In July 2011, violence erupted between ethnic political parties in the megacity of Karachi, Pakistan. Fierce gun battles broke out throughout the city as party members and supporters of the Muhajir-centric Muttahida Qaumi Movement (MQM) and the majority-Pashtun Awami National Party (ANP) took to the streets. Unable to stem the attacks, the Karachi police were given orders to “shoot on sight” anyone involved in the violence. Nearly 100 people were killed over the course of just four days, contributing to the over 1,700 people killed that year in ethnic and political violence, 500 of whom were party activists (Gazdar & Mallah Reference Gazdar and Mallah2013). But even this large number underestimates the extent to which violence had engulfed the city; as one news story recorded, “a grisly new feature of the carnage is that people are not just being shot. They are being abducted and tortured …” (The Economist 2011). When I visited Karachi in 2013, no conversation was complete without reference to the possibility of violence. Ethnic tension had transformed many neighborhoods into no-go areas. Crime was rampant. Drive-by target killings were no longer a rare occurrence. And citywide shutdowns called by political parties were commonplace.

While the ANP contributed to these high levels of violence in Karachi, to the north in Khyber Pakhtunkhwa (KP) province, the party had a different reputation and largely refrained from violence during this period. In 2008, the Christian Science Monitor reported on the ANP’s role in tackling extremism there, arguing: “As disciples of the nonviolence espoused by its late founder, Abdul Ghaffar Khan – the so-called ‘Frontier Gandhi’ and follower of the Mahatma – the ANP is uniquely qualified to attempt peacemaking” (Sappenfield Reference Sappenfield2008). For many Karachi residents, this description of the political party would have held little resonance.

In addition to engaging in violence themselves, parties in Pakistan also ally with violent actors for purposes of electoral gain. During approximately the same period in Punjab province, for example, a politician thought to be affiliated with a banned Islamist group – one who had previously been convicted on murder charges – contested national elections on behalf of Punjab’s largest political party, the Pakistan Muslim League-Nawaz (PML-N), and won. While generally considered right of center, the PML-N is not avowedly Islamist, sectarian, or ideological. In fact, it is not unlike political parties that dot much of the developing world: centrist, organizationally weak, and focused on patronage rather than programmatic appeals. Yet, the deadly effects of this electoral alignment were evident within months of the election: A local spiritual leader espousing views antithetical to the extremist group was gunned down along with a number of his followers. Shia leaders publicly blamed the PML-N’s ties to the sectarian groups for the violence, while officials privately expressed concern that the district police officer would be discouraged from pursuing the case (Kaleem et al. Reference Kaleem, Choudhry and Ashraf Butt2015).

Violence implicating political parties is hardly unique to Pakistan. Numerous and diverse democratic actors in Africa, Asia, Latin America, and eastern Europe engage in coercion and even outright violence as forms of electoral irregularities alongside the more commonly studied repertoires of vote buying and patronage. In Nigeria, political parties turn to banned ethnic militias to intimidate and eliminate their opponents during national elections. In Brazilian cities, the intersection of crime and politics has long affected daily life; between 1985 and 2016, politicians were victims of 455 assassinations or assassination attempts (Albarracín Reference Albarracín2021). In Guatemala, the 2007 elections saw numerous attacks against candidates and party activists, including at least 40 assassinations. Levels of interparty violence in Bangladesh in the same year were such that a state of emergency was declared, leading to the suspension of all political activities. Today, Bangladesh’s ruling Awami League frequently uses the police to attack opposition party activists and supporters. In the Philippines, landed elites representing political dynasties – and instrumentally attached to weakly organized political parties – campaign openly with handguns and grenades. Electoral politics also interact with conflict and post-conflict settings in complex, often violent, ways, as in Colombia where paramilitaries provide votes to aligned candidates (Acemoglu et al. Reference Acemoglu, Robinson and Santos2013; Steinert et al. Reference Steinert, Steinert and Carey2019). In Sierra Leone, too, political parties have “strategically remobilized ex-combatants into ‘security squads’ in order both to protect themselves and to mobilize votes” (Christensen & Utas Reference Christensen and Utas2008, 515).

Cross-country data backs up these examples. Eighteen percent of all national elections between 1945 and 2015 experienced violence involving civilian deaths immediately before, during, or after the election (Hyde & Marinov Reference Hyde and Marinov2012). This number was magnified for new democracies, and since 1990, over 30 percent of all elections held outside of advanced democracies have experienced deadly violence (Birch et al. Reference Birch, Daxecker and Hoglund2020; Daxecker et al. Reference Daxecker, Amicarelli and Jung2019). Election-related violence is now one of the most common forms of violence on the African continent; 60 percent of African elections between 1990 and 2008 saw violent intimidation or worse, with about 20 percent involving high levels of violence (Straus Reference Straus2012). Party violence in between election periods is also far from unusual, as in Bangladesh, for example, where levels of interparty violence are high even in nonelection years, and violence often accompanies day-to-day political interactions.

The above examples raise a number of questions about the routine manner in which violence and democracy can intersect. Under what conditions do democratic actors such as political parties engage in, or facilitate, violence? Why do parties refrain from violence in some locales but engage in it in others? What determines the strategy of violence that the party employs, whether directly, by outsourcing, or through alliances with violent actors? What do these nonstate armed actors – what I refer to in this book as violence specialists – gain from their relationship with parties? How do voters respond to such strategies? And, importantly, what are the effects of such violence on prospects for democratic transition and consolidation? That is, when do political parties impede democratization and the process of state-building rather than strengthen it? This book seeks to answer these questions through an examination of party violence in Pakistan.

1.1 Why Parties use Violence, and How They do so: the Argument in Brief

Political parties are an integral component of democratic systems. “[M]odern democracy is unthinkable,” in E. E. Schattschneider’s famous formulation (1942), “save in terms of the parties.” Yet, as the above examples demonstrate, alongside the emergence and consolidation of electoral democracy around the world, we have witnessed both significant episodes of electoral violence and simmering levels of violence that pervade ordinary, everyday political interactions. Indeed, party violence covers a broad set of actions, ranging from intimidation and threats to more overt acts such as assassinations (Straus and Taylor Reference Straus, Taylor and Bekoe2012; Staniland Reference Staniland2014). Parties may also ally with violent nonstate actors, use state forces to carry out violence, and incite their supporters to riot. Such acts take place before, during, and in periods between elections.

This book develops an argument for why and how political parties employ violence by focusing deliberately on the party itself. This “meso-level” analysis complements and bridges the gap between existing works that look at individual microlevel factors as well as cross-national macrolevel analyses. It argues that party violence and electoral alliances with violence specialists are not simple manifestations of weak state capacity; rather, political and economic conditions structure the incentives that political parties have to maintain violence specialists either within their party apparatus or externally. States are then, at least in part, unable to establish a monopoly of violence because nominally democratic actors – competing parties seeking to be electorally viable – have incentives to further inhibit state coercive capacity by maintaining strategic partnerships with violence specialists rather than eliminating them altogether. What is commonly understood to be state failure due to a lack of state capacity is instead the intentional product of particular political incentives, further complicating the process of democratization.

More specifically, this book argues that a party’s particular violence strategy depends on the incentives it faces in the subnational political landscape in which it operates, the cost it incurs from its voters for doing so, and its capacity for violence, highlighting the party apparatus itself. The possible benefits a party can accrue from engaging in violence or allying with violent actors depend in large part on the nature of subnational state capacity where it competes. Where state reach is limited and control is shared – formally or informally – with strongmen who have localized hegemony over violence, weakly organized parties have the incentive to work with such actors because they frequently control votes. In other landscapes of weak state capacity where there is no obvious hegemon and instead multiple competing sovereigns, the benefits to be gained from accessing the state and gaining control over the formal and informal economy raise the stakes of electoral competition. Here, a party’s incentives to outsource violence or to use party members to undertake violence directly remain high.

But violence is not cost-free – voters can, and do, hold parties accountable. Where parties have a captive support base – that is, where electoral support is relatively inelastic and does not change drastically in response to short-term strategies or specific events – they will be less likely to be punished for violence. Such a captive support base is more common in polarized environments, where parties have incentives to organize along certain cleavages or where membership in an identity group allows for access to scarce state resources. Such identities can range from ethnicity, religion, or sect to ideology or even partisanship. In such settings, parties that rely on the support of only part of the electorate and form linkages with voters primarily on the basis of this shared identity, excluding others in the process, will be less likely to incur costs from their core voters for engaging in violence because the barriers to switching parties are high (absence of a penalty). Violence can also further polarize a populace along the salient cleavage, in turn increasing the party’s support among its core voters (an active benefit). While voter costs are paramount, political parties may also have to contend with potential backlash from state institutions, such as the judiciary and military, if violence reaches an unacceptably high level.

Of course, having incentives for violence is not enough; parties do not all have equal capacity to inflict violence. A party’s organizational structure determines both whether a party is able to carry out violence itself or whether it must rely on violence specialists for this task, but also whether the party needs to rely on external actors – violent or nonviolent – for purposes of vote mobilization. I classify parties as organizationally strong when they have a robust party apparatus, local presence, and rely primarily on socialized party workers to contest elections, and organizationally weak when they have little presence, few workers at the local level, and rely on local elites for voter mobilization.

Together these three factors – nature of subnational state capacity, party support base, and party organizational strength – allow us to predict whether a party will engage in violence at all, and if so, one of three possible strategies it will employ (Figure 1.1).

Figure 1.1 Graphical depiction of violence strategies

Direct party violence is carried out by party members and workers at the behest of party leadership against opposing parties, their supporters, or state actors. Direct violence is most likely to take place where the incentives are high (in a political landscape with multiple competing sovereigns), costs are low (where the party has a captive support base), and where the party has a strong organizational structure. In Karachi, the organizationally strong, ethnic MQM is one example of a party that wielded violence through its own party cadres for purposes of coercive gain, economic rent, and to further polarize an already divided ethnic populace.

If parties do not have the organizational structure to carry out violence themselves but have incentives to do so, they must rely on distinct militias, gangs, or armed organizations to implement their directives through the process of outsourcing, as the Pakistan Peoples Party (PPP) did with the People’s Aman Committee (PAC), an ethnic militia in Karachi. Here, the party relies on violence specialists to carry out violence on its behalf. Violence specialists are autonomous actors and may agree to engage in violent acts because they are promised material spoils, such as jobs or contracts; aspire to form connections with state and bureaucratic actors with access to resources; or are promised immunity for their involvement in criminal activity or other acts of violence. For example, the Kenya African National Union (KANU) has in the past outsourced violence to the Mungiki, a religiopolitical movement and criminal gang. In 2002, the Mungiki leadership announced that it would back KANU and even fielded electoral candidates on KANU tickets, ultimately evolving into an “urban vigilante group for hire” (LeBas Reference LeBas2013, 248). In exchange, “they were allowed to take over certain transport routes … hold public processions and threaten opposition opponents with impunity” (Katumanga Reference Katumanga2005, 512).

Organizational capacity does not only affect parties’ ability to engage in violence, but it also affects parties’ ability to mobilize voters. Where local strongmen wield influence and act as patrons to communities, organizationally weak parties may need to rely on them to gather votes and rally constituents on their behalf. These elites therefore function as electoral intermediaries between the party, which lacks party cadres and a localized party presence, and voters. Where the elites possess the means of carrying out violence and do so in order to reach their goals, they are also violence specialists. Organizationally weak parties will form electoral alliances with such actors where and when it suits the parties electorally, in return providing them a permissive environment to carry out violence.Footnote ¹ Paramilitaries, such as militias in Rio de Janeiro, have allied with political parties, “‘contracting’ with traditional politicians to deliver votes in exchange for policy favors and access to confidential information” (Hidalgo & Lessing n.d.); such militias often employ violence to expand their own territorial control or to pursue economic rents. In Punjab, the organizationally weak PML-N has historically relied on landed elites to mobilize voters and provide patronage on their behalf; in recent years, sectarian actors affiliated with violent nonstate armed groups have become increasingly influential in some districts, proving to be important local electoral allies. Unlike the process of violence outsourcing, in alliance formation, the violence is a by-product of a vote-getting strategy rather than a desired outcome in its own right.

Finally, parties may choose not to engage in violence nor tolerate the violent activities of violence specialists in exchange for votes and support. Where organizationally strong parties do not benefit from diminished costs due to a captive support base or where the gains to violence are otherwise muted, they will have little incentive to engage in violence, and, because they are in control of their own microlevel clientelistic structures, they are not forced to ally with violence specialists at the local level. Such parties will tend to refrain from most violence, as does the ANP in Khyber Pakhtunkhwa province.

1.2 This Book’s Contributions and the Limits of Existing Explanations

The presence and variation in party violence that we see in Pakistan cannot be accounted for by conventional explanations emerging from the literature on either civil conflict or political parties. In part, this is because few existing studies have sought to explain the diversity of violent options available to political parties and their determinants, a key goal of this book. Past works have tended to discuss electoral violence and party-armed group relations separately, with the former focusing on the incentives for party-orchestrated violence around election time, particularly vis-à-vis voters (Birch et al. Reference Birch, Daxecker and Hoglund2020; Collier & Vicente Reference Collier and Vicente2012, Reference Collier2014; Lebas Reference LeBas2010; Rosenzweig Reference Rosenzweig2021; Gutierrez-Romero & Lebas Reference Gutiérrez-Romero and LeBas2020; Gutierrez-Romero Reference Gutiérrez-Romero2014; Mares & Young Reference Mares and Young2019; Bekoe Reference Bekoe2013) and the latter exploring such important questions as the relationship of state actors to armed groups, ethnic militias, and criminal gangs (Staniland Reference Staniland2015b; Acemoglu et al. Reference Acemoglu, Robinson and Santos2013; Reno Reference Reno2011; Biberman Reference Biberman and Koch2016, Reference Biberman2019; Arias Reference Arias2017; Lessing Reference Lessing2017; Arias & Barnes Reference Arias and Barnes2017; Turnbull Reference Turnbull2020; Matanock Reference Matanock2016), or on the transition from armed groups to political parties (Berti Reference Berti2013; Weinberg et al. Reference Weinberg, Pedahzur and Perliger2008).

Yet, there are numerous benefits to thinking of party violence as falling along a spectrum. It allows us to better understand the mechanics whereby violence becomes a party strategy as well as the particular goals – electoral, economic, or organizational – that these violent acts seek to serve. In focusing on party strategy as a primary outcome of interest, this book is able to draw attention to organizational factors that have tended to be ignored in favor of other microlevel and macrolevel determinants and hence brings to the forefront the role of political parties in contributing to violence. Indeed, what is missing from our understanding of political violence is how parties themselves organize and structure violence, given the limits and constraints of the political landscapes in which they operate. I contribute to studies that aim to “go within” political parties, examining their socialization and recruitment procedures, decision-making processes, and internal culture. Not all personalistic parties or internally undemocratic parties are the same; rather we see considerable diversity in the socialization of party cadre, presence of party offices and frequency of party events, and the nature of electoral candidates. These organizational differences produce meaningfully different party behavior, including in the realm of violence.

In focusing on party organization, this book deliberately marries the literatures on conflict and civil war, on the one hand, and that on political parties and electoral politics, on the other. Factors such as organizational control (Manekin Reference Manekin2012), indoctrination mechanisms or political education (Hoover Green Reference Green2018), recruitment (Daly Reference Daly2016), and unit discipline and cohesion (Humphreys & Weinstein Reference Humphreys and Weinstein2006) have been highlighted as the best predictors of an organization’s involvement in civilian abuse or the repertoire of violent practices in which it engages (Arjona Reference Arjona2016; Cohen Reference Cohen2016; Wood Reference Wood2009), but are relatively absent in discussions of parties and violence (see Fjelde Reference Fjelde2020 for an exception). While studies have focused on parties’ roles in fomenting ethnic riots (Wilkinson Reference Wilkinson2004; Brass Reference Brass1997), considerably less attention has been paid to political assassinations, intimidation and targeting of opposition parties, and violence accompanying criminal activity. As the case of West Bengal in India – where sitting members of parliament have been injured and killed – shows, even such “extreme” forms of party violence co-exist with otherwise functioning democratic activity.

This book, therefore, offers a new theoretical basis from which to study and analyze the various actors involved in electoral irregularities and which contribute to violence in democratic and hybrid settings: a spectrum from gangs and ethnic militias to landed elites, religious clerics, and militant groups. Understanding the link between parties and violent specialists adds a crucial and understudied dimension to work on countering violent extremism, particularly in so far as it demonstrates the continuation of local-level dynamics in explaining national or even international events. I find that local cleavages and intra-community dynamics – in Kalyvas’s words, the joint “action of local and supralocal actors, civilians, and armies, whose alliance results in violence that aggregates yet still reflects their diverse goals” (2003, 475) – are critical in explaining generalized political violence. Importantly, this book highlights the manner in which coercion accompanies many patron–client relations and other electoral tactics, explaining the conditions under which these various strategies may be complements – working in tandem with one another – rather than substitutes (Dunning Reference Dunning2011). While recent literature on patron–client relations has emphasized their coercive nature (Mares & Young Reference Mares and Young2019; Frye et al. Reference Frye, John Reuter and Szakonyi2019; Hidalgo & Lessing n.d.; Berenschot Reference Berenschot2011; Gallego Reference Gallego2018), that emphasis has focused primarily on the local strongman, instead of the political party that relies on him and his coercive tools for electoral success.

This dual role played by many local patrons – controlling both voter mobilization and the local means of violence – has ramifications for both local-level governance strategies and overall levels of violence. Their presence also raises important questions about the efficacy of democratic institutions, in particular when underlying illiberal institutions are coopted rather than replaced. This book’s findings, thus, advance overall understandings of states that are unable – or unwilling – to effectively maintain the legitimate use of violence within their borders. However, it departs from works that exclusively focus on the role played by state coercive capacity in determining political violence. Rather, by highlighting how different parties – with their own organizational structures and support bases – make distinct decisions about the use of violence, it is able to account for variation in party strategy within the same electoral arena, while still allowing for the same party to use different strategies in different locales.

This book highlights the interaction between four key actors – parties, voters, violence specialists, and state actors. Works that look at just one relationship, such as between parties and voters, miss how that relationship might be conditioned by the presence of violence specialists or shaped by varying levels of state capacity. All the while, this book takes a deliberate approach of centering the party as it seeks to navigate these complicated interrelationships to secure its political goals. In doing so, this study helps make sense of strategies that at first glance may appear counterintuitive and which existing theories are insufficiently able to explain. For example, by highlighting the ways in which parties can benefit economically from the use of violence, this book is able to explain why strong parties, otherwise poised for electoral victory, may nonetheless engage in high levels of violence. Indeed, theories of electoral competition are alone unable to sufficiently account for violence, including criminal and ideologically motivated violence, that takes place in between election cycles. One strand of the electoral competition theory predicts that when competition is high, the benefits of violence outweigh its costs and parties will choose to attack political opponents and intimidate opposition voters. However, greater electoral competition does not necessarily imply that the benefits of violence are also high. It may very well be the case, as other scholars (Wilkinson Reference Wilkinson2004) have argued, that electoral competition forces parties to reach out to minority groups – or form coalitions with parties that rely on minority votes – and hence dampens the prospect of party violence. There is also little evidence that electoral competition is able to account fully for subnational variation in party violence in Pakistan. While Punjab, for example, has seen relatively more close elections in recent years (Nellis & Siddiqui Reference Nellis and Siddiqui2018), this province has neither seen the high levels of violence observed in Karachi nor is violence entirely absent as parties in Punjab compete to ally with violence specialists.

Existing literature has also examined the ways in which polarizing ethnic violence can lead to political gain for political elites and parties who instrumentally use violence to drum up support (Wilkinson Reference Wilkinson2004; Horowitz Reference Horowitz1985; Brass Reference Brass1997). These explanations have tended to see violence as the result of hatred and grievances that politicians are able to easily rile up, but insufficient attention is paid to why such sentiments are open to manipulation. Where explanations for political violence in Pakistan point to the role of ethnicity, the precise nature of this role is similarly left unpacked. For example, Rollier writes that criminal life in Lahore “remains to a large extent distinct from electoral politics” in part because, unlike Karachi, Lahore has a “relatively homogenous ethnic and linguistic composition,” but doesn’t explain why underlying demography matters to this outcome (Michelutti et al. Reference Michelutti, Hoque, Martin, Picherit, Rollier, Ruud and Still2019, 132). At the other end of the spectrum are studies that show that party violence is rarely cost-free and often punished by voters, even when carried out by co-ethnics (Bratton Reference Bratton2008; Rosenzweig Reference Rosenzweig2021; Gutiérrez-Romero & Lebas Reference Gutiérrez-Romero and LeBas2020; LeBas Reference LeBas2010; Banerjee et al. Reference Banerjee, Green, McManus and Pande2014; De la Calle & Sanchez-Cuenca Reference De La Calle and Sánchez-Cuenca2013). Many other explanations of violence in Karachi similarly downplay the role of ethnicity, highlighting instead the combination of the informal economy, a weak state, the presence of arms, and demographic pressures as creating incentives for political actors to engage in violence.

While primordialist accounts of ethnicity and ethnic violence have – rightly – fallen out of vogue, many accounts now go too far in suggesting that ethnicity does not matter to political violence at all. In contrast to both explanations that highlight the inherent importance of ethnicity or disregard it completely, I demonstrate that co-ethnicity matters most where it translates into a captive support base for an ethnic political party. In situations where ethnicity is polarized, voters may perceive few alternative options for political presentation and as such are less likely to hold co-ethnic parties accountable for their acts. In contrast, in environments where such support is more elastic because voters believe they can move to a different party, shared identity will not cushion parties and politicians from voters’ displeasure. This account allows for ethnic parties to use violence in some locales while abstaining from it in others, as does the ANP, an ethnic party which had the capacity for violence in Karachi, but which refrains from employing violence in KP, despite being targeted by the militant Tehreek-e-Taliban Pakistan. Of course, a support base may be captive for reasons other than shared ethnicity. By focusing on options for voter exit, I am also able to explain why nonethnic or nonascriptive parties may sometimes employ high levels of violence, as does the Awami League in Bangladesh, where partisan identity plays a central role in polarizing the populace.

Finally, security dilemma theories argue that in dangerous areas, all actors will eventually take up arms because violence will create new hatreds, resulting in even more violence (Posen Reference Posen1993). Yet, such theories inadequately explain the behaviors described in this volume. These theories assume violence begets violence but do not explain why parties begin the violent action–reaction cycle at the outset given its potential costs. Nor can such theories explain why many parties remain nonviolent even in the face of escalating violence. Finally, even when parties do engage in violence, security dilemma explanations cannot explain why or how they respond to violence, whether directly or by delegating the task to distinct actors. For example, such theories are unable to explain why the PPP chose to outsource violence to an ethnic gang rather than take up arms directly against the MQM in Karachi.

Thus, while my theory falls within a broader group of explanations that emphasize electoral incentives as a primary reason for the use of violence by parties and that put “politics, politicians, and the state” at the center of its analysis of political violence (Wilkinson Reference Wilkinson2013), I differ from these existing explanations in key ways. I argue that violence is varyingly beneficial to parties depending on the type of party carrying out the violence and the nature of its support base. In situations where weak parties ally with violence specialists in order to get access to their vote bank, much of the specialists’ violence is, in fact, a cost that must be borne by political parties who otherwise rely on the votes which such alliances bring to them. Parties also frequently engage in violence for nonelectoral reasons, particularly for economic benefit. While parties still need to ensure that they do not lose core voter support through the use of criminal activity, party violence is not always aimed at achieving narrow electoral goals.

1.3 Scope Conditions

Examples of political parties engaging in violence and violent intimidation, employing violent rhetoric, and working together with thugs, militias, and other unsavory characters come from numerous countries around the world. The majority of countries have experienced some level of electoral and party violence and many have experienced pronounced periods of violent democracy (Daxecker & Jung Reference Daxecker and Jung2018). Party violence is also not a recent phenomenon nor one exclusive to the non-Western world. Scholars of the United States will see familiar echoes in the descriptions that follow to, for example, the nativist Know-Nothing Party that achieved electoral success in the 1850s. Similarly, white supremacist vigilantes targeted newly enfranchised African-American Republicans through the late 1800s, and even today, collusion between local-level politicians and criminal gangs who mobilize vote blocs continues in cities like Chicago (Bernstein & Isackson Reference Bernstein and Isackson2011). However, party violence is multifaceted and complex, and my argument will perform better in some contexts than in others.

In particular, the theory outlined in this book is more likely to apply to a subset of developing democracies characterized by incomplete control of the state. Such states do not possess complete monopoly over the legitimate use of violence; rather, this monopoly is contested by a slew of actors who possess economic and military means to contest the state. In hybrid regimes where control has oscillated between democratic or semi-democratic governments and military administrations, decades of military rule have embedded power in the hands of various local actors, many of whom exist outside of the formal political system. Similarly, in neopatrimonial regimes – which remain the predominant regime type among states in Africa – the existence of oligopolies of violence has prevented the state from effectively establishing a legitimate monopoly on violence. Traditional leaders who maintained local coercive control during the colonial era continue to do so in many parts today, by virtue of their power and resources and sometimes because states have chosen not to enforce their monopoly (Tilly Reference Tilly1992; Acemoglu et al. Reference Acemoglu, Robinson and Santos2013; Carey & Mitchell Reference Carey and Mitchell2017).

Thus, such states are characterized by the presence of groups or individuals who possess the ability to carry out violence and do so in order to obtain some ideological or material benefit either in exchange for or as a result of the employment or threat of violence – that is, violence specialists. Such specialists are able to exercise control over specific jurisdictions and local populations due to a combination of coercive and economic power. These actors exist outside of the state’s ambit but, because of their relationship to political parties and state institutions, occupy the “gray zone” between institutional and noninstitutional politics (Auyero Reference Auyero2007). Michelutti et al. write that these actors “inherently upset the neat boundaries between the legal and the illegal, state actors and nonstate actor, democratic and undemocratic, formal and informal economies, and legitimate and illegitimate violence” (2019, 11).

In such settings, then, routine and everyday democratic politics frequently contain elements of violence. Indeed, violence “may be a common mode of political competition … more common than consensus or co-operation” (Mehler Reference Mehler, Basedau, Erdmann and Mehler2007, 209). In Indonesia, the 1980s and 1990s saw the emergence of preman (gangsters); one such group, the Pemuda Pancasila, helped deliver votes to the political party Golkar and, as a result, was able to influence government policy to its own advantage (Sidel Reference Sidel, Harriss, Stokke and Tornquist2004). In Jamaica, party violence manifested in the political cleansing of neighborhoods, with gangs seeking to “purify” neighborhoods of opposition party supporters, particularly in the 1960s and 1970s. Political identities have crystalized around the leadership of the two primary political parties, the Jamaican Labour Party and the People’s National Party, and have become “the predominant marker of identity” (Sives Reference Sives2010, 2). In Kenya, where parties have outsourced violence to affiliated militias and instigated ethnic riots, 2,000 people were killed and 400,000 displaced in the 1990s (Rosenzweig Reference Rosenzweig2021). This violence was largely controlled by political leaders, who had the “ability to tamp down on local conflict should they so choose” (Rosenzweig Reference Rosenzweig2021).

Even within this broader category of weak state capacity, however, there is considerable room for diversity. Indeed, states may have incomplete control for a number of reasons, and how this incomplete control manifests can vary subnationally (for more on varieties of governance, see Naseemullah & Staniland Reference Naseemullah and Staniland2016). I focus in particular on two types of subnational political landscapes: shared sovereignty, where power is shared with long-standing local strongmen who possess control over local fiefdoms, and multiple competing sovereigns, where no single actor has monopoly over violence. In landscapes of shared sovereignty, state reach is relatively limited, allowing for the continued presence of local elites with whom the state has reached tacit agreements on the perpetuation of control and coercive power. In these contexts, police officials and other local state institutions serve more as agents of individual local elites than as impartial arbiters. In landscapes of multiple competing sovereigns, meanwhile, state penetration is greater, but the state does not possess complete monopoly and there is no other clear sovereign with whom power is shared; in this absence, political actors actively compete with one another for this control and for the associated access to rent and resources.Footnote ²

While this book explores party violence occurring in just these two landscapes, party violence occurs in other political landscapes as well. Within Pakistan’s neighborhood, for example, Bangladesh has seen very high levels of party violence since the restoration of democracy in 1991. Here, the ruling Awami League has control over the levers of state power and is therefore able to operate with relative impunity through the police who can target opposition parties on its behalf. This situation of party capture where a party has appropriated state power is sometimes referred to as a partyarchy, “a system in which the ruling party directs the administration, the police, schools, and budgets through its in loco political boss – the MP” (Michelutti et al. Reference Michelutti, Hoque, Martin, Picherit, Rollier, Ruud and Still2019, 45). In this context, election winners usurp all political power leaving no institutional role available to the opposition, ensuring that the stakes of winning elections are high. Because the ruling party has available at its disposal the institutions of the state, “the police become a political arm of the ruling party, which uses them to harass the opposition, break up opposition rallies while protecting its own, and so on” (Blair Reference Blair and Brass2010, 104). Unlike other weak state contexts where parties seek to skirt the state, here political parties – at least those in government – can utilize the arms of the state to enact violence. While the dynamics in this landscape are distinct, the incentives for violence remain high, and party organizational structure and support base remain predictive factors for party violence.Footnote ³

Finally, in addition to different landscapes of state capacity, different electoral systems may result in different types of violence. The first-past-the-post (FPTP), or single member plurality, system – which makes up about one-third of legislatures globally – returns one representative from each geographical constituency, with the candidate with the most votes winning the seat. Because a small shift in vote can be the difference between winning and losing, election-day and pre-polling violence will likely be more common than in other systems (Rief Reference Rief2009). This is not to say, however, that countries with different electoral systems will not see party alliances with violence specialists or parties engaging in violence and criminal activity for economic gain. However, the particular dynamics of election violence described herein would be most likely to extend to other FPTP contexts.

1.4 Empirical Approach

This book utilizes a within-case design by examining political party behavior and strategies across time and space in Pakistan. Rich subnational variation within Pakistan allows me to construct controlled comparisons of different political parties through an empirically diverse set of cases (Table 1.1). By showcasing where the same party uses different violence strategies in different locations, where ideologically distinct parties engage in similar alliances with violent actors, as well as variation in party strategy in the same city, I am able to keep constant party-specific and location-specific factors. In examining the “demand” side of the equation, I focus on how voters respond to party strategy at the lowest level of voting, the constituency, and how they react to local candidates running in party-based elections to the provincial and national legislatures.

My research approach is deliberately multi-method; I rely on qualitative, experimental, survey, and observational data to answer different parts of my research questions (Figure 1.2). My qualitative fieldwork took place in Pakistan over the course of 22 months between 2011 and 2020, and included semi-structured interviews with about 150 elected officials, party leaders, party members, law enforcement officials, civil society members, potential voters, and journalists; visits to party headquarters, events, and offices; and observation of meetings between party representatives and voters in both rural and urban constituencies.

Figure 1.2 Empirical approach

This qualitative research allowed me a deeper understanding of party strategy, providing me with several specific advantages. First, speaking with party members, attending party meetings, and witnessing interactions between elected officials and voters allowed me to better understand the parties’ internal processes and cultures – a task that would have been incomplete had I merely relied on party mandates and publications. This insight in turn allowed me to develop better measures of party organization – an advantage of field research outlined by Adcock and Collier (Reference Adcock and Collier2001). Second, when studying something as controversial and contentious as violence, actors have reason to obscure their true preferences or beliefs. One-on-one meetings – often numerous times with the same individual – allowed me to probe inconsistencies, ask the same question in different ways, and push further on complex subjects (Wood Reference Wood, Boix and Stokes2007).

I also rely heavily on local news coverage, including Urdu- and English-language sources and both conventional news sources and online sites. These sources provide valuable information about candidate linkages with violence specialists, such as whether party members were seen campaigning with militant actors – unlikely to be admitted to by party members in interviews – as well as party press statements and interviews pertaining to party strategy and policies. Doing so also enabled me to quickly triangulate with other sources – whether other interviewees or news reports and secondary sources – which allowed me to develop appropriate interview questions and approach my research subjects up-to-date.

To help provide microlevel evidence of voters’ responses to violence, I turn to quantitative analysis of a conjoint experiment embedded in a survey carried out among 1990 potential voters in three provinces in Pakistan (Sindh, Punjab, and KP) in 2015. Using an experimental approach allowed me to capture the causal effect of ethnicity, partisanship, clientelistic policies, and, most importantly, party violence, on voter preferences, as well as examine how these preferences varied across various electoral arenas. To borrow a phrase used by Auerbach and Thachil (Reference Auerbach and Thachil2018), the survey was “ethnographically informed.” I used the insights I had gathered during my time in the field to ask relevant questions related to party support as well as to operationalize locally salient candidate characteristics. This task was all the more important given the low literacy rates in Pakistan, which led to the survey team and myself creating pictorial representations of each candidate characteristic in the conjoint experiment (cf. Meyer & Rosenzweig Reference Meyer and Rosenzweig2016).

Conjoint experiments have several advantages over other types of survey experiments. First, they allow us to estimate the causal effects of multiple treatment components and assess multiple causal hypotheses simultaneously (Hainmueller et al. Reference Hainmueller, Hopkins and Yamamoto2014). Second, conjoint experiments mimic realistic choices for voters by, for example, presenting hypothetical pairings of candidates with numerous different characteristics. Rather than asking individuals if they are more likely to vote for one or another parties’ candidate, respondents are asked to choose between two candidates who have not only different partisan affiliations but also different ethnicities or experience in office – just as they would be faced with in real life. Finally, conjoint experiments can help reduce social desirability bias, even if they are unable to eliminate it altogether (Horiuchi et al. Reference Horiuchi, Smith and Yamamoto2020; Franchino & Zucchini Reference Franchino and Zucchini2015). Indeed, one challenge that existing research on voters and violence faces is the possibility of social desirability bias. While conjoint analysis enables researchers to know which components of the manipulation are producing the effects we observe, they do not ask the respondent to explain which of the characteristics is doing the greatest work in determining his/her choice. The bundling together of possible characteristics should help reduce the salience of each individual characteristic in the respondent’s mind. Additionally, respondents may be more willing to be truthful if they believe that their opinion regarding violence can be masked to some extent.

Second, I carried out a survey among 1,805 voters in the violent metropolis of Karachi in the weeks leading up to the 2018 elections. This survey allowed me to collect valuable descriptive data on party-voter linkages, expectations of party violence, and the impact of violence on political behavior. Carrying out the survey in the days just prior to national elections further provided me insight into political views at a time when voters were surrounded by election campaigns, rallies, and breaking news, and when the elections were most salient in their minds.

These voter surveys were coupled with a survey conducted among 251 elected politicians in Pakistan, including members of National Assembly and the four provincial assembles, over the phone in April 2020. Elected legislators were asked about their relationships with the parties that they were representing in parliament, their assessment of party–voter linkages and the party’s support base, their assessment of how voters and party leadership would react to various violent strategies, and their sense of the reputational costs incurred as a result of the violence. This data, which included numerous open-ended questions, provides us valuable insight into politicians’ cost–benefit analysis when making such decisions. Descriptive statistics and research design details of all surveys are provided in the appendix to this book, and the text of the survey questions are included in a supplementary online appendix.

Finally, I make use of a number of datasets to support my qualitative and survey findings and to help establish province-wide trends, including two original datasets. The first, a dataset on the characteristics of politicians contesting national and provincial elections in Punjab, helped me code parties there as organizationally weak or strong. The second, original data on the linkages between politicians and sectarian militants in individual constituencies in Punjab province, allowed me to demonstrate the existence of electoral alliances with such actors and also permitted me to analyze the effect of these relationships. I also make use of existing datasets on violence in Pakistan (Bueno de Mesquita et al. Reference Bueno de Mesquita, Fair, Jordan, Rais and Shapiro2014), exit poll data collected by the Pakistan Institute of Public Opinion (an affiliate of Gallup International in Pakistdaan, hereafter Gallup Pakistan), and data on the urban–rural nature of each constituency shared by Ali Cheema (LUMS and Institute of Development and Economic Alternatives) and Farooq Naseer (LUMS).

The advantages of a multi-method approach notwithstanding, studying violence is never a straightforward task. In addition to the complexities of speaking with party personnel, relying on the testimonies of law enforcement officials carries with it its own challenges. I soon discovered, for example, that it was problematic to assume the direction of loyalties or the nature of power dynamics. The individual who offered to put me in touch with Muhammad Ahmed Ludhianvi, the head of the banned anti-Shia militant group, Ahle Sunnt Wal Jamaat (ASWJ), was a retired senior police officer. As I sat in the police official’s home, he called Ludhianvi on his cell phone. His tone was friendly as he explained that a researcher from the United States was in town and wanted to speak with Ludhianvi about the latter’s organization and political career. I was surprised at the ease and familiarity with which the conversation took place (admittedly, I was only privy to one side of it). On the one hand, Ludhianvi had just contested elections in 2013, so nothing explicitly untoward was taking place. The police official himself had dismissed talk of the ASWJ being militant as “propaganda” and insisted that the “present leadership of ASWJ… [is] not for militant actions.”Footnote ⁴ On the other hand, the ASWJ is believed to be the slightly more palatable political face of the overtly militant Lashkar-e-Jhangvi (LeJ) – and is banned by the state. This one incident demonstrated to me just how murky many of these relationships were. Given the complicated reality of research on such a topic, I attempt throughout this book to indicate where the existing evidence is merely suggestive of a relationship or where the sample size is too small to reach definitive conclusions, and where I think the evidence is more convincing. Ultimately, taken together across different modalities, the evidence points us toward a particular story which forms the basis of this book.

1.5 Why Pakistan?

Political parties in Pakistan utilize violence against a fairly violent backdrop. Indeed, they are just one of many actors engaged in violence, and parties often find themselves at the receiving end of attacks carried out by militant groups or suffer at the hands of the state. Pakistan’s history has been checkered by instability, starting from its bloody partition from India and independence from the British Raj in 1947. In recent decades, it has been beset by at least four types of violence, many of which interact and overlap with party violence – sectarian violence, manifesting largely in the form of anti-Shia attacks; Islamist or jihadist violence which gained currency in the days following the attacks of September 11, 2001; ethnic violence, primarily seen in Karachi; and violence emanating from the long-standing separatist insurgency in Balochistan province.

However, while suicide attacks, assassinations, and violent riots are described as an endemic part of political life in Pakistan, we see significant subnational variation in the frequency and repertoires of violence employed by parties. Indeed, despite the ubiquity of weaponry in the country, not all parties choose to engage in violence and even fewer take up arms themselves. The geography of party violence raises important questions about its underlying logic. KP, for example, remained one of the most violent provinces in the country throughout the early 2000s, but this was not accompanied by equivalent levels of party violence.

Given the range of party involvement in violence across time and space, Pakistan provides a particularly compelling case for shedding light on key questions about violence as an electoral and political tactic in hybrid or unstable democracies. It is, of course, an important case in its own right, as a nuclear power and the world’s sixth-most populous country in a neighborhood of geostrategic importance. It is also important in what it can teach us about patterns of party violence in other contexts. Indeed, not unlike political parties in many parts of the developing world, Pakistani parties have faced decades of regime and institutional uncertainty (Lupu & Riedl Reference Lupu and Riedl2013). The country has oscillated between dictatorship and democracy during its existence, having spent many decades under military rule (1958–71, 1977–88, and 1999–2008). Given the country’s history of civilian acquiescence to military leaders, existing scholarship on Pakistan has tended to focus on the institution of the military. Where political parties are studied, the focus has been on their role as the junior governing partner. There is no doubt that the military plays an outsized role in Pakistan, with parties’ relationships with local elites and power holders in part the result of parameters that have been established by the military. However, while the relationship between parties, violence specialists, and the army has varied over the years, it has not done so in a systematic way that would invalidate the hypothesized relationships presented in this study. The army may get involved when violence starts to spin out of control, but at the “simmering” level, it has remained largely hands off. While I argue that the background distribution of local power brokers affected the initial organizational trajectories of the parties (see Chapter 2), parties have rarely permanently changed their organizational structure in response to army crackdowns.

Pakistan’s democratic credentials are, of course, far from impeccable. Over the years since independence, Pakistan has been varyingly characterized by the Polity dataset as an autocracy, an anocracy, and a democracy (see Chapter 3 for more). And while the military’s role, as will be further outlined, is not to be underemphasized, understanding why and when political parties engage in violence has critical consequences for democratic consolidation. That it is (nominally) democratic actors, operating within a framework of democratic procedures, however flawed, that employ the violence that forms the basis of inquiry of this book, has ramifications for studies beyond hybrid regimes. Indeed, as I show in Chapter 8, even relatively more well-established democracies, such as India, routinely see political parties employ violence.

Finally, not unlike other developing democracies, Pakistan’s politics are quickly and constantly changing. When I began research on this topic, the stakes of the game in Karachi, in particular, were high, with violent incidents happening on an almost daily basis. Since that time, the main actors in Karachi have changed and existing actors have modified their tactics and strategies. Brand new parties have also arrived on the scene, such as the violent far-right Tehreek-e-Labbaik Pakistan (TLP). These developments, however, do not make the core questions motivating the book any less important. If anything – as I explain in Chapter 8 with regard to the Pakistan Tehreek-e-Insaaf (PTI), who governed Pakistan from 2018 till 2022 – the lessons gleaned from the experiences of the parties discussed herein are valuable precisely because they can be applied beyond the immediate case studies. Indeed, the theory presented here identifies combinations of structural conditions and party characteristics that permit us to predict which strategies new actors will adopt. And as discussed in the conclusion of this book, this also means that so long as the underlying conditions do not change, party strategies involving violence will continue.

1.6 Book Plan

The book proceeds as follows. Chapter 2 presents my theory of party violence in more detail. I explain why parties choose to engage in violence despite its costs, which parties are most likely to do so, and the strategy of violence they will employ. I focus on two types of political landscapes of weak state capacity – landscapes of shared sovereignty and landscapes of multiple competing sovereigns – where parties face differing incentives for violence. I explain that the extent to which voters impose costs on parties depends on whether the party has a captive support base. Assuming a party does engage in violence, the party’s organizational structure is key to whether it will do so directly through its own party members or whether it will outsource the task to violence specialists. I also highlight a third way in which parties engage in violence: through electoral alliances with violence specialists. Each of these strategies of violence has different predictions for the nature of the violence that follows. This chapter also explains the origins of these key variables and argues that they are exogenous to party strategy.

In Chapter 3, I provide an overview of the political climate in Pakistan, focusing on the country’s civil–military balance and the nature of subnational state capacity. I explain in more detail the manner in which state institutions, including the police and military, interact with political parties and local strongmen to determine the cost and incentive structure available to parties and how these structures vary across distinct political landscapes. I introduce the four political parties examined in this book, with a focus on their organizational structures and how these structures came to be.

Chapters 4–7 provide empirical tests of my theory. Each chapter focuses on one type of party strategy – direct party violence, violence outsourcing, alliances with violence specialists, and no violence. I deliberately present each chapter as a self-contained whole, believing that there is a story to be told for each strategy and that such a narrative follows the logic of the argument it presents. Throughout these chapters, I make use of my qualitative interviews as well as descriptive and experimental survey and microlevel data to support my arguments.

Chapter 4 examines the phenomenon of direct party violence, examining why and how the MQM engaged in violence in Karachi between 1986 and 2016. I show that the MQM was able to reap the numerous benefits of violence in the ethnically polarized, Hobbesian landscape of Karachi – including generating revenue through land capture and extortion and controlling city resources – without losing the support of its core, Muhajir constituency. The MQM maintained a captive support base among the plurality Muhajir ethnic group, who, until about 2018, perceived few alternative options available for purposes of political representation and who did not therefore punish the party electorally for its involvement in violence. Perpetuating the notion that the Muhajir community was under attack also allowed the party to portray itself as necessary to defend the group’s rights. Survey experimental results based on an original conjoint survey are striking: The likelihood of Muhajirs supporting a violent MQM candidate is identical to the probability of their supporting a peaceful MQM candidate. Relying on my own fieldwork and secondary ethnographic accounts, I show that the MQM used its own militant cadres to target the opposition and engage in turf wars with rival ethnic groups. It was able to do so because it was an organizationally strong political party with committed and socialized party workers willing to engage in risky action. Because its primary support base was relegated to Karachi – and specifically, to Muhajirs in Karachi – it did not have to worry about reputational costs affecting its electoral performance in other regions of the country. Since 2016, two events – a military operation targeting the MQM’s organizational structure and the entry of the multi-ethnic Pakistan Tehreek-e-Insaaf (PTI) – have served to alter both the party’s incentives for violence and its capacity to engage in it.

Chapter 5 turns to the strategy of violence outsourcing by examining a party that shared many of the same incentives as the MQM for carrying out violence but that did not possess the capacity to do so itself. I focus on how the PPP relied on the PAC, an ethnic militia in Karachi, both to meet the demands of its constituents in the neighborhood of Lyari and to carry out the “dirty work” of violence, intimidation, and extortion. It benefited in a number of ways from the use of violence: Through the PAC, it was able to engage in voter fraud and intimidation; engage in criminal activity and “turf wars” with the MQM and ANP over valuable economic property; and maintain support among its co-ethnic base by further polarizing the electorate along ethnic lines. In Karachi’s polarized environment, the PPP had a generally captive support base of Sindhi and Baloch voters in the city who, for a period of time, imposed minimal electoral costs on the party given their lack of alternative options. Despite evidenced principal–agent problems – whereby the PAC challenged the PPP for dominance in Lyari – the PPP relied on the gang because it did not have the local-level organizational capacity to carry out violence – or gather votes – itself.

Chapter 6 focuses on the electoral alliances that the PML-N forms with violent sectarian actors in Punjab province. Using a combination of extensive fieldwork and data on party organization and electoral candidates, I demonstrate that the PML-N is an organizationally weak party lacking a captive support base in Punjab’s political landscape of shared sovereignty. Because it has limited presence at the local level, the PML-N must rely on preexisting, influential patrons who manage microlevel clientelistic structures and can function effectively as electoral intermediaries between the party and potential voters. Historically, these patrons have been landed elites. The PML-N has allied with them, often giving them party tickets on which to contest elections, in exchange for their vote bank. In recent years, anti-Shia sectarian actors associated with violent nonstate armed groups have started to challenge the influence of these traditional elites in many constituencies. As these violent actors have gained local power, they have replaced traditional elites as patrons and electoral intermediaries for the PML-N. In exchange for their help getting local votes, the PML-N turns a “blind eye” to their violence, resulting in their further entrenchment and empowerment. Original data on the linkages between parties and sectarian actors demonstrates that these alliances are electorally beneficial for parties. I end the chapter by looking briefly at another case, that of the PPP in rural Sindh, where similar dynamics hold despite the PPP’s center-left leanings and Shia leadership.

The Pashtun ANP may have engaged in violence in Karachi, but in Chapter 7, I explain why it refrains from violent acts in KP province. KP is a heavily armed province that has faced the brunt of militant violence over the last decade. Despite this, the ANP has chosen not to engage in violence itself or ally with any violence specialists in the area. I rule out explanations for this divergence that center on levels of electoral competition, ideology, or features exclusive to Karachi. Instead, I suggest that the ANP’s distinct violence strategies are a function of the dissimilar nature of ANP support bases in these two areas and the nature of state coercive capacity which affects the party’s incentives for violence. I use a combination of qualitative evidence from Peshawar, Islamabad, and Karachi along with survey data to showcase key differences between Pashtun voters in Karachi and KP.

In Chapter 8, I look beyond these four parties to see what we can learn about other cases through the lens of their experience. First, I look at an out-of-sample Pakistan case, the PTI, a party that was beginning to succeed electorally during the years I was conducting research. I show how the PTI fits my definition of an organizationally weak party lacking a captive support base and engages in violence accordingly. I then provide more detailed explanations of party violence in two countries other than Pakistan – Nigeria and the Philippines – to help demonstrate my theory’s utility in contexts far from Pakistan. Finally, I examine an organizationally strong, ethnic party, the Shiv Sena in India, and assess why and how it engages in violence in Mumbai. In each case, my party-centric variables of organizational structure and party support base are effective in explaining much of the variation that we see in these cases, helping demonstrate external validity beyond Pakistan.

Finally, in Chapter 9, I conclude the book by exploring some of the scholarly, policy, and normative implications of the research topic. I assess the implications of the book’s findings for democratization and democratic consolidation, political party development, and political violence in Pakistan and beyond. I explore limitations of my research and end with numerous policy implications that arise from my findings, in the process outlining several possible avenues for future research.