Marginalization and Riots: A Rationalistic Explanation of Urban Unrest

Urban riots are typically carried out by individuals who live in residential areas that are relatively marginalized socially, economically and politically. Previous research has discussed several aspects of deprivation that may help explain this relationship. Contributing further to this research, we aim to explain why marginalization produces riots by developing a rationalistic specification of social mechanisms. The utility of our model is demonstrated by a case study of the 2013 Stockholm riots. The model consists of (a) general local incentives that appeal to individual motives, but only lead to participation in riots when (b) the delicate local equilibrium is destabilized by an event that (c) makes riots appear justified, riskfree and thrilling. The advantage of this rationalistic model is that it shows why other people, in other places, would have reason to act in much the same way under similar circumstances. ARTICLE HISTORY Received 16 July 2018 Accepted 31 January 2019

advantages. It explains why marginalization produces riots sometimes, but not always, 1 and it shows why other people, in other places, would have reason to act in much the same way under similar circumstances. Hence, it helps us focus on the more specific circumstances linked with marginalization that make riots more likely in some contexts, sometimes.
Our model consists, first, of general local incentives that appeal to individual motives, but only lead to participation in riots when, second, an event destabilizes the local equilibrium and thus, third, makes riots appear justified, risk-free and thrilling. The first part, the local incentive structure, lowers the threshold for accepting and participating in riots. The incentive structure consists of five elements: (1) low institutional legitimacy; (2) resentment towards the police; (3) lacking opportunities for education and work; (4) opportunities to gain status through illegal activities; and (5) limited social control. 2 The second part, the destabilizing event, upsets an existing but delicate local equilibriums. Such events may be, for example, acts of police brutality towards a member of a marginalized and discriminated group. The third part is the impact this event has for individuals who, affected by the local incentive structure, are "potential participants" in a riot. Three ideal-type social mechanisms explain why some individuals participate in a riot once the equilibrium has been upset: (1) the event provides a justification for rioting, so that rules and norms of orderly conduct are overruled; (2) the collective aspect of riots make it less risky to engage in actions that lead to confrontations with the police, or in direct attacks on the police; and (3) several aspects, including collective action, norm-breaking and media attention, make riots thrilling to participate in.
To demonstrate the utility of this model, we use case study research on the Stockholm riots of 2013. Based on interviews with local civil society actors, social workers and police officers, as well as news reports and documents from governmental agencies, we aim to understand the local incentive structure and mechanisms that provided reasons to participate in the riot. Developing our general model, we reconstruct the local narratives in "thinly rationalized" form (i.e. starting from the basic assumption that rioters, like other actors, do things for a reason). If we can understand why some people in Stockholm found it rational to participate in the riots, we might recognize why other people, in other places, might do the same thing under similar circumstances (Bengtsson and Hertting 2014;cf. Elster 1983). Moreover, by focusing on local incentive structures and relevant social mechanisms we want to understand the reasons why people might participate in riots, rather than making any moralistic assumptions about their actions.
The paper begins with a discussion of previous research and theories of riots. In particular, we discuss two influential explanations for the relationship between marginalization and riots: the so-called "flashpoints model" and what we call the "protest model." Rather than rejecting these explanations, we build on them, aiming to further improve our collective understanding of the underlying social and political problems. In the second section, we describe the methodology of our case study of the Stockholm riots of 2013 and the logic of generalization to other cases. In the third section, we present the results of our empirical research. We structure our findings in a manner that demonstrates the logic of each step of our theoretical model. We conclude the paper with a discussion about the portability of our findings to other contexts.

Previous Research
Research on riots indicates several factors that help understand why rioting is more likely to attract persons living in poorer and marginalized residential areas (Lightowlers 2015;Kawalerowicz and Biggs 2015). Case studies stress contextual factors, such as intergroup relations and frequency of inter-group violence (Sen 2008, 10), competition over scarce resources, such as housing and employment (Olzak, Shanahan, and McEneaney 1996), and the consequences of "resignation" (Abbink, Masclet, and Mirza 2011, 19). The aim of this paper is to contribute to this research by explaining why marginalization causes riots, rather than investigating again whether it does. Our contribution is, more precisely, to construct a general actor-focused model to explain why marginalization produces riots sometimes, but not always.
In the literature, we find two dominant explanations of urban riots, both of which inform our analysis. The first is the well-known "flashpoints model," which suggests that factors on different levelsstructural, political, organizational, cultural, contextual, situational and interactionalcause frustration, creating explosive situations in which one provocative event or act could cause riots (Waddington, Jones, and Critcher 1989;Klein and Maxson 2006;Moran and Waddington 2016). By seeing such events as "flashpoints," the theory acknowledges that the macro-level "push factors," on their own, do not explain riots; those factors are constant, yet riots are quite rare (see also Haddock and Polsby 1994). Thus, the addition of provocative "flashpoints" (e.g. instances of police brutality) conditions the impact of structural factors.
However, flashpoints are indeterminate, which complicates the use of this theory alone to explain riots. As Owens puts it, "the effects of precipitating events on ensuing action do not necessarily inhere from any objective content of the events themselves" (Owens 2013, p. 2). Police brutality, for example, would likely lead to strong reactions anywhere, but, at least in Europe, only in some contexts have such acts been interpreted as part of a pattern of oppression and racism, which would make violent confrontations more likely. Hence, while there are obviously many causes of frustration in socioeconomically and politically marginalized areas, and while brutal acts and tragic incidents, such as the killing or mistreatment of local residents, can have significant emotional impact, such conditions and provocations do not necessarily push people into participation in riots. Moreover, the background factors of the flashpoints model are too general and too varied to explain variation in the reactions to a flashpoint within the same context: only some of the factors of deprivation potentially affect individuals directly as incentives to actually participate in riots (see also Newburn 2016a). In contrast, adding on to the flashpoint model's contributions, the actor-related model that we propose makes possible a more fine-grained analysis of rioters' incentives and the social mechanisms of riots.
Another explanation suggested in the literature is that riots are a form of political protest: protests that may take a violent form when carried out by persons and groups that have little chance to influence political agendas and decisions through more conventional kinds of political action (e.g. Wacquant 2008;Akram 2014). In the case of the Stockholm riots of 2013our focus in this articleresearchers have called the event a "rebellion," a "moment for social justice" (Schierup, Ålund, and Kings 2014) and an "uprising" (Sernhede, Thörn, and Thörn 2016). This explanation connects real, legitimate causes of frustration (segregation, socioeconomic inequality, discrimination etc.) to the inability to influence political priorities by claiming that these conditions motivated mobilization to protest violently in response to the lack of opportunities for meaningful, non-violent political action.
We agree with advocates of this explanation that socioeconomic inequalities are to blame for many problems, thus providing incentives for rioting. However, we are not convinced that always seeing riots as protests is the most fruitful way to understand and explain riots. First, analytically, scholars need to separate one phenomenon from another to make it possible to ask questions about the relationship between them. Following such conventions, we think it is important to reserve the term "protest" for events that can be described as "joint (i.e. collective) action of individuals aimed at achieving their goal or goals by influencing decisions of a target" (Opp 2009, p. 39). "Riots," by contrast, should then be used for events "in which a large number of people deliberately damage property and attack others" (Kawalerowicz and Biggs 2015, 675;cf. Bohstedt 1983, 5;Wilkinson 2009, 330). Making a distinction between protests and riots has analytical advantages, since it makes it possible to ask questions such as: "When do people riot rather than participate in more . . . orderly forms of collective contention?" (Kawalerowicz and Biggs 2015, 693). And: "When are riots initiated with the aim of influencing decisions of a target?" Second, empirically protests usually require resources, capacity for mobilization and the expectation that decision makers may respond in a certain way. By contrast, riots often occur where resources are scarce, the level of organization is low and individuals have reason not to think highly of the system's responsiveness (Kawalerowicz and Biggs 2015). That said, riots and protests may be empirically connected; for example, riots sometimes start after peaceful demonstrations are found to be unproductive, as in London 2011 and Stockholm 2013. A violent response or an absence of response to protests may indeed help explain a particular riot. But clearly, it is only if we treat "protest" and "riot" as separate concepts that we can formulate, and test the validity of, such explanations. It is thus important to distinguish between protests and riots, while at the same time acknowledging that empirically they may often coincide, partly, as we argue, because some of the conditions that contribute to making riots rational, also give reasons for people to protest politically.
The empirical focus of this paper is on recent European riots. We consider the connection between protesting and riots to be different there than in the United States, which has been the focus of most research on riots (see Kawalerowicz and Biggs 2015). While riots on both continents are typically related to urban marginalization, structural racism, discrimination and police violence, there are also significant differences in the history and consequences of urban marginalization in the two cases (Wacquant 2008). Some historically crucial uprisings against injusticessuch as the Civil Rights Movement in the United Statesinvolved violent confrontation following violent suppression of protests by police and other state agencies (e.g. McAdam 2010). 3 Riots in Europe do not, in general, follow the logic of these extraordinary episodes in American history. In our general model for explaining riots, we emphasize that structural discrimination, as well as experiences of humiliation, must be included as important background factors. Moreover, we consider police violence to be the typical trigger for riots. But it is important to acknowledge that in many European cases, including our case study, riots were neither a last resort of protest action, nor part of a major social and political mobilization. We suggest instead that depending on people's positions within the local incentive structure that we describe, they may respond either to incentives to protest or incentives to participate in a riot, or neither or possibly both. The important point is that the incentive structure will be partly different for different people. For example, while there is generally a stronger resentment towards the police in marginalized residential areas, not all people feel equally strong resentment. And not all people will be attracted to join groups that have a conflict with the police, even if this opportunity may exist locally. Similarly, if we sought to explain protests rather than riots, we might find that organizational capacity is a key factor explaining successful mobilization. And for those who have such organizational capacity, riots will seem less justified or less meaningful. In other words, these are distinct phenomena, with partly different and partly overlapping incentives. To explain one particular kind of action, rather than action as such, it is most important to focus on the incentives that are specific to that action.
In this paper, we investigate the relationship between problems of marginalization and riots to identify local incentive structures as well as the driving mechanisms on the actor level. In relation to previous research, the actor-focused theory we propose sees potential rioters as being thinly rational and acting within a local incentive structure. Sociologists have generally noted that such incentives are understudied (Tilly 2003, 40;Levine and Rosich 1996, 70;cf., however Collins 2009). These scholars' arguments suggest that strategic aspects can help explain participation in riots. Our analysis in this paper will follow that lead. We agree that relatively deprived and marginalized areas are hotbeds for potential riots, but to make the analysis portable to other similar contexts we focus on "potential rioters" and translate the general social and economic conditions into a specified incentive structure that may make some people ("potential rioters" actually take part in riots in certain circumstances. 4 Depending on position in local structures, some individuals will be more likely to participate than others. We thus aim to specify why marginalization leads to riots in some cases. By reconstructing the incentive structure that faces potential rioters, we can tie these conditions to three specific social mechanisms that are activated at moments of destabilization: justification, decrease of risk and thrill. Our model shares important aspects with previous explanations of riots, which identify several of the factors explaining social unrest (e.g. Kawalerowicz and Biggs 2015;Moran and Waddington 2016;Newburn 2016b;Hörnqvist 2016). Our contribution is constructing one coherent actor-related explanation that deepens our understanding of the interaction between different contextual factors and the mechanisms they set in motion in a situation of disrupted equilibrium. Seeing potential rioters as thinly rational actors has the important analytical advantage that if the perspective proves fruitful for understanding one case of rioting, we will have "reasonable expectations of finding a similar mechanism at work in other similar settings" (Bengtsson and Hertting 2014, 726;cf. Hedström and Swedberg 1998). Translating general physical and social characteristics to elements of an incentive structure makes it possible to distinguish between five structural elements that feed directly into potential rioters' incentives and other elements that do not. Some aspects often related to deprived urban estates, like meagre physical environment and lack of local associational life, do not seem to directly incentivize to rioting the same way that the five aspects included in our model do, at least not judging from the Stockholm case.

Methodology
With the aim to understand the background, outbreak and process of the riots in Stockholm 2013, we conducted in-depth interviews with 18 strategically selected individuals with firsthand knowledge of the conditions and issues related to the riots. The interviewees included civil society activists, social workers, youth centre employees and local police officers. 5 These locals are key informants with important knowledge of their communities. The interviewees live, work or both in the suburb of Husby, where the riots originated, and neighbouring Rinkeby, where the riots spread on the second day. Several of the interviewees were in the streets during the riots (trying, as concerned citizens, community leaders and police officers, to stop them). Due to their roles in the affected residential areas, they also have large social networks and understanding of social problems in the area that help understand the mechanisms of riots. In addition to interviews, we used archival material and secondary material. 6 We see interviewing people who are part of the local civil society or otherwise involved with youths as the best option to gain insights into the local conditions that help us understand the general logic of riots. Why, one may ask, did we not interview the rioters themselves? In part, our decision reflected the practical difficulties of interviewing rioters. For example, compared to the English riots of 2011, which generated a large material based on interviews with arrested participants (Clarke 2012), the Husby riots resulted in few arrests, and those who were arrested would likely have various reasons to either diminish or exaggerate their roles, as well as giving unreliable accounts of their reasons. Moreover, our purpose here is not to identify the driving forces of each individual rioter, but rather to understand, first, the more general incentive structure that may at some points make it seem "rational" for some to take part, depending on their positions in the local structures. This approach is different from process-tracing approaches that require knowledge of intentions, or minute detail of the process that led up to a decisions and action. We did not select persons to interview based on their proximity to the events that we explain or their special knowledge of the specific people that participated in the riots. Instead we aimed to interview people with insights about the incentives people may have for rioting, based on their experience of working with youths in the area or of seeking to prevent social unrest and related problems. With this purpose, we need to understand local circumstances more broadly, and our interviewees are particularly well-positioned to provide us with important insights in this regard. 7 We conducted the interviews between 2 April 2015 and 5 October 2017, between two and four years after the riots took place. For practical purposes, it was not possible to begin data collection earlier, but this seems to us short enough for such dramatic events to be vivid in these individuals' memories and perhaps long enough for strong emotions to fade and reflection to take place.
Our explanation of the Stockholm riots is based on a rationalistic form of process tracing (Bengtsson and Ruonavaara 2017) and is presented as an "analytic narrative" (Bates et al. 1998). This means that we work "abductively," going beyond deduction and induction by moving between theory and empirical findings to develop in parallel both the theory and the understanding of the empirical case (Alvesson and Sköldberg [2000] 2018, 4-8; cf. Timmermans and Tavory 2012). We construct a cohesive story of the Stockholm riots based on our data with the goal of articulating an empirically grounded, middle-range theory on riots. We focus on incentive structures, events and social mechanisms that help make sense of individuals' choices to participate. It is based on the "thinly rationalistic" assumption that "people do things for a reason" (Bengtsson and Hertting 2014)including rioters and potential rioters. Our aim is to construct a model that helps us understand how a specific incentive structure can make it reasonable, under certain circumstances, to take part in riots. By specifying those conditions, circumstances and rationales, we aim to show why people in other similar contexts may act in similar ways. This logic thus allows for generalization based on one single case in terms of an expectation that similar mechanisms and patterns are relevant to understanding riots also in other similar contexts. Such expectations are based not on statistical generalization, but on the analytical idea that what is seen as (thinly) rational by one actor would also be rational to other similarly situated actors in a given type of context (Bengtsson and Hertting 2014;cf. Johnson 1991).

The Stockholm Riots of 2013
On 13 May 2013, 69-year-old Lenine Relvas-Martins was shot by police in his apartment in the Stockholm suburb of Husby and subsequently died from the injuries. According to the police's initial statement, Relvas-Martins had been taken to the hospital where he had died from the shots fired by the police. But this statement was soon proven incorrect. Local residents provided photo evidence that the man was still in his home several hours after the police reported taking him to the hospital and, instead, a body bag was carried out from the building (Radio Sweden 2013). Rumours started circulating about the circumstances of the man's death.
The day after the shooting, a demonstration took place initiated by the local youth organization Megaphone (Megafonen) that called the incident a murder and claimed racist motives. 8 In a written statement, one of the members questioned the increased presence of police in the area and asked: "Since when is it alright for the police to enter our suburbs and kill our citizens?" He explained: "The police are constantly becoming more aggressive and prone to violence in their work. This violence is expressed against those at the bottom of society, not the least suburban youths that are constantly harassed and subject to the police's excessive violence" (Gerecci 2013). The protest called for an independent investigation of the man's death. The police neither met with the protestors nor responded to their demands.
On Sunday night, May 19, at 10:00 p.m., the police were called out to Husby after reports of several vehicles having been set on fire. That night more than 100 cars were destroyed, a shopping centre was vandalized and a garage was set on fire, forcing the evacuation of a whole apartment block. Three police officers were injured by stones thrown by participants in the riots, which continued until 5:30 in the morning. Each night until 24 May there were more incidents, more cars set on fire, and new confrontations with the police, as the riots spread to neighbouring suburbs as well as to other cities. According to reports, the confrontations were further fuelled by rumours of racist utterances from the police (de los Reyes and Hörnqvist 2017).

A Local Incentive Structure for Riots
The suburb of Husby, where the riots principally took place, is among Sweden's most marginalized and deprived residential areas. While it is often noted that Husby has a relatively vibrant civil society and strong sense of identity (e.g. Lundström 2017; González 2017), it is also, along with neighbouring Rinkeby and Tensta, often listed among residential areas that have particular problems with poverty, criminality, low school performance and relatively high unemployment. The Swedish police classifies Husby as 1 of 15 "particularly vulnerable" (särskilt utsatta) areas, where people's sense of safety is lower than in other areas and the police's capacity to satisfactorily do their job is more limited. According to the police, these areas are characterized by risk factors related to "the condition of the welfare system, unemployment, ethnic segregation, stigmatization and overcrowded housing" (Swedish Police 2015, 4).
In the Rinkeby-Kista borough, where Husby is located, the percentage of children who live under the poverty line is 37 percent, compared to 12 percent nationally (Save the Children 2014). One-fifth of the youths (aged 16-29) in these areas neither study nor work, which is twice as high as the number for Stockholm as a whole (Nilsson and Svärd 2013, 49). Moreover, a large share of Husby's population, close to 90 percent, has an immigrant background (defined as having immigrated or having at least one parent born outside of Sweden), most often Asian or African, and are therefore more likely than others to experience discrimination (Stockholm Statistics 2018).
Beyond these general conditions of marginalization, we find five elements that together form an incentive structure that feeds directly into individuals' drivers for taking part in rioting.

Low Institutional Legitimacy
By institutional legitimacy we mean the kind of "diffuse support" for the way the society is ordered that makes citizens inclined to accept present institutions (Easton 1965). Legitimacy significantly affects persons' inclination to abide by laws and societal norms. According to social-psychological research, most people in neglected and relatively deprived areas still have this diffuse support for present institutions. However, youths from such areas are more likely to be confronted with discourses in which the legitimacy of the system is questioned or attacked, as well as with discrimination, segregation and economic inequality, which erode legitimacy (Tyler 2004). Low sense of legitimacy is an important element in the incentive structure for riots because it undermines habitual compliance.
Lacking institutional legitimacy as part of the explanation for the Stockholm riots has been accentuated by some local activists, politicians and academics (for a discussion, see Boréus 2017; Hörnqvist 2016). The local youth activist group Megaphone, for example, explained the riots by focusing "on the increased poverty, the destructive environment that many youths grow up in, the increased police presence in the area and the lack of social measures." 9 Reflecting this pattern, a few of our interviewees mentioned that there is a tendency among youths in the area to characterize their relationship to state authorities as similar to that to an occupational power, using terms such as "colonialism" to describe the actions of the state in the area. 10

Resentment towards the Police
Relations between police and community are often an important part of explanations of riots, and policing methods are often seen to cause frustration and anger among residents in marginalized areas (e.g. Moran and Waddington 2016;Sernhede, Thörn, and Thörn 2016). Research on police authority stresses two different factors: the police's capacity to make citizens see that they are treated equally and not discriminated against on the basis of ethnicity, class, gender, abilities or other social distinctions, and the police's capacity to be present and available when citizens require their involvement (see Tyler 2004).
In both regards, the police have particular difficulties in marginalized areas. On the one hand, the resentment stems from discriminatory practices. In a report on "vulnerable areas" (Swedish Police 2015), the police note that the measures taken to decrease crime have been mostly reactive and that criminal networks profit from the absence of deeper and more consistent engagement from the police, which is verified by our interviewees. Adding to the resentment towards the police, the months before the riots began controversies arose about the police's alleged use of ethnic profiling.
On the other hand, the resentment is due to a sense of neglect. According to a survey conducted shortly before the riots, 86 percent of residents in the marginalized residential neighbourhoods of the so-called Million Programme disagreed, or strongly disagreed, with the statement that there is too much police presence, and almost 60 percent felt safer when they saw police or police cars in their neighbourhood (Rojas and Batist 2013). 11 Several residents in such neighbourhoods had experienced difficulties when trying to contact the police, and persons of immigrant background were almost twice as likely as others to give this answer (23 percent compared to 13 percent in the rest of the population).

Lacking Opportunities for Education and Work
Lacking opportunities in the labour and housing markets and in the educational system tend to increase the likelihood of criminal association. Several of the interviewees described a sense of powerlessness that comes from a feeling of systemic injustice, which is confirmed and reinforced by reports of discrimination in the labour and housing markets. They noted that "Stockholm is a divided city," 12 and lament the absence of substantial efforts to integrate youths in the labour market.
Several interviewees also saw a connection between the riots and criminal activities, and part of that connection lay in how criminal activities were associated with a certain jargon about legitimate job opportunities and other legal paths. A discourse of cynicism further strengthens the sense of socioeconomic exclusion, particularly among young men, they reasoned. Some saw this discourse as imposed on young men from somewhat older men that they look up to, that youths in the area hear from their own role models that in eyes of the society they are worthless and will never be successful. In this context, illegal ways to income appear more realistic: "the possibility to get fast cash and be somebody, feel important. That's what gives you status, said a person working for a local youth center." 13 Moreover, when other options seem unrealistic and even naïve to entertain, the potential costs appear smaller: "It's like 'if a prison sentence is the worst that can happen, what do I have now, what do I have to lose? I don't have a place, I can't move, I don't have a job,'" 14

A Social Context Where Illegal Activities Offer Status and Identity
Research on youth affiliation with criminal networks and gangs has demonstrated the high importance of social factors, such as offering social recognition and friendship (Klein and Maxson 2006). Social factors may incentivize confrontations with the police, if they function as opportunities to gain entrance to criminal networks. In Husby, an employee of the city administration who is working on crime prevention said: "The problem is that they recruit these younger talents . . . that look up to the older guys, who have nice cars and gold chains around their necks and so on." 15 Criminal groups can offer a sense of identity and status, and they reinforce this feeling through "a certain jargon, a certain culture," said a youth centre employee who works with youths who have left criminality. 16 Criminal groups are seen to offer a sense of importance. According to some interviewees, leaders of criminal networks take advantage of this view and seek to reinforce it. A local police officer said: "If they can strengthen their power positions and their influence, this may have the effect that they can get more youths to join." 17

Limited Social Control
By limited social control we refer to the insufficiency of social networks that might encourage youths to abide by the norms of a community. Social ties are a common explanation for low crime rates. For example, social pressure may impact on the decision to participate in illegal activity, and social ties may increase activity in public spaces, such as the streets of residential areas, which makes it more difficult to engage in criminal activities unseen (Groff 2014). Moreover, low density of organizations, which is taken to indicate lack of collective efficacy at the community level, is correlated with higher crime rates (e.g. Sampson 2012).
As mentioned Husby is often said to have a vibrant civil society, but obviously this did not serve as a strong enough antidote to rioting. Several voluntary associations, local leaders and youth centres seek to provide a sense of safety and security. For example, "citizen guards" and "youth guards," appointed by the city administration, organize nightly walks to show their presence as well as being "eyes on the street" and work with the city administration to reach youths at risk. Another example is a project run by the youth centre organization Fryshuset, which established a branch in Husby after the riots (called The Calm Street, Lugna Gatan) that engages with young former criminals and youths at risk. Moreover, by organizing activities that focus on demonstrating presence on the street and on the squares in areas where there have been illegal activities, the project is meant to discourage crime. However, our interviewees often stated that these initiatives are insufficient, that youths are still on the streets at night and that parents and other adults are unable to exercise authority over them, even when they have good relationships.
Part of the problem is overcrowded homes, which make it less attractive and comfortable to be indoors. 18 For a young man of 20-25 in this situation, there is not much to do but to walk, said one social activist. In an environment characterized by the presence of criminal networks, low confidence in the police and low legitimacy for societal institutions, overcrowding adds another risk factor for criminal involvement by making it more difficult to exert social pressure and show social presence.

Destabilized Equilibrium
Together these five elements constitute an incentive structure for rioting. However, even in a society characterized by these elements there are other factors that still make riots exceptional. For example, social ties and the likelihood of legal consequences or social sanctions may deter rioting. But under the structural conditions outlined above, this balance is fragile, and a sudden incident can easily tip it towards violence and disorder. In Husby, the killing of Relvas-Martins was such an incident.
We agree with previous research (e.g. Waddington, Jones, and Critcher 1989;Moran and Waddington 2016) that no explanation of riots can leave out the critical moment in which one act of provocation leads to a violent collective response. In previous research, such moments are called "flashpoints." However, the importance of such "flashpoints" becomes clearer, we argue, if the event is carefully situated in a social context and connected to specific individual reasons to participate. In other words, individuals are "pushed" to riot neither by objective conditions nor by triggers, not even by a combination of both. Instead, we argue, a comprehensive explanation must specify the social mechanisms that are triggered by the event. We suggest three distinct social mechanismsjustification for violence, opportunity to attack the police, and thrillthat explain how participation in rioting becomes a reasonable choice under the local incentive structure we have outlined above. These three mechanisms are social, because the individual choice to participate is tied to the way others act or are expected to act.
The pattern of interaction between individual incentives and collective action can be understood in terms of Granovetter's well-known threshold model of collective behaviour, where the costs and benefits of different actions that determine the threshold depend on how many other actors choose a certain alternative (Granovetter 1978). When more individuals participate, the cost goes down (e.g. the risk of being caught and of social sanctions), and the benefit goes up (e.g. the thrill of taking part).

Three Social Mechanisms of Rioting
We discuss three social mechanisms that explain why some people experiencing a local incentive structure for rioting might participate once a triggering event has occurred.

Justification for Violence
Violence seems justified when the injustices appear so grave that the norm of compliance no longer feels obligatory or even appropriate (see also Hörnqvist 2016). In Husby, we see examples of this mechanism, as our interviewees explained how the shooting of Relvas-Martins and the police's false information about the event was, for many, the last straw, a blatant reminder that the lives of people in the suburb do not count as much as the lives of others: "You had this shooting, of a member of their community, where the police had misinformed about what had happened. And where you felt already from the start that, regardless, this will never lead to prosecution. We don't matter as much. We are not of equal worth." 19 Several interviewees saw this general sense of neglect and structural racism as part of the explanation for the riots: "It was the result of enormous discontent and a sense of hopelessness. (. . .) Then comes this shooting [of Relvas-Martins] and nothing happens, it's wiped under the rug. . . and so 'enough!' There was so much behind this, more than the fact that the police shot this guy." 20 Several interviewees explain the riots in similar terms, suggesting that especially at the time before the riots there was much hatred towards the police among some groups of youths in the area. 21 Moreover, the police's response to the riots, which interviewees found unnecessarily brutal, is said to have contributed to the escalation of the situation (cf. Boréus and Flyghed 2017).

Opportunity to Attack the Police
Our second mechanism connects riots directly to criminal networks. The logic is quite simple: criminal networks gain from confrontations with the police and often seek to defend their "territory" through violent attacks (Peterson 2008). Riots may be started to attract police and gain an opportunity to attack them, or individuals involved in criminal networks may take that opportunity once it arises. This mechanism is connected to the incentives for being involved in criminal networks.
Most interviewees agreed that the riots were started by a small group of men involved in criminal networks that wanted a confrontation with the police. 22 A young social activist said the burning of cars and buildings was aimed to attract the police and lure them into residential areas where they could more easily be attacked: "The violence was really mostly targeting the police. Everything else that happened was just to attract the police. 'We are going to hit the police!' That was the focus in the riots." 23 A local police officer said, "It was not the usual youth thing that sometimes happens, when they throw stones and run away. In this case, it was planned, preparations had been made, not only for one night [of riots] but for a longer period, that was clear." People joined the riots from other suburbs and other cities. "You could only back down, because they were extraordinarily aggressive, with Molotov cocktails and stones and so on. They set things on fire and attacked the fire brigade." 24 As our interviewees suggest, the riots were organized by youths who took the opportunity to attack the police. Among 16 rioters that had been arrested up until 24 May 2013 were previously convicted (Aftonbladet 2016). 25 Thrill Previous research notes the importance of strong emotions to explain participation in riots, such as "camaraderie, jubilation, empowerment, pride, and a sense of accomplishment [combined with] anger and fear" (Moran and Waddington 2016, 15; see also Wilkinson 2009). The thrill of being part of a collective act of uproar is one important mechanism that helps us understand the action. In Husby, some interviewees claimed, smaller confrontations with the police can be a way of having fun: "There is nothing happening, nothing, and then suddenly something happens." 26 Similarly, one interviewee noted: "There are those who, when they are bored, might set a car on fire just to attract the police and get some action." 27 Several of our interviewees suggested that the riots created a rare opportunity to be part of a spectacle that got considerable attention. The enormous media attention, both nationally and internationally, added considerably to the drama: "People came from other areas already on the second day, 'Oh shit, there is something going on in Husby, let's go there!' Because it's fun. And . . . kids from here too, who just 'there is a fire in the center, let's go!'" 28 "The media's impact [. . .] contributed to the development. 'If we burn some cars we will be on the news!'" 29

Concluding Discussion
Why does marginalization produce riots? Building on existing research, we have constructed a rationalistic model that can account for the occurrence as well as the nonoccurrence of riots. The model includes a general local incentive structure with five elements. This incentive structure is most often counterbalanced by norms of peaceful conduct. However, this vulnerable local equilibrium is destabilized as an event occurs which makes riots appear justified, risk-free and thrilling. We argue that this model is generalizable across various urban contexts; other people, in other places, would have reason to act in much the same way if circumstances were similar.
We demonstrated the analytical utility of this model through a case study of the Stockholm riots of 2013, but we claim it helps make sense of riots more generally. Both the incentive structure for taking part in riots and the social mechanisms should be portable to other contexts of potential urban rioting, although the different elements may differ in strength between local settings.
Previous research on the English riots of 2011 illustrates this. Qualitative and quantitative data, including interviews with arrested rioters, indicate that participants in the English riots faced an incentive structure similar to the Stockholm case. First, indicating low institutional legitimacy, 80 percent of those interviewed by LSE/the Guardian stated that "government policy" was part of their motivation, and only 51 percent said they agreed with the statement "I feel I am part of British society" (compared to 92 percent nationally; Lewis et al. 2011, 24). Second, resentment towards the police was common among participants. Only 7 percent of rioters believed the police in their area were doing a "good" or "excellent" job, compared to 56 percent nationally (20). Regarding the third element in the incentive structure, lacking opportunities for education and work, data for rioters who were convicted in 2011 suggest that underage rioters had poor school performance. Almost two-thirds were unemployed at the time and three quarters had a history of unstable or no employment. As for opportunities to gain status through illegal activities, while only 13 percent said they were active members of criminal gangs, 68 percent had been previously convicted (Clarke 2012, 286-287). When it comes to limited social control, a quarter of those interviewed had no fixed home and another quarter were at risk of becoming homeless (286-287).
As in the Swedish case, the participants do not normally engage in rioting, because the incentives are not strong enough to outweigh the significant reasons that exist for orderly conduct. This equilibrium was destabilized in a way similar to the Stockholm case with the shooting of a local Tottenham resident outside a police station (Waddington 2012, 7). While the English case has not been analysed in such terms, we also find indications of the relevance of our three social mechanisms. A situation was created in which rioting seemed justified; the collective aspects made it relatively risk-free to engage in confrontations with the police; and the thrill of taking part is clearly indicated by interviews with rioters: "Everyone started joining in, different sides, different parts of town. They got bricks, started smashing windows, smashing shops. It was exciting, because we all had the intention to rob something for free, you know what I mean? . . . The first thing that came to my mind? Let's get wild, let's do it" (rioter interviewed in Lewis et al. 2011, 23).
These aspects of the English case further indicate that our model should be generally relevant to explain why marginalization produces riots sometimes, but not always. Each of the elements of our model has been observed in previous research and some have been used in previous explanatory models. However, we argue that integrating them all within one thinly rationalistic model improves possibilities to generalize findings to other contexts, thus contributing to our collective understanding of the underlying social and political problems. In accordance with our "thinly rationalistic" approach, we would expect the model to provide fruitful understanding also of other recent riots. Following this logic of generalization, we would also expect future urban riots, arising within similar incentive structures, to follow a similar pattern. Notes 1. Cf. Elster's well known formulation about mechanisms: "if p, then sometimes q" (Elster 1989, 10). 2. As will be demonstrated below, all elements of the suggested incentive structure for rioting are linked to housing, including segregation, overcrowded apartments, stigmatization related to suburbs, discrimination that makes it difficult to move, and a sense of territorial exclusion. 3. McAdam (2010, Ch. 8) shows that riots can be part of movement's repertoire of protest actions. 4. It is important to stress that persons who would normally not participate might, in certain situations, do so. This also means that it is not possible to estimate the number of potential rioters in a certain neighbourhood; there is no discernible "population" of potential rioters. Generally, only a fragment of the local population participates when riots erupt. 5. The interviews were conducted between lasted between 1 h and 20 min and 2 h and 35 min. The interviewees were chosen either based on our previous knowledge of civil society in the areas studied or suggestions from other interviewees, as well as our interest in including a variety of perspectives in the study. 6. We relied here on larger media outlets that covered the events, such as Dagens Nyheter, and Radio Sweden and used statistical analyses from Save the Children and The Swedish National Council for Crime Prevention. 7. This study is part of a larger research project, in which we examined the role of local civil society and street-level bureaucrats in working locally to address social problems after the riots 2013. Part of our interviews concerned the experience of the riots and the media's reporting on the events. The interviewees were asked whether they had their own views of what had happened and why. We developed our own thinly rationalistic interpretation based on this data, complemented with media, archival and secondary material, as we found that previous theories of riots did not appear to explain this case convincingly. 8. The police's internal investigation later concluded that the shots had been fired in selfdefense, but the handling of the situation has been starkly criticized, also by individual police officers. 9. Al-Khamisi (2015, 10). 10. Interviews 7, 14. 11. The reputed Swedish Million Programme was implemented in the period 1965-1974 when one million new dwellings were built over 10 years. Although the physical standard of dwellings was high, some of the estates were soon criticized for being barren and monotonous living environments. Today some of the Million Programme projects are among the most marginalized housing areas in Sweden.