Policing in divided societies: Officer inclusion,citizen cooperation,and crime prevention

Abstract

How does demographic inclusion in domestic security institutions affect security provision in divided societies? Police officers rely on information from citizens to identify problems and allocate resources efficiently. Where conflict along identity lines erodes trust between citizens and the state, the police face difficulty obtaining information, hindering their ability to provide public safety. I argue that inclusiveness in the police rank-and-file addresses this problem by fostering cooperation from previously excluded segments of society. I test this argument in Israel and its conflict between the Jewish majority and non-Jewish minority. First, a survey of 804 Israeli citizens shows that non-Jews who perceive the police as more inclusive are more willing to provide the police with information. I then use original panel data on police officer demographics at every police station in Israel over a six year period to show that increases in police inclusiveness are associated with decreases in crime.

Keywords

Civil conflict Israel Middle East police power sharing sectarian conflict

How can security forces provide effective public safety in divided societies? A rich line of research argues that the demographic makeup of political institutions affects peace, security, and public goods provision in societies with a history of conflict along identity lines (Roeder and Rothchild, 2005). Institutions which systematically exclude certain segments of the population are accused of delivering substandard services and motivating violent conflict (Horowitz, 1985; Weitzer, 1995), while inclusive institutions improve governance by representing the interests of all segments of society in the governing process (Lijphart, 1969, 1984; Norris, 2008; Strøm et al., 2017). Scholars pay special attention to questions of group representation in the domestic security forces and law enforcement, where the rank-and-file’s capacity for the use of force and discretion over enforcement play a critical role in shaping citizen–state relations (Bayley, 2008; Ben-Porat and Yuval, 2012; Enloe, 1980; Hoddie and Hartzell, 2003; Lipsky, 1980; Perito, 2011; Weitzer and Tuch, 2004).

This article argues that, in divided societies, inclusive policing institutions, and specifically those in which officers from all politically relevant identity groups serve in citizen-facing roles, provide more effective public safety by increasing citizens’ willingness to engage and cooperate with law enforcement. Police officers rely on information from citizens to develop procedures and distribute resources efficiently (Skogan, 1986; Weitzer and Tuch, 2006). In societies where conflict along group lines erodes trust between a particular group and the state, civilians who perceive their group to be marginalized are less willing to engage with institutions of state authority, forcing the police to provide public safety without a complete picture of citizens’ needs. Inclusion in the police rank-and-file addresses this problem by signaling to citizens that the police are more likely to treat them well, thereby increasing their willingness to cooperate with officers and provide them with information (Weitzer and Tuch, 2004). Officers then use this information to provide services more efficiently by identifying problem areas with greater accuracy and precision, and devoting appropriate resources to those areas.

Existing research on government service provision in divided societies focuses on the way that inclusion affects officers’ behavior. Officers are said to “actively represent” citizens from their own group by exerting a greater effort to serve them at the expense of members of other groups (Kennedy et al., 2017; Meier, 1975; Meier and Nicholson-Crotty, 2006), implying a distributive effect in which inclusion in the police improves outcomes for the newly included group at the expense of other groups. In contrast, the information-efficiency mechanism in this article implies a net improvement in the quality of security provision. Because public safety is provided non-excludably at the community level, more efficient service provision caused by better information should benefit all members of the community, regardless of group affiliation.

I test this argument in Israel, a compelling case because of its entrenched political conflict between the Jewish majority and non-Jewish minority, groups to which the law nominally affords identical rights and responsibilities but that in practice receive dramatically different levels of government service provision.¹ In the context of the Second Intifada, which claimed several thousand lives between 2000 and 2005, Israel provides a window into the role of institutional inclusion in post-conflict societies. While conventional wisdom suggests that the presence of more non-Jewish police officers would improve service provision specifically for non-Jewish citizens (Abraham Fund Initiatives, 2013; Ben-Porat and Yuval, 2012), I argue that increasing minority presence in the police should benefit all citizens by giving the police more complete information about the challenges facing their jurisdiction.

First, I test the information flows mechanism using a survey of 804 Israeli civilians. I find that those who perceive the police in their community as inclusive are more willing to report crimes to the police. As expected, the effect is strongest for non-Jews. However, there is no evidence that perceived inclusion reduces willingness to report crimes among Jews. Thus, as the police become more inclusive, information available to them increases, and the increases are largest in communities with large non-Jewish populations.

I then use an original dataset of yearly, local-level police officer demographics to test the effects of inclusiveness on public safety provision. I draw on a large-scale victimization survey to measure crime rates independent of reporting rates. Within a given location, increases in police inclusiveness are associated with decreases in the probability of crime. As expected, the effect is concentrated in locations with large non-Jewish populations, the same locations where increases in information flows should be largest. Within those locations, however, the effect does not differ significantly between Jewish and non-Jewish citizens, contradicting the conventional wisdom that officers provide higher-quality services to their coethnics.²

Most existing research on power sharing focuses on policy-making institutions like legislatures and executives (Lijphart, 2012; Norris, 2008; Strøm et al., 2017), and on the military (Enloe, 1980; Hoddie and Hartzell, 2003). Power sharing at the policy-making level is often viewed as zero-sum, creating incentives for groups that benefit from the status quo to resist its implementation and incentivizing ethnic elites to move toward ideological extremes (Roeder and Rothchild, 2005). The apparent net benefit of police inclusiveness on security provision addresses this problem by demonstrating that power sharing need not harm those privileged by the status quo, leading the way for future reforms. In post-conflict societies, police integration, or the process of hiring officers from previously under-represented groups into the police and empowering them to enforce laws throughout all segments of society, may present a first step toward broader institutional inclusion. Finally, I find that religious diversity among police officers leads to better, not worse, service provision, a result that contradicts conventional wisdom that heterogeneity harms public goods provision (Alesina et al., 1999; Habyarimana et al., 2007; Miguel and Gugerty, 2005) but is consistent with a growing body of evidence on police and military integration (Blair et al., 2016; Lasley, 1994). I suggest that the rigid structures of security institutions, including a strict chain of command, uniforms, rituals and traditions, and an “us against them” mentality in the struggle against criminals or other opponents, provide officers with many of the same heuristics as shared ascriptive identity and therefore counteract the negative effects of diversity on cooperation.

Police inclusiveness and security provision in divided societies

In divided societies, where identity significantly motivates political attitudes and behaviors, the demographic makeup of service-providing institutions substantially affects the quality and nature of the services those institutions provide. Courts hand down different rulings depending on the identity of judges (Grossman et al., 2015; Shayo and Zussman, 2011), schools with more minority teachers make minority-friendly policies (Meier and Stewart, 1992), and US police departments that are more racially representative of their jurisdictions receive fewer complaints of excessive force (Kennedy et al., 2017). Yet existing research is somewhat ambiguous about the mechanisms linking institutional inclusion with governance. One line of research suggests that civil servants “actively represent” ingroup coethnics (Andersen, 2017; Kennedy et al., 2017; Meier, 1975, 1993; Meier and Stewart Jr, 1992; Mosher, 1982)³ by exerting greater effort at providing services for members of their own group, potentially at the expense of effort to provide services for outgroups. A related argument suggests that bureaucrats and police officers are better suited to serve coethnic civilians as they are inherently better able to understand and identify with the community (Abraham Fund Initiatives, 2013; Ben-Porat and Yuval, 2012). Given the fixed resources typically available to the institutions in question (and officers’ limited supplies of effort), these arguments imply a zero-sum game in which representation for one group reduces the quality of service provision for other groups. This zero-sum outcome presents a problem for inclusion as a solution to conflict because it gives the group privileged by the status quo reason to oppose integration. Furthermore, a distributive effect on security provision by the police belies the argument that inclusion improves society’s security overall, undermining claims that inclusion is normatively desirable.

A second possible mechanism linking inclusion with service provision prevalent in research on the police and military is that serving in an inclusive institution alters officers’ attitudes about outgroups and the relative salience of identity. Samii (2013) finds that soldiers in Burundi who serve in mixed units exhibit less prejudice against outgroups, and Ostwald (2013) shows that soldiers assigned to live-in units in Singapore hold an increased sense of civic identity. This mechanism suggests that inclusion may smooth inequalities in security provision across groups by reducing police officers’ desires to differentiate service on the basis of recipient identity. However, it makes no clear prediction regarding the quality of security provision. Finally, diversity among police officers might reduce the quality of service provision (Banerjee et al., 2005; Habyarimana et al., 2007; Laitin, 2007). Shared identity helps solve collective action problems by improving community members’ abilities to monitor and sanction non-compliance (Berman, 2000; Besley et al., 1993; Fearon and Laitin, 1996; Miguel and Gugerty, 2005) or by providing a common language to aid communication (Hardin, 1997). Furthermore, if members of different groups have divergent preferences about the distribution of government services (Alesina et al., 1999), integrating the police could create disagreements between officers about how to engage in policing. On the other hand, the rigid institutional structures of policing like a strict chain of command, common uniforms and norms of appearance, and a threat to personal safety may counteract the inefficiencies of heterogeneity by providing a set of heuristics which aid cooperation (Gibson and Hoffman, 2013). Studies of the police in Los Angeles (Lasley, 1994) and Liberia (Blair et al., 2016) find no evidence that heterogeneity among officers impedes service provision. The highly structured nature of security institutions suggests that the inefficiencies of heterogeneity should be less of a concern among police officers than in most other settings.

Inclusion, information, and efficiency

This article shifts the focus from how inclusion affects officers’ behaviors to how it affects citizens’ reactions. Security forces rely on citizen-provided information to combat threats to public safety. Just as counterinsurgent forces use tips to combat insurgent activity (Berman et al., 2011; Kalyvas, 2006; Lyall et al., 2015), police officers use information from citizens about suspicious behavior and past crimes to allocate their limited personnel and equipment as efficiently as possible (Akerlof and Yellen, 1994; Skolnick and Bayley, 1988; Tyler, 2004; Ungar, 2011; Weitzer and Tuch, 2006). For example, if citizens report a series of vehicle break-ins on a specific street, the police can assign additional patrols to that street, deterring future break-ins and increasing the chances of catching the perpetrator. Without citizens’ reports, the police would be unaware that the street warranted extra attention, causing them to allocate patrols to areas where they are less useful.

The problem is that where historical conflict between identity groups erodes trust between citizens and the state, citizens from marginalized groups are less willing to engage with governing institutions, resulting in inefficient police service provision. Where identity is highly politically salient, and especially where violent conflict along identity lines is likely to occur, inclusion in the police affects citizens’ relationships with the institution. “Minority representation may convey a message of ownership to the minority and undermine the notion that the police are an alien, ‘occupying army’” (Weitzer and Hasisi, 2008). Shared identity provides citizens with a cognitive shortcut for how they should expect to be treated. Lyall et al. (2015) refer to this shortcutting as “coethnic bias.” Citizens need not view inclusion as normatively desirable or fair. Rather, the accumulation of lived experiences and experiences communicated by family and friends leads reasonable people to develop expectations that individuals from their own group will treat them better than individuals from other groups. Thus, a police force that includes officers from a citizen’s group reduces his expected costs of coming into contact with officers by signaling that the institution is trustworthy (Lyall, 2010) and committed to just treatment. In divided societies, where identity motivates political behavior, it is reasonable for a civilian to expect that he is less likely to be mistreated by an authority figure who shares his group identity. In contrast, exclusion from an institution signals that the institution privileges a particular group, so citizens from excluded groups may prefer to avoid contact with officers.

Coethnic bias also raises the expected benefits to citizens from reporting information when the police are inclusive. Citizens contact the police because they expect to benefit from reduced local crime or improved personal safety, or perhaps out of a sense of normative justice. The realization of these benefits depends on the effort the police exert in following up on their information. Coethnic bias makes citizens believe that outgroup police officers may not take their information seriously or exert effort to follow up on crimes occurring in their community. Thus, even if citizens do not fear overt mistreatment, they may be unwilling to incur mundane costs like waiting on hold for an emergency operator or walking to a police station to report a crime if they do not believe the police will reciprocate their effort.

Critically, citizens’ decisions to engage with police and security forces depend on their perceptions of inclusion. Officers’ visibility allows citizens to form perceptions of institutional inclusion based on the appearance and behavior of these officers. In most divided settings, the relevant identity cleavage has observable characteristics, including skin color, names, language or accents, and even clothing. Citizens form perceptions of the police based on direct observations of and interactions with officers, as well as media reports and conversations with friends and family (Mazerolle et al., 2013; Saunders et al., 2013). Online Appendix A shows that in Israel, civilians’ perceptions of police inclusiveness in their city correlate closely with actual levels of police inclusiveness. Thus, actual officer demographics drive citizens’ behaviors via the perceptions they form based on those demographics. Finally, the role of direct observations in forming perceptions implies that inclusiveness in domestic law enforcement should shape citizens’ behavior more than does inclusiveness in the military, as citizens are much more likely to directly observe police officers compared with soldiers.

These arguments lead to several expectations about the effects of police inclusion on citizens’ attitudes and behaviors:

H1: Citizens from marginalized groups will be more willing to report crimes to the police when they perceive the police as inclusive.

If increased crime reporting when the police are inclusive makes the police more effective at preventing crime, then,

H2a: Citizens will be less likely to experience crime when the police in their location are inclusive.

Given that police inclusiveness should increase reporting rates primarily among citizens from marginalized groups,

H2b: The effect of inclusion on crime will be strongest in locations with large populations of historically marginalized groups.

Finally, because public safety is a pseudo-public good which tends to be provided non-excludably at the local level, if the gains in police service provision come from increases in information available to the police rather than from officers exerting greater effort to serve coethnics compared with non-coethnics, those gains should apply to the entire community, not just those who provide information. For instance, all citizens within a community benefit from the police allocating more officers to crime hotspots. Thus, within a location, increased information will improve service provision for all citizens, regardless of whether they personally become more likely to report crimes.

H2c: Within a location, the effect of inclusion on crime should be constant across citizens of all groups.

Several assumptions govern these arguments. First, officers must want to provide effective public safety. If officers fail to act on information provided by citizens, then improved police–community relations should not influence crime. I make no claims about the effects of inclusion on officers’ biases toward members of different groups. While it is possible that working closely with outgroup officers may affect bias over the long term (Allport, 1954), I do not test that possibility in this article. Similarly, the institution and its officers must be capable enough not to squander information that they receive. Second, these hypotheses assume that officers do not intentionally and systematically mistreat civilians from a particular group; if they did, policing would not be a pseudo-public good as one group would become systematically worse off as policing services increase. For hypothesis 2c to hold, the police must treat each segment of society neutrally at worst.

Additionally, there must not be backlash against inclusion by citizens from the dominant group. While conventional wisdom suggests that group identity tends to be most salient for smaller groups (Lyall et al., 2015), implying that improved attitudes from inclusion among Arabs should outweigh negative reactions by Jews, some research finds that members of the majority exhibit backlash against outgroup members when their presence increases (Adida et al., 2016b). This backlash effect refers to changes in the overall population share; dominant-group members develop a distaste for outgroups when they become “surrounded” by them (141). The dynamic of police (or other institutional) integration is quite different, as the overall population share of the integrated group is held constant while it receives increased access to certain specific positions. With a ratio of just one police officer for every 285 civilians, and officers concentrated in dangerous or crime-ridden areas, integration of the police rank-and-file does not substantially affect most Jewish citizens’ exposure to non-Jews. Thus, the specific mechanisms that lead to backlash against changing population demographics are less of a concern in the context of the police rank-and-file integration.

There are, of course, other reasons why Jewish Israelis might oppose police integration. The dominant group might worry that integration removes job opportunities from their group, causing them to provide less information and reducing the overall quality of information available to the police. Furthermore, the extent of minority inclusion almost certainly affects the degree to which the dominant group feels threatened by integration. In cases like Israel, where plausible integration means moving from about 13% non-Jewish police officers to perhaps 20%, and political control remains in Jewish hands, there is little reason for Israeli Jews to worry that inclusion will erode their dominant status. Even so, the concern deserves further attention. The survey conducted for this project asked 403 Jewish Israelis about the importance of job loss and potential subversion from police integration. Online Appendix B shows that an overwhelming majority of respondents felt that these were not major concerns.

Public safety, citizen trust, and crime in Israel

I test these hypotheses in Israel. Ascriptive identity is highly politically salient, with ethnic and religious groups making up distinct partisan blocs. This analysis focuses on the broadest and arguably most salient of these cleavages, the division between Jewish and non-Jewish, primarily Arab, citizens. About 20% of Israeli citizens are Arabs.⁴ While Israeli governing institutions nominally provide equal benefits to all citizens, regardless of identity, in practice service provision is highly unequal. Schools in Arab areas receive only a fraction of the funding of schools in Jewish areas (New Israel Fund, 2005), while Arab citizens complain that the police dedicate inadequate resources to providing services in their neighborhoods (Abraham Fund Initiatives, 2013; Ben-Porat and Yuval, 2012).

The Israel Police are a national police department under the purview of the Ministry of Public Security. The country is divided into six policing districts corresponding to Israel’s main geographic regions, which are then served by officers distributed across approximately 70 police stations. As a full-service department, officers handle traffic enforcement, organized crime, and counterterrorism in addition to ordinary crime prevention. Approximately 28,000 full-time sworn officers serve Israel’s 8 million citizens.⁵ Like police in the US, but unlike in many European departments, all officers are armed.

In 2014, approximately 13% of all officers were non-Jewish (primarily Arab), compared with about 21% of the general population. When the non-Jewish category is broken into its constituent groups, considerable differences emerge. Most notably, Muslims make up only about 4% of full-time officers despite being more than 17% of the population. Druze,⁶ on the other hand, are over-represented with 7.4% of police officers compared with only 1.6% of the population. However, despite the important social and political differences between Arabs of different religions, as well as between Jews of different ancestries, the overarching dichotomy of “Arabs versus Jews” dominates political discourse and remains the most important division with regard to security. As such, I focus my analysis along this cleavage.

Since 2005, the police have engaged in a series of recruitment drives in an effort to increase the number of minority officers.⁷ These drives represent an explicit attempt by the police leadership to diversify the institution’s rank-and-file based on the belief that citizens in minority communities will respond positively to officers from their own community.⁸ These recruitment drives have produced only incremental changes in inclusion, however. Only 14% of the 2016 recruiting class were religious minorities, compared with 10.5% of non-Jewish officers in 2005.⁹

Tensions between Jews and non-Jews run high and periodically spill over into violence. When they do, the police take center stage. The Temple Mount (Haram ash-Sharif) in Jerusalem is a frequent flashpoint for riots, leading to conflict between Muslim residents of Jerusalem and the almost-exclusively Jewish police force assigned to the area. For example, in September 2009 a large group of Muslim worshipers attacked several non-Muslim tourists and police officers, leading to at least 35 reported injuries.¹⁰ Violence spilled into the surrounding neighborhoods, and clashes between Arab residents and the police continued for several weeks. Similar outbreaks of violence, which typically involve stone throwing by rioters and the use of tear gas and stun grenades by police, occurred in 2015 and 2016.¹¹

One of the most formative events in the relationship between Israeli Arabs and the police occurred in October 2000. Following weeks of protests, strikes, and riots by Arab citizens motivated by perceived government mistreatment, the police initiated a violent crackdown against the protesters and rioters.¹² They deployed thousands of extra officers throughout the country, including snipers using live ammunition.¹³ In the first eight days of October, police killed 12 Israeli-Arab citizens and one non-citizen Palestinian. The events heavily influenced Arab relations not just with the police but also with the government more generally, which many citizens viewed as responsible for the heavy-handed police tactics. A report released in 2003 following an official inquiry criticized the police for being unprepared for the riots and for using excessive force.¹⁴ Although a number of officers received disciplinary sanctions, the inquiry did not lead to any significant structural changes in the way that the Israel Police interacts with Arab citizens. Years later, Arabs frequently point to the October 2000 events as a prime example of heavy-handed police tactics against the non-Jewish population.¹⁵

Crime rates in Israel are on par with most Western European countries.¹⁶ As in most places, crime is more common in less affluent cities and neighborhoods.¹⁷ Official figures at the sub-national level are difficult to come by, but according to the Israel Police crime in Jerusalem has declined over the past several years.¹⁸ On the other hand, in large cities like Ashkelon, Modi’in, and Beer Sheva, residents have experienced increases in crime. Tel Aviv has the highest number of crimes per capita.¹⁹ Victimization surveys carried out by Israel’s Central Bureau of Statistics provide further insights into the frequency of each type of crime independently of reporting rates. In 2013, 5.4% of Israelis experienced a home break-in and 4.7% experienced a non-violent theft. Two per cent of Israelis had their vehicles stolen, and 1% were victims of violent crimes. Jews were more likely to be the victim of some crimes, including home break-ins and non-violent crimes, while Arabs were more likely to experience vehicle break-ins and assaults.²⁰

It is possible that shared identity between a criminal and victim might affect the victim’s willingness to contact the police. For instance, Arabs might be less likely to contact the police if attacked by another Arab, and this tendency could further interact with the identity of the responding police officer. Unfortunately, there is no systematic data on the religious identities of criminals compared with their victims in Israel. Generally speaking, however, criminals tend to commit crimes very close to home (Phillips, 1980; Rengert, 2004). Aside from a few exceptional neighborhoods in Akko, Jaffa, and Haifa, Israeli cities and towns are segregated, meaning that most crime in Israel is probably perpetrated by coethnics, and there is little variation in criminal–victim shared identity. Furthermore, victims do not always know the identity, much less the religious affiliation, of the perpetrator. Thus, while shared identity between criminals and victims is an important variable that should be explored directly in future work, I speculate that it is unlikely to explain a substantial portion of the variation in citizens’ willingness to report crimes in Israel.

Despite the police’s high levels of competence on objective measures like crime prevention relative to other developed countries, police–community relations in Israel are generally poor, and Israeli citizens express less trust in their police than do citizens in countries with similarly effective police forces (Saunders et al., 2013; Yogev, 2010). Israelis’ attitudes are shaped by both their perceptions of identity-based conflict and their personal ethnic and religious affiliations. An analysis of media accounts of the Israeli Police notes that different religious communities often interpret the same police actions differently (Saunders et al., 2013). Arab and other underrepresented citizens tend to have less trust in the police than do their Ashkenazi Jewish counterparts (Saunders et al., 2013; Ben-Porat and Yuval, 2012), a difference that impacts crime reporting. A 2008 study (Hasisi and Weisburd, 2014) finds that 70.1% of Jews said that they would call the police if they witnessed a crime, compared with 50.6% of non-Jews. Similarly, Figure 1 shows that among respondents surveyed for this article, Jews were much more likely than non-Jews to say that they would contact the police if they witnessed someone stealing from a shop.

Figure 1.

Willingness to report crimes.

The Israeli-Arab community expresses frustration at the level of policing services it receives, a problem citizens attribute to the Arab–Jewish division (Ben-Porat and Yuval, 2012). At a 2013 conference on policing held in the Knesset (Parliament), Arab Member of the Knesset (MK) Esawi Frij said, “In terms of [policing] results, the Arab sector lags behind. There is a gap and distance between the Arab citizen and the police. Violence and illegal firearms are extensive, and personal security is deteriorating. The police take action, but not enough” (Abraham Fund Initiatives, 2013: 1). Arab Israelis are concerned about both under- and over-policing. According to Salim Salibi, head of Majd al-Krum local council: “When there is violence or fighting in the village, we wait hours for the police but they don’t arrive. But when a company comes to collect debts, it is accompanied by police officers from the Special Patrol Unit. This creates a lack of trust with the police.” Salem Abu Ayash, mayor of the predominantly Arab town of Laqiya, says, “When they demolish a home or a shack, we see a whole convoy of police cars—two or three hundred, you can’t see the end of the convoy. When there is a fight in the village, they barely send a patrol car.”²¹

If subpar service provision in Arab communities results from intentional discrimination by the police, then more efficient policing will not improve outcomes for Arab communities. However, it is also possible that high crime in the Arab sector stems from the very information problem that I argue integration solves. Arab-Israelis experience poor treatment in many aspects of their interactions with the state and therefore hesitate to engage with institutions of state authority, including the police. In turn, the police receive less information about Arab communities than Jewish ones, which causes them to serve Arab communities less effectively, feeding into a cycle in which Arab citizens are unwilling to engage with the police and services deteriorate even further. If a significant portion of subpar public safety in the Arab sector stems from poor-quality information, not intentional discrimination, then Israel is within scope and inclusion should lead to better crime prevention in Arab communities.

Testing the mechanism: citizen–police information flows

To test the relationship between perceived police inclusiveness and citizen–police information flows, I conducted a survey of Israeli citizens in spring 2016. The survey included 804 respondents across 15 cities and neighborhoods which provide a cross-section of urban Israeli society. Table 1 shows the distribution of respondents by religious identity from each city. Cities were selected based on known religious demographics with the goal of achieving a politically diverse sample of half Jewish, half non-Jewish respondents. The religious demographics of each sampled location correspond roughly with the ratio of Jewish to non-Jewish respondents as reported in Table 1. Within each city, respondents were selected randomly from an exhaustive list of phone numbers, and respondents chose whether to take the survey in Hebrew or Arabic. Interviews in East Jerusalem were conducted via door-to-door enumeration, as no suitable sampling frame of phone numbers existed. Online Appendix C provides additional details about the sampling procedures.

Table 1.

Israel distribution of respondents

City	Non-Jews	Jews
Akko	34	36
Haifa	65	65
Jerusalem	0	100
East Jerusalem	104	0
Nazareth	123	2
Nazareth Illit	0	50
Rahat	50	0
Sderot	0	50
Tel Aviv-Jaffa	25	100

As in all survey-based research, social desirability bias presents a challenge to validity. Respondents may not answer sensitive items truthfully if they do not want to reveal their preference to the interviewer. While it is impossible to rule out social desirability bias entirely, certain characteristics of this survey guard against it. First, Adida et al. (2016a) find that bias is exacerbated when respondents and enumerators come from groups in conflict with one another. To avoid this possibility, phone interviews took place in the language selected by the subject with a fluent speaker of that language, and in-person interviews in Arab East Jerusalem used Arab enumerators. In-person interviews were conducted in respondents’ homes, and phone interviews in a location of the respondent’s choosing, in an effort to minimize discomfort and ensure a low-pressure environment. Finally, the design of the survey ensured that subjects’ responses, as well as their participation in the survey, remained confidential, decreasing their incentives to answer untruthfully in the face of social pressures. In the Online Appendix, I address the possibility that social desirability bias simultaneously inflates responses to both the independent and dependent variables, which could lead to a spurious relationship.

I measure perceptions of police inclusiveness at both the national and local levels. At the national level, the survey asks, “In general, how integrated would you say the Israeli police are? By integrated I mean that members of many different ethnic and religious groups serve together side by side. Throughout Israel in general, would you say that the police are very integrated, somewhat integrated, or not at all integrated?” Respondents were then asked about their perceptions of the police locally: “Now please think specifically about your neighborhood. How often do you see or hear about Jewish and non-Jewish officers working together? Often, sometimes, rarely, or never?” Figure 2 shows the distribution of responses to these two questions. The use of individual-level survey questions rather than community-level data as a predictor of reporting has several advantages. First, since the dependent variables are individual-level attitudes, individual-level predictors provide the greatest precision. Second, because willingness to report a crime is argued to be based on citizens’ beliefs that the police are inclusive, perceptions about police inclusiveness represent a more precise measure of the mechanism than does actual inclusiveness.

Figure 2.

Perceived police integration.

Results: perceived integration and willingness to report

How does perceived inclusiveness affect citizens’ willingness to provide information to the police? The survey asks respondents about their willingness to report a hypothetical crime: “If you witnessed someone stealing money from a shop, would you contact the police?” This type of crime is unlikely to come to the police’s attention unless reported by a witness or victim, reflecting the specific mechanism through which I argue non-inclusive policing harms crime prevention, and its less serious nature increases the likelihood that respondents might admit that they would not contact the police. I further combat any hesitance to admit to not contacting the police by asking the question as open-ended rather than multiple choice. Enumerators grouped responses into three categories: “yes” with no caveat, “maybe,” including “probably” or similar responses, and “no.” Whereas respondents may have felt pressure to select “yes” when presented with the option, the expectation was that they should feel fewer qualms about answering “probably” when this option was not juxtaposed against an even more positive response.

Control variables are especially important for these tests given the potential endogeneity between attitudinal measures. For example, people who are more positively predisposed in general may be more likely to perceive the police as fair and more likely to go out of their way to provide the police with information. Regression models control for respondents’ religion, age, gender, whether they voted in the last election, economic satisfaction, whether they work in the public sector, and the method of enumeration. All models include dummies for respondents’ city and ethnic group, and standard errors are clustered at the primary sampling unit. Because this specification provides only 15 clusters, I replicate the analysis in Online Appendix E using a wild cluster bootstrap (Cameron et al., 2008). All results hold. Models use ordered logistic regression, and a Brant test provides no evidence that the proportional odds assumption is violated. Given the potential for bias when using fixed effects with logistic regression, and to confirm that results are not sensitive to model choice, I replicate the analysis using ordinary least squares (OLS) in the Online Appendix. Results are consistent (Table 2).

Table 2.

If you witnessed someone stealing money from a shop, would you contact the police?

	(1)	(2)	(3)	(4)
	All	All	Non-Jews	Jews
Integration (national)	0.319**
	(0.126)
Integration (local)		0.270***	0.259**	0.00561
		(0.0933)	(0.109)	(0.128)
Jewish	1.530***	1.659***
	(0.541)	(0.620)
International (local) × Jewish		−0.253*
		(0.147)
Age	0.00471	0.00669	0.0151*	−0.00115
	(0.00681)	(0.00871)	(0.00809)	(0.0159)
Male	−0.00262	−0.0964	0.171	−0.700***
	(0.187)	(0.213)	(0.256)	(0.241)
Vote	0.376	0.285	0.258	−0.124
	(0.244)	(0.199)	(0.227)	(0.363)
Economic Satisfaction	0.0699	0.0708	0.0913	0.115
	(0.0829)	(0.0932)	(0.0988)	(0.125)
Work Public	0.479*	0.462*	0.812***	−0.0183
	(0.263)	(0.274)	(0.240)	(0.489)
In Person	−0.840**		−1.463***
	(0.343)		(0.352)
Observations	659	688	374	314

Ordered logistic regression. Standard errors clustered by Primary Sampling Unit.

Models include city and ethnic group dummies.

p = 0.10, **p = 0.05, ***p = 0.01.

Table 1 shows that respondents who perceive the police as more integrated at the national level (Column 1) are significantly more likely to say that they would report the hypothetical crime. However, citizens’ expectations of how the police will treat them should be shaped primarily by the inclusiveness of the officers they actually come into contact with, i.e. those in their neighborhood. Column 2 shows that, when it comes to perceived police inclusiveness at the local level, there is a significant difference in the effect of perceived inclusion on reporting between Jews and non-Jews. Non-Jews, the historically marginalized group, are significantly more likely to report the shop theft to the police when they perceive the police as integrated (Column 3), while Jewish respondents are neither more nor less likely to report the crime (Column 4). As predicted, members of the historically marginalized group who perceive the police in their community as inclusive are more likely to report information, but there is no evidence of a backlash by members of the dominant group against inclusion.

Controls generally have the expected effects. Non-Jews are less likely to report crimes than Jews. Those who work in the public sector or have a family member who works in the public sector are more likely to report a crime. Finally, non-Jewish respondents surveyed in-person rather than by phone are less likely to say that they would report the crime to the police. In all likelihood, this effect is caused not by survey method but by location: residents of East Jerusalem tend to have a worse relationship with the police.

The strong, positive association between perceived inclusiveness and willingness to report crimes suggests that inclusion substantially affects the quality of information available to the police, and that it does so primarily in locations with many Arab residents. The next section tests whether local inclusion is indeed associated with improvements in public safety provision.

Testing inclusion’s effects on public safety

The dependent variable is the probability of crime which, all else equal, should be lower when the police provide services more efficiently. Data comes from Israel’s Social Survey, a nationally representative survey of approximately 7000 respondents carried out by the Central Bureau of Statistics each year. All individuals ages 20 and over living in Israel are eligible to be included in the survey. The Social Survey covers a wide range of topics, including economic and employment status, language and technological skills, education, and basic household information. Most relevant to this analysis, it asks a set of questions about crime victimization, including whether the respondent experienced a vehicle break-in, home break-in, or non-violent theft in the past year. Unlike serious crimes like murder or vehicle theft, these crimes are unlikely to come to the police’s attention unless reported by citizens. I use two measures of crime victimization. First, I create an additive index ranging from 0 to 3, with 0 indicating that the respondent was not the victim of any of these crimes in the past year and 3 indicating that the respondent experienced all three types of crimes. I replicate all models using an alternative coding equal to 1 if the respondent experienced any of these crimes in the past year, or 0 if they did not.

Survey measures of crime victimization are preferable to official crime statistics reported by law enforcement (Banerjee et al., 2012). Official crime statistics include only the subset of crimes that were brought to the attention of the police, in most cases by being reported by citizens. The problem with using these numbers for analysis is that many of the same factors which make the police more effective at crime-fighting, for example increased citizen cooperation, also increase the proportion of crimes that are reported to the police. Thus, when trust in the police goes up, actual crime may decrease while increased reporting gives the appearance of increased crime. To the extent that respondents are willing to answer questions about victimization truthfully, the victimization survey achieves a more accurate measure of criminal activity. Answering a direct “yes or no” question from a person to whom they are already speaking requires considerably less effort than contacting the police to file an official report, making it more likely that respondents will reveal crimes that occurred, and any bias introduced by enumerators should be orthogonal to respondents’ trust in the police.

I test crime victimization against the key independent variable, police inclusion. While perceptions of inclusion would be the most direct measure of the construct of interest, the social survey does not contain such a measure, nor am I aware of any other measures of such perceptions at the sub-district-year level. Instead, this section analyzes actual levels of police inclusion using a unique dataset of police officer demographics. I collected yearly data on the number of police officers from each religious group at each police station between 2008 and 2014.²² I aggregate station-level data on officer demographics to the sub-district level to match the smallest available geographical unit in the Social Survey.²³ I then calculate police inclusiveness using the formula for ethno-linguistic fractionalization. The resulting measure is the probability that any two randomly selected police officers would be members of different religious groups. A measure of 1.00 on Police Integration indicates that all officers are from different religious groups, while a measure of 0.00 means that all officers are from the same group. Sub-district-level Police Integration ranges from 0.014 to 0.611, with a mean of 0.211. Given that there are no police stations at which a majority of officers are non-Jewish, an increase in Police Integration is conceptually identical to an increase in inclusion for non-Jews.

Models include several time-varying controls that may influence the likelihood of crime victimization. Individual-level controls come from the Social Survey and include gender, home ownership, whether or not the respondent is Jewish (a rough proxy for both economic status and neighborhood type), and military service. A dummy variable for urban–rural status of the respondent’s neighborhood also comes from the Social Survey. District-year unemployment estimates were collected from Israel’s Central Bureau of Statistics. Some models include sub-district indicator variables to account for systematic differences between sub-districts while others include Time, a count variable of the number of years since the start of the panel, to account for possible concurrent time trends in both police integration and crime rates. Summary statistics for all variables are available in Online Appendix D.

Results: police inclusion and crime

As Table 3 shows, the results of several regression models corroborate the hypothesis that local-level police integration is negatively associated with crime victimization. The analysis covers the years 2008–2014, but excludes 2010 owing to concerns about data quality.²⁴ Models 1 and 3 include sub-district dummy variables, meaning that they test the effects of change over time in police inclusion within a given locality. Standard errors are clustered by sub-district-year, the level of measurement for police integration, providing 76 clusters. Online Appendix E reproduces Model 4 but clusters standard errors by sub-district and uses a wild cluster bootstrap. Models 1 and 2 regress the additive crime index on integration using negative binomial regression as the dependent variable is skewed toward 0 (i.e. most respondents where not the victim of any crimes). Models 3 and 4 use the alternative specification of crime victimization, a dummy indicator of whether or not the respondent experienced at least one crime in the past year, and employ logistic regression.

Table 3.

Police inclusion and crime.

	(1)	(2)	(3)	(4)
	Crime	Crime	Crime (alternative)	Crime (alternative)
Police Integration	−3.449**	−0.0644	−4.083**	0.0359
	(1.635)	(0.169)	(1.720)	(0.202)
Arab Sub-district	0.224	1.510***	0.165	1.752***
	(0.210)	(0.303)	(0.202)	(0.361)
Police Integration ×		−3.483***		−4.014***
Arab Sub-district		(0.679)		(0.797)
Urban Sub-district	0.0574	0.0891*	0.0889*	0.127**
	(0.0498)	(0.0517)	(0.0540)	(0.0537)
Male	0.0646**	0.0623*	0.0703*	0.0671*
	(0.0319)	(0.0318)	(0.0362)	(0.0365)
Home Owner	−0.171***	−0.184***	−0.187***	−0.202***
	(0.0296)	(0.0297)	(0.0354)	(0.0358)
Non-Jew	0.0254	0.00324	0.0732	0.0417
	(0.0724)	(0.0699)	(0.0786)	(0.0738)
Military	0.0589*	0.0680**	0.0662*	0.0780*
	(0.0336)	(0.0339)	(0.0398)	(0.0399)
District Unemployment	5.916	−3.268	6.883	−3.631
	(4.399)	(2.006)	(4.572)	(2.305)
Time		−0.0995***		−0.111***
		(0.00999)		(0.0118)
Observations	22,904	22,904	22,904	22,904

Models 1 and 2, negative binomial regression; Models 3 and 4, logistic regression.

SE clustered by sub-district-year. Models 1 and 3 include sub-district dummies.

p = 0.10, **p = 0.05, ***p = 0.01.

Models 1 and 3 show that within a sub-district, an increase in police inclusiveness is associated with a decrease in the probability that an individual will be the victim of a vehicle break-in, home break-in, or non-violent theft. This finding is consistent with the argument that police inclusion improves the quality of service provision (Hypothesis 2a). The finding is robust to both measures of victimization, as well as the inclusion of various controls for standard determinants of crime. As expected, urban sub-districts have more crime (although the coefficient is only marginally significant), and men are more likely to be the victim of a crime than women. Homeowners, likely to be more affluent, are significantly less likely to be the victim of a crime. The coefficient on district-level unemployment is not statistically significant, perhaps owing to the higher level of aggregation (six districts rather than 15 sub-districts).

Models 2 and 4 add an interaction term between police inclusiveness and whether a sub-district has a substantial Arab (non-Jewish) population in a given year.²⁵ These models test Hypothesis 2b, that inclusion should have a stronger effect in communities with large populations of historically excluded groups, and Hypothesis 2c, that within community-type the effect will be constant across citizens of all groups. The results are best interpreted through the visualization in Figure 3, which displays the marginal effect of police inclusion on the probability of an individual experiencing at least one crime in the prior year calculated on Model 4.

Figure 3.

Effect by community and individual characteristics.

In Figure 3, the left-hand panels show the effect in predominantly Arab sub-districts, while the right-hand panels show the effect in predominantly Jewish sub-districts. Police inclusion is negatively and significantly associated with crime only in Arab sub-districts. This finding is consistent with the argument that police inclusion induces civilians from previously excluded groups to provide the police with information, allowing the police to provide crime prevention more efficiently. It is also consistent with the finding that non-Jews, but not Jews, who perceive the police in their location as more inclusive are more willing to report crimes. The top panels show the effect for non-Jewish (predominantly Arab) citizens, while the bottom panels show the effect for Jewish citizens. Within a community type (i.e. large Arab population vs small Arab population), there is no perceptible difference in inclusion’s effects on crime between Jewish and non-Jewish citizens. This finding is consistent with the argument that the police use information from citizens to provide crime prevention as efficiently as possible, and contradicts the alternative argument that police integration privileges only members of the previously excluded groups because officers provide preferential treatment for their coethnics.

Selection bias in officer assignments

Selection bias presents a major concern. If the police assign minority officers to areas with less crime as a matter of policy, the correlation between inclusiveness and crime may be spurious. While the sub-district dummies in the regression models account for inherent differences between locations, I further explore this possibility by testing whether communities that are crime prone are less likely to have inclusive police. According to existing research, crime is highest where unemployment and the proportion of male civilians are high (Raphael and Winter-Ebmer, 2001), while conventional wisdom in Israel suggests that Jewish cities are less crime-prone than Arab ones owing to socioeconomic factors.²⁶ However, I find a strong positive relationship between unemployment and police inclusiveness, as well as between percentage male and inclusiveness, and a strong negative relationship between percentage Jewish and police inclusiveness. In other words, the police are most integrated in areas with high baseline probabilities of crime, biasing against the finding that integration is associated with less crime (Table 4).

Table 4.

Predictors of police inclusiveness.

	(1)	(2)	(3)
Unemployment	8.017***
	(1.009)
Percentage Male		1.846**
		(0.765)
Percentage Jewish			−0.721***
			(0.0970)
Constant	−0.285***	−0.668*	0.809***
	(0.0634)	(0.377)	(0.0811)
Observations	76	82	82

OLS regression with robust standard errors in parentheses.

p = 0.10, **p = 0.05, ***p = 0.01.

To further explore the possibility of selection bias, I interviewed a Brigadier General in the Israel Police who is in charge of officer assignments.²⁷ According to this officer, the police do not systematically assign minority officers based on a location’s propensity for criminal activity. Rather, one of the top criteria for locating officers is to assign them close to, but not in, their hometown. Assigning officers close to home saves commuting and relocation expenses, while avoiding placing officers in their home towns, especially in rural areas, reduces corruption. The region in which a new officer is likely to be assigned is essentially a donut shape surrounding his or her home town, with low probability inside the hole or outside the ring. The police also try to avoid having too high a percentage of Arab officers in Arab communities so that the police are not coopted by the hamula (tribal) system. The police split up ethnic or religious minority officers when possible so that any given interaction is likely to have “at least one real policeman,” i.e. a full-time Jewish officer with experience. The concern, according to this officer, is not with the abilities of minority officers—who the police consider to be just as competent as any other officer—but with the way that citizens react to them. Thus, while officer assignment is far from random, there is no evidence that the police systematically integrate (or avoid integrating) officers in areas that are more or less susceptible to crime.

A closely related concern is that Arab communities which have a good relationship with the police are both more likely to cooperate with the police, improving crime prevention, and more likely to send recruits to the police, causing integration. There is little doubt that positive relations between Arabs and the police aid recruitment of Arab officers. The effect of this relationship on the outcomes in Table 3 depends on the way those officers are distributed once recruited. The police’s policy of not assigning officers (especially Arab officers) to their home towns for fear of corruption means that the areas where positive citizen–police relations cause a greater number of Arabs to join the police will not necessarily receive more Arab officers. Thus, while I cannot eliminate this source of endogeneity entirely, police procedures for officer assignment suggest that its impacts should be limited.

Alternative arguments: officer quality and institutional heterogeneity

The link between inclusion and crime assumes that changes in police demographics do not coincide with changes in the average quality of police officers. However, if officers from one group are more effective at crime fighting than officers from another group, the effects of integration proposed above may be endogenous to changes in the quality of police officers. The police may have difficulty recruiting officers from certain groups and therefore use lower hiring standards in order to meet recruiting targets (Lott, 2000). Alternatively, systematic discrimination in social or economic opportunities may lead to lower educational attainment for some groups compared with others, affecting the quality of available recruits.

Recruitment strategies used by the Israel Police counteract these societal inequalities. The police give hiring preference for university completion, which they view as a signal of maturity, independence, and the ability to follow through on tasks.²⁸ The use of this and other high-skills indicators as hiring qualifications narrows the educational gap between recruits of different religious groups. Furthermore, the inequalities in Israeli society imply that, on average, the pool of non-Jewish applicants might be less qualified, which biases against the finding that including more non-Jewish police officers improves service provision.

A related concern is that officers who are “too close” to the communities in which they serve might be less professional, engage in favoratism, or be more susceptible to corruption. These concerns motivate the assignment criteria described above in which officers are assigned near to, but not in, their home town. Policymakers implementing police integration must be sensitive to these types of challenges and balance the benefits of shared identity with the costs of officers who are overly familiar with civilians in their jurisdiction.

Another possibility is that forcing officers to work with outgroup partners may cause inefficiencies. Ethnic and religious heterogeneity within an institution are sometimes argued to have adverse effects for collective action (Alesina et al., 1999; Banerjee et al., 2005; Habyarimana et al., 2007; Laitin, 2007; Miguel and Gugerty, 2005). Yet, rather than impeding police service provision, communities in Israel with more diverse police forces benefit from lower crime. I suggest that concerns about inefficient cooperation between outgroup officers, while valid, are mitigated by the rigid structures of security institutions. Police officers, like members of other security institutions, operate under a strict chain of command and are expected to follow the orders of their superiors with little questioning. Officers may also develop a sense of “linked fate” as they work together to combat crime and other security threats, often risking their lives side-by-side to do so. Police officers develop shared rituals, especially surrounding their own mortality.²⁹ They foster a similar appearance by wearing uniforms and maintaining similar hair styles, and officers must meet the same training, educational, and physical fitness standards. Finally, officers develop a “subculture” which emphasizes fraternity and cohesion (Crank, 2014). These shared experiences and rigid formal institutional structures may replace shared group heuristics as a basis for cooperation, even while ascriptive group identity remains salient (Gibson and Hoffman, 2013). An experiment among police officers in Liberia finds no adverse effects of heterogeneity on collective action (Blair et al., 2016). Similarly, analysis of a survey of 2800 officers in the Los Angeles Police Department concludes that “police officer attitudes toward the community are more likely a product of occupational socialization than demographic background,” suggesting that institutional socialization mitigates the effects of preexisting identity cleavages (Lasley, 1994: 95–6). The findings in this article add to a growing body of evidence that institutional structures common in service-providing institutions, and especially those with rigid military-like structures, can overcome the negative aspects of heterogeneity.

Generalizing beyond Israel, policing

Israel’s highly entrenched identity conflict and the availability of unusually granular administrative data make it an excellent case in which to test this theory of security forces inclusiveness. However, the argument about inclusion, information, and security provision applies to a wide range of divided societies. Research and media accounts in countries as diverse as Northern Ireland, Lebanon, Iraq, the United States, and Liberia mention the makeup of the police as a potential determinant of police effectiveness (Bayley, 2008; Blair et al., 2016; McGarry, 2000; Perito, 2011; Robinson, 2009; Weitzer, 1995; Weitzer and Tuch, 2004, 2006). As discussed above, bureaucrats must want to provide services effectively, the police must be capable enough to take advantage of information citizens provide, and there must not be a backlash by the dominant group against inclusion. Beyond these scope conditions, the effects of inclusion on security should generalize to other divided societies. The positive record of police integration in at least one case matches the findings from Israel. Integration of Catholics into the police in Northern Ireland was a major component of the 1998 Good Friday Agreement (Bayley, 2008; Ellison, 2007; McGarry, 2000). Since then, a steady increase in the number of Catholic officers has coincided with increased public trust in the police and improved public safety provision, and the broad trends reported in research from Northern Ireland (Bayley, 2008; Byrne and Monaghan, 2008; Northern Ireland Policing Board, 2016) are consistent with the more granular quantitative evidence from Israel presented here.

Furthermore, the underlying argument about inclusion’s effects on citizens’ perceptions, information provision, and efficiency applies to all institutions which rely on community cooperation to provide public goods and services. A wide range of bureaucratic institutions depend on citizen buy-in and cooperation to serve citizens efficiently. Public works departments must locate potholes, school administrators must distinguish between effective and ineffective teachers, and healthcare providers must evaluate their current offerings against citizens’ needs. These institutions, and many others, rely on citizens to provide information which allows them to operate efficiently. In divided societies, institutional inclusion affects citizens’ expectations about how bureaucrats will treat them, and consequently their willingness to go out of their way to provide information to them. Future research on public goods provision in divided societies should consider the effects of representation on citizen engagement independently of its effects of bureaucrats’ behavior.

Conclusion

I find that inclusiveness in the police is associated with an improvement in the quality of public safety provision, and the evidence is consistent with the argument that it does so because marginalized groups are more willing to provide information to security institutions they perceive to be inclusive. This finding contrasts with the suggestion in existing research that the critical link between bureaucratic representation and service provision is “active representation” by bureaucrats who favor ingroup civilians. Whereas active representation implies a distributive effect, with well-represented groups receiving government services at the expense of poorly represented groups, I find that police inclusiveness improves public safety for all citizens living in communities where information flows increase, regardless of their personal group affiliation. By shifting the focus from the behavior of police officers to the behavior of citizens, I identify an important mechanism through which inclusion in domestic security institutions leads to net gains in public safety.

The net benefits of inclusion on public safety have important implications for institutional design as a solution to violent intergroup conflict. Existing research suggests that inclusion in policy-making institutions provides more effective, democratic governance in divided societies (Horowitz, 1985; Lijphart, 1969, 1984; Strøm et al., 2017). However, traditional power sharing can be difficult to implement in conflict or post-conflict settings because parties to the conflict conceptualize representation as zero sum, giving those who benefit from the status quo reason to oppose integration (Roeder and Rothchild, 2005). The finding that the effects of inclusiveness on service provision are not zero sum, and in fact lead to improvements for all members of a community, makes domestic security institutions an appealing setting for power sharing as a solution to conflict. Even if police integration disproportionately benefits communities where previously excluded groups live, it does not mean that the dominant group will be made worse off. Thus, police integration represents a plausible first step toward resolving hostilities where tensions between groups are too high to implement traditional power sharing. Inclusion in law enforcement can generate sufficient trust to open the door for future reforms, leading to a decrease in tensions and improvements in governance.

These findings contradict the argument that heterogeneity harms public goods provision (Alesina et al., 1999; Banerjee et al., 2005; Habyarimana et al., 2007). Consistent with another recent study on policing (Blair et al., 2016), I find no evidence that heterogeneity among police officers harms their ability to provide services. At least in Israel, the benefits to service provision realized through increased information flows from citizens vastly outweigh any potential harm caused by officer heterogeneity. I add to a growing body of evidence that the rigid institutional structures like those found in the police and military reduce the effects of heterogeneity on efficiency by providing alternative coordination points, a shared language, and shared experiences among institutional members. Future research might test the effects of specific institutional structures, and especially those found in security institutions, on the behavior of service providers like police officers and soldiers to determine how serving in such rigid settings interacts with demographic diversity.

Finally, more work is needed on the dynamics of inclusion at different levels of integration; that is, how does integration impact police behavior, citizen cooperation, and dominant group backlash at different levels of inclusion? Are there certain thresholds for inclusion which must be met before inclusion reduces crime? Is there a point beyond which members of the dominant group say “enough is enough”? Future research might explore institutional inclusion as a continuum with the potential for different outcomes at different levels.

Supplemental Material

CMP_802580__Online_Appendix – Supplemental material for Policing in divided societies: Officer inclusion, citizen cooperation, and crime prevention

Supplemental material, CMP_802580__Online_Appendix for Policing in divided societies: Officer inclusion, citizen cooperation, and crime prevention by Matthew J Nanes in Conflict Management and Peace Science

Footnotes

Acknowledgements

The author thanks Alexei Abrahams, Claire Adida, Konstantin Ash, Karen Ferree, David Laitin, Nazita Lajevardi, Brandon Merrell, Asfandiyar Mir, Phil Roeder, Deborah Seligsohn, the Stanford FSI Postdoc Working Group, and several anonymous reviewers for their comments on various drafts of this paper, and Badi Hasisi, Simon Perry, and David Weisburd for their invaluable introduction to policing in Israel during the early stages of this project.

Funding

This research was funded by an Israel Institute Doctoral Fellowship.

Supplemental material

All data, replication materials, and instructions regarding analytical materials upon which published claims rely, along with the Online Appendix, are available online through the SAGE CMPS website:

Notes

References

Abraham Fund Initiatives (2013) Conference—Policing in a Divided Society: Arab Society Police Relations in Israel. Available from: http://www.abrahamfund.org/webfiles/fck/Conference_–_Policing_in_a_Divided_Society.pdf.

Adida

Ferree

Posner

Robinson

(2016a) Who’s asking? Interviewer coethnicity effects in African survey data. Comparative Political Studies 49(12): 1630–1660.

Adida

Laitin

Valfort

M-A

(2016b), “One Muslim is enough!” evidence from a field experiment in France. Annals of Economics and Statistics/Annales d’Économie et de Statistique 121/122: 121–160.

Akerlof

Yellen

(1994) Gang Behavior, Law Enforcement, and Community Values. Canadian Institute for Advanced Research.

Alesina

Baqir

Easterly

(1999) Public goods and ethnic divisions. The Quarterly Journal of Economics 114(4): 1243–1284.

Allport

(1954) The Nature of Prejudice. Reading, MA: Addison-Wesley.

Andersen

(2017) From passive to active representation—Experimental evidence on the role of normative values in shaping white and minority bureaucrats’ policy attitudes. Journal of Public Administration Research and Theory 27(3): 400–414.

Banerjee

Iyer

Somanathan

(2005) History, social divisions, and public goods in rural India. Journal of the European Economic Association 3(2–3): 639–647.

Banerjee

Chattopadhyay

Duﬂo

Keniston

Singh

(2012) Can institutions be reformed from within? Evidence from a randomized experiment with the Rajasthan police. CEPR Discussion Paper (DP8869).

10.

Bayley

(2008) Post-conﬂict police reform: Is Northern Ireland a model? Policing 2(2): 233–240.

11.

Ben-Porat

Yuval

(2012) Minorities in democracy and policing policy: From alienation to cooperation. Policing and Society 22(2): 235–252.

12.

Berman

(2000) Sect, subsidy, and sacrifice: An economist’s view of ultra-orthodox Jews. The Quarterly Journal of Economics 115(3): 905–953.

13.

Berman

Shapiro

Felter

(2011) Can hearts and minds be bought? The economics of counterinsurgency in Iraq. Journal of Political Economy 119(4): 766–819.

14.

Besley

Coate

Loury

(1993) The economics of rotating savings and credit associations. The American Economic Review 83(4): 792–810.

15.

Blair

Karim

Gilligan

Beardsley

(2016) Policing ethnicity: Lab in-the-field evidence on discrimination, cooperation and ethnic balancing in the liberian national police. Social Science Research Network.

16.

Byrne

Monaghan

(2008) Policing Loyalist and Republican Communities: Understanding Key Issues for Local Communities and the PSNI. Institute for Conﬂict Research Belfast.

17.

Cameron

Gelbach

Miller

(2008) Bootstrap-based improvements for inference with clustered errors. The Review of Economics and Statistics 90(3): 414–427.

18.

Chandra

(2006) What is ethnic identity and does it matter? Annual Review of Political Science 9, 397–424.

19.

Crank

(2014) Understanding Police Culture. Abingdon: Routledge.

20.

Ellison

(2007) A blueprint for democratic policing anywhere in the world? Police reform, political transition, and conﬂict resolution in Northern Ireland. Police Quarterly 10(3): 243–269.

21.

Enloe

(1980) Ethnic Soldiers: State Security in Divided Societies. Athens, GA: University of Georgia Press.

22.

Fearon

Laitin

(1996) Explaining interethnic cooperation. American Political Science Review 90(4): 715–735.

23.

Gibson

Hoffman

(2013) Coalitions not conﬂicts: Ethnicity, political institutions, and expenditure in Africa. Comparative Politics 45(3): 273–290.

24.

Grossman

Gazal-Ayal

Pimentel

Weinstein

(2015) Descriptive representation and judicial outcomes in multiethnic societies. American Journal of Political Science 60(1): 44–69.

25.

Habyarimana

Humphreys

Posner

Weinstein

(2007) Why does ethnic diversity undermine public goods provision? American Political Science Review 101(4): 709–725.

26.

Hardin

(1997) One for All: The Logic of Group Conﬂict. Princeton, NJ: Princeton University Press.

27.

Hasisi

Weisburd

(2014) Policing terrorism and police–community relations: Views of the Arab minority in Israel. Police Practice and Research 15(2): 158–172.

28.

Hoddie

Hartzell

(2003) Civil war settlements and the implementation of military power-sharing arrangements. Journal of Peace Research 40(3): 303–320.

29.

Horowitz

(1985) Ethnic Groups in Conﬂict. Berkeley, CA: University of California Press.

30.

Kalyvas

(2006) The Logic of Violence in Civil War. Cambridge: Cambridge University Press.

31.

Kennedy

Butz

Lajevardi

Nanes

(2017) Race and Representative Bureaucracy in American Policing. Basingstoke: Palgrave Macmillan.

32.

Laitin

(2007) Nations, States, and Violence. Oxford: Oxford University Press.

33.

Lasley

(1994) Ethnicity, gender, and police-community attitudes. Social Science Quarterly 75(1): 85–97.

34.

Lijphart

(1969) Consociational democracy. World Politics 21(2): 207–225.

35.

Lijphart

(1984) Democracies. New Haven, CT: Yale University Press.

36.

Lijphart

(2012) Patterns of Democracy: Government Forms and Performance in Thirty Six Countries. New Haven, CT: Yale University Press.

37.

Lipsky

(1980) Street-level Bureaucracy: Dilemmas of the Individual in Public Services. New York: Russell Sage Foundation.

38.

Lott

(2000) Does a helping hand put others at risk?: Affirmative action, police departments, and crime. Economic Inquiry 38(2): 239–277.

39.

Lyall

(2010) Are coethnics more effective counterinsurgents? Evidence from the second Chechen war. American Political Science Review 104(1): 1–20.

40.

Lyall

Shiraito

Imai

(2015) Coethnic bias and wartime informing. The Journal of Politics 77(3): 833–848.

41.

Mazerolle

Antrobus

Bennett

Tyler

(2013) Shaping citizen perceptions of police legitimacy: A randomized field trial of procedural justice. Criminology 51(1): 33–63.

42.

McGarry

(2000) Police reform in Northern Ireland. Irish Political Studies 15(1): 173–182.

43.

Meier

(1975) Representative bureaucracy: An empirical analysis. American Political Science Review 69(2): 526–542.

44.

Meier

(1993) Latinos and representative bureaucracy testing the Thompson and Henderson hypotheses. Journal of Public Administration Research and Theory 3(4): 393–414.

45.

Meier

Nicholson-Crotty

(2006) Gender, representative bureaucracy, and law enforcement: The case of sexual assault. Public Administration Review 66(6): 850–860.

46.

Meier

Stewart

Jr (1992) The impact of representative bureaucracies: Educational systems and public policies. The American Review of Public Administration 22(3): 157–171.

47.

Miguel

Gugerty

(2005) Ethnic diversity, social sanctions, and public goods in Kenya. Journal of Public Economics 89(11): 2325–2368.

48.

Mosher

(1982) Democracy and the Public Service. Oxford: Oxford University Press.

49.

New Israel Fund (2005) Arab sector: NIF grantees fight discrimination in arab education. Available from: http://archive.li/UWPYy.

50.

Norris

(2008) Driving Democracy: Do Power-sharing Institutions Work? Cambridge: Cambridge University Press.

51.

Northern Ireland Policing Board (2016) Public Perceptions of the Police, PCSPs, and the Northern Ireland Policing Board: Report Based on the Northern Ireland Policing Board Module of the April 2016 Omnibus Survey.

52.

Ostwald

(2013) Using policy to engineer identity: How Singapore’s national service shapes ethnic and civic identities to reduce the divisive potential of ethnic diversity. Available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2369548.

53.

Perito

(2011) The Iraq Federal Police: US Police Building under Fire. US Institute of Peace.

54.

Phillips

(1980) Characteristics and typology of the journey to crime. Crime: A Spatial Perspective. New York: Columbia University Press, pp. 167–180.

55.

Raphael

Winter-Ebmer

(2001) Identifying the effect of unemployment on crime. Journal of Law and Economics 44(1): 259–283.

56.

Rengert

(2004) The journey to crime. In: Bruinsma

Elfers

de Keijser

(eds), Punishment, Places, and Perpetrators: Developments in Criminology and Criminal Justice Research. Abingdon: Routledge, chapter 11, pp. 169–181.

57.

Robinson

(2009) Tell Me How This Ends: General David Petraeus and the Search for a Way Out of Iraq. New York: PublicAffairs.

58.

Roeder

Rothchild

(2005) Sustainable Peace: Power and Democracy after Civil Wars. Ithaca, NY: Cornell University Press.

59.

Samii

(2013) Perils or promise of ethnic integration? Evidence from a hard case in Burundi. American Political Science Review 107(3): 558–573.

60.

Saunders

Popper

Morral

Davis

Berrebi

Leuschner

Efron

Segalovitz

Riley

(2013) Effective Policing for 21st-century Israel. New York: Rand Corporation.

61.

Shayo

Zussman

(2011) Judicial ingroup bias in the shadow of terrorism. The Quarterly Journal of Economics 126(3): 1447–1484.

62.

Skogan

(1986) Fear of crime and neighborhood change. Crime and Justice 8, 203–229.

63.

Skolnick

Bayley

(1988) The New Blue Line: Police Innovation in Six American Cities. New York: Simon and Schuster.

64.

Strøm

Gates

Graham

Strand

(2017) Inclusion, dispersion, and constraint: Powersharing in the world’s states, 1975–2010. British Journal of Political Science 47(1): 165–185.

65.

Taylor-Robinson

Heath

(2003) Do women legislators have different policy priorities than their male colleagues? A critical case test. Women & Politics 24(4): 77–101.

66.

Tyler

(2004) Enhancing police legitimacy. The Annals of the American Academy of Political and Social Science 593(1): 84–99.

67.

Ungar

(2011) Policing Democracy: Overcoming Obstacles to Citizen Security in Latin America. Baltimore, MD: Johns Hopkins University Press.

68.

Weitzer

Hasisi

(2008) Does ethnic composition make a difference? Citizens’ assessments of arab police officers in Israel. Policing & Society 18(4): 362–376.

69.

Weitzer

Tuch

(2004) Race and perceptions of police misconduct. Social Problems 51(3): 305–325.

70.

Weitzer

Tuch

(2006) Race and Policing in America: Conﬂict and Reform. Cambridge: Cambridge University Press.

71.

Weitzer

(1995) Policing under Fire: Ethnic Conﬂict and Police–Community Relations in Northern Ireland. Albany, NY: State University of New York Press.

72.

Yogev

(2010) Public attitudes towards the Israeli police. Israel Ministry of Public Security: Bureau of the Chief Scientist.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.37 MB

0.00 MB