post-conﬂict pay-offs would be decisive. The
core of the paper was the derivation and testing of the implication that high
post-conﬂict pay-offs would tend to justify long civil wars. We now recognize that
this assumption is untenable: rebel groups often more than cover their costs during
the conﬂict. Here we propose a more general theory which juxtaposes the opportunities
for rebellion against the constraints. Our previous empirical analysis
conﬂated the initiation and the duration of rebellion. We now treat this separately.
This paper focuses on the initiation of rebellion.1
Our sample is expanded from a
cross-section analysis of 98 countries during the period 1960–92, to a comprehensive
coverage of 750 ﬁve-year episodes over the period 1960–99, enabling us to
analyse double the number of war starts. Further, we expand from four explanatory
variables to a more extensive coverage of potential determinants, testing for robustness
to select a preferred speciﬁcation.
2. Rebellion: approaches and measures
2.1 Preferences, perceptions, and opportunities
Political science offers an account of conﬂict in terms of motive: rebellion occurs
when grievances are sufﬁciently acute that people want to engage in violent protest.
In marked contrast, a small economic theory literature, typiﬁed by Grossman
(1991, 1999), models rebellion as an industry that generates proﬁts from looting,
so that ‘the insurgents are indistinguishable from bandits or pirates’ (Grossman,
1999, p.269). Such rebellions are motivated by greed, which is presumably sufﬁciently
common that proﬁtable opportunities for rebellion will not be passed up.2
Hence, the incidence of rebellion is not explained by motive, but by the atypical
circumstances that generate proﬁtable opportunities. Thus, the political science and
economic approaches to rebellion have assumed both different rebel motivation—
grievance versus greed—and different explanations—atypical grievances versus
atypical opportunities.
Hirshleifer (1995, 2001) provides an important reﬁnement on the motiveopportunity
dichotomy. He classiﬁes the possible causes of conﬂict into preferences,
opportunities, and perceptions. The introduction of perceptions allows
for the possibility that both opportunities and grievances might be wrongly
perceived. If the perceived opportunity for rebellion is illusory—analogous to the
‘winners’ curse’—unproﬁtability will cause collapse, perhaps before reaching our
threshold for civil war. By contrast, when exaggerated grievances trigger rebellion,
ﬁghting does not dispel the misperception and indeed may generate genuine
grievances.
Misperceptions of grievances may be very common: all societies may have groups
with exaggerated grievances. In this case, as with greed-rebellion, motive would
not explain the incidence of rebellion. Societies that experienced civil war would
be distinguished by the atypical viability of rebellion. In such societies rebellions
564 p. collier and a. hoeffler
..........................................................................................................................................................................
1
On the analysis of the duration of civil war see Collier et al. (2004) and Fearon (2004).
2
By the ‘Machiavelli Theorem’ (Hirshleifer, 2001, p.10–11) no one will pass up a proﬁtable opportunity
to exploit someone else.
would be conducted by viable not-for-proﬁt organizations, pursuing misperceived
agendas by violent means.
Greed and misperceived grievance have important similarities as accounts
of rebellion. They provide a common explanation—‘opportunity’ and ‘viability’
describe the common conditions sufﬁcient for proﬁt-seeking, or not-for-proﬁt,
rebel organizations to exist. On our evidence they are observationally equivalent
since we cannot observe motives. They can jointly be contrasted with the political
account of conﬂict in which the grievances that both motivate and explain rebellion
are assumed to be well-grounded in objective circumstances such as unusually
high inequality, or unusually weak political rights. We now turn to the proxies
for opportunities and objective grievances.
2.2 Proxies for opportunity
The ﬁrst step in an empirical investigation of conﬂict is a clear and workable
deﬁnition of the phenomenon. We deﬁne civil war as an internal conﬂict with at
least 1,000 combat-related deaths per year. In order to distinguish wars from
massacres, both government forces and an identiﬁable rebel organization must
suffer at least 5% of these fatalities. This deﬁnition has become standard following
the seminal data collection of Small and Singer (1982) and Singer and Small (1994).
We use an expanded and updated version of their data set that covers 161 countries
over the period 1960–99 and identiﬁes 79 civil wars, listed in Table 1. Our task is to
explain the initiation of civil war using these data.3
We now consider quantitative indicators of opportunity, starting with opportunities
for ﬁnancing rebellion. We consider three common sources: extortion of natural
resources, donations from diasporas, and subventions from hostile governments.4
Klare (2001) provides a good discussion of natural resource extortion, such as
diamonds in West Africa, timber in Cambodia, and cocaine in Colombia. In Table 2,
we proxy natural resources by the ratio of primary commodity exports to GDP for
each of the 161 countries. As with our other variables, we measure at intervals of
ﬁve years, starting in 1960 and ending in 1995. We then consider the subsequent
ﬁve years as an ‘episode’ and compare those in which a civil war broke out (‘conﬂict
episodes’) with those that were conﬂict-free (‘peace episodes’). The descriptive
statistics give little support to the opportunity thesis: the conﬂict episodes were on
average slightly less dependent upon primary commodity exports than the peace
episodes. However, there is a substantial difference in the dispersion. The peace
episodes tended to have either markedly below-average or markedly above-average
dependence, while the conﬂict episodes were grouped around the mean.5
Possibly if
greed and grievance 565
..........................................................................................................................................................................
3
Later in the paper (Table 6) we examine whether our results are sensitive to this deﬁnition of civil war.
Sambanis (2002) comes to a similar conclusion.
4
We list the data sources and deﬁnitions in the Appendix.
5
The standard deviation of primary commodity exports is 0.11 for the conﬂict episodes and 0.19 for the
peace episodes.
566 p. collier and a. hoeffler
Table 1 Outbreaks of war
Country Start of
the war
End of
the war
Previous war Gdp
sample
Secondary
schooling sample
Afghanistan 04/78 02/92
Afghanistan 05/92 Ongoing *
Algeria 07/62 12/62 *
Algeria 05/91 Ongoing * * *
Angola 02/61 11/75
Angola 11/75 05/91 * * *
Angola 09/92 Ongoing * * *
Azerbaijan 04/91 10/94
Bosnia 03/92 11/95
Burma/Myanmar 68 10/80 * * *
Burma/Myanmar 02/83 07/95 * * *
Burundi 04/72 12/73 * *
Burundi 08/88 08/88 * * *
Burundi 11/91 ongoing * * *
Cambodia 03/70 10/91 *
Chad 03/80 08/88 *
China 01/67 09/68 * *
Columbia 04/84 ongoing * * *
Congo 97 10/97 * *
Cyprus 07/74 08/74 *
Dominican Rep. 04/65 09/65 * *
El Salvador 10/79 01/92 * *
Ethiopia 07/74 05/91 * *
Georgia 06/91 12/93
Guatemala 07/66 07/72 * * *
Guatemala 03/78 03/84 * * *
Guinea-Bissau 12/62 12/74
India 08/65 08/65 * * *
India 84 94 * * *
Indonesia 06/75 09/82 * * *
Iran 03/74 03/75 * *
Iran 09/78 12/79 * * *
Iran 06/81 05/82 * * *
Iraq 09/61 11/63 *
Iraq 07/74 03/75 * * *
Iraq 01/85 12/92 * * *
Jordan 09/70 09/70 *
Laos 07/60 02/73 *
Lebanon 05/75 09/92 *
Liberia 12/89 11/91 *
Liberia 10/92 11/96 *
Morocco 10/75 11/89 * * *
Mozambique 10/64 11/75
Mozambique 07/76 10/92 * * *
Nicaragua 10/78 07/79 * *
Nicaragua 03/82 04/90
Nigeria 01/66 01/70 * *
(continued )
natural resources are sufﬁciently abundant, as in Saudi Arabia, the government may
be so well-ﬁnanced that rebellion is militarily infeasible. This offsetting effect may
make the net effect of natural resources non-monotonic.6
The observed pattern
may also reﬂect differences between primary commodities (which we defer to
Section 3). Further, primary commodities are associated with other characteristics
that may cause civil war, such as poor public service provision, corruption and
economic mismanagement (Sachs and Warner, 2000). Potentially, any increase in
conﬂict risk may be due to rebel responses to such poor governance rather than to
ﬁnancial opportunities.
greed and grievance 567
..........................................................................................................................................................................
6
Collier (2000) provides an illustrative formal model of such a non-monotonic relationship.
Table 1 Continued
Country Start of
the war
End of
the war
Previous war Gdp
sample
Secondary
schooling sample
Nigeria 12/80 08/84 * * *
Pakistan 03/71 12/71 * *
Pakistan 01/73 07/77
Peru 03/82 12/96 * *
Philippines 09/72 12/96 * * *
Romania 12/89 12/89 * *
Russia 12/94 08/96
Russia 09/99 Ongoing *
Rwanda 11/63 02/64
Rwanda 10/90 07/94 * * *
Sierra Leone 03/91 11/96 * *
Sierra Leone 05/97 07/99 * *
Somalia 04/82 05/88 * *
Somalia 05/88 12/92 * * *
Sri Lanka 04/71 05/71 * *
Sri Lanka 07/83 ongoing * * *
Sudan 10/63 02/72
Sudan 07/83 ongoing * * *
Tajikistan 04/92 12/94
Turkey 07/91 ongoing * *
Uganda 05/66 06/66 * *
Uganda 10/80 04/88 * * *
Vietnam 01/60 04/75 *
Yemen 05/90 10/94
Yemen, Arab Rep. 11/62 09/69 *
Yemen, People’s Rep. 01/86 01/86 *
Yugoslavia 04/90 01/92
Yugoslavia 10/98 04/99 *
Zaı¨re/Dem. Rep. of Congo 07/60 09/65
Zaı¨re/Dem. Rep. of Congo 09/91 12/96 * * *
Zaı¨re/Dem. Rep. of Congo 09/97 09/99 * * *
Zimbabwe 12/72 12/79 * *
Note: Previous wars include war starts 1945–1994.
A second source of rebel ﬁnance is from diasporas. Angoustures and Pascal
(1996) review the evidence, an example being ﬁnance for the Tamil Tigers from
Tamils in north America. We proxy the size of a country’s diaspora by its emigrants
living in the US, as given in US Census data. Although this neglects diasporas living
in other countries, it ensures uniformity in the aggregate: all diasporas are in the
same legal, organizational, and economic environment. We then take this emigrant
population as a proportion of the population in the country of origin. In the formal
analysis we decompose the diaspora into that part induced by conﬂict and that
which is exogenous to conﬂict, but here we simply consider the crude numbers.
These do not support the opportunity thesis: diasporas are substantially smaller in
the conﬂict episodes.
A third source of rebel ﬁnance is from hostile governments. For example,
the government of Southern Rhodesia pump-primed the Renamo rebellion in
Mozambique. Our proxy for the willingness of foreign governments to ﬁnance
military opposition to the incumbent government is the Cold War. During the
Cold War each great power supported rebellions in countries allied to the opposing
568 p. collier and a. hoeffler
Table 2 Descriptive statistics
Sample No civil war Civil war
(n ¼ 1167) (n ¼ 1089) (n ¼ 78)
War starts 0.067 0 1
Primary commodity exports/GDP 0.168 0.169 0.149
GDP per capita (const. US$) 4061 4219 1645
Diaspora (relative to population of
country of origin)
0.017 0.018 0.004
Male secondary schooling (% in school) 43.42 44.39 30.3
GDP per capita growth (average for
previous 5 years)
1.62 1.74 À0.23
Previous war (% with war since 1945) 20.8 18.5 53.8
Peace duration (months since last conﬂict) 327 334 221
Forest cover (%) 31.11 31.33 27.81
Mountainous terrain (%) 15.82 15.17 24.93
Geographic concentration of the
population (Gini)
0.571 0.569 0.603
Population density (inhabitants per
square km)
150 156 62
Population in urban areas (%) 45.11 46.00 32.7
Ethnic fractionalization (index, 0–100) 39.57 38.64 52.63
Religious fractionalization (index, 0–100) 36.09 35.98 37.70
Polarization  ¼ 1.6 (index, 0–0.165) 0.077 0.077 0.076
Democracy (index, 0–10) 3.91 4.07 1.821
Ethnic dominance (% with main ethnic
group 45–90%)
0.465 0.465 0.452
Income inequality (Gini) 0.406 0.406 0.410
Land inequality (Gini) 0.641 0.641 0.631
Note: We examine 78—rather than the 79—war starts as listed in Table 1 because Pakistan experienced
two outbreaks of war during 1970–74. We only include one of these war starts to avoid double counting.
power. There is some support for the opportunity thesis: only eleven of the 79 wars
broke out during the 1990s.
We next consider opportunities arising from atypically low cost. Recruits must
be paid, and their cost may be related to the income forgone by enlisting as a rebel.
Rebellions may occur when foregone income is unusually low. Since non-economists
regard this as fanciful we give the example of the Russian civil war. Reds and
Whites, both rebel armies, had four million desertions (the obverse of the recruitment
problem). The desertion rate was ten times higher in summer than in winter:
the recruits being peasants, income foregone were much higher at harvest time
(Figes, 1996). We try three proxies for foregone income: mean income per capita,
male secondary schooling, and the growth rate of the economy. As shown in Table 2,
the conﬂict episodes started from less than half the mean income of the peace
episodes. However, so many characteristics are correlated with per capita income
that, depending upon what other variables are included, the proxy is open to other
interpretations. Our second proxy, male secondary school enrollment, has the
advantage of being focused on young males—the group from whom rebels are
recruited. The conﬂict episodes indeed started from lower school enrollment,
but this is again open to alternative interpretation: education may affect the risk
of conﬂict through changing attitudes. Our third measure, the growth rate of the
economy in the preceding period, is intended to proxy new income opportunities.
Conﬂict episodes were preceded by lower growth rates. This is consistent with
evidence that the lower is the rate of growth, the higher is the probability of
unconstitutional political change (Alesina et al., 1996).7
Although the three proxies
are all consistent with atypically low forgone income as an opportunity, low income
could also be interpreted as an objective economic grievance.
The opportunity for rebellion may be that conﬂict-speciﬁc capital (such as
military equipment) is unusually cheap. We proxy the cost of such capital by the
time since the most recent previous conﬂict: the legacy of weapon stocks, skills, and
organizational capital will gradually depreciate. Empirically, peace episodes are
preceded by far longer periods of peace than conﬂict episodes (Table 2). While
this supports the opportunity thesis, it could also be interpreted as reﬂecting the
gradual decay of conﬂict-induced grievances.
Another dimension of opportunity is an atypically weak government military
capability. An unambiguous indicator is if the terrain is favorable to rebels: forests
and mountains provide rebels with a safe haven. We measure the proportion of a
country’s terrain that is forested, using FAO data. We could ﬁnd no equivalent data
greed and grievance 569
..........................................................................................................................................................................
7
The economic growth literature concentrates on the analysis of political instability as a determinant of
economic growth (see for example Barro 1991, 1997). Alesina et al. (1996) estimate a simultaneous
equation system of economic growth and political instability. They present support for the hypothesis
that political instability reduces growth. Lower growth does not seem to cause political instability,
deﬁned as the number of government changes. However, when they deﬁne political instability more
narrowly as unconstitutional government changes they ﬁnd that lower growth rates are a causal factor of
political instability.
on mountainous terrain: proxies such as altitude tend to misclassify both plateaus
and rugged uplands. We therefore commissioned a new index from a specialist,
John Gerrard. The descriptive statistics (Table 2) suggest that terrain may matter: in
conﬂict episodes 25% of the terrain is mountainous, versus only 15% in peace
episodes, although there is no difference in forest cover. Geographic dispersion of
the population may also inhibit government capability: Herbst (2000) suggests that
Zaire is prone to rebellion because its population lives around the edges of the
country. We measure dispersion by calculating a Gini coefﬁcient of population
dispersion.8
In fact, the concentration of the population is slightly lower prior to
peace episodes (0.57) than prior to war episodes (0.6). Similarly, low population
density and low urbanization may inhibit government capability. Empirically, prior
to war episodes both population density and urbanization are low (Table 2).
A ﬁnal source of rebel military opportunity may be social cohesion. Ethnic and
religious diversity within organizations tends to reduce their ability to function
(Easterly and Levine, 1997, Alesina et al., 1999, Collier, 2001). A newly formed rebel
army may be in particular need of social cohesion, constraining recruitment to a
single ethnic or religious group. A diverse society might in this case reduce the
opportunity for rebellion by limiting the recruitment pool. The most widely used
measure of ethnic diversity is the index of ethno-linguistic fractionalization. This
measures the probability that two randomly drawn people will be from different
ethnic groups. We could ﬁnd no measure of religious fractionalization, but we
constructed one equivalent to that of ethnic fractionalization using data from
Barrett (1982). If ethnic and religious divisions are cross-cutting, social fractionalization
is multiplicative rather than additive. We could ﬁnd no data relating
religious and ethnic divisions and so we construct a proxy that measures the maximum
potential social fractionalization.9
The thesis that social cohesion enhances
opportunity is not supported by the descriptive statistics: conﬂict episodes have
atypically high fractionalization. This seems more consistent with a grievance
interpretation, to which we now turn.
2.3 Proxying objective grievances
We consider four objective measures of grievance: ethnic or religious hatred, political
repression, political exclusion, and economic inequality.
570 p. collier and a. hoeffler
..........................................................................................................................................................................
8
For the calculation of the Gini coefﬁcient we used the population data per 400 km2
cell. Analogous to
the income Gini coefﬁcient, the Gini coefﬁcient of population dispersion will be high if the population is
concentrated in a relatively small area of the country.
9
If there were e equally sized ethnic groups and r equally sized religious groups, maximum potential
social fractionalization would be measured simply by the product er. Since both the underlying indices of
ethnic and religious fractionalization range on the scale 0–100, their product is zero if there is either
religious or ethnic homogeneity whereas there is social homogeneity only if both indices are zero. We
therefore measure social fractionalization as the product of the underlying indices plus whichever index
is the greater.
Ethnic and religious hatreds are widely perceived as a cause of civil conﬂict.
Although such hatreds cannot be quantiﬁed, they can evidently only occur in
societies that are multi-ethnic or multi-religious and so our proxies measure
various dimensions of diversity. Our previously discussed measures of fractionalization
are pertinent: inter-group hatreds must be greater in societies that are
fractionalized than in those which are homogenous. However, arguably the
source of inter-group tension is not diversity but polarization.10
Fortunately, the
allowable class of measures of polarization is quite limited. We adopt a general
measure due to Esteban and Ray (1994)
P ¼ K
Xn
i¼1
Xn
j¼1
1þ
i jd ð1Þ
where i denotes the percentage of people that belong to group i in the total
population, i ¼ 1, . . . , n. This measure of polarization depends on the parameters
K and . K does not change the order, but is used for population normalization.
Esteban and Ray show that  is bounded between zero and 1.6. We calculate
the polarization measure for three different values of , 0, 0.8 and 1.6, using
primary data on ethnic composition.11
In addition we investigate the variant of
the Esteban-Ray measure proposed by Reynal-Querol. These measures indeed distinguish
polarization from fractionalization: their correlation coefﬁcient ranges
between 0.39 ( ¼ 1.6) and 1.0 ( ¼ 0).12
The descriptive data does not suggest
that polarization is important: conﬂict and peace episodes have very similar mean
values (Table 2).
We measure political repression using the Polity III data set (see Jaggers and
Gurr, 1995). This measure of political rights ranges 0–10 on an ascending ordinal
scale. Political rights differ considerably between conﬂict and peace episodes. We
also investigate the Polity III measure of autocracy, and a measure of political
openness published by Freedom House (the ‘Gastil Index’). The quantitative political
science literature has already applied these measures to conﬂict risk. Hegre
et al. (2001) ﬁnd that repression increases conﬂict except when it is severe.
Even in democracies a small group may fear permanent exclusion. A potentially
important instance is if political allegiance is based on ethnicity and one ethnic
group has a majority. The incentive to exploit the minority increases the larger is
the minority, since there is more to extract (Collier, 2001). Hence, a minority
greed and grievance 571
..........................................................................................................................................................................
10
The link from polarization to conﬂict is proposed by Esteban and Ray (1999) and Reynal-Querol
(2000) and is common in the popular literature.
11
Our data source was Atlas Narodov Mira, USSR (1964). The Esteban-Ray measure includes a coefﬁcient
d that denotes the degree of antagonism between two different ethnic groups. Obviously, in large
samples such as we are using this is not observed. Following Reynal-Querol (2000) we assume that the
distance between any two ethnic groups is unity whereas that within the group is zero, so that d has the
properties: d ¼ 1 if i 6¼ j and d ¼ 0 if i ¼ j.
12
For  ¼ 0 the polarization measure is equal to the Gini coefﬁcient.
may be most vulnerable if the largest ethnic group constitutes a small majority. We
term this ethnic dominance. In Table 2 we deﬁne it as occurring if the largest ethnic
group constitutes 45–90% of the population. On this deﬁnition it does not appear
important: it is as common in peace episodes as in conﬂict episodes.
The opening page of Sen’s On Economic Inequality (Sen, 1973) asserts that ‘the
relation between inequality and rebellion is indeed a close one’. The poor may rebel
to induce redistribution, and rich regions may mount secessionist rebellions to preempt
redistribution.13
We measure income inequality by the Gini coefﬁcient and
by the ratio of the top-to-bottom quintiles of income. We measure asset inequality
by the Gini coefﬁcient of land ownership. The data are from Deininger and Squire
(1996, 1998). Inequality is slightly higher prior to the conﬂict episodes.
2.4 Scale effects
Our measures of opportunity, such as primary commodity exports, income, and
school enrolment, are scaled by measures of country size. For given values of these
variables, opportunities should be approximately proportional to size. Grievance
might also increase with size: public choices diverge more from the preferences
of the average individual as heterogeneity increases.14
We are, however, able to
control for three aspects of heterogeneity: ethnic, religious and income diversity.
Empirically, the conﬂict episodes had markedly larger populations than the peace
episodes.
3. Regression analysis
As set out above, the proxies for opportunity and objective grievances are largely
distinct and so can be compared as two non-nested econometric models. There is,
however, no reason for the accounts to be exclusive and the aim of our econometric
tests is to arrive at an integrated model which gives an account of conﬂict risk in
terms of all those opportunities and grievances that are signiﬁcant.
We now attempt to predict the risk that a civil war will start during a ﬁve-year
episode, through a logit regression. Our dependent variable, civil war start, takes a
value of one if a civil war started during a ﬁve year episode (1965–69, . . . , 1994–99).
Episodes that were peaceful from the beginning until the end are coded zero.
Ongoing wars are coded as missing observations as to not conﬂate the analysis
of civil war initiation and duration.15
If a war ended and another one started in the
same period we coded these events as one. Some of our explanatory variables are
572 p. collier and a. hoeffler
..........................................................................................................................................................................
13
This is analogous to the theory of tax exit proposed by Buchanan and Faith (1987).
14
Mounting diversity is the offset to scale economies in the provision of public goods in the model of
optimal county size proposed by Alesina and Spolaore (1997).
15
This approach contrasts with our initial work (Collier and Hoefﬂer, 1998) in which we used a tobit
procedure to study the duration of civil war (on a much inferior data set) and argued that the same
factors that determined duration would determine the risk of initiation. Collier et al. (2004) establishes
that this is wrong: initiation and duration are radically different processes.
time invariant. Those that are not are measured either for the ﬁrst year of the
period (e.g. 1965) or during the preceding ﬁve years (e.g. growth during 1960–64)
in order to avoid endogeneity problems. Our results rest on how societies that
experienced an outbreak of war differed from those that sustained peace.
We start with the opportunity model (see Table 3). The ﬁrst regression (column 1)
excludes per capita income and diasporas. Because per capita income and enrollment
greed and grievance 573
Table 3 Opportunity model
1 2 3 4 5
Primary commodity 18.149 18.900 16.476 17.567 17.404
exports/GDP (6.006)*** (5.948)*** (5.207)*** (6.744)*** (6.750)***
(Primary commodity À27.445 À29.123 À23.017 À28.815 À28.456
exports/GDP)2
(11.996)*** (11.905)*** (9.972)** (15.351)* (15.366)*
Post-coldwar À0.326 À0.207 À0.454
(0.469) (0.450) (0.416)
Male secondary schooling À0.025 À0.024
(0.010)** (0.010)**
Ln GDP per capita À0.837 À1.237 À1.243
(0.253)*** (0.283)*** (0.284)***
GDP growth À0.117 À0.118 À0.105
(0.044)*** (0.044)*** (0.042)***
Peace duration À0.003 À0.004*** À0.004 À0.002 À0.002
(0.002)
p ¼ 0.128
(0.001) (0.001)*** (0.001) (0.001)
Previous war 0.464
(0.547)
p ¼ 0.396
Mountainous terrain 0.013 0.014 0.008
(0.009)
p ¼ 0.164
(0.009) (0.008)
Geographic dispersion À2.211 À2.129 À0.865
(1.038)** (1.032)** (0.948)
Social fractionalization À0.0002 À0.0002 À0.0002
(0.0001)
p ¼ 0.109
(0.0001)
p ¼ 0.122
(0.0001)**
Ln population 0.669 0.686 0.493 0.295 0.296
(0.163)*** (0.162)*** (0.129)*** (0.141)** (0.141)**
Diaspora/peace 700.931
(363.29)**
Diaspora corrected/peace 741.155
(387.636)*
(Diaspora-diaspora 823.941
corrected)/peace (556.024)
N 688 688 750 595 595
No of wars 46 46 52 29 29
Pseudo R2
0.24 0.24 0.22 0.25 0.25
Log likelihood À128.49 À128.85 À146.86 À93.27 À93.23
Notes: All regressions include a constant. Standard errors in parentheses. ***, **, * indicate signiﬁcance at
the 1, 5, and 10% level, respectively.
in secondary schooling are highly correlated, they cannot be used in the same
regression ( ¼ 0.8). Our diaspora measure is available only for 29 war episodes
and so we explore it as an addendum. The variables included in the ﬁrst regression
permit a sample of 688 episodes (from 123 countries), including 46 wars.
Primary commodity exports are highly signiﬁcant. Although their effect is nonlinear,
the risk of conﬂict peaks when they constitute around 33% of GDP,16
which
is a high level of dependence. The other proxy for ﬁnance, the end of the Cold War,
has the expected sign but is insigniﬁcant. The foregone earnings proxies are also
both signiﬁcant with the expected sign: secondary schooling and growth both
reduce conﬂict risk. Our proxy for the cost of conﬂict-speciﬁc capital is the
number of months since any previous conﬂict (back to 1945). To distinguish
between this interpretation and the danger that the proxy might be picking up
ﬁxed effects, we add a dummy variable that is unity if there was a previous conﬂict
post-1945. Our proxy has the expected sign and is on the borderline of signiﬁcance,
while the dummy variable is completely insigniﬁcant. When the dummy variable is
dropped (column 2) the proxy becomes highly signiﬁcant and no other results are
changed. The proxies for military advantage also have the expected sign and are
marginally signiﬁcant: mountainous terrain, population dispersion and social fractionalization.
Finally, the coefﬁcient on population is positive and highly signiﬁcant.
The third column replaces secondary schooling with per capita income. This
permits a larger sample—750 episodes (from 125 countries) including 52 wars.
Per capita income is highly signiﬁcant with the expected negative sign. However,
the change of variable and the expansion of sample have little effect on the other
results—social fractionalization becomes signiﬁcant and population dispersion
loses signiﬁcance. There is little to chose between these two variants of the
model—secondary schooling gives a slightly better ﬁt, but per capita income permits
a slightly larger sample.
In the last two columns of Table 3 we introduce our diaspora variable. Since
many observations are missing, the number of war episodes with complete data is
radically reduced. In order to preserve sample size we therefore retreat to a more
parsimonious version of the model. We drop four sample-constraining peripheral
explanatory variables: social fractionalization, population dispersion, mountainous
terrain, and the rate of growth in the previous episode. The remaining explanatory
variables are thus per capita GDP, primary commodity exports, population, and the
number of months since the previous conﬂict. Even with these data-restoring
deletions, the sample size is reduced to 29 war episodes (and 595 observations).
However, all the included explanatory variables remain signiﬁcant. The size of the
diaspora is not directly signiﬁcant in the initiation of conﬂict (column 4). However,
it is signiﬁcant when interacted with the number of months since the previous
conﬂict. ‘Diaspora/peace’ divides the size of the diaspora by the time since a
574 p. collier and a. hoeffler
..........................................................................................................................................................................
16
We differentiate the probability of civil war with respect to primary commodity exports and ﬁnd that
the risk is at its maximum at 33% of primary exports in GDP (18.149/(2*27.445) ¼ 0.33).
previous conﬂict. The variable is positive and signiﬁcant: a large diaspora considerably
increases the risk of repeat conﬂict.
While this result may indicate that diasporas increase the risk of conﬂict through
their ﬁnance of rebel organizations, it is also open to a more anodyne interpretation.
Diasporas are endogenous to the intensity of conﬂict: when civil war occurs,
people emigrate to the USA. Hence, the size of the diaspora might be proxying the
intensity of conﬂict. The result may therefore be spurious: intense conﬂicts may
have a higher risk of repetition. To test for this we decomposed observed diasporas
into a component which is exogenous to the intensity of conﬂict and a residual
endogenous component. For this decomposition we estimated a migration model,
reported in Appendix 1. The size of the diaspora in a census year is predicted to be
a function of its size in the previous census, time, per capita income in the country
of origin, and whether there was a war in the intervening period. This model
predicts the size of the diaspora with reasonable accuracy. For years subsequent
to a conﬂict we replace the actual data on the size of the diaspora with an estimate
from this regression. Thus, all post-conﬂict observations of diasporas are estimates
which are purged of any effect of the intensity of conﬂict. The difference between
actual and estimated ﬁgures is then used as an additional variable, measuring that
part of the diaspora which is potentially endogenous to the intensity of conﬂict.
Both of these measures are then introduced into the regression in place of the
previous single measure of the diaspora. The results are reported in the ﬁnal
column of Table 3. The purged measure of the diaspora remains signiﬁcant, and
the size of the coefﬁcient is only slightly altered (it is not signiﬁcantly different from
that on the endogenous diaspora measure). This suggests that there is indeed a
substantial causal effect of the diaspora on the risk of conﬂict renewal. The result
also guides our interpretation of why the risk of conﬂict repetition declines as peace
is maintained. Recall that in principle this could be either because hatreds gradually
fade, or because ‘rebellion-speciﬁc capital’ gradually depreciates. How might diasporas
slow these processes? Diasporas preserve their own hatreds: that is why they
ﬁnance rebellion. However, it is unlikely that the diaspora’s hatreds signiﬁcantly
inﬂuence attitudes among the much larger population in the country of origin.
By contrast, the ﬁnance provided by the diaspora can offset the depreciation of
rebellion-speciﬁc capital, thereby sustaining conﬂict risk.
In Table 4 we turn to objective grievance as the explanation of rebellion, dropping
all the economic measures of opportunity.17
We retain the number of months
since a previous conﬂict, since (subject to our discussion above) this can be interpreted
as proxying fading hatreds. In the ﬁrst column we also exclude the inequality
measures due to considerations of sample size. This enables a very large sample of
850 episodes and 59 civil wars.
greed and grievance 575
..........................................................................................................................................................................
17
We retain the two geographic measures, population dispersion and mountainous terrain. Although
their exclusion does not affect the results, non-economists often ﬁnd the proposition that geographic
opportunity affects conﬂict plausible and inoffensive, while contesting the role of economic opportunity.
We retain population size as a scale variable.
The four proxies for ethnic and religious tension are surprisingly unimportant in
view of the attention that the phenomenon attracts. Ethnic fractionalization is
signiﬁcant at 10% with the expected sign. Religious fractionalization and polarization
are insigniﬁcant with the wrong sign, and ethnic dominance is insigniﬁcant.
Nor are the three measures jointly signiﬁcant.18
Democracy is highly signiﬁcant
with the expected sign—repression increases conﬂict risk. The time since the
previous conﬂict is again highly signiﬁcant, but we have suggested that this is
more likely to be proxying rebellion-speciﬁc capital than grievance. In the
second and third columns we introduce income inequality and land inequality
respectively. Although the sample size is reduced, it is still substantial—over 600
576 p. collier and a. hoeffler
..........................................................................................................................................................................
18
At this stage we measure polarization with  ¼ 1.6 and deﬁne ethnic dominance as occurring when the
largest ethnic group constitutes 45–90% of the population. These speciﬁcations are justiﬁed in Section 4
where we investigate robustness to alternative deﬁnitions.
Table 4 Grievance model
1 2 3
Ethnic fractionalization 0.010 0.011 0.012
(0.006)* (0.007)* (0.008)
Religious fractionalization À0.003 À0.006 À0.004
(0.007) (0.008) (0.009)
Polarization  ¼ 1.6 À3.067 À4.682 À6.536
(7.021) (8.267) (8.579)
Ethnic dominance (45–90%) 0.414 0.575 1.084
(0.496) (0.586) (0.629)*
Democracy À0.109 À0.083 À0.121
(0.044)*** (0.051)* (0.053)**
Peace duration À0.004 À0.003 À0.004
(0.001)*** (0.001)*** (0.001)***
Mountainous terrain 0.011 0.007 À0.0001
(0.007) (0.009) (0.009)
Geographic dispersion À0.509 À0.763 À1.293
(0.856) (1.053) (0.102)
Ln population 0.221 0.246 0.300
(0.096)** (0.119)** (1.133)**
Income inequality 0.015
(0.018)
Land inequality 0.461
(1.305)
N 850 604 603
No of wars 59 41 38
Pseudo R2
0.13 0.11 0.17
Log likelihood À185.57 À133.46 À117.12
Notes: All regressions include a constant. Standard errors in parentheses. ***, **, * indicate signiﬁcance at
the 1, 5 and 10% level, respectively.
Column 1: the two measures of fractionalization and ethnic dominance are not jointly signiﬁcant.
episodes of which 41 (income) and 38 (land) are wars. Neither variable is close to
signiﬁcance.19
All three grievance models have very low explanatory power with a
pseudo R2
of 0.17 or lower.
We now turn to the question of which model, opportunity or grievance, provides
a better explanation of the risk of civil war. Since the two models are non-nested,
i.e. one model is not a special case of the other, we use the J-test as suggested by
Davidson and MacKinnon (1981). As shown in the ﬁrst two columns of Table 5, we
ﬁnd that we cannot reject one model in favor of the other.20
Thus, we conclude that
while the opportunity model is superior, some elements of the grievance model are
likely to add to its explanatory power. We therefore investigate the combination of
the two models as presented in column 3 of Table 5.
Since this combined model includes income inequality and a lagged term,
our sample size is much reduced (479 observations). In column 4 we drop inequality
(which is consistently insigniﬁcant). Omitting inequality increases the sample
size to 665. In this combined model neither democracy, ethnic and religious
fractionalization nor the post-Cold War dummy are signiﬁcant. Other variables
are statistically signiﬁcant or close to signiﬁcance and the overall ﬁt is reasonable
(pseudo R2
of 0.26). Since both the grievance and opportunity models
are nested in the combined model, we can use a likelihood ratio test to determine
whether the combined model is superior. We can reject the validity of
the restrictions proposed by the grievance model, but not by the opportunity
model.21
Although the combined model is superior to the opportunity and grievance
models, several variables are completely insigniﬁcant and we drop them sequentially.
First we exclude the post-Cold War dummy, then religious fractionalization,
greed and grievance 577
..........................................................................................................................................................................
19
We also tried the ratio of the income shares of the top to the bottom quintiles. This was also
insigniﬁcant.
20
The J-test is based on the following artiﬁcial nesting procedure. First we explain the risk of civil war, p,
in terms of the two different models, opportunity and grievance.
(1) p ¼ f(opportunity)
(2) p ¼ f(grievance)
Based on these logit regressions we calculate the predicted probabilities and add these predicted values,
^ppopportunity
and ^ppgrievance
to our alternative models.
(1) p ¼ f(opportunity, ^ppgrievance
)
(2) p ¼ f(grievance, ^ppopportunity
)
According to the J-test the signiﬁcance of the coefﬁcients of these added variables enables us to choose
between the two different models. If ^ppgrievance
is signiﬁcant in the opportunity model we reject the
opportunity model in favor of the grievance model. If ^ppopportunity
is signiﬁcant in the grievance model
we reject the grievance model in favor of the opportunity model. As can be seen in columns 1 and 2 of
Table 5, ^ppgrievance
is signiﬁcant in the opportunity model and ^ppopportunity
is signiﬁcant in the grievance
model.
21
Using the same sample as for the combined model (n ¼ 665) we obtain the following results:
Opportunity model versus combined model, 5 degrees of freedom, Likelihood Ratio Test (LRT) statistic
7.85 (p ¼ 0.165); grievance model versus combined model, 6 degrees of freedom, LRT statistic 29.64
(p ¼ 0.000).
578 p. collier and a. hoeffler
Table5Combinedopportunityandgrievancemodel
1234567
Primarycommodity19.10737.07223.38518.93716.77350.608
exports/GDP(5.996)***(10.293)***(6.692)***(5.865)***(5.206)***(14.09)***
(Primarycommodityexports/GDP)2
À30.262À69.270À36.335À29.443À23.800À131.00
(12.008)***(21.697)***(12.998)***(11.781)***(10.040)**(42.93)***
Post-coldwarÀ0.208À0.873À0.281
(0.457)(0.644)(0.459)
MalesecondaryschoolingÀ0.021À0.029À0.022À0.031À0.034
(0.011)**(0.013)**(0.011)**(0.010)***(0.011)***
LnGDPpercapitaÀ0.950
(0.245)***
(GDPgrowth)tÀ1À0.108À0.045À0.108À0.115À0.098À0.113
(0.044)***(0.062)(0.045)**(0.043)***(0.042)**(0.046)***
PeacedurationÀ0.00030.0005À0.0003À0.003À0.004À0.004À0.003
(0.002)(0.0014)(0.0015)(0.001)***(0.001)***(0.001)***(0.001)***
Mountainousterrain0.0050.0010.0050.015
(0.010)(0.008)(0.012)(0.009)
p¼0.11
GeographicdispersionÀ1.9760.053À4.032À1.962À2.487À0.992À2.871
(1.049)*(1.101)(1.490)***(1.149)*(1.005)**(0.909)(1.130)***
Lnpopulation0.489À0.0220.9270.6970.7680.5101.123
(0.193)**(0.136)(0.250)***(0.181)***(0.166)***(0.128)***(0.226)***
SocialfractionalizationÀ0.0002À0.0008À0.0005À0.0002À0.0002À0.0003
(0.0001)***(0.0003)**(0.0003)(0.0001)**(0.0001)***(0.0001)***
p¼0.11
Ethnicfractionalization0.0080.0410.023
(0.007)(0.019)**(0.015)
ReligiousfractionalizationÀ0.0050.0150.014
(0.008)(0.020)(0.019)
PolarizationÀ9.338À25.276À15.992
greed and grievance 579
(8.734)(13.390)*(10.518)
Ethnicdominance(45–90%)1.2102.0201.5920.6700.4800.769
(0.648)*(0.915)**(0.746)**(0.354)*(0.328)
p¼0.14
(0.369)**
DemocracyÀ0.036À0.018À0.042
(0.054)(0.062)(0.054)
Incomeinequality0.025
(0.024)
Grievancepredictedvalue0.765
(0.413)*
Opportunitypredictedvalue1.044
(0.211)***
Primarycommodityexports/GDP*
oildummy
À28.275
(9.351)***
(Primarycommodityexports/GDP)2
*
oildummy
106.459
(38.704)***
N665665479665688750654
Noofwars46463246465245
PseudoR2
0.240.250.240.260.240.220.30
LoglikelihoodÀ126.69À125.29À89.55À124.60À128.21À146.84À114.20
Notes:Allregressionsincludeaconstant.Standarderrorsinparentheses.***,**,*indicatesigniﬁcanceatthe1,5,and10%level,respectively.
then democracy,22
then polarization, then ethnic fractionalization and ﬁnally
mountainous terrain, yielding the baseline model of column 5 and its variant
with per capita income replacing secondary enrolment in column 6. No further
reduction in the model is accepted, and no additions of variables included in our
previous models are accepted. The baseline model and its variant yield very similar
results although the variant has less explanatory power and two variables lose
signiﬁcance (ethnic dominance and geographic dispersion).
Our baseline model allows us to calculate the change in the probability of warstarts
for different values of the explanatory variables. We present these calculations
in Appendix Table A2. At the mean of all variables the risk of a war-start is about
11.5%. Our model predicts that a hypothetical country with all the worst characteristics
found in our sample would have a near-certain risk of war, while one with
all the best characteristics would have a negligible risk. We now calculate how each
variable affects the risk of civil war (keeping all other variables at their mean
values).
The effect of primary commodity exports on conﬂict risk is both highly signiﬁcant
and considerable. At peak danger (primary commodity exports being 33% of
GDP), the risk of civil war is about 22%, while a country with no such exports has
a risk of only 1%. The effect is sufﬁciently important to warrant disaggregation
into different types of commodities. We categorized primary commodity exports
according to which type of product was dominant: food, non-food agriculture, oil,
other raw materials, and a residual category of ‘mixed’.23
Of the many potential
disaggregations of primary commodity exports permitted by this data, only one
was signiﬁcant when introduced into our baseline regression, namely oil versus
non-oil. The results are reported in column 7. We add variables that interact the
primary commodity export share and its square with a dummy variable that takes
the value of unity if the exports are predominantly oil. Both variables are signiﬁcant:
oil exports have a distinct effect on the risk of conﬂict. However, the effect is
modest: at the average value of primary commodity exports oil has the same effect
as other commodities. Low levels of oil dependence are somewhat less risky than
other commodities and high levels of dependence are somewhat more risky. The
disaggregation slightly reduces the sample size, does not change the signiﬁcance of
any of the other variables, and substantially improves the overall ﬁt of the model.24
Recall that the other proxies for ﬁnancial opportunities, the Cold War and
diasporas, are not included in this baseline. The end of the Cold War does not
580 p. collier and a. hoeffler
..........................................................................................................................................................................
22
We tried different speciﬁcations to test for the effect of political repression by investigating non-linear
effects, by including the autocracy score instead of the democracy score, and by using the difference
between the two variables as suggested by Londregan and Poole (1996). We also tried the Freedom
House measure of political freedom, but neither of these alternative political repression measures were
found to be signiﬁcant.
23
We would like to thank Jan Dehn for providing us with the data that enabled this disaggregation.
24
Furthermore, using data from Dehn (2000) we investigated whether contemporaneous export price
changes altered the risk of conﬂict. We could not ﬁnd any evidence to support this hypothesis.
appear to have had a signiﬁcant effect. Diasporas are excluded from the baseline
purely for considerations of sample size. In the parsimonious variant in which they
are included, their effect on the risk of repeat conﬂict is substantial: after ﬁve years
of peace, switching the size of the diaspora from the smallest to the largest found in
post-conﬂict episodes increases the risk of conﬂict six-fold.
The proxies for earnings foregone have substantial effects. If the enrollment rate
for secondary schooling is ten percentage points higher than the average the risk of
war is reduced by about three percentage points (a decline in the risk from 11.5%
to 8.6%). An additional percentage point on the growth rate reduces the risk of war
by about one percentage point (a decline from 11.5% to 10.4%). Our other proxy
for the cost of rebellion is also highly signiﬁcant and substantial. Directly after a
civil war there is a high probability of a re-start, the risk being about 32%. This risk
declines over time at around one percentage point per year.
The only measures of rebel military advantage that survive into the baseline are
population dispersion and social fractionalization. Consistent with Herbst’s
hypothesis, countries with a highly concentrated population have a very low risk
of conﬂict, whereas those with a highly dispersed population have a very high risk
(about 37%). Consistent with the hypothesis that cohesion is important for rebel
effectiveness, social fractionalization makes a society substantially safer: a maximally
fractionalized society has a conﬂict risk only one quarter that of a homogenous
society.
Only one of the proxies for grievance survives into the baseline regression,
namely ethnic dominance. If a country is characterized by ethnic dominance its
risk of conﬂict is nearly doubled. Thus, the net effect of increased social diversity is
the sum of its effect on social fractionalization and its effect on ethnic dominance.
Starting from homogeneity, as diversity increases the society is likely to become
characterized by ethnic dominance, although this will be reversed by further
increases in diversity. The risk of conﬂict would ﬁrst rise and then fall. Note that
while these measures in combination are superﬁcially similar to the hypothesized
effect of polarization, our measure of polarization itself is insigniﬁcant.
Finally, the coefﬁcient on the scale variable, population, is highly signiﬁcant and
close to unity: risk is approximately proportional to size. We have suggested that
proportionality is more likely if conﬂict is generated by opportunities than by
grievances.
4. Robustness checks
We now test these baseline results for robustness. We consider the sensitivity both
to data and to method. With respect to data, we investigate the effect of outlying
observations, and of different deﬁnitions of the dependent and independent variables.
With respect to method, we investigate random effects, ﬁxed effects and rare
events bias.
We investigate outlying observations using two different methods. First, we
inspect the characteristics of the 46 conﬂict episodes used in the baseline regression
greed and grievance 581
and second, we use a systematic analysis of inﬂuential data points. Since our sample
is unbalanced as between events and non-events, the potential problems of outliers
arises predominantly among the 46 conﬂict episodes. Of these conﬂict episodes, 24
were ﬁrst-time conﬂicts and 22 were repeat conﬂicts.
First, the classiﬁcation of events in Romania in 1989, and in Iran in 1974, 1978,
and 1981 as civil wars is in various respects questionable. They are, on our analysis
highly atypical of conﬂict episodes. Both had secondary school enrolments much
higher than the other conﬂict episodes, and Iran also had an atypically high
primary commodity export share. In Table 6 column 1 we drop these doubtful
observations. No results are overturned, but the performance of the regression
improves and all variables are now signiﬁcant at the 1% or 5% level.
There are four observations of highly negative growth: Angola in 1970–74, Zaı¨re
(now the Democratic Republic of the Congo) in 1990–95, Iran in 1975–79 and Iraq
in 1980–84. All of these growth collapses appear to be genuine, and they occur in
different countries. We now check whether the result that the growth rate affects
conﬂict risk is dependent upon these four observations, deleting them along with
Iran and Romania (Table 6, column 2). Growth remains signiﬁcant, and its coefﬁcient
is only slightly reduced. Hence, we can conclude that the increased risk of
conﬂict due to slow growth is not conﬁned to episodes of growth collapse, but is a
more continuous relationship.
We next analyse whether our regression results are sensitive to the inclusion of
inﬂuential data points. Based on the methods developed by Pregibon (1981)25
we
examined which observations may be inﬂuential and investigated whether omitting
these observations from our baseline model changed our results. We ﬁnd three
inﬂuential observations: Congo 1995–99, Iran 1970–74, and Romania 1985–89.
However, when we omitted these three observations from our regression, the overall
ﬁt of the regressions improved (from previously R2
¼ 0.24 to R2
¼ 0.29) and all
of the coefﬁcients remain statistically signiﬁcant. (Table 6, column 3).
We now investigate the possibility that a few countries with a high commodity
export ratio account for the non-monotonic relationship to conﬂict risk. This
might imply that the reduction in conﬂict risk only occurred at extreme values
of commodity dependence. Four peaceful countries have particularly high values
of primary dependence: Saudi Arabia, Guyana, Oman, and Trinidad and Tobago.
In Table 6 column 4 we present our baseline model excluding these four high
primary commodity exporters. The non-monotonic relationship between primary
commodity exports and the risk of conﬂict remains signiﬁcant, as do all other
results.
We next turn to questions of variable deﬁnition. The most contentious aspect of
the dependent variable is distinguishing between whether a country has a single
long war or multiple shorter wars interrupted by periods of peace. In the above
analysis we have been guided by the judgement of the political scientists who built
582 p. collier and a. hoeffler
..........................................................................................................................................................................
25
Long (1997) pp.98–101 provides a discussion of inﬂuence in limited dependent variable models.
the original data sets. Some peace periods are, however, quite short and it might be
better to conceptualize these as interludes in a single war. We ﬁrst reclassiﬁed all
those wars that were separated by peace periods of less than one month as continuous
wars (Table 6, column 5). The baseline results are not altered by this
redeﬁnition. We then reclassiﬁed those wars separated by less than a year as continuous
wars (Table 6, column 6). The only result to be affected is that the growth
rate becomes marginally insigniﬁcant (p ¼ 0.12), although its coefﬁcient is little
changed.26
We investigated how robust our results are to the deﬁnition of ethnic dominance
and social fractionalization. In the baseline we deﬁne ethnic dominance as the
largest ethnic group constituting 45–90% of the population. We investigate other
deﬁnitions that either vary the range of the population or use the share of the
largest group regardless of its size. As the range is changed from 45–90% the
signiﬁcance level and the coefﬁcient are both reduced, while if the deﬁnition is
changed more radically to being the population share of the largest group it is
completely insigniﬁcant. We also ﬁnd that ‘social fractionalization’, our measure of
cross-cutting cleavages, dominates the other possible aggregation procedures for
ethnic and religious diversity. When this measure of fractionalization is included
with the ethnic and religious diversity indices either together or individually, it is
signiﬁcant whereas the underlying indices are not signiﬁcant.
In the baseline we use only the most extreme measure of polarization over the
range proposed by Esteban and Ray (1994). However, if this measure is replaced by
either the lower bound ( ¼ 0), or the central measure ( ¼ 0.8) the results are
unaffected: polarization remains insigniﬁcant and the other variables remain signiﬁcant.
We also experimented with the alternative measure proposed by ReynalQuerol
(2002), and with the number of ethnic groups, but with the same result.27
In Table 7 we investigate a number of different estimation issues. We concentrate
on the analysis of random effects, ﬁxed effects, time effects, and a correction for
rare events. We re-estimated our models using random effects. For the baseline
model we ﬁnd that the panel data estimator is not different from the pooled
estimator, i.e. we accept the hypothesis that we can pool across the observations.28
The estimation of ﬁxed effects logits was only possible on a very small sub-sample
of the observations. The countries for which the dependent variable does not vary
over time (the majority of countries experienced only peace) cannot be included in
greed and grievance 583
..........................................................................................................................................................................
26
We also examined the effect of time since the previous conﬂict in more detail by including the natural
logarithm of the peace variable or its square, however, a linear decay term provides a better ﬁt. Note that
the measure of peace since the end of the civil war is somewhat imprecise since we only measure it from
the end of the war to the initial year of each sub-period. A duration model of post-war peace would allow
a more detailed analysis of this peace effect, however, the duration model results in Collier et al. (2004)
support the results presented in this paper.
27
All of these robustness checks are presented in Collier and Hoefﬂer (2002).
28
A LRT provides a 2
statistic of 0 (p ¼ 0.998). Thus, we cannot reject the null-hypothesis that the panel
data and pooled estimator provide the same results.
584 p. collier and a. hoeffler
Table6Robustnesschecks
123456
ExcludingExcludingExcludingExcludingPeaceperiodsPeaceperiods
IranandIranandRomaniainﬂuentialhighprimaryofshorterthanofshorterthan
Romaniaandgrowthdatapointscommodityonemonthareoneyearare
collapsesexporterstreatedastreatedas
continuouswarscontinuouswars
Primarycommodityexports/GDP19.69619.02928.74518.77119.14722.686
(6.608)***(6.671)***(7.862)***(6.063)***(5.939)***(6.718)***
(Prim.com.exports/GDP)2
À34.090À33.250À59.818À28.466À30.150À39.053
(14.356)**(14.609)**(17.781)***(12.299)**(12.031)***(14.405)***
MalesecondaryschoolingÀ0.035À0.037À0.041À0.031À0.031À0.031
(0.011)***(0.011)***(0.011)***(0.010)***(0.010)***(0.010)***
(GDPgrowth)tÀ1À0.140À0.100À0.137À0.122À0.102À0.071
(0.047)***(0.052)**(0.046)***(0.044)***(0.044)***(0.047)
PeacedurationÀ0.004À0.003À0.004À0.004À0.003À0.003
(0.001)***(0.001)***(0.0011)(0.001)***(0.001)***(0.001)***
GeographicdispersionÀ2.114À2.272À2.890À2.449À2.541À2.953
(1.080)**(1.090)**(1.136)***(1.008)**(1.012)***(1.049)***
greed and grievance 585
SocialfractionalizationÀ0.0002À0.0002À0.0003À0.0002À0.0002À0.0002
(0.0001)**(0.0001)**(0.0001)***(0.0001)**(0.0001)**(0.0001)***
Ethnicdominance0.7270.7320.6550.6470.7320.741
(0.368)**(0.370)**(0.372)*(0.354)*(0.357)**(0.362)**
Lnpopulation0.7470.7430.8990.7720.7820.832
(0.174)***(0.175)***(0.195)***(0.168)***(0.167)***(0.176)***
N674671685662686683
No.ofwars423943464441
PseudoR2
0.250.220.290.210.230.21
LoglikelihoodÀ118.40À116.17À114.04À122.23À126.33À122.23
Notes:Allregressionsincludeaconstant.Standarderrorsinparentheses.***,**,*indicatesigniﬁcanceatthe1,5,and10%level,respectively.
Column2:Weexcludethefollowingthreegrowthcollapses:Angola1970–74,Iraq1980–84,andZaire1990–94.
Column3:Weexcludethefollowingthreeinﬂuentialdatapoints:Iran1970–74,Romania1985–89,Congo1995–99.
Column4:WeexcludethecountrieswiththehighestprimarycommodityexporttoGDPratio,namelySaudiArabia,Guyana,Oman,andTrinidadandTobago.Their
averageprimarycommodityexporttoGDPratiois0.504(sampleaverage0.158).
Column5:Weexcludethefollowingwarstarts:Angola1975andSomalia1988.
Column6:Weexcludethefollowingwarstarts:Angola1975,Mozambique1976,SierraLeone1997,Somalia1988andZaı¨re/DemocraticRep.ofCongo1997.
the analysis. Although the ﬁxed effects test is very severe, the non-monotonic effect
of primary commodity exports remains signiﬁcant. Were the effect of primary
commodity exports dependent only upon cross-section data, it might suggest
that the variable was proxying some other characteristic such as geography.
586 p. collier and a. hoeffler
Table 7 Estimation issues
1 2 3 4
Random
effects
Fixed effects Pooled logit plus
time dummies
Rare events logit
Primary commodity
exports/GDP
18.937
(5.865)***
35.850
(14.436)***
18.895
(5.988)***
17.161
(6.535)***
(Primary commodity
exports/GDP)2
À29.443
(11.782)***
À65.967
(26.964)***
À29.815
(12.098)***
À25.594
(14.355)*
Male secondary schooling À0.032 0.007 À0.031 À0.029
(0.010)*** (0.033) (0.010)*** (0.010)***
(GDP growth) t-1 À0.115 À0.045 À0.129 À0.110
(0.043)*** (0.072) (0.047)*** (0.040)***
Peace duration À0.004 0.011 À0.004 À0.004
(0.001)*** (0.002)*** (0.001)*** (0.001)***
Geographic dispersion À2.487 115.363 À2.447 À2.394
(1.005)*** (74.562) (1.018)** (1.085)**
Social fractionalization À0.0002 À0.007 À0.0002 À0.0002
(0.0001)** (0.006) (0.0001)** (0.0001)**
Ethnic dominance
(45–90%)
0.670
(0.354)*
0.682
(0.359)*
0.644
(0.336)*
Ln population 0.768 0.010 0.762 0.726
(0.166)*** (1.410) (0.170)*** (0.151)***
T70-74 0.725
(0.602)
T75-79 0.578
(0.608)
T80-84 1.137
(0.602)*
T85-89 À0.013
(0.757)
T90-94 0.802
(0.677)
T95-99 À0.492
(0.921)
N 688 145 688 688
No of wars 46 44 46 46
Pseudo R2
0.26
Log likelihood À128.21 À38.18 À124.30
Notes: All regressions include a constant. Standard errors in parentheses. ***, **, * indicate signiﬁcance at
the 1, 5, and 10% level, respectively
However, the ﬁxed effects regression uses only changes in primary commodity
dependence, and so reduces the scope for alternative interpretations.29
We analysed whether time effects matter by including time dummies in the
model. Based on a log likelihood ratio test we cannot reject the hypothesis that
the time dummies are zero.30
Finally, in the last column of Table 7 we use a recently developed correction
method for rare events data (King and Zeng, 2001). The event we predict (war)
occurs in only about 7% of our observations. King and Zeng show that standard
logit estimation tends to underestimate the probability of rare events. We therefore
used their correction procedure. The differences between the standard logit results
and the rare events corrected results are negligible with all variables signiﬁcant at
the same levels. The mean of the predicted probabilities obtained from the rare
events logit regression is 0.072. Thus, we ﬁnd that the corrected results are very
similar to the logit results.
We examined a number of different model speciﬁcations. We found that none of
the following geographic and demographic characteristics were signiﬁcant: forest
coverage, population density and the proportion of young men aged 15 to 29.31
We
also investigated the potential endogeneity of income to civil war. Evidently since
we are measuring income prior to war the endogeneity only arises if a country has
more than one war. Since the ﬁrst war will have reduced income, for subsequent
wars the correlation between income and war could in principle reﬂect this reverse
causation. To control for this we re-estimated excluding repeat wars. The income
variable remained highly signiﬁcant.
5. Interpretation and conclusion
Using a comprehensive data set of civil wars over the period 1960–99 we used logit
regressions to predict the risk of the outbreak of war in each ﬁve-year episode. We
ﬁnd that a model that focuses on the opportunities for rebellion performs well,
greed and grievance 587
..........................................................................................................................................................................
29
We also investigated the effect of commodity prices. Since prices are exogenous, they can be entered
contemporaneous with the episode being predicted, whereas our value-based proxy has to be lagged. We
experimented with both the level of export prices and with the change in prices from the previous
period. However, in either form when added to the baseline regression the variable was insigniﬁcant. The
fact that lagged values of exports are signiﬁcant even in the ﬁxed effects regression suggests that rebels do
respond to changes in values, but the response is evidently not so rapid as to give rise to an in-period
price response. Potentially, the effect on conﬂict risk captures the ‘voracity effect’ predicted by Lane and
Tornell (1999) whereby an increase in the price of a natural resource export would induce more than the
increment in value to be devoted to conﬂict. Our results suggest that there may be such an effect but that
it is lagged.
30
The LRT statistic is 7.83, 6 restrictions (p ¼ 0.251).
31
The proportion of the population living in urban areas was statistically signiﬁcant when we excluded
the geographic concentration of the population. However, when we included both proxies for the
concentration of the population, the geographic concentration measure remained statistically signiﬁcant
while the proportion of the population living in urban areas was marginally insigniﬁcant (p ¼ 0.11).
whereas objective indicators of grievance add little explanatory power. The model is
robust to a range of tests for outliers, redeﬁnitions, and alternative speciﬁcations.
One factor inﬂuencing the opportunity for rebellion is the availability of ﬁnance.
We have shown that primary commodity exports substantially increase conﬂict
risk. We have interpreted this as being due to the opportunities such commodities
provide for extortion, making rebellion feasible and perhaps even attractive. An
alternative explanation would be that primary commodity dependence worsens
governance and so generates stronger grievances. However, we are controlling for
economic performance—the level, growth, and distribution of income—and for
political rights (which appear not to affect the risk of conﬂict). While we would not
wish to discount the possibility of an effect working through corruption (for which
we cannot control), there is plenty of case study evidence supporting the extortion
interpretation. Another source of ﬁnance for which there is good case study
evidence is diasporas. We have found that diasporas substantially increase the
risk of conﬂict renewal, and it is hard to ﬁnd an alternative explanation for this
result.
A second factor inﬂuencing opportunity is the cost of rebellion. Male secondary
education enrollment, per capita income, and the growth rate all have statistically
signiﬁcant and substantial effects that reduce conﬂict risk. We have interpreted
them as proxying earnings foregone in rebellion: low foregone earnings facilitate
conﬂict. Even if this is correct, low earnings might matter because they are a source
of grievance rather than because they make rebellion cheap. However, if rebellion
were a protest against low income, we might expect inequality to have strong
effects, which we do not ﬁnd.
A third aspect of opportunity is military advantage. We have found that a
dispersed population increases the risk of conﬂict, and there is weaker evidence
that mountainous terrain might also advantage rebels. It remains possible that these
are correlated with unmeasured grievances.
Most proxies for grievance were insigniﬁcant: inequality, political rights, ethnic
polarization, and religious fractionalization. Only ‘ethnic dominance’—one ethnic
group being a majority—had adverse effects. Even this has to be considered in
combination with the benign effects of social fractionalization: societies characterized
by ethnic and religious diversity are safer than homogenous societies as long as
they avoid dominance. We have suggested that diversity makes rebellion harder
because it makes rebel cohesion more costly. It would be difﬁcult to argue that
diversity reduced grievance.
Finally, the risk of conﬂict is proportional to a country’s population. We have
suggested that both opportunities and grievances increase with population. Thus,
the result is compatible with both the opportunity and grievance accounts.
However, grievances increase with population due to rising heterogeneity. Yet
those aspects of heterogeneity that we are able to measure are not associated
with an increased risk of conﬂict. Hence, a grievance account of the effect of
population would need to explain why unobserved, but not observed, heterogeneity
increases conﬂict risk.
588 p. collier and a. hoeffler
One variable, the time since a previous conﬂict, has substantial effects: time
heals. Potentially, this can be interpreted either as opportunity or grievance. It
may reﬂect the gradual depreciation of rebellion-speciﬁc capital, and hence an
increasing cost of rebellion, or the gradual erosion of hatred. However, we have
found that a large diaspora slows the ‘healing’ process. The known proclivity of
diasporas to ﬁnance rebel groups offsets the depreciation of rebellion-speciﬁc
capital, and so would be predicted to delay ‘healing’. The diaspora effect thus
lends support to the opportunity interpretation.
Opportunity as an explanation of conﬂict risk is consistent with the economic
interpretation of rebellion as greed-motivated. However, it is also consistent with
grievance motivation as long as perceived grievances are sufﬁciently widespread
to be common across societies and time. Opportunity can account for the existence
of either for-proﬁt, or not-for-proﬁt, rebel organizations. Our evidence does
not therefore imply that rebels are necessarily criminals. But the grievances that
motivate rebels may be substantially disconnected from the large social concerns
of inequality, political rights, and ethnic or religious identity.
Acknowledgements
We would like to thank Todd Sandler and two anonymous referees for very helpful
comments. Anke Hoefﬂer gratefully acknowledges funding from the Research Council of
Norway (NFR).
References
Alesina, A., Baqir, R., and Easterly, W. (1999). ‘Public goods and ethnic divisions’, Quarterly
Journal of Economics, 114, 1243–84.
Alesina, A., Oetzler, S., Roubini, N., and Swagel, P. (1996). ‘Political instability and economic
growth’, Journal of Economic Growth, 1, 189–211.
Alesina, A. and Spolaore, E. (1997). ‘On the number and the size of nations’, Quarterly
Journal of Economics, 112, 1027–56.
Angoustures, A. and Pascal, V. (1996). ‘Diasporas et ﬁnancement des conﬂits’, in F. Jean and
J.-C. Ruﬁn (eds), Economie des Guerres Civiles, Hachette, Paris.
Barro, R.J. (1991). ‘Economic growth in a cross section of countries’, The Quarterly Journal
of Economics, 106, 407–43.
Barro, R.J. (ed.) (1997). Determinants of Economic Growth, MIT Press, Cambridge, MA.
Barrett, D.B. (ed.) (1982). World Christian Encyclopedia, Oxford University Press, Oxford.
Buchanan, J.M. and Faith, R.L. (1987). ‘Secession and the limits of taxation: towards a
theory of internal exit’, American Economic Review, 77, 1023–31.
Collier, P. (2000). ‘Rebellion as a quasi-criminal activity’, Journal of Conﬂict Resolution, 44,
839–53.
Collier, P. (2001). ‘Ethnic diversity: an economic analysis of its implications’, Economic
Policy, 32, 129–66.
Collier, P. and Hoefﬂer, A. (1998). ‘On the economic causes of civil war’, Oxford Economic
Papers, 50, 563–73.
greed and grievance 589
Collier, P. and Hoefﬂer, A. (2002). ‘Greed and grievance in civil war’, CSAE Working Paper,
WPS 2002–01, Oxford University.
http://www.economics.ox.ac.uk/CSAEadmin/workingpapers/pdfs/2002–01text.pdf.
Collier, P., Hoefﬂer, A., and So¨derbom, M. (2004). ‘On the duration of civil war.’ Journal of
Peace Research, 41, 253–73.
Davidson R. and MacKinnon, J.G. (1981). ‘Several tests for model speciﬁcation in the
presence of alternative hypotheses’, Econometrica, 49, 781–93.
Dehn, J. (2000). ‘Commodity price uncertainty, investment and shocks: implications for
economic growth’, D.Phil. thesis, University of Oxford.
Deininger, K. and Squire, L. (1996). ‘A new data set measuring income inequality’, World
Bank Economic Review, 10, 565–91.
Deininger, K. and Squire, L. (1998). ‘New ways of looking at old issues: inequality and
growth’, Journal of Development Economics, 57, 249–87.
Easterly, W. and Levine, R. (1997). ‘Africa’s growth tragedy: policies and ethnic divisions.’
Quarterly Journal of Economics, 113, 203–49.
Esteban, J.-M. and Ray, D. (1994). ‘On the measurement of polarization.’ Econometrica, 62,
819–51.
Esteban, J.-M. and Ray, D. (1999). ‘Conﬂict and distribution’, Journal of Economic Theory,
87, 379–415.
Figes, O. (1996). A People’s Tragedy: The Russian Revolution 1891–1924, Pimlico, London.
Fearon, J. (2004) ‘Why do some civil wars last so much longer than others?’ Journal of Peace
Research, 41, 275–301.
Gerrard, A.J.W. (2000). ‘What is a mountain?’, mimeo, DECRG, World Bank,
Washington, DC.
Grossman, H.I. (1991). ‘A general equilibrium model of insurrections’, American Economic
Review, 81, 912–21.
Grossman, H.I. (1999). ‘Kleptocracy and revolutions’, Oxford Economic Papers, 51, 267–83.
Hegre, H., Ellingsen, T., Gates, S., and Gleditsch, N.-P. (2001). ‘Toward a democratic civil
peace? Democracy, political change, and civil war, 1816–1992’, American Political Science
Review, 95, 33–48.
Herbst, J. (2000). States and Power in Africa, Princeton University Press, Princeton, NJ.
Hirshleifer, J. (1995). ‘Theorizing about conﬂict’, in K. Hartley and T. Sandler (eds),
Handbook of Defense Economics, Vol. 1, Elsevier Science, Amsterdam, 165–89.
Hirshleifer, J. (2001). The Dark Side of the Force: Economic Foundations of Conﬂict Theory,
Cambridge University Press, Cambridge.
Jaggers, K. and Gurr, T.R. (1995). ‘Tracking democracy’s third wave with the Polity III data’,
Journal of Peace Research, 32, 469–82.
King, G. and Zeng, L. (2001). ‘Logistic regression in rare events data’, Political Analysis, 9,
137–63.
Klare, M.T. (2001). Natural Resource Wars: The New Landscape of Global Conﬂict,
Metropolitan Books, New York.
Lane A. and Tornell, P.R. (1999). ‘The voracity effect’, American Economic Review, 89,
22–46.
590 p. collier and a. hoeffler
Londregan, J.B. and Poole, K.T. (1996) ‘Does high income promote democracy?’, World
Politics, 49, 1–30.
Long, J.S. (1997). Regression Models for Categorical and Limited Dependent Variables,
Sage Publications, London.
Mauro, P. (1995). ‘Corruption and growth’, The Quarterly Journal of Economics, 110,
681–712.
Pregibon, D. (1981). ‘Logistic regression diagnostics’, The Annals of Statistics, 9, 705–24.
Reynal-Querol, M. (2000). ‘Religious conﬂict and growth: theory and evidence’. Ph.D. thesis,
London School of Economics and Political Science.
Reynal-Querol, M. (2002). ‘Ethnicity, political systems and civil war’, Journal of Conﬂict
Resolution, 46, 29–54.
Sachs, J. and Warner, A.M. (2000). ‘Natural resource abundance and economic growth’,
in G.M. Meier and J.E. Rauch (eds), Leading Issues in Economic Development, 7th ed., Oxford
University Press, Oxford.
Sambanis, N. (2002). ‘What is a civil war? Conceptual and empirical complexities of an
operational deﬁnition’, mimeo, Yale University.
Sen, A. (1973). On Economic Inequality, Clarendon Press, Oxford.
Singer, D.J. and Small, M. (1994). ‘Correlates of war project: international and civil war data,
1816–1992’, data ﬁle, Inter-University Consortium for Political and Social Research, Ann
Arbor, MI.
Small, M. and Singer, J.D. (1982). Resort to Arms: International and Civil War, 1816–1980,
Sage, Beverly Hills, CA.
The Stockholm International Peace Research Institute (2002). Yearbook of World
Armaments and Disarmaments, Oxford University Press, Oxford.
Summers, R. and Heston, A. (1991). ‘The Penn World Table (Mark 5): an expanded
set of international comparisons, 1950–1988’, The Quarterly Journal of Economics, 99,
327–68.
USSR (1964). Atlas Narodov Mira, Department of Geodesy and Cartography of the State
Geological Committee of the USSR. Moscow.
World Bank (2000). World Development Indicators, Washington DC, data ﬁle.
Appendix 1
1. A simple migration model
Our estimation of migration is based on the following model
diasit ¼1:163 Á diasit À 0:0002 Á lnGDPi;tÀ1 þ 0:003 Á wari;tÀ1 þ 0:003 Á T80 þ 0:005 Á T90þ 0:013
ð0:045ÞÃÃÃ
ð0:001ÞÃÃ
ð0:03Þ ð0:002Þ ð0:002Þ ð0:008Þ
Where dias denotes diaspora which is measured as the ratio of emigrants in the USA to the
total population of the country of origin. The variable war is a war dummy, measured at
t À 1 it takes a value of one if the country experienced a civil war in the previous period.
The method of estimation is OLS. The data is measured at the beginning of each decade,
greed and grievance 591
i.e. 1960, 1970, 1980, and 1990. The regression includes time dummies, T, which are jointly
signiﬁcant.
Based on this simple migration model we estimated the size of the diaspora at time t.
di^aasit ¼ xit Á ^
For countries which experienced a previous civil war we used these estimated values to
correct for a possible endogeneity problem. We replaced a total of 64 observations. For
countries which did not experience a civil war we use the actual diaspora data. In order
to obtain values for 1965 we took the averages of this corrected diaspora data measured in
1960 and 1970, and analogously for the values for 1975 and 1985. For 1995 we use the
observations measured in 1990.
Appendix 2
1. Calculating the marginal probabilities
In our regressions we estimate the probability of a war breaking out during a ﬁve-year
period, and the model can be written in the following general form
Yit ¼ a þ bXit þ cMi;tÀ1 þ dZi þ uit ðA2:1Þ
where t and i are time and country indicators. The dependent variable is a dummy variable
indicating whether a war broke out during the ﬁve-year period, so that Yit is the log odds of
war. The explanatory variables are either measured at the beginning of the period (for
example, income per capita, primary commodity exports/GDP, population), or during the
previous ﬁve-year period (for instance, per capita income growth, or are time invariant or
changing slowly over time (for example, social fractionalization).
The expected probability ^ppit of a war breaking out can be calculated by using the estimated
coefﬁcients obtained from equation (A1.1):
^aa þ ^bbXit þ ^ccMi;tÀ1 þ ^ddZi ¼ ^WWit ðA2:2Þ
^ppit ¼
e
^WWit
ð1 þ e ^WWit Þ
Á 100 ðA2:3Þ
Appendix 3
1. Data sources
Democracy The degree of openness of democratic institutions is measured on a scale of
zero (low) to ten (high). Source: http://www.cidcm.umd.edu/polity/index.html. The data are
described in Jaggers and Gurr (1995).
592 p. collier and a. hoeffler
greed and grievance 593
TableA2Marginalprobabilities
VariableCoefﬁcientMean(1)(2)(3)(4)(5)(6)(7)(8)(9)
ofXAttheWorstBestprimary10%extra1%extraMin.Max.Ethnic
meancommodity/meningrowthpeacefraction-dominance
GDP¼0.33schoolalization
Primarycommodity
exports/GDP
18.9370.1582.9926.06006.0602.9922.9922.9922.9922.992
(Primarycommodity
exports/GDP)2
À29.443À0.735À3.0150À3.015À0.735À0.735À0.735À0.735À0.735
MalesecondaryschoolingÀ0.03244.489À1.406À0.032À4.645À1.406À1.722À1.406À1.406À1.406À1.406
(GDPgrowth)tÀ1À0.1151.618À0.1861.508À1.660À0.186À0.186À0.417À0.186À0.186À0.186
PeacedurationÀ0.004347.5À1.286À0.004À2.19À1.286À1.286À1.286À0.004À1.286À1.286
GeographicdispersionÀ2.4870.602À1.4970.000À2.415À1.497À1.497À1.497À1.497À1.497À1.497
SocialfractionalizationÀ0.00021790À0.376À0.004À1.465À0.376À0.376À0.376À0.376À1.465À0.376
Ethnicdominance
(45–90%)
0.6700.4390.2940.67000.2940.2940.2940.2940.2940.670
Lnpopulation0.76830,500,00013.23016.0499.13613.23013.23013.23013.23013.23013.230
ConstantÀ13.073À13.073À13.073À13.073À13.073À13.073À13.073À13.073À13.073À13.073
^WWÀ2.0438.160À16.312À1.255À2.359À2.273À0.761À3.132À1.667
^pp0.1151.0000.0000.2220.0860.0930.3180.0420.159
Diaspora We used the data on the foreign born population from the U.S. Bureau of
the Census and divided these numbers by the total population in the country of origin.
http://www.census.gov/population/
Ethnic dominance Using the ethno-linguistic data from the original data source (USSR,
1964) we calculated an indicator of ethnic dominance. This variable takes the value of one
if one single ethno-linguistic group makes up 45 to 90% of the total population and zero
otherwise. We would like to thank Tomila Lankina for the translation of the original data
source.
Forest coverage We used the FAO measure of the proportion of a country’s terrain which
is covered in woods and forest. Source: http://www.fao.org/forestry
GDP per capita We measure income as real PPP adjusted GDP per capita. The primary
data set is the Penn World Tables 5.6 (Summers and Heston, 1991). Since the data are only
available from 1960–92 we used the growth rates of real PPP adjusted GDP per capita data
from the World Bank’s World Development Indicators 1998 in order to obtain income data
for the 1990s. These GDP per cpaita data were used to calculate the average annual growth
rate over the previous ﬁve years.
Geographic dispersion of the population We constructed a dispersion index of the
population on a country by country basis. Based on population data for 400 km2
cells we
generated a Gini coefﬁcient of population dispersion for each country. A value of 0 indicates
that the population is evenly distributed across the country and a value of 1 indicates that the
total population is concentrated in one area. Data is available for 1990 and 1995. For years
prior to 1990 we used the 1990 data. We would like to thank Uwe Deichman of the World
Bank’s Geographic Information System Unit for generating this data. He used the following
data sources: Center for International Earth Science Information Network (CIESIN),
Columbia University; International Food Policy Research Institute (IFPRI); and World
Resources Institute (WRI). 2000. Gridded Population of the World (GPW), Version 2.
Palisades, NY: IESIN, Columbia University. Available at http://sedac.ciesin.org/plue/gpw.
Inequality Inequality was either measured as income inquality (source: Deininger and
Squire, 1996) or as inequality in land ownership (source: Deininger and Squire, unpublished).
Both inequality measures are provided as a Gini coefﬁcient.
Male secondary school enrolment rates We measure male secondary school enrolment
rates as gross enrolment ratios, i.e. the ratio of total enrollment, regardless of age, to the
population of the age group that ofﬁcially corresponds to the level of education shown.
Secondary education completes the provision of basic education that began at the primary
level, and aims at laying the foundations for lifelong learning and human development,
by offering more subject- or skill-oriented instruction using more specialized teachers.
Source: World Bank Development Indicators, 1998.
Mountainous terrain The proportion of a country’s terrain which is mountainous was
measured by John Gerrard, a physical geographer specialized in mountainous terrain. His
measure is based not just on altitude but takes into account plateaus and rugged uplands.
The data are presented in Gerrard (2000).
594 p. collier and a. hoeffler
Peace duration This variable measures the length of the peace period (in months) since
the end of the previous civil war. For countries which never experienced a civil war we
measure the peace period since the end of World War II.
Population Population measures the total population, the data source is the World Bank’s
World Development Indicators 1998.
Primary commodity exports/GDP The ratio of primary commodity exports to GDP
proxies the abundance of natural resources. The data on primary commodity exports and
GDP were obtained from the World Bank. Export and GDP data are measured in current
US dollars.
Social, ethnolinguistic, and religious fractionalization We proxy social fractionalization
in a combined measure of ethnic and religious fractionalization. Ethnic fractionalization is
measured by the ethno-linguistic fractionalization index. It measures the probability that two
randomly drawn individuals from a given country do not speak the same language. Data are
only available for 1960. In the economics literature this measure was ﬁrst used by Mauro
(1995). Using data from Barrett (1982) on religious afﬁliations we constructed an analogous
religious fractionalization index. Following Barro (1997) we aggregated the various religious
afﬁliations into nine categories: Catholic, Protestant, Muslim, Jew, Hindu, Buddhist, Eastern
Religions (other than Buddhist), Indigenous Religions, and no religious afﬁliation.
The fractionalization indices range from zero to 100. A value of zero indicates that the
society is completely homogenous whereas a value of 100 would characterize a completely
heterogeneous society.
We calculated our social fractionalization index as the product of the ethno-linguistic
fractionalization and the religious fractionalization index plus the ethno-linguistic or the
religious fractionalization index, whichever is the greater. By adding either index we
avoid classifying a country as homogenous (a value of zero) if the country is ethnically
homogenous but religiously divers, or vice versa.
War data A civil war is deﬁned as an internal conﬂict in which at least 1,000 battle related
deaths (civilian and military) occurred per year. We use mainly the data collected by Small
and Singer (1992) and according to their deﬁnitions (Singer and Small, 1984) we updated
their data set for 1992–99.
greed and grievance 595