Legacies of Wartime Order: Punishment Attacks and Social Control in Northern Ireland

Abstract Many armed groups create informal institutions to maintain social order during conflict. The remnants of these informal institutions form a key challenge for governments in postconflict societies in their attempts to reestablish themselves as credible state authorities. The persistence of paramilitary groups’ informal “justice” systems in the form of so-called punishment attacks in Northern Ireland, more than twenty years on from the Good Friday Agreement, offers insights into the legacy of wartime institutions. We argue that armed actors can benefit from the social control wartime institutions grant them long after the conflict ends and both armed actors and civilians are socialized into relying on these institutions. Building on research on wartime institutions, criminal governance, and postwar state-building, we examine how the informal “justice” systems created during the Troubles (1968–98) remain at the fringes of postwar society, drawing on historical works, interviews with stakeholders, geocoded data on “punishment attacks,” and survey data.

they continue to carry out so-called punishment attacks within their own communities. The attacks are typically directed at young men whom the paramilitaries accuse of criminal or antisocial behavior. Most infamously, these attacks come in the form of "kneecappings," as described in the quote above. The attack is eerily reminiscent of the paramilitary groups' brutal informal "justice" system developed during the thirty-year-long conflict known as "the Troubles" , and it is not a one-off-which is a source of public concern. 2 Indeed, data from the Police Service of Northern Ireland (PSNI) show that these attacks, which they refer to as "paramilitary-style attacks," have been a consistent feature in certain locales across Northern Ireland since the Troubles, conducted by both Loyalist and Republican paramilitary groups. 3 Despite the reforms resulting from the Good Friday Agreement in 1998, which have succeeded in preventing a recurrence of the armed conflict, these violent practices persist. But why?
We argue that instead of going away, informal institutions that emerge in wartime have a sticky legacy, both because armed actors can benefit from the social control these institutions grant them long after the conflict ends and because both armed actors and civilians are socialized into relying on them. We develop our argument by drawing on research on wartime institutions, criminal governance, and postwar state-building. We explore how the informal "justice" systems created during the Troubles have longlasting legacies at the fringes of postwar society, drawing on historical works, interviews with stakeholders, geocoded data on paramilitary-style attacks, and survey data.
While empirically focusing on Northern Ireland, this study speaks to broader debates about governance, the rule of law, and state-society relations in postwar peace and state-building. Social control in the hands of nonstate armed groups is characteristic of many postconflict societies, potentially diminishing the credibility of the state as the authority holding the monopoly on the legitimate use of force, or at least the monopoly on the use of force. There is a growing body of work on armed groups' wartime governance, but researchers are only beginning to systematically explore these informal institutions' legacies. In the conclusion, we discuss how the case of Northern Ireland can inform future research. Our studyand the growing body of work on wartime institutions and the social legacies of conflict-points to the importance of reform efforts recognizing that postwar state-building does not necessarily happen in a governance vacuum.

Wartime Order and Its Legacy
Armed actors control populations both by coercing and governing them. Although the Islamic State, for example, exerted its authority over large parts of Syria and Iraq through the use of brutal tactics, it also built a governing apparatus and provided public services. 4 Indeed, a growing body of research explores how, in wartime, armed groups create informal institutions to maintain social and territorial control, sometimes replacing the formal institutions of the state or developing parallel informal institutions of governance. 5 To the degree that civilians in an area have turned to nonstate groups rather than the state for the provision of public goods in wartime, such informal institutions are likely to prevail into the postwar era.
Despite war's chaotic nature, order often emerges. Wartime institutions are the governance structures nonstate armed groups establish to maintain social control and undermine or replace the state. 6 Armed groups may develop and enforce the rules that citizens are expected to obey, and citizens may turn to these wartime institutions for public goods, to ensure their security, or to adjudicate disputes. Armed groups develop wartime institutions, or "rebel governance," to varying degrees, sometimes in competition with one another, often in competition with the state, though sometimes also in cooperation with it. 7 They can also form what Ana Arjona terms "rebelocracy," becoming the de facto rulers in a given territory and thus able to regulate political, economic, and social relationships. 8 At a minimal level of providing wartime order, armed groups attempt to hold the monopoly over the use of violence and provision of security within the communities they control. In Northern Ireland, the main governance function taken on by paramilitary groups was related to informal "justice" systems: punishing perceived criminal behavior within their communities.
The development of informal institutions serves several purposes for armed groups, just as they do for criminal groups. 9 Gangs and organized criminal groups are associated with violent activities and do not shy away from ruling by the gun, but their operations are less costly if they can rely on "their" community's voluntary support. As such, they may provide this community with both private and public goods and, most importantly, protection. 10 For an armed group, even a minimal level of governance, such as security provision and policing, can solidify the group's strength by providing a pool of potential recruits and a safe haven for both the rank and file and leaders. Policing their own community also helps avoid defection and deter informers. Groups fighting the state need to organize against an often much stronger enemy and convince both the population in whose name they are fighting-and even the international community-that they are a viable alternative to the state. As Robert W. McColl notes, the development of "insurgent states" signals the group's strength vis-a-vis the state and provides it with "at least an aura of legitimacy," likely to boost recruitment and support and prevent defection. 11 Indeed, although armed groups, not unlike criminal groups, are specialists in the production of violence, it is less costly to control a population not only based on coercion but also consent. Conflict-affected areas where the state does not exert control are rarely ungoverned spaces. As Joel S. Migdal argues, "Social control rests with the organizational ability to deliver key components for individuals' strategies of survival." 12 These strategies, stitched together by individuals' quests to fulfill material needs and the myths and symbols they use to make sense of the world, can be met by the state (or segments of it) as well as nonstate groups. Armed groups "may be coercive, even brutal, but they may nevertheless be perceived as legitimate by many in the populations under their control." 13 Research on criminal gangs reveals similar dynamics, where gang members are both feared and revered due to the order they provide. 14 9 Mart ın S anchez Jankowski, Islands in the Street: Gangs and American Urban Society (Berkeley: University of California Press, 1991). 10  Just as formal institutions can be "sticky," so can informal institutions. 15 We argue that wartime orders are hard to dislodge due to both top-down and bottom-up mechanisms. Armed groups and civilians maintain the informal institutions that developed during a conflict either because they are socialized to rely on these institutions or, more instrumentally, they benefit from them.

Top-Down: Armed Actors
Top-down, wartime institutions are likely to prevail into the postwar era as armed actors instrumentally benefit from them or are socialized to have a certain position of influence and uphold certain norms in the community. Instrumentally, armed actors can use the legacy of wartime orders to continue to undermine the state. They may do so for political reasons, because they are opposed to a peace agreement or dissatisfied with their own position of power in the postwar state, 16 or because they want to control a territory for criminal operations. Regardless of motive, by exercising social control in areas where their armed group was rooted historically, armed actors can better control information, mobilize support, and recruit new members. Indeed, as noted by McColl, Mao's first principle in the effective location and application of guerrilla base operations is that "revolutionary activity should be concentrated in areas with previous political or revolutionary activity." 17 It takes time for any armed actor to build up an effective reputation and monitoring capacity within an area; hence there is a reluctance to relocate. 18 Put simply, it is easier to control an area where informal institutions were developed during conflict.
Socialization-a "process of inducting actors into the norms and rules of a given community, the endpoint of which is internalization" 19 -may also be at work. Members of armed groups may have become used to positions of power or a certain standing within their communities. They may find it hard to let go, even after the groups are officially demobilized, particularly if they feel marginalized in the postwar political system. 20 Research on criminal groups points to how gang members "understand that their social status in life comes from criteria established in their community, not from criteria of the larger society, and that the best way to attain this status is to be helpful to the community." 21 Similarly, members of armed groups may have come to internalize that it is their role to uphold the norms of wartime institutions. 22 This is likely to be the case in areas where armed groups establish institutions over decades. As wartime institutions are likely to increase the pool of local recruits over time, the cadre will be predominantly local and have internalized the norms and rules in their communities. Once internalized over the course of a long conflict, such expectations and norms about one's role in society are sticky.

Bottom-Up: Civilians
Bottom-up, wartime institutions are likely to prevail into the postwar era as civilians become socialized to rely on informal institutions for their strategies of survival. In areas where armed groups developed public goods provision and did so for a long time, members of the community may have come to accept and internalize norms stipulating that governance, or certain aspects of governance, is done-and done better-not by the state but by nonstate groups. 23 There is a "normalization" of nonstate governance. It is not a given that civilians approve of the wartime order imposed by armed groups; 24 hence armed groups may have taken steps to actively build legitimacy not only by providing public goods but also by fostering inclusion, belonging, and identity. 25 Writing about the legacy of armed networks in Colombia, Sarah Zukerman Daly argues that, if recruitment into armed groups happened locally and former combatants stayed in the same areas after the war, not only are former combatants likely to remain embedded in the local community, but civilians are likely to continue to look to them as sources of authority. 26  course of the long conflict in Guatemala, a process of "civilian socialization" has fostered tolerance for extralegal security patrolling as a crimeprevention measure in the post-civil war period. 27 For many postwar states, the process of (re)establishing confidence in the state's authority presents an enduring challenge. 28 Agents of the state may be perceived as either directly responsible for the conflict and violence committed or, more indirectly, unable to uphold law and order, protect citizens' security, and provide public goods, none of which bodes well for people's confidence in the authority of the state. In societies where armed conflicts end through a peace settlement and governance reforms, one might expect that civilians (again) come to be socialized into seeing the state as the legitimate upholder of law and order. 29 However, where informal wartime institutions developed, the postwar state is exerting its authority in competition with informal authorities. Indeed, the targets of socialization are "never blank slates." 30 Thus, if armed groups governed and civilians relied on them for their strategies of survival in wartime, civilians may, despite the state's reform efforts, continue to do so into the postwar period.
Two propositions follow from these mechanisms. First, whether armed actors have instrumental reasons to control certain communities or are socialized into doing so, we expect persistence in the geographic locations of informal institutions from the conflict into the postconflict period. Postwar, armed actors will seek to exercise social control in areas where well-developed informal institutions existed during the war. Second, if the persistence of informal institutions from the conflict into the postconflict period is also driven by civilian socialization, we expect civilians living in areas controlled by armed groups to be likely to rate informal authorities highly and be skeptical of formal authorities. To the degree that informal institutions are sticky, both propositions should hold long after the armed conflict has ended.

Research Design: Assessing Legacies of Wartime Order
We assess these propositions-and the causal mechanisms underpinning them-in a mixed-methods case study of Northern Ireland. We proceed in three steps. First, we present a historical narrative that anchors our study within existing research on Northern Ireland and draws on fifteen semistructured interviews (conducted in 2018) with relevant stakeholders, including affected community members and restorative justice organizations (see the online appendix). 31 This step allows us to establish how and where wartime social control developed and what it looked like, informing the measures we use in subsequent analyses. Furthermore, it allows us to demonstrate that these informal institutions developed during the conflict itself, not before.
The second step of our analysis examines our first proposition, which suggests that there is a persistent geographic pattern in paramilitary groups' social control from the Troubles until today. We assess whether paramilitary groups' social control during the Troubles is associated with the location of paramilitary-style attacks today while also accounting for other variables that can account for the occurrence of such attacks today, not least deprivation. To do so, we rely on a dataset of in-group killings from the Troubles and a unique dataset of geolocated paramilitary-style attacks that have taken place after the Good Friday Agreement. As the historical narrative shows, key to paramilitary groups' effort at establishing social control both during and after the conflict is informal "justice" systems in the form of so-called punishment attacks-both beatings and shootingswithin their communities.
The third step in our analysis examines our second proposition, which suggests that civilians' socialization to rely on informal justice institutions persists into the postconflict period. To do so, we combine the geolocated data on paramilitary-style attacks in the postconflict period with data from a nationally representative survey (N ¼ 811) conducted in Northern Ireland in spring 2016 (see the online appendix). 32 We conduct statistical analyses examining respondents' rating of informal and formal authorities, probing whether their perceptions are associated with the prevalence of paramilitary-style attacks in their neighborhood. Although we cannot establish causality, these analyses allow us to explore whether communities experiencing paramilitary-style attacks today are simply coerced or, alternatively, whether people have come to perceive informal authorities as an effective alternative to the formal apparatus of the state, which is indicative of the persistence of a bottom-up socialization mechanism. The study is motivated by the Northern Ireland case and a quest to understand the persistence of paramilitary-style attacks there, which seems puzzling for several reasons. The Good Friday Agreement is a comprehensive peace agreement that took steps toward both disarmament, demobilization, and reintegration (DDR) and security sector reforms (SSR). Decommissioning of paramilitary groups was part of the Good Friday Agreement, though the official process was not complete until 2010, and there are still dissident groups that reject the peace process. 33 The police was a key and controversial actor during the Troubles but has undergone significant reform to enhance its legitimacy. 34 Arguably, these steps, along with significant political reforms to ensure power sharing between Nationalist and Unionist political parties, have helped prevent a recurrence of the armed conflict. 35 Yet they have not prevented paramilitary groups from still exercising social control within a strong and democratic state. Thus, at the outset, Northern Ireland seems like a least likely case for observing the continuation of armed groups' social control after the war ends. 36 Alternatively, these are precisely the conditions that allow armed groups to continue to exercise social control. Thus, in the conclusion, we turn to an in-depth discussion of whether conflict termination, state strength, and regime type are scope conditions of our argument, pointing to avenues for future research. Though there is a growing body of work on armed groups' wartime governance, researchers are only beginning to systematically explore the legacies of such informal institutions. 37 The Troubles: "No-Go" Areas and Informal "Justice" The communal violence that marked the beginning of the Troubles in the late 1960s paved the way for the emergence of new paramilitary groups and the resurgence of those that had been inactive since the partition of Ireland and the Irish Civil War in the early 1920s. In this section, we draw on historical works and interviews to show how, during the thirty years of the Troubles, both Republican and Loyalist groups developed informal "justice" systems in certain areas of Northern Ireland. Central to the conflict were economic difficulties both the Catholic/ Nationalist/Republican and Protestant/Unionist/Loyalist communities faced. Industrial stagnation in the 1960s undermined the historical economic dominance of the Unionist community, and grievances in Protestant working-class communities were often directed toward Catholic communities. Catholic working-class communities, on their end, faced unemployment and discrimination. The civil rights movement that emerged targeted the historical inequalities between the Protestant and Catholic communities. Although it was not intended to be a Catholic movement per se, it became one when Unionists reacted to the popular mobilizations they feared would undermine their community's position. Marches were faced with countermarches, leading to violent clashes, most infamously during the People's Democracy March from Belfast to Derry in January 1969. 38 The police, the Royal Ulster Constabulary (RUC), did little to defend the marchers and often sided with Unionists, becoming a symbol of an oppressive state for the Catholic community. 39 By 1969 it was clear the RUC had lost control, and the British Army was deployed to secure order. Initially welcomed by the Catholic community, the army's use of violence, most infamously during the 1972 peaceful protest march in Derry that became known as "Bloody Sunday," signaled to the Catholic community that the army was there to maintain the status quo. 40 From the early days of the Troubles, both Republican and Loyalist groups developed informal "justice" systems, particularly in the segregated working-class areas of Belfast and Derry. They did so in response to demands from their communities, as the RUC was either unable or unwelcome to provide for citizens' security, and as a strategy of control. Barricaded areas became known as "no-go" areas where the police did not enter, exacerbating a diminution of state authority. 41 No-go areas that 38 Members of the Nationalist community call the city "Derry," and members of the Unionist community call it "Londonderry." For simplicity, we write "Derry. emerged came to be policed by local defense groups even after the barricades were dismantled. Although paramilitaries did not necessarily set up the groups, the former quickly took on the community policing role.
No-go areas in Republican communities effectively came under the control of Republican armed groups, predominantly the Provisional IRA (PIRA). 42 In Belfast's Republican no-go areas of the late 1960s, community councils that included justice, development, and welfare committees were established. 43 Cases of minor offenses came before People's Courts, which decided on the punishment to be enforced. Paramilitary groups in both Republican and Loyalist communities developed a system of punishments based on the seriousness of the offense under consideration, 44 ranging from warnings, curfews, fines/victim restitution, acts of public humiliation (such as "tarring and feathering"), punishment beatings, punishment shootings (such as kneecappings), expulsions, and assassinations. 45 In Republican areas, paramilitaries would either call on people in their homes to issue warnings or "deliver" punishment, or those accused of having committed an offense-for example, a violent crime or antisocial behavior-would be called to a formal meeting, often held at a Sinn Fein office. 46 These structures were "rudimentary justice systems," 47 but there is little doubt they were informal governance structures.
Republican no-go areas ended after Operation Motorman in July 1972, when more than thirty thousand British Army personnel were employed to reoccupy no-go areas in Derry and Belfast. 48 The RUC was no longer deployed to police these areas, and policing was done almost entirely by the British Army. 49 The army's policing in Republican areas did not, however, include everyday forms of policing, such as, for example, addressing antisocial behavior. As a result, the social control exerted by Republican groups did not end with Operation Motorman. Areas under Republican paramilitary control became what Allen Feldman describes as "sanctuaries," areas "reserved for residence and kinship." 50 They included areas such as the Ardoyne, New Lodge, Falls, and Short Strand in Belfast, and Creggan and the Bogside in Derry. It is within this context that Republican governance structures continued to exert social control in Catholic working-class urban areas throughout the conflict. The main public good was community "policing" via the informal "justice" system-though, then as now, significant doubts exist as to how just this process was and whether it worked to prevent crime. 51 Other governance functions taken on by nonstate groups in Republican areas included Irish language schools, welfare provisions to the families of political prisoners, and, as the city's buses were prevented from serving these areas, the black taxicab service. Whereas Republican paramilitary groups emerged to defend Catholic communities and to establish a thirty-two-county Ireland, Loyalist groups emerged to defend Protestant communities from the PIRA and to maintain the Union. Although Loyalist groups were ostensibly on the side of the state, they, too, came to take on community policing roles, administering informal "justice" when they thought the state did not act quickly or go far enough. Unionist working-class communities-in area such as East Belfast, Sandy Row, Tiger's Bay, and Shankill Road-felt the absence of the RUC and looked to the paramilitary groups to take on community policing. 52 Loyalist groups' relationship with the state deteriorated with the Anglo-Irish Treaty of 1985, when the British government shifted its historic alliance with the Unionist community in exchange for a stronger partnership with the Republic of Ireland, which at the time still had a territorial claim on the six northern counties enshrined in its constitution. 53 Similarly, with the Good Friday Agreement, many working-class Protestant communities felt left behind and betrayed by the state and the Unionist elite. 54 In both communities, the development of informal "justice" systems was a direct result of the conflict. In Republican communities, the heavyhanded tactics of the RUC led people to mistrust the police and seek alternative authorities for addressing their everyday problems. As several of our interviews suggested, "you just did not call the police." This was either because the police was not present, you mistrusted them, or you feared being labeled a "tout," an informer and traitor within your community. 55 In a vicious-cycle dynamic, the strengthening of alternative authorities meant the police became unable or unwilling to enter certain areas. This strengthened paramilitary groups' control and a sense among the 51 Interviews with community workers, former paramilitaries, and members of affected communities, 28-31 August 2018. Many campaigners argue that the terminology of "justice" system and "punishment attacks" is misplaced. population that the police was simply absent. The RUC was also perceived to be absent in Loyalist areas, fostering distrust. 56 As described to us by one former Loyalist paramilitary, since 1969, there was simply no (formal) policing in Belfast. 57 In both communities, paramilitary groups were initially reluctant to take on policing duties but did so in response to demand from their communities. 58 Strategically, informal policing allowed paramilitary groups to control "their" communities. Politically, communities controlled by Republican paramilitaries served to undermine the legitimacy of the state. 59 For example, in one night in 1992 the PIRA killed a drug dealer and kneecapped another ten, an activity that allowed it to "project itself as the defender of the community." 60 Militarily, these areas provided Republican groups with sanctuary and became notorious as centers of IRA recruitment, organization, and arms concealment. 61 Control of information within the community was ensured through a combination of intimidation, punishment, and rewards, which identified and eliminated informers or "touts." 62 By the late 1970s, the British security forces confronted an insurgency that had built a deeply rooted microsociety within sections of the Nationalist community. 63 For Loyalist groups, "punishment attacks" were also a strategy to maintain discipline and control within their ranks. 64 Although armed groups came to benefit from controlling their communities through an informal policing role, it is simplistic to conclude that paramilitary groups imposed informal "justice" systems on these communities through coercion only. During the thirty-year-long conflict, parts of both communities demanded some form of justice system, and in time came to rely on it. This developed into a "folk memory."

Persistent Paramilitary Social Control
Elements within both communities rejected the Good Friday Agreement. In the Republican communities, some believed the PIRA had failed to create a united Ireland and remove the British presence on the island, whereas members of the Loyalist community felt that too many concessions were given to the PIRA and Sinn Fein. 66 Today, there are also groups that act in the name of paramilitary groups but employ violence for criminal reasons-for example, by running extortion rackets or controlling the drug trade. 67 In this section, we statistically test whether a pattern exists in where these groups exercise social control from the Troubles until today and provide evidence from interviews and existing research on how the pattern is maintained. While deprivation, particularly in urban areas, is associated with present-day paramilitary-style attacks, we show that armed groups also rely on the legacy of wartime institutions.
We rely on a unique dataset of geolocated paramilitary-style attacks from the PSNI to establish where they have taken place after the Good Friday Agreement and then assess whether these attacks happen where informal "justice" systems existed during the Troubles. The PSNI shared their data on paramilitary-style attacks from 1999 to October 2018 with us. The data show where and when attacks happened but include no personal information about perpetrators or victims. 68 Figure 1 shows that from 1999 through 2017, there were 2,789 paramilitary-style attacks in Northern Ireland. The PSNI code most (97.7 percent) as directed toward the paramilitary groups' "own" communities: Republican groups attacking members of the Catholic community and Loyalist groups attacking members of the Protestant community. The attacks include both assaults (involving "major or minor physical injury to the injured party typically involving a group of assailants armed with, for example, iron bars or baseball bats") and shootings (which "usually result in the injured party being shot in the knees, elbows, feet, ankles or thighs and the motive is supposedly to punish the person for anti-social activities"). 69 Whereas the immediate years after the Good Friday Agreement saw up to and more than three hundred attacks per year-some of them due to turf wars between armed groups-yearly numbers of attacks have declined over time and remained relatively steady for the last decade.
Indeed, what is remarkable is the persistence of paramilitary-style attacks from 2008 onward. These attacks are not merely an immediate leftover of the conflict. They continue to happen after the long process of reforming the police to address its lack of legitimacy, including the Sinn Fein leadership's historic vote to recognize the PSNI as the legitimate police force in January 2007, as well as the Ulster Volunteer Force's and Red Hand Commando's announcement of the end of their campaign in May 2007. 70 Similar data on paramilitary-style attacks do not exist for the period of the Troubles. Instead, as an indicator for the presence of informal "justice" systems during the conflict, we rely on data on deaths as a result of paramilitary groups attacking members of their own community. In-group deaths represent the tip of the iceberg in terms of the informal "justice" systems paramilitary groups developed during the Troubles-ignoring nonlethal forms of punishments-and are a conservative measure for their social control. The data come from the Sutton Index of Deaths, which traces all conflict-related deaths during (and after) the Troubles. 71 We rely on and update Neil T. N. Ferguson's compilation of the data, parsing out the in-group deaths-that is, Republican groups killing Catholics and Loyalist groups killing Protestants-and their location. 72 The data include 519 in-group killings between 1969 and 1998. We are not the first to systematically examine the legacies of the Troubles in Northern Ireland or expect a persistent pattern in paramilitary groups' activities across space and over time. 73 However, our study is the first to systematically link the present-day pattern of paramilitary groups' social control to their social control during the Troubles.
To get a sense of the geographic pattern, Figure 2 shows three maps of Belfast: community divisions attributed to the level of super output areas (SOAs), in-group killings during the Troubles (1969-98), and paramilitarystyle attacks over a decade on from the Good Friday Agreement (2008-18). Given the localized nature of paramilitary groups' control, we present the data and conduct the analysis at the level of SOAs. Northern Ireland has 890 SOAs, with an average population of 2,000. The maps illustrate how armed groups on both sides continue to undertake attacks directed at their own communities in specific areas of Belfast. Most illuminating for our argument, even ten to twenty years after the conflict ended, armed groups continue to carry out so-called punishment attacks in areas where paramilitary groups set up informal "justice" systems during the Troubles. They include areas such as the Falls, Springfield, Andersonstown, New Lodge, Shankill, and Sandy Row. Outside Belfast, this pattern is consistent in towns such as Bangor, Colerain, Derry, Lisburn, Newtownabbey, and  Strabane. 74 The overlap in areas of operation is clearer for Republican than for Loyalist groups. About 60 percent of Republican paramilitary-style attacks in the 2008-18 period occurred in areas that experienced Catholic in-group killings during the conflict. Loyalist groups, which operate in more SOAs than their Republican counterparts, 75 seem less likely to operate in areas where they historically developed informal "justice" systems (about 30 percent). Both in Belfast and across Northern Ireland, Loyalist groups today operate in more areas than Republican groups, branching out to some they did not control during the Troubles, typically deprived areas that did not get their share of the "peace dividend." 76 To probe our first proposition statistically, we examine whether our indicator for informal wartime institutions is associated with postconflict informal "justice" systems also when accounting for other variables that may explain whether an area is susceptible to armed groups' social control today.
The dependent variable is the number of paramilitary-style attacks in an SOA from 2016 to 2018. We have chosen this time period due to data availability and to give our argument a hard test: namely to see if wartime institutions have a legacy almost two decades into the postconflict period. 77 There were 245 attacks in this period, occurring in 16 percent of SOAs. Loyalist paramilitaries were responsible for 62 percent of the attacks, and Republicans, 38 percent.
The key independent variable is an SOA's number of in-group killings from the Troubles. This is not to say we believe the legacy of informal institutions from the Troubles is the only driver of paramilitary-style attacks today. Existing research argues that since the Good Friday Agreement dissidents and criminal elements among both Republican and Loyalist paramilitaries have found support in segregated and deprived communities, particularly in urban areas, left behind by the peace process. 78 We do not disagree that paramilitary groups today tap into deprivation in their respective communities, but our argument highlights the legacies of wartime institutions from three decades of conflict. To assess deprivation, and particularly urban deprivation, we include a measure for the proportion of the population in each SOA living in households whose equivalized income is below 60 percent of the country median well as a dummy indicator for whether an SOA is urban (taken from the 2011 census).
Notably, if we look at the most deprived urban areas (coded as urban SOAs in the top 10 percentile of income deprivation) more than half (52 percent) did not experience any paramilitary-style attacks in the 2016-18 period (if we include rural deprived areas as well, the share is 64 percent). Nor do paramilitary-style attacks take place only in the most deprived areas. That is, deprivation alone cannot account for the occurrence of these practices. Descriptively, when we take the history of in-group killings into account, the average number of attacks per SOA is much higher in the deprived areas that have such a history (0.89) compared to the deprived areas without such a history (0.37). 79 In the statistical analysis to follow, we also control for other variables likely to be associated with an SOA experiencing paramilitary-style attacks, including the rate of violence, sexual offense, robbery, and public order, as well as the rate of antisocial behavior incidents (both taken from the NIMDM indicators). Other controls include measures for total population and whether the area is a community stronghold, which we assess by indicating whether more than 90 percent of the area's Christian population is from either the Catholic or Protestant community (both taken from the 2011 census). 80 Tables 1 and 2 show the results of logit regressions and negative binomial models, respectively. The logit models examine whether in-group killings from the Troubles are associated with paramilitary-style attacks in the postconflict period, and the negative binomial models examine whether they are associated with the intensity of postconflict paramilitary-style attacks. 81 As a robustness check, we run the same models using the two alternative measures for paramilitary groups' wartime social control (see the online appendix).
Though there is some variation between Republican and Loyalist attacks, we find that areas with income deprivation, urban locales, and community strongholds are generally associated with a higher number of paramilitary-style attacks. This is consistent with existing research, though the findings for violent crime and antisocial behavior are less conclusive. We find that there is a legacy of wartime order. Areas where informal "justice" systems existed during the Troubles, as assessed by in-group killings, are more likely than areas without such a legacy of wartime social control to experience paramilitary-style 79 We see a similar pattern when we look not only at the most deprived areas (in the 10th percentile) but in the top half of income deprivation: relatively deprived areas have a higher average of paramilitary-style attacks a decade after the Good Friday Agreement if they have a history of in-group killings from the Troubles. 80 For more information on these indicators, see the online appendix. 81  attacks long after the conflict came to an end. Not only are they more likely to experience paramilitary-style attacks, the count models indicate that areas controlled during the conflict are also more likely to experience many of them. Existing research and our interviews indicate that both the top-down and bottom-up mechanisms drive the pattern we observe. Both Republican and Loyalist armed groups have instrumental incentives to exercise social  control. They do so to show they are politically relevant to their communities-either through compliance or fear-for operational reasons. The Republican Action Against Drugs, for example, targeted "suspected drug dealers, sex offenders, and other criminals within their communities." 82 As during the conflict, this strategy aimed to undermine the legitimacy of the state, while promoting themselves as the rightful guardians of peace to "gain the support of elements of the local population while simultaneously taking control of others through fear and retribution." 83 Neither side is free from groups acting with criminal intent, engaging in extortion or trying to control the drug trade. 84 Whether for political or criminal reasons, existing research suggests that armed actors maintain informal "justice" systems for instrumental reasons.
There is also some indication that armed actors do so as they have come to see themselves as having a certain role in the community. This was clear following the brutal murder of Paul Quinn in 2007, who former PIRA members accused of criminality. The Independent Monitoring Commission (IMC) described the suspected murderers as "accustomed over a substantial period of time to exercising considerable local influence, collectively and individually" and that they "would find it very difficult to accept any waning in this influence and respect." 85 After the conflict, the PIRA stopped policing communities and left behind a policing vacuum. Drugs, which had historically been kept out of most urban and poor neighborhoods in Northern Ireland due to the high-security situation and the "policing" armed actors undertook, became a major problem. A "new crop of Republicans" emerged after the conflict to fill the void, 86 and they ensured that paramilitary groups continued "to do what they had always done, namely doing violence against their own communities," 87 targeting those involved in drug dealing and antisocial behavior.
Our statistical analysis, combined with interviews and existing research, provide support for our first proposition. Deprivation, community strongholds, and urban areas are, as existing research suggests, associated with present-day paramilitary-style attacks, but we show that long after the conflict officially ended, armed groups also continue to rely on the legacy of wartime institutions. Indeed, among deprived areas, the average number of attacks is much higher where paramilitary groups can draw on a legacy of social control from the Troubles. Not only do paramilitary groups benefit from controlling the same areas they controlled during the Troubles for instrumental reasons; they were socialized into doing so over a long conflict.

Assessing Bottom-Up Support for Informal Institutions in the Postconflict Period
We proposed that if the persistence of informal institutions from the conflict into the postconflict period is driven also by civilian socialization, civilians living in areas controlled by armed groups are likely to rate informal authorities highly and be skeptical of formal authorities. Our interviews, specifically with restorative justice organizations in Belfast and Derry, provide initial evidence of such a bottom-up mechanism. People from both Republican and Loyalist communities have come to expect the "swift" or "speedy" justice conducted by paramilitary groups during the Troubles and may be disappointed by what is perceived as a "slow" criminal justice system that is too lenient and "not going far enough." 88 There is a "folk memory" of going to the IRA if there was a problem to be solved in Republican communities, and that is hard to change. 89 Kevin Bean argues that extended social and family networks, combined with local traditions and loyalties, have provided a base for electoral and other passive support for particular dissident groups. 90 Even if it is a small share of the population that holds on to deep-rooted practices of going to the local armed group rather than the police-for instance, to report an incident of antisocial behavior-this process of citizen socialization fuels armed groups' effort to exercise social control. Indeed, the fact that many community-based restorative justice organizations-which emerged after the Troubles to tackle "punishment attacks" and internal community disputesare led by former paramilitaries speaks to the enduring legacy of wartime social control. 91 In their capacity as former paramilitaries, they have a certain standing in the community and are, thus, in a position to mediate between still-active armed groups and potential "offenders." 92 To shed more systematic light on the perceptions of people living in areas where paramilitary groups are present, in this section, we rely on a 2016 nationally representative survey (N ¼ 811), 93 combining it with the geolocated data on paramilitary-style attacks. Due to the local nature of the social control paramilitary groups exert, the number of respondents who live in areas affected by paramilitary-style attacks is relatively low, but the nationally representative sample allows for analyses of perceptions that reflect those of the population as a whole. 94 To probe respondents' views of informal and formal authorities, we draw on a hypothetical scenario aimed at capturing which authorities people could turn to when faced with a law-and-order problem that formally is the domain of the state: antisocial behavior. As shown above, during the conflict, in certain neighborhoods it became common to report antisocial behavior to informal rather than formal authorities. Therefore, if the legacy of informal institutions from the conflict is maintained by a bottom-up socialization mechanism, we expect people in areas presently controlled by paramilitary groups to find informal authorities effective and be skeptical of formal authorities.
The respondents were presented with the following scenario: "A man lives in a neighborhood where there is a severe problem of anti-social behavior, such as vandalism and car thefts. What would he do to try to solve the problem?" The survey enumerators then asked the respondents to assess how useful it would be to turn to a particular authority for help, on a scale of (1) "this would make no difference"; (2) "this might help a little"; and (3) "this would help a lot." The question does not force the respondents to pick one authority over another but allows them to rank different authorities' effectiveness. Here, we examine how useful people would find it to turn to the police, our indicator for a formal authority, and whether they would find it useful to "contact a member of the community who has influence," our indicator for informal authorities. For the latter, if prompted, the survey enumerators were instructed to say this was an informal person of influence. 95 This is not a direct measure of respondents' perceptions of paramilitary groups but, rather, a measure of their views regarding informal authorities more generally. Overall, as shown in Tables 3 and 4, most respondents (81 percent) say it would help (either "a little" or "a lot") to contact the police, whereas 76 percent said it would help (either "a little" or "a lot") to contact "a member of the community who has influence." To isolate the association between paramilitary-style attacks and a respondent's rating of formal and informal authorities, we conduct multivariate regression analyses. The key independent variables are the number of Loyalist and Republican paramilitary-style attacks a decade on from the Good Friday Agreement in the SOA where each survey respondent lives. 94 Whereas 68.6 percent of respondents (556) live in areas that have not experienced any paramilitary-style attacks, 18.7 percent (152) live in areas that have experienced between one and two attacks, and 12.7 percent (103) in areas that have experienced three or more. 95 The answer options also included "contact a community leader, like social worker, youth group leader, or similar," which is not perfectly correlated with "a member of the community who has influence" (0.50), suggesting we are capturing something different.
The survey was conducted in 2016; hence we measure the independent variable, paramilitary-style attacks, from 2008 to 2016. There were 733 attacks over this period, with Republican groups responsible for 47 percent and Loyalist groups for 53 percent. Over 30 percent of survey respondents (N ¼ 255) live in an area that experienced at least one attack in this period. Although we cannot establish causality, a high rating of informal authorities in these areas would indicate that paramilitary groups do not operate based on coercion alone-and the possibility that these practices are also associated with a bottom-up socialization mechanism of relying on informal authorities. The analyses include models for Catholic respondents and Protestant respondents separately to examine the social control exercised by armed groups within their respective communities. We run three logistic regression models: (1) whether respondents rate the police as effective ("help a little" or "help a lot" as opposed to "make no difference"); (2) similarly, whether respondents rate informal authorities as effective; and (3) whether respondents rate the effectiveness of informal authorities higher than the effectiveness of the police. In Models 1 and 2, we create dichotomous variables, as we are interested in whether respondents have faith in the effectiveness of a certain authority. The third model is based on a binary variable constructed from the two survey questions presented in Tables 3 and 4. It is coded as 0 if a respondent rates the effectiveness of the police as equal to or higher than her or his rating of whether it would be useful to contact a member of the community who has influence. 96 Approximately 23 percent rate informal authorities as more effective than the police.  Although respondents were asked about the effectiveness of formal and informal authorities independently, they were first asked to rate the police. As such, they may consciously or subconsciously rate the effectiveness of the informal authority in light of their previous response regarding the police.
We control for several alternative explanations for people's rating of informal or formal authorities. We include age to capture whether respondents were socialized during the Troubles and, thus, are more likely to be used to turning to informal authorities. We include an indicator for whether respondents experienced violence at the hands of the state during the Troubles, expecting those who did to be more skeptical of formal authorities. 97 We control for gender, expecting women to be more wary of informal authorities. 98 We also control for trust in people in the neighborhood-those who do not trust their neighbors may prefer to go to the police, and we control for general levels of trust, as skepticism of either authority may be driving general (dis)trustfulness. 99 We control for whether respondents feel discriminated against on the basis of their community or religion, expecting those who do to be more skeptical of the state. We also include SOA-level controls. The first is a control for whether respondents live in a community stronghold, assessed as an SOA where more than 90 percent of the Christian respondents are either Catholic or Protestant. Overall, 31 percent of respondents live in community strongholds. We also control for levels of deprivation, measured in 2015-16 as the proportion of population living in households whose equivalized income is below 60 percent of the national median. On average, respondents live in SOAs where the level of income deprivation per this measure is 13 percent. 100 Finally, we control for whether respondents live in urban SOAs, 61 percent of whom do.
The statistical findings support the qualitative research, while revealing differences across the two communities. 101 The models in Table 5 show that there is a negative and significant association between Catholic respondents' rating of police effectiveness in SOAs affected by Republican attacks. Given the long-held concern about being seen as a "tout" among Catholics in Northern Ireland, particularly in areas controlled by paramilitary groups, this is as expected. However, as the results for Catholic respondents in Tables 6 and 7 show this skepticism of the effectiveness formal authorities is not matched by a positive view of informal authorities. Although Catholics respondents' rating of the effectiveness of informal authorities is positive in areas of paramilitary control, it does not attain traditional levels of significance. We find the opposite for Protestant respondents. Although there is a positive and significant association between Loyalist paramilitary-style attacks in an area and Protestant respondents' rating of the effectiveness of informal authorities, their rating of the police is negative but not significant. Predicted probabilities of the models (Figure 3) show that the effect of Protestant respondents' rating of informal authorities as effective (would "help a lot" or "help a little") when faced with an antisocial behavior scenario is consistently strong and increases as paramilitary groups' control of an area increases, though only marginally so. The models in Table 7 reinforce the findings in Tables 5 and 6, demonstrating that Protestant respondents in areas controlled by Loyalist paramilitary groups are more likely than those living in areas without such social control to rate informal authorities as more effective than the police. This is a nonnegligible effect according to the predicted probabilities in Figure 3.
There are a few reasons why Protestant respondents in areas of Loyalist paramilitary-style attacks may be more likely than their Catholic    counterparts to consider informal authorities as effective-and more effective than the police. For one, working-class Protestants feel that they have not received their share of the "peace dividend" and have been left to fend for themselves, without official representatives-neither the state nor politicians-representing their interests. 102 Two, unlike the main Loyalist paramilitary groups from the Troubles, the main Republican paramilitary groups have publicly rejected the use of violence, including "punishment attacks." Indeed, in the Catholic community, these attacks are attributed to Republican dissident groups, which may shape people's perceptions of informal authorities more generally. Finally, Catholic respondents may be more affected by social stigma when answering these questions. According to Michael McKeown's data, sixty-six Catholic civilians and paramilitaries were killed for informing during the Troubles compared to twenty-nine from the Protestant community. 103 As it was predominantly Catholic areas that were targeted by British counterinsurgency strategies, it is also these areas that have inherited a strong stigma against informing or being a "tout." In the Protestant community, the British Army-initially at leastpermitted informal "justice" activities. 104 As such, Catholic respondents may be more wary than Protestants to report that they would find informal authorities effective in handling a law-and-order problem. All in all, these models show there is an association between paramilitary groups' social control in the postwar period and people's perceptions of formal and informal authorities. The findings suggest that the paramilitary groups' informal "justice" systems more than twenty years on from the Good Friday Agreement are not only the result of these groups' top-down incentives. They are also associated with skepticism of the effectiveness of formal authorities (among Catholic respondents) and a positive view of the effectiveness of informal authorities (among Protestant respondents). Our findings speak directly to those of the Independent Reporting Commission (IRC) in 2018, which concluded that there was fear and anger about the continuing coercive control exercised by paramilitaries in some communities, but also that the paramilitaries were "regarded by some in the community as protecting their areas." 105 Though our statistical analyses cannot causally establish that there is a socialization mechanism at work, and the differences across the Catholic and Protestant communities merit further investigation, this study is a first cut at systematically assessing the legacy of paramilitary groups' social control a decade or more after the conflict ended, encouraging further research in Northern Ireland and beyond.

Legacies of Informal Justice beyond Northern Ireland
We know armed conflicts leave legacies, though most work on postconflict societies focuses on the legacies of armed actors only as armed actors. Yet, as a growing body of work on wartime institutions and rebel governance highlights, armed actors do not just fight. They also govern, providing people with public goods that are central to their strategies for survival, particularly security. We focus on the continued efforts of armed groups in Northern Ireland to impose social control two decades after the signing of the Good Friday Agreement. Indeed, both Loyalist and Republican paramilitary groups-or individuals acting in their name-continue to orchestrate so-called punishment attacks as a way to control "their" communities. Instead of going away, armed groups operate by drawing on deprivation and the legacy of wartime institutions from the Troubles. Our study suggests that the continued existence of these informal "justice" systems in the postwar period may not be driven exclusively by paramilitary groups' topdown attempts to impose social control; this persistence is also associated with skepticism of the effectiveness of formal authorities and a more favorable view of informal authorities within certain communities.
Our findings contribute to a growing body of work that emphasizes armed actors' socialization during conflict, which may hinder their reintegration into "normal" life after conflict ends. Similarly, there is a renewed focus on how communities become used to certain types of governance and "justice" by nonstate actors during conflict. This study suggests that these dynamics are mutually reinforcing. Certain communities continue to look to informal authorities to take on some governance functions, and armed actors see their position as justified.
To what extent can findings from Northern Ireland resonate elsewhere? Most work on informal and wartime institutions focus on settings where the state is not sufficiently strong to exercise control over (parts of) the territory that it claims. However, paramilitary-style attacks in Northern Ireland occur in a strong state where the police undertake concerted efforts to prevent paramilitary groups from employing violence. 106 This is why we were initially puzzled by the case of Northern Ireland. These attacks also happen in a democratic state where the armed conflict resulted in a comprehensive peace agreement, and in which the peace process has enjoyed high support, both financially and in terms of public opinion. 107 As such, Northern Ireland is potentially a least likely case for observing armed groups' continued social control in the postwar period. Alternatively, do these conditions make Northern Ireland a most likely case? Indeed, the legacies of informal "justice" in Northern Ireland suggest three possible scope conditions that can guide future work as the research community builds up comparative research on the legacy of informal wartime institutions: state strength, conflict termination, and regime type. First, what is particularly puzzling about the persistence of armed groups' informal "justice" systems in Northern Ireland is that these efforts at controlling communities are happening in a strong state. The state's ability to eliminate armed actors or challengers alone cannot explain the persistence of social control in Northern Ireland. Nor can it explain the variation in where paramilitary-style attacks take place. The police were a controversial and central actor during the Troubles, unable to police so-called no-go areas. But significant steps have been taken to reform the police to be representative of both Catholic and Protestant communities. Despite such efforts, the wartime orders created during the thirty years of the Troubles are hard to dislodge. The Northern Ireland case underscores that although state strength and legitimacy often go hand in hand, this is not always the case. 108 Our study speaks to the puzzling results of research that have examined international efforts to strengthen the formal institutions of the state to promote the rule of law in weak states. Robert A. Blair, Sabrina M. Karim, and Benjamin S. Morse emphasize that such programs require local community engagement that goes beyond bolstering the efficacy of the state to provide security or fight crime. 109 Our findings echo this implication. State strength may not be sufficient to explain why informal wartime institutions endure.
Second, to what extent is the manner in which an armed conflict ends consequential? Whereas our intuition is that informal institutions may be less likely to endure if there is a comprehensive peace agreement that tackles DDR, SSR, and political reforms-as these steps would weaken armed groups and boost the legitimacy of the state as the upholder of law and order-one could argue that a peace agreement that grants amnesties to both sides, as in Northern Ireland, ensures that paramilitary groups remain on the scene. Our study draws on Bateson's findings from Guatemala, 110 where the armed conflict also ended in a peace agreement. Early findings following the 2016 peace agreement in Colombia indicate that paramilitary and rebel groups retain control in certain areas, and that they are reluctant to relinquish lucrative control of the drug trade and illegal mineral extraction. 111 Future research should explore whether the mechanisms identified in our study also exist in conflicts where rebels were defeated militarily, such as the Liberation Tigers of Tamil Eelam in Sri Lanka.
Third, the persistence of paramilitary groups' social control in Northern Ireland also draws attention to the role of regime type. Whereas our intuition is that armed groups' social control would be less likely to endure in a postwar state that is democratic than one that is authoritarian-as democracies may have more legitimacy-the other side of the coin is that as a democratic state, Northern Ireland and the United Kingdom tolerate dissenting voices in ways authoritarian states may not. Thus, a democracy may be the ideal environment for such wartime institutions to persist long after the armed conflict ends. The extent to which these dynamics will persist in more repressive states that employ harsher measures to eliminate them is uncertain. However, we might still expect informal "justice" systems to persist through the same mechanisms identified in our work. Indeed, although the United Kingdom is a democracy, these practices emerged and remained in place during a conflict in which the state actively sought to eliminate the groups and reduce their public and local support through a counterinsurgency campaign. Future work should explore the extent to which these institutions are sticky in more authoritarian and repressive postconflict environments.
The legacies of wartime institutions have important implications for policymakers in the United Kingdom and other postconflict states. Northern Ireland is an important case in its own right, with both observers and paramilitaries suggesting that the Brexit process has jeopardized peace. 112 Social control by paramilitary groups combined with the increasingly uncertain future posed by Brexit vindicate political commentators who emphasize the fragility of peace in Northern Ireland-as became evident with the riots in March-April 2021. 113 There are actors in Northern Ireland who would like a return to the violence of the Troubles, 114 and they have some support at the fringes of both communities that have been socialized over decades of conflict. The Good Friday Agreement has been hailed as a success, with other peace agreements modelled on it. 115 However, our findings point to the importance of postwar reform efforts recognizing that both decommissioning and SSR require consideration of how wartime actors do (or do not) govern, an aspect of postconflict society with which Northern Ireland still struggles.
The research in this study points to the challenges facing SSR if reform efforts do not recognize the legacies of both wartime violence and wartime institutions. Both armed actors and civilians can perpetuate these legacies, even in a strong state. Postwar state-building does not necessarily happen in a governance vacuum, and wartime institutions may persist long after a conflict ends, undermining efforts to consolidate political order. In the context of Northern Ireland, our findings support the IRC, which recommends a "Twin Track" approach that combines "policing and justice responses alongside measures to tackle the deep, systemic, socio-economic issues in the communities most affected by paramilitarism." 116 Indeed, our findings emphasize the "socio" side of persistent paramilitary control. They speak to social control by both progovernment and antigovernment armed actors in postwar cities in Syria and Iraq, for example, as well as in Libya, where armed actors exert substantial political power at a local level, including basic public goods provision. 117 Across several contemporary conflicts armed groups opposed to and aligned with the state have developed wartime institutions, which will have implications for the effort of any postwar authority to govern after a war ends. numerous colleagues, including project partners Karin Dyrstad and Helga Malmin Binningsbø; our University College London colleagues in the Conflict & Change group; seminar presentations at King's College, the London School of Economics, University of Oslo, and University of Reading; and conference presentations at the annual meetings of the Conflict Research Society, the European Political Science Association, and the International Studies Association. In particular, the manuscript benefitted from comments by Zacharia Mampilly at the American Political Science Association's annual meeting in 2020. We are also grateful for the constructive feedback from the reviewers and editors at Security Studies.  Data availability statement The data and materials that support the findings of this study are available in the Security Studies Dataverse at https://doi.org/10.7910/DVN/XXQX2G.