PreprintPDF Available

Strategic virtues? Individual Logics of Nonviolent Civil Resistance

Authors:

Abstract and Figures

An emerging consensus holds that nonviolent resistance campaigns are more successful than violent campaigns because they have more participants. Yet, we lack an understanding of whether nonviolent tactics themselves (rather than correlated characteristics) attract mass participation , and how nonviolent resistance motivates supporters. We conduct a survey experiment probing these questions, focusing on two motivational logics: A strategic logic, whereby nonvi-olent resistance is preferred based on cost-benefit considerations, and an intrinsic logic where nonviolent resistance is preferred because of perceived inherent moral superiority. To elicit responses consistent with these logics, we conduct a multi-factorial vignette experiment among a convenience sample of more than 5000 respondents across 33 countries. We find that nonviolent tactics strongly increase movement support relative to violent tactics, and that the preference for nonviolent resistance is primarily driven by intrinsic commitments to the moral superiority of nonviolence and is highly robust to varying factors pertaining to instrumental considerations.
Content may be subject to copyright.
Strategic virtues? Individual Logics of Nonviolent Civil Resistance
Sirianne Dahlum1and Jonathan Pinckney2Tore Wig3,1
1Peace Research Institute Oslo (PRIO)
2Norwegian University of Science and Technology (NTNU)
3University of Oslo
September 9, 2020
Abstract
An emerging consensus holds that nonviolent resistance campaigns are more successful than
violent campaigns because they have more participants. Yet, we lack an understanding of
whether nonviolent tactics themselves (rather than correlated characteristics) attract mass par-
ticipation, and how nonviolent resistance motivates supporters. We conduct a survey experiment
probing these questions, focusing on two motivational logics: A strategic logic, whereby nonvi-
olent resistance is preferred based on cost-benefit considerations, and an intrinsic logic where
nonviolent resistance is preferred because of perceived inherent moral superiority. To elicit re-
sponses consistent with these logics, we conduct a multi-factorial vignette experiment among a
convenience sample of more than 5000 respondents across 33 countries. We find that nonviolent
tactics strongly increase movement support relative to violent tactics, and that the preference
for nonviolent resistance is primarily driven by intrinsic commitments to the moral superiority
of nonviolence and is highly robust to varying factors pertaining to instrumental considerations.
1 Introduction
Nonviolent resistance is a powerful driver of change. History offers many compelling examples
of successful nonviolent movements, from the anti-apartheid movement, and the anti-Communist
1
movements in Eastern Europe at end of the Cold War to the “Jasmine revolution” in Tunisia.
Comparative work indicates that nonviolent resistance movements are more likely than violent
movements to achieve their desired ends - including overthrowing corrupt and authoritarian regimes
(Chenoweth and Stephan, 2011; Bayer, Bethke and Lambach, 2016). One suggested explanation
for this highly cited and influential finding that “nonviolent resistance works” is that nonviolent
movements have greater success in recruiting large numbers of participants. Cross-national data
collection on resistance campaigns strongly supports this explanation, finding that nonviolent move-
ments are, on average, four times larger than violent movements (Chenoweth and Stephan, 2011,
33).
However, it remains unclear whether it is nonviolent (rather than violent) resistance in itself
that attracts participants, or whether it is other correlated features of these campaigns that increase
their appeal. Since existing research on the link between nonviolent resistance and success is
supported primarily by correlational evidence, we can not rule out, for instance, that primarily
nonviolent movements attract more participants because they cluster in specific areas (e.g. cities)
or tend to focus on political demands appealing to more of the population. Furthermore, we lack
a systematic understanding of the cognitive appeal of nonviolent resistance when contrasted with
violent uprisings with similar goals. In short, is nonviolent resistance more attractive for supporters,
and why?
To address these questions, we conduct a survey experiment exploring whether individual sup-
port for a pro-democracy demonstration against a dictatorship is affected by reported use of violent
or nonviolent tactics, and if so, why. We focus on two potential explanations for the appeal of non-
violent resistance. First, nonviolent resistance may be preferred because of its expected advantage
over violent resistance when it comes to efficiency, potential for success, and/or the level of risk
incurred by individual participants. We label this an “instrumentalist” logic, since its driving
forces are expected costs and benefits. A second motivation is present when individuals consider
nonviolent resistance to be intrinsically desirable. We label this an “intrinsic logic” since it values
nonviolent resistance as an inherent moral good, valuable irrespective of contextual factors.
As it is both unfeasible and unethical to experimentally induce participation in a real-world
2
anti-regime protest (particularly one involving violence and the risk of government repression), we
present respondents with a hypothetical scenario of an ongoing anti-regime movement. To gauge
different levels of support for the movement, we ask respondents whether they would support
the movement as a bystander, participate in a public demonstration, or contribute financially.
Although a measure of willingness to participate in a hypothetical social movement will never
perfectly capture real-world participation, we find that respondents are much more hesitant to
participate in a protest than support a movement from the sidelines, suggesting that potential costs
of protest participation are, to a certain extent, factored in. The movement’s use of nonviolent vs
violent tactics is randomly varied between respondents. To identify intrinsic and instrumentalist
logics behind individual support for protest, we also randomly vary the conditions faced by the
hypothetical protest movement – including whether it is repressed (thereby posing a physical threat
to participants) and whether it is reported to be making progress and expected to succeed. This
allows us to analyze, for instance, whether the support for nonviolent resistance is moderated or
mediated by expected success. Our sample consists of more than 5000 respondents in a convenience
sample, across a wide range of national contexts and background characteristics, which allows us
to explore the sensitivity of the results to various background conditions.
We find that nonviolent tactics have a strong and substantial positive effect on movement
support, even when conditioning on other aspects of the movement and strategic context. When it
comes to underlying mechanisms we find most support for our “intrinsic” logic; nonviolent resistance
is considered morally superior, irrespective of instrumental concerns. Instrumental considerations
do matter, but largely irrespective of nonviolent or violent tactics. Furthermore, much of the
nonviolent tactics’ effect is mediated through perceptions of normative worth while a comparatively
small portion of the effect is mediated through expectations of success.
2 Individual support for nonviolent and violent resistance
During the last decade, millions of citizens across the world have taken to the streets in a series of
anti-regime resistance waves, from the Arab Spring in 2011 and the global protest wave that took
place in 2019 and 2020 in countries such as the United States, Lebanon, Hong Kong and Chile.
3
Some commentators have described this as one of the largest waves of political contention since
WWII (Taub and Fisher, 2019). In many cases, such contention generates revolutions that oust
incumbents and set off regime change. In other cases, activists struggle to achieve their goals in
the face of government repression. Systematic comparisons of resistance campaigns from the last
century suggest that successful campaigns tend to share an important characteristic: primary tactics
of nonviolent resistance (Chenoweth and Stephan, 2011). Several studies indicate that movements
that are able and willing to stick to nonviolent resistance even when faced with stalled progress or
violent repression are more likely to both achieve their stated goals and promote democratization
(Bayer, Bethke and Lambach, 2016). One suggested explanation for this is nonviolent resistance
campaigns’ comparative advantage (over violent campaigns) for recruiting participants and gaining
popular support. Consistent with this, nonviolent resistance campaigns are on average larger than
violent campaign, and tend to have more diverse participants (Chenoweth and Stephan, 2011).
Yet, there is very little work considering the individual-level rationales for supporting and/or
participating in nonviolent protest when contrasted with its violent alternative. This is needed
to address many of the assumptions embedded in the macro-level literature. For example, does
nonviolent resistance have greater appeal to both bystanders and potential participants, ceteris
paribus, or do people support and participate in nonviolent resistance campaigns mainly because
they expect them to succeed? Relatedly, do (certain) individuals opt for participation in (or support
for) nonviolent resistance for (partly) normative reasons, as suggested in e.g. Dahlum (2018)?
To explain individual participation in resistance movements, the armed conflict literature of-
ten relies on models centering around the concept of opportunity costs, that have been influential
in research on individual participation in rebel groups and civil war. According to such models,
participation in violent resistance movements is a function of the costs and benefits of participa-
tion, especially compared to the costs and benefits of not participating. This model aligns with
numerous studies establishing a link between structural factors that reduce opportunity costs, such
as economic development, employment and education, and reductions in the risk of civil war or
individual participation in armed conflict (Humphreys and Weinstein, 2008; Blattman, 2009). Yet,
far from all instances of participation in resistance can be explained with reference to opportunity
4
cost models. Another strand of research highlights how individual participation can emerge with-
out feasible opportunity structures, driven by identities, emotions or grievances based on perceived
injustices (Wood, 2003; Tezcur, 2016, e.g.,). Similar models have also been applied to nonviolent
resistance, and there is evidence that individual participation in nonviolent protest is driven by
both opportunity structures, shaped by e.g. economic conditions (Schussman and Soule, 2005; Ver-
hulst and Walgrave, 2009; Dalton, Van Sickle and Weldon, 2010), or grievances rooted in identities
or ideology (e.g. Orbell, 1967; R¨udig and Karyotis, 2014).
While this literature offers important insights into potential explanations for individual partic-
ipation in resistance movements more generally, it does not, however, sufficiently account for how
individual participation may depend on the way that resistance is conducted. In particular, it does
not address the distinction between violent and nonviolent political resistance, typically considering
them two separate domains. This leaves us in the dark with respect to how variations in the use of
violence by protest movements influence individual decisions to support or participate. We do this
by exploring the specific appeal of nonviolent tactics, when holding all other protest/opposition
characteristics constant.
Our study also speaks to a literature on normative evaluations of different kinds of political
contention, which has looked at various forms of support for mass resistance, with an emphasis
on political violence (Hillesund, 2015; Blair et al., 2013; Fair et al., 2016). While this research
has made great strides, it does not address the specific issue of attitudes to tactical choice, ceteris
paribus, since it looks at the actions of specific political actors; specific militant groups in specific
conflicts and rarely varies the tactical choices within the group. Our analysis also speaks to social-
psychological work on attitudes to violence (and, by contrast, nonviolence) more generally, which
looks at causes such as honor cultures (Cohen, 1996), gender (Smith, 1984), media content (Rule
and Ferguson, 1986; Bensley and Van Eenwyk, 2001) and other factors.
We build on and contribute to an emerging strand of studies looking at the links between
resistance group tactics and public support. Focusing on rebel groups, Arves, Cunningham and
McCulloch (2019) explores the link between choice of strategies and American public opinion,
demonstrating that use of elections in the rebel organization, demonstrations, and hunger strikes
5
improve perceptions of rebels, while use of terrorism decreases support. Simpson, Willer and
Feinberg (2018) explores respondents’ perceptions of a scenario of nationalist protesters and anti-
racist counter-protesters. They find that use of violence reduces public identification and support
for anti-racist protesters, while increasing support for the nationalists. Finally, comparing surveys
conduced before and after the outbreak of violent riots in Barcelona by the anti-austerity movement
15-M, Mu˜noz and Anduiza (2019) finds that these violent riots led to decreases in support for
the movement among the general public, but less so for its core supporters. Another related
study is Lupu and Wallace (2019), who shows that violent opposition tactics increases support for
government repression and violations of human rights.
With the exception of Lupu and Wallace (2019), these studies do not consider nonviolent re-
sistance in dictatorships, but they suggest that nonviolent resistance is an attractive attribute of
resistance campaigns that can motivate greater support. While these studies have pushed the re-
search frontier on the question at hand, they have two other shortcomings which we improve on.
First, these studies do not distinguish between different levels of involvement (e.g., from stated
support, to stated participation and financial donation). Second, and most crucially, while these
studies indicate some support for nonviolent movements in specific political settings, they do not
probe different rationales and mechanisms accounting for greater support for nonviolent movements.
Different explanations have been put forward. For example, Thomas and Louis (2014) suggest that
the reason is that observers perceive nonviolent resistance as more effective, while Simpson, Willer
and Feinberg (2018) point to personal identification with protesters, but we lack tests on these
propositions. This necessitates additional research that does more than comparing nonviolent and
violent resistance, and also varies and probes the correlated characteristics of nonviolent resistance
that may motivate participation.
In sum, we currently lack a rich understanding of how and under what conditions variations in
the use of nonviolent and violent strategies in anti-regime resistance campaigns influences individual
participation and support. Our paper responds to this by constructing the survey experiment we
outline below. Before presenting this, we discuss two motivational logics that may produce support
for nonviolent resistance.
6
3 Mechanisms: Intrinsic and instrumentalist logics
We outline two cognitive“logics” that may underlie support for nonviolent resistance: an instru-
mentalist logic emphasizing costs and benefits, and an intrinsic logic that emphasizes nonviolent
resistance’s inherent moral superiority. We do not claim that only one of these operate - non-
violent resistance may often be driven by a combination of instrumentalist and intrinsic reasons.
Acknowledging this, we investigate whether one or both of these are a driving force for nonviolent
resistance, and gauge how strong they are.
Instrumentalist logic: An instrumentalist logic operates when an individual chooses to sup-
port or join nonviolent resistance because of its consequences. Individuals can value different kinds
of consequences: A crucial set of consequences relate to the effects of the campaign itself. If a
campaign is expected to be more successful because it uses nonviolent tactics, and an individual
participates in or supports the campaign because of this expectation, then s/he is motivated by an
instrumentalist logic that values nonviolent tactics for their effects on the probability of campaign
success. Another set of consequences relates to the individual. If a person thinks that nonviolent
resistance reduces their risk of harm, when compared to violent resistance, and this influences their
decision to participate, then this person is motivated by an instrumentalist logic emphasizing in-
dividual consequences. Hence, an instrumentalist logic can mix both individual-level cost/benefit
calculations and campaign-level consequences. The instrumentalist logic yields the following overall
hypothesis:
H1: Nonviolence increases individual preferences for participation in protest, and this effect is
moderated and mediated by positive expected campaign- and individual-level consequences
Intrinsic logic: An intrinsic logic operates when individuals consider nonviolent resistance
to be inherently beneficial. The individual derives intrinsic utility from the use of nonviolent
resistance as a form of political contention. This follows if individuals are motivated by deontological
concerns, viewing physical harm against others as intrinsically bad (e.g., Kant and Schneewind,
2002; Korsgaard and Korsgaard, 1996), or if they have an instinctive aversion to violence, even if
they expect it would be more effective in achieving their goals. According to an intrinsic logic,
we would expect individuals to prefer nonviolence to violence when expected outcomes are held
7
constant.
While there is likely a clear difference in the degree to which intrinsic or instrumental consid-
erations on average correlate with valuing consequences or intrinsic aspects, it will often be hard
to completely isolate the two. In our case, for example, it is hard to isolate “the act” fully from
its consequences. It is hard to envision a situation where violence is used, but where violence as
such is the only difference made, with no difference in consequences. After all, violence is partly
defined by a consequence; physical harm. So, it will not be easy to perfectly separate the two
categories. Nevertheless, we think the two categories of motivations will have differing tendencies.
If nonviolent resistance is a very strong factor when holding other relevant consequences constant,
we think this is evidences an intrinsic logic. The intrinsic logic yields the following hypothesis:
H2: Nonviolence increases preferences for participation regardless of campaign- and individual-
level consequences
These two logics can be operationalized in two different approaches. First, they can be op-
erationalized in terms of moderation. If the impact of nonviolent resistance is moderated by its
perceived likelihood of success and reduced likelihood of repression, then this indicates an instru-
mentalist logic. Second, they can be operationalized in terms of mediation. If the effect of nonviolent
resistance is mediated through impressions of success, such that an individual supports a nonviolent
movement because they think it is successful. We have no strong priors about which of these ways
to operationalize the mechanisms constitute stronger evidence, and suspect that exploring both is
a better alternative.
4 Research design
4.1 Survey design: Multi-factorial experiment
To test these expectations, we propose a multi-factorial vignette experiment.1This presents re-
spondents with a text with treatment components that vary simultaneously, while holding other
aspects of the text constant. Rhis allows us to compare the relative impact of different treatments
1Our experiment, including the main hypotheses, is pre-registered in the EGAP registry, and is available at
BLINDED FOR PEER REVIEW.
8
to evaluate their relative importance, and - by extension - the theories motivating them. Since
treatments vary randomly across respondents, we can estimate treatments’ impact by assuming
strong ignorability.
This design has several desirable features. Combined with an appropriately sized sample, it
enables us to investigate several hypotheses, allowing us to evaluate different theories (when holding
other factors of the experiment constant). While we cannot know for certain how responses to our
vignette correlate with real-world behavior recent studies indicate that the link between the two
is often quite robust. Hainmueller, Hangartner and Yamamoto (2015) show, for instance, that
attitudes towards the attributes of asylum seekers measured through survey experiments closely
match voter behavior on corresponding referendums in Switzerland.
Similar designs to ours have been used to investigate various questions, such as attitudes to
immigrants (Hainmueller and Hopkins, 2015), voter-preferences (Carlson, 2015), and attitudes to
international interventions (e.g. Johns and Davies, 2019). Below, we discuss in more detail to what
extent our responses correlate with real-world decisions.
4.2 Sample
We recruited our respondents on the Prolific platform. Prolific is a survey-participant recruitment
company aimed towards academic studies. While Prolific is relatively new, and thus there is not a
wide body of research using its platform,2Prolific’s participant pool has been found to outperform
several common alternatives such as Amazon Mechanical Turk (MTurk) on basic tests of honesty
and attention, and is also more diverse than the MTurk participant pool (Peer et al., 2017; Palan and
Schitter, 2018). Prolific’s platform also allows for the payment of individual respondent bonuses.
This gives greater credibility to one of our dependent variable questions where we ask if respondents
would like to donate money to the movement (see below). Since our vignette describes a fictional
movement, we do not, in the end, let them give away their donation.
2However, see recent examples such as Flynn and Stewart (2018).
9
4.3 Treatment variables
We present respondents with a fictional vignette described as a recent news article about a public
demonstration against an autocratic regime. The demonstrators accuse the regime of stealing the
latest election and engaging in corruption, and for these reasons demands that the head of the
regime resign.
We randomly assigned respondents to one of two versions of the vignette: One that describes
the events as taking place in a non-specific country and one that describes them as taking place in
Togo, a country with a current pro-democracy movement (Amegboh, 2018), but which is unlikely
to be familiar to our respondents. We included both versions because of twin concerns for potential
bias. A vignette describing a specific country (such as Togo) might lead to participants inferring
other characteristics of the described events not explicitly stated depending on their preconceived
ideas about that country. On the other hand, a vignette describing a non-specific country might
lead participants to treat the vignette with less seriousness and care. Our expectation, validated
through pre-testing, was that these sources of bias would not be substantive, but we included both
out of an abundance of caution and as a robustness check to show that our results are not dependent
on the vignette’s setting.
Both versions include three randomly-varying treatments:
Use of nonviolent resistance by protesters:The treatment group is presented with in-
formation about a protest movement that uses nonviolent resistance methods, including peaceful
marching, while the control group receives information about a movement that uses violent tactics,
including attacks with makeshift weapons, and espouses a commitment to violence. We chose to
focus on rioting with makeshift weapons as our “violent” condition to increase comparability be-
tween the violent and nonviolent conditions, and to speak specifically to research focusing on the
distinction between nonviolent resistance and “unarmed collective violence”(Kadivar and Ketchley,
2018).
The treatment has two corresponding text pieces – the reported use of violent/nonviolent tactics
and stated commitments to it. The first text piece describes the strategies that the movement has
used. The second includes a statement by a protester that emphasizes the movement’s commitment
10
to either nonviolent or violent tactics. We include the explicit statement of commitment to one
tactical repertoire to make very clear to our respondents that the tactics described in the event are
core to the movement’s identity and unlikely to change.
Use of repression by the regime:We vary whether the news story reports that the police
interfered violently with the protesters, resulting in casualties, or whether the police did not inter-
fere. This treatment is partly intended to capture the personal cost aspect of the instrumentalist
logic, that is whether people join movements because they perceive the risks to their personal safety
to be low. We ask respondents about whether they would join the movement, and assume that
there is, at least, a correlation between their responses and their actual (in this case unobservable)
behavior.
Progress:We also vary whether the protest movement is reported to be making strategic
progress. This component consists of two corresponding pieces of text. First, we present information
on whether or not the movement has received partial government concessions in the recent past.
Second, we vary information offering insights into whether the movement can be expected to succeed
in the future, in the form of a statement by an expert. This treatment is intended to capture the
political aspect of the instrumentalist logic, that is whether people join movements because they
believe they are more likely to succeed.
The Togo version of the full survey-vignette with the different treatment arms is outlined in
Figure 1 below.
11
Figure 1: The survey-vignette
12
4.4 Outcome variables
We test the impact of our varying vignettes on four outcome variables:
1. Based on this news report, do you support or oppose the opposition movement? On a scale
from 0 to 5, where 0 indicates “Not at all” and 5 indicates “Very strong support”
2. Imagine that you are a [citizen of Togo / citizen of this country], how likely would you be
to support this movement? On a scale from 0 to 5, where 0 indicates “Not at all” and 5
indicates “Very strong support”
3. Imagine that you are a [citizen of Togo / citizen of this country], if a friend asked you to join
one of the opposition movement’s protests, how likely would you be to do it? On a scale from
0 to 5, where 0 indicates “Definitely would not join” and 5 indicates “Definitely would join”
4. What percentage of your payment for completing the survey would you be willing to donate
to help this movement? Numeric scale 0 to 100.
These questions are intended to capture different levels of commitment to the movement in
question. For example, one might ideologically support the movement as an external observer
living in another country, but may be more hesistant to support it as a citizen of the country
where the movement is located. Similarly, even local citizens may support the movement but
not want to participate for fear of potential personal costs. These are crucial distinctions, with
real-world consequences. As Kuran’s (1991) work indicates, support for an opposition movement
may be quite widespread and yet the movement may fail if such support remains nascent, never
manifesting in actual participation. We frame the participation question in terms of a friend asking
the respondent to participate both to enhance the question’s realism and to capture one of the
most common avenues of recruitment into activism (McAdam, 1986).
The fourth question moves respondents’ answers from the theoretical to the practical. If re-
spondents are willing to give up actual money that they just earned this sends a strong signal that
the experimental stimulus has indeed powerfully influenced their level of support for the movement.
13
Rather than tapping abstract beliefs about the efficacy of nonviolent resistance, we seek to
get directly at respondents’ perception of whether this movement is likely to succeed and if they
themselves would be likely to face physical repression were they to join. While individuals may
vary in the degree to which they believe nonviolent resistance in the abstract may or may not work,
they are likely to update this empirical belief as they observe a movement in progress.
All four questions situate respondents as observers of a previously-existing movement with the
option to support, join, and donate or not support, join, or donate. We choose this set-up because
this is the situation faced by most individuals who join a resistance movement. Movements are
typically initiated by ideological entrepreneurs whose commitment to the cause is based on relatively
stochastic, contingent factors. Thus, we would not expect to be able to gain significant insight into
the process of movement formation through this experiment. Almost all movement participants,
however, tend to join when a movement has already been initiated, after observing a period of
strategic interaction between the movement and the state. This is the situation that our vignette
is intended to capture. This focus also aligns with the paper’s ambition to explore the mechanisms
behind the potential advantages of nonviolent resistance in recruiting participants to an already
ongoing movement.
4.5 Background variables
While our primary hypotheses of interest relate to the effects of nonviolent resistance, personal
risk, and long-term success on the likelihood and level of support, we also consider it highly likely
that these treatment effects will be heterogeneous across different populations. We do not present
specific hypotheses for these conditional effects, but will present results in an exploratory fashion.
First, we include demographic questions on age, sex, education, student status, and marital status
of the respondents, to see if they are related to the outcome variables, and also whether they affect
the impact of each of our three treatments. We also include the following two questions:
14
Have you ever personally participated in a political protest (for example a march, rally, or
demonstration about a political issue)? Yes/no
“To what extent do you support the current government of the country that you reside in”?
5 level Llkert scale ranging 1-5, from 0 =“strongly oppose” to 5 =“strongly support”
This is to get at whether effects vary with the degree to which respondents are politically active,
and/or hold dissenting views. It is likely that respondents with previous protest history should be
more likely, on average, to support both violent and nonviolent movements.
4.6 Post-treatment questions
An additional implication of the theory discussed above, in addition to expecting an interaction
effect, is that we should see particular patterns of causal mediation. For example, if the intrinsic
logic operates, we expect people who support nonviolent movements to do so because they view
the protesters as having justice on their side. In contrast, we expect support to be mediated by
expectations of success if the instrumental logic dominates. To test this, we include the following
post-treatment questions, to test the mediation hypotheses, and to serve as serve as validation
checks for the different treatment combinations:
15
“How much do you agree with the following statements?:”
“The government in this article has justice on its side” 1-5 Likert Scale, 1 =“strongly
disagree” to 5 =“strongly agree.”
“The protesters in this article have justice on their side” 1-5 Likert Scale, 1 =“strongly
disagree” to 5 =“strongly agree.”
“The government in this article is likely to win” 1-5 Likert Scale, 1 =“strongly disagree”
to 5 =“strongly agree.”
“The protesters in this article are likely to win” 1-5 Likert Scale, 1 =“strongly disagree”
to 5 =“strongly agree.”
“To what extent do you think the protesters used the following strategies?:” “Only
nonviolent strategies”, “a mix between nonviolent and violent strategies”, “only violent
strategies”, “don’t know”
“Participating in an event like this one would threaten my personal safety” Five-level
Likert Scale from “strongly disagree” to “strongly agree.”
4.7 Attention checks and comprehension questions
We include comprehension and attention checks to ensure our respondents have paid attention to
and understand the vignette. Our core attention check asks respondents to recall a fact from the
vignette.3In addition we test comprehension in the pre-test by including a question on whether
protesters used violent methods, nonviolent methods, or a mixture of the two.
3For the Togo version of the vignette we ask respondents the name of the President of Togo, and for the generic
country we ask respondents the name of the square in which the protest took place.
16
4.8 Information equivalence
One core concern with any survey experiment is ensuring that one does not violate the “information
equivalence” assumption (Dafoe, Zhang and Caughey, 2018), that is to say that the treatments only
trigger differences in the respondents’ perception of the actual concept of interest. Dafoe, Zhang and
Caughey (2018) suggest several strategies for ensuring information equivalence, including specifying
additional background details to limit respondent inferences and embedding a natural experiment
in the survey vignette. We were concerned that an embedded natural experiment might confuse
respondents or reduce respondent attention to detail in our treatments. Thus, we followed the route
of specifying the most relevant background details.
Reflecting this, all versions of the vignette describe a public demonstration by a pro-democracy
opposition movement engaged in a resistance campaign against a “dictatorial president” who we de-
scribe as “notorious for high levels of political corruption” and engaged in election fraud. We chose a
movement pursuing these goals to hold relatively constant what inferences about a movements’ goals
respondents might make, and in particular whether our respondents would ideologically support
such goals. Thus, when we ask questions about their willingness to support or join the movement,
their responses will be informed primarily not by the movement’s goals but by the varying tactics
and government responses captured in our different treatments. Extensive cross-national research
shows that political corruption reduces regime support and public trust in institutions across a wide
variety of contexts (Anderson and Tverdova, 2003; Seligson, 2002), and election fraud is a common
and highly-resonant frame for collective action across political contexts (Tucker, 2007). Thus we
considered corruption and election fraud to be easily relatable and widely-supported goals for our
hypothetical opposition movement.
While these strategies help ensure information equivalence, there are also trade-offs between
such specificity and the generalizability of our particular scenario. We emphasize that the vignette
broadly describes many of the most transformative nonviolent (and violent) movements of the last
several decades, and thus inferences from this vignette are of important theoretical and practical
significance. However, making inferences about the logic of support for resistance campaigns outside
of the context we describe in our vignette should only be done very cautiously.
17
4.9 Pre-testing
We conducted several rounds of pre-testing the vignette prior to deploying the final survey ex-
periment to ensure that the treatments clearly captured our underlying concepts of interest. We
significantly revised the vignette based on results from our initial rounds of pre-testing, most promi-
nently by including the explicit statements of commitment to nonviolent or violent resistance by an
interviewed protester. The final round of pre-testing indicated clear understanding and distinction
between the different treatments, for instance with respondents clearly identifying the protesters
in the “nonviolent” condition as employing solely nonviolent strategies, and respondents clearly
identifying the government’s use of violence in the “repression” condition. We present detailed
results from and discussion of our pre-testing in the appendix.
4.10 Estimation
Our estimand of interest is the Average Treatment Effect (ATE) of each vignette treatment. We
can estimate the ATE based on observed data, since we randomize the assignment of treatments
to vignettes. Randomization occurs at the individual level. The link provided by Prolific directs
respondents to an IP address containing a short Javascript that generates a uniformly-distributed
random number and then, based on the value of that random number, redirects respondents to one
of 16 distinct online surveys containing the various vignette versions (Togo/nondescript country *
violent/nonviolent * repression/no repression * success/no success). Since we are interested in how
the effect of nonviolent resistance is conditional on the different “logics” identified above, we will
be interested in the interaction between violent/nonviolent resistance and different attributes.
We will here present the full model we estimate. We include the following treatment variables:
Nonviolent/violent (N), repression/no repression (R), success/no success (S). Our outcome Yis
any of the four post-vignette questions presented above. With this set of variables, we want to
estimate the average treatment effects (ATEs) of the different treatments (and combinations). We
will do this by estimating OLS models. We estimate a separate OLS model for all of our different
outcomes, and one for a composite outcome of the four outcome variables. This composite outcome
will be a simple additive index, where we take the mean score across the measures (the “donation”
18
variable is divided by 20 and ranges between 0 and 5 to conform with the other measures). We
estimate the following linear model, using Ordinary Least Squares:
Y=β0+β1N+β2R+β3S+β4N·R+β5N·S+(1)
As discussed above, our hypotheses H1 and H2 pertain to the conditional and unconditional
effects of nonviolent resistance. H1 can be broken down into:
H1a: There is a positive interaction effect between nonviolent resistance and success
and
H1b: There is a negative interaction effect between nonviolent resistance and repression
while H2 can be formulated in the same terms as it is formulated above, namely that there is an
unconditional relationship between nonviolent resistance and support. Table 1 below summarizes
our concrete expectation for the different coefficients, and how they relate to the hypotheses above.
Table 1: Concrete expectations under the different hypotheses
Hypothesis Label Coefficient Sign
H1a Nonviolence/success β5N·S+
H1b Nonviolence/repression β4N·R
H2 Nonviolence β1N+
We estimate 5 treatment effects plus the intercept (6 parameters in total). Since we test multiple
hypotheses, we apply the Bonferroni correction when calculating significance levels. We take .05
as the overall significance level, and adjust for 3 hypotheses. Each individual hypothesis is thus
tested at .05/3 = .017.
19
5 Results
We received 5,040 responses to our survey. Our attention check question revealed a high degree
of competence among respondents, with only 40 (or roughly 0.8 per cent) failing the check. We
removed these respondents from our tests. 22 respondents failed to complete the survey for technical
reasons, resulting in a final sample size of 4,978. The sample was relatively gender-balanced with a
slight edge towards female respondents (51.5 per cent female)4, and skews somewhat young, with
a median age of 30. While a majority of respondents (2,915) were American or British, the total
sample contains 86 reported nationalities and 33 countries of residence (see appendix for details).
5.1 Support for nonviolence
We first analyze the overall effect of nonviolent resistance, for all the outcome variables measuring
levels of support listed above. The coefficients from these models are presented in Figure 2. The
first coefficients (starting at the top) describe the average treatment effect of nonviolent resistance,
when holding all else constant. This indicates a substantial effect of nonviolent resistance on the
tendency to support, support as imagined citizen, join, and donate to the movement. Consequently,
the monviolent resistance treatment yields an increase in the aggregate index of almost 1 point.
This is quite substantial, as the index ranges from 0 to 5, with an empirical mean of 3.99 (median
of 4) and a standard deviation of about 1.21. The resulting coefficient represents more than one
standard deviation.
The coefficients for success are mostly indistinguishable from zero in this baseline model. This
suggests that the overall decision to support a protest movement (regardless of use of nonviolent
protest strategies) is not conditioned by expectations about potential success. Repression, on the
other hand, is negatively associated with most indicators of protest support, including stated sup-
port, stated support as citizen and willingness to join the movement. This suggests that individuals
do factor in costs in the form of (expected) repression when deciding whether to support a protest
movement. Moreover, it provides some assurance to our research design, as it suggests that protest
4Prolific’s demographic data provides only a binary gender variable, thus we are unable to say whether there are
any non-binary individuals in our sample.
20
Figure 2: Baseline results
Repression
Success
Non_Violence
0.0 0.5 1.0
Coefficient
Variables
Outcomes
Index
Support
Support as citizen
Join
Donate
Main results
risk is factored in even when considering hypothetical (as opposed to real) protest support and/or
participation. Repression is most strongly negatively linked to willingness to join, but matters less
for stated support. The only indicator of support which is not negatively affected by repression is
willingness to donate, which increases when protest movements are reported to be repressed. This
also aligns with the interpretation that reported repression decreases support when repression is
considered as a risk factor, which will not be the case for donation.
5.2 Instrumentalist and intrinsic motivations:
Next, we turn to exploring the conditions under which individuals prefer nonviolent resistance. To
do this, we first estimate interaction effects between reported success and nonviolent resistance,
testing hypothesis H1a above, which suggests that support for nonviolent resistance will increase
when its use is reported to lead to progress. The results are presented in figure 3, which shows
the predicted difference in support between violent to nonviolent tactics, distinguishing between
expected support when movements are succeeding and failing. Contrary to hypothesis H1a, there
21
is very little evidence that support for nonviolent resistance is strongly conditional on expectations
about success. As seen in figure 3, the lines showing the effect of nonviolence on support in both
the success and non-success condition overlap. This is the case for all five indicators of support.
Figure 3: Effect of nonviolence on support. Conditioning on success.
3.75
4.00
4.25
4.50
4.75
0.00 0.25 0.50 0.75 1.00
Non_Violence
index
Success
0
1
Predicted values of index
4.8
5.2
5.6
0.00 0.25 0.50 0.75 1.00
Non_Violence
support_oppo
Success
0
1
Predicted values of support_oppo
5.0
5.5
0.00 0.25 0.50 0.75 1.00
Non_Violence
support_oppo_cit
Success
0
1
Predicted values of support_oppo_cit
3.75
4.00
4.25
4.50
4.75
0.00 0.25 0.50 0.75 1.00
Non_Violence
join_protest
Success
0
1
Predicted values of join_protest
2.0
2.5
3.0
0.00 0.25 0.50 0.75 1.00
Non_Violence
donate2
Success
0
1
Predicted values of donate2
We now turn to the interaction between repression and nonviolence. H1b predicts that nonvi-
olent resistance is less likely to lead to support if it is met with repression. However, across all the
different measures of support, a shift from violent to nonviolent resistance increases support even
when the movement is met with repression. When considering general support, support as citizen
and the support index, there is little difference between repressed and non-repressed movements
when it comes to the extent to which nonviolence increases support. Willingness to participate is
22
slightly higher when the movement is not repressed compared to when it is repressed, but repres-
sion also reduces reported willingness to participate in violent movements. This reflects repression’s
general discouraging effect, which is illustrated by the lines being parallel. This suggest that re-
spondents possibly view repressed nonviolent movements as more deserving of support from a moral
point of view, and that this is balanced by the threat to physical safety for the willingness to join.
Figure 4: Effect of nonviolence on support. Conditioning on repression.
4.0
4.5
0.00 0.25 0.50 0.75 1.00
Non_Violence
index
Repression
0
1
Predicted values of index
4.4
4.8
5.2
5.6
0.00 0.25 0.50 0.75 1.00
Non_Violence
support_oppo
Repression
0
1
Predicted values of support_oppo
4.8
5.2
5.6
0.00 0.25 0.50 0.75 1.00
Non_Violence
support_oppo_cit
Repression
0
1
Predicted values of support_oppo_cit
4.0
4.5
0.00 0.25 0.50 0.75 1.00
Non_Violence
join_protest
Repression
0
1
Predicted values of join_protest
2.0
2.5
3.0
3.5
0.00 0.25 0.50 0.75 1.00
Non_Violence
donate2
Repression
0
1
Predicted values of donate2
23
5.3 Mediation analysis
As discussed, an alternative potential implication of our theoretical framework is that the effect
of nonviolent resistance on support is mediated by different factors. If preferences for nonviolent
resistance are driven by the instrumentalist logic, we might expect that the nonviolent resistance’s
effect on support is mediated by expectations about success. If the intrinsic logic is strong, we
expect support for nonviolent resistance to be mediated by opinions of the moral superiority of
nonviolent protesters. To gauge this, we draw on post-treatment question asking respondents about
their perceptions of movement success, and who has justice on their side. We use the mediation
analysis framework from Imai et al. (2011). This construes causal mediation in terms of a treatment
variable T, a mediator variable Mand the outcome Y. The goal is to disaggregate the Average
Treatment Effect (ATE) into the Average Direct Effect (ADE) and the Average Causal Mediation
Effect (ACME). The former is the proportion of the effect of Ton Ythat does not work through
M, while the ACME represents the mediation effect; the proportion of the ATE that is explained
by the relationship between T,Mand Y.
In Figure 5 we present the mediation analysis results, drawing on two post-treatment questions.
The first measures whether respondents believe that “the opposition is likely to win”. We assume
that this mediator question proxies for the instrumentalist logic. Next, we present mediation
analysis exploring whether the effect of nonviolence on support is mediated by whether respondents
believe that the “opposition has justice on its side.” In contrast to the previous question which
taps into expectations about success, this question captures general moral support for nonviolence.
The mediation analysis shows three things. First, there is a mediation effect (ACME) for both
success perception (left) and the perception of opposition morality (Right). Second, the ACME for
opposition morality (39 per cent) is almost twice as large as that of perceptions of success (20 per
cent). Furthermore, both analyses show that most of the effect of nonviolent resistance is direct,
not mediated by expectations of success nor by perceptions of justice. We show mediation analyses
for the other (disaggregated) outcomes in the appendix.
In summary mediation analysis partly confirms the pattern uncovered in the interaction anal-
yses. Nonviolent resistance has a strong effect on support. This is only weakly related to in-
24
Figure 5: Mediation analysis (Outcome: Support index)
0.0 0.5 1.0 1.5
Mediation effect on support through ``opposition likely to win
Total
Effect
ADE
ACME
0.0 0.5 1.0 1.5
Mediation effect on support through ``opposition has justice on their side
Total
Effect
ADE
ACME
Left: M=“Opposition is likely to succeed” (Prop. mediated= 20%), Right: M=“Opposition has justice on their
side” (Prop. mediated= 20%)
strumentalist concerns. Respondents are highly likely to support nonviolent resistance movements
regardless of their perceptions of their success. Perceptions of the moral value of nonviolent protest
yields a much stronger causal mediation effect than expectations of success.
5.4 Additional tests
We explore several heterogeneous effects inductively. First, does nationality significantly affect our
findings? Figure 6 shows how the effects vary by one major dimension. Since the majority of
respondents are in the US and UK, we split the sample into US/UK respondents (N=2915) and
non-US/UK respondents (N=2063). Results are similar across groups, although the interactive
effects are somewhat weaker for the US/UK respondents. Finally, we conduct a split-sample test
looking at differential effects among women and men (Figure 7). Women appear to be much more
swayed by nonviolent resistance than men. The ATE for women is almost double that for men
across most measures. This conforms with studies showing that women are, on average, more
averse to violence (Yablon, 2009; Regan and Paskeviciute, 2003). The comparison also shows that
men are more likely to support and join a nonviolent movement that is violently repressed. This
is consistent with existing evidence that men are less averse to violence and more risk-seeking (e.g.
25
Non_Violence:Repression
Non_Violence:Success
Repression
Success
Non_Violence
−0.5 0.0 0.5 1.0 1.5
Coefficient
Variables
Outcomes
Index
Support
Support as citizen
Join
Donate
Main results: Non UK/US respondents
(a) Results for respondents outside the US or UK
Non_Violence:Repression
Non_Violence:Success
Repression
Success
Non_Violence
−0.5 0.0 0.5 1.0 1.5
Coefficient
Variables
Outcomes
Index
Support
Support as citizen
Join
Donate
Main results: UK/US respondents
(b) Results for US or UK respondents
Figure 6: Comparing results when splitting on US/UK vs. non-US/UK
Sapienza, Zingales and Maestripieri, 2009).
In addition to these explorations of heterogeneity, we conduct several robustness tests. Table
2 shows results from two tests of particular interest. The first column conducts a balance test,
regressing the nonviolent resistance outcome on covariates. Most covariates are balanced with
respect to the treatment, with the exception of prior protest participation. We therefore run tests
where we condition on that variable, with no substantive difference in results. The second column
shows correlations between our background variables and the outcome (measured using the additive
index) of support for the protest movement, across different conditions. It shows that men are less
likely to support the movement, non-UK/US respondents more likely to support, people who have
protested before are more likely to support, and people with positive attitudes of their government
less likely to support.
We also re-ran our models limiting the population to the vignette version describing a non-
specific country or describing Togo. Results are substantively similar across these two populations.
This can be found in the appendix. We also conducted several validation checks. We provided
respondents the opportunity to give open-ended text responses in pre-testing to describe respon-
26
Table 2: Balance check and correlations between background factors and outcomes
Dependent variable:
Non Violence Support index
(1) (2)
Married 0.015 0.111
(0.017) (0.195)
Gender 0.003 0.363∗∗
(0.015) (0.179)
Age 0.0004 0.024
(0.004) (0.051)
Age sq 0.00000 0.0004
(0.0001) (0.001)
Non UK/US resident 0.003 0.805∗∗∗
(0.016) (0.191)
Student 0.004 0.270
(0.020) (0.235)
Education 0.005 0.106
(0.006) (0.069)
Personal protest 0.033∗∗ 2.842∗∗∗
(0.017) (0.197)
Govt attitude 0.003 0.336∗∗∗
(0.006) (0.075)
Constant 0.523∗∗∗ 16.714∗∗∗
(0.082) (0.964)
Observations 4,593 4,582
R20.001 0.065
Note: p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
27
Non_Violence:Repression
Non_Violence:Success
Repression
Success
Non_Violence
−0.5 0.0 0.5 1.0 1.5
Coefficient
Variables
Outcomes
Index
Support
Support as citizen
Join
Donate
Main results: Men
(a) Results for men
Non_Violence:Repression
Non_Violence:Success
Repression
Success
Non_Violence
−0.5 0.0 0.5 1.0 1.5
Coefficient
Variables
Outcomes
Index
Support
Support as citizen
Join
Donate
Main results: Women
(b) Results for women
Figure 7: Comparing results when splitting on Men vs. Women
dents’ common impressions of the movement, regime, and country where the events take place
across treatment conditions. This validated the informational content of the treatments. Including
the other treatment variables as covariates (e.g., conditioning on government repression when we
estimate the effects of nonviolent resistance) will also capture some of the potential covariation
between our treatments of interest and correlated associations. If our results are highly unstable
to the inclusion of such controls, then this will indicate a potential violation of the assumption of
informational equivalence. But, this does not seem to be the case, as coefficients are quite stable
across the inclusion of these controls.
6 Conclusion
Our survey experiment builds on the growing literature on nonviolent resistance by gauging the
attitudinal underpinnings of resistance campaigns. Our main ambition is to test whether nonviolent
resistance, holding other aspects constant, induces different kinds of support, and second to explore
the different potential motivations for supporting nonviolent resistance. We propose that individuals
28
can support nonviolent resistance for intrinsic reasons relating to its perceived moral rightness, and
for instrumental reasons, relating to its costs and benefits.
Our main finding is that nonviolent resistance is a strong pull factor in attracting movement
support. Nonviolent tactics are a powerful source of movement support even when other factors
would discourage support on instrumental grounds, suggesting a strong intrinsic logic. We also find
that the effect of nonviolent resistance is strongly mediated by considerations of its moral value.
While there is a mediation effect of expectations of success, the mediation effect of perceptions of
justice is nearly twice as large. This finding could help explain why nonviolent resistance campaigns
are often larger than violent ones in similar settings. These results are highly robust across different
model specifications and sample sub-groups.
Our exploratory findings on the differential impact of our treatments across different variables
also uncovered interesting variation. The differences across gender speak to the growing literature
on the impact of gendered participation in movements (Principe, 2017; Chenoweth, 2019). Both
men and women are significantly more likely to support, join, or donate to a nonviolent movement
relative to a violent movement, yet the logic of nonviolence appears to be much more compelling
to women than men. Men in our sample, however, exhibited an “anti-instrumentalist” logic when
it came to our repression treatment, becoming significantly more likely to donate and support a
nonviolent movement when that movement was violently repressed.
Our results suggest that nonviolent resistance strongly increases movements’ levels of support
relative to violent resistance, supporting recent observational research suggesting the positive im-
pact of nonviolent tactics on movements’ abilities to achieve their goals (Wasow, 2020). This in
turn implies that movements interested in maximizing public support and achieving their strategic
goals would do well to pursue strategies that reduce the likelihood of violent outbreaks.
It is important to emphasize that our research does not speak to the critically important question
of how violent outbreaks in primarily nonviolent resistance campaigns occur, nor should our findings
be interpreted as placing undue moral responsibility for maintaining nonviolence in the face of
violent provocation on protesters’ shoulders. Existing research suggests that government repression
is the most consistent factor leading to violence in primarily nonviolent movements (Pinckney,
29
2016), and indeed many repressive governments explicitly seek to provoke violence by protesters at
otherwise nonviolent demonstrations (Marx, 1979).
Our findings have implications for the growing literature on the micro- and macro-dynamics
of nonviolent protest (Chenoweth and Stephan, 2011; Bayer, Bethke and Lambach, 2016; Arves,
Cunningham and McCulloch, 2019), protest and social movements more generally (e.g. Tarrow,
1994), democratization and mobilization (Acemoglu and Robinson, 2001; Aidt and Leon, 2015),
and the political psychology literature on attitudes to political violence (Hillesund, 2015; Blair
et al., 2013; Fair et al., 2016). We encourage further work on this research agenda, looking into the
microfoundations of support for- and participation in nonviolent resistance and political violence.
30
References
Acemoglu, Daron and James A. Robinson. 2001. “A Theory of Political Transitions.” American
Economic Review 91(4):938–963.
Aidt, Toke S and Gabriel Leon. 2015. “The Democratic Window of Opportunity Evidence from
Riots in Sub-Saharan Africa.” Journal of Conflict Resolution p. 0022002714564014.
Amegboh, Joel. 2018. “Togo Protests Continue to Demand Enforcement of Term Limits.” Africa
Center for Strategic Studies .
URL: https://africacenter.org/spotlight/protests-grow-ahead-of-togo-term-limit-referendum/
Anderson, Christopher J. and Yuliya V. Tverdova. 2003. “Corruption, Political Allegiances, and
Attitudes Toward Government in Contemporary Democracies.” American Journal of Political
Science 47(1):91–109.
Arves, Stephen, Kathleen Gallagher Cunningham and Caitlin McCulloch. 2019. “Rebel tactics and
external public opinion.” Research & Politics 6(3):2053168019877032.
Bayer, Markus, Felix S Bethke and Daniel Lambach. 2016. “The democratic dividend of nonviolent
resistance.” Journal of Peace Research .
URL: http://jpr.sagepub.com/content/early/2016/09/30/0022343316658090.abstract
Bensley, Lillian and Juliet Van Eenwyk. 2001. “Video games and real-life aggression: Review of
the literature.” Journal of adolescent health 29(4):244–257.
Blair, Graeme, C Christine Fair, Neil Malhotra and Jacob N Shapiro. 2013. “Poverty and support
for militant politics: Evidence from Pakistan.” American Journal of Political Science 57(1):30–
48.
Blattman, Christopher. 2009. “From Violence to Voting: War and Political Participation in
Uganda.” American Political Science Review 103(2):231–247.
Carlson, Elizabeth. 2015. “Ethnic voting and accountability in Africa: A choice experiment in
Uganda.” World Politics 67(2):353–385.
Chenoweth, Erica. 2019. Women’s Participation and the Fate of Nonviolent Campaigns: A Report
on the Women in Resistance (WiRe) Data Set. Broomfield, CO: One Earth Future Foundation.
Chenoweth, Erica and Maria J. Stephan. 2011. Why Civil Resistance Works: The Strategic Logic
of Nonviolent Conflict. Columbia Studies in Terrorism and Irregular Warfare.
Cohen, Benjamin J. 1996. “Phoenix Risen: The Resurrection of Global Finance.” World Politics
48(2):268–296.
Dafoe, Allan, Baobao Zhang and Devin Caughey. 2018. “Information equivalence in survey exper-
iments.” Political Analysis 26(4):399–416.
Dahlum, Sirianne. 2018. “Students in the Streets: Education and Nonviolent Protest.” Comparative
Political Studies p. 0010414018758761.
31
Dalton, Russell, Alix Van Sickle and Steven Weldon. 2010. “The individual–institutional nexus of
protest behaviour.” British journal of political science 40(1):51–73.
Fair, C Christine, Rebecca Littman, Neil Malhotra and Jacob N Shapiro. 2016. “Relative poverty,
perceived violence, and support for militant politics: Evidence from pakistan.” Political Science
Research and Methods pp. 1–25.
Flynn, DJ and Megan A Stewart. 2018. “Secessionist social services reduce the public costs of civil-
ian killings: Experimental evidence from the United States and the United Kingdom.” Research
& Politics 5(4):2053168018810077.
Hainmueller, Jens and Daniel J Hopkins. 2015. “The hidden American immigration consensus:
A conjoint analysis of attitudes toward immigrants.” American Journal of Political Science
59(3):529–548.
Hainmueller, Jens, Dominik Hangartner and Teppei Yamamoto. 2015. “Validating vignette and
conjoint survey experiments against real-world behavior.” Proceedings of the National Academy
of Sciences 112(8):2395–2400.
Hillesund, Solveig. 2015. “A dangerous discrepancy: Testing the micro-dynamics of horizontal
inequality on Palestinian support for armed resistance.” Journal of Peace Research 52(1):76–90.
Humphreys, Macarthan and Jeremy M. Weinstein. 2008. “Who fights? The determinants of par-
ticipation in civil war.” American Journal of Political Science 52(2):436–455.
Imai, Kosuke, Luke Keele, Dustin Tingley and Teppei Yamamoto. 2011. “Unpacking the black box
of causality: Learning about causal mechanisms from experimental and observational studies.”
American Political Science Review pp. 765–789.
Johns, Robert and Graeme AM Davies. 2019. “Civilian casualties and public support for military
action: Experimental evidence.” Journal of Conflict Resolution 63(1):251–281.
Kadivar, Mohammad Ali and Neil Ketchley. 2018. “Sticks, Stones, and Molotov Cocktails: Unarmed
Collective Violence and Democratization.” Socius 4:2378023118773614.
Kant, Immanuel and Jerome B Schneewind. 2002. Groundwork for the Metaphysics of Morals. Yale
University Press.
Korsgaard, Christine M and Christine Marion Korsgaard. 1996. Creating the kingdom of ends.
Cambridge University Press.
Kuran, Timur. 1991. “Now out of Never: The Element of Surprise in the East European Revolution
of 1989.” World Politics 44(1):7–48.
Lupu, Yonatan and Geoffrey PR Wallace. 2019. “Violence, Nonviolence, and the Effects of Inter-
national Human Rights Law.” American Journal of Political Science 63(2):411–426.
Marx, Gary T. 1979. External Efforts to Damage or Facilitate Social Movements: Some Patterns,
Explanations, Outcomes and Complications. Vol. The Dynamics of Social Movements Cambridge,
MA: Winthrop pp. 94–125.
32
McAdam, Doug. 1986. “Recruitment to high-risk activism: The case of freedom summer.” American
journal of sociology 92(1):64–90.
Mu˜noz, Jordi and Eva Anduiza. 2019. “‘If a fight starts, watch the crowd’: The effect of violence
on popular support for social movements.” Journal of Peace Research p. 0022343318820575.
Orbell, John M. 1967. “Protest participation among Southern Negro college students.” American
Political Science Review 61(2):446–456.
Palan, Stefan and Christian Schitter. 2018. “Prolific. ac—A subject pool for online experiments.”
Journal of Behavioral and Experimental Finance 17:22–27.
Peer, Eyal, Laura Brandimarte, Sonam Samat and Alessandro Acquisti. 2017. “Beyond the Turk:
Alternative platforms for crowdsourcing behavioral research.” Journal of Experimental Social
Psychology 70:153–163.
Pinckney, Jonathan. 2016. Making or Breaking Nonviolent Discipline in Civil Resistance Move-
ments. Washington,DC: ICNC Press.
Principe, Marie A. 2017. Women in nonviolent Movements. Washington,DC: United States Insti-
tute of Peace.
Regan, Patrick M and Aida Paskeviciute. 2003. “Women’s access to politics and peaceful states.”
Journal of Peace Research 40(3):287–302.
udig, Wolfgang and Georgios Karyotis. 2014. “Who protests in Greece? Mass opposition to
austerity.” British Journal of Political Science 44(3):487–513.
Rule, Brendan Gail and Tamara J Ferguson. 1986. “The effects of media violence on attitudes,
emotions, and cognitions.” Journal of Social Issues 42(3):29–50.
Sapienza, Paola, Luigi Zingales and Dario Maestripieri. 2009. “Gender differences in financial risk
aversion and career choices are affected by testosterone.” Proceedings of the National Academy
of Sciences 106(36):15268–15273.
Schussman, Alan and Sarah A Soule. 2005. “Process and protest: Accounting for individual protest
participation.” Social forces 84(2):1083–1108.
Seligson, Mitchell A. 2002. “The Impact of Corruption on Regime Legitimacy: A Comparative
Study of Four Latin American Countries.” Journal of Politics 64(2):408–433.
Simpson, Brent, Robb Willer and Matthew Feinberg. 2018. “Does Violent Protest Backfire? Testing
a Theory of Public Reactions to Activist Violence.” Socius: Sociological Research for a Dynamic
World 4:1–14.
Smith, Tom W. 1984. “The polls: Gender and attitudes toward violence.” The Public Opinion
Quarterly 48(1):384–396.
Tarrow, Sidney. 1994. Power in Movement: Social Movements, Collective Action and Politics.
Cambridge, MA: Cambridge University Press.
33
Taub, Amanda and Max Fisher. 2019. “The Global Protest Wave, Explained.” The New York
Times.
Tezcur, Gunes Murat. 2016. “Ordinary People, Extraordinary Risks: Participation in an Ethnic
Rebellion.” American Political Science Review 110(2):247–264.
Thomas, Emma F. and Winnifred R. Louis. 2014. “When Will Collective Action Be Effective?
Violent and Non-Violent Protests differentially Influence Perceptions of Legitimacy and Efficacy
Among Sympathizers.” Personality and Social Psychology Bulletin 40:263–276.
Tucker, Joshua A. 2007. “Enough! Electoral fraud, collective action problems, and post-Communist
colored revolutions.” Perspectives on Politics 5(3):535–551.
Verhulst, Joris and Stefaan Walgrave. 2009. “The first time is the hardest? A cross-national and
cross-issue comparison of first-time protest participants.” Political Behavior 31(3):455–484.
Wasow, Omar. 2020. “Agenda Seeding: How 1960s Black Protests Moved Elites, Public Opinion
and Voting.” American Political Science Review OnlineFirst.
Wood, Elisabeth Jean. 2003. Insurgent collective action and civil war in El Salvador. Cambridge
studies in comparative politics New York, NY: Cambridge University Press.
Yablon, Yaacov Boaz. 2009. Gender differences in the consistency of attitudes towards peace. In
Women’s Studies International Forum. Vol. 32 Elsevier pp. 305–310.
34
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
Rebel groups employ a number of strategies beyond violence, and these alternative tactics are often thought to improve the reputation and legitimacy of rebel actors. How powerful states (and their publics) view rebels can affect their chances of international recognition, inclusion in peace talks, and whether they are eventually successful at achieving their objectives. This study employs two experiments to test the link between rebel tactics and opinions of these rebels held by external audiences. We examine the impact of six rebel behaviors on American public opinion: (a) nonviolent demonstrations, (b) nonviolent interventions (such as blockades and sit ins), (c) social noncooperation (such as hunger strikes), (d) terrorism, (e) stone throwing, and (f) the use of local democratic practice (elections) in rebel groups. We find that the use of elections within rebel actors, demonstrations, and hunger strikes improve positive perceptions of rebels, whereas rebel use of terrorism decreases support.
Article
Full-text available
How do international audiences evaluate the legitimacy of secessionist insurgencies? Although secessionists often propagate their behavioral choices, such as state-building and non-violence, to international audiences in the hopes of generating support, scholars know little about the effects of this information. In this article, we use survey experiments in the United States and the United Kingdom to examine how international audiences respond to two commonly used strategies of secession: civilian killings and social service provision. We find that international audiences view secessionists who avoid civilian killings and provide social services as more legitimate than secessionists who kill civilians and do not provide services, respectively. Further, we show that service provision can allow secessionists to reduce—and, in some cases, eliminate—the public costs of civilian killings. These findings have important implications for ongoing secessionist conflicts across the globe.
Article
Full-text available
How do people respond to violent political protest? The authors present a theory proposing that the use of violence leads the general public to view a protest group as less reasonable, a perception that reduces identification with the group. This reduced identification in turn reduces public support for the violent group. Furthermore, the authors argue that violence also leads to more support for groups that are perceived as opposing the violent group. The authors test this theory using a large (n = 800) Internet-based survey experiment with a politically diverse sample. Participants responded to an experimental scenario based on recent violent confrontations between white nationalist protesters and antiracist counter-protesters, allowing the authors to study whether violent protest would reduce public support even when used against a widely reviled group. The authors found that the use of violence by an antiracist group against white nationalists led to decreased support for the antiracist group and increased support for the white nationalist group. Furthermore, the results were consistent with the theorized causal process: violence led to perceptions of unreasonableness, which reduced identification with and support for the protest group. Importantly, the results revealed a striking asymmetry: although acts of violence eroded support for an antiracist group, support for white nationalist groups was not reduced by the use of violence, perhaps because the public already perceives these groups as very unreasonable and identifies with them at low levels. Consistent with this interpretation, the authors found that self-identified Republicans, a subset of the sample that reported less extremely negative views of white nationalists, showed reduced support for white nationalists when they engaged in violence.
Article
Full-text available
The number of online experiments conducted with subjects recruited via online platforms has grown considerably in the recent past. While one commercial crowdworking platform - Amazon's Mechanical Turk - basically has established and since dominated this field, new alternatives offer services explicitly targeted at researchers. In this article, we present www.prolific.ac and lay out its suitability for recruiting subjects for social and economic science experiments. After briefly discussing key advantages and challenges of online experiments relative to lab experiments, we trace the platform's historical development, present its features, and contrast them with requirements for different types of social and economic experiments.
Article
How do stigmatized minorities advance agendas when confronted with hostile majorities? Elite theories of influence posit marginal groups exert little power. I propose the concept of agenda seeding to describe how activists use methods like disruption to capture the attention of media and overcome political asymmetries. Further, I hypothesize protest tactics influence how news organizations frame demands. Evaluating black-led protests between 1960 and 1972, I find nonviolent activism, particularly when met with state or vigilante repression, drove media coverage, framing, congressional speech, and public opinion on civil rights. Counties proximate to nonviolent protests saw presidential Democratic vote share increase 1.6–2.5%. Protester-initiated violence, by contrast, helped move news agendas, frames, elite discourse, and public concern toward “social control.” In 1968, using rainfall as an instrument, I find violent protests likely caused a 1.5–7.9% shift among whites toward Republicans and tipped the election. Elites may dominate political communication but hold no monopoly.
Article
Social movements often face tactic diversification. In otherwise nonviolent movements, some groups or radical flanks may resort to violent actions such as street rioting. This article analyzes the impact that these violent episodes can have on popular support for the movement as a whole. To estimate the causal effect of violence, it exploits an unexpected riot outbreak that occurred during the fieldwork of a face-to-face survey in Barcelona in May 2016, led by a squat group linked to the anti-austerity movement known as the 15-M or indignados that emerged during the financial crisis. By comparing respondents interviewed before and after the riots, it finds that the street violence episode reduced support for the 15-M movement by 12 percentage points on average. However, the magnitude of the effect is highly conditional on the respondents’ predispositions towards the movement. Core supporters, that are expected to share the frame of the movement in justifying violent actions, are the least affected by the violent outbreak. On the other extreme, weak supporters, opposers, and non-aligned citizens reduce their support to a larger extent. Results are robust to different specifications and a wide range of robustness checks. These findings have potentially important implications for movements concerned with broadening their support base.
Article
Under what conditions are individuals more likely to approve of human rights abuses by their governments? While various theoretical expectations have been offered about public approval of repression, many of them have not been directly tested. We analyze the effects of differing opposition tactics, differing government tactics, and legal constraints on approval of repression through a series of survey experiments in India, Israel, and Argentina. Our results indicate that violent action by opposition groups consistently increases support for government repression. In the context of contentious politics, we find that the effects of international law vary by national context. While our respondents in India were less likely to approve of their government when told the government violated international law, the same information likely increased approval of the government in our Israel experiment. The findings provide insights into the microfoundations of existing theories and suggest areas for theory refinement.
Article
Survey experiments often manipulate the description of attributes in a hypothetical scenario, with the goal of learning about those attributes’ real-world effects. Such inferences rely on an underappreciated assumption: experimental conditions must be information equivalent (IE) with respect to background features of the scenario. IE is often violated because subjects, when presented with information about one attribute, update their beliefs about others too. Labeling a country “a democracy,” for example, affects subjects’ beliefs about the country’s geographic location. When IE is violated, the effect of the manipulation need not correspond to the quantity of interest (the effect of beliefs about the focal attribute). We formally define the IE assumption, relating it to the exclusion restriction in instrumental-variable analysis. We show how to predict IE violations ex ante and diagnose them ex post with placebo tests. We evaluate three strategies for achieving IE. Abstract encouragement is ineffective. Specifying background details reduces imbalance on the specified details and highly correlated details, but not others. Embedding a natural experiment in the scenario can reduce imbalance on all background beliefs, but raises other issues. We illustrate with four survey experiments, focusing on an extension of a prominent study of the democratic peace.
Article
This study investigates whether protest movements consisting of students and educated protesters are more likely to (a) use nonviolent rather than violent resistance and (b) successfully reach their goals. Extant literature suggests that education is negatively linked to violent conflict, and the commonly assumed mechanism is that educated groups are less likely to resort to violence. Moreover, many argue that education is a force for regime change and democratization, by inducing successful protest movements. This article is the first to systematically test implications of these mechanisms at the protest level. The empirical analysis builds on original data on the educational background of participants in all protest campaigns aiming for regime change from 1900 to 2006 identified in the Nonviolent and Violent Campaigns and Outcomes Dataset 1.0. I find robust evidence that protest movements with a high degree of involvement by students and graduates are more likely to turn nonviolent. Moreover, there is some (although weaker) evidence that these movements are more likely to achieve their goals, but only due to their nonviolent dispositions. This adds to the literature explaining why some movements resort to nonviolence (and succeed), by establishing that the identity and socioeconomic background of protesters matter.