Download full-text PDF

Receptivity to Violence in Ethnically Divided Societies: A Micro-Level Mechanism of Perceived Horizontal Inequalities

Article · October 2015with183 Reads
DOI: 10.1080/1057610X.2015.1084162
Abstract
Although past scholarship shows that group inequalities in economic and political power (“Horizontal Inequalities”) correlate with dissent, violence, and civil wars, there is no direct empirical test of the perceptual explanation for this relationship at the individual level. Such explanation is vital to understanding how integration, inclusion in power-sharing agreements, and exclusion from political power filter down to mass publics. Moreover, subjective perceptions of group conditions do not always correspond to objective group realities. We hypothesize subjective perceptions attenuate the effect of objective exclusion on support for violence in ethnically divided societies. Cross-national comparative multilevel analyses of the 2005/6 Afrobarometer dataset (N D 19,278) confirm that subjective perceptions both amplify the effect of exclusion on acceptance of violence and alter the readiness of included groups to dissent. These findings carry implications for research, state-building, and conflict management.
Figures
Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=uter20
Download by: [Stanford University] Date: 30 October 2015, At: 12:48
Studies in Conflict & Terrorism
ISSN: 1057-610X (Print) 1521-0731 (Online) Journal homepage: http://www.tandfonline.com/loi/uter20
Receptivity to Violence in Ethnically Divided
Societies: A Micro-Level Mechanism of Perceived
Horizontal Inequalities
Dan Miodownik & Lilach Nir
To cite this article: Dan Miodownik & Lilach Nir (2015): Receptivity to Violence in Ethnically
Divided Societies: A Micro-Level Mechanism of Perceived Horizontal Inequalities, Studies in
Conflict & Terrorism, DOI: 10.1080/1057610X.2015.1084162
To link to this article: http://dx.doi.org/10.1080/1057610X.2015.1084162
View supplementary material
Published online: 16 Oct 2015.
Submit your article to this journal
Article views: 23
View related articles
View Crossmark data
Receptivity to Violence in Ethnically Divided
Societies: A Micro-Level Mechanism of Perceived
Horizontal Inequalities
DAN MIODOWNIK
Department of Political Science and Department of International Relations
The Hebrew University of Jerusalem
Jerusalem, Israel
LILACH NIR
Department of Political Science and Department of Comm unication
The Hebrew University of Jerusalem
Jerusalem, Israel
Although past scholarship shows that group inequalities in economic and political
power (“Horizontal Inequalities”) correlate with dissent, violence, and civil wars,
there is no direct empirical test of the perceptual explanation for this relationship at
the individual level. Such explanation is vital to understanding how integration,
inclusion in power-sharing agreements, and exclusion from political power filter
down to mass publics. Moreover, subjective perceptions of group conditions do not
always correspond to objective group realities. We hypothesize subjective perceptions
attenuate the effect of objective exclusion on support for violence in ethnically
divided societies. Cross-national comparative multilevel analyses of the 2005/6
Afrobarometer dataset (N D 19,278) confirm that subjective perceptions both amplify
the effect of exclusion on acceptance of violence and alter the readiness of included
groups to dissent. These findings carry implications for research, state-building, and
conflict management.
The recent eruptions of widespread violence and armed conflicts in Libya, Syria, and
Mali; mass unrest in Tahrir Square in Cairo, and in Athens; the Occupy Wall Street move-
ment in the United States; and election-related violence in Togo, Kenya, and Nigeria have
attracted followers from dif ferent social strata. Citizens from mainstream society, with
access to political and economic resources, as well as those who lacked such access, sup-
ported violence and dissent. This reflected widespread attitudinal legitimacy—across
social groups—for articulating dissatisfaction in more or less violent means. But why do
both privil eged and deprived membe rs of society support violence?
Received 19 April 2015; accepted 13 August 2015.
Address correspondence to Dan Miodownik, Department of Political Science, The Hebrew
University of Jerusalem, Mt. Scopus, Jerusalem 91905, Israel. E-mail: dan.miodownik@mail.huji.
ac.il
1
Studies in Conflict & Terrorism, 0:1–24, 2015
Copyright ! Taylor and Francis Group, LLC
ISSN: 1057-610X print / 1521-0731 online
DOI: 10.1080/1057610X.2015.1084162
Downloaded by [Stanford University] at 12:48 30 October 2015
To answer this question, we examine past studies linking inequalities and outbreak of
conflict, and test a key theoretical mechanism—individual subjective perceptions of
inequalities—in explaining the effect of objective horizontal inequalities.
1
Perceptions
are crucial to understanding how macro-level conditions, such as inclusion in government
power-sharing agreements, diffuse through society and are interpreted by individual
members of various groups.
People often misperceive their group condition, as our evidence will show, and these
misperceptions carry two profound consequences for research. These mismatches reveal
it is problematic to assume uniformity in the way different members of a group perceive
their group situation. First, not all members perceive their group is disadvantaged, and
consequently, not all members are susceptible to mobilization. Assuming group unifor-
mity thus overestimates the likelihood of mobilization among members of excluded
groups. Second, ignoring these mismatches might lead to underestimate the potential of
powerful groups to instigate violence and unrest. The propensity for supporting violence
among powerful groups is affected not by objective exclusion, but rather by a mispercep-
tion that their group is deprived. As we show, subjective perceptions of inequalities mat-
ter; they explain why individuals in either politically powerful or powerless groups
support dissent and accept violenc e.
This article is organized as follows: Af ter distinguishing between objective condi-
tions and subjective perceptions, we review theoretical expectations for the role of per-
ceptions in the literature on ethnic conflict and consider how both objective conditions
and subjective perceptions affect related outcomes, pro-violent attitudes and behavior .
Following the hypotheses statement, we detail the data and measures and then present the
main findings from our analyses. In the concluding section, we discuss the implications
of this work for future research, state-building, and conflict management.
Objective Inequalities, Subjective Perceptions, and Unrest
Objective Horizontal Inequalities
The effect of horizontal inequalities—political and economic—on political instability is
well established. Previous research has connected several forms of inequality with civil
unrest and violence. Inequalities in access to political power, such as exclusion from
power, note group exclusion from government decision making, power-sharing agree-
ments, and discrimination. Political inequalities are significant predictors of ethnic con-
flict in gener al
2
and onset of civil war.
3
Group exclusion from political power increases
the likelihoo d of group involvement in conflict.
4
Economic inequalities precipitate political instabil ity and conflict as well, although
findings are mixed. Economic disparity between culturally defined groups was identified
as one inequality dimension that preceded civil conflict.
5
However, others
6
did not find
that the disparities across individuals within a country precipitated civil war onset.
7
However, objective deprivation and subjective perceptions of deprivation are not
identical; inferring perceptions from objective conditions might lead to different conclu-
sions.
8
Moreover, perceptions of inequality do not always correspond to the observed
reality; in fact can these two can be substantially different.
9
For example, a study based
on Ghana and Nigeria found that while minority groups in Ghana accur ately perceived
their condition, the dominant group tended to misperceive its objective political status.
Furthermore, in Nigeria, members of the two dominant groups tended to perceive their
group as having less power than the other dominant group.
10
Other studies find that some
2 D. Miodownik and L. Nir
Downloaded by [Stanford University] at 12:48 30 October 2015
privileged objective conditions, such as high social status, education, and employment—
correlate positively with grievances, contrary to expectations.
11
Together, these findings
suggest that systematic mis perceptions, rather than accurate perceptions of power, tend to
occur among dominant groups.
Another striking illustration of the mismatch between objective and subjective condi-
tions emerges from examining objective indicators of group political status (Ethnic Power
Relations data
12
) together with individual perceptions of their group condition (obtained
from the Afrobarom eter Round 3, thirteen countries, over 17,500 respondents
13
). The
respondents evaluat ed whether their group fared politically better, worse, or similar to
other groups in the country. Cross-tabulating exclusion from power with subjective per-
ceptions of exclusion reveals that fully 35 percent of the respondents misperceive their
group’s political status. Similarly, respondents misperceive the economic status of their
group: nearly half, or 48.3 perc ent, of individuals’ perceptions mismatch with the group’s
objective condition. What is more, whereas 29.3 percent overestimate the group’s status,
a striking fifth of all respondents think that their group is disadvantaged, contrary to their
group status in reality.
Due to this gap between perceptions and observed reality, then, it is not clear
whether we can make valid inferences about the consequences of percei v ed condi-
tions from studies that focus on or measure only objective indi cators. The implica-
tion is that we should distinguish both analytically and empirically between
objective and s ubjec tive inequalities . Next, we re view past claims about the lin k
between perceptions and unrest.
Subjective Perceptions of Inequalities
How individuals perceive their group status in a country is an important mechanism for
understanding support for violence. Past research sugges ts that these perceptions are
instrumental in linking groups’ objective conditions and conflict. Group grievances are
“intersubjectively perceived phenomena.”
14
Perceptual mechanisms for understanding
group behavior are important, because people often act on the basis of a socially mediated
understanding of their conditions, rather than the conditions themselves: “[P]eople take
actions on the basis of their perceptions of others and of their relative position rather than
actual inequality.”
15
Perceptions breed discontent. A perception of relative deprivation can emerge when
individuals compare their actual and ideal group situation to other groups: that is, “the
feeling of an individual who lacks some status or condition that he thinks he should have
... by reference to what some other group or person has.
16
This perceived gap between
the current situation and the situation group members believe they should be in creates
discontent and increases the potential for violence.
17
Successful mobilization to action
hinges on people’s awareness that their group is maltreated and the diffusion of resent-
ment toward their oppressors.
18
Objective political and economic asymmetries or differences between groups can
thus be “transfor med into grievances through a process of gr oup comparison.
19
Sub-
jective social comparisons “produ ce very real pe rce ptio ns o f inequity and feeli ngs of
discontent”
20
Social comparison with other groups triggers resentment, a sense of
grievance or i nsult: a “feeling of being politically dominated by a group that has no
right to be in a superior position.”
21
Feelings o f u njust treatment,
22
or unfair punish-
ment, lead to strong emotional reaction, “making people more willin g to undertake
risky actions.”
23
Framing group conditions in terms of injustice thus plays an
Receptivity to Violence 3
Downloaded by [Stanford University] at 12:48 30 October 2015
important role in crystallizing ind iv i dua l readi ne s s for action.
24
Perceptions of
inequality and injustice arethereforenotsimplygrievanceindicators,butarea
“mobilizational resource.”
25
Perceptions thus figure in major works that theorize the connection between group
grievances and opportunities, on the one hand, and violent support, on the other. Percep-
tions operate as trigge rs of both affective and cognitive mechanisms that alter the readi-
ness for action. Perceived benefits, in the form of selective incentives to participate in
violent action, are pivotal both in recruiting and in cementing social suppor t for guerrilla
armies.
26
An important point in Collier and Hoeffler’s contribution is that perceiving
(accurately or not) the opportunity for profit-making drives the dynamics of civil war
insurgency.
27
People often act on misperceptions rather than objective opportunity for
rebellion. This “illusory” opportunity for rebellion may have devastating outcomes as a
result of a miscalculation.
Whereas Collier and Hoeffler’s work theorizes on perceived benefits, others have
focused on the perceived cost of risky behavior. Willingness to take costly action fluctu-
ates with the signals people receive about the number of people who share that view and
are willing to engage in that action.
28
The perception of a greater number of participants
increases the readine ss of each individual to take part in an action, minimizing the per-
ceived risk of such engagement.
29
Problem Statement
Theoretical arguments on the importance of perceptions thus seem fairly consensual. To
be mobilized, individuals need to perceive either that their group has suffered some injus-
tice or that some profit can be gained, individually or collectively. Yet empirical evidence
for the role of perceptions in triggering mobiliza tion to violence is scarce and inconsis-
tent. A review of the existing evidence reveals three problems arising from past work:
(1) no direct test; (2) ecological fallacy; (3) use of objective conditions as proxy for sub-
jective perceptions.
First, major contributions to the literature on civil wars employ the theoretical mech-
anism of perceptions without directly testing it, leaving it as an underlying assumpt ion.
One work, for example, argues that the greater the share of natural resources in a country,
the greater the perceived benefit associated with seizing the resources, and consequently
the higher the probability of civil war onset.
30
However, percept ions are theoretically
assumed, rather than tested.
31
Second, past studies have tested individual-level mechanisms at a higher level of
analysis. Classic studies of group mobilization theorized on individuals’ receptivity to
rebellion as a function of perceived deprivation.
32
Whereas the mechanism (perceived
deprivation) was theorized at the individual and group levels, the empirical test was at the
polity level. Aggregate correlations were used in testing the main proposition on per-
ceived deprivation and participation in violence, with the expectation that the “proportion
of a population that participates in violence ought to vary with the average intensity of
perceived deprivation.”
33
The third potential problem with past studies is the use of objective conditions as a
proxy for subjective individual-level perceptions. For example, researchers point out that
they “investigate the relationship between externally measured inequalities, rather than
self-perceived [inequalities], and conflict.”
34
The rationale, in most cases, is assumed
rather than empirically tested: “The validity of this approach rests on the assumption that
perceptions broadly reflect the observed reality.”
35
Many researchers follow this
4 D. Miodownik and L. Nir
Downloaded by [Stanford University] at 12:48 30 October 2015
approach, and do not see the distinction between perceptions and observed reality as
problematic. Others caution, “subj ective deprivation (the gap between expectations and
experience), ... may or may not correlate with actual deprivation.”
36
We concur that this
assumption is highly problematic, since there are considerable mismatches, as the afore-
mentioned evidence clearly shows.
How Objective and Subjective Inequalities Contribute to Violence
This brief review underscores the importance of subjective assessments of the group con-
dition. We suggest the term “subjective inequalities” to distinguish these perceptions
from objective conditions. Two subjective judgments are central to the processes of
arousing resentment, willingness to take risk, and subsequent mobilization: judgments
about the group’s status relative to others, and about equitable treatment of the group by
the government.
Perception that one’s group is worse o ff than other groups may spark off resent-
ment a nd mobili zation. G roup- minded (“fra ternal”) judgments , rather th an individual
aspirations for self-improvement, generate feelings of collective relative deprivation.
37
People assess the distance between the group’s situation and future aspirations
throu gh perceptions about the group’s political and economic st at us relative to other
groups. Despite dearth of direct evidence about the role of perceived lower status on
violence, indirect evidence reveals that such perceptions undermine satisfaction with
democracy: A recent article found that the moreindividualsperceivedtheirgroupto
be politically and economically disadvantaged, the less satisfied they were with
democracy in their country.
38
In contrast, perceptions about the equitable and fair treatment by government reflect
judgments about justice and wrongdoing with regard to one’s group. These judgments
hold politically relevant consequences in assessments of distributive justice: that is, the
fair allocation of collective goods. Equity theory in social psychology “holds that in ren-
dering judgments about distributive justice, people seek to determine whether there is a
proportional relationship between their inputs ... and the outcomes they receive.”
39
Those who are under-benefited from the outcomes “feel angry and resentful”; and anger
“is generally associated with protest.”
40
Whereas past literature on equity theory focused on judgments concerning individu-
als’ well-being—for example, in business , the workplace, or intimat e relations
41
—others
made the explicit connection to collective and politically minded outcomes.
42
For exam-
ple, experimental evidence shows that participants in a perceived unjust (vs. just) condi-
tion exhibited a significantly increased behavioral intention to protest and take part in
collective action.
43
Indeed, two different Afrobarometer working papers show that per-
ceptions of inequity are associated with greater acceptability of political violence and a
higher likelihood of participation in violent actions and political protest.
44
Hypotheses
Our review of the literature suggests three main hypotheses. First, consistent with past
studies, we expect that objective horizontal inequalities will correlate positively with
social unrest (H1). Second, consistent with past theorizing, subjective feelings of depriva-
tion increase susceptibility to violence (H2). Taking these two statements together, we
expect that the effect of objective inequalities on violent outcomes will be even stronger
when the subjective perception is that the group has been disadvantaged (politically,
Receptivity to Violence 5
Downloaded by [Stanford University] at 12:48 30 October 2015
economically, and unfairly treated). We expect a positive interaction term between objec-
tive and subjective inequalities (H3). Formally, we test whether:
H1: Horizontal Inequalities (objective) increase support for violence;
H2: Perceptions of inequalities (subjec tive) increase support for violence; and
H3: Perceptions amplify the effect of Horizontal Inequalities on support for violence.
Method
Data
The analysis is based on data drawn from the final edition of the third round of the Afro-
barometer opinion survey series .
45
The survey was conducted in 18 African countries
between March 2005 and February 2006. In each country, a rando m (probability) sample
was drawn, with comparable estimates based on full coverage of all citizens of voting age
in participa ting countries (ns between 1,200 and 2,400). The final analysis includes
responses from 13 countries (N D 19,278).
46
Measures
Dependent Variables. Two dependent variables were used in the analysis: violence and
protest. Ideally, our dependent measures would have included both an attitudinal and a
behavioral measure of suppor t for violence. AB3 includes only attitudes toward violent
behavior, not participation in violence or rioting. The closest behavioral indicator that fit
our research aims was a self-report of protest behavior. We therefore use protest behavior
as the second dependent variable. Respondents were asked which of the following state-
ments they agreed with: “A: The use of violence is never justified in [respondent’s coun-
try] politics. B: In this country, it is sometimes necessary to use violence in support of a
just cause.” Responses ranged from 1 to 4 (1 D strong agreement with A; 4 D strong
agreement with B). Respondents who agreed with neither statement were assigned 5. The
variable violence was recoded to a 5-point index ranging from 0 (strong objection) to
4 (strong support) for the use of violence, with middle-of-the-road respondents as the
mid-category (M D 1.02; SD D 1.24; Range 0.44 [Benin] to 1.56 [Namibia]). Respond-
ents were also asked whether they “attended a demonstration or protest march,” or
whether they would intend to if they had a chance. Response categories of protest ranged
from 0 to 4 (“No, would never do this” to “Yes, often”; M D 0.74; SD D 0.96; Range
M D 0.42 [Ghana] to 1.01 [Mozambique]).
Political Inequalities: Exclusion. Respondents were posed the open-ended question,
“What is your tribe?”—asking them explicitly about their “ethnic or cultural group.”
Answers were compared to a list of ethnic groups derived from the Ethnic Power Rela-
tions dataset (EPR).
47
EPR includes a list of politically relevant ethnic groups within
countries. Although overall there is considerable overlap, the list of ethnic groups in the
Afrobarometer is longer than EPR’s. In some cases, respondents used different names to
refer to a group included in EPR; in others, respondents mentioned ethnic groups that
were politically irrelevant and hence excluded from EPR. We used information collected
from Joshua Project, a repository of ethnologic data, and Ethnologue
48
to resolve mis-
matches, successfully matching 87.7 percent (n D 16,740) of the respondents to one of
6 D. Miodownik and L. Nir
Downloaded by [Stanford University] at 12:48 30 October 2015
EPR’s groups. In line with past work, respondents were coded as excluded from power if
EPR classified their group as Irrelevant, Discriminated, Powerless, or enjoying Regional
Autonomy.
49
Members of groups that were absent from the EPR list or could not be clas-
sified to any category in the list (n D 1,954) were coded as excluded from power.
Responses endorsing a nationa l (rather than ethnic) identity and refusals were removed
from the analysis (n D 386). The final variable in the analysis, political exclusion, is a
dichotomy: 1 D respondents whos e ethnic group is excluded from power (M D 0.24;
SD D 0.43; Range 0 [Ghana, all included] to 0.73 [Mozambique]).
Economic Inequalities. To operationalize economic inequalities, we drew on Afrobar-
ometer interviewers’ assessment of the living conditions of respondents, such as the pres-
ence of basic services in the primary sampling unit/enumeration area. Examples of such
services include electricity grid, piped water system, a sewage system, and paved roads.
Because these sampling units (respondents’ districts in most cases) were quite homoge-
nous, ethnically, and external observers provided the answers, we are confident these are
reliable and valid proxies of the group’s objective economic conditions. Answers to these
questions were combined into a single, internally consistent economic factor (Cronbach’s
alpha D .78). Each respondent was assigned a group economic condition value corre-
sponding to the average econom ic factor of its members, and each country was assigned a
country economic condition value corresponding to the mean economic factor of its citi-
zens. The group economic condition measure was normalized, dividing it by the country’s
economic condition value. This resulted in values smaller and larger than 1, denoting
groups’ economic advantage (or disadvantage) relative to the country mean. Categoriza-
tion of these values to five quintiles yielded the final scale in the analysis, whose
values ranged from 0 (very weak economic conditions) to 4 (very strong economic condi-
tions). Groups’ mean economic condition varied across countries, ranging from 1.59
(Mozambique, with 71 percent of the p opulation in economically weak or very weak
groups) to 2.27 (Benin, with 57.7 percent of the population in economically strong or
very strong groups) ( M D1.97, SD D 1.41).
50
Subjective Inequalities (Perceptions). Perceptions of (low) relative group status and
unfair governmental treatment of one’s group were operationalized as subjective inequal-
ities. Respondents’ perceptions of political and economic lack of status were derived
from two questions. The first asked respondents to think about the group’s economic con-
ditions, and the second asked whether the group’s political conditions were “worse, the
same as, or better than other groups in this country.” Answers ranged on a 5-point scale,
from 0 (much better) to 4 (much worse). Perceptions of political status ranged between
1.47 in Namibia to 2.42 in Uganda (M D 2.08, SD D 1.03). Average perceptions of eco-
nomic status ranged between 1.57 in Namibia and 2.79 in Malawi (M D 2.18, SD D
1.02). With rega rd to government unfairness, respondents’ percept ions of equity (fair-
ness) were derived from their responses assessing whether their identity group was
“treated unfairly by the government.” Answers ranked on a 4-poi nt scale, from 0 (never)
to 3 (always; M D 0.89; SD D 1.01). Perceptions of unfair treatment across countri es
ranged from weak (M D 0.39, Senegal) to strong (M D 1.40, Nigeria).
Control Variables. To rule out the possibility that perceptions of group conditions and
willingness to use violence are an outcome of personal experiences (endogeneity prob-
lem), we measured whether respondents or their family members have been victimized,
as well as their potential exposure to violence in proximity to their place of residence.
Receptivity to Violence 7
Downloaded by [Stanford University] at 12:48 30 October 2015
Victim was the average of responses provided to two questions probing physical attacks
(“Over the past year, how often (if ever) have you or anyone in your family: Been physi-
cally attacked?”) and thefts (“Had something stolen from your house?”) Answers on a
5-point scale ranged from 0 (never) to 4 (always) (M D 0.37; SD D 0.63).
Violence proximity. Afrobarometer provides information on respondents’ district.
We assigned to each respondent the geographical coordinates of the major populated
city/town in the district. We used the Armed Conflict Locat ion and Event Dataset
(ACLED)
51
to obtain information concerning violence against civilians at any time
between 1997 (the earliest year included in ACLED) and 2004 (the year before the
administration of AB3). We then used ArcG IS to measure the distance from the
respondents’ location to the nearest violent location. The final variable in the analyses is
a dichotomy, coded 1 if the respondent resided within a 5-kilomet er radius from a violent
location, 0 otherwise. This distance cutoff made violence both visible (i.e., within
respondents’ vicinity) and proximate (i.e., having a real effect on respondents’ potential
behavior). The distribution ranges between the low 1.3 percent of the respondents living
close to a violent location in Botswana, and 67.8 percent in proximity to violence in
Uganda ( M D 0.30, SD D 0.46).
In addition, we controlled for respondents’ socioeconomic status: education, was
measured on the Afrobarometer questionnaire on a 10-point continuum ranging from “No
formal schooling” (0) to “Post-graduate” (9). Responses were recoded into a 3-point scale
(0 for no formal schooling to some primary schooling; 1 for primary schooling and some
secondary schooling; and 2 for respondents who completed secondary education and
beyond; M D 0.88, SD D 0.81). Livi ng conditions were measured on a 5-point scale rang-
ing from 0 (very bad) to 4 (very good) (M D 1.06, SD D 1.46). Unemployment, which
was measured on a 6-point scale from 1 (unemployed, not looking) to 6 (employed, full
time), was collapsed into a 4-point scale ranging from 0 for unemployed (not looking),
1 for unemployed (looking), 2 employed part time (looking and not), and 3 employed full
time (looking and not) (M D 1.30, SD D 1.17). Age (reversed) was recoded into a 4-point
scale: 0 (65C), 1 (45–64), 2 (30–44), 3 (18–29) (M D 2.10, SD D 0.91; non-scaled age
M D 36.01, SD D 14.57). Gender (50.1 percent male, 49.9 percent female) was coded 1
for females and 0 for males. Residence was dichotomized as urban (D 1) or rural (40.2
percent urban, 59.8 percent rural).
Country level. Additional country-level controls were taken into account as alterna-
tive explanations; these are based on known predictors in the literature of civil war onset.
52
Political regimes that are neither democratic nor autocratic have been shown to be more
susceptible to conflict. To preserve consistency with previous work, a dichotomous control
variable anocracy was created, with values of “1” noting countries whose Polity IV score
was between C5 and –5.
53
In addition, we controlled for ethnic fractionalization of the
country. We measured fractionalization based on the politically relevant ethnic group
(PREG),
54
a fractionalization index that is similar to ethno-linguistic fractionalization
(ELF), except that it excludes all groups that are irrelevant to the major political cleav-
age in a given country.
55
Gross Domestic Product (GDP) per capita and population size
may also play an important role explaining support for violence and dissent. Data on
GDP Per Capita in U.S. dollars for 2005 and population size (millions) was obtained
from the United Nations Statistical Database.
56
To account for the possibility that ter-
rain type provides opportunities for hideouts and facilitates violence and war, we
employed a measure of the percent of the country covered by mountains.
57
Oil produc-
tion was dichotomized: oil-producing countries (D 1, else 0) are more likely to experi-
ence civil war and violence.
58
8 D. Miodownik and L. Nir
Downloaded by [Stanford University] at 12:48 30 October 2015
Lastly, to account for possible country level endogeneity (i.e., that citizens of coun-
tries that experienced armed conflict may differ from citizens in more peaceful societies),
we included a dummy variable indicating if a country had gone through periods of armed
conflicts with at least 25 deaths since 1945 (D 1, else 0).
59
Table A1 in the Appendix
details univariate descriptives of all the variables in the analysis.
Results
How do subjective inequalities alter receptivity to violence and political dissent? To
answer this question, we used Multi-Level Modeling (MLM)—a statistical modeling
technique that is most suitable to analyze individual-level and country-level properties
simultaneously. Unlike linear regression, MLM is optimal to estimate individuals-within-
countries nested models without violating the assumption of observations independence.
We began by fitting two unconditional models (identical to one-way ANOVAs) to test
systematic cross-country variability in each of the two dependent variables. The models
are formally expressed as:
Level 1: Y
ij
D b
0j
C r
ij
Level 2: b
0j
D g
00
C u
0j
Combined : Y
ij
D g
00
C u
0j
C r
ij
The level 1 model indicates the i-th respondent ’s value on the depend ent variable (i.e.,
receptivity to violence) as a function of mean violence receptivity in country j (b
0j
), plus
a residual (r
ij
/ that reflects a respondent’s difference around its own country’s mean.
Level 2 indicates a country’s mean violence receptivity (b
0j
) as a function of the grand
mean (g
00
/ plus country-specific deviation (u
0j
/ from that mean. The combined equation
results from substituting the level 2 equation into level 1.
Table 1 shows the results of the unconditional means estimation. In the main text we
discussed results of a multilevel linear maximum likelihood models with country random
effects, estimated using the xtmied command in Stata 13. To check the robustness of the
results we ran the same models treating the dependent variables as either ordinal or linear,
with country-level fixed effects. Importantly, these checks suggest that alternative
Table 1.
Unconditional models (ANOVA) results
Model 1 Protest behavior Model 2 Violence receptivity
Fixed effects
Intercept
.g
00
)
.745
****
(.051)
.999
****
(.082)
Random Effects
Country level .t
00
) .033
**
.087
**
Individual level .s
2
) .893
****
1.471
****
log likelihood ¡25622.9 ¡30388.7
Observations 18794 18833
Countries 13 13
These unconditional linear models use a five-point scale of protest behavior and violence recep-
tivity as dependent variables. The results are linear maximum likelihood with country random
effects, estimated using the xtmied command in Stata 13.
*
p < .1,
**
p < .05,
***
p < .01,
****
p < .001
Receptivity to Violence 9
Downloaded by [Stanford University] at 12:48 30 October 2015
measurement levels of the dependent variables and respective statistical analyses did not
alter the substantive inferences of the analysis (see Table A2 in the Appendix).
60
Results
in Table 1 indicate that variability in the individual support for protest and acceptability
of violence can be attributed to cross-country differences. Grand mean support for protest
(Model 1) is g
00
D 0.745, and the variance component indicates that mean support for
protest significantly varies within and across countries (t
00
D 0.033, s
2
D 0.893). The
intra-class correlation (ICC), the proportion of variability between countries, indicates
that about 3.6 percent ((0.033/[0.033 C 0.893]) D 0.036) of the variation in respondents’
protest behavior occurred between countries.
Model 2 indicates that the grand mean of violence receptivity is g
00
D 0.999, and the
variance components suggest a statistically significant variability within and between
countries (t
00
D 0.087, s
2
D 1.471). ICC indicates that 5.6 percent ((0.087/[0.087 C
1.471]) D 0056) of the variation in respondents’ receptivity of violence is between coun-
tries. Country-specific conditions clearly account for only a small portion of the variation
in the dependent variables. Still, there are good empirical and substantive reasons to esti-
mate a multilevel model that takes into account both individual and country-level
factors.
61
Next we turn to the main effects of objective and subjective political and economic
inequalities on protest behavior and receptivity to violenc e. Table 2 presents the estima-
tion for Protest (Models 3–5) and Violence (Models 6–8). Models 3 and 6 test both H1
and H2. Consistent with our expectations, we found that objective inequalities increased
the likelihood of protest behavior; political exclusion increased the likelihood of protest
behavior (g D .150, p < .0001), and economic advantage decreased protest behavior
(g D –.021, p < .0001). Objective inequalities, however, did not affect violence receptiv-
ity. Con sistent with H2, we found that subjective perc eptions significantly correlated with
both violence and unrest; the association, however, was more complex than hypothesized.
As perceptions of inequity increased, both the likelihood of protest behavior (g D .046,
p < .0001) and the acceptability of violence increased (g D .092, p < .0001). Individuals
who perceived their group to be unfairly treated by the government were more likely to
protest and to consider violence as sometimes necessary.
On the other hand, we observed a more complex association between perceptions of
group status and the outcomes. Models 3 and 6 attest to a significant but negative asso-
ciation between perceptions of group disadvantage and protest behavior (g D –.034,
p < .0001) and negative but insignificant association with violence (g D –.021,
p D ns). In other words, contrary to expectations, individuals who perceived their
groups’ political status as inferior relative to other groups were less likely to support
protest and marginally less likely to accept violence. Moreover, the analysis indicated
that perceptions of a group’s economic inferiority did not affect the outcomes signifi-
cantly. Further below we will discuss the change in significance levels when interaction
terms were added to the models.
The four following models in Table 2 test the main effects of subjective and objec-
tive inequalities, their interactions or contingent effects (Models 4 and 7), and control for
other known predictors of violence and protest support (Models 5 and 8). Several of the
individual- and country-level controls correlated as expected with the outcomes, in line
with past research. For instance, moun tainous terrain, anocratic regime, and oil produc -
tion all correlated positively with protest support. Young individuals, men, and those who
were victi mized were more likely to support protest and violence. Proximity to violence
against civilians decreased suppor t for violence and had no effect on protest behavior.
62
A history of armed conflicts in the country had no effect on individual support for protest
10 D. Miodownik and L. Nir
Downloaded by [Stanford University] at 12:48 30 October 2015
Table 2
Individual and national predictors of protest behavior and violence receptivity
Protest behavior Violence receptivity
Model 3 Model 4 Model 5 Model 6 Model 7 Model 8
Fixed effects
Political Exclusion .150
****
.189
****
.181
****
.002 ¡.038 ¡.037
(.017) (.020) (.020) (.022) (.026) (.026)
Economic Advantage ¡.021
****
¡.022
****
¡.025
****
.0002 .006 .011
(.005) (.006) (.006) (.006) (.007) (.007)
Unfairness Perceptions .046
****
.035
****
.029
****
.092
****
.091
****
.086
****
(.008) (.008) (.008) (.010) (.011) (.011)
Pol Status Perceptions ¡.034
****
¡.043
****
¡.038
****
¡.015 ¡.018
*
¡.017
(.008) (.008) (.008) (.010) (.011) (.011)
Econ Status Perceptions ¡.020
**
¡.011 ¡.010 ¡.005 ¡.005 ¡.006
(.008) (.008) (.008) (.010) (.011) (.011)
Pol Exclusion £ Unfair ¡.084
****
¡.087
****
¡.074
***
¡.070
***
(.018) (.018) (.024) (.024)
Econ Advantage £ Unfair ¡.024
****
¡.027
****
¡.012* ¡.011
(.006) (.006) (.007) (.007)
Pol Excl £ Pol Status ¡.014 ¡.015 .089
****
.089
****
(.018) (.018) (.023) (.023)
Econ Adv £ Econ Status .014
**
.015
***
.022
***
.022
***
(.005) (.005) (.007) (.007)
(Continued on next page)
11
Downloaded by [Stanford University] at 12:48 30 October 2015
Table 2
Individual and national predictors of protest behavior and violence receptivity (Continued)
Protest behavior Violence receptivity
Model 3 Model 4 Model 5 Model 6 Model 7 Model 8
Individual Level Controls
Victim .098
****
.066
****
(.012) (.015)
Violence Proximity ¡.017 ¡.101
****
(.019) (.024)
Living Conditions .002 ¡.066
****
(.013) (.017)
Education .079
****
¡.026
*
(.011) (.015)
Employment .040
**
.054
**
(.017) (.022)
Female ¡.165
****
¡.031
(.015) (.020)
Age .044
****
.040
***
(.009) (.011)
Urban ¡.043
**
.034
(.017) (.022)
(Continued on next page)
12
Downloaded by [Stanford University] at 12:48 30 October 2015
Country Level Controls
Mountains .025
***
¡.016
(.006) (.014)
Population ¡.001 ¡.0001
(.001) (.003)
PREG Index ¡1.215
****
1.122
**
(.230) (.533)
Anocracy .458
****
.009
(.100) (.232)
Oil Producer .278
****
¡.257
(.076) (.177)
GDP Per Capita 4.34 £ 10
¡5**
7.87 £ 10
¡5*
(1.69 £ 10
¡5
) (3.96 £ 10
¡5
)
Armed Conflict .074 .224
(.081) (.189)
Intercept .753
****
.772
****
.713
****
1.009
****
1.009
****
.505
**
(.048) (.048) (.103) (.081) (.081) (.240)
Random Effects
Country level ðt
00
) .885 .891 .870 1.464 1.484 1.476
Individual level ðs
2
) .029 .029 .007 .084 .085 .039
log likelihood ¡25544.0 ¡22093.2 ¡21893.3 ¡30343.0 ¡26349.7 ¡26308.7
Observations 18794 16206 16026 18833 16290 16290
Countries 13 13 13 13 13 13
These linear models use a five point scale of protest behavior and violence receptivity as dependent variables. The results are linear maximum likelihood with
country random effects, estimated using the xtmied command in Stata 13.
*
p < .1,
**
p < .05,
***
p < .01,
****
p < .001
13
Downloaded by [Stanford University] at 12:48 30 October 2015
or violence. Overall, despite the strong controls we applied in these models, the focal sub-
stantive effects remained significant—the main and interactive effects of inequalities. The
rest of this section will focus on these effects.
In Models 4 and 7 we added interaction terms to Models 3 and 6, respectively. As
seen in these models, adding interaction terms did not affect, for the most part, the magni-
tude and direction of the reported main effects. For example, members of excluded groups
were more likely to protest (g D .181, p < .0001), and perceptions of unfair governmental
treatment correlated with higher protest (g D .029, p < .0001) and pro-violence attitudes
(g D .086, p < .0001).
To test Hypothesis 3, that subjective perceptions of group status and equity amplify
the effect of objective inequalities on v iolence and unrest, we took a closer look at the
interaction terms. As seen in Table 2, six of the eight interactions terms had statistically
significant coefficients. To aid the substantive interpretation of the coefficients, we
graphed the estimated effect of objective and subjective inequalities on the outcome vari-
ables, holding all other variables at their means.
63
Figure 1 illustrates in three different
panels the pattern of findings consistent with Hypothesis 3.
Panel 1A shows the effects of equity perceptions and objective political conditions on
protest behavior. As seen in the panel, exclusion from power increased participation in
protest. Moreover, and consistent with Hypothesis 3, subj ective perceptions that the gov-
ernment treats one’s group unfairly also contributed to protest behavior; the likelihood of
support for protest increased among members of included groups who perceived they
were treated unfairly (dark bar).
Panel 1B shows additional evidence in support of Hypothesis 3. Objective economic
disadvantage increased the likelihood of protest behavior. Moreover, subjective percep-
tions that one’s group was treated unfairly amplified the likelihood of expressing support
for protest (dark bar on the right-hand side). Among those who perceived the group was
treated fairly by the government, objective economic disadvantage did not make any dif-
ference in support for protest (gray bars).
Panel 1C demonstrates a similar pattern supportive of Hypothesis 3; the outcome is
acceptance of violence as necessary in one’s country politics. Panel 1C shows that mem-
bers of excluded groups who also perceived their group as politically disadvantaged were
more likely to support violence (dark bars). It is interesting to note a counterintuitive find-
ing: members of included groups who perceived their group to be politically advantaged
were also more likely to support violence (among included, the gray bar is higher than
the black bar). We revisit this finding in the concluding discussion.
Figure 2, on the other hand, charts a surprising effect of subjective perceptions.
Whereas the panels in Figure 1 illustrate that subjective perceptions amplified the effect
of objective grievances on unrest, Figure 2 illustrates the role of misperceptions in moti-
vating unrest among the least expected group members.
Panel 2A shows that subjective percepti ons incre as e d s up po r t f or v i ol e nce , ev e n
in contexts where political exclusion was not a motivator for unrest. Specifically,
Panel 2A shows perceptions of unfair governmental treatment of the group increased
support for violence even (especially) among members of included groups. Similarly,
Panel 2B demonstrates that members of economically adv a nt ag ed groups, who mis-
perceived their group to be economically disadvantaged, increased support for vio-
lence. A similar pattern emerges in Panel3C,whichchartstheeffectofeconomic
conditions and perceptions of economic status on protest behavior. It i s advantaged
group members, who misperceived their group to be disadvantaged, who protest.
Taken together, these findings show that subjective misperceptions motivated unrest.
14 D. Miodownik and L. Nir
Downloaded by [Stanford University] at 12:48 30 October 2015
Political inclusion does not assuage support for violence, if members of this group
perceive that the government treats them unfairly. The effect of economic disadvan-
tages on violence and protest is also drivenbysubjectiveperceptions.Ultimately,
subjective inequalit ies help explain the conditions under which objective inequalities
contribute to social instabil ity and unres t. We discuss the implications of these find-
ings in the next section.
Figure 1. Interactive effects of subjective and objective inequalities on protest and violence
(95 percent confidence intervals).
Receptivity to Violence 15
Downloaded by [Stanford University] at 12:48 30 October 2015
Discussion
Subjective perceptions of group conditions are an important factor in understanding
the link between objective group horizontalinequalitiesandviolence.Although
major works in the literature on civil wars and violence acknowledge the theoretical
importance of perceptions, scantevidencesupportsthisimplicit assumption directly.
Figure 2. Conditional effects of subjective and objective inequalities on protest and violence
(95 percent confidence intervals).
16 D. Miodownik and L. Nir
Downloaded by [Stanford University] at 12:48 30 October 2015
This article fills the gap by t esting objective horizontal inequalities, subjective per-
ceptions of inequalities, and their combined effects on receptivity to violence and
protest behavior. Our findings provide direct evidence to the role of perceptions as
moderators of objective conditions.
Our study complements past studies in several ways. First, our analyses of the Afro-
barometer survey data provide strong support to inferences that were based on aggregate,
large-N country-year, and group-year datasets. We found that both objective political and
economic inequalities (H1) and subjective perceptions of group status (H2) correlate with
greater social unrest.
64
Second, we provide direct supportive evidence to implicit concep-
tualizations on the role of subjective judgments as motivators of unrest. Specifically, our
findings support the argument that perceived grievances about one’s group conditions
crystallize the effects of horizontal inequalities on violence.
65
Third, we show horizontal
inequalities increase the likelihood of social unrest, and especially if individuals perceive
their group as disadvantaged.
In addition to supportive findings with regard to past intuitively appealing argumen ts,
we als o found evidence to complement less-intuitive arguments in the literature. For
example, the fact that a group’s exclusion from power makes it more likely to rebel is
well-established and the reasons for it are well-understood. However, recent work finds
that under certain circumstances, included groups (members of power sharing agree-
ments) are likely to instigate armed conflict. Perhaps, as an example from Nigeria sug-
gests, this might occur because members of dominant groups misperceive their status.
66
Included groups engage in violence for at least three reasons: (a) the group has lost its for-
mer status (“downgraded”), (b) the group is underrepresented in the power arrangement,
and (c) an increased number of power sharing elites has destabilized the alliance
structure.
67
Our current study finds that people’s suppor t for violence “as necessary in politics” is
at its highest among members of politically included groups who perceive that their group
enjoys a higher political status (see Figure 1C). Similarly, we found that support for vio-
lence is highest among members of included groups who perceive the government treats
the group unfairly (Figure 2A). These findings highlight the potential contribution of mis-
matched perceptions to increasing suppor t for violence. Taken together, these findings
also attest to the importance of misperceptions in driving unrest. Moreover, our study
lends supportive evidence to a process by which ambitious elite leaders mobilize their fol-
lowers by cueing perceptions of group grievances or group entitlement to a larger share of
the political pie.
What is more, this study provides a more nuanced explanation of the purported effect
of objective economic conditions on violence. Economic disadvantage—a factor in vio-
lence instigat ion—was found to both increase (Figure 1B) and decrease (Figure 2B)
social unrest. This seemingly contradictory find ing can be reco nciled as a result of the
focal subjective perception that moderates the effect of group economic disadvantages on
the outcome: that is, whether the moderator is perceived government equity or perceived
economic status. Economic disadvantage coupled with percept ions of government unfair-
ness increased the likelihood of unrest (1B). These group members, who live in deprived
areas and depend more heavily on central government transfers, were motivated to protest
in demand for greater equality in the distribution of collective goods. On the other hand,
economic disadvantage decreased the likelihood of unrest for groups who correctly per-
ceived economic disadvantage (2B). These could be groups so disaffected and disempow-
ered that members feel helpless to procure the necessary means to utilize violence
effectively to ameliora te their condition.
Receptivity to Violence 17
Downloaded by [Stanford University] at 12:48 30 October 2015
Another noteworthy finding was that the alternative explanations (control variables at
the country level) for the occurrence of violence and protest predicted protest better than
support for violence (Table 2). Whereas five of the seven predictors were significantly
correlated with protest behavior, only two (PREG and GDP Per Capita) were marginally
significant predictors of violence receptivity.
68
The finding is noteworthy because the
controls were taken from the literature on civil wars and violent conflict, with the expecta-
tion they explain violence better than protest. Perhaps that expectation was too generous
given that the item wording on the questionnaire concerned attitudes toward violence,
rather than self-report of actual violent behavior.
An important counterargument to ours is that Afrobarometer respondents’ attitudes,
such as perceived government responsiveness or its corruption levels, influence protest
and violence rece ptivity, not misperceptions of group deprivation. We therefore tested
the robustness of our results controlling for such assessments: respondents’ perceptions
of national economic conditions, their satisfaction with democracy, elected officials
responsiveness, the state of corruption, and whether people have had to pay bribes. Add-
ing these controls did not fundamentally change the results discussed above (see Appe n-
dix, models 12 and 17, Tables S1 and S2 in the supplemental material online). The
additional controls affected the outcomes in predictable ways: perceptions of corrupt ion,
experience of bribe and dissatisfaction with democracy contributed to protest behavior
and violence receptivity, whereas officials’ responsiveness diminished support for dem-
onstrations and violence.
69
Importantly, however, results we reported above remained
significant even after applying these stringent controls.
Our study is not without limitations. First, although we found evidence to support
Hypothesis 3 (interactions, objective £ subjective inequalities), only six of the eight
terms were statistically sign ificant. We found no evidence that perceptions of government
unfairness moderate the effect of economic advantage on violence acceptability, nor that
perceptions of political inequalities moderate the effect of exclusion on protest behavior.
Future research would benefit from replicating and extending this study to determine
whether the lack of effect is substantively meaningful or merely byproduct of the data col-
lection and analysis. Second, we readily acknowledge the limitations of a cross-sectional
design in making causal inferences. Despite this limitation, we suppose it is safe to
assume that institutional and economic contexts are causally antecedent to individuals’
perceptions and attitudes. Third, while the study employed data from 13 countries, there
may be limits to the extent one can generalize from the results to regions outside sub-
Saharan Africa. Future research would do well to test the insights from the current study
on subjective and objective inequalities in other regions of the world.
70
To conclude, this article underscores the role of subjective perceptions of group sta-
tus and equity as important moderators, connecting objective political and economic
inequalities to protest and violence outcomes. These findings add to the literature, as dis-
cussed above, and carry two broad normative implications. The first concerns the risks of
misinformation and misperceptions, particularly in societies undergoing democratization
while struggling with the challenges of development and internal strife. Such conditions
might be conducive to mobilization attempts by irresponsible, self-interested leaders who
agitate discontent among group members by harping on grievances, even when objective
conditions of the group are better than others.
The second normative issue concerns genuine inclusivity and representation in
power-sharing agreements. The feeling that the governme nt treats one’s group unfairly,
even—or especially—if this group is included in the power structure, leads to greater
grievances and readiness for violence. These are groups, which, despite political
18 D. Miodownik and L. Nir
Downloaded by [Stanford University] at 12:48 30 October 2015
advancement, perceive that their inclusion is mere tokenism and that they have little to
lose by risking societal stability. The normative implications of subjective perceptions
thus extend beyond violence and protest to fundamental questions of democratic repre-
sentation, transparency, and accountability.
Supplemental Material
Supplemental data for this article can be accessed on the publisher’s website.
Notes
1. Lars-Erik Cederman, Nils B. Weidmann, and Kristian Skrede Gleditsch, “Horizontal Inequal-
ities and Ethnonationalist Civil War: A Global Comparison, American Political Science Review 105
(August 2011), pp. 478–495; Lars-Erik Cederman, Kristian Skrede Gleditsch, and Halvard Buhaug,
Inequality, Grievances, and Civil War (Cambridge: Cambridge University Press, 2013).
2. Ted R. Gurr, “Why Minorities Rebel: A Global Analysis of Communal Mobilization and
Conflict since 1945,” International Political Science Review 14 (April 1993), pp. 161–201; Andreas
Wimmer, Nationalist Exclusion and Ethnic Conflicts: Shadows of Modernity (Cambridge, UK:
Cambridge University Press, 2002); Susan Olzak, The Global Dynamics of Race and Ethnic Mobili-
zation (Stanford, CA: Stanford University Press, 2006).
3. Andreas Wimmer, Lars-Erik Cederman, and Brian Min, “Ethnic Politics and Armed Con-
flict: A Configurational Analysis,” American Sociological Review 74 (April 2009), pp. 316–337;
Lars-Erik Cederman, Andreas Wimmer, and Brian Min, “Why Do Ethnic Groups Rebel? New Data
and Analysis,” World Politics 62 (January 2010), pp. 87–119.
4. Gurr, “Why Minorities Rebel; Olzak, The Global Dynamics of Race and Ethnic Mobiliza-
tion; Cederman, Wimmer, and Min, “Why Do Ethnic Groups Rebel?”; Cederman, Weidmann and
Gleditsch, “Horizontal Inequalities and Ethnonationalist Civil War.”
5. Frances Stewart, “Horizontal Inequalities and Conflict: An Introduction and Some Hypoth-
eses,” in Frances Stewart, ed., Horizontal Inequalities and Conflict: Understanding Group Violence
in Multiethnic Societies (Houndmills, UK: Palgrave Macmillan, 2008), pp. 3–24; Gudrun Østby,
“Polarization, Horizontal Inequalities, and Violent Civil Conflict,” Journal of Peace Research 45
(March 2008), pp. 143–162; Gudrun Østby, “Inequalities, the Political Environment, and Civil Con-
flict: Evidence from 55 Developing Countries,” in Frances Stewart, ed., Horizontal Inequalities and
Conflict: Understanding Group Violence in Multiethnic Societies (Houndmills, UK: Palgrave Mac-
millan, 2008), pp. 136–159; Gudrun Østby, Ragnhild Norda
"
s, and Jan Ketil Rød, “Regional Inequal-
ities and Civil Conflict in Sub-Saharan Africa,” International Studies Quarterly 53 (June 2009), pp.
301–324.
6. James D. Fearon and David D. Laitin, “Ethnicity, Insurgency, and Civil War,” American
Political Science Review 97 (February 2003), pp. 75–90; Ha
"
vard Hegre, Ranveig Gissinger, and
Nils Petter Gleditsch, “Globalization and Internal Conflict,” in Gerald Schneider, Katherine Bar-
bieri, and Nils Petter Gleditsch, eds., Globalization and Armed Conflict (Lanham, MD: Rowman &
Littlefield, 2003), pp. 251–275; Paul Collier and Anke Hoeffler, “Greed and Grievance in Civil
Wars,” Oxford Economic Papers 56 (October 2004), pp. 563–595.
7. Horizontal inequalities are also distinguished in the literature from vertical inequalities (see
Stewart, “Horizontal Inequalities and Conflict”). The latter are typically individual-level measures
(e.g., income disparities) are aggregated to larger units of analysis (e.g., a state, see Collier and
Hoeffler, “Greed and Grievance in Civil Wars”). In contrast, differences between groups (or
regions) in access to political power, economic welfare, or social goods were termed horizontal
inequalities. This article focuses on perceived differences between groups, not individuals, that is,
horizontal inequalities.
8. On this point see Donald M. Taylor, Michael J. Wohl, Michael King, and Persia Etemadi,
The Psychology of Violent Conflict in Failing States: A Review of the Scientific Literature (Defence
Research and Development Canada Contract Report. Toronto, Canada: DRDC, 2010), p. 12.
9. Arnim Langer and Kristien Smedts, “Seeing is Not Believing: Perceptions of Horizontal
Inequalities in Africa,” CRPD Working Paper No. 16 (2013).
Receptivity to Violence 19
Downloaded by [Stanford University] at 12:48 30 October 2015
10. Arnim Langer and Unoha Ukiwo, “Ethnicity, Religion and the State in Ghana and Nigeria:
Perceptions from the Street,” in Frances Stewart, ed., Horizontal Inequalities and Conflict: Under-
standing Group Violence in Multiethnic Societies (Houndmills, UK: Palgrave Macmillan, 2008),
pp. 205–226.
11. Anderoju Oyefusi, “Oil and the Probability of Rebel Participation among Youths
in the Niger Delta of Nigeria,” Journal of Peace Research 45 (July 2008), pp. 539–555, 55 2
(Table IV).
12. The Ethnic Power Relations (EPR) dataset includes a list of politically relevant ethnic
groups within countries and classifies them by their relative political status (more details below, see
political inequalities: exclusion). See Wimmer, Cederman, and Min, “Ethnic Politics and Armed
Conflict”; Cederman, Wimmer, and Min, “Why Do Ethnic Groups Rebel?”
13. Michael Bratton, E. Gyimah-Boadi, and Robert Mattes, Afrobarometer Round 3: The
Quality of Democracy and Governance in 18 African Countries, 2005–2006 [computer file]. Study
No. 22981. Ann Arbor, MI: Inter-University Consortium for Political and Social Research [distribu-
tor] Version 2009-11-08. Available at http://dx.doi.org/10.3886/ICPSR22981.v1
14. Cederman, Weidmann and Gleditsch, “Horizontal Inequalities and Ethnonationalist Civil
War,” p. 481 (italics added).
15. Stewart, “Horizontal Inequalities and Conflict,” p. 18 (italics original).
16. Ted R. Gurr, Why Men Rebel (Princeton, NJ: Princeton University Press, 1970), p. 24.
17. Stephen G. Brush, “Dynamics of Theory Change in the Social Science: Relative Deprivation
and Collective Violence,” Journal of Conflict Resolution 40 (December 1996), pp. 523545, 527.
18. Roger D. Petersen, Understanding Ethnic Violence: Fear, Hatred, and Resentment in
Twentieth-Century Eastern Europe (Cambridge, MA: Cambridge University Press, 2002).
19. Cederman, Weidmann and Gleditsch, “Horizontal Inequalities and Ethnonationalist Civil
War,” p. 481.
20. Taylor et al., The Psychology of Violent Conflict in Failing States, p. 11.
21. Petersen, pp. 40–41.
22. Ibid., p. 56.
23. Stathis N. Kalyvas, The Logic of Violence in Civil War (Cambridge, MA: Cambridge Uni-
versity Press, 2006), p. 154.
24. William A. Gamson, Talking Politics (New York, Cambridge University Press, 1992);
Bert Klandermans, The Social Psychology of Protest (Oxford: Blackwell, 1997).
25. Cederman, Weidmann and Gleditsch, “Horizontal Inequalities and Ethnonationalist Civil
War,” p. 482.
26. Jeff Goodwin and Theda Skocpol, “Explaining Revolutions in the Contemporary Third
World,” Politics and Society 17 (December 1989), pp. 489–509; Mark I. Lichbach, “What Makes
Rational Peasants Revolutionary? Dilemma, Paradox and Irony in Peasant Collective Action,”
World Politics 46 (April 1994), pp. 382–417; Ana M. Arjona and Stathis Kalyvas, Preliminary
Results of a Survey of Demobilized Combatants in Colombia (Unpublished Manuscript, Yale Uni-
versity, 2006); Jeremy M. Weinstein, Inside Rebellion: The Politics of Insurgent Violence (Cam-
bridge, MA, Cambridge University Press, 2007); Macartan Humphreys and Jeremy M. Weinstein,
“Who Fights? The Determinants of Participation in Civil War,” American Journal of Political Sci-
ence 52 (April 2008), pp. 436–455.
27. Political entrepreneurs often appeal to others’ perceptions of opportunities to profit from
supporting insurgency (Collier and Hoeffler, “Greed and Grievance in Civil Wars”). Citing Hirsh-
leifer they classify “possible causes of conflicts into preferences, opportunities, and perceptions”
(Collier and Hoeffler, “Greed and Grievance in Civil Wars, p. 564).
28. Susanne Lohmann, “Dynamics of Informational Cascades: The Monday Demonstrations
in Leipzig, East Germany, 1989–91,” World Politics 47 (October 1994), pp. 42–101; Terri Mannar-
ini, Michele Roccato, Angela Fedi, and Alberto Rovere, “Six Factors Fostering Protest: Predicting
Participation in Locally Unwanted Land Uses Movements,” Political Psychology 30 (December
2009), pp. 895–920.
29. Anthony Oberschall, “Loosely Structured Collective Conflict: A Theory and an
Application,” Research in Social Movements, Conflict and Change 3 (1980), pp. 45–68; Timur
Kuran, “Now Out of Never: The Element of Surprise in the East European Revolution of 1989,”
World Politics 44 (October 1991), pp. 7–48; Lilach Nir, “Motivated Reasoning and Public Opinion
Perception,” Public Opinion Quarterly 75 (Fall 2011), pp. 504–532.
30. Collier and Hoeffler, “Greed and Grievance in Civil Wars.”
20 D. Miodownik and L. Nir
Downloaded by [Stanford University] at 12:48 30 October 2015
31. Ibid., p. 565. See also Cederman, Weidmann and Gleditsch, “Horizontal Inequalities and
Ethnonationalist Civil War,” p. 481, on assuming rather than testing subjective perceptions.
32. Ted R. Gurr, “A Causal Model of Civil Strife: A Comparative Analysis Using New Indices,”
American Political Science Review 62 (December 1968), pp. 1104–1124; Gurr, Why Men Rebel.
33. Gurr, Why Men Rebel, p. 9. While Gurr’s publication Why Men Rebel initially received
scholarly acclaim, prizes, and additional empirical validation (e.g., Ivo K. Feierabend, Rosalind L.
Feierabend, and Betty A. Nesvold, “The Comparative Study of Revolution and Violence,” Compar-
ative Politics 5 (April 1973), pp. 393–424; see Brush, “Dynamics of Theory Change in the Social
Science,” p. 534), his findings were subsequently criticized on both empirical and theoretical
grounds (e.g., Clarck McPhail, “Civil Disorder Participation: A Critical Examination of Recent
Research,” American Sociological Review 36 (December 1971), pp. 1058–1073; Edward N. Muller,
“A Test of a Partial Theory of Potential for Political Violence,” American Political Science Review
66 (September 1972), pp. 928–959; David Snyder and Charles Tilly, “Hardship and Collective Vio-
lence in France, 1830 to 1960,”American Sociological Review 37 (October 1972), pp. 520–532.
Much of the criticism concerned the blurred boundaries between levels of analysis, the “use of
aggregate measures of deprivation as indicators of individual motivation ... along with his failure
to show exactly how individual motivations are translated into social movements” (Brush,
“Dynamics of Theory Change in the Social Science,” p. 529).
34. Stewart, “Horizontal Inequalities and Conflict,” p. 18.
35. Ibid.
36. Humphreys and Weinstein, “Who Fights?,” p. 445, fn. 12.
37. Walter G. Runciman, Relative Deprivation and Social Justice: A Study of Attitudes to
Social Inequity in Twentieth Century England (Berkeley: University of California Press, 1966).
38. Valentin Gold, “Partitioning Ethnic Groups and Their Members: Explaining Variations in
Satisfaction with Democracy in Africa,” Peace Economics, Peace Science and Public Policy 8
(December 2012), pp. 1–13.
39. John T. Jost and Aaron C. Kay, “Social Justice: History, Theory, and Research,” in Susan
T. Fiske, Daniel T. Gilbert, and Gardner Lindzey, eds., Handbook of Social Psychology, Vol. 2, 5th
Ed. (Hoboken, NJ: Wiley & Sons, 2010), pp. 1122–1165, 1130.
40. Ibid., p. 1130.
41. J. Stacy Adams, “Inequity in Social Exchange,” in Leonard Berkowitz, ed., Advances in
Experimental Social Psychology, Vol. 2 (San Diego, CA, Academic Press, 2010), pp. 267–299;
Elaine Walster, G. William Walster, and Ellen Berscheid, Equity: Theory and Research (Boston,
MA: Allyn and Bacon, 1978).
42. For example, Donald. M. Taylor, Fathali M. Moghaddam, Ian Gamble and Evelyn Zell-
erer, “Disadvantaged Group Responses to Perceived Inequality: From Passive Acceptance to Col-
lective Action,” Journal of Social Psychology 127 (June 1987), pp. 259–272.
43. Ibid., experiment 2.
44. Ravi Bhavnani and David Backer, “Social Capital and Political Violence in Sub-Saharan
Africa,” Afrobarometer Working Paper 90 (2007); Matthew F. Kirwin and Wonbin Cho, “Weak
States and Political Violence in Sub-Saharan Africa,” Afrobarometer Working Paper 111 (2009).
45. Bratton, G yimah-Boadi, and Mattes, Afrobarometer Round 3. We use t he third of the
four publicly available Afrobarometer’s rounds because it included both et hnic group identifica-
ti on, an d items on protest and violence. The fir st round, AB1 included ite ms about protest, but
no t on violence, nor ethnic group identifiers. AB2 included items on violence and protest,
but not on ethnic identity. AB4 inc luded items on ethnic group, on protest behavior but not on
violence.
46. Five countries were omitted from this study: Cape Verde, Lesotho, Madagascar, Tanzania,
and Zimbabwe. The first four countries were either absent from, or coded as countries in which the
ethnic dimension is irrelevant, in the EPR dataset; Cederman, Wimmer, and Min, “Why Do Ethnic
Groups Rebel?” Respondents from Zimbabwe were not asked about perceptions of their own group.
The final dataset includes: Benin, Botswana, Ghana, Kenya, Malawi, Mali, Mozambique, Namibia,
Nigeria, Senegal, South Africa, Uganda and Zambia.
47. Wimmer, Cederman, and Min, “Ethnic Politics and Armed Conflict”; Cederman,
Wimmer, and Min, “Why Do Ethnic Groups Rebel?”
48. Joshua Project (2011). M. Paul Lewis, ed., Ethnologue: Languages of the World, 16th ed.
(Dallas, TX: SIL International, 2009). Available at http://www.joshuaproject.net (accessed 1
March 2012).
Receptivity to Violence 21
Downloaded by [Stanford University] at 12:48 30 October 2015
49. Wimmer, Cederman, and Min, “Ethnic Politics and Armed Conflict”; Cederman,
Wimmer, and Min, “Why Do Ethnic Groups Rebel?”
50. We considered alternative measures of local economic conditions, such as Demographic
and Health Surveys data (Østby, “Polarization, Horizontal Inequalities and Violent Civil Conflict”),
but some countries did not include the necessary local disaggregated data. We also constructed an
alternative measure of objective economic inequalities based on Afrobarometer’s Lived Poverty
Index (LPI). See Robert Mattes and Michael Bratton, “Learning about Democracy in Africa: Aware-
ness, Performance, and Experience,” American Journal of Political Science 51 (January 2007),
pp. 192–217. Our final measurement is chosen for two reasons: (a) the objective evaluation of living
conditions provided by the interviewers is more reliable and valid than respondents’ potentially
skewed subjective evaluation of their living status; (b) empirically speaking, using the alternative
measure does not affect the results (see Table 2 and Tables S1 and S2 [models 13 and 18] in the sup-
plemental material online for variable description and additional results).
51. Clionadh Raleigh, Andrew Linke, Ha
"
vard Hegre and Joakim Karlsen, “Introducing
ACLED: An Armed Conflict Location and Event Data,” Journal of Peace Research 47(September
2010), pp. 1–10.
52. See meta-analysis in Ha
"
vard Hegre and Nicholas Sambanis, “Sensitivity Analysis of
Empirical Results on Civil War Onset,” Journal of Conflict Resolution 50 (August), pp. 508–535.
53. Ha
"
vard Hegre, Tanja Ellingsen, Scott Gates, and Nils Petter Gleditsch, “Toward a Democratic
Civil Peace? Democracy, Political Change, and Civil War, 1816–1992, American Political Science
Review 95 (March 2001), pp. 33–48; Cederman, Weidmann and Gleditsch, “Horizontal Inequalities
and Ethnonationalist Civil War. But see James R Vreeland, “The Effect of Political Regime on Civil
War: Unpacking Anocracy,” Journal of Conflict Resolution 52 (June 2008), pp. 401–425.
54. Daniel N. Posner, “Measuring Ethnic Fractionalization in Africa,” American Journal of
Political Science 48 (October 2004), pp. 849–863.
55. Measures of ethnic fragmentation in general, and the popular ELF index, have been shown
to increase, decrease, or be unrelated to the likelihood of conflict (see review in Ravi Bhavnani and
Dan Miodownik, “Ethnic Polarization, Ethnic Salience, and Civil War,” Journal of Conflict Resolu-
tion 53 [February 2009], pp. 30–49).
56. United Nations Statistics Division, “Per Capita GDP at Current Prices—US Dollars,”
National Accounts Estimates of Main Aggregates (2010). Available at http://data.un.org; United
Nations Statistics Division, “Total Population, Both Sexes Combined,” World Population Pros-
pects: The 2010 Revision (2010). Available at http://data.un.org
57. Fearon and Laitin, “Ethnicity, Insurgency, and Civil War,” replication data.
58. Fearon and Laitin, “Ethnicity, Insurgency, and Civil War,”; Michael Ross, “A Closer Look
at Oil, Diamonds, and Civil War,” Annual Review of Political Science 10 (2006), pp. 265–300.
59. See Wimmer, Cederman, and Min, “Ethnic Politics and Armed Conflict.”
60. See also Peter Sandholt Jensen and Mogens K. Justesen, “Poverty and Vote Buying: Sur-
vey-based evidence from Africa,” Electoral Studies 33 (March 2014), pp. 220–232.
61. Andrew F. Hayes, “A Primer on Multilevel Modeling,” Human Communication Research
32 (October 2006), pp. 385–410, 349.
62. Follow-up sensitivity analyses, with different indicators of violence (battles, riots/pro-
tests), and varied distances between the respondent and the location of violence (20, 50, or 100 kilo-
meters) yielded results very similar to those reported here.
63. Interaction coefficients in MLM should not be interpreted in the same manner as linear regres-
sion coefficients (see Daniel J. Bauer and Patrick J. Curran, “Probing Interactions in Fixed and Multi-
level Regression: Inferential and Graphical Techniques,” Multivariate Behavioral Research 40(3)
[2005]), pp. 373–400; Kristopher Preacher, Patrick J. Curran, and Daniel J. Bauer, “Computational
Tools for Probing Interactions in Multiple Linear Regression, Multilevel Modeling, and Latent Curve
Analysis,” Journal of Educational and Behavioral Statistics 31 [Winter 2006], pp. 437–448).
64. For example, Why Men Rebel; Østby, “Polarization, Horizontal Inequalities and Violent
Civil Conflict”; Cederman, Weidmann and Gleditsch, “Horizontal Inequalities and Ethnonationalist
Civil War.”
65. Cederman, Weidmann and Gleditsch, “Horizontal Inequalities and Ethnonationalist Civil
War.”
66. Langer and Ukiwo, “Ethnicity, Religion and the State in Ghana and Nigeria.”
67. Wimmer, Cederman, and Min, “Ethnic Politics and Armed Conflict”; Cederman,
Wimmer, and Min, “Why Do Ethnic Groups Rebel?”
22 D. Miodownik and L. Nir
Downloaded by [Stanford University] at 12:48 30 October 2015
68. We examined a number of alternative country-level controls as indicators of previous vio-
lence: ethnic conflicts, length of conflict, civil war, and years since last conflict. None correlated sig-
nificantly with the outcomes.
69. Interestingly, perception of dire national economic conditions decreased support for protest
and did not affect violence receptivity. This echoes the positive and significant association between
GDP per capita and protest, contrasted to rather week association between GDP/capita and violence.
70. We thank the anonymous reviewer who suggested these insights complement studies of
militias and radical right movements of White supremacists in the United States.
Appendix
Table A1
Descriptive statistics
n Median Mean SD Min Max
Protest Behavior 18,598 0 0.7396 0.96131 0 4
Violence Receptivity 18,639 1 1.0268 1.24438 0 4
Political Exclusion 18,694 0 0.2446 0.42988 0 1
Economic Advantage 19,08 0 2 1.9655 1.41099 0 4
Unfairness Perfections 17,414 1 0.8939 1.00744 0 3
Pol. Statu s Perfections 17,706 2 2.0838 1.02545 0 4
Eco. Status Perfections 18,038 2 2.1799 1.02406 0 4
Victim 19,030 0 0.3671 0.62787 0 4
Living 19,021 0 1.0613 1.45671 0 4
Education 18,999 1 0.877 0.81163 0 2
Employment 19,021 1 1.3014 1.1722 0 3
Female 19,080 0 0.499 0.50001 0 1
Age 18,846 2 2.1044 0.90656 0 3
Urban 19,080 0 0.402 0.49032 0 1
Violent Proximity 19,080 0 0.2990 0.45783 0 1
Mountains 13 7.7 7.34 8.34 0 26.3
Population (Millions) 13 13.5 26.25 34.08 1.6 131.5
PREG 13 0.465 0.415 0.231 0 1
Anocracy 13 1 0.5385 0.5189 0 1
Oil Producer 13 0 0.4615 0.5189 0 1
GDP Per Capita (US$) 13 634 1504.77 1887.18 218 5468
Armed Conflict 13 1 0.6154 0.50637 0 1
Receptivity to Violence 23
Downloaded by [Stanford University] at 12:48 30 October 2015
Table A2
Individual and national predictors of protest behavior and violence
receptivity—Robustness checks
Protest behavior Violence receptivity
Ordered
logistic
Linear
FE
Ordered
logistic
Linear
FE
Fixed effects
Political Exclusion
.308
****
(.040) .181
****
(.020) ¡.050 (.040) ¡.036 (.026)
Economic Advantage
¡.059
****
(.011) ¡.024
****
(.006) .020
*
(.011) .011 (.007)
Unfairness Perceptions
.048
***
(.017) .029
****
(.008) .121
****
(.017) .086
****
(.011)
Pol Status Perceptions
¡.079
****
(.017) ¡.037
****
(.008) ¡.011 (.017) ¡.017 (.011)
Econ Status Perceptions
¡.003 (.024) ¡.010 (.008) ¡.009 (¡.017) ¡.005 (.011)
Pol Exclusion £ Unfair
¡.111
***
(.038) ¡.084
****
(.018) ¡.077
**
(.037) ¡.071
***
(.024)
Econ Advantage £ Unfair
¡.054
****
(.012) ¡.027
****
(.006) ¡.027
**
(.011) ¡.011 (.007)
Pol Excl £ Pol Status
¡.003 (.037) ¡.014 (.018) .181
****
(.037) .089
****
(.023)
Econ Adv £ Econ Status
.031
***
(.011) .015
***
(.005) .032
***
(.011) .021
***
(.007)
Individual Level
Victim
.191
****
(.024) .097
****
(.012) .098
****
(.023) .065
****
(.015)
Violence Proximity
.037 (.037) ¡.019 (.019) ¡.114
***
(.038) ¡.097
****
(.024)
Living Conditions
¡.030 (.027) .002 (.013) ¡.090
***
(.026) ¡.065
****
(.017)
Education
.192
****
(.023) .079
****
(.011) ¡.066
***
(.023) ¡.027
*
(.015)
Employment
.114
***
(.034) .042
**
(.017) .072
**
(.033) .054
**
(.022)
Female
¡.364
****
(.031) ¡.164
****
(.015) ¡.009 (.030) ¡.031 (.020)
Age
.132
****
(.018) .044
****
(.009) .066
****
(.018) .040
***
(.011)
Urban
¡.135
****
(.035) ¡.043
**
(.017) .064
*
(.034) .033 (.022)
Country Level
Mountains
.062
****
(.004) ¡.030
****
(.004)
Population
¡.003 (.001) ¡.001
*
(.001)
PREG Index
¡2.375
****
(.168) 1.486
****
(.176)
Anocracy
.735
****
(.078) .345
****
(.080)
Oil Producer
.818
****
(.055) ¡.621
****
(.056)
GDP Per Capita
8.65£10
¡5****
(1.16£10
¡5
)
1.70£10
¡4****
(1.14£10
¡5
)
Armed Conflict
¡.065 (.057) .455
***
(.057)
Intercept
.757
****
(.009) 1.063
****
(.012)
Cut 1
¡.158
**
(.075) .468
****
(.072)
Cut 2
1.640
****
(.077) 2.123
****
(.073)
Cut 3
2.530
****
(.080) 2.302
****
(.074)
Cut 4
3.581
****
(.088) 3.439
****
(.078)
Random Effects
Country level ðt
00
) x ðp # 2=3Þ .872 x ðp # 2=3Þ 1.479
Individual level ðs
2
) .475 .035 .089 .090
log likelihood ¡18103.3 ¡19874.8
Observations 16206 16206 16290 16290
Countries 13 13 13 13
The linear model and ordered logistic models use a five point scale of protest behavior and violence receptivity as dependent
variables. The results are generated with State 13. Ordered logistic denotes random effects ordered logistic regressio n estimated
using xtologit command. Linear denotes fixed effects regressions, implemented using xtreg. In the ordered logistic model the
coefficients are log-odds, and the level 1 standard deviation follows the logistic distribution and is defined as x ðp
2
=3Þ.
*
p < .1,
**
p < .05,
***
p < .01,
****
p < .001.
24 D. Miodownik and L. Nir
Downloaded by [Stanford University] at 12:48 30 October 2015
Project
http://ir.huji.ac.il/sites/default/files/internationalrelations/files/ir-call_for_candidates_2017.pdf
Project
To develop a robust, multi-level, cross-national model to explain relative regional political distinctiveness. The project begins by building on Hearl et al. (1996), exploring social and economic …" [more]
Discover more