Content uploaded by Ches Thurber
Author content
All content in this area was uploaded by Ches Thurber on May 23, 2024
Content may be subject to copyright.
Ethnic Barriers to Civil Resistance
Ches Thurber
Northern Illinois University
PREPRINT VERSION (EARLY 2018)
Published version available from author or at:
https://doi.org/10.1093/jogss/ogy018
Please cite as:
Ches Thurber, “Ethnic Barriers to Civil Resistance,”Journal of Global
Security Studies 3 (3), July 2018: 255–270.
Abstract
While ethnic cleavages have featured prominently in our understanding of civil wars,
attention to ethnic and social structures has been surprisingly absent from the recent
wave of research on civil resistance. Yet these structures likely have an important
impact on when and where we see nonviolent campaigns occur. This article argues that
the strategic logic of civil resistance presents high barriers to entry for politically
excluded ethnic minorities. Constraints on these groups’ ability to activate mechanisms
central to an exclusively nonviolent strategy either prevent them from getting a civil
resistance campaign off the ground, or deter them from ever attempting to do so. Using
original data on the ethnic composition of nonviolent and violent campaigns, I show
that nonviolent campaigns are less likely than violent ones to include participants from
politically disadvantaged ethnic groups, and also less likely to feature ethnic political
claims. Furthermore, I find that political exclusion and small group size reduce the
likelihood that members of an ethnic group will initiate a campaign of civil resistance.
Introduction
When they occur, civil wars play out along ethnic lines with astounding frequency.
According to a recent study by Denny and Walter (2014), 64 percent of civil wars
between 1946 and 2005 in the PRIO Battle Deaths Dataset Version 3.0 (Lacina and
Gleditsch, 2005) broke down along ethnic lines were initiated by groups expressing
political grievances on behalf of their self-identified ethnic group. Buhaug (2006)
similarly finds that ethnic conflicts make up a majority of cases in four major civil war
datasets. Do civil resistance campaigns follow a similar pattern?
Despite a few iconic cases of racially- or ethnically-motivated campaigns of nonviolent
resistance, such as the Anti-Apartheid struggle in South Africa, the civil resistance
literature has paid comparatively little attention to the role of ethnicity in shaping this
form of conflict. Instead, research on nonviolent conflict has focused primarily on
identifying factors related to movement agency – strategic planning (Ackerman and
Kruegler 1994), the use of diverse tactics (Schock 2005), and the maintenance of
nonviolent discipline (Sharp 1973) – that these scholars argue matter most when it
comes to explaining the trajectories and outcomes of such campaigns. Some studies that
seek to identify predictors of the onset nonviolent campaigns include measures of
ethnicity in their list of covariates, but generally focus more on measures of economic
and political structure, with less theoretical attention to the role that social structure
might play (Butcher and Svensson 2016; Chenoweth and Ulfelder 2017).
Svensson and Lindgren (2011) are an important exception: they show that the success
rate for nonviolence in non-democracies is lower when opponents and the regime are
divided along ethnic lines. Still, this does not address how ethnic cleavages might
inform when and where campaigns employing civil resistance materialize in the first
place.
In this study, I argue that an ethnic group’s level of political inclusion and demographic
size have a decisive impact on the likelihood that its members will initiate a civil
resistance campaign. Following the strategic logic of nonviolent action elaborated in the
civil resistance literature (see e.g. Chenoweth and Stephan 2011), I expect that small and
excluded ethnic groups face high hurdles when considering nonviolent resistance.
Specifically, the size of the ethnic group places a ceiling on its ability to generate mass
popular mobilization, and a lack of social ties to members of the ruling regime likely
makes inducing loyalty shifts from elites particularly challenging. Additionally, ethnic
cleavages between the opposition and ruling elites may make it easier for the state to
call upon its security forces to exact brutal repression on nonviolent activists.
As a result, movements whose supporters come from smaller, politically excluded
ethnic groups will be more pessimistic about the potential viability of a nonviolent
strategy, relative to movements whose supporters come from high-status, majority
groups. They will therefore be less likely to choose to even attempt a campaign using a
strategy of civil resistance, or if they do, will be less likely to get such a campaign off the
ground.
Using new data that links campaigns in the NAVCO 2.0 dataset (Chenoweth and Lewis
2013) to politically relevant ethnic groups in the EPR dataset (Vogt et al. 2015), I show
that politically excluded ethnic groups initiate only 21 percent of nonviolent conflicts
globally. Conversely, these groups initiate 52 percent of violent conflicts. Furthermore,
while 57 percent of violent campaigns feature an ethnic political claim, only 20 percent
of nonviolent ones do so. Using politically relevant ethnic groups as the level of
analysis, I find that politically excluded groups are less than one third as likely to
initiate a nonviolent campaign, compared to high-status groups. In addition, group size
is strongly correlated with the likelihood of nonviolent campaign initiation.
The findings challenge the conventional wisdom among civil resistance scholars and
advocates asserting that structural conditions matter little for civil resistance
movements. In fact, social structures, as opposed to the more frequently examined
economic and institutional structures, play a major role in conditioning the emergence
of nonviolent campaigns. By demonstrating how group characteristics raise and lower
barriers to the adoption of exclusively nonviolent repertoires of contention, the study
also highlights the importance of disaggregating actors to better understand the
dynamics of nonviolent resistance.
The rest of the article proceeds as follows. After first briefly reviewing scholarship on
the role of ethnicity in violent and nonviolent conflict, I then present a theory for how
ethnic structures and practices of political exclusion shape opportunities for civil
resistance. Next, I introduce the Ethnic Groups in Contention (EGC) dataset and
compare the ethnic composition of violent and nonviolent campaigns. I then present a
series of multinomial logistical regressions with the group-year as the unit of analysis
and examine the impact of power and demographic structures on the probability of
ethnic groups initiating nonviolent and violent campaigns. I conclude with a discussion
of the findings as well as implications for research, policy, and advocacy.
Ethnicity and Civil Resistance
Over the last three decades, scholarship on ethnic conflict and civil war has frequently
intersected (see e.g. Horowitz 1985; Posen 1993; and Harff and Gurr 1994). Some
scholars have, of course challenged whether ethnic grievances truly drive large-scale
armed conflicts (Fearon and Laitin 2003; Mueller 2000), while others have questioned
the concept of ethnicity altogether, observing that rather than being fixed, rigid, and
primordial, ethnic identities are fluid, fragmented, multidimensional, and often partly
shaped by political institutions and processes (Kasfir 1979; Waters 1990; Laitin 1998;
Posner 2005). But even if we allow for the fact that the salience of ethnic identity is
highly contingent and that diverse ethnic groups cooperate far more often than they
fight (Fearon and Laitin 1996), civil wars are nevertheless fought along ethnic lines more
often than along any other type of cleavage (Denny and Walter 2014).
Explanations for ethnic conflict fall into two broad camps. The first stresses the
relevance of ethnic group grievances as a motivating factor leading to high-risk
collective action. These explanations largely emphasize patterns of discrimination and
repression (Horowitz 1985; Gurr 1994), political exclusion (Wimmer, Cederman, and
Min 2009; Cederman, Wimmer, and Min 2010), and inter-group economic inequalities
(Cederman, Weidmann, and Gleditsch 2011) as triggers leading to the outbreak of
ethnic conflict.
A second tradition highlights the ways in which ethnic social structures may be
favorable for organizing rebellion. Ethnic ties may form a basis for incentivizing
recruitment, maintaining trust, and enforcing discipline (Gates 2002; Humphreys and
Weinstein 2006; Gubler and Selway 2012). Additionally, transnational kinship ties can
be a key source of material support or safe hiding, and ultimately facilitate ethnic
conflict (Salehyan, Gleditsch, and Cunningham 2011; Cederman, Girardin, and
Gleditsch 2009).
Scholars of civil resistance have paid comparatively less attention to how ethnicity
might influence the emergence of nonviolent strategies of resistance. Explanations of
nonviolent campaign onset have focused instead on processes of modernization and
globalization (Butcher and Svensson 2016; Karakaya, Forthcoming), political
opportunity structures linked to electoral cycles or civil liberties (Chenoweth and
Ulfelder 2017), and regional diffusion dynamics (Beissinger 2007; Gleditsch and Rivera
2017). While some of these studies include measures of ethnicity in their cross-national
analyses, they rarely grant ethnicity variables explicit theoretical attention. Perhaps
most importantly, a limitation to all of these studies is their focus on states as the unit of
analysis, which overlooks the drastically different positions that different groups
occupy within a given state. As a result, these studies cannot account for sub-national
variation, such as we have seen in Indonesia, Myanmar, Nepal, or the Philippines,
where we see members of different ethnic groups embrace different strategies–violent
and nonviolent–within the same state.
Better understanding the role of ethnicity in nonviolent conflict requires turning away
from the state as the unit of analysis to examine the unique characteristics of ethnic
groups. Indeed, a similar turn took place in the civil war literature, away from what
Chandra and Wilkinson (2008) call measures of “ethnic structure” and toward measures
of “ethnic practice.” The turn yielded important findings. While state-level measures of
ethnic diversity were only infrequently and inconsistently correlated to civil war onset,
the Ethnic Power Relations project, by examining the political status of ethnic groups,
demonstrated that ethnic groups excluded from executive-level state power were
significantly more likely to launch insurgencies. This was especially true when excluded
groups were larger and had recently experienced a downgrade in their political status
(Cederman, Wimmer, and Min 2010).
By examining the dynamics of nonviolent resistance at the level of groups, this article
adopts a similar disaggregated analysis of the impact of ethnicity on nonviolent
campaigns. In principle, civil resistance has the potential to provide aggrieved ethnic
groups an alternative means through which to pursue their political goals. However,
whereas the civil war literature argues that specific opportunity structures create
conditions favorable to violent insurgency for excluded ethnic groups, these same
structures raise barriers when it comes to the adoption of civil resistance strategies.
Ethnic Cleavages and the Strategic Logic of Nonviolent Action
While large-scale protests often appear to be chaotic and spontaneous, the literatures on
civil resistance and social movements demonstrate that collective action is often the
result of coordinated efforts by social movement organizations (Morris 1984; Schock
2005; Cunningham, Dahl, and Frugé 2017). In fact, there is ample evidence that
organization leaders engage in processes of strategic calculation, anticipating the likely
consequences of alternative courses of action and ultimately choosing the strategy they
deem most likely to offer success. Barrell (1992), for example, recounts the African
National Congress’s “strategic review” of 1978 and 1979, in which it formally decided to
transition from a primarily military strategy to one of political organizing. This set the
foundation for the United Democratic Front’s nonviolent campaign in the 1980s.
MacLeod (2015, 71) describes how the West Papuan self-determination movement
conducted “a cost-benefit analysis of the relative effectiveness of different methods.” My
own research has examined how Nepal’s Nepali Congress and Maoist parties each held
secret meetings in which they deliberated the relative advantages of violent and
nonviolent approaches (Thurber 2016).
If movements engage in a rational or quasi-rational decision-making process, what
variables affect their assessments of the viability of a nonviolent strategy? Scholarship
on civil resistance consistently emphasizes three key “comparative advantages” of
nonviolent resistance campaigns that help them achieve higher rates of success, relative
to violent resistance: (1) nonviolent movements create higher costs of repression; (2)
they can win over defections from elites within the regime; and (3) they generate higher
rates of participation. Movement leaders likely understand the importance of these
mechanisms, and are likely to ask themselves whether they can mobilize a large
number of individuals for active participation; whether they can recruit powerful actors
to their cause; and whether they will be repressed for mobilizing.
In the case of ethnic groups, the dynamics of ethnic politics are likely to influence how
movements evaluate the answers to these questions, and consequently their likelihoods
of attempting civil resistance campaigns.
One of the greatest obstacles nonviolent movements face is the prospect of brutal
repression by the state (Goodwin 2001; Davenport 2015; Chenoweth, Perkoski, and
Kang 2017). Indeed, by openly challenging a regime while foregoing weapons,
nonviolent activists appear to leave themselves largely defenseless against state
repression. However, ordering security forces to engage in the repression (especially
lethal repression) of unarmed protesters is a very difficult task. Professional codes of
conduct, fears of harming friends, family, or neighbors, and basic morality can make
soldiers and police reluctant to use force. As Sharon Nepstad (2013, 344) describes in
her account of military defectors in Syria: "…they are faced with a moral dilemma:
loyalty to the regime requires them to repress their own people. Repeatedly, in
interviews, many defectors have stated that they simply could not do this." They may
opt instead to shirk orders or even abandon their posts. Even when security forces do
engage in repression, such actions often “backfire” and create moral outrage against the
regime that has used force on unarmed protesters (Martin 2007). The outcome is often
greater popular support for the movement, and defections away from the state by key
regime elites.
Excluded ethnic groups, however, might be dubious of the backfire mechanism
applying to them. Numerous psychological studies show that individuals are more
willing to inflict physical harm on those they perceive as members of a different social
group. For example, in numerous follow-ups to Milgram’s (1974) infamous shock
experiments, scholars have consistently found that human subjects are more reticent to
harm another individual when that individual is observably different in terms of race,
linguistic dialect, or accent (Brant 1980). In the context of protests, Davenport,
Armstrong, and Soule (2011) find that white police officers in the United States were
more likely to make arrests during demonstrations in which a majority of participants
were black, relative to protests in which the participants were predominantly white.
And Johnson (2017) finds that when security forces are more “socially distant” from
protesters, they are more likely to use repressive force. In another example from my
own research, prior experiences with police brutality caused leaders of Nepal’s Magar
ethnic community to believe that attempts at civil resistance would be met with severe
repression (Thurber 2016).
Politically excluded ethnic groups will also be more skeptical that they can activate
another key mechanism of nonviolent success – elite defections. A lack of co-ethnics in
state institutions makes winning over loyalty-shifts difficult. Elites within the regime
may see the political agenda of another ethnic group as counter to their own interests,
and a threat to a favorable status quo. Furthermore, ethnic cleavages between
movement members and regime elites reduce the likelihood of direct interpersonal
relationships across sides, which civil resistance scholarship cites as key to convincing
elites to commit to the risky act of defection (Binnendijk and Marovic 2009; Jaafar and
Stephan 2010).
Finally, mass participation is the sine qua non of campaign effectiveness for nonviolent
movements (Chenoweth and Stephan 2011). The logic behind this is straightforward:
more people engaging in nonviolent anti-regime mobilization – from street protests to
strikes to boycotts – increases the effectiveness of those tactics by raising the costs
imposed on the regime and heightening the threat to regime power. Simply put, the
more people involved in carrying out acts of nonviolent resistance, the greater the
disruption those tactics cause. In fact, as DeNardo (1985, 35) points out, “it is nearly
impossible to imagine political circumstances where the disruptiveness of dissident
activity would diminish as its scope increased.”
There may be some advantages to overcoming barriers to collective action within an
ethnic group. Gubler and Selway (2012) theorize that common nationalist purpose, a
greater ability to exert social pressure and control, and enhanced communication help
ethnic-based movements mobilize in the context of violent insurgencies. While these
factors may increase the rate of participation among members of the ethnic group, the
overall size of the ethnic group likely places a ceiling on the overall mobilization
potential of the movement. This is especially likely to be the case if the grievances
around which people have mobilized are group-specific. In short, though maximizing
rates of participation in high-risk activity may be very important for armed rebels, the
need for high absolute numbers is far greater for ethnic groups organizing civil
resistance campaigns. Excluded ethnic groups, especially small ones, are likely to
perceive of the possibility of filling city squares with mass crowds as nearly impossible.
On the other hand, an excluded ethnic group that is sufficiently large may be more
likely to achieve sufficient levels of mass mobilization. In fact, this may help explain
some prominent cases of ethnic-based nonviolent resistance, such as the anti-Apartheid
struggle in South Africa. These cases are nevertheless rare, as demographic size and
political power often go hand in hand. Furthermore, cases such as the first Palestinian
intifada and the initial nonviolent resistance in Syria show that large excluded ethnic
groups still face formidable challenges in terms of repression and eliciting elite
defection. In sum, excluded ethnic groups face particularly high barriers when it comes
to mobilizing large crowds, wining over defections from regime elites, and thwarting
state repression. This may be particularly true when the excluded group is small. As a
result, these types of ethnic groups are less likely to get a civil resistance campaign off
the ground, and also less likely to attempt a civil resistance strategy in the first place.
Scholars of nonviolent action might object to the idea that ethnicity presents a barrier to
civil resistance by noting that an adherence to nonviolent tactics presents an
opportunity for groups to build support from across social divides. For example,
Chenoweth and Stephan (2011) specifically argue that a strategic advantage of civil
resistance is that it attracts participants from a broader and more diverse array of social
groups. It is likely true that nonviolent tactics more easily facilitate the crossing of social
divides than violent ones do. However, eschewing violence may be insufficient when it
comes to bridging these cleavages. Ethnic-based grievances that fuel a desire for
political change may not be shared by other groups, while leaders’ ability to inspire (or
coerce) collective action is unlikely to travel beyond members of the ethnic group.
Furthermore, McLauchlin (2017) demonstrates that even when excluded groups attempt
to transcend ethnic cleavages by framing goals in non-ethnic terms, fears about loss of
status and even reprisal among privileged groups will make them unlikely to join the
campaign. In the case of Syria, concerted efforts by the predominantly Sunni resistance
movement failed to produce substantial Allawi participation, especially as the Assad
regime strategically stoked sectarian violence in an effort to exacerbate this “loyalty
trap.”
Civil resistance campaigns may in fact be more diverse, in terms of number of groups
participating, but due to the constraints described above, I predict that this is most
likely to take the form of excluded ethnic groups joining along in campaigns led by
politically privileged ethnic groups where ethnic political claims are not present. Civil
resistance campaigns initiated by excluded ethnic groups will be rare, as will be
privileged groups participating in campaigns where ethnic claims are made.
In sum, I build on the horizontal inequalities literature that predicts that ethnic groups
facing political exclusion will be more likely to mobilize towards contentious action.
However, opportunity structures that are ill-suited to the strategic logic of civil
resistance will reduce the likelihood that excluded ethnic groups turn to nonviolent
contention. My logic leads to the following set of hypotheses:
H1: Politically excluded ethnic groups will be less likely to initiate civil resistance
campaigns relative to groups that enjoy high levels of access to state power.
H2: An ethnic group ‘s relative demographic size will be positively correlated with its
probability of initiating a civil resistance campaign.
Introducing the Ethnic Groups in Contention (EGC) Dataset
To better understand the relationship between ethnicity and insurgency, civil war
researchers have collected data on patterns of ethnic recruitment and claims-making by
armed conflict actors. Examples include the Ethnic Armed Conflict dataset (Wimmer,
Cederman, and Min 2009) and the ACD2EPR dataset (Wucherpfennig et al. 2012). Until
now, however, no equivalent data existed on ethnicity across nonviolent conflicts. The
Nonviolent and Violent Campaign Outcomes 2.0 (NAVCO 2.0) dataset (Chenoweth and
Lewis 2013) includes binary measures of movement “diversity” along multiple social
dimensions, including ethnicity. Movements are coded as ethnically diverse if they
include participants from more than one ethnic group, but no information exists on
which groups participated. It is therefore difficult to discern whether or not a conflict
included in NAVCO 2.0 is characterized by ethnic cleavages. For example, a movement
comprised only of members of an excluded minority, such as the Tibetan uprising of
1987, and one occurring within an ethnically homogenous society, such as the “June
Struggle” in South Korea the same year, are both labeled as not ethnically diverse.
Conversely, a movement comprising multiple excluded groups, such as the First
Defiance Campaign in South Africa (1952), and one that includes the participation of
multiple groups with access to political power, such as the Philippines People Power
campaign (1983), are both be coded as diverse despite having very different ethnic
dynamics. New data is thus required to better understand the role of ethnicity in
nonviolent campaigns, and to compare its role across different technologies of contention.
This article introduces data that fills this gap, under the name “Ethnic Groups in
Contention” dataset. The data consists of an effort to link ethnic groups to conflicts in a
manner that is conceptually and operationally similar to the ACD2EPR, providing data on
nonviolent as well as violent campaigns.
I use the NAVCO 2.0 dataset as the source of nonviolent and violent campaigns.
NAVCO 2.0 collects campaign-year information on campaigns that target the state with
maximalist political goals (defined as regime-change, secession, or anti-occupation). In
order to be included in the dataset, nonviolent campaigns must achieve a baseline
threshold of 1,000 observably mobilized participants, and last for at least one week.
Violent campaigns in NAVCO come from a variety of sources, including Gleditsch’s
(2004) updates to the Correlates of War (COW) database.1 I collected each unique
campaign onset from NAVCO, defined by the name and political goal of the opposition
movement. This results in 251 unique campaigns.
Next, I collected data on the ethnicity of participants involved at the outset of
campaigns. This information comes from multiple sources, including the NAVCO 2.0
and ACD2EPR datasets, case narratives from the online Appendix to Chenoweth and
Stephan (2011), the Swarthmore Global Nonviolent Action Database, and numerous
secondary sources. Using this information, I attribute one or more ethnic groups to each
NAVCO campaign. Ethnic groups come from the set of groups identified by the Ethnic
Power Relations (EPR) dataset as “politically relevant,” meaning that at least one
organization claims to represent the group in national politics, or the group experiences
discrimination by the state (Cederman, Wimmer, and Min 2010, 99). Similar to datasets
that provide information on the role of ethnicity in armed conflicts, I do not apply a
fixed participation threshold to link an ethnic group to a campaign.
1 For a full description of the inclusion criteria for the violent and nonviolent campaigns included in
NAVCO 2.0, see Chenoweth and Lewis (2013).
According to Wucherpfennig et al. (2012, 95): “it is difficult to impossible to obtain
reliable numbers of recruitment, especially for conflicts that are long past and not well
documented.” Instead, I code a movement as including members of an ethnic group
when there is sufficient evidence in the secondary literature of significant involvement
by members of an ethnic group.
In addition to coding all politically relevant ethnic groups that participate in the onset of
each campaign, I code exactly one ethnic group as the initiator of the campaign. This
refers to the ethnic group that constitutes a plurality among the group of original
campaign participants. Such a distinction is important for testing my hypotheses, as a
group’s calculus in deciding whether to launch a campaign is different from its
consideration of whether or not to join efforts led primarily by another group. For
example, while the 1988 pro-democracy movement in Myanmar included participants
from many of the country’s ethnic groups, I code it as being initiated by the Bamar.
Finally, I also record whether or not the campaign makes an explicit ethnic claim as a
political goal. Unlike most studies of ethnic civil war, however, I analyze campaigns in
which ethnic claims are present as well as those in which they are absent. The theory
presented above predicts that ethnic cleavages may present an obstacle to nonviolent
action, even if explicit ethnic claims are not made. My approach allows for an
examination of the dynamics of ethnic participation even when ethnic grievances are
not an explicitly stated goal of the campaign. A complete codebook is available in the
appendix to this article.
Table 1 presents selected summary statistics of the ethnic composition of nonviolent
and violent campaigns in the NAVCO 2.0 dataset based on the new Ethnic Groups in
Contention data. Consistent with arguments that nonviolent strategies attract more
diverse participants, I find that 47 percent of nonviolent campaigns include more than
one ethnic group. In contrast, only 25 percent of primarily violent campaigns include
multiple ethnic groups.2 Two additional measures from the new group-specific
ethnicity data further corroborate this pattern. The average number of groups
participating in a movement is higher for nonviolent campaigns (2.17) than for violent
ones (1.44). Additionally, the average cumulative size of groups as a share of the
country’s population is higher for nonviolent campaigns (.71) than it is for violent
insurgencies (.41).
However, an analysis of the size and power structure of ethnic groups participating in
resistance campaigns reveals a more nuanced picture. While nonviolent campaigns are
on average more ethnically diverse in terms of number of participating groups, these
movements primarily draw participants from privileged ethnic groups. Indeed, 69
percent of nonviolent campaigns included an ethnic group ranked by EPR as a Senior
Partner or higher (Senior Partner, Dominant, or Monopoly). Examples include many of
the most prominent and frequently studied civil resistance campaigns, such as the
People Power Movement in the Philippines, the Iranian Revolution, the anti-Pinochet
campaign in Chile, and the Color Revolutions of the early 2000s. Only 50 percent of
nonviolent campaigns included a group coded as “excluded” from power
(Discriminated, Powerless, or Self-Exclusion). Conversely, violent campaigns are more
likely to include excluded groups, with 61 percent of the total experiencing the
participation of at least one excluded group, and 34 percent including the participation
of a high status group. Junior Partners, which the EPR data describes as enjoying some
inclusion in the executive governance of the state but occupying a subordinate position,
participate in both nonviolent and violent campaigns at relatively low rates, 17 percent
and 11 percent respectively. Hezbollah (Shia) in Lebanon and the campaign for
Slovakian independence from Czechoslovakia are examples of this relatively rare
occurrence.
The difference in ethnic participation across nonviolent and violent conflicts is even
starker when examining the characteristics of groups coded as the “initiator” of a
campaign (i.e. those that constituted a plurality of the original set of participants). High-
status ethnic groups initiated 65 percent of nonviolent campaigns, while excluded
groups initiated only 21 percent. Examples of these cases include campaigns such as the
2 A similar analysis could be conducted using the ethnic diversity measure in the NAVCO 2.0 dataset
alone. That measure codes a greater number of both nonviolent and violent campaigns as “diverse,” 64
percent and 41 percent respectively, but a similar pattern of greater diversity in primarily nonviolent
campaigns. The difference is likely a result of NAVCO 2.0 counting the participation of groups not
included in the EPR dataset, either because they are deemed not politically relevant or are considered to
be part of a single “parent” ethnic group.
struggle against Apartheid in South Africa, the IRA in Northern Ireland (in 1968), and
the first Intifada in Palestine.
In contrast, excluded groups initiated over half (52 percent) of violent campaigns, while
high status groups initiated less than a third (31 percent) of violent insurgencies.
Groups initiating nonviolent campaigns also tend to be demographic majorities,
constituting on average 59 percent of the country’s population. Groups that initiate
violent campaigns, on the other hand, tend to be minorities, representing only an
average of 39 percent of the total population. The smallest group to initiate a violent
campaign is the Mizo of India, comprising only 0.1 percent of the country’s population.
The smallest group to initiate a nonviolent campaign is the Tibetans, accounting for 0.4
percent of China’s population. Finally, while 57 percent of violent campaigns involve an
ethnic claim, only 20 percent of nonviolent campaigns do so.
The last column in Table 1 presents a comparative analysis of the ethnic composition of
violent campaigns included in the ACD dataset based on the ACD2EPR data. The main
differences between the campaigns included in the Ethnic Groups in Contention and
ACD2EPR datasets are that the ACD includes only violent campaigns, uses a 25 battle-
related death criterion, and disaggregates some conflicts into dyads that otherwise
represent a single campaign in NAVCO 2.0. The ACD2EPR data therefore include a
much greater number of campaigns, but ethnic participation patterns largely mirror
those observed in the Ethnic Groups in Contention data. In fact, multiple groups and
privileged groups participate in violent campaigns at lower rates according to the
ACD2EPR than they do in the Ethnic Groups in Contention data.
In sum, these summary statistics show that nonviolent and violent campaigns look very
different in terms of ethnic make-up. While nonviolent movements are more likely to
attract the participation of multiple ethnic groups, they tend to be initiated by the
largest, most politically privileged groups. Unsurprisingly, ethnic claims are rarely
expressed in nonviolent campaigns. Violent campaigns are more likely to be initiated by
smaller and politically excluded groups. And more often than not, they feature ethnic
claims.
Testing the Relationship Between Movement Ethnicity and
Conflict Strategy
While the campaign-level data presented above reveal important differences in the
ethnic compositions of violent and nonviolent campaigns, they cannot fully test the
relationship between ethnic group characteristics and the likelihood of initiating
nonviolent and violent campaigns. This requires looking at the full set of ethnic groups,
including cases that do not mobilize into campaigns with maximalist political goals.
I begin by using EPR’s group-year data. This consists of over 30,000 group-year
observations from 1946 to 2006 (truncated by the end of NAVCO 2.0’s coverage in
2006). The Ethnic Groups in Contention data becomes the basis for coding a series of
dependent variables indicating whether a group initiated or participated in the onset of
a nonviolent or violent campaign in a given year.
Table 2 presents the rates at which various categories of ethnic groups initiate violent
and nonviolent campaigns. Groups with the greatest access to political power (Senior
Partners, Dominant, and Monopoly statuses) have initiated nonviolent campaigns at the
highest rate, in 0.91 percent of group-years. Junior Partners and Excluded groups
initiate nonviolent campaigns at far lower rates, in 0.07 percent and 0.10 percent of
group-years respectively.3
Next, I perform a series of regressions to further test the relationship between ethnicity
and nonviolent and violent conflict. I use the initiation variable as the dependent
variable, which takes a positive value if the ethnic group initiated (i.e. made up a
plurality of the initial participants) a campaign in that year. Cases in which a group
participated either in an onset or in an ongoing campaign initiated by a different group
are dropped.
As in the descriptive statistics above, I collapse the EPR ethnic status variable into three
discrete levels to test my theoretical framework: Excluded Groups (Self Exclusion,
Powerless, and Discriminated), Junior Partners, and Senior Partners and Above. The
last level serves as the reference category. I keep Junior Partners separate as several
studies have shown that these groups behave differently than either Excluded or more
privileged groups when it comes to participating in civil wars or coup attempts
(Cederman, Wimmer, and Min 2010; Roessler 2011; Johnson and Thurber,
Forthcoming). That said, my theory remains agnostic as to how this particular category
of ethnic group might behave in civil resistance campaigns. My prediction is simply
that Excluded groups will be less likely to initiate a nonviolent campaign, relative to
Senior Partners and Above. I also use EPR’s estimates of ethnic group size to test the
hypothesis that smaller groups will be less likely to initiate nonviolent campaigns.
3 A couple notable cases of Self-Exclusion groups, the West Papuan and East Timorese resistance
movements, later engage in prolonged nonviolent action for self-determination, but are coded by
NAVCO 2.0 as violent in their first year. For a detailed analysis of the use of nonviolence in self-
determination disputes — a category that largely overlaps with EPR’s concept of Self-Exclusion — see
Cunningham, Dahl, and Frugé (2017).
I control for a number of factors that might be correlated with movement strategy and
the social structure of a state. First, I use a logged measure of a country’s population,
lagged one year, to account for positive correlations between the onset of violent and
nonviolent movements and population size (Chenoweth and Lewis 2013; Butcher and
Svensson 2016). Next, I include a measure for per capita Gross Domestic Product (GDP)
to control for state capacity in terms of overall economic productivity (Fearon and
Laitin 2003). GDP is also likely correlated to modernization, which Butcher and
Svensson (2016) suggest produces stronger social networks and environments more
conducive to nonviolent mobilization. Per capita GDP values are lagged one year and
logged.4 I also include a count variable for the number of times an ethnic group has
participated in a nonviolent and violent campaign in the past. This accounts for path
dependencies in group repertoires of contention (Tilly 2008).
State-centric approaches to the study of conflict would suggest that highly repressive
regimes push aggrieved groups away from nonviolent forms of contention (Goodwin
2001) and may also exclude ethnic minorities (Roessler 2011). To account for regime
attributes, I include two indices from the V-DEM project (Coppedge et al. 2017): the
Polyarchy measure of degree of electoral democracy, and the Physical Integrity Index as
a measure of the regime’s prior use of repression against civilians. Each of these is
lagged one year.
All models include cubic polynomials (Carter and Signorino 2010) for the number of
years since the end of the last violent campaign, and the number of years since the end
of the last nonviolent campaign in each country. This controls for the path dependence
of contentious repertoires, enduring campaign infrastructures, and other potentially
time-dependent factors. I do not report coefficients for these variables in the results
tables, but they are available in the online appendix.
I use multinomial logit models to assess the relationship between these covariates and
the onset of violent and nonviolent campaigns. While my theoretical framework
pertains primarily to the onset of nonviolent campaigns, I follow Cunningham (2013)
and Butcher and Svensson (2016) in employing a multinomial model. This helps
account for the possibility that that the decision to choose one strategy over another
likely influences future use of the non-chosen strategy, often to the point of exclusion.
Furthermore, the multinomial model allows me to evaluate the impact of different
variables simultaneously on both violent and nonviolent outcomes. This provides a
more complete understanding of how various factors affect the broader landscape of
contentious politics. It is, however, possible that members of the same ethnic group are
involved in both types of campaigns within the same year. This occurs only once in the
data, in 1958 in Venezuela. I drop this case in the multinomial models as it makes little
sense to include a fourth outcome variable for a single case. However, I include results
for ordinary logit models (in which violent and nonviolent outcomes are modeled
4 Population and GDP data come from the 2013 update to Gleditsch’s (2002) “Expanded Trade and GDP
Data.” The original source for the GDP data is Feenstra, Inlkar, and Timmer’s (2015) “Penn World Tables
v. 8.0.”
independently) in the online appendix to show that dropping this case does not
significantly alter the results.
Results
The analysis shows a strong negative relationship between political exclusion and the
likelihood of a group initiating nonviolent campaigns. Figure 1 presents simulated risk
ratios from the base model, and Table 3 includes coefficients and standard errors. In
terms of odds, excluded groups are a little less than one third as likely to initiate a
nonviolent campaign, compared to groups with Senior Partner or higher status. By
contrast, the relationship between excluded status and onset of a violent campaign is
positive, though this result falls just short of statistical significance.5
Group size is positively correlated with both nonviolent and violent campaign
initiation. Figure 1 shows that when group size increases from the mean to one
standard deviation above the mean (from 18 percent to 44 percent of the population),
the likelihood of a group initiating a nonviolent campaign nearly doubles.
5 This is a slight contrast to the results presented in Cederman, Wimmer, and Min (2010), which shows a
stronger relationship between exclusion and violent conflict. This discrepancy is likely a result of three
differences between the two studies: 1) their use of the 25 battle-related death threshold, which produces
a greater number of conflicts, particularly those launched by excluded ethnic groups; 2) their restricted
examination of those conflicts where ethnic claims are made; 3) their inclusion of Junior Partners in the
reference category, alongside higher status groups. As Model 3 in Table 3 shows, including Junior
Partners in the reference category on its own yields a statistically significant positive relationship between
exclusion and violent onsets in this study.
Figure 1: Risk ratios of initiating a nonviolent or violent campaign as covariates are increased from 0 to 1
(EPR Status) or from their mean to one standard deviation above the mean. Risk ratios for nonviolent
onsets in green, violent onsets in red.
Figure 2 illustrates the results of Model 1 through a series of expected probability plots.
The plots in the top row depict a strong positive correlation between group size and the
onset of nonviolent campaigns, ranging from expected probabilities near zero for tiny
Excluded groups to as high as 0.031 for Senior Partners and Above that make up nearly
all of a state’s population. Importantly, group status has a considerable effect on the
initiation of nonviolent campaigns. For even the largest Excluded groups, as shown in
the top-right plot, the expected probability of initiating a nonviolent campaign is less
than .01. Conversely, the top-left plot shows that Senior Partners and Above need only
make up 52 percent of the population to have the same probability of initiating a
nonviolent campaign. When comparing the likelihood of initiating nonviolent versus
violent campaigns among groups with similar status, the pair of plots in the leftmost
column of Figure 2 show that for Senior Partners and Above, the predicted probability
of initiating a nonviolent campaign is over twice that of initiating a violent campaign
irrespective of group size. Among Excluded groups, the two plots in the rightmost
column show the reverse relationship: the predicted probability of violence is over
twice that of nonviolence across the range of group sizes. Interestingly, Junior Partners
are unlikely to initiate either nonviolent or violent campaigns. Such groups may fear
that initiating either type of campaign would jeopardize their already tenuous status.
Figure 2: Expected probability of initiating a nonviolent or violent campaign by ethnic group size and
status.
Figure 1 and Table 3 show that the control variables behave largely as expected.
Country population is positively correlated with nonviolent onsets, while GDP is
negatively correlated with violent onsets. Counts of prior group participation in violent
and nonviolent conflicts show no significant correlation to the likelihood of new violent
or nonviolent onsets. The V-Dem Electoral Democracy Index is positively correlated
with the initiation of violent campaigns, a result that is consistent with some findings
from the literature on terrorism and democracy (Chenoweth 2010). Conversely, the
electoral index is negatively correlated to the initiation of nonviolent campaigns,
consistent with recent work that examines the relationship between regime type,
contentious claims, and the onset of violent and nonviolent conflict (Cunningham et al.
2017). V-Dem’s Physical Integrity Index is negatively correlated with violent onsets,
indicating that prior repression decreases the likelihood of violence. In fact, moving
from the mean physical integrity score to one standard deviation above the mean
decreases the likelihood of a violent campaign by nearly one fifth. However, there is no
statistically significant relationship between physical integrity scores and nonviolent
onsets, consistent with previous findings in the civil resistance literature that prior state
repression does not have an effect on when and where we see nonviolent campaigns
emerge (Chenoweth and Stephan 2011; Chenoweth and Lewis 2013).
Table 3 includes two additional models to assess the robustness of the main findings.
Model 2 adds additional group-level covariates available in the EPR project (Vogt et al.
2015). A dichotomous variable measures whether transnational kin occupy a position of
power in a neighboring country. Groups with neighboring kin might be more likely to
be excluded out of fears of forming a “fifth column,” and may also have enhanced
abilities to mount an armed insurgency. Including this variable allows me to evaluate
whether excluded ethnic groups’ lower likelihoods of engaging in nonviolent resistance
results from greater opportunities for violent rebellion. I also include a measure of
whether the group has been downgraded in status in the past five years, as well as a
measure of economic horizontal inequality. These measures help capture conditions
that are likely to generate group grievances, but that unlike access to political power,
should not significantly impact a group’s opportunity to mobilize towards nonviolent
action. Including these variables tests the validity of relative deprivation arguments,
which contend that ethnic grievances incite psycho-social responses that lead groups to
violence. The results from Model 2 only report a statistically significant negative
correlation between downgraded status and the onset of violence.6 The significance and
direction of the study’s main independent variables remain unchanged in Model 2.
Model 3 collapses the categorical political inclusion variable into a single binary
distinction between “included” and “excluded” groups, eliminating Junior Partner as
its own level and rolling it into the reference category along with the Senior Partner,
Dominant, and Monopoly levels. This slightly weakens the relationship between
excluded status and nonviolence, though the coefficient for Excluded is still significant at
the p < 0.05 level. The positive relationship between exclusion and violence also
becomes statistically significant in this model, consistent with prior studies on ethnic
conflict that include a dichotomous measure of exclusion (Cederman, Wimmer, and
Min 2010).
6 While the lack of a statistically significant relationship for Horizontal Inequalities and Neighboring Kin
in Power differs from prior work on ethnic conflict, it is likely a consequence of the different threshold
standards of the NAVCO 2.0 and ACD datasets. The direction of the relationship exhibited in Model 2b is
consistent with prior studies.
Table 4 presents two additional models that include varied operationalizations of the
dependent variable. Model 4 includes only regime-change campaigns, dropping
NAVCO 2.0 secessionist and anti-colonial campaigns. A large number of insurgencies
launched by excluded ethnic groups have the objective of securing degrees of territorial
autonomy. Arguably, autonomy campaigns may rely on a different strategic logic for
which civil resistance may be less well-suited. As such, an examination of only those
campaigns that seek the overthrow of a central government provides stronger
assurance that any observed relationship between political exclusion and a nonviolent
strategy is not simply a function of the nature of the goal being pursued.
Finally, Model 5 includes all cases where an ethnic group participated in the onset of a
campaign, not simply those where ethnic groups were coded as “initiators.” The logic
elaborated in my theoretical framework applies primarily to the act of initiating a
campaign: small excluded ethnic groups may feel comfortable joining a nonviolent
campaign under the umbrella of larger, more privileged groups. However, the theory
would have little value if it were easy for such groups to form a coalition with larger,
higher status, partners. Furthermore, the definition of initiators as constituting a
plurality of first movers could be privileging larger groups in a way that is a function of
the operationalization and not the theory. The findings presented in Table 4 indicate
that the effects of ethnic group political status and size on civil resistance are both
robust to these alternative specifications.
Additional models including regional and decade fixed effects, simple logits, rare
events logits, probits, and models with robust standard errors clustered at the group
and state levels are presented in the online appendix. The results from these models
continue to provide support to my main hypotheses.
Discussion
The findings from both the comparative descriptive analysis of the ethnic composition
of campaigns as well as the group-level regressions offer substantial evidence that
ethnic structures – both access to political power and demographic size – affect the
propensities of groups to initiate campaigns of nonviolent resistance. Political exclusion
is consistently negatively correlated with nonviolent onsets in all of the quantitative
models: excluded ethnic groups are only one third as likely as groups with Senior
Partner or higher status to initiate a civil resistance campaign. This stands in contrast to
violent campaigns, where participation by excluded groups is the norm and where a
strong positive relationship between exclusion and violence has been established in the
literature (Cederman, Wimmer, and Min 2010). Though the relationship between ethnic
status and violent insurgency is somewhat weaker in this study compared to previous
ones, this is likely a result of the different thresholds for entry used by the NAVCO 2.0
and the ACD datasets, and of prior studies’ more narrow focus on conflicts where
groups make ethnic claims.
It is also interesting to compare the relative impact of group size on nonviolent and
violent campaign onsets. In four out of five models presented, the group size coefficient
for the outcome of a nonviolent campaign is larger than that for a violent campaign.
However, a test for difference between coefficients (Paternoster et al. 1998) shows that
only the group size coefficients in Model 3 are statistically distinguishable from each
other, and even then, only at the p < 0.1 level. Thus, while larger group size clearly
increases the likelihood of both nonviolent and violent campaigns, there is insufficient
evidence to conclude that it affects the likelihood of civil resistance campaigns to a
greater degree than it does violent campaigns.
The findings also show that alternative explanations are inadequate in explaining the
relationship between exclusion and the onset of nonviolent campaigns. The diversity
advantage of nonviolent resistance does not seem to apply as strongly for excluded
groups. In fact, excluded groups participated more frequently in violent, as opposed to
nonviolent, campaigns. Moreover, when excluded groups do participate in nonviolent
campaigns, they are rarely the initiating group (21 percent of cases) and also rarely
make ethnic claims (20 percent of cases).
Furthermore, the negative relationship between political exclusion and nonviolent
campaigns holds across different specifications of the dependent variable, for example
when it is expanded to include all participating ethnic groups, not just initiators.
Excluded groups are on average less likely to participate in nonviolent campaigns, and
when they do, it is often under the umbrella of more privileged groups espousing
political goals not specific to ethnicity.
The empirical analysis also finds substantial support for the role of state-level structures
in influencing the choices available to ethnic groups. Regime type and repressive
behavior appear to shape the opportunities available to movements, though there is no
observed relationship between a regime’s prior repressive behavior and the likelihood
of an ethnic group initiating a nonviolent campaign. This is consistent with findings
from the extant literature. Additionally, group-level measures of grievance and
opportunity performed more poorly in Model 2, with only downgraded status having a
statistically significant effect on the likelihood of ethnic groups initiating a violent
campaign.
More importantly, measures of political status and group size remain significantly
correlated to the initiation of nonviolent campaigns even when group-level controls are
included in the analysis. This increases confidence in the finding that the observed
relationships are in fact the result of constraints resulting from political exclusion and
small group size, rather than the product of statewide patterns of repression or
frustration/aggression urges sparked by group grievances.
Some limitations of this study are important to note. Because the NAVCO 2.0 data only
includes campaigns that reach a 1,000 participant threshold, we do not know whether
the relationships identified in this study hold at lower levels of contention. Indeed,
disaggregated data on smaller social disruptions might highlight different patterns. For
example, the current data do not allow us to disaggregate between movements that
reject civil resistance a priori, versus those that attempt civil resistance but fail to
achieve participation thresholds that include them in NAVCO 2.0. In all likelihood,
there are examples of both types of cases, and both pathways are consistent with the
logic put forth in this article: structural constraints prevent certain ethnic groups from
getting a civil resistance campaign off the ground.
Finally, it is possible that a psychological frustration-aggression response (Davies 1962;
Gurr 1970) to inter-group inequalities could also explain the reduced likelihood of
nonviolent contention from excluded groups. In this scenario, groups turn to arms as a
violent emotional response to injustice. Rational and psychological mechanisms could
work in tandem, and the theory presented in this article does not exclude the possibility
that rational deliberations could be colored by psychological biases. But the robustness
of the main quantitative findings even when group-level measures of grievance were
included in Model 2 combined with the numerous qualitative accounts of social
movements engaging in organized strategic deliberations described earlier provides
evidence in support of my opportunity-based theoretical framework.
Conclusion
This article has sought to apply insights from scholarship on ethnic conflicts to the
study of an alternative mode of political contention: nonviolent civil resistance. In so
doing, it has challenged a central tenet of civil resistance scholarship that agency
matters more than structure in shaping the course of nonviolent campaigns. The very
logic through which nonviolence has proven effective for so many campaigns around
the globe in fact poses particular challenges for ethnic groups that are politically
excluded, especially when those ethnic groups are small in size. Consequently, we see
relatively few civil resistance campaigns launched by such groups, especially when
compared to violent campaigns.
The argument and findings presented in this study are agnostic as to whether violence
or nonviolence is more effective in helping marginalized ethnic groups achieve their
goals. The historical track record for violent campaigns revealed in Chenoweth and
Stephan (2011) suggests that turning to arms is rarely an effective strategy. I argue
simply that ethnic groups and their movements face greater obstacles to nonviolent
mobilization relative to those that recruit from politically privileged ethnic groups.
They are thus relatively less likely to be able to (or to anticipate being able to) initiate a
civil resistance campaign.
Admittedly, the specific mechanisms through which excluded ethnic groups make
strategic decisions cannot be ascertained through the statistical methods used in this
study. In particular, the study has not answered the following questions: To what
degree are movements consciously and rationally deliberating strategic alternatives? To
what degree are they being influenced by psychological or emotional dynamics? Are
movements attempting tactics more or less at random, with some gaining traction and
others failing before they ever take off? Future qualitative research may be able to shed
greater light on these processes and pathways.
This study also provides potential new insights for the study of violent conflict.
Scholarship on civil war has generally considered aggrieved groups, especially ethnic
groups, as having but two choices: violent rebellion or accepting the political status quo.
As the empirical and scholarly records show, however, a third option exists. Research
on civil wars therefore needs to ask not just why and how certain groups seek to
overthrow a regime, but why they rule out nonviolent strategies as a possible means of
achieving political ends. My argument that the strategic logic of civil resistance poses
unique challenges to excluded ethnic groups, as opposed to groups bound by grievance
or belief, may help explain why civil wars play out along ethnic cleavages so much
more frequently than along any other type of politically salient division.
The findings also have important implications for policymakers and practitioners. They
may be of particular interest to those involved in promoting nonviolent methods of
activism and/or deterring the use of violence. This study highlights a particular type of
social movement that may be more skeptical of nonviolent action, and thus also more
likely embrace the alternative of violent insurgency. Rather than simply asserting that
nonviolence is effective in all conditions, policymakers and practitioners may wish to
directly confront the particular challenges excluded ethnic groups face when attempting
to utilize this strategy.
Acknowledgements
I would like to thank Nils Borman, Zeynep Bulutgil, Erica Chenoweth, Kelly Greenhill,
Benjamin Lessing, Robert Pape, Evan Perkoski, Jonathan Pinckney, Ivan Rasmussen,
Dan Slater, Paul Staniland, workshop participants at the University of Chicago and
Harvard Kennedy School’s Belfer Center for Science and International Affairs, three
anonymous reviewers, and the editorial team at JoGSS for valuable feedback on
previous versions of this article.
Sources Cited
Ackerman, Peter, and Christopher Kruegler. 1994. Strategic Nonviolent Conflict: The Dy-
namics of People Power in the Twentieth Century. Westport, CT: Praeger Publishers.
Arreguin-Toft, Ivan. 2001. “How the Weak Win Wars: A Theory of Asymmetric
Conflict.” International Security 26 (1): 93–128.
Barrell, Howard. 1992. “The Turn to the Masses: The African National Congress’
Strategic Review of 1978-79.” Journal of Southern African Studies 18 (1): 64–92.
Beissinger, Mark R. 2007. “Structure and Example in Modular Political Phenomena: The
Diffusion of Bulldozer/Rose/Orange/Tulip Revolutions.” Perspectives on Politics 5 (2):
259–276.
Binnendijk, Annika Locke, and Ivan Marovic. 2009. “Power and Persuasion: Nonviolent
Strategies to Influence State Security Forces in Serbia (2000) and Ukraine (2004).” Com-
munist and Post-Communist Studies 39 (3): 411–429.
Brant, William D. 1980. “The Effects of Race and Social Distance on Obedience.” Journal
of Social Psychology 112 (Dec.): 229–235.
Buhaug, Halvard. 2006. “Relative Capability and Rebel Objective in Civil War.” Journal
of Peace Research 43(6): 691–708.
Butcher, Charles, and Isak Svensson. 2016. “Manufacturing Dissent: Modernization and
the Onset of Major Nonviolent Resistance Campaigns.” Journal of Conflict Resolution 60
(2): 311–339.
Byman, Daniel. 2008. “Understanding Proto-Insurgencies.” Journal of Strategic Studies 31,
(2): 165–200.
Carter, David B, and Curtis S Signorino. 2010. “Back to the Future: Modeling Time De-
pendence in Binary Data.” Political Analysis 18 (3): 271–292.
Cederman, Lars-Erik, Nils B. Weidmann, and Kristian Skrede Gleditsch. 2011.
“Horizontal Inequalities and Ethnonationalist Civil War: A Global Comparison.”
American Political Science Review 105 (3): 478-495.
Cederman, Lars-Erik, Andreas Wimmer, and Brian Min. 2010. “Why Do Ethnic Groups
Rebel?: New Data and Analysis.” World Politics 62 (1): 87–119.
Cederman, Lars-Erik., Luc Girardin, and Kristian Skrede. Gleditsch. 2009.
“Ethnonationalist Triads: Assessing the Influence of Kin Groups on Civil Wars.” World
Politics 61 (3): 403–437.
Chandra, Kanchan, and Steven Wilkinson. 2008. “Measuring the Effect of "Ethnicity".”
Comparative Political Studies 41, (4/5): 515–563.
Chenoweth, Erica. 2010. “Democratic Competition and Terrorist Activity.” The Journal of
Politics 72 (1): 16–30.
Chenoweth, Erica, and Orion A. Lewis. 2013. “Unpacking Nonviolent Campaigns:
Intro- ducing the NAVCO 2.0 Dataset.” Journal of Peace Research 50, (3): 415–423.
Chenoweth, Erica, Evan Perkoski, and Sooyeon Kang. 2017. “State Repression and
Nonviolent Resistance.” Journal of Conflict Resolution 61 (9): 1950–1969.
Chenoweth, Erica, and Maria J. Stephan. 2011. Why Civil Resistance Works: The Strategic
Logic of Nonviolent Conflict. New York: Columbia University Press.
Chenoweth, Erica, and Jay Ulfelder. 2017. “Can Structural Conditions Explain the Onset
of Nonviolent Uprisings?” Journal of Conflict Resolution 61 (2): 298–324.
Coppedge, Michael, et al. 2017. V- Dem Country-Year Dataset v7.1. Varieties of
Democracy (V-Dem) Project.
Cunningham, David E., et al. 2017. “Words and Deeds: From Incompatibilities to
Outcomes in Anti-Government Disputes.” Journal of Peace Research 54, no. 4 (July): 468–
483.
Cunningham, Kathleen Gallagher. 2013. “Understanding Strategic Choice.” Journal of
Peace Research 50 (3): 291–304.
Cunningham, Kathleen Gallagher, Marianne Dahl, and Anne Frugé. 2017. “Strategies of
Resistance: Diversification and Diffusion.” American Journal of Political Science 61: 591-
605.
Davenport, Christian. 2015. How Social Movements Die : Repression and Demobilization of
the Republic of New Africa. New York: Cambridge University Press.
Davenport, Christian, David A. Armstrong II, and Sarah A. Soule. 2011. “Protesting
While Black? The Differential Policing of American Activism, 1960 to 1990.” American
Sociological Review 76 (1): 152–178.
Davies, James C. 1962. “Toward a Theory of Revolution.” American Sociological
Review27, no. 1 (1962): 5-19.
DeNardo, James. 1985. Power in Numbers: The Political Strategy of Protest and Rebellion.
Princeton: Princeton University Press.
Denny, Elaine K, and Barbara F Walter. 2014. “Ethnicity and Civil War.” Journal of Peace
Research 51 (2): 199–212.
Fearon, James D., and David D. Laitin. 2003. “Ethnicity, Insurgency, and Civil War.”
American Political Science Review 97 (1): 75-90.
— .1996.“Explaining Interethnic Cooperation.”American Political Science Review 90 (4):
715–735.
Feenstra, Robert C., Robert Inklaar and Marcel P. Timmer (2015), "The Next Generation
of the Penn World Table" American Economic Review, 105(10), 3150-3182.
Gates, Scott. 2002. “Recruitment and Allegiance: The Microfoundations of Rebellion.”
Journal of Conflict Resolution 46 (1): 111–130.
Gleditsch, Kristian S. 2002. "Expanded Trade and GDP Data,"Journal of Conflict
Resolution46: 712-24.
Gleditsch, Kristian S. 2004. “A Revised List of Wars Between and Within Independent
States, 1816-2002. International Organization 30(3): 231-262.
Gleditsch, Kristian S., and Mauricio Rivera. 2017. “The Diffusion of Nonviolent
Campaigns.” Journal of Conflict Resolution 61 (5): 1120–1145.
Goodwin, Jeff. 2001. No Other Way Out: States and Revolutionary Movements, 1945-1991.
New York: Cambridge University Press.
Gubler, Joshua R., and Joel Sawat Selway. 2012. “Horizontal Inequality, Crosscutting
Cleavages, and Civil War.” Journal of Conflict Resolution 56 (2): 206–232.
Gurr, Ted Robert. 1994. “People Against States: Ethnopolitical Conflict and the
Changing World System.” International Studies Quarterly 38 (3) :347–377.
— .1970.Why Men Rebel. Princeton: Princeton University Press.
Harff, Barbara, and Ted Robert Gurr. 1994. Ethnic Conflict in World Politics. Boulder:
Westview Press.
Horowitz, Donald L. 1985. Ethnic Groups in Conflict. Berkeley: University of California
Press.
Humphreys, Macartan, and Jeremy M. Weinstein. 2006. “Handling and Manhandling
Civilians in Civil War.” American Political Science Review 100 (3): 429-447.
Jaafar, Rudy, and Maria J. Stephan. 2010. “Lebanon’s Independence Intifada: How Un-
armed Insurrection Expelled Syrian Forces.” In Civilian Jihad: Nonviolent Struggle, De-
mocratization, and Governance in the Middle East, ed. by Maria J. Stephan, 169–84. New
York: Palgrave Macmillan.
Johnson, Paul Lorenzo. 2017. “Virtuous Shirking: Social Identity, Military Recruitment,
and Unwillingness to Repress.” Doctoral Dissertation, University of California, Davis.
Johnson, Paul Lorenzo, and Ches Thurber. Forthcoming. “The Security-Force Ethnicity
(SFE) Project: Introducing a new dataset.” Conflict Management and Peace Science.
Karakaya, Suveyda. Forthcoming. “Globalization and Contentious Politics: A
Comparative Analysis of Nonviolent and Violent Campaigns.” Conflict Management and
Peace Science.
Kasfir, Nelson. 1979. “Explaining Ethnic Political Participation.” World Politics 31 (3):
365–388.
Lacina, Bethany & Nils Petter Gleditsch. 2005. “Monitoring Trends in Global Combat: A
New Dataset of Battle Deaths. European Journal of Population 21(2–3): 145–166.
Laitin, David D. 1998. Identity in Formation: the Russian Speaking Populations in the Near
Abroad. Ithaca, NY: Cornell University Press.
MacLeod, Jason. 2015. “From the Mountains and Jungles to the Villages and Streets:
Transitions from Violent to Nonviolent Resistance in West Papua.” Chap. 3 in Civil
Resistance and Conflict Transformation: Transitions from Armed to Nonviolent Struggle, ed.
by Véronique Dudouet. New York: Routledge.
Martin, Brian. 2007. Justice Ignited: the Dynamics of Backfire. New York: Rowman &
Littlefield.
McLauchlin, Theodore. Forthcoming. “The Loyalty Trap: Regime Ethnic Exclusion,
Commitment Problems, and Civil War Duration in Syria and Beyond.” Security Studies.
Milgram, Stanley. 1974. Obedience to Authority. New York: Harper & Row.
Morris, Aldon D. 1984. The Origins of the Civil Rights Movement : Black Communities
Organizing for Change. New York: The Free Press.
Mueller, John. 2000. “The Banality of ‘Ethnic War.’” International Security 25 (1): 42–70.
Nepstad, Sharon Erickson. 2013. “Mutiny and Nonviolence in the Arab Spring:
Exploring Military Defections and Loyalty in Egypt, Bahrain, and Syria.” Journal of Peace
Research 50 (3): 337–349.
— . 2011. Nonviolent Revolutions: Civil Resistance in the Late 20th Century. New York:
Oxford University Press.
Paternoster, Raymond, et al. 1998. “Using the Correct Statistical Test for the Equality of
Regression Coefficients.” Criminology 36 (4): 859–866.
Posen, Barry. 1993. “The Security Dilemma and Ethnic Conflict.” Survival 35 (1): 27–47.
Posner, Daniel N. 2005. Institutions and Ethnic Politics in Africa. 337. New York: Cam-
bridge University Press.
Roessler, Philip G. 2011. “The Enemy Within: Personal rule, Coups, and Civil War in
Africa.” World Politics63 (2), 300-346.
Salehyan, Idean, Kristian Skrede Gleditsch, and David Cunningham. 2011. “Explaining
External Support for Insurgent Groups.”International Organization,65(4), 709-744.
Schock, Kurt. 2005. Unarmed Insurrections: People Power Movements in Nondemocracies.
Minneapolis: University of Minnesota Press.
Sharp, Gene. 1973. The Politics of Nonviolent Action. Boston: Porter Sargent Publishers.
Svensson, Isak, and Mathilda Lindgren. 2011. “Community and Consent: Unarmed
Insurrections in Non-democracies.” European Journal of International Relations 17 (1): 97–
120.
Thurber, Ches. 2016. “Social Roots of Revolution : Gandhians , Maoists , and Insurgent
Strategy in Nepal.” Presented at the Annual Meeting of the American Political Science
Association, Philadelphia, PA.
Tilly, Charles. 2008. Contentious Performances. New York: Cambridge University Press.
Vogt, Manuel et al. 2015. “Integrating Data on Ethnicity, Geography, and Conflict: The
Ethnic Power Relations Data Set Family.” Journal of Conflict Resolution 59 (7): 1327– 1342.
Waters, Mary C. 1990. Ethnic Options: Choosing Identities in America. Berkeley: University
of California Press.
Wimmer, Andreas, Lars-Erik Cederman, and Brian Min. 2009. “Ethnic Politics and
Armed Conflict: A Configurational Analysis of a New Global Data Set.” American
Sociological Review 74 (2): 316–337.
Wucherpfennig, Julian, Nils W. Metternich, Lars-Erik Cederman, and Kristian S.
Gleditsch. 2012. “Ethnicity, the State, and the Duration of Civil War.” World Politics 64
(1): 79–115.
.
