ArticlePDF Available

The inequality-conflict nexus re-examined: Income, education and popular rebellions

Authors:

Abstract

The impact of inequality on the outbreak of intrastate armed conflicts or civil wars has recently attracted considerable interest in conflict research. In contrast to previous studies that have focused on inequality in the total population (vertical inequality), recent studies have analysed inequality between certain groups of people (horizontal inequality), and found that inequality significantly increases the likelihood of conflict onset. However, most of the recent studies on the inequality-conflict nexus have focused on conflicts fought between ethnic groups. The relation between inequality and other (non-ethnic) categories of conflicts has attracted less attention. The present study aims to address this gap: it implements a theoretical and empirical analysis of the relation between inequality and popular rebellions, a subset of conflicts where mobilization transcends ethnic boundaries and hostilities involve popular participation. Based on a sample of 77 popular rebellions and new global data on vertical inequality in income and education, this study shows that inequality significantly increases the likelihood of popular rebellion onset. In addition, the study reveals that inequality proxies (income and education Gini indices) outperform proxies of the absolute level of income (GDP per capita) in the model of popular rebellion onset, suggesting that it is relative, not absolute, well-being that ultimately motivates people to rise up in arms.
1
The inequality-conflict nexus re-examined: Income, education and
popular rebellions
Henrikas Bartusevičius, Department of Political Science and Government, Aarhus University
Manuscript accepted for publication in the Journal of Peace Research. The final, definitive
version of this paper has been published Journal of Peace Research 51(1), Jan/2014 by SAGE
Publications Ltd, All rights reserved. © Henrikas Bartusevičius. Link to the published version:
doi: 10.1177/0022343313503179
Abstract
The impact of inequality on the outbreak of intrastate armed conflicts or civil wars has recently
attracted considerable interest in conflict research. In contrast to previous studies that have
focused on inequality in the total population (vertical inequality), recent studies have analysed
inequality between certain groups of people (horizontal inequality), and found that inequality
significantly increases the likelihood of conflict onset. However, most of the recent studies on
the inequality-conflict nexus have focused on conflicts fought between ethnic groups. The
relation between inequality and other (non-ethnic) categories of conflicts has attracted less
attention. The present study aims to address this gap: it implements a theoretical and empirical
analysis of the relation between inequality and popular rebellions, a subset of conflicts where
mobilization transcends ethnic boundaries and hostilities involve popular participation. Based on
a sample of 77 popular rebellions and new global data on inequality in income and education,
this study shows that inequality significantly increases the likelihood of popular rebellion onset.
In addition, the study reveals that inequality proxies (income and education Gini indices)
outperform proxies of the absolute level of income (GDP per capita) in the model of popular
rebellion onset, suggesting that it is relative, not absolute, well-being that ultimately motivates
people to rise up in arms.
Keywords: income inequality, educational inequality, civil war onset, ethnic conflict, non-ethnic
conflict, popular rebellion
Corresponding author: henrikas@ps.au.dk
2
Introduction
Important developments have taken place recently in the study of intrastate armed conflicts or
civil wars (hereafter conflicts). The renewed focus on the relation between inequality and
outbreak of conflict is one of these (Østby, 2008). In contrast to previous research that has
focused on the distribution of resources in the total population (vertical inequality), recent studies
have focused on the distribution of resources among certain groups of people (horizontal
inequality) (Stewart, 2002; 2008). The recent studies have found that inequality significantly
increases the likelihood of conflict onset (Cederman, Weidmann & Gleditsch, 2011; Østby,
2008), a finding that stands in contrast to a number of previous studies that have largely
dismissed the role of inequality in conflict (e.g., Collier & Hoeffler, 2004; Fearon & Laitin,
2003).
However, most of the recent studies of the inequality-conflict nexus have focused on
conflicts fought between ethnic groups.1 The relation between inequality and other (non-ethnic)
categories of conflicts has attracted less attention. Non-ethnic and ethnic conflicts are almost
equally common – out of 331 conflict onsets since the end of the Second World War, 144 were
non-ethnic. Non-ethnic and ethnic conflicts are also almost equally protracted on average, the
former last four and a half years and the latter five and a half years. Finally, and perhaps most
importantly, non-ethnic and ethnic conflicts are almost equally violent – 53% of non-ethnic
conflicts and 51% of ethnic conflicts cross the threshold of 1000 battle-related deaths.2 One
could, therefore, argue that an exclusive focus on ethnic conflicts is not warranted.
The present study addresses this gap by analysing a global sample of popular rebellions –
armed conflicts where mobilization transcends ethnic boundaries and hostilities involve popular
participation. It employs two recently introduced and previously unexploited datasets on vertical
income and educational inequality: the Standardized World Income Inequality Database (Solt,
2009) and the Data Set of Educational Inequality in the World, 1950–2010 (Benaabdelaali,
Hanchane & Kamal, 2012). The study demonstrates that inequality significantly predicts popular
rebellion onset. In addition, the study reveals that proxies of inequality consistently outperform
1 Exceptions include Tadjoeddin, Chowdhury & Murshed (2012), who have analysed the relation between vertical
inequality and non-ethnic routine violence. Several studies have also analysed the relation between spatial
inequalities and conflict onset (Østby, Nordås & Rød, 2009) and intensity (Murshed & Gates, 2005).
2 These numbers are based on Bartusevičius (2013) and include conflicts recorded in 19462010 (see the sub-section
‘Outcome Variable’).
3
proxies of the absolute level of income – the effect of GDP per capita becomes insignificant
when income and educational inequality are controlled for in the model of rebellion onset. This
finding challenges the widely-established ‘opportunity’ approach (Collier & Hoeffler, 2004) and
corroborates the theory of relative deprivation (Gurr, 1970), suggesting that it is relative, not
absolute, well-being that ultimately motivates people to rise up in arms.
The study proceeds as follows. The next section reviews previous research on the inequality-
conflict nexus, elaborates on the concepts of popular rebellion and inequality, and describes
potential mechanisms through which inequality leads to popular rebellion. The subsequent
section presents the empirical analysis. The final section summarizes the main results and
discusses their probable implications, followed by a conclusion offering suggestions for future
research.
The inequality-conflict nexus
Does unequal distribution of resources increase the risk of conflicts? This question has worried
conflict researchers ever since Bruce Russet published his Inequality and Instability (1964). The
results of the early research on the inequality-conflict nexus were mixed: the studies found a
positive relationship between inequality in income (or land tenure) and conflict (Nagel, 1976;
Prosterman, 1976; Russet, 1964; Sigelman & Simpson, 1977), no relationship (Hardy, 1979;
Weede, 1981, 1987), a negative relationship with conflicts most likely in egalitarian societies
(Mitchell, 1968) and a concave (inverted-U) with conflicts most likely at intermediate levels of
inequality (Nagel, 1974).
With the end of the Cold War, the focus of conflict researchers shifted to other variables such
as ethnic diversity (Ellingsen, 2000), natural resources (Collier & Hoeffler, 2004), economic
prosperity (Fearon & Laitin, 2003) and regime type (Hegre et al., 2001). Nevertheless, some
studies in the 1990s and 2000s analysed the role of inequality as well. Once again, findings were
mixed: Alesina & Perotti (1996) and Auvinen & Nafziger (1999) found a positive relationship
between inequality and conflict, while Hegre, Gissinger & Gleditsch (2003) found no such
relationship. On the whole, quantitative (aggregate) cross-country research has thus failed to
establish a robust relation between inequality and conflict.
Meanwhile, case-based research has demonstrated that sharp inequalities have had a major
4
impact on the outbreak of a number of present-day conflicts (e.g., Booth, 1991; Stewart, 2008).
Some scholars have argued that the non-findings of the cross-country studies are caused by
methodological flaws such as measurement error, omitted variable bias or poor quality of
inequality data (Lichbach, 1989; Sambanis, 2005). Further, previous research has mainly used
proxies of inequality in the total population (i.e., vertical inequality), which does not necessarily
overlap with inequality between particular groups:
In practice, a country can have large income inequalities between groups (His), despite the fact that
the overall (vertical) income inequality is rather low (as is the case in Rwanda), and vice versa; a
country can have a high vertical inequality score, even though the structural differences between
groups might be low (e.g. Brazil). Besides, a country can have both strong vertical and horizontal
inequality at the same time (e.g. South Africa), or it can score low on both (e.g. Switzerland) (Østby,
2011: 9).
This suggests that vertical and horizontal inequalities potentially have non-uniform effects on
ethnic and non-ethnic conflicts. Non-ethnic conflicts, unlike ethnic ones, transcend ethnic
boundaries. Such conflicts often involve participation of non-ethnically differentiated masses
whose share of resources is closely related to the overall distribution of resources, which – as
stated above – does not necessarily apply to ethnic groups. Thus, as Sambanis puts it:
There may exist a relationship between [vertical] inequality and popular revolutions or class conflict,
which is another reason to consider disaggregating the cases of civil war. But ethnic or secessionist
wars should, in theory, be driven more by group-based inequality (which I refer to here as horizontal
inequality) than by interpersonal inequality. High levels of interpersonal inequality in all ethnic
groups may actually reduce the ability to coordinate an ethnic rebellion as they can erode group
solidarity (2005: 328).
In fact, Besançon (2005) has found that vertical income inequality is positively related to the
onset of (non-ethnic) ‘revolutions’, but negatively to the onset of ‘ethnic wars’, while vertical
educational inequality is negatively (though insignificantly) related to the former, but positively
to the latter. This suggests that the study of the inequality-conflict nexus must consider the
distinction between horizontal and vertical inequalities on the one hand and ethnic and non-
5
ethnic conflicts on the other. Recently, considerable attention has been focused on horizontal
inequalities and ethnic conflicts (Cederman, Weidmann & Gleditsch, 2011; Østby, 2008),
whereas the relation between vertical inequalities and non-ethnic conflicts has largely been
neglected. The present study aims to address this gap it implements an analysis of the relation
between inequality and one particular (non-ethnic) category of conflict – popular rebellion.
Popular rebellions
Popular rebellions constitute a significant proportion of present-day conflicts.3 The Salvadoran
Civil War (1979–1992) is a good example of these. Like all (broadly-defined) intrastate armed
conflicts or civil wars (Gleditsch et al., 2002), the rebellion in El Salvador involved domestic
armed hostilities between politically organized actors, one of which was the government of a
state, included effective resistance and resulted in certain number of battle-related deaths. What
made this rebellion distinct from other categories of conflicts was (1) the composition of
conflicting parties and (2) popular participation.
Unlike ethnic conflicts, the rebellion in El Salvador was not limited to particular ethnic
groups – the rebels did not represent any specific religion or language. This suggests that
conflicts like these are fought over issues that affect the whole (or significant parts of) society,
not its particular segments as is often the case in ethnic conflicts. For comparison, consider the
conflict between Nationalists/Catholics and Loyalists/Protestants in the United Kingdom. The
conflict was largely limited to the territory of Northern Ireland and involved hostilities between
ethnically-defined actors. The conflict did not submerge the majority of British society and
concerned specific issues related to the status of Northern Ireland and its Protestant and Catholic
communities. In contrast, the conflict in El Salvador – in one or other form – engulfed significant
part of the country’s territory and involved participation of heterogeneous groups united in
common struggle against the government.
Further, the rebellion in El Salvador, unlike a typical violent coup, was not limited to the
elitist struggle among the incumbents and involved popular participation.4 For contrast, compare
this rebellion with the 1951 Manhattan Rebellion in Thailand. Whereas the former involved
3 The term ‘popular rebellion’ is chosen to emphasize the distinctive attributes of the concept non-ethnic
mobilization and popular participation and thus to distinguish it from violent coups and ethnic rebellions.
4 Popular participation is often used as a main criterion to distinguish armed conflicts from violent coups (e.g.,
Powell & Thyne, 2011).
6
active popular participation, the latter was a 36-hour military confrontation between mutinied
navy officers on the one hand and army and police officers sided with the government on the
other (Fineman, 1997). Though a number of civilians were killed (due to an undisciplined use of
force), the conflict was limited to the elitists struggle, and – unlike in El Salvador – involved no
active popular participation.
Thus, using the rebellion in El Salvador as an illustration5, we could define popular rebellion
as an intrastate armed conflict between two or more politically organized actors, one of which is
the government of a state, where hostilities involve: (a) confrontation between members of the
same ethnic group; (b) popular participation; (c) effective resistance; and (d) certain levels of
violence. 6 ‘Popular participation’ does not necessarily imply involvement of the whole
population (nor does it imply a widespread popular support). Rather, it implies open participation
of various socioeconomic groups (workers, peasants, students, etc.) in armed hostilities, whose
members played no direct role in the state apparatus during the period prior to the onset of
conflict.
Inequality
The fact that popular rebellions transcend ethnic boundaries and involve popular participation
suggests that they are fought over issues that affect significant parts of societies, not their
particular segments vertical inequality (as opposed to horizontal inequality between particular
groups) likely being one of these. Hereby, inequality is defined as an unequal (asymmetric)
distribution of certain goods within a given society. The goods over which inequalities exist are
defined broadly; they can include material assets (e.g., income or land), but also opportunities
(e.g., to participate in politics) or access (e.g., to education, health or social services). The
criterion for including a good in the definition employed in this study is its widely-acknowledged
value and the possession/lack of which significantly influences individuals’ socioeconomic
status. This criterion helps to distinguish goods whose possession/lack is potentially linked to the
individuals’ predisposition to be involved in collective violence from those goods whose
possession/lack does not influence such a predisposition (see below).
5 Other examples of popular rebellions include the Guatemalan Civil War (19601996) or the Peruvian Civil War
(19802000) (a full list is provided in an Online Appendix) (measurement issues are discussed below).
6 Thus, the concept of popular rebellion is the concept of a broadly-defined intrastate armed conflict with limited
extension (delimited by two additional, secondary level attributes a and b) (Goertz, 2006). In other words, popular
rebellion is just a sub-category of a broadly defined intrastate armed conflict.
7
This concept of inequality reflects the distribution of certain goods in the total population.
Therefore, it primarily deals with vertical inequality. But indirectly, it also accounts for
inequality between groups. Individuals possessing a comparable share of certain goods (such as
income or land) typically form socioeconomic groups (or ‘social classes’), which – like other
social groups (e.g., ethnic groups) – share peculiar attributes such as geographic concentration
and common social identity (characteristics that play important role in the mobilization of
would-be rebels a point I elaborate below). While vertical inequality in a country cannot
directly account for the actual level of inequality between groups taking part in armed hostilities,
it indicates the distance between the ‘advantaged’ and the ‘disadvantaged’, who typically
represent different sides in non-ethnic conflicts (the government and the rebels).7, 8
What are the potential mechanisms through which inequality leads to popular rebellions?
Previous literature has identified at least two of these. The first relates to the behaviour of
solitary individuals (‘the psychological mechanism’) and the second to the formation and
behaviour of groups (‘the social mechanism’).
The psychological mechanism
High inequality generates ‘relative deprivation’ (Gurr, 1970; Runciman, 1966). ‘A person’,
Runciman writes,
is relatively deprived of X when (i) he does not have X; (ii) he sees some other person or persons,
which may include himself at some previous or expected time, as having X, (iii) he wants X, and (iv)
he sees it as feasible that he should have X (1966: 10).
Such a discrepancy between one’s wishes and capabilities could trigger aggressive response
because of a psychological mechanism summarized in the ‘frustration-aggression hypothesis’
(Dollard et al., 1939: 1): ‘the occurrence of aggressive behaviour always presupposes the
existence of frustration’ and ‘the existence of frustration always leads to some form of
7 As Lichbach notes: ‘Conflict protagonists in a society are often divided into two groups: the challenging groups,
i.e., the have-nots or the disadvantaged, who seek economic equality by attacking the status quo distribution of
resources; and the established groups, i.e., the haves or the advantaged, who perpetuate economic inequality by
defending the status quo distribution of resources’ (1989: 432).
8 Nevertheless, inequality in the total population is only treated here as a proxy of the inequality between conflicting
parties which, in some cases, may not accurately account for the actual inequality between them.
8
aggression’.9 Frustration is here defined as ‘an interference with the occurrence of an instigated
goal-response at its proper time in the behaviour sequence’ (Ibid.: 7) and aggression as ‘sequence
of behaviour, the goal-response to which is the injury of the person toward whom it is directed’
(Ibid.: 9).
Needless to say, deprivation of certain goods does not necessarily lead to frustration in every
individual. The chance of frustration depends on the value one attaches to the good one is
deprived of. Thus, frustration is only likely when individuals are deprived of widely-
acknowledged and highly valued goods or, as Gurr states, ‘goods and conditions of the life to
which they believe they are justifiably entitled’ (1968: 1104). Likewise, the presence of
frustration does not necessarily lead every individual to respond aggressively. The chance of
aggression depends on intensity of frustration – the more intense the frustration, the higher the
likelihood of aggressive response: ‘men who are frustrated have an innate disposition to do
violence to its source in proportion to the intensity of their frustrations’ (Gurr, 1970: 37).
Intensity of frustration depends on the value attached to the good one is deprived of and on the
scale of the deprivation itself. The scale of the deprivation (the distance between the
disadvantaged and the advantaged in respect of possession of certain goods) must be sufficiently
large to result in frustration intense enough to predispose one to commit violence.
The social mechanism
Individuals whose goals (they think they are rightfully entitled to) have been blocked are
predisposed to commit violence. The chance of violence rises as the deprivations increase. Not
only because of the intensity of the frustration itself, but also because of the increase in the
strength of the sense of ‘we-ness’.
‘We-ness’ (or social identity) refers to ‘social categorizations of self and others, self-
categories that define the individual in terms of his or her shared similarities with members of
certain social categories in contrast to other social categories’ (Turner et al., 1994: 454). The
social groups around which ‘we-ness’ develops are often described in terms of ‘race’, language
or religion. But it may also develop around other attributes such as share of income, occupation
9 This statement was later amended by Miller (1941). He noted that the frustration-aggression link should not be
understood in deterministic terms frustration does not necessarily lead to aggression, and aggression is not
necessarily the only response to frustration. For further elaboration of the frustration-aggression hypothesis, see
Rummel (1977). For an updated version of the hypothesis, see Berkowitz (1989).
9
or – more broadly – ‘social class’10 (Jackman & Jackman, 1983; Wright, 1997).11 Indeed, ‘class
has a subjective meaning that transcends the economic sphere and incorporates factors normally
associated with status groups’ (Jackman & Jackman, 1983: 41). This is not surprising since
members of socioeconomic groups – just like members of ethnic groups – tend to live around,
form social networks with and marry members of the same socioeconomic group (Argyle, 1994:
66–92).12
Social identity, therefore, is not only relevant in the context of ethnic conflicts but also in the
context of conflicts involving socioeconomic groups. It is relevant because it facilitates
mobilization of solitary individuals for collective action by increasing commitment and solidarity
among members of the same group (Hunt & Benford, 2004: 448). This in turn reduces the
likelihood of free riding and helps to overcome the collective action problem (Olson, 1965) – a
major obstacle in collective actions such as armed rebellions, where participants are exposed to
high risks. Indeed, case studies suggests that party activists, union workers, ‘revolutionaries’
whoever took the organizational role13used socioeconomic statuses of peasants, landless
workers, working-class, etc., as mobilizational devices in a number of rebellions (e.g., Booth,
1991).
The stronger the social identity, the easier it is for leaders to mobilize the group’s members
for collective action (Gurr, 2000: 66; Van Zomeren, Postmes & Spears, 2008). The strength of
the social identity significantly depends on the extent of collective grievances (Gurr, 2000: 68),
which, in turn, depends on the extent of inequalities: the higher the inequality, the stronger the
grievances.
This discussion thus suggests a linear relationship between inequality and onset of popular
rebellion: the higher the inequality – the higher the individuals’ predisposition to commit
10 Class identity is ‘the ways in which people consider themselves “members” of different classes…it constitutes one
of many ways in which people define what is salient about their lives and what differentiates them from others…it is
rooted in one’s personal history and in the ways in which that personal history is tied to the history of communities
and social groups (Wright, 1997: 495).
11 It is commonly assumed, however, that ethnic identities are stronger that the socioeconomic ones (Smith, 1991: 1
18; McPherson, Smith-Lovin & Cook, 2001), which potentially explains why ethnic conflicts are somewhat more
common that the non-ethnic ones.
12 It is important to note that the concept of ‘class’ and ‘class identity’ has been challenged in sociology see, most
notably, Pakulski & Waters (1996). For an overview of recent debates on class identity, see Bottero (2004) or
Crompton (2008).
13 I acknowledge the central role leadership plays in mobilization processes (Gurr, 2000: 7879); I bypass, however,
this factor in this study, as my focus on the state-level factors precludes empirical analysis of individual
characteristics. On the other hand, I suggest that social identity, itself irrespective of the leadership characteristics
constitutes a mobilizational advantage, which makes rebellions more likely.
10
violence; similarly, the higher the inequality, the higher the likelihood of successful mobilization
of solitary individuals for collective action. Thus,
H1: The higher the inequality, the higher the risk of popular rebellion onset, ceteris paribus.
Obviously, popular rebellion is not the only possible outcome of sharp inequalities. Popular
rebellion only represents an extreme on a scale that includes other, ‘more moderate’ responses
such as demonstration or protest. But popular rebellion certainly remains an option in the
repertoire of the disadvantaged as long as inequality results in intense resentment and the
distance between the disadvantaged and the advantaged remains large.
Likewise, inequality is not the only possible cause of popular rebellion. Rebellions, just like
other types of collective action, depend on a number of other variables and contextual factors.
Yet, inequality – as the theory suggests – is certainly one of the viable causes or (at least)
contributing variables to the onset of violence directed towards the source of inequality, and the
higher the inequality, the higher the chance that it develops into violent collective action.
It is likely, however, that inequality at high levels – may have opposite effects. Initiation of
conflict significantly depends on the would-be rebels’ chances of success, which depends on the
estimation of the would-be rebels’ strength vis-à-vis the government (Hendrix, 2010: 274).
Would-be rebels’ strength vis-à-vis the government, in turn, depends on the levels of inequality,
as inequality often benefits those in control of government and disadvantages those willing to
rebel (see fn. 7). The more well-off the government, the higher its ability to recruit, equip and
retain soldiers, and, conversely, the more well-off the would-be rebels, the higher their ability to
recruit, equip and retain fighters. Thus, inequality, at high levels, regardless of the intensity of
the discontent and strength of ‘we-ness’, may lead to inactivity on the part of the disadvantaged,
as military confrontation with a state could be deemed too costly (see also Mitchell, 1968).
In addition, social psychological research has suggested that peoples’ tendency to compare
themselves with others decreases as the perceived differences between themselves and the
reference individuals (or groups) increase. In his seminal work on social comparison, Festinger
writes:
A person does not tend to evaluate his opinions or his abilities by comparison with others who are
11
too divergent from himself. If some other person’s ability is too far from his own, either above or
below, it is not possible to evaluate his own ability accurately by comparison with this other
person. There is then a tendency not to make the comparison (Festinger, 1954: 120).
While Festinger’s ‘similarity hypothesis’ has been challenged in subsequent research (showing
that people do sometimes compare themselves with ‘different others’), peoples’ comparison with
‘similar others’ has been found to have a much stronger impact on their self-evaluation than
comparison with ‘different others’.14 As Crosby points out, ‘the wages of manual workers are
more important in determining whether or not miners feel entitled to a pay increase than are the
salaries of white-collar workers (1976: 95). Thus, inequalities – at high levels – may not
necessarily lead to higher levels of relative deprivation, as people may simply stop comparing
their own socioeconomic position with the ‘different others’ as the gap between the two gets too
large.15
The two arguments, then, lead to the following hypothesis:
H2: The relationship between inequality and popular rebellion is concave (inverted-U),
ceteris paribus.
Empirical analysis
Outcome variable
The hypotheses were tested in a standard country-year logit regression analysis covering 1961–
2009. The sample included all annual observations of states as defined by Gleditsch & Ward
(1999). The data on the outcome variable was taken from the Categorically Disaggregated
Conflict dataset (CDC) (Bartusevičius, 2013). The CDC classifies conflicts into four categories:
1) ethnic governmental, 2) ethnic territorial, 3) non-ethnic governmental and 4) non-ethnic
territorial. This study employed category 3 as a proxy of popular rebellions. Category 3 includes
14 For an extensive overview of the subsequent research and experimental evidence on Festinger’s similarity
hypothesis, see Wood (1989, especially pp. 233–238). For a more recent overview of research on social comparison
see Suls, Martin & Wheeler (2002).
15 Nagel (1974) was first to apply Festinger’s similarity hypothesis in the study of rebellions. He has found empirical
support for an inverted-U relation between inequality in landholdings and support for rebellion in South Vietnam
(however, he found no support for the inverted-U relation between inequality and political instability in a cross-
national setting).
12
all conflict onsets listed in UCDP/PRIO Armed Conflict Dataset v.4-2013, 1946 – 2012
(Gleditsch et al., 2002; Themnér & Wallensteen, 2013)16 that were: (a) fought between members
of the same ethnic group17 (b) and concerned government. The CDC contains 124 onsets of non-
ethnic governmental conflicts recorded in 1946–2009. As the time span of the present study was
restricted to 1961–2009, 24 cases of popular rebellions were dropped. Subsequently, the analysis
dropped 23 cases of non-ethnic governmental conflicts as cases of violent coups/coup attempts.18
The empirical analysis thus encompassed 77 onsets of popular rebellions (see the Online
Appendix for a full list).
Predictor variables
Inequality was proxied by two indices representing the distribution of income and educational
attainment in the total population: the Gini Index of Net Income Inequality (t-1) (hereafter
income Gini) and the Gini Index of Educational Inequality (t-1) (education Gini) (the two indices
were taken from the datasets described below). The Gini index represents the difference between
the ‘ideal’ cumulative distribution of certain resources where every person gets the same share
of the resources and the actual distribution of the resources. It is estimated as the ratio of the
area bounded by the Lorenz curve (representing the actual cumulative distribution of the
resources) and the diagonal line of equality (representing the ‘ideal’ cumulative distribution of
the resources) to the whole area under the line of equality. The values of the indices range
between 0 and 1, 0 indicates a perfect equality and 1 – a perfect inequality. Income Gini thus
16 Note, therefore, that the CDC employs the UCDP/PRIO definition of intrastate armed conflict (‘a contested
incompatibility which concerns government and/or territory where the use of armed force between two parties, of
which at least one is the government of a state, results in at least 25 battle-related deaths’ (Themnér, 2013: 1)) as a
base for the definition of the non-ethnic governmental category.
17 The CDC classifies ‘ethnic groups’ on the basis of ‘race’, language and religion. Therefore, the non-ethnic
governmental category includes conflicts that were fought between members of the same ‘race’, language and
religion. For further details on coding of ethnic/non-ethnic categories in the CDC (2013) see the Online Appendix.
18 Following the conceptual framework introduced above, I used popular participation as an operational difference
between popular rebellions and violent coups. Thus, conflicts limited to the elitist struggle within the state
apparatus were treated as coups, and conflicts involving popular participation as popular rebellions. I have not
applied any numeric threshold for popular participation; therefore, analysis includes several ‘borderline’ cases (e.g.,
Government of Bolivia vs. National Liberation Army, ID: 1), where popular participation was limited (from several
dozens to hundreds of individuals). The analysis also includes several cases of sectarian conflicts (e.g., Government
of Nigeria vs. Boko Haram, ID: 100). These cases could not have been excluded without additional ad hoc coding
rules. Note that sectarian conflicts and rebellions with limited popular participation constitute only a small fraction
of cases in the sample the vast majority were political in nature and involved extensive popular participation.
Including coups or removing cases with limited participation or sectarian conflicts does not significantly affect the
results of this study.
13
represents inequality in the net income in the total population and education Gini inequality in
the educational attainment (proxied by schooling years) in the total adult population (age 15 and
above). For further details on the construction of the indices see the original sources described
below.
While these indices capture only part of inequality, they represent inequality in goods whose
value is almost universally acknowledged and whose possession, in most cases, has a major
impact on individuals’ socioeconomic position, making these indices particularly suitable to
proxy for goods whose unequal distribution is likely to be linked to intensity of resentment and
strength of ‘we-ness’ among the disadvantaged.
Data for the indices was taken from two recently introduced datasets: the Standardized World
Income Inequality Database (Solt, 2009) (hereafter SWIID) and the Data Set of Educational
Inequality in the World, 1950–2010 (Benaabdelaali, Hanchane & Kamal, 2012) (DEIW). 19 The
datasets substantially improve previous time-series cross-sectional data on inequality. SWIID
encompasses 173 countries – the largest sample of countries included in publicly available
datasets on income inequality. For example, Deininger & Squire (1996) (hereafter DS) – the
most commonly used data source within conflict literature (Besançon, 2005; Collier & Hoeffler,
2004; Fearon & Laitin, 2003; Hegre, Gissinger & Gleditsch, 2003) – includes only 138 countries.
Even its update, the World Income Inequality Database (WIID) (UNU-WIDER, 2008),
encompasses significantly fewer (159) countries than SWIID. SWIID also has a larger time span
than the other datasets (1960–2009 compared to 1960–1996 in DS20 or 1960–2006 in WIID).
The number of observations in SWIID exceeds 4500, almost twice the number of observations in
DS (1996).21 Finally, and perhaps most importantly, SWIID contains standardized observations
for maximum comparability across countries. Whereas DS’s (and WIID’s) observations are
estimated based on different concepts of income (net, gross, household, individual, etc.)22,
SWIID’s observations are estimated using a common point of reference (see Solt, 2009, for
details on the data standardization), making the data particularly suitable for cross-country
comparisons.
19 To my knowledge, neither of these datasets has been used in previous large-N studies of conflict onset.
20 The full time span in DS is longer (1890–1996), but the vast majority of observations are only available for 1960
1996.
21 Note that out of more than 2600 observation in DS’s dataset, only 682 meet the standard (set up by the authors
themselves) of ‘high-quality’
22 See Atkinson & Brandolini (2001) for critical evaluation of DS’s dataset, which can also be extended to WIID.
14
The data for education Gini was taken from DEIW. Like SWIID, DEIW outperforms
previously introduced datasets on educational inequality. It is based on Barro & Lee’s (2010)
most recent version and contains time-series cross-sectional data on the distribution of
educational attainability in the total population of 146 countries for 1950–2010. For comparison,
consider Thomas, Wang & Fan (2003), the second best data source on educational inequality in
terms of coverage and time span, which encompasses 140 countries for 1960–2000, or Castello
& Domenech (2002) – the only data source on vertical educational inequality previously
employed in the conflict literature (Besançon, 2005) – which contains observations for just 108
countries for 1960–2000. The data in DEIW is quinquennial. To adjust it to the country year
regression (see below), I linearly imputed annual observations with ‘ipolate’ command in Stata.
Thus, the final dataset contains 6200 observations for the period 1961–2009.
Control variables
The analysis aimed to isolate the effect of inequality proxies on the outcome variable and not to
identify variables explaining most of the variation in the outcome variable. Therefore, the main
model was largely limited to the set of variables that were likely to be related to both inequality
and popular rebellion (i.e., likely confounders): regime type, the absolute level of income and
economic growth.
The relationship between inequality and popular rebellion could be confounded by regime
type. Democratic countries typically enjoy higher development levels (Boix & Stokes, 2003) and
have less income inequality than non-democratic countries (Muller, 1995). It has been argued
that democratic countries are also less likely to experience conflicts (Gurr, 2000). To control for
regime type effects, I introduced xpolity scores (t-1) (Vreeland, 2008) transformed to a positive
14-point scale (hereafter xpolity scores).
Economic literature has also suggested that relationship between regime type and inequality
could follow the pattern of an inverted-U (Acemoglu & Robinson, 1998). The inverted-U
relationship between regime type and conflict is also suggested in the conflict literature (Hegre et
al., 2001). To capture the likely non-linear pattern between regime type and popular rebellion, I
introduced xpolity scores squared (xpolity scores^2).
Further, economic literature has suggested that income inequality is negatively related to the
15
absolute level of income and economic growth (Deininger & Squire, 1998; Ravallion, 1997).23
Conflict researchers have also shown that the absolute level of income and economic growth are
negatively related to the outbreak of conflicts (Hegre & Sambanis, 2006; Miguel, Satyanath &
Sergenti, 2004). To control for likely spuriousness, I introduced a natural log of GDP per capita
(t-1) (GDP per capita) and GDP per capita growth (t-1) (GDP per capita growth). Data on GDP
per capita was taken from Maddison (2010).
Finally, following standard practice (e.g., Hegre & Sambanis, 2006), I controlled for
population size (ln) (with data from the National Material Capabilities dataset (v.4.0) (Singer,
1987))24 and time dependency using peace years with a decay function (( /)).25
------------------
Table I in here
------------------
Results
Table II presents the logit regression estimates. As shown in Model 1.1, income Gini
significantly affects popular rebellion onset when other variables are not controlled for.26 The
same is true for education Gini (Model 1.2). The subsequent model includes income Gini and
education Gini in one block. The two indices are moderately correlated (r = .40) suggesting a
possible overlap in their effect on the outcome variable. Yet, as shown in Model 1.3, the effect
sizes of the two indices remain almost unchanged, though the p-values slightly increase (from
.027 to .079 for income Gini and from .001 to .003 for education Gini). Models 1.4 and 1.5
indicate that regime type proxies have virtually no confounding effect on the relation between
inequality indices and popular rebellion onset. The same is true for GDP per capita and GDP per
23 Note that the link between inequality, the absolute level of income and economic growth is highly debated in the
economic literature, with some researchers arguing exactly the opposite that is, that inequality positively affects
economic growth (e.g., Forbes, 2000).
24 As the data on population size in Singer (1987) is limited to 2007, Missing observations for 2008 and 2009 were
linearly extrapolated using ‘ipolate’ command (with ‘epolate’ option) in Stata.
25 Peace years stand for the number of years since the last conflict (or 1946). X determines the rate of decay.
Following Hegre et al. (2001), X was set to 4, which halved the effect of the peace years with every additional three
years in peace. As onset of popular rebellion potentially depends on any kind of previous conflict (e.g., because of
‘the legacy of weapon stocks’ (Collier & Hoeffler, 2004:569)), I used peace years since conflicts of all categories
listed in the CDC.
26 ‘Significant’ refers to estimates significant at least at the 10% level.
16
capita growth (the p-values for income and education Gini in the full model (1.7) amount to .053
and .051, respectively).27
------------------
Table II in here
------------------
Interestingly, the effect of GDP per capita – one of the most robust predictors of (aggregate)
conflict (Hegre & Sambanis, 2006) – is highly insignificant (p = .850). This finding suggests that
inequality indices confound the GDP per capita effect on popular rebellion. Indeed, the p-value
for GDP per capita in the full model drops to .306 when income Gini is removed from the block,
and to .118 when both income Gini and education Gini are removed from the block.28
To test the non-linear hypothesis, I regressed quadratic terms, income Gini^2 and education
Gini^2 (Models 1.8b and 1.9b). The quadratic term of income Gini has an expected negative
sign, but the estimate is insignificant (p = .143). In contrast, the quadratic term of education Gini
is significant (p = .040),29 suggesting that the relation between educational inequality and
popular rebellion is indeed non-linear an issue I return to below.
To assess the robustness of these estimates, I performed a number of additional tests. SWIID
and DEIW contain a number of missing observations,30 which (combined with missing data on
control variables) could significantly have affected the estimates. In an attempt to address this, I
imputed missing data on all predictor variables with the multiple imputation software Amelia II
(Honaker, King & Blackwell, 2011) and performed the same regression once again (Table
III).31, 32
-------------------
Table III in here
--------------------
The results based on the imputed data are largely in line with the previous estimates: while the
27 In addition, I tested for interaction between GDP per capita growth and the inequality indices, but found no
significant effects (not reported here).
28 When I use imputed data the corresponding p-values drop to .327 and .036.
29 Note that the coefficient for the linear term of education Gini is insignificant in Model 1.9a. This coefficient,
however, becomes significant in an identic model based on imputed data (2.9a).
30 ~45% and ~17% respectively (see Table I).
31 For details on the imputation model and imputation diagnostics see the Online Appendix.
32 Multiple imputation cannot substitute complete datasets; yet, it can (and, under general conditions, does)
outperform listwise deletion (King et al., 2001).
17
effect sizes of the inequality proxies in some models (2.32.7) are noticeably smaller, the p-
values throughout all models follow almost an identical pattern.33
Further, I tested the robustness of the estimates using alternative coding of the dependent
variable (a. including coups, b. excluding cases with limited popular participation, c. excluding
cases of sectarian conflicts, see fn. 18) – and found no significant changes. The results of the
robustness tests are reported in the Online Appendix.
Discussion
What specific implications could these results have? First and foremost, the results demonstrate
that inequality in income and education significantly affect popular rebellion onset. As suggested
by H1, the relation between income inequality and popular rebellion is linear – the higher the
income inequality, the higher the likelihood of popular rebellion. In contrast, the relation
between educational inequality and popular rebellion is non-linear. Yet, in disagreement with
H2, this relation follows the shape of an inverted-J (not an inverted-U). As can be seen in Figure
1, the risk of popular rebellion increases linearly for most (75%) of the education Gini values.
-------------------
Figure 1 in here
--------------------
The findings of the present study thus stand in contrast to Besançon (2005), who – as
mentioned above – found that educational and income inequality have opposing effects on
‘revolutionary’ and ‘ethnic wars’. Using substantially expanded data, this study finds that effects
of income and educational inequality (on popular rebellions) largely go in the same direction.
Does the fact that education Gini, at extreme values, affects popular rebellion in an opposite
direction support the ‘relative strength’ and ‘social comparison’ arguments presented above? The
answer is most likely no, as rebels’ strength vis-à-vis the government, on the one hand, and
peoples’ tendency not to compare their socioeconomic position with ‘different others’, on the
33 The only noticeable differences are reported in Models 2.6 and 2.7, where p-values for education Gini drop
marginally below the level of significance (.110 and .112 respectively).
18
other, should be more a function of inequality in material goods (such as income) rather than
inequality in schooling years; yet, as indicated in the models based on both original and imputed
data income inequality affects popular rebellions linearly.
The finding that the risk of popular rebellion drops when educational inequality reaches
extreme levels is more likely a consequence of too low levels of education among the
disadvantaged themselves. Educated people may be more aware of social injustice and thus be
more aggravated by their implications (Berrebi, 2007: 8). Education may also strengthen
peoples’ sense of social responsibility and civil engagement (Ibid.), which in turn may facilitate
their mobilization. Hence, a certain minimum level of education may be needed for people to
become aware of (and resented over) their disadvantageous position, which, in turn, is needed for
collective action. Further research, however, is necessary to assess these considerations.
Second, the results indicate that the relation between inequality and popular rebellion is
largely independent of regime type, the absolute level of income and population size. This
suggests that inequality has a non-spurious effect on states’ proneness to popular rebellions.
While inequality could ultimately be rooted in one of these factors (e.g., Boix, 2009), the
immediate effects of inequality seem to outperform the immediate effects of regime type, the
absolute level of income and population size.
Finally, the findings indicate that distribution of income plays a more important role in
popular rebellion onset than the absolute level of income. This, in turn, suggests that would-be
rebels’ decision to join a rebellion may depend on their relative and not on their absolute level of
income. In their path-breaking study, Collier & Hoeffler have suggested that it is not grievances
over economic inequalities, political repression or ethnic discrimination that play the crucial role
in conflict onset rather it is opportunities:
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
(2004: 564).
One such opportunity is a large pool of potential recruits. The size of this, the argument goes,
depends on the absolute level of income. Potential rebels with a low income (in absolute terms)
19
are more willing to join a rebellion than those with a high income because of lower opportunity
costs: the former have less to lose than the latter. Thus, Collier & Hoeffler have hypothesized
that states with low absolute income have higher chances of rebel mobilization and, in turn,
higher risks of conflicts. Yet, the results of the present study indicate that low-income states are
not significantly more prone to popular rebellions than high-income states, when controlled for
distribution of income, regime type and population size. In contrast, the results show that states
with a highly skewed distribution of income are significantly more prone to popular rebellions
than those with a more equal distribution of income, when controlled for the absolute level of
income, regime type and population size. These findings thus suggest that what may ultimately
motivate people to rise up in arms is not lower opportunity costs (as proxied by GDP per capita),
but grievances over unequal distribution of income (as proxied by income Gini) (Gurr, 1970).
Conclusion
Recent research on the inequality-conflict nexus has mainly focused on conflicts fought between
ethnic groups. The relation between vertical inequality and non-ethnic conflicts has attracted less
attention. This study has attempted to address this gap by implementing an analysis of the
relation between vertical inequality and popular rebellions, armed conflicts where mobilization
transcends ethnic boundaries and hostilities involve popular participation. The study has shown
that vertical inequality proxies income and education Gini indices significantly predict
popular rebellion onset. Furthermore, the study has revealed that GDP per capita – one of the
most robust predictors of (broadly-defined) conflict– fails to account for popular rebellion onset
when inequality proxies are controlled for.
Based on the theoretical discussion and empirical results, this study offers three broad
suggestions for conflict research. First, conflict researchers should not disregard vertical
inequality as one of the potential sources of present-days conflicts. While this has largely been
the case, specified tests based on improved data suggest that vertical inequality could be as
important a predictor of non-ethnic conflicts as horizontal inequality of ethnic ones.
Second, conflict researchers should consider that vertical and horizontal inequalities play
different roles in ethnic and non-ethnic conflicts. This idea potentially explains why previous
research on the inequality-conflict nexus has failed to find a robust relation between vertical
20
inequality and an aggregate category of conflict (i.e., including both, ethnic and non-ethnic
conflicts). Therefore, further research on the inequality-conflict nexus should consider
disaggregating ethnic conflicts and non-ethnic conflicts into two separate categories.
Finally, conflict researchers should take into account the possibility that the well-established
relation between GDP per capita and risk of conflict could be confounded by the distribution of
income. While the results are not yet conclusive, it seems that the measure of the distribution of
income largely outperforms the measure of the absolute level of income in the models of popular
rebellion onset. Thus, further studies should consider controlling for the distribution of income
whenever the role of the absolute income in conflict is analysed.
Replication data
The empirical analyses were conducted in Stata 11.2. The replication data for the empirical
analysis in this article can be found at http://www.prio.no/jpr/datasets.
Acknowledgments
I thank Derek Beach, Halvard Buhaug, Scott Gates, Kristian Skrede Gleditsch, Nils Petter
Gleditsch, Tomas Janeliūnas, Mansoob Murshed, Gudrun Østby, Svend-Erik Skaaning, two
anonymous reviewers, and the editor for helpful feedback on earlier versions of this paper.
21
References
Acemoglu, Daron & James A Robinson (1998) Why did the West extend the franchise?
Democracy, inequality and growth in historical perspective. The Quarterly Journal of
Economics 115(4): 1167–1199.
Alesina, Alberto & Roberto Perotti (1996) Income distribution, political instability, and
investment. European Economic Review 40(6): 1203–1228.
Argyle, Michael (1994) The Psychology of Social Class. London: Routledge.
Atkinson, Barnes Anthony & Andrea Brandolini (2001) Promise and pitfalls in the use of
‘secondary’ data-sets: Income inequality in OECD countries as a case study. Journal of
Economic Literature 39(3): 771–799.
Auvinen, Juha & Wayne Nafziger (1999) The sources of humanitarian emergencies. Journal of
Conflict Resolution 43(3): 267–290.
Barro, Robert & Jong-Wha Lee (2010) A new data set on educational attainment in the world,
1950–2010. NBER Working Paper No. 15902.
Bartusevičius, Henrikas (2013) The Categorically Disaggregated Conflict Dataset, Version 1.0
(http://pure.au.dk/portal/en/henrikas@ps.au.dk).
Benaabdelaali, Wail; Said Hachane & Abdelhak Kamal (2012) Educational inequality in the
world, 1950-2010: Estimates from a new dataset. In: John A Bishop & Rafael Salas (eds)
Inequality, Mobility and Segregation: Essays in Honor of Jacques Silber (Research on
Economic Inequality, Volume 20). Bingley: Emerald.
Berrebi, Claude (2007) Evidence about the link between education, poverty and terrorism among
Palestinians. Peace Economics, Peace Science and Public Policy 13(1): 1–36.
Berkowitz, Leonard (1989) Frustration-aggression hypothesis: Examination and reformulation.
Psychological Bulletin 106(1): 59–73.
22
Besançon, Marie (2005) Relative resources: Inequality in ethnic wars, revolutions, and
genocides. Journal of Peace Research 42(4): 393–415.
Boix, Charles (2009) The conditional relationship between inequality and development. PS:
Political Science and Politics 42(4): 645–649.
Boix, Charles & Susan C Stokes (2003) Endogenous democratization. World Politics 55(4):
517–549.
Booth, John A (1991) Socioeconomic and political roots of national revolts in Central America.
Latin American Research Review 26(1): 33–73.
Bottero, Wendy (2004) Class identities and the identity of class. Sociology 38(5): 985–1003.
Castello, Amparo & Rafael Domenech (2002) Human capital inequality and economic growth:
Some new evidence. The Economic Journal 112(478): 187–200.
Cederman, Lars-Erik; Nils B Weidmann & Kristian Skrede Gleditsch (2011) Horizontal
inequalities and ethno-nationalist civil war: A global comparison. American Political
Science Review 105(3): 478–495.
Collier, Paul & Anke Hoeffler (2004) Greed and grievance in civil war. Oxford Economic
Papers 56(4): 563–595.
Crompton, Rosemary (2008) Class and Stratification, 3rd Edn. Cambridge: Polity.
Crosby, Faye (1976) A model of egoistical relative deprivation. Psychological Review 83(2):
85–113.
Deininger, Klaus & Lyn Squire (1996) A new data set measuring income inequality. World Bank
Economic Review 10(3): 565–591.
Deininger, Klaus & Lyn Squire (1998) New ways of looking at old issues: Inequality and
growth. Journal of Development Economics 57(2): 259–287.
Dollard, John; Neal E Miller, Leonard W Doob, O H Mowrer & Robert R Sears (1939)
23
Frustration and Aggression. New Haven, CT: Yale University Press.
Ellingsen, Tanja (2000) Colourful community or ethnic witches’ brew? Multiethnicity and
domestic conflict during and after the Cold War. Journal of Conflict Resolution 44(2): 228–
249.
Fearon, James & David Laitin (2003) Ethnicity, insurgency, and civil war. American Political
Science Review 97(1): 75–90.
Festinger, Leon (1954) A theory of social comparison processes. Human Relations 7(2): 117–
140.
Fineman, M Daniel (1997) A Special Relationship: The United States and Military Government
in Thailand, 1947–1958. University of Hawai’i Press.
Forbes, Kristin J (2000) A reassessment of the relationship between inequality and growth.
American Economic Review 90(4): 869–897.
Gleditsch, Kristian Skrede & Michael D Ward (1999) Interstate system membership: A revised
list of the independent states since 1816. International Interactions 25(4): 393–413.
Gleditsch, Nils Petter; Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg & Håvard
Strand (2002) Armed conflict 1946–2001: A new dataset. Journal of Peace Research 39(5):
615–637.
Goertz, Gary (2006) Social Science Concepts: A user’s guide. Princeton, NJ: Princeton
University Press.
Gurr, Ted Robert (1968) A causal model of civil strife: A comparative analysis using new
indices. The American Political Science Review 64(4): 1104–1124.
Gurr, Ted Robert (1970) Why Men Rebel. Princeton, NJ: Princeton University Press.
Gurr, Ted Robert (2000) Peoples versus States: Minorities at Risk in the New Century.
Washington, DC: United Institute of Peace.
24
Hardy, Melissa (1979) Economic growth, distributional inequality, and political conflict in
industrial societies. Journal of Political and Military Sociology 7(2): 209–227.
Hegre, Håvard; Tanja Ellingsen, Scott Gates & Nils Petter Gleditsch (2001) Towards a
democratic civil peace? Democracy, political change, and civil war, 1816–1992. American
Political Science Review 95(1): 33–48.
Hegre, Håvard; Ranveig Gissinger & Nils Petter Gleditsch (2003) Globalization and internal
conflict. In: Gerald Schneider; Katherine Barbieri & Nils Petter Gleditsch (eds)
Globalization and Armed Conflict. Oxford: Rowan & Littlefield, 251–276.
Hegre, Håvard & Nicholas Sambanis (2006) Sensitivity analysis of empirical results on civil war
onset. Journal of Conflict Resolution 50(4): 508–535.
Hendrix, Cullen S (2010) Measuring state capacity: Theoretical and empirical implications for
the study of civil conflict. Journal of Peace Research 47(3): 273–285.
Honaker, James; Gary King & Matthew Blackwell (2011) Amelia II: A program for missing
data. Journal of Statistical Software 45(7): 1–47.
Hunt, Scott A & Robert D Benford (2004) Collective identity, solidarity, and commitment. In:
David A Snow; Sarah A Soule & Hanspeter Kriesi (eds) The Blackwell Companion to Social
Movements. Malden: Blackwell, 433–457.
Jackman, Mary R & Robert W Jackman (1983) Class Awareness in the United States. Berkeley,
CA: University of California Press.
King, Gary; James Honaker, Anne Joseph & Kenneth Scheve (2001) Analyzing incomplete
political science data: An alternative algorithm for multiple imputation. American Political
Science Review 95(1): 49–69.
Lichbach, Mark (1989) An evaluation of ‘does economic inequality breed political conflict’.
World Politics 41(4): 431–472.
Maddison, Angus (2010) Historical Statistics: Statistics on World Population. GDP and Per
25
Capita GDP. 1–2008 AD (http://www.ggdc.net/maddison/).
McPherson, Miller; Lynn Smith-Lovin and James M Cook (2001) Birds of a feather: Homophily
in social networks. Annual Review of Sociology 27: 415–444.
Miguel, Edward; Shanker Santyanath & Ernest Sergenti (2004) Economic shocks and civil
conflict: An instrumental variables approach. Journal of Political Economy 112(4): 725–
753.
Miller, Neal E (1941) The frustration-aggression hypothesis. Psychological Review 48(4): 337–
342.
Mitchell, Edward (1968) Inequality and insurgency: A statistical study of South Vietnam. World
Politics 20(3): 421–453.
Muller, Edward N (1995) Economic determinants of democracy. American Sociological Review
60(6): 966–982.
Murshed, S Mansoob & Scott Gates (2005) Spatial-horizontal inequality and the Maoist
insurgency in Nepal. Review of Development Economics 9(1): 121–134.
Nagel, Jack (1974) Inequality and discontent: A nonlinear hypothesis. World Politics 26(4): 453–
472.
Nagel, Jack (1976) Erratum. World Politics 28(1): 315.
Olson, Mancur (1965) The Logic of Collective Action. Cambridge, MA: Harvard University
Press.
Østby, Gudrun (2008) Polarization, horizontal inequalities and violent civil conflict. Journal of
Peace Research 45(2): 143–162.
Østby, Gudrun; Ragnhild Nordås & Jan Ketil Rød (2009) Regional inequalities and civil conflict
in Sub-Saharan Africa. International Studies Quarterly 53(2): 301–324.
Østby, Gudrun (2011) Horizontal Inequalities and Political Violence. Doctoral Dissertation.
26
University of Oslo.
Pakulski, Jan & Malcolm Waters (1996) The Death of Class. London: Sage.
Powell, M Jonathan & Clayton L Thyne (2011) Global instances of coups from 1950 to 2010: A
new dataset. Journal of Peace Research 45(2): 249–259.
Prosterman, Roy (1976) IRI: A simplified predictive index of rural instability. Comparative
Politics 8(3): 339–353.
Ravallion, Martin (1997) Can high inequality developing countries escape absolute poverty?
Economic Letters 56(1): 51–57.
Rummel, Rudolph (1977) Understanding Conflict and War Vol. 3. Conflict in Perspective.
Beverly Hills, CA: Sage Publications.
Runciman, Walter (1966) Relative Deprivation and Social Justice. Berkeley: University of
California Press.
Russet, Bruce (1964) Inequality and instability: The relation of land tenure to politics. World
Politics 16(3): 442–454.
Sambanis, Nicholas (2005) Using case studies to refine and expand the theory of civil war. In:
Paul Collier & Nicholas Sambanis (eds) Understanding Civil War, Vol. 1. Washington DC:
The World Bank.
Sigelman, Lee & Miles Simpson (1977) A cross-national test of the linkage between economic
inequality and political violence. Journal of Conflict Resolution 21(1): 105–128.
Singer, David J (1987) Reconstructing the Correlates of War dataset on material capabilities of
states, 1916–1985. International Interactions 14(2): 115–132.
Smith, Anthony D (1991) National Identity. Reno, NV: University of Nevada Press.
Solt, Frederick (2009) Standardizing the world income inequality database. Social Science
Quarterly 90(2): 231–242. SWIID Version 3.1, December 2011.
27
Stewart, Frances (2002) Horizontal inequalities: A neglected dimension of development.
Working Paper Number 81, QEH Working Paper Series.
Stewart, Frances, ed. (2008) Horizontal Inequalities and Conflict: Understanding Group
Violence in Multiethnic Societies. London: Palgrave Macmillan.
Suls, Jerry; Rene Martin & Ladd Wheeler (2002) Social comparison: Why, with whom, and with
what effect? Current Directions in Psychological Science 11(5): 159–163.
Tadjoeddin, Mohammad Z; Anis Chowdhury & S Mansoob Murshed (2012) Routine violence in
Java, Indonesia: Neo-Malthusian and social justice perspectives. In: Jürgen Scheffran,
Michael Brzoska, Hans Günter Brauch, Peter Michael Link and Janpeter Schilling (eds)
Climate Change, Human Security and Violent Conflict: Challenges for Societal Stability.
Berlin: Springer, 633–655.
Themnér, Lotta (2013) UCDP/PRIO Armed Conflict Dataset Codebook, Version 4-2013
(http://www.pcr.uu.se/digitalAssets/63/63324_Codebook_UCDP_PRIO_Armed_Conflict_Dataset_v
4_2011.pdf).
Themnér, Lotta & Peter Wallensteen (2013) Armed Conflict, 1946–2012. Journal of Peace
Research 50(4): 509–521.
Thomas, Vinod; Yan Wang & Xibo Fan (2003). Measuring education inequality: Gini
coefficients of education for 140 countries, 1960–2000. Journal of Education Planning and
Administration, 17 (1): 5–33.
Tomz, Michael; Jason Wittenberg & Gary King (2003) CLARIFY: Software for interpreting and
presenting statistical results. Version 2.1. Stanford University, University of Wisconsin, and
Harvard University (http://gking.harvard.edu/).
Turner, C John; Penelope J Oakes, S Alexander Haslam & Craig McGarty (1994) Self and
collective: Cognition and social context. Personality and Social Psychology Bulletin 20(5):
454–463.
UNU-WIDER (2008) World Income Inequality Database, Version 2.0c, May 2008.
28
(http://www.wider.unu.edu/research/Database/en_GB/database/).
Van Zomeren M; Tom Postmes & Russell Spears (2008) Toward an integrative social identity
model of collective action: A quantitative research synthesis of three socio-psychological
perspectives. Psychological Bulletin 134(4): 504–535.
Vreeland, James Raymond (2008) The effect of political regime on civil war: Unpacking
anocracy. Journal of Conflict Resolution 52(3): 401–425.
Weede, Erich (1981) Income inequality, average income, and domestic violence. Journal of
Conflict Resolution 25(4): 639–654.
Weede, Erich (1987) Some new evidence on correlates of political violence; Income inequality,
regime repressiveness, and economic development. European Sociological Review 3(2): 97–
108.
Wood, Joanne V (1989) Theory and research concerning social comparisons of personal
attributes. Psychological Bulletin 106(2): 231–248.
Wright, Erik Olin (1997) Class Counts: Comparative Studies in Class Analysis. New York:
Cambridge University Press.
HENRIKAS BARTUSEVIČIUS, b. 1986, PhD student in Political Science, Aarhus University
(2010- ); previously visiting researcher at Center for the Study of Civil War, Peace Research
Institute Oslo (PRIO); research interests include: inequality and armed conflict.
29
Figures
Figure 1. Estimated probabilities (with 95% confidence intervals) of popular rebellion onset as a
function of income and education Gini, holding other variables at their mean values. The
probabilities were estimated using CLARIFY software (Tomz, Wittenberg & King, 2003).
30
Tables
Table I. Summary statistics
Observations
Mean
S.E.
Min
Max
N=7471
.010
.101
0
1
N=4130
.381
.106
.154
.713
N=6200
.459
.223
.09
.99
N=6099
7.900
4.834
1
14
N=6099
85.777
81.037
1
196
N=6675
1.081
1.098
-1.575
3.759
N=6655
1.831
6.146
-61.473
76.901
N=7471
15.781
.1664
11.599
21.025
N=7471
.261
.378
0
1
31
Table II. Logistic regression estimates of popular rebellion onset
(1.1) (1.2) (1.3) (1.4) (1.5) (1.6) (1.7) (1.8a) (1.8b) (1.9a) (1.9b)
Income Gini
3.276*
(1.482)
3.186†
(1.817)
3.403†
(1.846)
3.372†
(1.859)
3.411†
(1.939)
3.685†
(1.901)
3.076†
(1.781)
29.805†
(17.568)
Income Gini^2
-30.679
(20.945)
Education Gini
1.833***
(.494)
2.330**
(.776)
2.050*
(.933)
2.089*
(.914)
2.341*
(1.133)
2.224†
(1.138)
1.203
(1.022)
11.683*
(5.458)
Education Gini^2
-9.1873*
(4.470)
Xpolity scores
-.006
(.046)
.109
(.245)
.127
(.244)
.113
(.235)
-.106
(.253)
-.167
(.244)
.225
(.212)
.180
(.203)
Xpolity scores^2
-.007
(.014)
-.008
(.014)
-.007
(.014)
.005
(.015)
.009
(.014)
-.014
(.013)
-.011
(.012)
GDP per capita(ln)
.086
(.285)
.053
(.283)
-.253
(.247)
-.215
(.241)
-.054
(.244)
-.004
(.226)
GDP per capita growth
.063
(.044)
.052
(.034)
.050
(.034)
.039
(.037)
.038
(.036)
Population size(ln)
.216*
(.094)
.265***
(.060)
.265*
(.106)
.270*
(.110)
.271*
(.111)
.266*
(.117)
.259*
(.108)
.204*
(.104)
.202†
(.104)
.214**
(.079)
.193**
(.075)
Peace years
.638†
(.098)
.131
(.322)
.449
(.429)
.509
(.480)
.502
(.474)
.516
(.478)
.554
(.489)
.502
(.430)
.433
(.416)
.282
(.389)
.154
(.412)
Constant
-9.819***
(1.842)
-9.947***
(1.024)
-11.693***
(2.199)
-11.705***
(2.573)
-12.062***
(2.694)
-12.255***
(3.209)
-12.330***
(3.048)
-8.875***
(2.434)
-14.221***
(4.406)
-9.474***
(1.888)
-11.678***
( 2.263)
Wald χ2 18.87 38.24 36.86 29.63 28.89 37.17 38.84 50.29 46.59 29.50 30.29
N 4130 6200 3682 3199 3199 3149 3149 3462 3462 5004 5004
N of popular rebellions 36 57 31 28 28 28 28 32 32 46 46
Coefficients (β) with robust standard errors in parentheses. †p<.10; *p<.05; **p<.01; ***p<.001.
32
Table III. Logistic regression estimates of popular rebellion onset (imputed data)
(2.1) (2.2) (2.3) (2.4) (2.5) (2.6) (2.7) (2.8a) (2.8b) (2.9a) (2.9b)
Income Gini
2.590**
(.825)
1.922*
(.900)
1.995*
(.902)
1.900*
(.905)
1.773†
(.951)
1.811†
(.950)
1.959*
(.951)
12.890
(9.450)
Income Gini^2
-11.940
(11.056)
Education Gini
1.813***
(.455)
1.595***
(.489)
1.303*
(.586)
1.302*
(.579)
1.107
(.692)
1.095
(.689)
1.192†
(.681)
11.782**
(4.320)
Education Gini^2
-9.150*
(3.756)
Xpolity scores
-.032
(.0314)
.209
(.183)
.193
(.187)
.184
(.185)
.149
(.181)
.144
(.174)
.188
(.188)
.178
(.183)
Xpolity scores^2
-.015
(.011)
-.014
(.011)
-.013
(.011)
-.012
(.011)
-.011
(.010)
-.013
(.011)
-.012
(.011)
GDP per capita(ln)
-.094
(.167)
-.116
(.174)
-.226
(.152)
-.212
(.148)
-.165
(.169)
-.085
(.158)
GDP per capita growth
.028
(.024)
.028
(.023)
.028
(.023)
.027
(.023)
.026
(.022)
Population size(ln)
.236***
(.054)
.232***
(.053)
.260***
(.056)
.267***
(.055)
.271***
(.054)
.267***
(.055)
.263***
(.054)
.251***
(.052)
.248***
(.053)
.236***
(.051)
.227***
(.050)
Peace years
.240
(.259)
.079
(.274)
.061
(.275)
.090
(.276)
.051
(.278)
.020
(.287)
.040
(.290)
.091
(.287)
.047
(.285)
.030
(.292)
-.072
(.299)
Constant
-9.476***
(1.022)
-9.226***
(.917)
-10.349***
(1.082)
-10.125***
(1.158)
-10.801***
(1.262)
-10.473***
(1.446)
-10.452***
(1.436)
-9.534***
(1.206)
-11.829***
(2.392)
-9.312***
(1.243)
-11.903***
(1.628)
F 8.99 12.47 10.38 8.73 8.26 8.08 7.10 7.84 5.74 7.63 6.33
Average RVI .008 .034 .020 .072 .094 .085 .081 .082 .168 .084 .079
Imputed datasets 5 5 5 5 5 5 5 5 5 5 5
N 7471 7471 7471 7471 7471 7471 7471 7471 7471 7471 7471
N of popular rebellions 77 77 77 77 77 77 77 77 77 77 77
Coefficients (β) with robust standard errors in parentheses. †p<.10; *p<.05; **p<.01; ***p<.001.
33
THE INEQUALITY-CONFLICT NEXUS RE-EXAMINED: INCOME, EDUCATION AND
POPULAR REBELLIONS
ONLINE APPENDIX
Henrikas Bartusevičius
Department of Political Science and Government
Aarhus University
34
This Online Appendix provides:
1. Conceptualization and coding criteria used to code popular rebellions in the paper;
2. A full list of popular rebellions (Table I);
3. Details and diagnostics of the multiple imputation model employed in the paper;
4. Robustness tests of the results reported in the paper’s Tables II and III using alternative
coding of the dependent variable:
4.1. Including coups (Tables IIa and IIb);
4.2. Excluding sectarian conflicts (Tables IIIa and IIIb);
4.3. Excluding cases with limited popular participation (Tables IVa and IVb).
35
1. Conceptualization and coding criteria used to code popular rebellions in the paper
As described in the paper, to code popular rebellions I used the Categorically Disaggregated
Conflict Dataset (CDC) (Bartusevičius, 2013) as a base. To explain the coding of popular
rebellions, I first explain how conflicts are categorized in the CDC.
The CDC Dataset34
The CDC categorizes conflicts based on two distinctions:
(i) The aims over which conflicts are fought (governmental or territorial);
(ii) The ethnic composition of conflicting parties.
The two distinctions applied at once result in the following categorization:
Fought over
Government Territory
Fought between
Groups of different ethnicity
(1)
Ethnic governmental
(2)
Ethnic territorial
Groups of the same ethnicity
(3)
Non-ethnic governmental
(4)
Non-ethnic territorial
The CDC uses UCDP/PRIO dataset as a base; therefore, it employs the UCDP/PRIO’s
definition of an aggregate conflict:
‘a contested incompatibility that concerns government and/or territory where the use of
armed force between two parties, of which at least one is the government of a state, results in
at least 25 battle-related deaths’ (Themnér, 2011: 1).
The CDC also relies on the UCDP/PRIO’s coding of incompatibility (‘Incomp’) to distinguish
between governmental conflicts (‘incompatibility concerning type of political system, the
replacement of the central government, or the change of its composition’) and territorial
conflicts (‘incompatibility concerning status of a territory, [...] e.g., secession or autonomy’).
34 The following description of the CDC dataset is a condensed version of a description of the CDC dataset
provided in Bartusevičius (2013). The rationale behind coding decisions (and the need for such a dataset) cannot
be explained here in full. Interested readers should consult the original source.
36
For full definitions see Themnér (2011).
Subsequently, the CDC classifies these conflicts as ethnic or non-ethnic. Following Gurr and
Harff the CDC defines ‘ethnic groups’ as those
composed of people who share a distinctive and enduring collective identity based on shared
experiences and cultural traits…[who]…may define themselves, and be defined by others, in
terms of any or all of the following traits: life ways, religious beliefs, language, physical
appearance, region of residence, traditional occupations, and a history of conquest and
repression by culturally different peoples’ (1994: 190).
Based on this definition, the CDC classifies every conflict between two or more groups whose
majorities represent different ethnicities as ‘ethnic’ and every conflict between groups whose
majorities represent the same ethnicity as ‘non-ethnic’. Conflicts in which individuals of one
ethnicity compose a substantial part of both opposing groups are considered ‘non-ethnic’.
As the CDC was primarily established to study the conflict onset, conflicts in the CDC are
classified based on their characteristics recorded in the initial phase. Thus, conflicts that start
between non-ethnic groups but develop into inter-ethnic clashes are considered ‘non-ethnic
conflicts’ and conflicts that start between ethnic groups but develop into clashes between
members of the same ethnicity are considered ‘ethnic conflicts’.
For the reasons described in Bartusevičius (2013), the CDC disregards the motivation of
conflicting groups and focus solely on their composition. No matter the reason a conflict is
fought, it is considered an ‘ethnic conflict’ if the conflicting parties are composed of different
ethnic groups. The CDC assumes that conflicts fought between ethnic groups are, indeed,
fought over ethnic issues. Equally, the CDC assumes that conflicts involving systematic
fighting and killing between individuals of the same ethnicity are non-ethnic conflicts.
The following is the exact description of how the coding of ‘ethnic’ and ‘non-ethnic’ conflicts
was implemented in the CDC:
- First, identification of the parties to a conflict. For this purpose, the CDC used the
UCDP/PRIO’s ‘SideA’ and ‘SideB’ variables.
- Second, determination of the composition of the parties to a conflict. The coding of the
37
composition of SideA and SideB was based on author’s own reading of primary and
secondary sources. SideA is always the government of a state, and to determine its
composition the CDC focused on (i) the executive branch (i.e., presidents, prime
ministers, members of the cabinet), (ii) military leadership and (iii) foot soldiers taking
part in the conflict. In many cases, the CDC coded the composition of SideA based on
the composition of its de facto leaders, assuming that the formal composition of the
government (especially in autocracies) may not represent the actual power distribution
within the executive. Similarly, to determine the composition of SideB, the CDC
focused on the composition of (i) political and/or military leadership and (ii) foot
soldiers.
- Third, determination of the ethnic differences of the parties to a conflict. ‘Ethnic
differences’ in the CDC were operationalized by differences in language, religion and
‘race’.35 The CDC coded SideA and SideB as ‘ethnically different’ if their members
were distinct in at least one of the three characteristics.36 It is important to note that
the CDC treats language, religion and ‘race’ merely as proxies of the concept of
‘ethnicity’ and not as constitutive parts of ‘ethnicity’. The CDC concurs with the
widely-held view that ethnic affiliations are, to a great extent, socially constructed, and
acknowledges that some groups may define their ethnicity on the basis of other
attributes. Yet language, religion and ‘race’ are the qualities that can be more or less
unambiguously observed and used as proxies of the directly unobservable ‘ethnicity’.
These are also the qualities that (taken separately) help, according to the CDC –
surprisingly well – to empirically distinguish between groups widely perceived as
distinct ‘ethnic groups’.37
- Fourth, determination of the pattern of confrontation between parties to a conflict. In
this step, the CDC attempted to ascertain whether a conflict involved systematic
fighting between (and killing of) members of the same ethnic group.
35 Note, therefore, that conflict coding into ‘ethnic’ and ‘non-ethnic’ in the CDC did not merely rely on the labels
attached to groups (e.g. ‘Christian Maronites’ or ‘Kurds’), but on actual linguistic, religious and ‘racial’
characteristics.
36 To distinguish between separate languages and two dialects of the same language the CDC used Ethnologue
(Lewis, Simons & Fennig, 2013). To determine the religion of particular ethnic groups the CDC used World
Christian Database (Johnson, 2007). Note that followers of the main branches of Islam (Shia and Sunni) as well
as members of the main groupings of Christianity (Catholicism, Protestantism and Orthodoxy) are considered
members of different ethnic groups. The CDC provides coding for all three characteristics so that potential users
could easily apply other combinations of the three characteristics to match alternative definitions of ethnic
conflicts.
37 Indeed, there are very few groups widely considered to be ‘distinct ethnic groups’ that the coding criteria
introduced in the CDC fail to distinguish most notably, Tutsis and Hutus in Rwanda and Burundi and, less-well
known, the Lulua and Luba in DRC.
38
To make the coding explicit, the CDC provides coding descriptions (along with references to
the primary and secondary literature), documenting coding choices along the four steps for
every conflict.
The following are the primary advantages of the CDC as identified in Bartusevičius (2013):
1. The CDC is based on explicit definition and coding criteria, allowing systematic
empirical comparison of the conflict categories deemed important in recent conflict
research.
2. It provides explicit coding of the three principal marks of ethnicity (language, religion
and ‘race’), allowing potential users to quickly alter the coding of ethnic/non-ethnic
categories to fit alternative definitions.
3. It documents coding decisions, allowing users to track (and, if needed, update)
individual coding choices.
4. It categorizes conflicts based on ‘who actually fought whom’ – not just on ‘who were
the members of conflicting parties’ (as is the case previously introduced categorically
disaggregated datasets). This allows more precise coding of conflicts where conflicting
parties despite claiming to represent different ethnic groups engage in systematic
intra-ethnic fighting. This, in turn, results in a more ‘conservative’ list of ethnic
conflicts, allowing statistical analysis of the non-ethnic territorial category.
The UCDP/PRIO dataset (v.4-2011) contains 368 separate onsets of aggregate intrastate
armed conflicts and internationalized intrastate armed conflicts. In line with previous research,
the CDC applied the two-year intermittency rule. Therefore, the final number of onsets in the
CDC amounts to 331: 59 ethnic governmental, 128 ethnic territorial, 124 non-ethnic
governmental and 20 non-ethnic territorial.
The CDC comes in two formats:
.xls file (full list of variables together with the codebook);
.pdf file (a more reader-friendly version of the dataset that contains the codebook (p.15), key
variables and Coding Descriptions).
39
Popular rebellions
As indicated in the paper (p. 11), the CDC’s third category – ‘non-ethnic governmental’
was employed as a base for coding popular rebellions. The CDC contains 124 onsets of non-
ethnic governmental conflicts recorded in 1946–2009. As the time span in the paper was
restricted to 1961–2009, 24 cases of popular rebellions were dropped. Subsequently, 23
cases of non-ethnic governmental conflicts were dropped as cases of violent coups/coup
attempts. Following the conceptual framework introduced in the paper, popular
participation was used as an operational difference between popular rebellions and violent
coups/coup attempts. Thus, conflicts limited to the elitist struggle within the state apparatus
were treated as coups, and conflicts involving popular participation were coded as popular
rebellions.
I have not applied any numeric threshold for popular participation; therefore, popular
rebellions include several ‘border’ cases (e.g., Government of Bolivia vs. National
Liberation Army, ID: 1), where popular participation was rather limited (from several
dozens to hundreds of individuals). Moreover, the analysis includes several cases of
sectarian conflicts (e.g., Government of Nigeria vs. Boko Haram, ID: 100).38
These cases could not have been left out without additional ad hoc coding rules. However, as
can be seen from the table below, sectarian conflicts (marked with ), as well as rebellions
with limited popular participation (marked with *), constitute relatively small fraction of
cases in the sample – the vast majority of popular rebellions were political in nature and
involved extensive popular participation. As demonstrated in robustness tests below,
including coups, removing cases with limited participation or sectarian conflicts does not
significantly affect the results reported in the paper.
38 Potential users are welcome to enquire about individual coding decisions and references used to inform these
coding decisions.
40
2. Full list of popular rebellions (Table 1)
Table 1. Popular Rebellions 1961–2009
ID
Side A
Side B
Onset year
1
Government of Bolivia
National Liberation Army
1967*
10
Government of
Philippines
Communist Party of the Philippines
1969
22
Government of
Paraguay
Military faction (forces of Andres Rodriguez)
1989
24
Government of
Myanmar
All-Burma Students Democratic Front
1990
29
Government of India
Communist Party of India Marxist-Leninist
1969
29
Government of India
People’s War Group
1990
33
Government of North
Yemen
Royalists
1962
33
Government of North
Yemen
National Democratic Front
1979
33
Government of Yemen
al-Qaida in the Arabia Peninsula
2009
36
Government of
Guatemala
Rebel Armed Forces
1963
43
Government of Thailand
Communist Party of Thailand
1974
45
Government of Cuba
Cuban Revolutionary Council
1961
50
Government of
Argentina
Military faction (colorados)
1963
50
Government of
Argentina
People’s Revolutionary Army
1974
62
Government of Iraq
Military faction (forces of Abd as-Salam Arif); National Council of the Revolutionary
Command
1963
62
Government of Iraq
Al-Mahdi Army; Supporters of Muslims; The Monotheism and Jihad Group
2004
63
Government of Lebanon
Lebanese National Movement
1975
63
Government of Lebanon
Lebanese National Movement
1982
64
Government of Malaysia
Communist Party of Malaya
1974
64
Government of Malaysia
Communist Party of Malaya
1981
70
Government of Ethiopia
Ethiopian People’s Revolutionary Party; Tigrean People’s Liberation Front
1976
72
Government of Nepal
Communist Party of NepalMaoist
1996
73
Government of France
Secret Army Organization
1961*
80
Government of
Venezuela
Military faction (navy)
1962
80
Government of
Venezuela
Bandera Roja (Red Flag)
1982*
80
Government of
Venezuela
Military faction (forces of Hugo Chávez)
1992
87
Government of Gabon
Military faction (forces loyal to Léon M'Ba)
1964
90
Government of Burundi
Military faction (forces loyal to Gervais Nyangoma)
1965
90
Government of Burundi
Party for the Liberation of the Hutu People
1991
90
Government of Burundi
National Council for the Defense of Democracy
1994
90
Government of Burundi
Party for the Liberation of the Hutu People Forces for National Liberation
2008
92
Government of
Colombia
Revolutionary Armed Forces of Colombia
1964
93
Government of
Dominican Republic
Military faction (constitutionalists)
1965
95
Government of Peru
National Liberation Army; Movement of the Revolutionary Left
1965
95
Government of Peru
Shining Path
1982
95
Government of Peru
Shining Path
2007*
98
Government of Ghana
National Liberation Council
1966
98
Government of Ghana
Military faction (forces of Jerry John Rawlings)
1981
98
Government of Ghana
Military faction (forces of Ekow Dennis and Edward Adjei-Ampofo)
1983
100
Government of Nigeria
Boko Haram
2009
103
Government of
Cambodia
Red Khmers
1967
103
Government of
Cambodia
Kampuchean National United Front for National Salvation
1978
111
Government of Guinea
Rally of Democratic Forces of Guinea
2000*
113
Government of Sudan
Sudanese Communist Party
1971
113
Government of Sudan
Islamic Charter Front
1976
114
Government of
National Movement for the Independence of Madagascar
1971
41
Madagascar
115
Government of Morocco
Military faction (forces of Mohamed Madbouh)
1971
117
Government of Sri Lanka
People’s Liberation Front
1971
117
Government of Sri Lanka
People’s Liberation Front
1989
118
Government of Uganda
Military faction (forces of Charles Arube)
1974
120
Government of El
Salvador
Military faction (forces of Benjamin Mejia)
1972
120
Government of El
Salvador
People’s Revolutionary Army; Farabundo Marti Popular Liberation Forces
1979
123
Government of Uruguay
Movement of National Liberation/Tupamaros
1972
125
Government of Chile
Forces of Augusto Pinochet, Toribio Merino and Leigh Guzman
1973
130
Government of Eritrea
Eritrean Muslimsic Jihad Movement Abu Suhail faction
1997
130
Government of Eritrea
Eritrean Muslimsic Jihad Movement Abu Suhail faction
2003
137
Government of
Afghanistan
People's Democratic Republic of Afghanistan
1978
137
Government of
Afghanistan
Taleban
2003
140
Government of
Nicaragua
Sandinista National Liberation Front
1977
140
Government of
Nicaragua
Contras/Nicaraguan Democratic Forces
1982
141
Government of Somalia
Somali Salvation Democratic Front
1982
141
Government of Somalia
Somali National Movement
1986
141
Government of Somalia
Somali Reconciliation and Restoration Council
2001
141
Government of Somalia
Supreme Islamic Council of Somalia
2006
143
Government of Iran
People's Mujahideen
1979
143
Government of Iran
People's Mujahideen
1986
143
Government of Iran
People's Mujahideen
1991
143
Government of Iran
People's Mujahideen
1997
145
Government of Saudi
Arabia
The Salafi groups which practice hisba
1979
148
Government of Tunisia
Tunisian Armed Resistance
1980*
149
Government of Gambia
National Revolutionary Council
1981*
158
Government of
Cameroon
Military faction (forces of Ibrahim Saleh)
1984
163
Government of Togo
Togolese Movement for Democracy
1986
164
Government of South
Yemen
Yemenite Socialist Party - Abdul Fattah Ismail faction
1986
165
Government of Burkina
Faso
Popular front
1987
167
Government of Comoros
Presidential guard
1989
172
Government of Panama
Military faction (forces of Moises Giroldi)
1989
175
Government of Romania
National Salvation Front
1989
179
Government of Rwanda
Rwandan Patriotic Front
1990
179
Government of Rwanda
Armed People for the Liberation of Rwanda
1996
179
Government of Rwanda
Armed People for the Liberation of Rwanda
2009
185
Government of Georgia
National Guard and Mkhedrioni
1991
186
Government of Haiti
Military faction (forces of Himmler Rebu and Guy Francois)
1989
186
Government of Haiti
Military faction (forces of Raol Cédras)
1991
186
Government of Haiti
National Front for the Liberation of Haiti, OP Lavalas (Chimères)
2004
187
Government of Sierra
Leone
Revolutionary United Front
1991
188
Government of Turkey
Revolutionary Left
1991*
188
Government of Turkey
Maoist Communist Party
2005
191
Government of Algeria
Exile and Redemption
1991
196
Government of Egypt
Islamic Group
1993
200
Government of
Tajikistan
United Tajik Opposition
1992
201
Government of
Azerbaijan
Military faction (forces of Suret Husseinov)
1993
201
Government of
Azerbaijan
Special Police Brigade
1995
204
Government of Russia
(Soviet Union)
Parliamentary Forces
1993
205
Government of Mexico
Popular Revolutionary Army
1996
209
Government of Pakistan
Movement for the Enforcement of Islamic Laws
2007
216
Government of Guinea-
Bissau
Military Junta for the Consolidation of Democracy, Peace and Justice
1998
217
Government of Lesotho
Military faction
1998
221
Government of
Uzbekistan
Islamic Movement of Uzbekistan
1999
42
221
Government of
Uzbekistan
Jihad Islamic Group
2004
Grey colour indicates cases excluded from the main analysis in the paper (i.e., coups/coup attempts). *Cases with limited popular
participation; † Sectarian conflicts.
43
3. Details and diagnostics of the multiple imputation model employed in the paper
Multiple imputation was implemented using Amelia II (Honaker, King & Blackwell, 2011).
The imputation model was set up largely using default options recommended in the Amelia II
manual. More specifically:
The imputation model contained all the predictor variables reported in Model 1.7;
location was set as the cross-sectional variable;
year was set as the time-series variable (one knot);
Cross-section was interacted with time-series;
The model included lag and lead versions of income Gini and education Gini;
The imputation model was set to generate five imputed datasets;
The imputations were exported in ‘stacked’ Stata file.39
The imputed data was declared as imputed in Stata using:
mi import flong, m(imp) id(location year) imputed(iginit
eginit xpolt14 xpolt142 gdp_pcgt lngdp_pct iginit2
eginit2) passive(epyears lnpops)
The analysis was implemented using mi estimate (see the do-replication file for further
details).
Below I provide some basic imputation diagnostics graphs together with a number of
examples of imputed income Gini values for particular countries.
39 The following is an excerpt from the output log reporting the settings of the imputation model in R:
amelia(x = getAmelia("amelia.data"), m = 5, idvars = NULL, ts = "year",
cs = "location", priors = NULL, lags = c("iginit", "eginit"
), empri = 0, intercs = TRUE, leads = c("iginit", "eginit"
), splinetime = 1, logs = NULL, sqrts = NULL, lgstc = NULL,
ords = "xpolt14", noms = NULL, bounds = c(3, 4, 5, 0, 0,
0, 1, 1, 14), max.resample = 1000, tolerance = 1e-04)
44
45
46
47
48
4. Robustness tests of the results reported in the paper’s Tables II and III using alternative coding of the dependent variable
Table IIa. Logistic regression estimates of popular rebellion onset (including coups)
(1.1) (1.2) (1.3) (1.4) (1.5) (1.6) (1.7) (1.8a) (1.8b) (1.9a) (1.9b)
Income Gini
4.107***
(1.235)
3.950**
(1.509)
4.076**
(1.519)
4.066**
(1.531)
4.141*
(1.622)
4.259**
(1.573)
3.588*
(1.479)
34.034*
(14.156)
Income Gini^2
-34.532*
(16.595)
Education Gini
2.167***
(.428)
2.842***
(.679)
2.477**
(.816)
2.497**
(.797)
2.772**
(.916)
2.688**
(.942)
1.598
(.829)
11.569**
(4.464)
Education Gini^2
-8.603*
(3.659)
Xpolity scores
-.018
(.040)
.057
(.208)
.078
(.210)
.074
(.206)
-.065
(.232)
-.133
(.221)
.308
(.189)
.261
(.180)
Xpolity scores^2
-.005
(.012)
-.006
(.013)
-.005
(.013)
.001
(.014)
.005
(.013)
-.019
(.011)
-.015
(.011)
GDP per capita (ln)
.104
(.233)
.076
(.249)
-.178
(.223)
-.134
(.217)
-.016
(.191)
.019
(.173)
GDP per capita growth
.033
(.062)
-.005
(.057)
-.006
(.055)
-.004
(.035)
-.004
(.035)
Population size (ln) .138
(.083)
.206***
(.052)
.202*
(.095)
.197*
(.101)
.197
(.101)
.190
(.105)
.184
(.104)
.111
(.107)
.111
(.105)
.154*
(.073)
.139
*(.069)
Peace years
.562
(.357)
.015
(.281)
.131
(.394)
.194
(.425)
.190
(.421)
.205
(.421)
.226
(.433)
.458
(.385)
.382
(.376)
.020
(.340)
-.098
(.354)
Constant
-8.594***
(1.560)
-8.812***
(.883)
-10.874***
(1.927)
-10.515***
(2.275)
-10.729***
(2.258)
-10.942***
(2.608)
-10.889***
(2.564)
-7.135***
(2.132)
-13.355***
(3.722)
-8.464***
( 1.611)
-10.657***
(1.920)
Wald χ2 28.57 46.19 46.12 38.46 40.57 43.81 43.73 55.65 52.02 35.69 35.29
N 4130 6200 3682 3199 3199 3149 3149 3462 3462 5004 5004
N of popular rebellions 46 77 39 36 36 36 36 42 42 64 64
Coefficients (β) with robust standard errors in parentheses. †p<.10; *p<.05; **p<.01; ***p<.001.
49
Table IIb. Logistic regression estimates of popular rebellion onset (including coups) (imputed data)
(2.1)
(2.2)
(2.3)
(2.4)
(2.5)
(2.6)
(2.7)
(2.8a)
(2.8b)
(2.9a)
(2.9b)
Income Gini
2.674**
(.827)
1.900*
(.865)
1.973*
(.856)
1.879*
(.856)
1.817*
(.894)
1.808*
(.895)
2.026*
(.892)
14.666
(9.495)
Income Gini^2
-13.766
(10.668)
Education Gini
2.045***
(.403)
1.828***
(.432)
1.561**
(.509)
1.547**
(.502)
1.454*
(.590)
1.458*
(.592)
1.560**
(.586)
11.111***
(3.493)
Education Gini^2
-8.159**
(3.003)
Xpolity scores
-.030
(.027)
.256
(.158)
.248
(.162)
.252
(.163)
.206
(.159)
.200
(.152)
.256
(.165)
.245
(.161)
Xpolity scores^2
-.018
(.009)
-.017†
(.010)
-.017†
(.010)
-.016
(.009)
-.015
(.009)
-.017†
(.010)
-.016
(.010)
GDP per capita (ln)
-.046
(.141)
-.039
(.144)
-.183
(.129)
-.168
(.127)
-.090
(.139)
-.026
(.129)
GDP per capita growth
-.011
(.025)
-.010
(.025)
-.010
(.025)
-.011
(.024)
-.010
(.023)
Population size (ln)
.180***
(.047)
.180***
(.050)
.207***
(.053)
.213***
(.052)
.220***
(.050)
.218***
(.051)
.221***
(.052)
.204***
(.047)
.200***
(.048)
.195***
(.048)
.186***
(.047)
Peace years
.254
(.234)
.068
(.246)
.052
(.248)
.077
(.249)
.028
(.249)
.013
(.255)
-.002
(.258)
.066
(.254)
.019
(.252)
-.015
(.259)
-.103
(.264)
Constant
-8.339***
(-8.339)
-8.237***
(.844)
-9.333***
(1.051)
-9.124***
(1.117)
-9.953***
(1.197)
-9.792***
(1.349)
-9.837***
(1.361)
-8.610***
(1.076)
-11.264***
(2.424)
-8.717***
(1.152)
-11.084***
(1.436)
F 8.82 14.26 11.45 9.88 9.90 9.33 8.42 9.19 6.91 9.08 7.51
Average RVI .115 .063 .088 .101 .118 .105 .096 .100 .155 .081 .070
Imputed datasets 5 5 5 5 5 5 5 5 5 5 5
N 7471 7471 7471 7471 7471 7471 7471 7471 7471 7471 7471
N of popular rebellions 77 77 77 77 77 77 77 77 77 77 77
Coefficients (β) with robust standard errors in parentheses. †p<.10; *p<.05; **p<.01; ***p<.001.
50
Table IIIa. Logistic regression estimates of popular rebellion onset (excluding sectarian conflicts)
(1.1)
(1.2)
(1.3)
(1.4)
(1.5)
(1.6)
(1.7)
(1.8a)
(1.8b)
(1.9a)
(1.9b)
Income Gini
3.744*
(1.629)
3.679†
(1.882)
3.956*
(1.899)
3.938*
(1.909)
3.938*
(2.005)
4.312*
(1.958)
3.731†
(1.933)
27.248
(18.397)
Income Gini^2
-26.565
(21.239)
Education Gini
1.642**
(.531)
2.126*
(.832)
2.002*
(1.021)
2.033*
(1.003)
2.187†
(1.246)
2.050†
(1.246)
.915
(1.060)
9.269†
(5.239)
Education Gini^2
-7.337†
(4.310)
Xpolity scores
.011
(.048)
.105
(.272)
.113
(.271)
.096
(.260)
.064
(.296)
-.011
(.289)
.286
(.226)
.247
(.218)
Xpolity scores^2
-.006
(.015)
-.006
(.016)
-.005
(.015)
-.002
(.017)
.002
(.017)
-.017
(.013)
-.014
(.013)
GDP per capita (ln)
.050
(.307)
.018
(.300)
-.322
(.266)
-.287
(.264)
-.166
(.248)
-.117
(.234)
GDP per capita growth
.074*
(.037)
.059†
(.031)
.058†
(.031)
.048
(.033)
.047
(.033)
Population size (ln)
.188†
(.107)
.227***
(.063)
.250*
(.115)
.250*
(.120)
.251*
(.121)
.244†
(.126)
.242*
(.114)
.183
(.118)
.184
(.118)
.180*
(.079)
.161*
(.076)
Peace years
.809*
(.408)
.269
(.325)
.622
(.432)
.678
(.494)
.672
(.488)
.680
(.491)
.722
(.504)
.633
(.461)
.583
(.446)
.415
(.392)
.313
(.411)
Constant
-9.747***
(2.113)
-9.351***
(1.060)
-11.686***
(2.380)
-11.875***
(2.791)
-12.186***
(2.955)
-12.213***
(3.495)
-12.450***
(3.323)
-9.737***
(2.877)
-14.493**
(4.963)
-9.093***
(1.959)
-10.786***
( 2.238)
Wald χ2 23.48 28.98 36.40 32.21 30.92 38.06 40.78 45.44 43.22 31.06 32.71
N 4130 6200 3682 3199 3199 3149 3149 3462 3462 5004 5004
N of popular rebellions 32 52 30 26 26 26 26 28 28 43 43
Coefficients (β) with robust standard errors in parentheses. †p<.10; *p<.05; **p<.01; ***p<.001.
51
Table IIIb. Logistic regression estimates of popular rebellion onset (excluding sectarian conflicts) (imputed data)
(2.1)
(2.2)
(2.3)
(2.4)
(2.5)
(2.6)
(2.7)
(2.8a)
(2.8b)
(2.9a)
(2.9b)
Income Gini
2.911***
(.906)
2.287*
(.974)
2.314*
(.964)
2.209*
(.966)
1.937†
(1.01933)
1.992†
(1.025)
2.120*
(1.029)
11.370
(8.914)
Income Gini^2
-9.842
(9.925)
Education Gini
1.778***
(.506)
1.509**
(.549)
1.420*
(.636)
1.416*
(.627)
.980
(.753)
.964
(.749)
1.074
(.743)
8.976*
(4.387)
Education Gini^2
-6.803†
(3.835)
Xpolity scores
-.010
(.031)
.264
(.187)
.228
(.194)
.216
(.192)
.186
(.188)
.180
(.181)
.222
(.194)
.214
(.190)
Xpolity scores^2
-.017
(.011)
-.014
(.011)
-.013
(.011)
-.012
(.011)
-.011
(.011)
-.013
(.011)
-.012
(.011)
GDP per capita (ln)
-.212
(.182)
-.242
(.189)
-.341*
(.161)
-.327*
(.159)
-.297
(.182)
-.225
(.173)
GDP per capita growth
.036
(.023)
.036
(.023)
.035
(.023)
.035
(.023)
.034
(.022)
Population size (ln)
.204***
(.057)
.193***
(.056)
.227***
(.059)
.229***
(.058)
.236***
(.057)
.229***
(.058)
.227***
(.057)
.217***
(.054)
.214***
(.055)
.196***
(.052)
.186***
(.051)
Peace years
.304
(.282)
.160
(.295)
.139
(.296)
.148
(.300)
.101
(.302)
.034
(.313)
.056
(.316)
.098
(.314)
.066
(.311)
.042
(.320)
-.035
(.326)
Constant
-9.257***
(1.095)
-8.751***
(.955)
-10.087***
(1.167)
-10.021***
(1.245)
-10.841***
(1.345)
-10.141***
(1.564)
-10.157***
(1.553)
-9.381***
(1.325)
-11.356***
(2.449)
-8.884***
(1.282)
-10.773***
(1.651)
F 7.57 9.27 8.32 6.96 6.96 7.56 6.52 7.22 5.34 7.03 5.72
Average RVI .023 .032 .019 .040 .053 .044 .046 .049 .098 .049 .059
Imputed datasets 5 5 5 5 5 5 5 5 5 5 5
N 7471 7471 7471 7471 7471 7471 7471 7471 7471 7471 7471
N of popular rebellions 77 77 77 77 77 77 77 77 77 77 77
Coefficients (β) with robust standard errors in parentheses. †p<.10; *p<.05; **p<.01; ***p<.001.
52
Table IVa. Logistic regression estimates of popular rebellion onset (excluding cases with limited popular participation)
(1.1)
(1.2)
(1.3)
(1.4)
(1.5)
(1.6)
(1.7)
(1.8a)
(1.8b)
(1.9a)
(1.9b)
Income Gini
2.924†
(1.715)
2.793
(2.065)
3.512†
(2.053)
3.448†
(2.099)
3.376
(2.173)
3.696†
(2.140)
3.178
(1.987)
19.528
(15.354)
Income Gini^2
-18.688
(18.148)
Education Gini
1.842***
(.524)
2.502**
(.828)
1.913†
(1.041)
2.005*
(1.012)
1.834
(1.207)
1.733
(1.210)
.603
(1.087)
10.889†
( 5.938)
Education Gini^2
-9.060†
(4.852)
Xpolity scores
-.014
(.048)
.240
(.261)
.224
(.264)
.202
(.253)
-.060
(.270)
-.103
(.265)
.151
(.217)
.115
(.207)
Xpolity scores^2
-.015
(.015)
-.014
(.016)
-.012
(.015)
.002
(.016)
.005
(.015)
-.011
(.013)
-.008
(.013)
GDP per capita (ln)
-.060
(.298)
-.090
(.294)
-.265
(.263)
-.241
(.260)
-.177
(.274)
-.121
(.255)
GDP per capita growth
.064
(.043)
.054
(.035)
.053
(.035)
.047
(.035)
.047
(.034)
Population size (ln)
.215*
(.107)
.276***
(.064)
.260*
(.119)
.288*
(.122)
.291*
(.124)
.282*
(.129)
.275*
(.118)
.226*
(.115)
.225†
(.115)
.231**
(.080)
.209**
(.076)
Peace years
.788†
(.413)
.252
(.335)
.538
(.460)
.563
(.507)
.548
(.495)
.541
(.501)
.583
(.5121)
.577
(.454)
.531
(.439)
.423
(.406)
.303
(.434)
Constant
-9.883***
(2.084)
-10.299***
(1.126)
-11.708***
(2.495)
-12.054***
(2.835)
-12.873***
(2.993)
-12.509***
(3.507)
-12.583***
(3.336)
-9.589***
( 2.685)
-12.837***
(4.250)
-9.263***
(1.994)
-11.417***
(2.428)
Wald χ2 18.07 6200 3682 3199 3199 3149 3149 3462 3462 5004 5004
N 4130 33.49 33.32 25.64 24.45 51.33 54.18 58.80 53.82 33.36 36.80
N of popular rebellions 31 50 27 25 25 25 25 28 28 40 40
Coefficients (β) with robust standard errors in parentheses. †p<.10; *p<.05; **p<.01; ***p<.001.
53
Table IVb. Logistic regression estimates of popular rebellion onset (excluding cases with limited popular participation) (imputed data)
(2.1)
(2.2)
(2.3)
(2.4)
(2.5)
(2.6)
(2.7)
(2.8a)
(2.8b)
(2.9a)
(2.9b)
Income Gini
2.143*
(.9112)
1.454
(.993)
1.553
(1.000)
1.473
(1.000)
1.272
(1.031)
1.312
(1.035)
1.434
(1.026)
13.921
(9.346)
Income Gini^2
-13.935
(10.894)
Education Gini
1.812***
(.480)
1.657***
(.518)
1.238*
(.622)
1.236*
(.617)
.927
(.730)
.913
(.726)
.979
(.716)
11.937*
(4.827)
Education Gini^2
-9.449*
(4.171)
Xpolity scores
-.047
(.034)
.170
(.187)
.144
(.193)
.134
(.190)
.106
(.185)
.103
(.179)
.136
(.192)
.129
(.187)
Xpolity scores^2
-.014
(.011)
-.012
(.011)
-.011
(.011)
-.010
(.011)
-.009
(.011)
-.011
(.011)
-.009
(.011)
GDP per capita (ln)
-.153
(.178)
-.178
(.185)
-.270†
(.162)
-.255
(.157)
-.212
(.182)
-.128
(.170)
GDP per capita growth
.031
(.024)
.031
(.024)
.030
(.023)
.030
(.024)
.029
(.023)
Population size (ln)
.231***
(.057)
.234***
(.057)
.255***
(.060)
.267***
(.060)
.269***
(.059)
.263***
(.059)
.259***
(.058)
.249***
( .056)
.245***
(.057)
.240***
(.054)
.230***
(.053)
Peace years
.370
(.266)
.198
(.282)
.183
(.283)
.223
(.283)
.187
(.283)
.135
(.295)
.158
(.298)
.198
(.296)
.146
(.293)
.151
(.300)
.048
(.310)
Constant
-9.367***
(1.106)
-9.422***
(.996)
-10.259***
(1.198)
-9.949***
(1.282)
-10.547***
(1.366)
-10.022***
(1.533)
-10.001***
(1.521)
-9.249***
(1.274)
-11.805***
(2.418)
-9.184***
(1.306)
-11.883***
(1.781)
F 7.40 11.09 8.96 7.48 6.91 7.28 6.39 7.23 5.81 7.07 5.98
Average RVI .021 .031 .026 .080 .103 .090 .085 .088 .120 .091 .086
Imputed datasets 5 5 5 5 5 5 5 5 5 5 5
N 7471 7471 7471 7471 7471 7471 7471 7471 7471 7471 7471
N of popular rebellions 77 77 77 77 77 77 77 77 77 77 77
Coefficients (β) with robust standard errors in parentheses. †p<.10; *p<.05; **p<.01; ***p<.001.
54
References
Bartusevičius, Henrikas (2013) The Categorically Disaggregated Conflict Dataset, Version 1.0
(http://pure.au.dk/portal/en/henrikas@ps.au.dk).
Gleditsch, Nils Peter; Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg & Håvard Strand
(2002) Armed conflict 1946–2001: A new dataset. Journal of Peace Research 39(5): 615637.
Gurr, Ted Robert & Barbara Harff (1994) Ethnic Conflicts in World Politics. Boulder, CO: Westview
Press.
Honaker, James; Gary King & Matthew Blackwell (2011) Amelia II: A program for missing data. Journal
of Statistical Software 45(7): 147.
Johnson, Todd M. (ed.) (2007) World Christian Database. Leiden/Boston: Brill.
(http://www.worldchristiandatabase.org).
Lewis, M. Paul; Gary F. Simons, and Charles D. Fennig (eds) (2013) Ethnologue: Languages of the
World, Seventeenth edition. Dallas, Texas: SIL International. (http://www.ethnologue.com)
Themnér, Lotta (2011) UCDP/PRIO Armed Conflict Dataset Codebook, Version 4-2011
(http://www.pcr.uu.se/digitalAssets/63/63324_Codebook_UCDP_PRIO_Armed_Conflict_Datase
t_v4_2011.pdf).
... Empirical research lends considerable support to Guterres' claim. High levels of economic inequality within societies have been shown to increase the probability of riots, antigovernment demonstrations, coups, revolutions, and the onset of civil war (Baten and Mumme 2013;Bartusevičius 2019;Houle 2019). Inequality may also slow economic growth, which can produce or exacerbate domestic unrest (Alesina and Perotti 1996;Roe and Siegel 2011). ...
... Studies have shown that horizontal inequality, particularly between ethnic groups, is often associated with the outbreak of ethnic conflict (Østby 2008;Cederman, Weidmann, and Gleditsch 2011). Vertical inequality among individuals, what is more commonly thought of as economic inequality within society, has been shown to be positively related to nonethnic civil conflict, nonethnic revolutions, and popular rebellions (Besançon 2005;Bartusevičius 2014Bartusevičius , 2019. Using a measure of height inequality within cohorts as a proxy for economic disparities over the past two centuries , Baten and Mumme (2013) present evidence suggesting that inequality is associated with civil war onset. ...
... Studies have shown that horizontal inequality, particularly between ethnic groups, is often associated with the outbreak of ethnic conflict (Østby 2008;Cederman, Weidmann, and Gleditsch 2011). Vertical inequality among individuals, what is more commonly thought of as economic inequality within society, has been shown to be positively related to nonethnic civil conflict, nonethnic revolutions, and popular rebellions (Besançon 2005;Bartusevičius 2014Bartusevičius , 2019. Using a measure of height inequality within cohorts as a proxy for economic disparities over the past two centuries , Baten and Mumme (2013) present evidence suggesting that inequality is associated with civil war onset. ...
Article
Scholarship has demonstrated that domestic economic inequality is related to a number of forms of intrastate conflict, such as civil wars and rebellions. There are good reasons to believe that it also has an impact on the initiation of militarized interstate disputes for diversionary reasons. Such use of external force may refocus popular attention and may reinforce the strong nationalist sentiment that tends to prevail in societies with substantial economic inequality. Our empirical results support this contention in democracies but, as expected, not in autocracies. At a time when domestic economic inequality is rising across the world, our findings may be timely.
... However, the elemental question of whether inequality-related grievances indeed motivate violence remains underexplored. State-or group-level research reports that individual-based (vertical) inequalities predict non-ethnic conflict (Bartusevičius 2014;Buhaug, Cederman, and Gleditsch 2014) and that horizontal inequalities predict ethnic conflict (Buhaug, Cederman, and Gleditsch 2014;Cederman, Gleditsch, and Buhaug 2013). However, macro-or meso-level relationships say little about individual motivations. ...
... Several macro-level studies have identified a positive association between (static) vertical inequalities and civil conflict (e.g., Bartusevičius 2014;Boix 2008;Buhaug, Cederman, and Gleditsch 2014; see also Bartusevičius 2019). Why have we not found an association between inequality and violence? ...
Article
Full-text available
Despite extensive scholarly interest in the association between economic inequality and political violence, the micro-level mechanisms through which the former influences the latter are not well understood. Drawing on pioneering theories of political violence, social psychological research on relative deprivation, and prospect theory from behavioral economics, we examine individual-level processes that underpin the relationship between inequality and political violence. We present two arguments: despite being a key explanatory variable in existing research, perceived lower economic status vis-à-vis other individuals (an indicator of relative deprivation) is unlikely to motivate people to participate in violence; by contrast, although virtually unexplored, a projected decrease in one’s own economic status (prospective decremental deprivation) is likely to motivate violence. Multilevel analyses of probability samples from many African countries provide evidence to support these claims. Based on this, we posit that focusing on changes in living conditions, rather than the status quo, is key for understanding political violence.
... Agbiboa and Maiangwa (2013) in their work on Boko Haram, religion and violence, seek to study the non-killing approach and resolving the trend of violence in the Northern part of the country, they found out that Given these points, Agbiboa and Maiangwa (2013) noted that, many diplomats and Nigerian citizens were killed in the bombing of the UN building, in August, 2011, as they were affected. The work of Bartusevicius (2013) observed that, some of the issues that motivate people in carrying arms against the government or a particular society is as a result of the feeling that they are unequal with other people of the same society due to lack of access to education, health facilities, and as well as social services. To some, there is no opportunity for them to participate in politics and as such they become available in participating or taking arms against the government. ...
... The work of Bartusevicius (2013) opined that, people are motivated to carry arms when they are unequal in terms of lack of access to social amenities such as health facilities and education. In the same way, Onapajo, Uzodike and Whetho (2012) Salaam (2012) revealed its findings that, poverty, mass illiteracy, endemic corruption, and unemployment as well as socio-political marginalization are the factors which necessitated the availability of the youths to engage in insurgency activities in Nigeria. ...
... The main argument is that inequalities between individuals in education increase the risk of political violence due to individuals' lack of opportunities. (Bartusevičius 2014;Nepal et al. 2011;Østby 2008;. Therefore, higher educated individuals have a lower risk of PVJ (Brockhoff et al. 2015). ...
... Finally, Gini coefficients of education have been used in many single-country studies not included here for obvious reasons of space. Gini coefficients of education are an important item in the broader analysis of inequalities of education, health, income, wealth, and many others, and the related issues of development strategies (Clemens 2004), conflicts (Østby et al. 2009;Bartusevičius 2014), trust and tolerance problems (Borgonovi 2012), electoral rights (Castelló-Climent 2008), institutions (Berthélemy 2006;Braga et al. 2011), social cohesion (Green et al. 2003), crime (Kelly 2000), and policies in other fields than economics (Unterhalter 2021). ...
... It could potentially lead to more conflict, as has been the case in recent decades. More inequality across religious groups, where some religious groups are at different ends of the income distribution, can lead to greater likelihood of social upheaval, political instability, tensions and violent conflict (Bartusevičius, 2014;Nordås, 2014). ...
Book
Full-text available
With around 23% of the global population, democratic countries of the former Indian sub-continent—India, Pakistan and Bangladesh—have an important role in shaping the future of global demography. Population continues to grow in these countries, although at varied pace and demographic diversity is wide. This paper explores the demographic changes in these Asian super-size economies and their political repercussions, with a special focus on the world’s largest democracy—India. To put the political consequences of varied demographic changes in perspective, we discuss the past and future demographic profile in these countries and the factors leading to such changes. In addition, we highlight the regional and religious diversity within these Asian democracies. The chapter elucidates that while India and Bangladesh have similar patterns of changes in age structure and fertility transition, Pakistan is comparatively at earlier stages of such demographic transitions. Consequently, in the coming three decades, while India and Bangladesh will show signs of ageing society, Pakistan will remain a young country. Striking differentials in regional and religious patterns of demographic heterogeneity are observed within India and Pakistan. Such demographic sketches have significant political repercussions at various levels. The chapter opens wider discussions on the political challenges of various demographic changes and illuminates the enormous importance of demographic patterns on the political order among these Asian Giants in population size.
... It could potentially lead to more conflict, as has been the case in recent decades. More inequality across religious groups, where some religious groups are at different ends of the income distribution, can lead to greater likelihood of social upheaval, political instability, tensions and violent conflict (Bartusevičius, 2014;Nordås, 2014). ...
Chapter
Full-text available
In this chapter, we provide a novel description of the religious composition of the poorest 20% of global society, around a billion citizens in the lowest part of the income distribution globally. The aim is to characterize a group of people with shared identity and sometimes political interests and behaviours that transcend national borders. Religions traditionally offer ways of coping with poverty, including offering meaning and hope for people in need. It could also steer financial wealth flows. The religious composition of a country or a region may also affect the level of social support and level of financial transfers, as welfare and social welfare systems can be organized through religious groups and indirectly affected by the degree of social welfare. Some religious groups have relatively large shares living in poverty. These religions may motivate different political behaviour if a large share of their compatriots is poor. For instance, this could motivate a stronger preference for transfers within the particular religions—but may also lower the ability to implement an effective, universal and sufficiently generous social security scheme as the economic burden would be too high on the rest of the community. We end the chapter by proposing poverty alleviating measures that take into account religious differences, and that may become thereby more effective.
... Beside the cultural determinants of protest, we find that economic development is not associated with demonstrations and riots, but less developed countries are more likely to engage in strikes. While these results seem to contradict some micro-level studies (Verba et al. 1995;Dalton et al, 2010;Welzel & Deutsch, 2012), Chenoweth and Lewis (2013), Cunningham (2013), Bartusevičius (2014) show similar findings at the macro level. Countries with a greater welfare states are more likely to report demonstrations and riots, but not riots. ...
Article
The literature considers nonviolent protests among the most important predictors of transitions towards democracy and democratic reforms. This study addresses the conditionsmaking countries more likely to experience nonviolent instead of violent forms of protest. While the literature emphasizes economic and political predictors of protest at the country level, we expand the study of nonviolent-vs-violent protest by incorporating cultural predictors. To do so, we use a newly developed time-pooled cross-sectional database covering an established set of orientations from the World Values Survey, known as “emancipative values”. Estimating the prevalence of these values at the country level as a time-varying predictor of nonviolent-vs-violent protests, we present the first comprehensively longitudinal analysis of the determinants of protests. Taking into account time-varying unobserved heterogeneity, persistence, the excess number of zeros and over-dispersion in protest event data, we find that emancipative values are a significant and positive predictor of the countries’ nonviolent protest levels, yet not violent protest levels.
Chapter
Full-text available
The Boko Haram conflict started in Nigeria from a city in the north-eastern part of the country known as Maiduguri in 2009 during the administration of President Umaru Musa Yar'Adua. It was a new thing to Nigeria considering the fact that the country was yet to witness terrorism and its activities. Boko Haram later continued after the death of President Yar'Adua, and they became stronger during the President Jonathan's administration. Boko Haram later transformed from an insurgent group into a terrorist group where they spread from Nigeria to other parts of African countries such as Chad, Niger, and Cameroon. This has it effects, impacts, and consequences on trade in Africa. The objective of this chapter is to draw the impact of terrorism on trade with specific attention to Nigeria and Chad.
Article
This paper investigates the impact of economic crises on income inequality. Important evidence has emerged that in the aftermath of crises politics becomes polarized and economists have linked this to greater differences in income due to crises. However, the evidence on whether crises are linked to divergent incomes is weak and plagued by i) the possibility of a reverse effect going from great disparity in incomes to major economic crises; ii) the persistent nature of income inequality; and iii) important measurement error in both the dependent and independent variables. We use the longest time stretch of available data on crises and types of crises (Reinhart and Rogoff, 2011) and income inequality (Solt, 2009), as well as decade averaged data, general method of moments and error-correction models to more credibly estimate the complex relationship between crises and inequality. We find strong evidence that currency, banking, inflation and debt crises increase inequality, particularly in the long run.
Chapter
Full-text available
Current thinking about development places individuals firmly at the centre of concern, the basic building block for analysis and policy. This is as true of the innovations led by Amartya Sen, which move us away from a focus purely on incomes to incorporate wider perspectives on well-being, as of the more traditional neoclassical welfare analysis which underpins most development policy. The present overriding concerns with reduced poverty and inequality, which stem from both types of analysis, are equally individual-focused. The Millennium Development Goals (MDGs), for example, are concerned with the numbers of individuals in poverty in the world as a whole, not with who they are, or where they live. Measures of inequality relate to the ranking of individuals (or households) within a country (or sometimes the globe). The issues of individuals’ poverty and inequality are, of course, extremely important, but they neglect a vital dimension of human well-being and of social stability: that is, the group dimension.
Chapter
Full-text available
This chapter examines the role of population pressure and inequality and their possible joint effects on routine violence across districts in Java. It looks at violence from the perspectives of neo-Malthusian and social justice. These issues are highly relevant for Java. Inhabited by 128 million people, it is the most populous island on earth and the most densely populated island in Indonesia, making it classically Malthusian. Furthermore, the effects of global climate change (e.g. of rising sea level) and environmental degradations are likely to deepen the Malthusian scenario for Java.
Article
Full-text available
Conflict researchers have increasingly stressed the importance of distinguishing between different categories of civil conflict, such as ethnic vs non-ethnic. However, the data on conflict categories has remained limited. This paper introduces the Categorically Disaggregated Conflict (CDC) dataset, which categorizes conflicts based on the two most commonly used distinctions, ethnic-vs-non-ethnic and governmental-vs-territorial, resulting in four conflict categories: ethnic governmental, ethnic territorial, non-ethnic governmental and non-ethnic territorial. While not the first of its kind, the CDC contains a number of novel features. Aside from its unique conceptualization of ethnic conflict, the CDC provides coding of the key component variables (language, religion and “race”), allowing users to re-code ethnic/non-ethnic conflicts into several alternative lists (e.g. religious/non-religious). Furthermore, the CDC provides detailed descriptions documenting coding choices for every single conflict, allowing users to track individual coding decisions. To demonstrate the value of the CDC, this paper replicates a recent study by Cederman, Gelditsch and Buhaug, based on the ACD2EPR—the only extant alternative to the CDC. The findings of the replication analysis challenge some of the key conclusions of the original study, substantiating the need for alternative categorically disaggregated datasets.
Article
Draws on theories of sociopolitical violence and revolution in an attempt to explain the origin of the widespread popular political mobilization that has played a major part in Central America's recent rebellions. The study employs aggregate data from the 1950s through the 1980s as well as descriptive data to explore how differences in the rate and nature of economic growth, income and wealth distribution, and governmental response to unrest may have contributed to rebellions in Somoza's Nicaragua (1977-1979), El Salvador (since 1979), and Guatemala (since 1978) and to the lack of rebellions in Honduras and Costa Rica. The evidence strongly suggests that Central America's rapid growth of export agriculture after 1950 and industrialization after 1960 markedly reduced the relative and absolute living standards of the working class, who then mobilized to demand redress of their grievances. Where the state responded accommodatingly and with limited repression (in Costa Rica and Honduras), opposition mobilization stagnated or subsided. Where the state did not ameliorate growing inequality and employed heavy repression, opposition mobilization and unity increased and led to a broad, rebellious challenge to regime sovereignty. -from Author
Article
Coherent democracies and harshly authoritarian states have few civil wars, and intermediate regimes are the most conflict-prone. Domestic violence also seems to be associated with political change, whether toward greater democracy or greater autocracy. Is the greater violence of intermediate regimes equivalent to the finding that states in political transition experience more violence? If both level of democracy and political change are relevant, to what extent is civil violence related to each? Based on an analysis of the period 1816-1992, we conclude that intermediate regimes are most prone to civil war, even when they have had time to stabilize from a regime change. In the long run, since intermediate regimes are less stable than autocracies, which in turn are less stable than democracies, durable democracy is the most probable end-point of the democratization process. The democratic civil peace is not only more just than the autocratic peace but also more stable.