When Does Lethal Repression Fail? Unarmed Militancy and Backfire in Bolivia, 1982–2021

Abstract Repressive state violence, intended to tamp down collective mobilisation, sometimes inspires greater participation by protesters. When popular and/or elite reactions cause the repressing party to concede, civil resistance scholars define the failure of state repression as ‘backfire’. Some have proposed that movements’ nonviolent discipline is essential to backfire. This article demonstrates that movements that practise ‘unarmed militancy’ – forceful, combative tactics less damaging than armed violence – can also succeed through backfire, achieving policy concessions and even presidential resignations, and presents a qualitative comparative analysis of the outcomes of 48 protest events with multiple deaths in Bolivia between 1982 and 2019, and a case-based analysis of how either movements or repressors prevailed. Movements that confronted deadly repression succeeded in 57–8 per cent of cases. Whether or not protesters engaged in lethal defensive violence did not affect their likelihood of success. However, state repression of guerrillas and paramilitary groups, and during polarised partisan conflicts, was consistently successful. Current understandings of backfire need to be reconsidered in light of successful unarmed militant protest in Bolivia and numerous other locations worldwide.

Sometimes repression using lethal force, the ultimate recourse of governments faced with protest, simply does not work.Gonzalo Sánchez de Lozada's second term as president of Bolivia, which spanned the 14 months from 6 August 2002 to 17 October 2003, exemplifies how lethal violence against protest movements often proves futile for the governments that order it.Police and soldiers killed a total of 101 Bolivians during this period. 1Nevertheless, the government conceded to the demands of protesters in the January 2003 pensioners' strike, in the February 2003 anti-tax protests known as the tarifazo, and in the September-October 2003 protests against gas privatisation and for greater Indigenous rights, or 'Gas War' (to name only the deadliest examples).2Such rebounds from lethal state repression to social movement successtermed 'backfire' by scholars of protest have occurred frequently over the last four decades of Bolivian political life.
How did this happen?Consider two examples.
In July 2003, President Sánchez de Lozada mobilised 130 National Police officers and 220 Army soldiers to the rural town of Santa Rosa del Sara.Residents had blockaded the regional highway and shut off the valves of the Transredes gas pipeline to press demands for local development and settlement of a land dispute. 3oad blockades are a long-standing protest tactic in Bolivian grassroots movements, 4 and interrupting commerce and transport can be particularly effective at arousing a response from regional and national governments.The troops seized control of the pipeline valve but were soon confronted by some 1,500 of the town's 4,000 residents.In a three-hour clash, government troops with tear gas and live ammunition confronted residents wielding sticks and stones.Eight civilians and seven troops were wounded; protester Luis Zelaya Márquez was killed by multiple gunshots to his torso. 5Zelaya's death served only to anger the crowd, who proceeded to retake the pipeline valve.The military/police intervention at Santa Rosa del Sara was quickly characterised as a failure and even a moral atrocity by the national media and opposition legislators.The day after the intervention the Santa Cruz regional government agreed to pay for road upgrades and to name the resulting highway after the slain protester. 6Black Monday' at Santa Rosa del Sara prefigured a vastly larger 'Black October' (or 'Gas War') of protests against gas privatisation and for greater Indigenous rights.During his last three months in office, Sánchez de Lozada's government attempted to prohibit, to police and to militarily repress road blockades, only to have this repression dramatically backfire, leading him to resign from office.7 Soldiers deployed to break up road blockades shot five civilians in rural areas of La Paz department on 20 September and killed 53 civilians (largely in the cities of El Alto and La Paz) between 9 and 16 October 2003.8 Undeterred, protesters rebuilt and expanded their blockades day after day.The killings prompted Vice President Carlos Mesa to publicly distance himself from the government and Human Rights Ombudswoman Ana María Romero de Campero to lead a nationwide hunger strike campaign calling for the president's resignation.Over 1,500 people swore off food at 83 different sites.9 Sánchez de Lozada yielded to massive crowds demanding his resignation, and fled the country on 17 October.
The 'paradox of repression' is that violence against social movements can either tamp down collective mobilisation or inspire greater participation and risk-taking by protesters. 10Santa Rosa del Sara and the Gas War are exemplars of the latter, where violent repression 'backfires '. 11 In both cases, the massive deployment of governmental force, including lethal violence, failed to demobilise the protesters, failed to deter them from further direct action, and failed to demonstrate control over public space.In Santa Rosa, the government rapidly backtracked from its considerable investment in repression and opted to yield to the protesters' demands.During the Gas War, the climbdown came after five weeks of confrontation and through a major political crisis.In both cases, the very lethality of the military/ police repression contributed to its failure on the ground, galvanising opposition while fracturing supporters of the government.
Nonviolence theorists, 12 scholars of civil resistance 13 and strategists of backfire 14 have proposed that massive participation and rigorous commitment to nonviolent tactics offer the best way for protesters to transform a situation of deadly repression into one that backfires upon the repressive government.In such cases, as Doug McAdam and William Sewell summarise, 'The condition of the success of such "people power" revolution is that the regimes in power be unwilling to use their superior military force in putting the demonstrations down.' 15 In these accounts, del discurso neoliberal (La Paz: CEDLA, 2006); Christian Jiménez Kanahuaty, Maquinaria andante: Historia, poder y movilizaciones sociales en la ciudad de El Alto (Quito: Abya-Yala, 2015).movements succeed primarily by demonstrating their moral superiority or worth, typically by 'maintaining nonviolent discipline', and thereby dissuading governments from repression, often by alienating the security forces, national elite and/ or international supporters from the repressing government. 16olivian protesters' militant, if unarmed, response throws these formulas into question.As in our two examples, the government is often willing to use lethal force, but unable to deter people from continuing to mobilise by doing so.At times of escalated mobilisation and repression, most Bolivian social movements adopt a stance of 'unarmed militancy'fierce, confrontational, property-destroying and property-repurposing forms of mass actionand engage in battles for public space. 17f, as this article will show, protesters engaged in unarmed militancy succeed through backfire, we need to rethink under exactly which circumstances backfire occurs.And if, as the data examined here demonstrate, backfire is no less likely for unarmed militant protesters, we must also rethink the centrality of 'nonviolent discipline' to the narrative of backfire.
Drawing on a nearly comprehensive database of deaths in Bolivian political conflict from 1982 to 2021 (see the online Supplement), this article assesses the role of state violence and protester tactics.It centres on a qualitative comparative analysis of the outcomes of 48 events over a 39-year period that involved three or more deaths.Episodes of lower-level repression may also spur backfire, as the above example of Santa Rosa del Sara illustrates, but because it had only one death, it is not included in the sample analysed here.Qualitative comparative analysis seeks to connect outcomes (here, the success or failure of movement demands) with causal factors.Using the 48 events as cases, I conducted an iterative analysis of their outcomes, seeking to develop an if-then path through which analytical variables could generally predict the outcomes of protest events. 18To analyse the role of one-sided or unanswered violence, I split the cases into four quadrants based on the presence or absence of deadly state violence and deadly violence against state security forces (see Table 1).Twenty-eight of the events involved deaths inflicted by state security forces (Quadrants 1 and 2), and these were analysed to seek the explanatory factors that predict movement success or failure.I separately consider the remaining cases to analyse the phenomenon of backfire around acts of private (non-state) lethal violence.Four events, involving unanswered lethal violence by protesters against state security forces (Quadrant 3), showed mixed results.And in 16 events, deaths were neither inflicted nor suffered by security forces (Quadrant 4), but disparities in violence did affect outcomes.16 Chenoweth and Stephan, Why Civil Resistance Works, pp.46-51; Sharon Erickson Nepstad,  'Nonviolent Resistance in the Arab Spring: The Critical Role of Military-Opposition Alliances ', Swiss  Political Science Review, 17: 4 (2011), pp.485-91, https://doi.org/10.1111/j.1662-6370.2011.02043.x.17   Carwil Bjork-James, 'Unarmed Militancy: Tactical Victories, Subjectivity, and Legitimacy in Bolivian Street Protest ', American Anthropologist, 122: 3 (2020), pp.514-27.18   Examining the related phenomenon of security forces defecting from the repression of dissenting movements using a computational Boolean analysis, Alexei Anisin, 'Unravelling the Complex Nature of Security Force Defection', Global Change, Peace and Security, 32: 2 (2020), pp.135-55, https://doi.org/10.1080/14781158.2020.1767046,also found that 'contrary to theoretical expectations … violent oppositional campaigns' were capable of causing a breakdown in repression.
Overall, repression tended to backfire.In a majority of cases (57-8 per cent; see Conclusion), protesters won their desired outcomes after suffering deadly repression.The civil resistance literature would predict that events with unanswered state violence (in Quadrant 1) would benefit from backfire, whereas those with deadly violence by movements (Quadrants 2 and 3) would usually not.This turned out not to be correct: backfire did not depend on the nonviolence or nonlethality of protesters.As I argue in the next section, unarmed but forceful protests (including many that ended up in Quadrant 2) often benefit from the same dynamics of backfire as those in Quadrant 1.
The analysis below explores additional causal recipes to explain when repression is more likely to succeed.According to civil resistance scholarship, repression of armed actors such as left-wing guerrillas and right-wing paramilitaries would not generate backfire against the state repressing them, and nor would repression of the coca grower movement, given its willingness to fight back violently.The analysis here shows that repression of organised armed groups was successful, but that the (mass-based, but combative) coca grower movement had no less success than other movements.One new hypothesis emerged from my analysis: during a two-sided partisan conflict, even unanswered state killings may not arouse enough outrage against repression for them to backfire.This hypothesis explains an important cluster of cases.
The article first summarises the literature on tactics and backfire as well as democracy and repression.Next, I present a summary of the Bolivian case and some overall trends in repression that frame my choice to focus on tactics in this article.Third, I introduce the Ultimate Consequences dataset of lethal events in Bolivian political conflict and explain how events were defined and outcomes coded.The dataset is described in greater detail in the online Supplement for the article.The queries, data tables and code used to create them are published online at The article then presents the results of the qualitative comparative analysis, as well as a look at how cases of deadly private (non-state) violence (Quadrant 4) can be seen in relation to patterns found in the analysis.

Untangling the Paradox of Repression
Governments' use of violence, and deadly violence in particular, is only one of a variety of possible strategies designed to repress and control dissent, which may range from surveillance and harassment to torture and mass killings. 19Scholars agree that states repress in response to dissent, particularly dissent seeking a change in rulers or political and economic structure, 20 but the outcome of repression, including deadly state violence, is uneven and difficult to predict.There are important cases of deadly violence ending a protest campaign, and human rights movements often point to the dissuasive or chilling effects of state violence on political participation. 21On the other hand, strategists of nonviolence and civil resistance scholars have also described the phenomenon of 'backfire': 'when an unjust actoften violent repressionrecoils against its originators, leading to power shifts by increasing the internal solidarity of the resistance campaign, creating dissent and conflicts among the opponents' supporters, increasing external support for the resistance campaign, and decreasing external support for the opponent'. 22While rooted in injustice, backfire requires activation by movements that highlight the act and mobilise a response. 23Ronald Francisco found that after 31 massacres of protesters in undemocratic countries, people joined response protests in even larger numbers, but that these 'backlash protesters' could still be deterred by further repression. 24onfrontations between large-scale unarmed protest movements and the governments they challenge are critical events in the political evolution of contemporary societies.During these conflicts, the use of deadly force can tip the balance between dramatic political openings and a retrenchment of the status quo, albeit in sometimes unexpected ways.Scholars have identified the frequent success of 'people power' revolutionsunarmed mobilisations that win moral and/or practical leverage over authoritarian statessince the 1980s as a significant shift in how revolutionary political changes are realised. 25This change in political praxis has prompted a shift in scholarship towards studying the interactions of unarmed political revolutionaries and the states they confront. 26he peace studies and civil resistance literature developed against a background presupposition, particularly in political science and history, that violent revolution was the modal way in which political change occurred. 27In a first stage of the scholarship, advocates proposed theoretical arguments for the utility of nonviolence to make political change 28 and aimed to characterise the methods involved. 29In the second stage, successful cases of nonviolent action surviving repression and resulting in political change were characterised and analysed. 30Sharon Erickson Nepstad and Kurt Schock used people power revolutions of the 1980s and 1990s as comparative case studies. 31In the third stage, Erica Chenoweth and Maria Stephan performed quantitative analysis on a dataset of twentieth-century movements attempting to end colonial occupations or overthrow governments. 32As these stages have proceeded, nonviolence advocate Mohandas Gandhi's description of nonviolent resistance as 'a force which is more positive than electricity, and more powerful than even ether' 33a statement that was as much spiritual as it was strategichas been embodied by Gene Sharp's notion of political jiu-jitsu, that 'by combining nonviolent discipline with solidarity and persistence in struggle, the nonviolent actionists cause the violence of the opponent's repression to be exposed in the worst possible light.This, in turn, may lead to shifts in opinion and then to shifts in power relationships favorable to the nonviolent group.' 34 Chenoweth and Stephan's Why Civil Resistance Works completed a paradigm shift from nonviolence as moral precept to nonviolence as strategic imperative.Their core finding was that 'nonviolent resistance has been strategically superior to violent resistance during the twentieth and twenty-first centuries'. 35Echoing nonviolence theorists' arguments, Chenoweth and Stephan argue that both the likelihood and the strength of backfire are amplified when a movement maintains nonviolent discipline and clear contrast between their nonviolent means and the violence used by their state opponents.They further propose that backfire is conditional on nonviolence: 'The broader population's tolerance of government crackdowns may depend on whether the resistance campaign is nonviolent or violent, as repressing nonviolent campaigns may backfire.' 36Their explanation emphasises elite dissent, security force defections and international isolation as mechanisms for backfire.
While these works categorise protesters' tactics according to a violence-nonviolence dichotomy, the actual tactics used by people power revolutionaries blur this boundary, even in the paradigmatic cases highlighted by Chenoweth and  Stephan: the South African anti-apartheid struggle and the first Palestinian Intifada.Like recent Bolivian protests, these movements interwove mass rallies and strikes with iconic street confrontations.All three of these use what I have termed unarmed militancy: 37 the use of forceful, combative tacticssuch as barricades, property destruction, hands-on pushes and projectilesby unarmed crowds in political mobilisation to serve symbolic, tactical and strategic goals.Despite some assumptions to the contrary, unarmed militantsparticularly in mass movements of the Global Southoften maintain cooperative, even immersive, relationships with larger mass movements.By disrupting the economic life of apartheid South Africa in the 1980s, seizing Argentina's streets in the December 2001 political crisis, and securing and defending access to Egypt's Tahrir Square in January 2011, unarmed militants played pivotal roles in major political transitions.My recent scholarship suggests that unarmed militancy needs to be treated as occupying a space between and beyond nonviolence and armed struggle, and examined as a strategically distinct form of praxis. 38n alternative way of understanding backfire scenarios is to see them as primarily tactical rather than moral victories.When a government decides to attack a movement with force, it puts its credibility and reputation on the line: if the movement is not quelled, and even expands its mobilisation, then the government's inability to control the population is publicly revealed.Movements succeed by demonstrating irrepressibility, or resilience to repression, rather than moral superiority over their repressors.it and to restore interrupted workdays, transportation flows and circulation of goods. 39Deadly repression does not just backfire, it fails on its own terms.
This pattern of failed repression and resilient protest movements dramatically weakening the state has been explored in nonviolent and unarmed militant scenarios alike.Nonviolent strategist Sharp argues that the application of repressive violence is also a critical test of a social movement: If a political non-cooperation campaign, or a massive strike, collapses as soon as there are arrests, beatings, or deaths, there is no time for the resistance to have an effect.All sacrifices will have been in vain.If, however, the opponents' use of repression fails and the resisters are willing to persist, widespread noncooperation has the potential of gaining the objectives of the struggle and even of disintegrating the oppressive system. 40alysing the Egyptian uprising in January 2011, Mona El-Ghobashy described the proliferation of protest 'signaling the unwillingness or incapacity of the coercive apparatus to suppress demonstrations' as among 'the worst fears of an authoritarian regime'. 41In Egypt as in Bolivia, protesters' resilience to repression was achieved through unarmed and combative resistance.
In short, scholars agree that state repression has a clear intentto dissuade protesters from mobilising for their demandsbut ambiguous results: sometimes quelling mobilisation and at others inspiring broader involvement and greater commitment.Many mobilisations ultimately succeed despite (or indirectly because of) deadly repression.What is sharply debated is whether such backfire emerges when protesters morally distinguish themselves from government repressors, or rather when protesters demonstrate persistence despite deadly repression.This distinction is of critical strategic importance to social movements operating in a context of deadly violence.
Comparative, and sometimes quantitative, studies of multiple cases have been an important method for considering the success and failure of mass nonviolent protest, of civil resistance and of violent rebellion.Scholars have created pairwise comparisons (small clusters of similar cases), 42 and assembled global datasets 43  support of these analyses.The premise of this data collection has been that more complete and globally representative samples best facilitate rigorous analysis.In this article, I introduce a different dataset that is limited to one nation, but which provides exhaustive coverage of social movement conflicts over a four-decade period.The nation in question, Bolivia, is highly politically mobilised and recognised for the extraordinary level of mass participation in political life. 44I consider both government-unseating protests and those with more limited demand sets over this period.As with Harvard's Nonviolent and Violent Campaigns and Outcomes (NAVCO) dataset used by Chenoweth and Stephan,45 the effort at comprehensive coverage serves to minimise selection bias.Regardless of whether the conclusions reached here about Bolivia are generalisable (a point to which I return in the Conclusion), they illustrate the potential for unarmed militants to benefit from backfire while flouting or breaking the presumed rules of nonviolence.

Protest and Repression in Democratic Bolivia
This article analyses moments of deadly conflict across nearly four decades in a onecountry-multiple-cases comparative study. 46The Ultimate Consequences dataset begins with the restoration of electoral democracy in Bolivia in October 1982 and data is still being collected in 2024.At the time of submission, the last coded death occurred in July 2021 and the last event with three or more deaths occurred in 2019.The data covers 12 presidencies and two days of interim military rule, and includes three unelected interim presidents.Governments during this period can be identified as socialist and conservative, as pro-and anti-neoliberal globalisation, and as supporting and opposing a military role in policing dissent.The number of deaths perpetrated by state security forces during these presidencies varied from zero (Eduardo  Rodríguez, 2005-6) to 101 (Sánchez de Lozada, 2002-3).Conflict-related deaths occurred in all but four years in the study period (1982, 1983, 1999 and 2020) and state security forces have been responsible for deaths in 31 of 39 calendar years.
Political violence in Bolivia differs from that in its South American neighbours in scale and kind.Mass grassroots politics in Bolivia has used highly contentious forms of action that are nonetheless distinct from conventional military conflict.
Its longer history, marked by Indigenous uprisings, labour militancy and frequent military rule has been described in terms of blood, fire, dynamite and massacres. 47he social movement traditions that have resulted include proclamations of fearlessness (even protesting high school students shout, 'Rifle, machine gun, we will not be silenced!')and vows to carry on with struggles 'until the final consequences'.
Guerrilla and paramilitary forces have been small and marginal, especially by comparison with those in the prolonged civil wars in Peru and Colombia, but also with the urban guerrilla movements in Chile, Argentina and Uruguay.During the years of military dictatorship (1964-82), many hundreds of people were killed, and thousands imprisoned or exiled.Most of these deaths took the form of massacres of protesters, though many were also killed in detention. 48he number of deaths per year during the study period (1982-2021) was less than one-sixth of that during 1964-82.Spectacular incidents of backfire in Bolivia date back to the reversal of Alberto Natusch Busch's November 1979 coup following the deaths of scores of protesters.Politicians across the political spectrum have disavowed massacre as a means of governing and made distinguishing themselves from the dictatorship years a rhetorical priority. 49uring this entire period, Bolivian presidents were either elected or faced an upcoming election.Bolivia might therefore be expected to see lower levels of repression than in places or times without democratic constraints. 50The Polity IV Project has consistently rated Bolivia as a democracy (7, 8, 9 on a scale where scores above 6 are classed as democracies); 51 Christian Davenport and David Armstrong found that Polity IV scores above 7 are associated with declining levels of repressive violence. 52Nonetheless, the consolidation of democracy in Bolivia has often been seen as incomplete. 53Protests prompted four presidents to leave office before their elected terms ended (in 1985, 2003, 2005 and 2019), and deadly partisan violence accompanied regime transitions in 2006-9 and 2019.
Long-term protest monitoring data collected by the Bolivian Centro de Estudios de la Realidad Económica y Social (Centre for Studies on Economic and Social Reality, CERES) shows two major waves of mass action (see Figure 1): a significant protest boom after the restoration of democracy (from 1982 to 1985), and a larger tide of protest from the end of the 1990s to 2012 (their latest published data).Later data from Fundación UNIR Bolivia records a peak number of conflicts in 2014, and the frequency of conflict remaining above that for 2010 in all but one year of the decade 2010-20. 54The first protest wave challenged the government of Hernán Siles Zuazo from the Left, using general strikes and coordinated blockades to demand new elections in 1985.Labour unions affiliated with the country's main confederation (the Central Obrera Boliviana) and peasant movements were the primary actors in this period.However, Víctor Paz Estenssoro, who won the 1985 elections, instead led the country sharply to the Right, initiating a crash programme of neoliberal economic restructuring.After mass arrests and military intimidation put down miners' protests against the plan, direct action protestsuch as 'confrontations, takeovers, and riots', urban blockades and rural blockadesnearly disappeared until 1997; expressive protest like 'marches and demonstrations', on the other hand, never stopped (see Figure 1).During this relative lull, the Chapare coca growers and the lowland Indigenous movement emerged as new protest actors.A new protest cycle, challenging neoliberal policies and the mestizo/creole domination of political life, is usually dated to 2000 to 2005.This cycle included urban antiprivatisation protests and greater activity by peasant and coca growers' unions.The increased pace and intensity of protest did not end with the election of Evo Morales (leader of the Movimiento al Socialismo, Movement towards Socialism, MAS) in December 2005 but continued an upswing until at least 2012.Over the long term, active and confrontational forms of protest have become the most common acts by social movements in Bolivia. 55Left-leaning grassroots movementslabour unions, peasant movements and movements of the urban poorhave been the primary protagonists of protest through much of this period, 56 but significant mobilisations by largely urban and economically well-off movements on the Right challenged the Morales government from 2006 to 2019, and the Arce government since its accession in October 2020.
A chronological look at the large protest events examined in this article (Annex 1) shows two streaks of movement success, the first consisting of four events between 1987 and 1989 (including a partial success) and a second between 2000 and 2004, comprising 11 movement successes and one repression success.The 1987-9 streak occurred during a period remembered better for the defeat of labour movements as neoliberal shock therapy was imposed.The second streak matches the widely discussed 2000-5 protest cycle. 57  political leadership saw very different levels of political violence.Presidents on the political Right governed for 54 per cent of days from October 1982, but were in office during 75 per cent of deaths, 88 per cent of killings by the security forces and 82 per cent of deadly acts against the security forces. 58Consistent with this disparity, the analysis here of 28 events covers 23 involving state repression by right-wing governments and five by left-wing governments. 59All of the 23 involved protest by traditionally Left constituencies, but two protest events challenging a left-wing government were by Right movements (the Sucre constitution protest, ID no.35, and the Santa Cruz raid on the Rósza group, ID no.37) and three by Left movements (the Yapacaní Mayoral Dispute, ID no.38; Cooperative miner strike, ID no.43; and the Asunta coca conflict, ID no.44).Despite the relevance of political orientation to success, this small and heterogenous set of events seems too small to allow separate conclusions to be drawn about the effect of left-wing governments on protest outcomes.Moreover, governments of all political orientations have faced mass protests, made political concessions after repression and even left office under popular pressure.Hence, in the analysis that follows, I focus my search for causal explanations on repression and movement outcomes, rather than on government ideology.

Analysis
The qualitative comparative analysis presented here seeks to specify if-then paths through which analytical variables can generally predict the outcomes of protest events.It is paired with a case-based examination of how moral outrage, widened popular support and elite defections (that is, backfire) contribute to these outcomes.I focus here on explanatory variables related to social movements' tactics and campaigns, rather than to their strategic efforts to generate backfire in the aftermath of violence. 60I treated contradictory cases as inherent to the size of the dataset and handled them in terms of proportions of outcomes, 61 rather than seeking additional variables that could explain every success and failure.While the analysis formally treats the presence or absence of deaths perpetrated by the state and against the state as binary variables, my narrative intentionally gives greater weight to events with more deaths.
The analysis uses 'events' rather than individual deaths as its cases.Bolivian mass mobilisation is organised in a routine sequencecollective organisation, mass mobilisation, escalation of tactics, negotiation and concessionsand organisers work to propel the process forward repeatedly until the target, usually the government, gives in.Repressive force seeks to interrupt this process, usually by punishing movements in the mobilisation or escalation stages.For purposes of the database, I treat a relatively continuous series of actions as a single episode or event, unless and until it is interrupted by demobilisation.
For each of these events, I surveyed news reports and academic writings on the conflicts involved and described the results of the protest event from the point of view of the protesting group.In Quadrants 1-3 these are coloured as having a binary result: movement success (up arrow; mid-grey; green online) or repression success (down arrow; black; dark red online), while the text describes the outcome; in Quadrant 4 outcomes are not coloured as success by the movement or the state, but rather classified as success by the (quantitatively and qualitatively) less violent party or as attracting mediation by the state.Among the several that were coded as movement successes, I note a '(partial) agreement' on the part of the government.One case, the 1988 Villa Tunari massacre (ID no.5), is coded as having a mixed outcome.Brief narratives for all cases involving state repression are included in the online Supplement.
Altogether, the 48 qualifying events (i.e., those with three or more deaths; see Annex 1 and Table 2) caused 422 confirmed deaths, 74 per cent of all conflict-related deaths between 1982 and 2021.(Of the remaining events in the database, 98 involved one death each, and 26 involved two.)The qualitative comparative analysis began by splitting the data into four quadrants based on the presence or absence of deadly violence.A first cut separated cases in which state repression was lethal from those where it was not.The second cut addressed whether state security forces suffered deadly losses during the event.(In the terms of the database this is whether the number of 'state perpetrator deaths' and 'state victim deaths', respectively, is zero or nonzero.)I hypothesise three Quadrant 1 presents outcomes of protest events in cases with deadly state repression and zero state victims.In 15 events, resulting in 87 deaths, state security forces were responsible for all killings.At least five of these events are commonly referred to as 'massacres'62 and all are examples of unanswered repression, the presumed prerequisite for backfire.The government conceded major movement demands during the 1987 and 2001 coca growers' protests (ID nos.3, 21) and 1987 education strike (in which striking teachers made wage demands, ID no.4); and the September 2000 peasant mobilisation (ID no.18) succeeded after suffering nine state-perpetrated deaths.In the 1989 Santa Ana de Yucuma event (ID no.6), a town's widespread participation in resisting a narcotics raid prevented detentions and, in 2012, Yapacaní's mayor tendered his resignation following the deadly repression of protesters demanding it (ID no.38).
Nonetheless, in five of the 15 cases, state repression succeeded.In three of the four deadliest cases (the exception is the September 2000 peasant mobilisation, ID no.18), protesters did not achieve their goals following repression: the June 1988 Villa Tunari massacre and regionwide militarisation effectively repressed a coca growers' campaign against a new law prohibiting the crop, although protesters did stop plans to use chemical defoliants as a method of eradication.The November 2019 massacres of pro-Evo Morales protesters in Sacaba (10 deaths, Cochabamba department, ID no.47) and Senkata (11 deaths, La Paz department, four days later, ID no.48) resulted in the demobilisation of a widespread campaign to demand the return of the ousted president Morales.Smaller losses of three to five people at the hands of state violence saw no concessions granted by the state (Sucre protests against the Constituent Assembly, 2007, ID no.35 and Coca eradication, 1995, ID no.11) and the collapse of guerrilla and paramilitary groups (raid on militant Marxist-Leninist Comisión Néstor Paz Zamora (CNPZ), La Paz, 1990,  ID no.8; raid on the right-wing Rósza group, Santa Cruz, 2009, ID no.37).
To explain the cases where unanswered state repression did not result in movement victories, I propose three working hypotheses, which will be tested in the remainder of the analysis:  Hypothesis A: Repression of armed actors such as guerrilla and paramilitary groups immunises the government from backfire over their repression.This reasonable hypothesis could explain why raids on the leftist CNPZ guerrillas and the rightist Rósza paramilitaries aroused little public outrage.
Hypothesis B: The coca eradication conflict is different from other campaigns, either because international pressures on the state from the United States make it unable to make major concessions to coca growers or because (especially after 1997) it is an arena in which the anti-government side also used deadly force.This would explain one case of successful repression (coca eradication, 1995, ID no.11) and the mixed results of the Villa Tunari massacre (ID no.5).
Hypothesis C: When both sides are highly mobilised in a two-sided partisan conflict, outrage against repression is insufficient for repression to backfire.This hypothesis can explain the success of violent repression at Senkata and Sacaba in 2019 and in the 2007 Sucre protests against the Constituent Assembly.All these events were part of protracted conflicts in which government allies also suffered losses, including deaths, at the hands of government opponents.
Quadrant 2: Cases with Deadly State Repression, but where State Security Forces Were Also Killed Quadrant 2 presents outcomes of protest events in cases with deadly state repression and state victims.This second quadrant comprises 13 events that resulted in 198 deaths.Over two dozen members of the security forces were killed in these events. 63Seven of these 13 cases resulted in success for the movement involved, while five saw the government provide no concessions to the movement in violently (if not one-sidedly) repressed conflicts.Counted by event, then, state repression is slightly more likely than not to backfire, and the presence or absence of violent and deadly resistance to it seems to do little to change these odds.
This time, however, the larger cases reveal a different picture: movement success.The Gas War protests of September and October 2003 (ID no.29) were a historically pivotal failure of state repression.Over six weeks of protests, state security forces killed at least 59 people, many of them nonprotesting bystanders; this total includes three people who died of their wounds after 2003 and one two-month-old child crushed as his mother fell fleeing from gunfire.There were at least 12 additional fatalities for which the security forces were not the proximate cause of death.On the other hand, protesters were responsible for the deaths of two members of the military.The second deadliest case is the February 2003 protest against a suddenly announced increase in taxes on public services (known as the tarifazo; ID no.28).In this case, police in high-profile units staged a mutiny in solidarity with the protests and confronted the military in downtown La Paz.
Ten police officers and five soldiers were killed in clashes between these two security forces near the presidential palace in the capital's central square.Funerals held for the police hailed them as heroes of the movement.Altogether there were 29 deaths from state repression, five deaths targeting state security forces loyal to the president, and one killing of a protester committed either by security forces or a private security guard. 64oth these 2003 events involved significant violence and property destruction by nonstate actors, including deaths of seven soldiers.Even as protesters sometimes crossed the line into deadly force in the Gas War, their tactics remained a notch below those of the state security forces they confronted.Some combination of widespread support for the protesters' policy demands and outrage at government's violence against civilians explains the moral outrage that state actions generated during these events.Contravening nonviolence theorists' assumptions, backfire occurred when the public and politicians made distinctions between degrees and types of violence, rather than choosing the purely nonviolent side.
Protesters also won the government's agreement to some or all of their demands in five other cases where the security forces suffered deadly casualties.Two of these were part of the coca growers' movement: the January 2002 Sacaba-Chapare coca market conflict and the January 2003 clashes (ID nos.23, 26).In each of two other cases, the April 2000 protests by Bolivia's largest peasant union, the Confederación Sindical Única de Trabajadores Campesinos de Bolivia (Sole Trade Union Confederation of Rural Workers of Bolivia, CSUTCB) and the January 2003 pensioners' strike (ID nos.17, 27), a member of the security forces was beaten to death in reprisal for prior shootings of protesters.Finally, peasant protests involving the blockade of the Trinidad-Santa Cruz highway (ID no.32) led to the dismissal of the prefect of Beni Department in 2004, but responsibility for the deaths of two protesters is disputed.
There were five further cases in which protesters did not win any concessions.These were the violent crackdown on miners' strikes in Chayanta in December 1996 (ID no.12), remembered as the 'Christmas Massacre' (security forces killed 11, but suffered a fatal ambush that killed a military officer), the resistance to coca eradication in Eterazama in 1997 (ID no.13), a long series of deaths in the Chapare coca conflict from April to October 1998 resulting from protests against the government's national anti-drug 'Dignity Plan'65  (ID no.14), the 2016 strike by members of the miners' cooperative union (ID no.43), and the 2018 coca eradication in Asunta (ID no.44).In the latter two cases, the government used public outrage over protesters' use of deadly violencethe assassination of a deputy minister in 2016 and the killing of three troops charged with coca eradication in 2018as the reason to refuse any concessions.
Quadrant 3: Cases with No Deaths Perpetrated by the State, but Where State Forces Were Killed Quadrant 3 presents outcomes of protest events in cases with no deaths perpetrated by the state, but where state forces were killed.This third quadrant consists of four events, 64 For purposes of analysis, the ten police shot dead while in mutiny in February 2003 and one conscript shot dead when he refused to join in deadly repression in October 2003 are treated as state-perpetrated deaths, while the five soldiers killed defending the president in February 2003 are treated as state victims of deadly non-state violence.
accounting for 17 deaths, in which state security forces were killed but did not perpetrate deadly violence themselves.These cases might easily be overlooked in a study of state repression but can help clarify whether backfire offers the right explanatory framework for understanding protesters' success.Since protesters were responsible for all of the deaths, backfire is thought unlikely.
Nonetheless, in two of the four cases, both involving dramatic violence, concessions were granted to protesters despite their responsibility for deaths.Although six deaths stemmed from coca growers' kidnappings in September-October 2000 (ID no.19), the government and growers concluded an agreement to scale back eradication.In March 2004, Eustaquio Picachuri, a retired miner who had spent years pressing a claim for his pension, strapped dynamite around his waist and entered the National Congress building in La Paz.He committed suicide by detonating the explosives, killing two soldiers in the process (ID no.31).Picachuri's act was interpreted sympathetically as proof of the desperation of Bolivia's retired miners, an emblematic symbol of the country. 66his too led the government to open negotiations and concede a more transparent system for accessing pensions.This case should remind us that a binary division between victim and perpetrator, and dichotomised assumptions about public reactions to violence, may mask important truths about public reaction and official concessions.
In two further cases, the late 2003 coca deaths and the 2013 resistance to coca eradication in Apolo (ID nos.30, 39), violence on the part of the community led to no concessions.As might be expected by theories of both governance and nonviolence, the government responded to these deadly incidents by increasing nonlethal repression in the affected regions.Quadrant 4 presents outcomes of protest events in cases in which the state did not directly kill members of social movements, nor suffer losses to them: 16 events (with 120 deaths).Since the role of the state was different from that in the previous three quadrants, I did not classify these outcomes as success by the movement or the state, but rather classified them as success by the (quantitatively and qualitatively) less violent party or as attracting mediation by the state.(There are no instances of success by the more violent party.)A review of these cases revealed several recurrent patterns: • In six cases, deadly nonstate violence led to public sympathy and state action to meet the demands of the less violent party and/or to protect them against the more violent party: the 1984 shooting at a San Julián blockade by agrarian colonists (ID no.1), the 2001 raid on the Movimiento Sin Tierra (Landless Workers' Movement, MST) at Panantí (ID no.22), the 2008 Porvenir massacre (ID no.36), 67 the 2015 Cruz del Sur mining conflict (ID no.41), the 2016 arson attack at El Alto City Hall (ID no.42), and the 2019 preresignation protests (ID no.45).The first three of these cases are instances of private actors attempting to repress protest.In these cases, the mechanisms of backfire may have worked against private violence.• In two cases, violence between feuding parties led the state to mediate between them.In one of those cases, the 2006 Huanuni clashes (ID no.33), the government acted to merge the feuding parties into a single mining workforce.• In one case, a bus crash during the 2003 pension protest by retired miners, state intervention led to a tragic (but not violent) loss of life: marchers forced by security forces to board buses back to the departure point of the march were involved in a road traffic accident at Panduro (ID no.23).These deaths are similar to those suffered by participants in the 2011 march against the highway that was planned to go through the Territorio Indígena y Parque Nacional Isiboro Sécure (Isiboro Sécure National Park and Indigenous Territory, TIPNIS). 68In both these cases, loss of life (other than acts of deadly violence) followed nonlethal repression by states.In their wake, protesters regrouped and continued to protest, winning state concessions.These cases are another form of backfire to state repression.

67
The government intervention in response to the Porvenir massacre (the militarisation of Cobija; see 'Hypothesis C: Two-sided Partisan Conflicts' in the data page for this article, https://ultimateconsequences.github.io/ultimate-consequences/WLRF-Tables.html#hypothesis-c-two-sided-partisan-conflicts)appears in the larger database as separate event, not analysed here because only two deaths resulted during a brief shootout.The outcome of the combined event was that the governor was removed from office and many members of his armed faction fled across the border to Brazil.The political crisis of October and November 2019 (ID nos.45, 46) is a significant, but difficult to characterise, part of this dataset. 69In late 2019, following highly disputed elections on 20 October, Morales's government encountered a new and challenging mass mobilisation, which ultimately demanded his resignation.Following a police mutiny and pressure from the military High Command, Morales resigned his office on 10 November 2019.With 38 deaths in 30 days, this was the seconddeadliest period of the democratic era.The complex series of events poses delicate questions about lumping or splitting deaths.I opted to characterise the anti-Morales protest wave (which coincided with counterprotests in support of the president) as one event up to his resignation on 10 November (in Quadrant 4; ID no.45).One plausible interpretation of the crisis is that deadly violence by Morales's supporters bolstered the movement against him and contributed to security force defections: a police mutiny and the military leadership's public call for him to resign.
The violence that followed Morales's pressured resignation (on 10-13 November) in various locations forms a second event (in Quadrant 2; ID no.46), whose result was inconclusive.I also chose to treat the Sacaba and Senkata massacres of 15 and 19 November 2019 (ID nos.47, 48) as two further events of one-sided state repression (Quadrant 1).An alternate approach (see Table S5 in the online Supplement) would have clustered all of the violence after Morales's resignation into a single event (in Quadrant 2) in which security forces caused at least 26, and as many as 29 deaths, while suffering the loss of two of their number.In the online Supplement, I include a robustness analysis that shows that these categorisation choices have only small effects on the overall outcomes of the study.
However we group these deaths, as the separate Senkata and Sacaba massacres or as the combined violence of 11-19 November, we might expect such lopsided repression to generate backfire.But these events also match the description of Hypothesis C that has emerged in this analysis, to the effect that repression in partisan conflicts, where both sides are already highly mobilised, is often immune from such backfire.At best, we can see governments in such situations opting to de-escalate their use of deadly repression when public, elite or foreign outrage arises, as clearly happened after the two massacres.This was the choice of Jeanine Áñez's government after Senkata.The repression succeeded on its own terms, even if international outrage may have prevented further killings. 70

Conclusion
Governments engage in repressive violence in order to deter protests, that is to demobilise protesters without conceding their demands.This article's 69 Bjork-James, 'Mass Protest and State Repression' and the International Human Rights Commission (IHRC) et al., '"They Shot Us Like Animals": Black November and Bolivia's Interim Government', IHRC, 2020, http://hrp.law.harvard.edu/wp-content/uploads/2020/07/Black-November-English-Final_Accessible.pdfoffer detailed descriptions of the violence and human rights violations in the 2019 political crisis.comprehensive examination of significant protest confrontations (as marked by three or more deaths) in Bolivia from 1982 to 2021 shows that this effort was usually unsuccessful.I surveyed 48 events in recent Bolivian history to find instances when deadly repression backfired against its perpetrators, ending in success for social movements rather than quelling their mobilisations.Out of 28 cases of deadly state repression (Quadrants 1 and 2), movements succeeded in 13 cases and achieved partial success in one more, and failed in ten.That is, in 57 or 58 per cent of the 23 or 24 decisive cases, movements prevailed despite deadly repression.The cases of movement success (or equivalently of backfire) involved 169 deaths (62 per cent of those killed in decisive cases), and security forces killed at least 136 people in failed bids to quash protests (see Quadrants 1 and 2).
In Bolivia's democratic era, lethal attempts at repression were frequently frustrated by persistent mobilisation, defections in the elite or security forces and/or public outrage.So, backfire happened as it is predicted by its theorists.However, backfire did not only happen when it was predicted.Table 3 summarises the outcomes of causal recipes proposed in the qualitative comparative analysis conducted here.Backfire did not depend on the nonviolence or nonlethality of social movement protesters.The proportions of movement success changed little in the presence of protester counterviolence (i.e., between Quadrants 1 and 2); and movement success was possible even when there was deadly violence against the state but no deadly violence perpetrated by the state (Quadrant 3).
In the analysis of the cases presented above, I proposed three alternative conditions (Hypotheses A-C) to explain when backfire did, and did not, occur.Here I summarise the results, adding in information from smaller (one-or two-death) events in the Ultimate Consequences database.
A. State violence to repress armed groups (guerrillas and paramilitaries) may be immune to backfire.Organised armed opposition forces could not mobilise moral outrage when they suffered fatalities.Both cases of deadly repression targeting guerrillas and paramilitaries (La Paz raid on CNPZ,  1990, ID no.8; raid on the right-wing Rósza group, Santa Cruz, 2009, ID no.37) succeeded.The only smaller event involving armed groups, the 1990 shooting of a (wrongly) suspected member of the Fuerzas Armadas de Liberación Zárate Willca (Zárate Willca Armed Liberation Forces, FAL-ZW) guerrilla group, did not result in backfire.
C. Partisan political conflict (with both sides mobilising) may be immune to backfire.Intense periods of two-sided partisan political conflict were also exempt from backfire.All three instances of deadly government violence in partisan conflict were successful: the repression of anti-Constituent Assembly protesters in Sucre in 2007 (ID no.35), and the Sacaba and Senkata massacres in 2019 (ID nos.47, 48). 71This explanation is consistent with a recent finding that democratic governments have a freer hand to repress 71 While state repression was effective in episodes of partisan conflict, one-sided partisan violence in the 2008 Porvenir massacre (ID no.36) and the 2019 pre-and post-resignation electoral protests (ID nos.45,  46; see above, 'A Note on the 2019 Political Crisis') rebounded to the detriment of the more violent side.concessions despite the deaths of security forces.In all of these cases, popular sympathy, movement persistence and (in all but one case) broad popular participation in protest helped to sway the government towards concessions.This article has offered a qualitative comparative analysis of a sizable number of conflict events.While the causal patterns identified here are significant, they are not enough to comprehensively predict success or failure of the movements involved.Future research using this dataset could further characterise the events involved in terms of the tactical choices of participants, the scale of mobilisation 75 and the movements' efforts to mobilise outrage at repression. 76This analysis is intended as a complement to my prior work on how successful Bolivian mass mobilisations exert pressure on the state, claim the right to represent the public, and even compel incumbent governments to resign. 77Comprehensive coding of smaller events (with one or two deaths each) in the Ultimate Consequences dataset could also provide further clarity on how movement tactics and backfire function on a smaller scale.
Broader Implications and Suggestions for Future Research Recent Bolivian history undermines the conventional story that backfire is closely linked with unanswered state repression.Where civil resistance scholars have focused on the single variable of nonviolent tactics, these protests succeeded through demonstrating their endurance despite repression (or irrepressibility) as well as through diversity and large numbers.Significantly, unarmed militants avoid the limitation, as described by Chenoweth and Stephan, that 'Violent campaigns … cannot gain large numbers of diverse participants.' 78Chenoweth's term for irrepressibility is 'staying power'.In 2021, she defined it as 'cultivating resilience, maintaining discipline, and sustaining mass involvement even as the government cracks down on them [the movements]', and the discipline involved is specifically nonviolent: 'neither fighting back with their own counterviolence or … retreating in disarray'. 79Unarmed militants' irrepressibility represents a different, but still efficacious, form of persistence.To return to the opening examples of this article, the protesters in Santa Rosa del Sara and the Gas War fought back and held (or regained) their ground.Physical resistance characterised all seven cases of (full or mixed) movement success in Quadrant 1 (where no security forces were killed) as well as the seven cases of movement success in Quadrant 2 (where between one and six members of the security forces died in each event in episodes of state repression).As I found ethnographically and in interviews with Bolivian protesters, this on-the-ground failure of repression symbolised the government's vulnerability and often presaged its political defeat.
The evidence shown here reaffirms some existing ideas about the failure of repression, while calling others into question.Civil resistance scholars and nonviolence theorists correctly predict that deadly repression is often ineffective due to backfire: most instances (57-8 per cent) of lethal repression causing three or more deaths failed.They also correctly predict that organised armed actors, whether guerrillas or paramilitaries, cannot claim the same kind of moral outrage when they are killed as can civilian movements.Still, backfire remains an uncertain result and the data eludes a purely Boolean if-then analysis.The paradox of repression remains not that deadly force is useless to quell mass movements, but rather that it fails to meet the expectations of those who wield it.Sometimes dramatic deadly repression fails spectacularly to achieve its desired results, while politicians' choice to employ it puts their political futures at stake.This diversity in outcome shows that causing repression to backfire is a challenging and elusive goal for opposition movements.We should think of backfire as an art, rather than a science, which movements must learn to practice to survive.
However, the current scholarly description of backfire is inadequate.We must refine our view of which movements can succeed by winning moral and mobilisational victories over repression, and of how they do so.This article has found that, in democratic Bolivia, while armed actors are unlikely to benefit from backfire, unarmed militant protesters often do, even when their resistance claims the lives of members of the security forces.Unions of peasants and miners, urban uprisings against privatisation, teachers and pensioners each cultivated a combative profile and harkened back to histories of revolt in their public image (even as they rarely inflicted deadly casualties).Coca growers in particular were cast by successive governments as both criminals and terrorists, but maintained a mass base for mobilisation and allies in civil society, often prevailing after deadly repression.Strikingly, it was this movement that provided leadership to the emergent MAS that was elected to govern Bolivia following the turbulent protests of 2000-5.
These findings may have wider significance given the prominent role of unarmed militancy in Latin American, Middle Eastern and eastern European contexts.Argentina's December 2001 uprising resulted in 39 protester deaths (and none among security forces) during widespread street clashes.In the 2011 Egyptian revolution, 846 people were killed, including 26 police officers.And in Ukraine's 2013-14 revolution, 108 civilians and 13 members of the security forces were killed. 80The common result in these cases was the removal of the head of government as demanded by protesters.In each case, protesters benefitted from backfire despite their use of combative tactics (even including gunfire in the Ukrainian case).Conversely, this article's other conclusionpartisan division can prevent backfiremay also be generalisable.Jean Lachapelle argues that Egypt's new military government in 2013 was able to leverage partisan divisions between Islamists 80 and non-Islamists to avoid backfire following the August 2013 massacre of over 900 Islamist demonstrators. 81n the other hand, popular acceptance of unarmed militancy is an element of national or regional political culture that can vary across geography or be built up through historical experience.In Bolivia, both politicians and the public may compare present-day tactics against past precedents, including Indigenous and peasant uprisings, workers' militias and military rule.This process of local historical comparison can sometimes legitimise current combative action as a continuation of valorised struggles, and sometimes delegitimise government violence as a return to dictatorship.82Across Latin America, traditions of nationalism grounded in independence from European powers, of leftist armed struggle, of mass organising by workers and peasants, of Indigenous revolt and of urban uprisings can all provide cultural referents for contemporary movements that fight back.Further comparative and local studies are needed to find out to what extent this article's conclusions can be generalised to unarmed militant protesters in other national contexts.
When unarmed opposition movements challenge a government willing to use deadly force, they have tactical choices: should they fight uphill battles to hold the streets, or maintain nonviolent discipline to claim a moral victory?Recent Bolivian experience suggests that this choice is less fraught than other scholars have argued.Unarmed militants who engage in deadly confrontations can both prevent their mobilisations from being quelled and win their demands.Moreover, the tragic circumstances of a desperate retired miner or of beaten Indigenous protesters can prove just as morally compelling as actual deaths.Mass participation in unarmed combative resistance, nonpartisan mobilisation, tragedy and moral standing are all key elements that allow movements to succeed despite deadly repression.La violencia estatal represiva, buscando neutralizar a la movilización colectiva, a veces genera una participación mayor de los manifestantes.Cuando la reacción popular y/o la de las élites hace que la parte represora haga concesiones, los académicos trabajando sobre la resistencia civil definen el fracaso de la represión estatal como 'acto contraproducente'.Algunos han propuesto que la disciplina de la no violencia por parte de los movimientos es esencial para lograr los efectos contrarios de los represores.Este artículo demuestra que los movimientos que practican una 'militancia desarmada', es decir, tácticas contundentes y combativas menos dañinas que la violencia armada, pueden también tener éxito y superar a los represores al lograr concesiones políticas e incluso renuncias presidenciales.Este artículo presenta un análisis cualitativo comparado de los resultados de 48 protestas con múltiples muertes en Bolivia entre 1982 y 2019 y un análisis basado en casos sobre cómo los movimientos o sus represores tuvieron éxito.Los movimientos que confrontaron una represión violenta tuvieron éxito en el 57-8% de los casos.El hecho de que los manifestantes participaran o no en una violencia defensiva letal no afectó sus probabilidades de éxito.Sin embargo, la represión estatal hacia las guerrillas o a grupos paramilitares, y durante conflictos polarizados, fue consistentemente exitosa.Los entendimientos actuales del acto contraproducente necesitan ser reconsiderados a la luz de los éxitos de las protestas militantes desarmadas en Bolivia y en otros muchos lugares en todo el mundo.

Figure 1 .
Figure 1.Protests in Bolivia, 1982-2012Note: The two lines on the graph illustrate the distinct trends for expressive and confrontational protests, the latter dropping to near zero between 1986 and 1998.There were many protests in other categories.Source: Author with data from Laserna and CERES, 43 años de conflictos sociales en Bolivia.

60
The focus of Martin, Justice Ignited.61 Charles C. Ragin, The Comparative Method: Moving beyond Qualitative and Quantitative Strategies (Berkeley, CA: University of California Press, 2014), pp.113-18.furthercausal factors in the course of the analysis.In the Conclusion, I reformulate these into a synthetic set of conditions that lead from lethal repression to backfire.Quadrant 1: Cases with Deadly State Repression and Zero State Victims

Notes
In the 'Deaths' columns the shades of grey indicate the number of deaths (the darker the shade, the higher the number of deaths).In the 'Outcome summary' column mid-grey (green online; up arrow) indicates movement success; black (dark red online; down arrow) indicates repression success; other outcomes are indicated by light grey and white, and explained in the respective text boxes CNPZ Comisión Néstor Paz Zamora (Marxist-Leninist guerrilla group) CSUTCB Confederación Sindical Única de Trabajadores Campesinos de Bolivia (Sole Trade Union Confederation of Rural Workers of Bolivia) DEA Drug Enforcement Agency (US) UMOPAR Unidad Móvil Policial para Áreas Rurales (Mobile Police Unit for Rural Areas)

Quadrant 2 .
Cases with Deadly State Repression, but Where State Security Forces Were Also Killed Note In the 'Deaths' columns the shades of grey indicate the number of deaths (the darker the shade, the higher the number of deaths).In the 'Outcome summary' column mid-grey (green online; up arrow) indicates movement success; black (dark red online; down arrow) indicates repression success; another outcome is indicated by light grey, and explained in the text box.CSUTCB Confederación Sindical Única de Trabajadores Campesinos de Bolivia (Sole Trade Union Confederation of Rural Workers of Bolivia)

Quadrant 3 .
Cases with No Deaths Perpetrated by the State, but Where State Forces Were Killed Note In the 'Deaths' columns the shades of grey indicate the number of deaths (the darker the shade, the higher the number of deaths).In the 'Outcome summary' column mid-grey (green online; up arrow) indicates movement success; black (dark red online; down arrow) indicates repression success.

Quadrant 4 :
Cases with No Deaths Directly Caused or Suffered by State Security Forces

Quadrant 4 .
Cases with No Deaths Directly Caused or Suffered by State Security ForcesNoteIn the 'Deaths' columns the shades of grey indicate the number of deaths (the darker the shade, the higher the number of deaths).

70
The Morales government made similar choices after the Sucre Constituent Assembly clash of November 2007 and the Cochabamba ('11 de enero') clashes of January 2007 (ID nos.35, 34).

Table 3 .
Results of Qualitative Comparative Analysis