If Repression Makes Protests Grow, Why do

Governments Repress?

S. Erdem Aytac, Luis Schiumerini, and Susan Stokes

Koc University and Yale University
Word Count: 12,045

We thank Angela Alonso, Kate Baldwin, Chitralekha Basu, Gretchen Helmke, Edgar Kiser,
Melis Laebens, Adria Lawrence, Fernando Limongi, Leonid Peisakhin, Tiago Peterlevitz, Hari
Ramesh, Anastasia Rosovskaya, Milan Svolik, G
ulay T
urkmen-Dervisoglu, Tariq Thachil, Andres
Vargas, Libby Wood, and audiences at Yale and Rochester for comments and assistance. We
are grateful to Yales MacMillan Center for International and Area Studies, and its director, Ian
Shapiro, for financial support. The research contained here was reviewed by the Yale University
Human Subjects Review Committee, approval #1404013794.

Abstract
Governments sometimes respond to protests with aggressive police actions. But this
strategy can be self-defeating, sparking large backlash movements. Why do governments
deploy levels of repression that make protests grow, and how do they adjust, once they face
a backlash? Our answer is that governments may miscalculate actual and potential
demonstrators response to police actions and therefore deploy inefficient levels of
repression, causing a backlash. How they then extricate themselves from the backlash
depends on how secure their hold is on office, conditional on increased levels of repression:
secure government use heightened repression to end the protests, whereas less secure ones
reduce repression and offer concessions. We interviewed police officials, government
authorities, and activists in three new democracies that recently experienced mass protests
Turkey, Brazil, and Ukraine and use original survey data and third-party surveys to
test the implications of our model of government decision-making under uncertainty about
citizens reactions to repression.

A relatively small group of protesters gathers in city streets or squares, demanding

something of the government. The authorities try to break up the protests by sending in
the police, who use less-lethal weapons water cannons, tear gas, pepper spray, sound
bombs, batons, rubber bullets. But rather than breaking up the protests, the police actions
incite others to join. Backlash movements of these kinds have been documented by
journalists and scholars in settings as diverse as the U.S. during the Civil Rights
movement,1 the Eastern bloc countries and Soviet regions in the 1990s,2 and the advanced
democracies facing anti-globalization movements.3 At the time of this writing, the most
recent example is the pro-democracy movement in Hong Kong. The appearance there of
police with tear gas and long-barrel rifles, in the autumn of 2014, seemed to galvanize the
movement, bringing many more people into the streets.
Social scientists have offered a number of explanations for the publics puzzling
response of participating more when repression increases.4 But they have been virtually
silent on a second puzzle: if repression can make protests grow, why do governments
repress?5 That is one question we take up in this paper. We also offer answers to the
questions, What strategies do they use to extricate themselves once they have caused a
backlash movement and what explains variation in extrication strategies?
In the next section we offer a decision-theoretic model of government
decision-making under uncertainty about the publics sensitivity to repression. In the third
section we offer accounts, based on interviews with elected authorities, police officials, and
activists, and on survey data and published materials, about recent backlash movements in
three new democracies, Turkey, Brazil, and Ukraine. The evidence supports our theoretical
prediction that governments unintentionally spark mass uprisings when they underestimate
the publics sensitivity to repression. Once they observe the mass response to early
1

See Chong 1991.

See Opp 1994; Beissinger 2002.
3
See della Porta 2013.
4
See, e.g., Lichbach 1987; Francisco 1995, 2005; Brockett 1995; della Porta and Tarrow 2011; Opp 1989,
1994. For a review, see Davenport 2007.
5
For partial exceptions, see Earl 2003; Davenport 1995; Moore 2000.
2

repression, their decision to increase repression or reduce it depends on the key variable,
conditional security of office, identified in our model. In the fourth section we consider rival
explanations. Regarding the initial missteps, we consider whether, instead of elected
authorities miscalculating, their calculations were accurate but they had incomplete
control over the police. Regarding the strategies that governments used to extricate
themselves from the crisis, we consider whether, instead of differences in the elected
authorities security in office, they reflected distinct levels of centralization or of the
consolidation of democracy, or differences in the ideological orientation of the government,
or in how costly policy concessions would be for governments, or in the class composition of
the protesters. We end with some concluding thoughts about the applicability of our model
to other new democracies and to protests in advanced democracies and in authoritarian
regimes.
In addition to providing general insights, each of these particular recent movements
was important, nationally and internationally. The international press can scarcely
mention the Justice and Development Party (AKP) government and Prime Minister (now
President) Recep Tayyip Erdogan without also mentioning the brutality they deployed
during the 2013 Gezi Park uprising.6 A year after the Brazilian protests, the authorities
were on tenterhooks during the soccer World Cup, fearing a replay of June, 2013. The
EuroMaidan protests set off a sequence of events from the fall of the Yanukovych
government, to Russias annexation of Crimea, to rebellion and civil war in Eastern
Ukraine. We offer new insights into these events with original interviews and survey data.
Our selection of these cases reflects not just their inherent importance but also their
variation on outcomes of interest and on key explanatory variables. The first outcome of
interest is the governments willingness to use mid-level repression against relatively small
protests, and the relevant explanatory factor is its perception of the risk that doing so will
instigate a large backlash movement. Here the variation in causes and outcomes was
6

See O
guzlu 2013.

inter-temporal: all three governments initially perceived the risk as low and then updated
their perceptions upwards, once the backlash had begun. As predicted, they then changed
course. Exactly what strategy they changed to is the second outcome of interest whether
they ramped up repression to a more lethal level or reduced it and offered concessions. In
this case the key explanatory variable (electoral risk conditional on very high repression)
varied, cross-sectionally, across our cases.7

A Theoretical Approach to Governments Strategies toward

Protests
Our model of governments actions draws on our answer to a prior question of why
repression sometimes draws more people into protests. In a separate paper, we note that
when would-be protesters observe others being subjected to rough treatment by police,
they frequently have a bifurcated response. They experience fear and intimidation, but at
the same time anger and moral outrage.8 The fear response drives up peoples costs of
participation, the anger response drives up their costs of abstention. By costs of abstention
we mean the psychic discomfort and social shame people can experience when they stay
away from collective actions, especially when they shares the protesters goals.
Figure 1 illustrates how both kinds of costs can rise as the level of repression
increases. We depict the costs of abstention (A) as rising in a linear way while those of
participation (C) rise at an accelerated rate at high levels of repression. This is the region
in which crowd-control techniques graduate from relatively mild to extreme, e.g., when
rubber bullets are replaced with live ones. Even the courageous and reckless usually depart
7

We do not claim that our theory explains all observed instances of backlash movements. For instance,
though we model such movements as a result of governments underestimating the costs of abstention sparked
by repression, and demonstrate that such a miscalculation rather than police indiscipline lay behind the
actions of the governments we study, in other settings police indiscipline may well cause the backlash. Chong
(1991) documents the latter dynamic in the U.S. Civil Rights Movement. Rather than a universal theory,
we offer one that is internally consistent and the best explanation for the cases we study. At the end of the
paper we indicate that our theory may apply to backlash movements in other settings, as well.
8
Aytac, Schiumerini, and Stokes 2015; see also Opp 1989, 1994, and della Porta 2013.

in reaction to very high levels of repression.9 The curve tracing the vertical distance
between A and C relates the level of repression to the probability of a persons joining a
protest; its form is of an inverted-U.
[Figure 1]
Figure 1 is suggestive of why governments might find themselves unintentionally
igniting mass uprisings: when deploying repressive force, they may be uncertain where they
are on the repression-protest curve. Is the force being applied likely to increase protests or
suppress them? And how much more heavy a hand will be needed before the point is
reached at which intimidation and injury (C) start increasing at a faster pace than anger
(A)?
We model governments as uncertain about the publics response to repression. Once
the government acts, and in particular if it misfires and spawns a mass uprising, it learns
from its mistake and tries to adjust the level of repression. In terms of Figure 1, it might
displace itself toward the right on the horizontal axis, shifting, for instance, from rubber
bullets to live ammunition. Or it might displace itself leftward, to what has been called
negotiated management of demonstrations by the police.10
In the cases under consideration, how governments responded to the backlash
depended on the political price they expected to pay later, were they to undertake very high
levels of repression. The scope condition here is elected governments in new democracies
places where incumbent office-holders would have to face the voters in the future, and
where norms of free assembly may be less entrenched than in advanced democracies.11 The
Turkish government was the most secure among the three we consider. It enjoyed an
electoral support base that was unlikely to punish it for unleashing brutal force against
protesters. After having inadvertently helped spark the Gezi Park uprising in May-June
9

Though, as Lichbach (1987) notes, rather than being completely demobilized by extreme repression,
activists may shift into another form of collective action.
10
McCarthy and McPhail 1998.
11
Though aggressive police tactics toward protesters are by no means absent in advanced democracies;
see Gillham 2011.

2013, the Turkish authorities ended the movement by upping the level of repression
displacing themselves far to the right on the horizontal axis in Figure 1. The Brazilian and
Ukrainian governments found themselves in weaker positions, realistically fearing that their
tenure in office might be shortened should they not retreat to a more conciliatory stance.
In hindsight, both the secure and insecure governments miscalculated in their early
response to protests. The Turkish authorities would have done better, by their own
calculus, by unleashing full brutality before the protests grew.12 Their response to protests
in the post-Gezi Park period indicates that this is the lesson the authorities have learned:
meet any signs of demonstrations with overwhelming force.
The less-secure governments would have fared better had they avoided repressive
measures. Once the Brazilian authorities found themselves in the midst of a mass uprising,
they banned the use of shock troops and rubber bullets and made painful policy
concessions to the protesters. The Yanukovych government tried to retreat from a highly
repressive initial response to protests on the Maidan, but was confounded by problematic
decision-making and collapsed three months after the demonstrations began.

The Model
Our model bears some resemblance to those of Przeworski (1991) and Blaydes and
Lo (2012), in which governments interact with pro-democracy movements but lack full
information about the protesters capacity for mobilization. We model a simplified,
decision-theoretic situation that highlights the governments problem of choosing an
efficient level of repression. Civil society reacts to the governments actions but is not a
strategic actor in our model. The reason is that, in the cases we consider, the governments
dilemma is how to react to highly decentralized choices of a multitude of ordinary citizens,
rather than to the strategies of movement leaders. Our fieldwork suggests that, in these
12

This statement should not be construed as a normative endorsement of heavy-handed repression.

Turkish and international human rights organizations have made strong cases that the government violated
international law and treaties to which it is a signatory in the measures used during the Gezi Park protests.

cases, movement leaders try to elicit certain responses from the government and from the
broader citizenry and they do so with a strategic orientation. But the protest leaders were
often just as surprised by the repression-sparked uprising as was the government, and saw
themselves as having limited ability to shape the actions of the citizenry. As a Turkish
activist told us, We dont know how the protests got that big. If we knew, we would do it
again, immediately.13
Likewise, a movement leader, a veteran of many smaller protests, responded this way
to our question about why Brazils June, 2013 protests grew to be so large: In general, at
least in Brazil, the police arrive, beat people up, and everyone leaves. Sometimes the next
demonstration is bigger but in general when the police decide that they are going to end a
demonstration, they end it. This wasnt the case in the June protests. When we asked her
why the 2013 protests were different, she shrugged and turned the question back to us:
There are some things that are hard to explain... Perhaps researchers can explain it.14
[Figure 2]
In our model, the government makes decisions in two periods (see Figure 2). The
first one is in response to a relatively small movement that demands concessions which the
authorities are loath to make. The government has three choices: make (early) concessions
(C), authorize an aggressive police response (R), or inflict a truly brutal response right
away (R+EARLY ). If the government makes concessions or if it unleashes full brutality, the
protests end.
If the government chooses heavy-handed tactics (R) in the first period, it faces two
possible reactions from the citizenry. Repression might (with probability (1 p)) increase
the costs of participation of active and potential protesters enough that the protests end.
Or it might (with probability p) provoke anger and hence drive up the costs of abstention
13

Interviewed by G
ulay T
urkmen-Dervisoglu and Susan Stokes, Istanbul, July 18, 2014. In what follows,
only public figures who said they were willing to be identified and quoted are named. We do not identify
people who are not public figures or public figures who requested anonymity.
14
Interview with MPL member conducted by Luis Schiumerini and Susan Stokes, ao Paulo, May 26, 2014.

among enough people that the protests grow indeed, they may explode into a full-scale
uprising. In this case, in the second period the government decides whether to scale back
the repression and make late concessions (LC), or scale up repression to a higher level
(R+LAT E ).
Regarding the governments level of security, some governments anticipate that even
if they repress protesters harshly (R+EARLY or R+LAT E ), this action will have relatively
little effect on their tenure in office, while others have to worry that it may leave them
adrift in public opinion and reduce their chances of winning future elections. We call the
former type a secure government and the latter an insecure government, and reflect this
distinction in the costs they incur when applying harsh repression. Secure governments
incur a cost of rs and insecure ones incur ri , and rs < 1 < ri . Thus, in our model, secure
governments always prefer harsh repression (R+EARLY or R+LAT E ) over (early or late)
concessions, while the opposite is true for insecure governments.
In the appendix we show that preference orders at the first decision node depend on
the governments level of security and on the probability it ascribes to early repression
quelling the protests, as summarized in Table 1. The key point is that if the net effect of
repression is to mobilize if early repression pushes the costs of abstention higher than the
costs of participation and enough people join the protests then secure governments
should repress more harshly early on and insecure ones should make early concessions. In
neither case does an uprising occur. Uprisings occur, our model suggests, if the authorities
misconstrue the publics sensitivities.
[Table 1]
Readers might wonder about the secure government: How can it instigate a major
uprising but not feel threatened by mass opinion at the next election? The answer goes to
the structure of opinion in some polities. When mass opinion is sharply divided across a
single, overlapping dimension, and when protests draw people from one side of this divide,
the anti-government segment of the polity may be so outraged by the governments actions
7

that many bear heavy costs of abstention and join in. But repression does not erode
support from among the pro-government silent majority indeed, these sectors may
embrace the government all the more enthusiastically when they see it challenged by
riotous fellow citizens whose cause the governments supporters find anathema.
The Turkish and Brazilian cases illustrate the difference between secure and
insecure governments. As we will show later with polling data, virtually no government
supporters joined the Gezi Park protests in Istanbul. In Turkey, reactions to the protests
and to the governments handling of them were sharply structured by party affinities.
Therefore few government supporters were likely to turn against the government or ruling
party in the wake of the crackdown. By contrast, in Brazil, party affinities are weak and
they did not strongly structure peoples attitudes toward the protests or toward the
governments handling of them. We will show that support for the protests ran high among
partisans of all the parties in power, all of which saw themselves as running electoral risks
if they did not dial down the repression and quickly resolve the crisis.
Ukraine approximates Turkey more than Brazil in the depth and unidimensionality
of the divide between pro-Western and pro-Russian populations; the EuroMaidan protests
drew mainly from the pro-Western segment of the populace. But there was no chance of
governing the country in the midst of a mass upheaval in Kiev and other Western cities.
Yanukovych did not sit atop a silent majority, like Erdogans, that would give him the
freedom to end the uprising with a ferocious crackdown. His government therefore chose an
extrication strategy more like the Brazilian one. The Ukrainian government implemented
the strategy less deftly than was the case in Brazil.

Repression, Uprisings, and Extrication in Three New Democracies

Turkeys Gezi Park Protests
Overview. In Istanbul, in May-June, 2013, harsh police repression of a small group
of activists was followed by massive, national protests. Within four days of the initial
repression, hundreds of thousands of people had gone to the streets. Faced with the
uprising, Prime Minister Erdogan ratcheted up the level of repression to a point where the
protests finally subsided. The prime ministers implacable leadership style, as well as his
desire not to repeat past experiences of protests, contributed to this strategy of
extrication-through-increased-repression (R+LAT E ). But what allowed him to pursue this
course was the confidence that his conservative and devout voting base would not punish
him at subsequent elections for the harsh police tactics. Since the Gezi Park uprising the
government, now well aware of the risk that mid-level repression will instigate a backlash,
has dealt extremely harshly with protesters, and the AKP has been victorious in municipal
(March 2014) and presidential (August 2014) elections.
Phase 1 (May 28-30): Early repression. In the early morning of May 28, 2013,
a force from the municipal police (Zabta) used pepper spray, tear gas, and water cannons
to try to dislodge members of the Taksim Solidarity Coordination Committee out of Gezi
Park. Located at the center of Taksim Square, the park had been occupied by activists
trying to block its destruction, in particular the felling of its sycamore trees a rare green
area in central Istanbul. The planned demolition was to make room for a shopping mall,
housed inside a reconstruction of an Ottoman-period barracks. The scene that morning
was the source of a Reuters photograph that would come to be known as the Woman in
Red a young woman, garbed in a red dress, standing still while a police officer about
four feet away aimed a stream of pepper spray at her face. The image, circulated widely on
the Internet, was among the most notorious of the protests, and seemed to draw many
people to the park.

Prime Minister Erdogan made very clear that there would be no concessions to the
protesters. On May 29, he announced, regarding the Taksim building project, we have
already decided; we will go ahead whatever they do.15
Phase 2 (May 31-June 1): Uprising. A second phase is marked by a sharp
increase in the number of protesters and ends with the police beating a tactical retreat
from Taksim Square. In the early morning of Friday, May 31st, the police again raided the
park, and again used tear gas, pepper spray, and water cannons against the protesters.
Now the national police had joined the Zabta, and the police actions intensified. They set
up barricades on the streets surrounding Taksim Square to keep protesters out. They
sprayed so much tear gas that scores of canisters covered the streets around Taksim, as well
as in the Besiktas district and elsewhere in Istanbul. Among the injured was an opposition

member of parliament, Srr S

ureyya Onder,
who was hospitalized after being hit in the
shoulder with a tear-gas canister.
This was the moment when the protests began to swell to massive proportions.
Estimates put the number of demonstrators on Istiklal Caddesi, a pedestrian avenue that
ends at Taksim, at around ten thousand by Friday afternoon, May 31. The protests grew
as word of police actions spread through social media. The number of tweets per day that
included the word eylem (protest) surged to around 50,000 on May 31, to around 90,000
on June 1, and remained above 30,000 until June 8 (Janys Analytics 2013:5). On June 6-8,
a survey research team asked more than 4,000 people in the park, What was the most
important reason for you to join the protests? The most frequent answer, offered by
nearly half of respondents, was seeing the repressive acts of the police.16
The intensity of the police actions rained down international condemnation on the
Turkish authorities and created some tensions within the government and between the
government and the business community. Yet Erdogan remained intransigent. In a June 1
speech, he threatened to meet popular force with popular force: where they gather 20, I
15
16

See H
urriyet newspaper, May 30, 2013.
See Konda (2013).

10

will get up and gather 200,000 people. Where they gather 100,000, I will bring together
one million from my party.17 Using a derogatory term (capulcu) he would repeat many
times to describe the protesters, the prime minister declared on June 2, I will not seek
permission from hoodlums to implement my plans for Taksim.18
Phase 3 (June 2-June 10): Gezi without the police. Having failed to contain
the protests or stop their spread, on Saturday, June 2, on orders of the interior minister,
the police retreated from Gezi Park. As the quotes from the prime minister just cited
suggest, this was a tactical retreat, aimed at refreshing the beleaguered police forces and
changing the dynamics of public opinion.
Phase 4 (June 11-15): Negotiations and heightened repression. The Gezi
Park protests came to an end with batons, torrents from water cannons and enough tear
gas to keep the air around Taksim toxic and discolored for most of the day. On the
morning of June 11, police entered Taksim, took down protesters barricades, and forced
them from the square; most demonstrators dispersed or retreated into the park. The police
returned, on June 15, with a massive deployment and show of force that cleared the park
and ended the protests.
In between, on June 13-14, the prime minister held talks with Taksim Solidarity
leaders. These discussions seemed unlikely to stave off the coming Armageddon. In
contrast to Brazil, where as we shall see the government made unilateral concessions,
the Turkish government offered to soften its stance if the movement first dispersed. The
Taksim Solidarity leaders were not in a position to end the protests. They could have
encouraged a retreat, but there was no guarantee that the demonstrators would have
complied. And the offers made by the prime minister were not credible, since they relied on
future decisions of judges or on referendums.19
On June 15, the police entered the park with a massive show of force. Ahmet Sk, a
17

Radikal newspaper, June 2, 2013.

Aksam newspaper, June 3, 2013.
19
See the account in H
urriyet Daily News, June 13, 2013.
18

11

prominent Turkish journalist, told us that he had covered several war zones in his career
but had never faced a scene as frightening as the one he encountered on June 15 in Gezi.
Sk took refuge in a building nearby and dared not venture out for many hours.20 A medical
doctor who helped organize the care of injured protesters noted that not until the renewed
attacks of June 11 and 15 did they see widespread evidence of the use of batons and of
beatings by police.21
In the end, the police response to the protests left a grim toll. According to the
Turkish Medical Association, by mid-July, 8,000 people had been injured at the scenes of
demonstrations, 61 of them seriously. Eleven people lost eyes after being hit tear gas
canisters. Four civilians and one police officer were killed at the site of demonstrations.22
Two more demonstrators died later of injuries sustained at the protests. Exposure to
chemical toxins also caused numerous asthma attacks, was believed to be linked to several
heart attacks, and may have induced chronic reactive airwave conditions in protesters after
repeated exposure.23 Medical personnel reported treating numerous cases of burns,
apparently from chemical irritants added to the water shot from cannons.
Aftermath in Turkey. The park remained closed to the public until early July,
2013. Since then, the police have maintained a constant presence in Taksim Square, with
the now-defunct Atat
urk Cultural Center next to it serving as a de facto police station.
Political organizations have been vigilantly kept out of the park. In March 2014, upon the
death of Berkin Elvan, a 15-year-old boy who was injured during the protests and had been
in a coma for nine months, the police used tear gas and water cannons to keep the crowds
attending his funeral out of Taksim.

20

Interviewed by T
urkmen-Derviso
glu and Susan Stokes, Istanbul, July 11, 2014.
Interviewed by T
urkmen-Derviso
glu and Susan Stokes, Istanbul, July 18, 2014.
22
One protester was shot in the head by a police officer with live ammunition, one was beaten to death,
and a third sustained head injuries. See Amnesty International (2013, 15).
23
Detailed descriptions were offered by a medical doctor who treated the injured in a make-shift
infirmary. Interviewed by T
urkmen-Dervisoglu and Susan Stokes, Istanbul, July 18, 2014. See also Amnesty
International (2013, 15).
21

12

Understanding the Governments Actions

The miscalculation. The Gezi Park protests were a major crisis for the Erdogan
government, provoking internal divisions in the AKP, criticisms from some wealthy and
influential actors who normally side with it, and inviting lasting derision from the
international press and foreign governments. This was a self-inflicted wound; yet the
miscalculation was easy to make. An uprising on this scale was a once-in-a-generation
event in Turkey, if that. As one activist put it, Gezi was our 68.24 The phrase of the
Taksim Solidarity activist cited earlier resonates: we have no idea what made it so big.
The evidence that police actions played a crucial role in transforming a protest into
a mass uprising is strong,25 and this is the widely accepted view in Turkish public opinion,
even among those who saw the repression as justified.
Extrication. A common perception is that the harsh measures and
uncompromising language with which the protests were met reflected the personality and
leadership style of Prime Minister Erdogan. Indeed, Erdogans default tactic in many
situations is to inflame the passions of his base by insisting on the illegitimacy of his
opponents. This tactic has yielded much success, and Erdogan has stayed on top of
Turkish politics longer than any leader since Atat
urk. Others attribute the ferocity of the
June 11th and 15th crackdowns to an earlier episode of prolonged protests which Erdogan
perceived as a challenge to his authority and policies.26
Still, many political leaders would prefer to deal harshly with opponents and to
bend reality to suit their ambitions and policy goals. What gave Erdogan the freedom to
respond to the Gezi Park protesters so harshly was the near-certainty that repression
would have no serious electoral repercussions. This was a level of security of office that the
other governments we consider lacked. Where Erdogan could be confident that his
24

Interviewed by T
urkmen-Derviso
glu and Susan Stokes, Istanbul, July 13, 2014.
Elsewhere we offer observational and experimental evidence that repression increased many peoples
propensity to protest (Authors, 2015).
26
These were protests by dismissed workers from TEKEL, a privatized state enterprise, which took place
in Ankara in late 2009. See Todays Zaman, March 14, 2014.
25

13

constituents support would not waver when their government dealt harshly with
protesters, others whose actions we consider had to worry about an electoral backlash.
One indication of Erdogans greater insulation from this backlash can be seen in the
contrasting partisan compositions of the Turkish and Brazilian protests. Almost none of
the many thousands of protesters who flooded the Taksim Square area (or who protested in
elsewhere in Turkey) were supporters of the ruling party, the AKP. In Brazil, protesters
included a sizeable minority of supporters of the ruling parties (the Workers Party
Partido dos Trabalhadores, PT nationally and in the city of Sao Paulo, and the Brazilian
Social Democratic Party Partido da Social Democracia Brasileira, PSDB in the State of
Sao Paulo). These contrasts are displayed in Figure 3.
[Figure 3]
In sum, at the moment corresponding to the first node in our decision tree, the
Erdogan government underestimated the mobilizing effect of everyday levels of police
repression against protesters, and undersupplied violence. Once it had a chance to observe
the actual result of mid-level repression, it upped its estimate of p and changed course. At
the second node, its insulation from future electoral risk should it resolve the crisis through
more force allowed it to do just that. It escalated the level of repression to the point at
which the costs of participation overtook those of abstention for the vast majority of
would-be demonstrators.

Brazils June of Fury

Overview. The largest street demonstrations that Brazil had experienced in two
decades began in June, 2013, after a small organization took to the streets to oppose
increases in public transportation fares. The perception of scholars who have studied the
protests is that they went from small to massive after the public became aware of excessive
use of force by the police. Alonso and Mische (2014, 8) write that the Brazilian scenario

14

was like others in which a disproportionate police response . . . captured on social as well
as mainstream media sources, provokes indignation and anger among a broader swath of
the population and generates a scale shift [in] the movement.
The parallels with Turkey (and Ukraine), where state violence also encouraged a
transition from small protests to massive demonstrations, are striking. But the authorities
extrication strategy was remarkably different from that pursued in Turkey. Rather than
upping the level of repression, the Brazilian authorities stopped using rubber bullets, sent
Shock Troops back to their barracks, and made a key policy concession.
The explanation for this divergence lies in contrasting levels of polarization and
dimensionality of politics in the two countries, which produced a greater risk of electoral
backlash, should the Brazilian authorities have attempted to up the level of repression. In
Turkey, government supporters were unlikely to turn against the ruling party no matter
how harshly it treated the protesters. The Brazilian protests tapped into a more complex
set of political identities. Many protesters had weak or non-existent affinities with any
political parties; the proportions of those supporting the nationally ruling PT were roughly
identical among protesters and non-protesters. And, in Brazil, opposition political leaders,
in power in some state and local governments, were just as reviled in public opinion as the
PT for their early handling of the crisis. None of the parties felt shielded by a large group
of die-hard supporters whose opinions would be insensitive to ending the crisis with all-out
police aggresssion.
These differences come across in Tables 2 and 3. Table 2 indicates the powerful way
in which partisanship structured Turks opinions of the Gezi Park protests. When asked
whether the protesters were making legitimate demands or represented a foreign plot
against Turkey (a claim often repeated by the prime minister), the overall response tracks
the pro-government/anti-government split in the country. Among supporters of the ruling
AKP, those who believed in the foreign plot outnumber non-believers by about eight to
one. The proportions are reversed among supporters of the main opposition party, the

15

Republican Peoples Party (CHP). By contrast, in Brazil partisanship played a negligible

role in shaping responses to the protests, or the governments handling of them. Table 3
shows similar opinions of the Brazilian protests among supporters of the leftist PT (in
power nationally and locally) and of the center-right Brazilian PSDB (in power in the Sao
Paulo state government).
[Table 2 here]
[Table 3 here]
Phase 1 (June 3-12): Early protests and initial repression. In Sao Paulo, a
small protest took place on June 3, 2013, in opposition to recently announced public
transportation fare increases. The protests were not a spontaneous response of low-income
riders. They were part of the regular strategy of the Free Fare Movement (Movimento
Passe Livre, MPL), a small organization comprised mainly of graduate students from the
University of Sao Paulo, a leading public university.
At the outset of the protests, Sao Paulos elected civilian leadership was univocal in
its support of a tough police response. On June 8, Mayor Haddad, from the leftist PT,
declared that the protesters rejected the democratic rule of law.27 The day before, June
7, the governor of the state of Sao Paulo, Geraldo Alckmin, had called the protests
absurd; on June 12th, he called the protesters hooligans and vandals.28 Alckmin is
from the center-right PSDB. The federal Minister of Justice, Jose Eduardo Cardozo (PT),
also condemned the protests on June 12, saying, Unfortunately, we have seen this taking
place in Sao Paulo. I think its absurd, this is not the way that a demand will be met.29
The traditional paulista press also called for strong police action. On June 8, an
editorial in the newspaper Estadao lamented, The authorities in the area in charge of
public security should have demanded more rigorous police action from the beginning of
27

Cited in Locatelli (2014, 10).

See Estad
ao, June 7th, 2013; and Estad
ao, June 12nd, 2013.
29
See Estad
ao, June 12, 2013.

28

16

the protests.30 A geograhical point of contention was the Avenida Paulista, a central
artery where protesters wanted to march and the authorities wanted to keep open for
traffic. On June 12, the Folha de Sao Paulo, in an editorial entitled Retake the Avenida
Paulista, demanded, it is time to put a full stop to this. The municipality and the
Military Police have to enforce the restrictions on protests in the Avenida Paulista.31
Phase 2 (June 13): The crackdown. Heeding these calls for strong measures,
the Minister of Public Security of Sao Paulo State, Fernando Grella Vieira, authorized the
use of Military Police Shock Troops (Tropas de Choque), a specialized force that deals with
protests and crowd control.32 The key moment of repression came on June 13, 2013. That
day was for Sao Paulo what May 31 was for Istanbul and November 30, 2013 for Kiev. As
in these other cities, police actions in Sao Paulo would be self-defeating. Images of masses
of Shock Troops aiming tear gas canisters and rubber bullets at unarmed civilians took the
public aback. Photos and videoclips of injured and frightened protesters sailed through the
Internet and social media, and shifted opinion strongly in sympathy with the protesters. A
survey conducted on June 13 by Datafolha, just before the demonstration and the news of
police violence, reported 55% support for protests. A second Datafolha survey, conducted
on June 18, found support had risen to 77%.33
Having the day before called for stronger police actions, the newspaper Estadao
decried the actions of the police: Bombs and rubber bullets were shot without restraint.
Policemen shot even when they were caught on camera by newspapers and TV. Journalists
were injured, in addition to more than one hundred demonstrators. The cowardice and
excesses by the police, shown time and again in the Internet and TV, changed the game.34
The medias attitude toward the demonstrations shifted after June 13 not least
30

See Estad
ao, June 8th, 2013.
See Folha de S
ao Paulo, June 13th, 2013.
32
The Military Police are not connected to the Brazilian armed forces. They are the main police force of
Brazilian states.
33
See Estad
ao, June 14th, 2013 and June 20th, 2013. The number of respondents in the first survey 815,
in the second one, 805. Both sampled S
ao Paulo residents aged 16 or older.
34
See Estad
ao, June 18, 2013.
31

17

because several journalists were injured and arrested. A much-viewed image from that day
was of a 26-year-old TV Folha reporter, Giuliana Vallone, sitting on a curb, her face
bloody and her right eye swollen shut, having been hit in the face with a rubber bullet.
The civilian authorities also began to sound a new note. Justice Minister Cardozo,
having called the protests absurd on June 12, declared on June 14: Beginning
yesterday, we had a situation that we cannot, evidently, accept. . . . [T]here were
situations of police violence that I consider unacceptable. I dont think its correct for the
police to treat people as the images showed yesterday.35
In the aftermath of the crackdown, there followed a war of narratives and images
between the police and protesters, a war that the police were still fighting one year later.
When two of the authors arrived at the headquarters of the Central Area Command of the
Military Police on May 26, 2014, they were greeted by a Lt. Colonel who immediately
pulled up a photo on his iPhone which showed him in the June 2013 protests, bleeding
above his lip. He then showed the interviewers the scar that the wound had left. His
superior, Colonel Celso Luiz (who moved into this position after the protests), started our
interview by playing video clips on his desktop computer of acts of vandalism during the
protests people smashing ATM machines and store windows, a car on fire.
But the police had difficulty winning the war of images and narratives. The
asymmetry in the level of weaponry worked against them. So did their long-standing
reputation for excessive force (see Brinks, 2003). One Military Police colonel whom we
interviewed offered an historical explanation for distrust of the Brazilian police: We
supported the Getulio regime, it was a dictatorship; we gave support to the [1964-89]
military dictatorship. They remember because they are historians, they are social
scientists, they are sociologists, law students. We want to turn the page but they remember
the page.36
Phase 3 (June 14-20): Concessions and reduced repression. The June 13
35
36

See Estad
ao, June 14, 2012.
Interviewed by Schiumerini and Stokes, Sao Paulo, May 26, 2014.

18

actions were followed by a change in strategy of the civilian authorities, who now reined in
the police. Governor Alckmin announced that rubber bullets would no longer be used. At
a press conference on June 16, the Minister of Public Security Grella ruled out further
deployment of Shock Troops.
The police recognized that civilian authorities were bending to popular sentiment.
But despite misgivings, police officers at the protests mainly complied with the mandated
shift to a more passive policing of the demonstrations. Soon after, the elected authorities
made a key concession on bus fares. On June 18th, less than a week after the crackdown,
Mayor Haddad announced that transit fares would revert to their earlier level. The
protests peaked two days later and then subsided.
Rather than persisting, as Erdogan did, in questioning the legitimacy of the
protesters, the Brazilian authorities abandoned the language of
protesters-as-lawless-hooligans and shifted to one that exalted free speech and collective
action as essential for democracy. President Dilma Rousseff (PT) made favorable
statements, if belated ones, about the protests. In a speech on June 18, for instance, the
president said, Today Brazil woke up stronger. The greatness of yesterdays
demonstrations proves the energy of our democracy, the strength of the voice from the
street, the civility of our population.37
Aftermath in Brazil. Once the crisis had passed, there were signs that the
government feared repression leading to larger protests, and tried to adjust strategies. No
individual police officers were prosecuted for excessive use of force. But after the protests,
some Military Police leaders were shifted out of central urban districts and moved to
locations where protesters would be less likely to gather. And in some instances, new
leadership was brought in to places that had been protest hot-spots (such as Colonel Celso
Luiz, mentioned earlier).
Some changes in police procedures suggested that lessons had been learned and
37

See Estad
ao, June 18, 2013.

19

attempts were being made to institutionalize them. The crisis instigated an expansion of
therepertoire of policing strategies used by the Brazilian police.38 Police officials whom
we interviewed credited Celso Luiz with introducing non-repressive crowd-control methods.
A new tactic deployed before the 2014 World Cup was to try to stop protests from
happening, in advance. The Civil Police of Sao Paulo undertook sweeps of anarchist
organizations. Protests did take place, in all host cities, though they were generally small.
The police were heavily armed and on occasion used force.39

Explaining the Brazilian Governments Actions

The miscalculation. There can be little doubt that the civilian authorities
ordered strong police measures in Brazil not anticipating that the result would be a mass
uprising. Like the MPL organizer quoted at the outset, there was little reason for them to
expect other than that the police would arrive, beat people up, and everyone leaves. Our
interviews with police officers and officials made clear that they, too, were taken aback by
the scale shift in the dimensions of the protest.
A case can be made that civilian authorities in Brazil, more than their counterparts
in Turkey or Ukraine, failed to anticipate not just the effect of police brutality on the
public, but the nature and extremity of that brutality. Yet, often the real surprise was not
that the police were so brutal but that their brutality was so public. The Estadao
editorialists dismay was that policemen shot even when they were caught on camera by
newspapers and TV.40 In the new world of omnipresent cameras and immediate
production and diffusion of images, there was little chance of remaining ignorant of
extreme police actions.
Extrication in a federal system. Once the police had acted and the public had
reacted, the authorities chose to reduce the level of repression, as we have seen, and to
38

See della Porta and Tarrow 2011

See, e.g., Protesto termina em confronto no Rio. Estad
ao, June 16, 2014.
40
op cit., Estad
ao, June 18, 2013, emphasis added.

39

20

make a key, and painful, policy concession. Our contention is that electoral sensitivities
concerns about the impact of persistent movements and police repression on incumbents
future electoral prospects encouraged the Brazilian authorities to pull back the police;
whereas the Turkish authorities were relatively insulated from these pressures. Anticipating
the aftermath of the protests, one official told us, The [2014] World Cup generates political
interest. What happens after the World Cup? Elections. For the state government, for the
Brazilian presidency. Everyone is worried. If the police act, [the civilians] can lose votes; if
the police dont act, they can also lose votes. Thats the dilemma.41
But another salient difference between the two countries is the decentralization of
the Brazilian system and, linked to this, the division of political leadership among
competing political parties. In Turkey, though the opposition CHP controls some
municipal governments, the structure of government is centralized and the AKP controls
most important offices and elected bodies. Adding to this centralization is the towering
figure of Erdogan himself.
Does this greater decentralization explain Brazils more conciliatory extrication
strategy? For instance, was strategy the result of bargaining among key political actors at
different levels of government? We find little evidence that this was the case. What is
striking, instead, is the shared sense of crisis that the movement instilled. Mayor Haddad,
his counterparts in other cities (such as Eduardo Paes, the mayor of Rio de Janeiro, from
the Brazilian Democratic Movement Party, Partido do Movimento Democratico Brasileiro
PDMB), governors and public security ministers from Sao Paulo and other states, and
the federal authorities; all were sent scrambling, post-June 13, to reverse course on the
demonstrations. And, as Figure 4 suggests, as of June 18 in Sao Paulo, all of the relevant
political leaders had reason to fear a public-opinion backlash from their perceived handling
of the protests.42
41

Interview with Colonel Morelli, Commander of Sao Paulo Military Police in Sorocaba. Conducted by
Schiumerini and Stokes, Sorocaba, May 27, 2014.
42
Mayor Haddads approval was statistically worse than President Rousseffs and Governor Alckmins,
but the difference is small.

21

[Figure 4]
In sum, the Brazilian authorities in June 2013 underestimated the mobilizing effects
of tough police measures against protesters. Once this effect was revealed to them, and
given incumbents fears of electoral consequences, they reduced the level of repression and
conceded to the protest organizers key demand.

Ukraines EuroMaidan Protests

Overview. In 2010, Viktor Yanukovych became the fourth president of
independent Ukraine, six years after the Orange Revolution had removed him and his
Party of Regions from power. Yanukovych and his entourage had come up through the
Soviet system in Eastern Ukraine; he had served in the early 1990s as governor of the
Donetsk oblast. As a political leader in independent Ukraine, Yanukovych maintained
strong links with Russia. But in 2012, dissatisfied with the terms Vladimir Putin was
offering for Ukraines entry into a Eurasian customs union, Yanukovych entered talks about
a possible Association Agreement with the European Union. This possibility was greeted
with enthusiasm in Western Ukraine.
Yet negotiations with the EU were also difficult, and on Friday, November 29, 2013,
Yanukovych left an Eastern Partnership summit in Vilnius without signing an Association
Agreement. An estimated 10,000 protesters came into central Kiev, both to Maidan or
Independence Square and to the nearby European Square.
The brutal repression of a small, lingering crowd in the Maidan, early the next
morning, November 30, set in motion a series of events that ended three months later in
the fall of Yanukovychs government. So surprising and counter-productive were the police
actions that they spawned a number of conspiracy theories, involving, for instance, a covert
role for the Russians or a plot by some in Yanukovychs inner circle to discredit other
insiders. Yet we have seen in other settings that this kind of mistake is not hard for
governments to make. What stands out in the Ukrainian case is the clumsiness of the
22

governments subsequent efforts at self-extrication from the repression-induced crisis.

Recognizing the error it had made, the authorities instincts were to restrain further police
actions, as the Brazilians had done, rather than ramping up the public repression of
protesters, as in Turkey. But they repeatedly took actions that reenergized the movement:
allowing police forces to appear near the Maidan or tolerating discrete forms of brutality
against activists that subsequently became widely publicized. Activists, who had the
difficult task of keeping the movement alive from Sunday to Sunday over a period of many
weeks, claim that the government came to their rescue, time and again. As one protest
leader put it, Maidan was waning every time that it was left alone.43
Phase 1(November 21-December 1, 2013): Early repression. Early
Saturday morning, November 30, about 1,000 people remained from the previous days
protests in the Maidan. The sound equipment that organizers had used was being taken
down and workers had arrived to put up the squares traditional Christmas tree. At 4:30
a.m., the Berkut, a special police force under the direction of the Ministry for Internal
Affairs, spilled into the square. With batons and boots they beat everyone they could find
student protesters, municipal workers, visitors, and journalists and pursued those who
fled down nearby streets. Thirty-five people required medical attention, among them a
cameraman and a photographer, both from Reuters; thirty-six people were arrested.44
If the aim of the attack was to clear the square and end the protests, the result was
the opposite. Estimates for the size of the crowds who choked the Maidan the following
day Sunday, December 1 ran as high as 800,000.45 Certainly the crowds dwarfed those
that had appeared in support of the Association Agreement the previous weekend. People
close to the government and in the parliament agreed that the turnout was massive and
43

Interview with Ihor Lusenko, by Leonid Peisakhin and Anastasia Rozovskaya, Kiev, June 28, 2014.
See BBC, December 1, and Kiev Post, November 30.
45
Vitaliy Portnikov, a freelance journalist, offered this figure in an interview (June 26, 2014). RT cites a
figure of 700,000 (RT, December 1). In an interview conducted in Kiev on July 2, 2014, by Leonid Peisakhin
and Anastasia Rozovskaya, the editor of a major newspaper (who preferred to remain anonymous) said that
intelligence experts put the number on December 1 at more than 700,000. The Kiev Post cited estimates
of 350,000 to 500,000 protesters.
44

23

that the government was surprised.46

Public opinion polls confirm that many protesters came to the Maidan out of anger
about the Berkuts actions. On December 7-8, 2013, the Kiev International Institute of
Sociology conducted a poll of a convenience sample of about 1,000 protesters on the
Maidan. They asked, Why are you here? The modal response was because of the brutal
beating of demonstrators at the Maidan on the night of November 30. This was the
response of 66% of the more than one thousand respondents.47
Phase 2 (December 2-25, 2013): Mixed signals about repression and
concessions. In the days following the November 30 Berkut attacks in the Maidan and
the huge December 1 rallies, the Yanukovych government, newly sensitive about possible
backlash from police repression, gave signs that it would back down: lessons had been
learned. On December 2, the chief of the Kiev police, Valeriy Koryak, resigned. The
Minister of Internal Affairs, Vitaly Zakharchenko, declared publicly that riot police
abused their power, and promised an investigation.48
On December 11, the Interior Ministry again sent the Berkut into the area around
the Maidan, with orders to remove barricades but not to touch the demonstrators. In
public statements, Interior Affairs Minister Zakharchenko made clear that this was not a
return to the repression of November 30: I want everyone to calm down. There will be no
storming of the square. No one will violate your rights to protest peacefully, but do not
ignore the rights ... of other citizens.49 The security forces mainly complied with the
governments orders for restraint. Tetiana Chornovol, a journalist, activist, and harsh critic
of the Yanukovych government, explained in an interview that the police were given
orders not to hurt people. And [the opposition] exaggerated . . . they said people were
46

As told to Leonid Peisakhin and Anastasia Rozovskaya in Kiev by Portnikov and Bohdana Babych, in
separate interviews conducted on June 26, 2014.
47
See Aytac, Schiumerini, and Stokes 2015 for details.
48
Though he also told state television that if there are calls for mass disturbances, then we will react to
this harshly. (CNN online, December 2, 2013, Pro-EU Protesters in Kiev Call for Governments Ouster.
Available at http://edition.cnn.com/2013/12/01/us/ukraine-eu-protest-sunday/).
49
Quoted in Shaun Walter, The Guardian newspaper December 11, 2013, italics added.

24

being killed. No one was killed, some people were beaten, but [in general] the police acted
very peacefully at that moment.50
Yet the wound left from the November 30 attack was too raw. As Chornovols words
suggest, it was easy for protest organizers to send texts and Tweets saying that the Berkut
had again set upon the demonstrators. Along with the messages came images of phalanxes
of officers, holding three-quarter-length body shields, their faces hidden behind protective
visors. As activist and radio broadcaster Vitali Pornikov explained, all this looked horrible
when people saw the picture. The Berkut did not beat, but they looked ominous.51
The police actions were counter-productive. People streamed back into the square
and barricades were quickly re-erected. The Guardian reported that With the return of
the Berkut, the protests were reenergized.52
But soon after the December 11 events, organizers faced the same problem of
waning crowds. At this point they were helped by a different kind of repressive action: the
abduction, beating, and in one case killing of movement leaders. The first such abduction
was on December 25, when men in a luxury car forced Tetiana Chornovol off the road, near
the Kiev airport. They beat her brutally and left her for dead. After she reappeared, the
traditional and social media displayed graphic photos of her injuries, which seemed to have
an effect like those of Gezi Parks Woman in Red or Sao Paulos Giuliana Vallone. In the
view of many observers, a flagging Maidan movement was again revived. The following
Sunday, December 29, the Maidan crowds swelled to 50,000 people, an increase over
previous days, which the press attributed to the attack on Chornovol.53
The movement ebbed and flowed for two more months, until Yanukovych fled to
Russia on February 21, 2014, denouncing the coup that ousted him from power.
Aftermath in Ukraine. Within a week of Yanukovychs departure, the interim
interior minister announced that the Berkut would be disbanded. A small contingent of
50

Interviewed by Leonid Peisakhin and Anastasia Rozovskaya, Kiev, June 25, 2013.
Interviewed by Leonid Peisakhin and Anastasia Rozovskaya, Kiev, June 26, 2014.
52
Shaun Walter, The Guardian, December 11, 2013.
53
BBC Online, December 29, 2013.
51

25

protesters remained encamped in the Maidan through early May, 2014, at which point a
combination of face-to-face persuasion by erstwhile activist, now mayor Kitschko, and
gentle police actions finally cleared the square.

Explaining Government Actions in Ukraine

The initial miscalculation and failed extrication. The decision to send the
Berkut into Maidan Square in the early morning hours of November 30, 2013, with
instructions to use harsh measures and break up the protest, was universally viewed, after
the fact, as a blunder. But we have seen that such mistakes are not uncommon. What is
surprising in the Ukrainian case is not the initial police actions and the backlash
movement, but the ham-handedness with which Yanukovychs inner circle attempted to
resolve the crisis. In general outlines, the extrication strategy was supposed to be like the
Brazilian one, featuring a retreat from harsh police repression. Yet where the Brazilian
authorities basically got things right in the second phase, the Ukrainian ones lurched from
misstep to misstep. They allowed themselves to appear to be undertaking another round of
repression by sending the Berkut back into the Maidan on December 11, though their
intention was to avoid violence against demonstrators. And while the government sent
signals that it would eschew repression, thugs, presumed by many to have links to the
government, abducted high profile activists, including Chornovol, Automaidan founder
Dmytro Bulatov, and civil society leader Ihor Lutsenko.54 Presumably these violent acts
were meant to be discrete, just to intimidate opposition leaders. But the social media make
it hard to repress discretely, once repressive actions have the publics attention.55
With so much at stake, what explains this clumsiness? In general terms, the
Yanukovych governments decision-making was hampered by an inner circle of advisors
who filtered out unwanted information. It was also hampered by decision-makers mistaken
beliefs about the motivations of the protesters, such as that the protesters were paid and
54
55

See Amnesty International, January 31, 2014.

In Ukraine, much of the traditional media was also critical of the government.

26

that they were the pawns of Ukrainian oligarchs. In sum, an isolated president drew on
erroneous beliefs about the identities and motives of the protesters, and therefore
miscalculated the publics reactions to successive strategies of repression and extrication.

Rival Explanations
The initial miscalculation. Regarding each governments initial decision to
deploy repression at levels which turned out to be inefficient, we have argued that all three
underestimated the likelihood of a backlash movement p, in our model. The
counterfactual is that, had they interpreted p as large and the risk of a backlash high, they
would have avoided the middle-level repression/no-concessions early response.
Though we do not observe cross-national variation in governments assessment of p,
we do observe longititudinal variation in all three cases. Once the backlash occured, they
reassessed p and shifted strategies.
Still, we do not observe variation in the early assessment of p, so its worthwhile to
consider whether our account stands up to the possibility that, even if they had accurately
assessed the risk of a backlash at the outset, we might still observe inefficient levels of
repression at the first decision-point. Police indiscipline could explain such a pattern. One
could imagine that the Turkish government tried to instigate truly brutal repression early
on but could not get the police to carry it out, or that the excessively high levels of early
repression in Brazil and Ukraine reflected police indiscipline and an oversupply of
repression.
We find little evidence to support such a scenario. In the Turkish case, there is no
evidence of police insubordination or indiscipline and of their unwillingness to act
aggressively against the Gezi Park protesters. We interviewed Turkish police officers and
journalists who covered the events. Neither gave any indication that the government had
asked for higher levels of repression early on but had been rebuked by the police. Just a
few weeks later, furthermore, the police demonstrated a ready willingness to intensify
27

repression.
In Brazil as well, we found little evidence of insubordination in the Military Police
and Shock Troop actions against protesters, before, during, or after the June 13 crackdown.
Police officials whom we interviewed insisted that they acted on orders of the civilian
authorities. The government said, Dont allow people to stay in the streets. We obeyed.
In the beginning, Mayor Haddad said, I cant accept [the blockage of the Avenida
Paulista]. One week later he reversed his decision.56 After June 13, they also complied
with civilian orders to hold back on the use of rubber bullets, Shock Troops, and other
aggressive measures.
Of course the police would prefer to shift the blame for the repression onto their
civilian superiors. Yet the pre-June 13 public statements of mayor Haddad, governor
Alckmin, minister Cardozo, and others were consistent with police officials claim that they
were not acting against orders with the initial repression. Though there is little doubt from
their later statements that some civilian leaders were shocked by the degree of police
violence and its public quality, still this was not the first time that such tactics were used
nor the first time the authorities had ordered them.
Finally, on Ukraine: the debate there about where responsibility lay for the crucial
November 30, 2013 Berkut attacks revolves around whether it was ordered by Yanukovych
himself (the majority view among those we interviewed) or was the result of infighting and
intrigue within his inner circle or even, on some views, Russian pressure.57 But there are
no claims of Berkut indiscipline in the early stages of the repression, and most evidence
points to a direct role for Yanukovych for the November 30 attack, and to his son,
Aleksandr, for the December 11 events.58
Extrication strategies. The crucial factor shaping extrication strategies, we have
56

Interview with Colonel Glauco Silva de Carvalho, head of the Military Police Human Rights Directorate.
Conducted by Schiumerini and Stokes, S
ao Paulo, May 26, 2014.
57
For a further discussion, see Peisakhin 2014.
58
Loss of civilian control over the Berkut and the army did occur later in the protests, in late January
and February, 2014.

28

argued, is the security of the governments hold on office, specifically, its ability to maintain
its popular support base even if it undertakes very high levels of repression. For these
governments, varying degrees of security of office in turn reflected the nature of social
cleavages and public opinion about the government and the protesters. We now briefly
consider several rival accounts.
Decentralization. Brazil stands out among our cases for its decentralized structure
of government. And we saw that office-holders at several levels of government
sometimes from competing parties played a role in the extrication process there.
We saw little evidence that federalism was related to their choice of concessions and
restraint. Furthermore, the Ukrainian system was centralized and yet the strategic
instincts of the Yanukovych government, in the extrication phase, were more like the
Brazilians than the Turks.
Democratic Consolidation. Perhaps the penetration of democratic norms, such as
of the physical integrity of the person, was the crucial factor that distinguished, say,
Brazil from Turkey. Yet, what makes it hard to imagine the Brazilian police
repressing protesters with the level of brutality eventually used in Turkey is certainly
not a lack of police violence inflicted regularly on Brazilian citizens. Instead, harsh,
publicly visible actions against middle-class protesters, and against the journalists
who reported on them, would likely have reverberated powerfully against the
Brazilian authorities in contemporaneous public opinion and, possibly, at the next
elections. In addition, our case studies demonstrate that even in new and
unconsolidated democracies, such as Ukraine, norms against the physical abuse of
citizens are widely shared.
Ideological Orientation of the Government. Perhaps leftist governments are
less likely to crack down on protests, right-wing or conservative ones more likely. It is
no easy matter to characterize the ideological orientation of the governments in
29

question on a left-right, conservative-liberal, or any single dimension; we opt in Table

4 for a leftist-conservative dimension. But these orientations do not account for
different extrication strategies. The best evidence is from within-case variation in
Brazil. Conservative and leftist governments alike favored mid-level repression at the
first node, and both ideological types shifted to a non-repression, conciliatory stance
at the second node.
The Nature of Demonstrators Demands. Perhaps differences in the costs that
governments would have to bear should they make policy concessions explain the
different strategies of extrication they pursued. If this were the key explanatory
variable for the cases we have studied, it would have to be the case that the most
exigent demands were those of protesters in Turkey, with those in Brazil and Ukraine
making demands that were easier for their respective governments to meet. In any
objective sense, the opposite is true: the demand that the Ukrainian government
rejoin talks with the EU must be considered more substantial than that bus fares be
reduced or, even more so, that an urban development project be set aside.
Social Class of Protesters. The allusion, earlier, to the class composition of
protesters suggests another possible explanation. Perhaps governments are wary of
wielding batons against highly educated, middle class protesters, less so when the
protests are composed of the ill-educated and the poor. Though this proposition
might be true in general, it does not explain variation among the countries we
studied. Gezi Park protesters skewed high in income and education, as did those on
the streets of Sao Paulo and Kiev.59
Table 4 summarizes our cases in terms of our favored explanatory factor, factors
related to the five rival explanations just discussed, and the outcome the extrication
method that each government attempted. What emerges is the similarity of the Turkish
59

For more details, see Aytac, Schiumerini, and Stokes 2015.

30

and Ukrainian cases on most of the rival explanatory variables. What made Ukraine look
more like Brazil and less like Turkey was the lack of security of office that the leaders of
this tumultuous and revolution-prone country faced.
[Table 4]

Conclusions and Discussion

We have studied recent experiences in which the governments of new democracies
sparked major protests by underestimating the publics reaction to early repression. Why
did they get p wrong? The experiences recounted here point repeatedly to
post-digital-revolution technologies allowing the public to quickly become aware of police
violence. Certainly there were backlash protests in eras before the invention of the social
media and before cameras became ubiquitous. But, as Arab Spring movements in Tunisia
and Egypt also demonstrate, there is less room now for discrete or hidden violence than
there once was.60 Governments seem not to have adjusted fully to the world of quick
images flying through the Internet, a failure that is likely to have contributed to their
miscalculations.
The key observed variation in our study came in the extrication stage. Why did the
authorities embark on such sharply diverging routes out of the crisis? Our model led to the
prediction that secure governments would extricate themselves by displacing the level of
repression to the right on Figure 1, insecure ones, to the left. Our case studies support this
prediction.
Does our model travel to other new democracies, to advanced democracies, or to
authoritarian governments? Certainly government and protest dynamics in other new
democracies appear to follow the patterns identified here. For instance, Venezuelas
60

See Howard and Hussain (2013) for evidence that hand-held cameras played a key role in these
movements. Lawrence (2014) makes the case that such discrete violence was important in limiting the
impact of Moroccos Arab Spring movement.

31

experience in 2014 runs parallel to Turkeys in 2013, with small protests growing into a
mass movement after a heavy-handed police response against university students. As in
Turkey, the setting was highly polarized, the population of protesters was drawn nearly
exclusively from among the governments opponents, and the ruling party did not fear a
loss of support among its own constituents if it increased the level of repression, which it
did.61 A less secure leader was Chiles Sebastian Pi
nera (2011-14), who experienced a sharp
decline in public approval during massive student protests in 2011. Rather than ratcheting
up repression, Pi
nera reshuffled his cabinet and offered educational reforms.
These experiences raise the question: What difference does it make that the context
is a new rather than an established democracy? Our hunch is that the Chilean scenario is
typical of democracies everywhere but that repression like that meted out to protesters in
Turkey or Venezuela is uncommon in established democracies. Research suggests that in
the U.S. and several European countries, the police use intense (R) but not extreme (R+)
levels of repression. Scholars use terms like command and control or strategic
incapacitation to describe police actions in these settings, and have documented a
hardening of police tactics against demonstrators, in response to anti-globalization protests
and in the shadow of terrorist attacks.62 But rarely are large numbers of deaths or severe
injuries endured by protesters.63 Our model suggests that this lower level on the upper
bound of police violence may mean that elected governments in older democracies find
themselves in the situation of insecure governments in developing ones: they anticipate a
powerful erosion of support, should they deploy truly lethal force against unarmed
protesters.
Turning to authoritarian regimes, a growing literature indicates that they, too, may
unintentionally exacerbate protests by deploying excessive force. Excessive use of police
61

By late March, 37 people had been killed and more than 550 people injured (Amnesty International,
April 2014).
62
See della Porta and Reiter 2006; della Porta and Tarrow 2011; Gillham 2011; Vitale 2005.
63
This relative lack of severity contrasts, in the U.S., with the frequent use of lethal force by the police
against individuals.

32

force helped spark the 2011 movement that toppled Egypts Mubarak regime. Organizers
of the first large protest in Tahrir Square, Cairo, on January 25, 2011, kindled
anti-government activism by posting get out the protest clips on YouTube, which strung
together notorious scenes of police brutality captured by cell phone video cameras.64
Thus, public outrage at police violence can spark backlash movements in
authoritarian settings as it can in new democracies. But if the publics reaction to
repression can be hard to predict in democracies, it may be even harder to predict in
authoritarian systems, where people have incentives to falsify their preferences65 and where
the absence of free elections and party competition mean that the regimes lack valuable
sources of information about the publics mood. Once uprisings are under way, we suspect
that dictators perceptions of the security of their hold on office is not irrelevant to their
choice of extrication strategies. For instance, when Moroccos King Mohamed VI faced
large rallies for reform in February, 2011, it meted out relatively mild repression. It is likely
that this mild and concilatory response was shaped by fresh memories of revolutions in
Tunisia and Egypt.66 Yet, authoritarian regimes calculations about their hold on office are
also shaped by intra-elite competition, which can also be difficult to gauge.67 In addition to
asking themselves Will a harsh crackdown cost us too much support in public opinion?,
authoritarian rulers are more likely to ask, Will a harsh crackdown erode support too
sharply among our softline allies? Will our military leaders be willing to carry it off? And,
like their counterparts in new democracies, the answers they arrive at can turn out to be
consequential miscalculations.

64

El-Ghobashy 2011.
Kuran 1991.
66
See Lawrence 2014.
67
See Svolik 2012.
65

33

References
Alonso, Angela, and Ann Mische. 2014. Changing Repertoires and Partisan Ambivalence
in the New Brazilian Protests. Unpublished typescript, University of Notre Dame.
Amnesty International. 2013. Gezi Park Protests: Brutal Denial of the Right to Peaceful
Assembly in Turkey. London: Amnesty International.
Amnesty International. 2014. Venezuela: Human Rights at Risk amid Protests. London:
Amnesty International.
Aytac, Erdem, Luis Schiumerini, and Susan Stokes. 2015. Social Movements and the
Paradox of Repression: Lessons from Turkey, Brazil, and Ukraine. Working Paper, Yale
University.
Beissinger, Mark. 2002. Nationalist Mobilization and the Collapse of the Soviet State.
Cambridge: Cambridge University Press.
Blaydes, Lisa, and James Lo. 2012. One Man, One Vote, One Time? A Model of
Democratization in the Middle East. Journal of Theoretical Politics 24(1): 110146.
Brinks, Daniel. 2003. Informal Institutions and the Rule of Law: The Judicial Response
to State Killings in Buenos Aires and Sao Paulo in the 1990s. Comparative Politics
36(01): 119.
Brockett, Charles. 1995. A Protest-Cycle Resolution of the Repression-Popular Protest
Paradox. In Repertoires and Cycles of Collective Action, ed. Mark Traugott. Durham:
Duke University Press pp. 117144.
Chong, Dennis. 1991. Collective Action and the Civil Rights Movement. Chicago:
University of Chicago Press.
Datafolha. 2013. Violencia em Sao Paulo, Aumento da Tarifa do Transporte P
ublico,
Emprego II. Datafolha, June 18, 2013.
34

Davenport, Christian. 1995. Multi-dimensional Threat Perception and State Repression:

An Inquiry into Why States Apply Negative Sanctions. American Journal of Political
Science 39(3): 68312.
Davenport, Christian. 2007. State Repression and Political Order. Annual Review of
Political Science 10(01): 123.
della Porta, Donatella. 2013. Can Democracy be Saved? Oxford: Polity Press.
della Porta, Donatella, and Herbert Reiter. 2006. The Policing of Transnational Protest:
A Conclusion. In The Policing of Transnational Protests. Burlington VT: Ashgate
pp. 175189.
della Porta, Donatella, and Sidney Tarrow. 2011. Interactive Diffusion: The Coevolution
of Police and Protest Behavior With an Application to Transnational Contention.
Comparative Political Studies 45(01): 119152.
Earl, Jennifer. 2003. Tanks, Tear Gas, and Taxes: Toward a Theory of Movement
Repression. Sociological Theory 21(1): 4468.
El-Ghobashy, Mona. 2011. The Praxis of the Egyptian Revolution. Middle East Research
and Information Project 41(258).
Francisco, Ronald. 1995. The Relationship between Coercion and Protest: An Empirical
Evaluation in Three Coercive States. Journal of Conflict Resolution 39(2): 263282.
Francisco, Ronald. 2005. A Dictators Dilemma. In Repression and Mobilization, ed.
Christian Davenport, Hank Johnston, and Carol Mueller. Minneapolis: University of
Minnesota Press pp. 5881.
Gillham, Patrick. 2011. Securitizing America: Strategic Incapacitation and the Policing of
Protest Since the 11 September 2001 Terrorist Attacks. Sociology Compass 5(7):
636652.
35

Howard, Philip N., and Muzammil M. Hussain. 2013. Democracys Fourth Wave? Digital
Media and the Arab Spring. Oxford: Oxford University Press.
Konda. 2013. Gezi Park Survey: Who is Involved, What do They Want, and Why are They
There? Istanbul. Istanbul: Konda Research & Consultancy.
Kuran, Timur. 1991. Now Out of Never: The Element of Surprise in the East European
Revolution of 1989. World Politics 44(01): 748.
Lawrence, Adria. 2014. Repression and Activism among the Arab Springs First Movers.
Manuscript, Yale University, July 2014.
Lichbach, Mark. 1987. Deterrence or Escalation? The Puzzle of Aggregate Studies of
Repression and Dissent. Journal of Conflict Resolution 31(2): 266297.
Locatelli, Piero. 2014. Vem Pra Rua: As Revoltas de Junho Contadas pelo Jovem Rep
orter
que Recebeu Passe Livre para Contar a Historia do Movimento [Come Out into the
Street: The June Revolts as Recounted by a Young Reporter who Received the Free Pass
to Tell the Story of the Movement]. Sao Paulo: Companhia Das Letras.
McCarthy, John, and Clark McPhail. 1998. The Institutionalization of Protest in the
United States. In The Movement Society: Contentious Politics for a New Century, ed.
David Meyer, and Sidney Tarrow. Boulder CO: Rowman and Littlefield pp. 83110.
Moore, Will H. 2000. The Repression of Dissent: A Substitution Model of Government
Coercion. Journal of Conflict Resolution 44(1): 107127.
Opp, Karl-Dieter. 1989. The Rationality of Political Protest: A Comparative Analysis of
Rational Choice Theory. Boulder, CO: Westview.
Opp, Karl-Dieter. 1994. Repression and Revolutionary Action: East Germany in 1989.
Rationality and Society 6(1): 101138.

36

Oguzlu, Tarik. 2013. The Gezi Park Protests and its Impact on Turkeys Soft-Power
Abroad. Middle Eastern Analysis/Ortadogu Analiz 5(55): 1015.
Peisakhin, Leonid. 2014. The Logic of Maidan Revisited: Causes of Violent Regime Change
in Ukraine. Unpublished manuscript, NYU Abu Dhabi.
Przeworski, Adam. 1991. Democracy and the Market. Cambridge University Press.
Svolik, Milan. 2012. The Politics of Authoritarian Rule. Cambridge University Press.
Vitale, Alex. 2005. From Negotiated Management to Command and Control: How the
New York Police Department Polices Protests. Policing and Society 15(3): 283304.

37

Tables and Figures

Table 1: Security in Office, Perceived Public Sensitivity to Repression, and
Government Preferences over Strategies
Public sensitive to
repression (high p)

Public insensitive to
repression (low p)

Secure government

R+EARLY > R > C

R > R+EARLY > C

Insecure government

C > R > R+EARLY

R > C > R+EARLY

38

Table 2: Turkish Public Opinion of Gezi Park Protests

Which statement best explains your
opinion of the Gezi Park protests?

Turkey
overall

AKP
supporters

CHP
supporters

Protesters demanded their rights

and freedom in a democratic manner

40%

11%

87%

The protests are part of a plot against Turkey

54%

82%

10%

Dont know/no answer

6%

7%

2%

Source: Konda July 2013 Barometer (N=2,629). Konda Research & Consultancy, July 6-7, 2013.

39

Table 3: Sao Paulo Public Opinion of Protests

Do you support or oppose the protests
against the increase in bus fares
in the city of Sao Paulo?

S
ao Paulo
overall

PT
supporters

PSDB
supporters

Support

77%

73%

74%

Oppose

18%

22%

23%

No answer

5%

5%

4%

Source: Violencia in S
ao Paulo, Aumento da Tarifa do Transporte P
ublico, Emprego II (N=805).
Datafolha, June 18, 2013. The PT is the leftist party that, at the time of the protests, headed
the municipal government of S
ao Paulo as well as the national government. The PSDB is a
conservative party that headed the government of the State of Sao Paulo.

40

Table 4: Extrication Strategies: Where the Cases Fall on Favored and Rival Explanations
Security Centralof Office ization
Turkey
High
High
Brazil
Low
Low
Ukraine
Low
High

Democ.
Ideology
Consol.
of govt
Low
Conservative
High
Leftist
Low
Conservative

41

Costliness of Social Class Extrication

Demands
of Protesters
Strategy
Low
High
Repression
Middle
High
Restraint
High
High
Restraint

Probability that individual i joins

AC
Level of repression

Figure 1: Illustration of the Impact of Repression on the Costs of Abstention (A) and
Participation (C )

42

G
C

R+EARLY

(0)
1p

Secure gov: (1 rs)

Insecure gov: (1 ri)

Back down

Mobilize

(1)

G
LC

R+LATE

(0 )

Secure gov: (1 rs)

Insecure gov: (1 ri)

Figure 2: Decision Tree for Governments

43

Figure 3: Percent of Protesters and Non-Protesters Who Said They

Would Vote for the Ruling Parties, Sao Paulo and Istanbul

Percentage who would vote for ruling parties

45
40
35
30
Respondent
Protesters
NonProtesters

25
20
15
10
5
0
PT

PSDB

AKP

Note: At the time when the survey was conducted in Brazil (November 20-December 23)
the PT (Workers Party) was in power nationally and in the city of Sao Paulo; the PSDB
was in power in the state government of Sao Paulo. At the time the survey was conducted
in Turkey (November 20-December 15), the AKP (Justice and Development Party) was in
power nationally and in the city of Istanbul. Source: Authors Surveys, N = 2000 (Sao
Paulo) and 1214 (Istanbul).

44

Figure 4: Responses to Question: What is Your Opinion of [President Rousseffs

/Governor Alckmins/Mayor Haddads] Handling of the Recent Protests?

400

Politician
Rousseff
Alckmin
Haddad

Number of Responses

300

200

100

0
V.GoodGood

Fair

BadV.Bad

Opinion

No answer

Note: Source: Datafolha survey, June 18, 2013, N=805.

45

Supporting Information
Appendix A: Surveys
Turkey. We utilize data from three surveys conducted in Turkey. In the first one
we interviewed a probability sample of 1,214 adults in the city of Istanbul between
November 20 and December 15, 2013, five months after the Gezi protests. The selection of
households started with the random selection of one hundred and one neighborhoods from
the districts of Istanbul with a probability-proportional-to-size principle. Streets within
these neighborhoods and households were selected using a random selection table. In each
household the individual to be interviewed was selected according to their first names.
When the interview could not be completed with the selected respondent, the interviewers
tried to reach the individual for a second time. In the case of an unsuccessful second
attempt a new household was selected randomly, and this process was repeated until an
interview is completed. The response rate was 21%. The fieldwork was conducted by
Infakto Research Workshop.
An important contemporaneous source about the Gezi Park protests is a study by
Konda Research & Consultancy, a private polling firm (Konda, 2013). It was carried out
on June 6-8, 2013, a time when the police had retreated from the Taksim Square area and
the park served as a center of assembly for protesters and civil society organizations. The
goal of the Konda survey was to provide a profile of the protesters. They divided the park
into 10 zones of equal size and interviewed 4,393 protesters, in roughly equal numbers
across these zones, in a non-stop shift over the two days.
Additional information cited about the Turkish publics perceptions of the Gezi
protests draw on Konda Research & Consultancys July 2013 Barometer survey (Konda,
2014). Konda Research has been conducting monthly field surveys since 2010 on a sample
population that is representative of Turkish population. The July 2013 Barometer focused
on perceptions of the Gezi protests, and the fieldwork was conducted on July 6-7. Within

46

the scope of the survey, 2,629 respondents were interviewed face-to-face in 150
neighborhoods and villages of 98 districts of 28 provinces.
Brazil. As a contemporaneous source on public opinion about the protests, we
draw on a survey of the overall Sao Paulo population conducted by Datafolha on June 18
(Datafolha, 2013). Datafolha interviewed 805 residents of Sao Paulo at several points of
high concentration spread across the city. Respondents were randomly selected, and were
screened using quotas for gender and age.
We also conducted a representative sample survey in the city of Sao Paulo. This
survey provides insights on the opinions of both participants and non participants of the
protests. The fieldwork was conducted by the firm Analise, Pesquisa e Planejamento de
Mercado between November 20 and December 23. Interviews were conducted via phone.
Households were selected using random digit dialing, with quotas for neighborhood,
gender, age, and level of education. Individuals were selected according to their first
names. When the interview could not be completed with the selected respondent, the
interviewers tried to reach the individual for a second time. In the case of an unsuccessful
second attempt a new household was selected randomly, and this process was repeated
until an interview is completed. The response rate was 10%.
Kiev. An important source about the protesters in the EuroMaidan is the survey
conducted by the Kiev International Institute of Sociology (KIIS). Similar to Kondas and
Datafolhas surveys in Istanbul and Sao Paulo, respectively, the KIIS survey was a
contemporaneous convenience sample of protesters who were present at the EuroMaidan.
It was conducted with 1,037 respondents on December 7-8, 2013 (the weekend of large
demonstrations), a week after the initial Berkut attack.

47

Appendix B: A Decision-Theoretic Model of Government

Repression of Protests
In our model the government first decides how to respond to a relatively small
movement that demands concessions. The government can choose to make (early)
concessions (C), use aggressive police tactics to try to break up the protest (R), or inflict a
truly brutal response (R+EARLY ).
If the government makes concessions or goes for full brutality, the protests end. It
receives a payoff of 0 if it makes concessions and 1 if it suppresses the protests with harsh
repression and denies any concessions to the movement. If the government chooses tactics
that are heavy-handed but not fully brutal, it faces the possibility (1 p) that the
movement will back down and protests end (the government again receives a payoff of 1) or
the possibility (p) that the citizenry pours into the streets and the government has an
uprising on its hands. In the latter case, the governments next decision is whether to scale
back the repression and make late concessions (LC), or scale up repression to a higher level
(R+LAT E ).
The cost of applying high level repression (R+EARLY or R+LAT E ) differs across
government types, with secure governments incurring a cost of rs and insecure ones
incurring ri . Insecure governments expect to pay a higher political price for applying high
level of repression than secure governments, and therefore we posit that rs < 1 < ri .
Accordingly, secure governments always prefer early high repression (R+EARLY ) over
(early) concessions while the opposite is true for insecure governments.
In the case of an uprising (second node), the payoff of late concessions differs from
early concessions by , which is the price paid for enduring massive protests.68
Alternatively, the government can ramp up the level of police brutality (R+LAT E ). Again,
68

As modeled, the government pays a price for an uprising occurring, apart from its possible effect on
the governments future election prospects, which are in the background here. As we shall see, during
the protests, even secure government, like the Turkish one, faced pressure from the business community,
international actors, and its own party, to end them.

48

secure governments prefer late, intense repression over late concessions while insecure ones
prefer concessions. The payoff to secure governments that choose late, intense repression is
(1  rs ); to insecure governments, it is (0  ri ).
Which strategy governments deploy at the first node depends on the magnitudes of
ri , rs , , and p. For a secure government, the expected payoff for the early, middle-level
repression (R) is:
(1 p) 1 + p (1  rs )

(1)

1 p ( + rs )

(2)

or, equivalently,

If a secure government is certain that early, middle-level repression will work in

shutting down protests (p = 0), its preference order is R > R+EARLY > C. If a secure
government is certain that repression will spark an uprising (p = 1), its preference order is
R+EARLY > R > C. A secure government that sees p as falling between 0 and 1 will prefer
R > R+EARLY when the following inequality holds:

1 p ( + rs ) > 1 rs

(3)

or, equivalently, when

p<

rs
 + rs

(4)

If the cost of repression (rs ) is high, the secure government will be willing to pursue
a strategy of early, modestly intense repression at higher levels of p, tolerating a greater
risk that the repression will cause a backlash. Its willingness to use this strategy is also
sensitive to , the cost of having to endure the uprising. The lower the anticipated costs of
protests (), the greater the probability of an uprising (p) that the government will tolerate
the more willing it will be to the gamble and not repress at the highest level of intensity,
early on. Here the government can be thought of as trading off the possibility of lost future

49

electoral support against a chance of ending the protests and not making concessions.
Turning to insecure governments, R+EARLY is always the least attractive option.
Between early concessions and repression (R), if it is certain that repression will instigate a
revolt (p = 1) it will make concessions; if it is certain that repression will end the protests
(p = 0), it will repress. If it sees p as falling between 0 and 1, it prefers repression when:

p<

1
+1

(5)

Inequality [5] indicates that the greater the costs that the protests impose on
insecure governments, the more willing they will be to make early concessions. And,
obviously, for both kinds of governments, the smaller p the more likely R is to provoke a
backlash the more they will favor a different strategy: for the insecure, concessions, for
the secure, an early Armageddon. Hence if the government represses and there is a
backlash, it is likely that the government has misperceived the magnitude of p.
A final point is on the activists incentives to mobilize against a secure government.
The model predicts that a secure government would never make concessions; activists
would face either (i) early harsh repression, R+EARLY , or (ii) late harsh repression,
(R+LAT E ), when the government underestimates p and its initial response R sparks an
uprising. If in both cases the protests end without concessions, why should activists
mobilize against a secure government in the first place? Although we do not model the
activists payoffs explicitly, we can plausibly assume that activists would get value from
imposing costs on the government. The payoffs associated with early and late harsh
repression to a secure government are (1 rs ) and (1  rs ), respectively, and less than
1, so that in both cases the government is worse off than if the activists had not mobilized.
This should give activists a reason to mobilize even in the absence of possibility of
immediate concessions, especially if we think of the protests as part of an on-going struggle
between the government and its opponents. In Turkey, for example, even though the
government swiftly and repeatedly suppresses any protest attempts in the aftermath of
50

Gezi, opposition activists still try to stage protests sporadically. This observation is in line
with the assumption that activists see value in demonstrating public opposition to the
government beyond any immediate effects (e.g., concessions).

51