jerusalem_smallFrom the Bottom-Up: Policing and Sectarian Conflict in Divided Societies

*2016-2017 UCSD Lijphart/Quon Dissertation Prize – Best Department Dissertation of the Year*

How does the design of the police affect motives for sectarian conflict in divided societies? The rank-and-file bureaucrats and law enforcement officers who make up law enforcement institutions have considerable discretion over the way that government laws and policies affect citizens in real terms (Lipsky 1980). In divided societies, where ascriptive identity like ethnicity or religion significantly motivates political attitudes and behaviors, this discretion assigns great importance to the makeup and structure of these rank-and-file officers. The importance of this discretion is amplified by the police’s capacity to use force against civilians, making them a particularly-important institution in communities where sectarian conflict is likely to occur.

I develop a theory of bottom-up integration in which integrating the police rank-and-file reduces the motives of civilians from marginalized groups to use violence against the government. I define integration as the inclusion of sufficient officers from each relevant group to affect the operations of the institution, and the distribution of those officers so that non-coethnic officers work side-by-side and serve civilians from all groups. Integration balances power and influence within the institution and forces interactions between officers and civilians from different groups. Integration is “bottom-up” when the positions being integrated are at the entry-level, and officers move up through the ranks based on their qualifications. Thus, whereas most existing research on institutions in divided societies focuses on the makeup of the individuals who make laws and policies (Lijphart 1969, 2012; Norris 2008; Horowitz 2004), bottom-up integration refers to the citizen-facing bureaucrats and officers responsible for interpreting and implementing laws.

I develop and test two mechanisms through which police integration affects motives for anti-state violence along group lines. First, bottom-up integration reduces biases in police service provision which form the basis for grievances. Integration reduces officers’ motives for sectarian bias by improving attitudes towards outgroups. Patrol officers serve closely with one another as status equals working towards a common goal which requires cooperation and personal interaction to achieve, conditions which existing research suggests should improve attitudes towards outgroups (Allport 1954). However, even if integration fails to reduce preferences for bias, it restricts officers’ abilities to act on those biases. Integration positions officers from different groups to monitor one another’s behavior and either deter mistreatment of their own coethnics or sanction mistreatment when it does occur. Patrol officers frequently work in pairs or groups, allowing officers to monitor the way that other officers treat their own coethnics. By reducing biases in police treatment of civilians, integration addresses an importance source of grievances which motivate anti-state violence.

Second, integration addresses a commitment problem in which the powerful state is unable to credibly-commit not to take advantage of a weaker group (Fearon 1995, de Figueiredo and Weingast 1997, Lake and Rothchild 1996). Integration raises the costs of future repression by arming the outgroup, providing them with access to state resources and information, and positioning officers from weaker groups to impose costs on the dominant group by withholding services from them. Recognizing that a state intending to repress would be unlikely to first integrate the police, members of the marginalized group who observe integration should be feel more secure, reducing the likelihood that they misinterpret ambiguous state actions as aggressive and decreasing incentives to engage in preemptive violence.

Research Design

I explore bottom-up integration in two very different divided societies, Iraq and Israel. In Iraq, I focus on conflict between Shia and Sunni Arabs; in Israel, I deal with conflict between Jewish and non-Jewish citizens. I leverage sub-national and over-time variation on police integration within each of these two cases to analyze the effects of police integration. At the same time, the vast differences between the two cases on levels of violence, police institutionalization, and the entrenchment of citizens’ political attitudes provide insights into scope conditions and generalizability.

I use a mixed-methods approach to test a number of hypotheses related to bottom-up integration. The mechanisms described above depend on individuals’ reactions to their perceptions of police integration, therefore I conduct original surveys in both Iraq and Israel to measure perceptions about police integration as well as the main outcome of interest, support for anti-government violence. Because the sensitive nature of support for violence makes a direct measure both unsafe and unreliable, I use a list experiment to estimate the proportion of citizens who would consider using violence against the government. To gain traction on the causal relationship between perceived integration and violence, I embed an experiment within the survey which provides a random subset of respondents with information that the national police are integrated. Finally, I ask a series of questions about the mechanisms described above, including perceived bias in police and government service provision and fears of future repression.

In addition to the survey, I conduct interviews with police officers, trainers, and officials to uncover the processes through which officers in both countries are recruited, assigned, and distributed. These interviews help paint a picture of the role of sectarian identity in policing in Iraq and Israel. They also help ensure that regional differences in police integration are not themselves driven by the same factors driving the relationship between citizens and the police. Finally, I draw on administrative data on the religious makeup of officers at each police station in Israel over a six year period alongside a multi-year crime victimization survey to test the effects of local integration on police effectiveness.


 The main finding is that Sunni Iraqis who received the randomized prime that the Iraqi Police are integrated are significantly less likely to say they would consider using anti-state violence compared to Sunni Iraqis who received a control prime. This finding supports the argument that members of marginalized groups who perceive the police as integrated are less likely to take up arms against the state. I also find a decrease in support for anti-state violence among Arab Israelis who received the prime; however, the difference is much smaller and not statistically significant. I attribute this difference between the two cases to Iraq’s much younger law enforcement institutions which have undergone significant changes over the last two decades, leading to very malleable perceptions about those institutions that are easily shaped by new information in the experimental prime. In contrast, the Israel Police are highly institutionalized, and Israelis’ highly entrenched attitudes about the police were likely not affected to the same degree by the prime.

I then test the two mechanisms directly using direct questions about perceived integration. Sunni Iraqis and Arab Israelis who perceive their respective police forces to be more integrated are significantly more likely to say that police and government services are provided fairly on the basis of identity. These trends hold when controlling for individual- and local-level factors. Similarly, both Sunni Iraqis and Arab Israelis who perceive the police as more integrated are less fearful of future repression by either the police or the government. In Iraq, Sunnis who receive the experimental prime about integration are less fearful of future repression, while in Israel, the prime again has no significant effect on attitudes.

I also test the difference between integration and another form of institutional inclusiveness, group-based autonomy. Whereas integration causes officers from different groups to work together and serve civilians of all groups, autonomy over policing means that officers serve only civilians from their own group. Integration should be more effective than autonomy at reducing conflict motives for two reasons. First, it positions non-coethnic officers to monitor one another, allowing them to monitor one another’s behavior allowing officers from all groups to access all information and resources. By separating officers from different groups, autonomy prevents officers from marginalized groups from monitoring the behavior of outgroup officers and makes it easier for the state to withhold resources and information from them. Second, integration empowers officers from the weaker group to impose costs on the dominant group by withholding service provision from them in the event of a conflict, providing further credibility to the state’s claim that it does not intend to repress in the future.

I test the difference between integration and autonomy by asking survey respondents whether the police in their neighborhood are mostly officers from another group, mostly officers from their group (autonomy), or a mix between the two (integration). Respondents who said that officers in their neighborhood are mixed were significantly less fearful of future repression and significantly less likely to perceive service provision as unequal compared to those who said the police were mostly outgroups. However, there was no such effect for those who said officers are mostly people from their own group. In other words, integration, but not autonomy, is associated with a reduction in these two motives for conflict.

Finally, I test the effects of police integration on service provision in the form of crime prevention. Using my contacts in the Israeli police, I collected data on the demographic makeup of police officers at every police station in Israel between 2008 and 2014 which I used to calculate local-level integration. I test integration against crime victimization from a yearly, large-scale victimization survey conducted by Israel’s Central Bureau of Statistics, along with a full set of controls for individual and local characteristics. I find that within a given subdistrict, increases in police integration are associated with decreases in crime victimization. The result holds for both Jewish and non-Jewish citizens, indicating that integration causes a net benefit in the quality of policing rather than a zero-sum redistributive effect. Using data from my original survey, I show that citizens who perceive the police as more integrated are less likely to say that police service provision is biased, implying that integration leads to a more efficient use of police resources. At the same time, citizens who perceive the police as more integrated are more likely to say they would report a crime to the police, providing officers with better quality information which could lead to more effective crime prevention.

Conclusion and Future Research

My dissertation has two main takeaways which I will continue to develop as the project transitions into a book manuscript. First, policy-implementing institutions play a critical role in determining the citizen-state relationship, and the rank-and-file bureaucrats and officers who make up these institutions are responsible for the way that government laws and policies affect citizens in real terms. While this observation is not new, policy-implementing institutions like the police, as well as education, social services, and public works receive little attention from research on institutions in divided societies. Researchers and policymakers seeking to build institutions in the shadow of potential sectarian conflict must account not only for the institutions that make laws but also those that implement and enforce them.

Second, institutional inclusiveness comes in many forms, each of which suggests dramatic differences in the way that inclusion affects the quality and distribution of service provision, the power dynamic within and beyond the institution, and citizens’ responses to the institution and the government it represents. For example, I show that integration but not autonomy is associated with fewer grievances over unequal service provision and reduced fears of future government repression. Similarly, integration differs from proportionality in the way that it conceptualizes inclusion; integration focuses on balancing power and influence while proportional representation conceptualizes inclusion as a normative goal in its own right. Future research will explore the ways in which these different configurations of institutional inclusiveness affect outcomes like police service provision and sectarian violence.