CRITICAL REVIEW 2

206 Public Administration Review • March | April 2017

Public Administration Review,

Vol. 77, Iss. 2, pp. 206–216. © 2017 by

The American Society for Public Administration.

DOI: 10.1111/puar.12734.

Sergio Fernandez is associate

professor in the School of Public and

Environmental Affairs at Indiana University,

Bloomington, and visiting professor in

the Department of Public Management

and Governance at the University of

Johannesburg, South Africa. His research

focuses on organizational behavior in the

public sector, government contracting

and procurement, and representative

bureaucracy.

E-mail: [email protected]

Jill Nicholson-Crotty is associate

professor in the School of Public and

Environmental Affairs at Indiana University,

Bloomington. Her research focuses on

the management of public and nonprofit

organizations and on the impact of race,

gender, and representation on the outcomes

of public programs.

E-mail: [email protected]

Sean Nicholson-Crotty is professor

in the School of Public and Environmental

Affairs at Indiana University, Bloomington.

His research focuses on the management

of public organizations, intergovernmental

relations, and the diffusion of policy

innovations among governments.

E-mail: [email protected]

Abstract : In response to police-involved homicides of black citizens in Ferguson, Missouri, and elsewhere, some have suggested that more black police officers could reduce the number of these events. The authors offer an empirical test of this assertion. The literature offers conflicting expectations: some studies suggest that increased representation reduces discrimination, while others suggest that it increases discrimination. The authors reconcile these perspectives using the concept of critical mass, which leads to the expectation that an increase in black officers will reduce the number of black citizens killed in encounters with police, but only once the proportion of black officers is sufficiently large. We test this expectation in analyses of recently compiled data on police-involved homicides in 2014 and 2015 in large U.S. cities.

Practitioner Point • Increasing the proportion of the force that is black does not appear to be an effective strategy for reducing

police-involved homicides of black citizens in most large cities.

Sean Nicholson-Crotty Jill Nicholson-Crotty Indiana University, Bloomington

Sergio Fernandez Indiana University, Bloomington

University of Johannesburg, South Africa

Will More Black Cops Matter? Officer Race and Police-Involved Homicides of Black Citizens

On August 9, 2014, Michael Brown was shot to death by Officer Darren Wilson of the Ferguson, Missouri, Police Department. The incident touched off days of protest in Ferguson and a heated national debate about the treatment of black Americans by the police, thanks in part to a number of now well-known specifics. The most salient of these are that Officer Wilson was white; Michael Brown was black and unarmed; there was a lack of clear evidence that Mr. Brown was doing anything that would have justified his killing; videos of his body lying in the street were replayed repeatedly by media around the nation; and a grand jury failed to indict Officer Wilson for Mr. Brown ’ s death. 1

In the 12 months after Michael Brown ’ s death, 15 similar incidents of an unarmed black person being killed by police, or dying in police custody under suspicious circumstances, received national media coverage. In some of these cases the officers were indicted, in others they were exonerated, and in others investigations are still on going. Nonetheless, the steady stream of stories about black deaths at the hands of the police has further eroded already strained relationships between police and black citizens in many urban areas. It has also convinced more white Americans that this is a major policy and/ or administrative problem. 2 Most importantly, it has sparked a sustained debate among members of both races about potential solutions.

One of the most frequently proposed remedies to the problem of police-involved homicides of black citizens has been the creation of more diverse police forces or ones that are more representative of the communities they serve. Black and white community members around the country have demanded in written and public forums that city governments hire more police of color (see, e.g., Jackson 2014 ; Ouriel 2014 ). Communities across the country, from tiny Brick Township, New York, to Los Angeles, California, have announced initiatives to do so. The New York Times ran a major story suggesting that these efforts are warranted in even the largest cities, where they claim that police forces are typically 30 percent more white than the communities they patrol (Ashkenas and Park 2015 ).

More black officers are obviously seen, in part, as a way to directly reduce unnecessary violence between police and citizens. As a Ferguson resident said in the days after Michael Brown ’ s shooting, “We want answers, we want justice in our community, we want diversity” (Bouie 2014 ). Increased diversity or representation of minorities is also proposed as a way to indirectly reduce violence by enhancing the legitimacy of the police force within communities. Two months after the shooting of Michael Brown, the mayor of Ferguson announced that the city would be hiring additional black officers in an effort to make the force “more reflective of the demographics of

Will More Black Cops Matter? Offi cer Race and Police-Involved Homicides of Black Citizens 207

Ferguson” and to help appease residents “who feel there is a divide between them and law enforcement” (Giegerich 2014 ). Claims that greater police diversity or representation of minorities may increases legitimacy and reduces events like police-involved homicides are not new (see, e.g., Fyfe 1980 ), but they have certainly been reinvigorated by recent events.

The claim that more black cops would result in fewer homicides of black citizens has not been without critics. Some of these simply reject the narrative that race is an issue in the interaction between police and citizens (see, e.g., Scarry 2015 ). Others, including some black police chiefs, readily admit that there is a problem but warn that more black officers will not be a “panacea.” For example, Chief Calvin Williams of Cleveland argued that “if you come from the premise that only an African American can police other African Americans, then we ’ re all doomed to failure.” 3 Others have been more pointed in their criticism, suggesting that black officers may be more willing to abuse power because “brutality complaints can be shrugged off … because the racial aspect is neutralized” (Simon and Burns 1998 ).

Nonetheless, increased black representation on police forces has been a prominent policy solution offered in the wake of Ferguson, and it is the primary focus of this article. Beyond the obvious significance of the policy problem, we feel that such an investigation is necessary and warranted for a number of other reasons. The first is that no such analysis exists because, until very recently, no data existed that allowed systematic study of the relationship between black officers on a police force and fatal encounters with black citizens. Second, existing research on the other types of policing outcomes, including the use of lethal force when victims are not distinguished by race, does not generate clear predictions about whether more black officers should reduce police- involved homicides of black citizens.

The remainder of the article proceeds in four sections: The first reviews the literature on the impact of bureaucratic representation of minorities on interactions with the community. The second section draws on the theory of representative bureaucracy to develop expectations about the degree to which more black officers might influence police-involved homicides of black citizens in U.S. communities. We offer the expectation that greater representation of blacks on the police force may in fact reduce the number of police- involved homicides of black citizens, but only once the achievement of a critical mass of black officers makes it possible for them to actively represent or advocate for black citizens. The article uses data on police-involved homicides in 2014 and 2015, compiled by two different organizations using two different techniques, to test for that relationship in the 100 largest American cities by population. In the final section of the article, we draw conclusions about the efficacy of black representation on police forces as a solution to police-involved homicides of black citizens.

Literature on Police Force Demographics There exists a fairly extensive literature on the use of lethal force by police, without accounting for race, which can provide a

beginning for this inquiry. These studies have been conducted at the state (Jacobs and Britt 1979 ; Kania and Mackey 1977 ; Liska and Yu 1992 ), municipal (Jacobs 1998 ; Sorensen, Marquart, and Brock 1993 ), and neighborhood levels (Fyfe 1980 ; Langworthy 1986 ) using Supplementary Homicide Report data provided to the Federal Bureau of Investigation (FBI) or data provided directly by the New York City Police Department. Generally speaking, these studies are tests of the threat hypothesis, which asserts that police violence is a mechanism for controlling groups that threaten the social order, and the conflict hypothesis, which claims that police violence is a direct response to the violence that police are exposed to in the community. The results from these studies are mixed, but taken together, they provide some evidence that a large minority population, significant income inequality, and the amount of violence in a community all correlate positively with police killings of citizens.

Smith ( 2003 ) examines the impact of police force demographics on killings. This study also uses the Supplementary Homicide Data, and thus it cannot measure the race of the victim, but it does measures the racial composition of the police force. It models

the number of police killings of felons in 179 U.S. cities while controlling for wealth inequality, racial composition of the citizenry, and other factors. The results suggest that there is not a significant correlation between the racial composition of the police force and the number of reported killings of felons.

Subsequent research on the impact of police force demographics on police-involved

homicides not delineated by victim race has reached mixed conclusions. In some cases, studies have confirmed the finding that hiring more minority officers does not correlate with a reduction in the killing of citizens (see, e.g., Holmes and Smith 2008 ; Smith and Holmes 2014 ). Alternatively, some work has shown that the use of lethal force increases with the proportion of black officers (Ochs 2011 ). Finally, other studies have found that white officers are more likely to use deadly force than their black or Hispanic counterparts (see, e.g., McElvain and Kposowa 2008 ).

Studies of the impact of race on other policing outcomes, where the race of the citizen can be taken explicitly into account, have also reached inconsistent conclusions. In the case of arrest behavior, some studies suggest that black officers may treat black citizens more harshly (Brown and Frank 2006 ; Thompson 1976 ), while others find no correlation between race congruence and officer behavior (Holmes and Smith 2008 ). Similarly, in the case of racial profiling, some work finds that white officers are more likely to stop and search black motorists (Antonovics and Knight 2009 ), while other studies suggest that the opposite is true and that black motorists are more likely to be profiled when the proportion of black officers is higher (Wilkins and Williams 2008 ).

The existing literature does not offer a direct test of the relationship between black officers and police-involved homicides of black citizens, and therefore it does not speak to the potential of increased force diversity or representation of minorities to reduce incidents like those in Ferguson, North Charleston, Cleveland, Cincinnati, and so

Increased black representa- tion on police forces has been a prominent policy solution

off ered in the wake of Ferguson, and it is the primary focus of

this article.

208 Public Administration Review • March | April 2017

many other places. Additionally, as noted earlier, existing work has produced largely inconsistent findings regarding the impact of officer race or officer–citizen race congruence in other contexts.

Critical Mass and Police-Involved Homicides of Black Citizens In this article, we are interested in understanding the relationship between the percentage of a police force that is black and the incidence of police-involved homicide of black citizens. We utilize the theory of representative bureaucracy, and more specifically the concept of critical mass, in an attempt to reconcile previous work which finds that more black officers may be associated with both more and less discrimination.

The Theory of Representative Bureaucracy Very early in the development of the literature, scholars identified two types of bureaucratic representation: passive or demographic representation of a social group and active representation, which is conceived of as administrative decisions, actions, and outputs that benefit members of the same group (Mosher 1968 ). Since then, a proliferation of studies have examined the link between these two types of representation. Recent research even suggests that both passive and active representation in the bureaucracy are necessary for a healthy democracy (Riccucci and Van Ryzin 2017 ).

The relationship between passive representation and active representation is believed to be contingent on certain preconditions being met (Keiser et al. 2002 ; Meier 1993 ; Meier and Nicholson- Crotty 2006 ; Meier and Stewart 1992 ; Thompson 1976 ). First, bureaucrats should have discretion when implementing policies. As Meier and Bohte explain, “For representation to occur, bureaucrats must have a sphere of influence to take actions that reflect the specific values they hold” (2001, 457). Second, the policies that bureaucrats implement should be salient or significant to the demographic group in question. This means that the issue that bureaucrats are dealing with has clear and direct implications for them and members of their social group. Finally, the decisions and actions of bureaucrats should impact the demographic group in question. That is, the bureaucrat is in a position—including the right agency, organizational level, and job—to make decisions that influence the well-being of those belonging to his or her social group.

A growing number of studies across numerous policy domains have empirically demonstrated a link between passive racial representation and bureaucratic decisions and outputs favoring the represented racial group when the foregoing preconditions are met (e.g., Bradbury and Kellough 2011 ; Hindera 1993 ; Meier 1993 ; Meier and Stewart 1992 ; Meier, Stewart, and England 1989 ; Meier, Wrinkle, and Polinard 1999 ; Selden 1997 ). For example, Meier, Wrinkle, and Polinard ( 1999 ) find that standardized test scores of minority and nonminority students increase as school districts hire more minority teachers (see also Meier 1993 ; Meier and Stewart 1992 ). Also in the area of education, Nicholson-Crotty, Grissom,

and Nicholson-Crotty ( 2011 ) find that black teachers are more likely to refer black students to gifted and talented programs than their white counterparts (see also Grissom, Nicholson-Crotty, and Nicholson-Crotty 2009 ). In the U.S. federal procurement arena, Fernandez, Malatesta, and Smith ( 2013 ) find that increasing racial minority representation enables agencies to become more effective at achieving the legislatively mandated goal of promoting minority- owned small business participating in federal contracting. And

Selden ( 1997 ) finds that an increase in the percentage of administrators in the U.S. Department of Agriculture ’ s Farmers Home Administration who were black, Hispanic, and Asian American led to an increase in the number of loan eligibility awards favoring black, Hispanic, and Asian American loan applicants, respectively.

In the area of law enforcement, which is the primary concern of this article, the results have been decidedly more mixed. Taking a slightly different approach to

the consequences of representation, Theobald and Haider- Markel ( 2009 ) find that black citizens perceive stops to be more legitimate when the officer is black. Modeling a more traditional outcome variable, Hong ( 2017 ) finds that an increase in black officers is correlated with a decrease in complaints filed by black citizens in an analysis of policing in the United Kingdom and Wales.

However, numerous studies of representative bureaucracy in policing have failed to find a positive relationship. Sharp ( 2014 ) finds that racial representation in elected positions decreases the amount of racial bias in order maintenance arrests, but that increased representation on the police force does not and may even be positively associated with bias, although it is important to note that the positive relationship is not statistically significant. Also, as noted earlier, Wilkins and Williams ( 2008 ) fail to uncover a link between black representation among police officers and a reduction in racial disparities in vehicle stops. Indeed, they show that an increase in the percentage of police officers who are black leads to an increase in racial profiling. They explain this as a product of organizational socialization in law enforcement agencies that discourages minority police officers from advocating for the interests of minorities.

Bradbury and Kellough ’ s ( 2011 ) extensive review of the criminal justice literature through a representative bureaucracy lens confirms the existence of mixed findings. They highlight some results that are consistent with active representation, such as those of Close and Mason ( 2006 , 2007 ) and Gilliard-Matthews, Kowalski, and Lundman ( 2008 ), who find that minority officers treated minority drivers more favorably compared with white officers. However, they also point to other policing studies (e.g., Anbarci and Lee 2008 ; Brown and Frank 2006 ) in which the results show that minority officers were less lenient toward minorities and more likely to arrest them. Importantly, Bradbury and Kellough note that minority officers may be tougher on minority citizens at least in part to protect minority communities, thus making their behavior consistent with active representation.

We utilize the theory of rep- resentative bureaucracy, and

more specifi cally the concept of critical mass, in an attempt to reconcile previous work which fi nds that more black offi cers may be associated with both more and less discrimination.

Will More Black Cops Matter? Offi cer Race and Police-Involved Homicides of Black Citizens 209

Critical Mass and Active Representation The concept of critical mass offers a potential way to reconcile these inconsistent results regarding the relatiohsip between black police officers and outcomes for black citizens. Scholars have used this concept when attempting to understand the active representation of minority groups within institutions, particularly those characterized by strong norms or high levels of socialization. It suggests that individuals may only begin to act in ways that help other minorities, either in that organization or within the community that they serve, when they are empowered by a large enough number of individuals from similar backgrounds within their own organization.

Rosabeth Moss Kanter ( 1977a , 1977b ) is often credited with developing the concept of critical mass in her studies of work groups in private firms. She notes that minority members of these groups, specifically women, were largely silent until their numbers reached a certain percentage (~15 percent), at which point they became more active participants. She suggests, however, that it was not until they made up a much higher proportion of the group (35 percent to 40 percent) that they became active advocates for their preferences. Although some question the rigor of Kanter ’ s early studies and point to weaknesses, particularly questioning the specific thresholds at which one might see these changes in behavior, many still turn to her early work as a frame for their own examination of the phenomenon.

The idea of thresholds at which the behavior of minority group members changes has been an important feature of representative bureaucracy studies and studies of group behavior since at least the 1970s. Henderson ( 1979 ) and Thompson ( 1976 ) both argue that active representation by minorities within the bureaucracy will not be observed until a critical mass of individuals is employed. More recent work has demonstrated empirically that a critical mass is necessary for passive representation to translate into active representation by bureaucrats (see Keiser et al. 2002 ; Meier 1993 , Meier, Wrinkle, and Polinard 1999 ; Nicholson-Crotty, Grissom, and Nicholson-Crotty 2011 ). These effects have been shown in both education and criminal justice organizations and for both race and gender representation. They have also been shown to influence a variety of representation behaviors and outcomes, including disciplinary actions, test scores, and client behavior.

Organizational research on conformity to norms offers insights into why achieving a critical mass can lead to active representation. Organizations and groups can exert considerable pressure on members to conform to norms, especially norms that regulate behavior, safeguard the prestige and legitimacy of the organization or group, and help ensure its survival (Feldman 1984 ; Hackman 1992 ). The reasons members acquiesce to such pressures include being accepted by others, preserving and promoting one ’ s standing, and gaining access to resources and opportunities for advancement.

The pressure to conform to norms appears to be especially strong in police departments, many of which place a premium on loyalty and cohesion among police officers (Brown 1988 ; Cordner and Scarborough 2010 ). In police departments that are predominantly

staffed and led by white police officers, advocating for the interests of minority citizens may be perceived as violating prevailing norms, some of which condone, if not promote, negative images of racial minorities and behavior detrimental to their interests, including racial profiling (Wilkins and Williams 2008 , 2009 ). Achieving a critical mass of minority representation may enable minority officers to overcome pressures to conform and to advocate for minority citizens—or at least not to engage in discriminatory behavior toward them—by making such behavior appear more acceptable or less deviant, thereby shielding them from the repercussions of not conforming.

Expectations Police–citizen interactions that could result in violence are clearly an example of a case in which the bureaucratic actor has significant discretion regarding his or her actions and the history of tension between police and communities of color makes the outcome of those interactions extremely salient to the affected group. On the other hand, police organizations are extremely effective at socializing members to certain values and norms and enforcing compliance with them, factors that previous research suggests may suppress representation behavior on the part of black officers (Wilkins and Williams 2008 ). For all of these reasons, we believe that the critical mass corollary of representative bureaucracy theory can be used to generate expectations about the impact of black officers on police-involved homicides of black citizens. Specifically, our main expectation is that the relationship between the percentage of black officers and police-involved homicides of black citizens will take a quadratic form, with the relationship turning negative only after the proportion of black officers has reached a relatively high level.

Importantly, the work of Wilkins and Williams ( 2008 , 2009 ), which shows that greater minority representation on police forces actually increases racial profiling, suggests that the pressure to conform may be so strong in police departments that even beyond a critical mass, black representation in police departments never translates into active representation. That is, increasing black representation beyond a

critical mass may not lead to additional police- involved homicides of black citizens, but it will not reduce the number of such incidents. Either way, we have reason to believe that the relationship between black representation and police-involved homicides of black citizens will change once a critical mass is achieved, but the relationship may never go from positive to negative. We are not offering a specific prediction regarding the inflection point, although we note that previous work has consistently identified a critical mass of roughly

30 percent to 40 percent of a work group or organization.

Data, Variables, and Methods Dependent Variable The main dependent variable is the count of police-involved homicides of black citizens in a city. Local law enforcement agencies historically have not been compelled to report deaths in custody, and thus no federal source for these data exists. Instead, we use data gathered by advocacy groups and major news organizations. In order to ensure that our results are not an artifact of coding decisions or other choices made by one of these groups, we estimate models

We have reason to believe that the relationship between black

representation and police- involved homicides of black citizens will change once a

critical mass is achieved, but the relationship may never go from

positive to negative.

210 Public Administration Review • March | April 2017

on two counts of police-involved homicides collected by different organizations, using different methodologies, and in different years.

For the first measure, we use data gathered from Mapping Police Violence, an advocacy organization that has created a sample of police homicides in America ’ s largest cities. According to Mapping Police Violence, a police-involved homicide is a case in which an on-duty or off-duty police officer, either intentionally or accidently, kills a person in the process of chasing, beating, arresting, restraining, shooting, pepper spraying, Tasering, or otherwise harming the person. Each police homicide is coded by police department and by the gender and race of the victim.

Mapping Police Violence has developed a data set of police homicides occurring in 2014 in the 100 largest American cities by population, ranging from slightly more than 200,000 to about 8,000,000 residents. 4 Its data set was compiled using data from the three largest, most comprehensive crowdsourced databases on police killings in the country: FatalEncounters.org, the U.S. Police Shootings Database, and KilledbyPolice.net. The majority of the data come from KilledbyPolice.net. Mapping Police Violence specifically selected KilledbyPolice.net because it was fact-checked by Nate Silver ’ s FiveThirtyEight research organization. KilledByPolice.net aggregates links from major social media sites to news articles on police-related killings, and FiveThirtyEight selected a random sample of approximately 10 percent of the links from the original news stories (Fischer-Baum and Johri 2014 ). That study found that 100 percent of the links were accurate and that 93 percent of reported incidents were either incidents in which the victims were fatally shot by on-duty officers or the victims died after being restrained in some way (e.g., Taser, choke hold, etc.) by law enforcement.

Mapping Police Violence undertook extensive data gathering of its own to improve the accuracy and completeness of the data from the three crowdsourced databases on police killings in the United States. Using cases from FatalEncounters.org, the U.S. Police Shootings Database, and KilledbyPolice.net as the starting point, it used social media, newspaper accounts, obituaries, police reports, and other sources to confirm the occurrence of a death and identify the race of 91 percent of victims. In addition, it used other crowdsourced databases, media sources, and criminal records databases and performed additional inquiries to fill in additional fields in its database, such as whether the victim was armed or unarmed. 5

The data from Mapping Police Violence, and the sources from which they were gathered, represent the most complete record of police killings of citizens by race in the United States. 6 Moreover, at the time of this research, they were the most comprehensive data available for 2014, which is the closest year to 2013, when we can measure the racial composition of police forces. The Mapping Police Violence variable has a mean of 1.34, a minimum value of 0, and a maximum value of 16.

Our second measurement of the dependent variable is created from data collected by the Washington Post on police-involved homicides in 2015. 7 The Washington Post compiles data on cases in which an on-duty police officer shoots a citizen. This is a more limited definition of police-involved homicides in that, unlike

Mapping Police Violence, it does not account for cases involving off-duty police officers and homicides resulting from violence other than gunshots (e.g., beatings). The Washington Post relies on a range of sources to identify police homicides, including local news reports, its own reporting, law enforcement websites, social media, and independent databases such as FatalEncounters.org and KilledbyPolice.net (both of which are also used by Mapping Police Violence to create its database). The Washington Post conducted follow-up investigations of these cases by requesting comments from the police departments involved in the cases, filing open records requests with these departments, and soliciting information from the public.

To create this second count dependent variable, we coded all the cases in the Washington Post database that occurred in 2015 by race of the victim and by the police department for all municipalities in our original sample of the 100 largest American cities by population. The Washington Post variable has a mean of 1.18, a minimum value of 0, and a maximum value of 8. That variable and the one gathered from Mapping Police Violence correlate at .66.

Limiting the analysis to just these 100 largest cities in 2014 or 2015 obviously means that we can only safely draw inference to encounters between police and citizens in larger communities. However, police-involved homicides in smaller communities appear to be extremely rare events.

Independent Variables The key independent variables for hypothesis testing are the percentage of black officers in a department and the square of that term. Previous studies of police violence have used the ratio of force and population composition (i.e., a representation index), but representative bureaucracy theory assumes that active representation is stimulated by shared experiences between the bureaucrat and the client and not by the makeup of the bureaucracy relative to society as a whole. As a result, the percentage of black officers is a more theoretically appropriate variable for our test.

To measure our independent variable, we used 2013 Law Enforcement Management and Administration Statistics (LEMAS) published by the Bureau of Justice Statistics. The 2013 data contain information about personnel reported to the federal government by individual departments. This allows us to measure the percent of officers that are black in each municipal police department. The mean percentage of black officers in our sample of cities is 14.43, with a minimum value of 0.23 and a maximum of 62.58. 8

Control Variables Models discussed here include several control variables. The first of these captures the fact that some departments may be more likely to use lethal or excessive force against all citizens, regardless of race. Specifically, we include the number of white citizens killed by police in 2014 or 2015.

The next set of control variables reflect the threat and conflict hypotheses that have typically guided research on police violence. We include the percentage of the population below the poverty level and the percentage of the population that is black to capture the expectation that police violence is a means of social control for groups that threaten the social order. We include the violent

Will More Black Cops Matter? Offi cer Race and Police-Involved Homicides of Black Citizens 211

crime rate and the murder rate in order to account for the assertion that police violence is a direct response to the degree of conflict or violence they encounter in the environment. These data were gathered from the FBI’s Uniform Crime Report (UCR) for 2013. The violent crime rate includes murders, rapes, sexual assaults, robberies, and assaults.

Finally, both models control for the total number of sworn officers per capita within a jurisdiction, gathered from the 2013 LEMAS. They also control for the percentage of citizens within a jurisdiction that are black and the total black population. These are included because the count of black homicides is likely to be highly correlated with the likelihood that police officers come in contact with black citizens, which is likely to be a function of the density of both parties within a jurisdiction. We can estimate models with total population rather than total black population and the results remain unchanged, but we believe that the latter is a better measure because it is well documented that policing resources are dedicated more heavily to areas with high minority populations and that those populations are by no means evenly distributed across urban jurisdictions. Means and other characteristics of all variables are presented in table 1 . A correlation matrix of variables is presented in table 2 . 9

Estimation Because of the nature of the data, we use a count model. Descriptive statistics and a significant likelihood ratio test comparing the distribution with the Poisson suggest overdispersion. The data are also characterized by a significant number of zeroes, but fit statistics and assessment of predicted versus actual probabilities suggest that the negative binomial model is the best estimator. 10 In order to deal with the problems that can arise from estimating a negative binomial regression in a small sample, we report bootstrapped standard errors clustered at the state level.

Findings The results from our primary analyses predicting the total number of police-involved homicides of black citizens are presented in table 3 . The model of homicides gathered from the Mapping Police Violence data is presented in the first column, while the one modeling the Washington Post data is in the second column.

Turning first to the control variables, the number of white citizen homicides by police is positively associated with the number of black fatal encounters with police in both models. This suggests that, all else being equal, some jurisdictions just use lethal force more often than do others.

The real variables of interest are the percentage of officers within a department that are black and that percentage squared. In both models, the main effect is positive, while the quadratic term is negatively associated with police-involved homicides of black citizens. In the 2015 model the relationship appears statistically significant. The pattern is consistent with our nonlinear hypothesis regarding black representation on the police force and the number of black citizen deaths.

Because the quadratic term is an interaction, the actual shape and significance of that relationship, and thus the substantive impact of officer race on police killings of black citizens, must be determined graphically. Figure 1 plots the predicted counts from the model using 2014 Mapping Police Violence data (see table 3 , first column), and those from the model using 2015 Washington Post data are plotted in figure 2 (see table 3 , second column). In the first case (figure 1 ), the relationship is positive until the inflection point of approximately 35 percent, at which point additional black officers begin to decrease the number of black citizens killed by police. However, despite the shape of the relationship, the confidence interval is wide enough that it is impossible to say that

Table 1 Descriptive Statistics

Variable Obs Mean SD Min . Max .

Black citizens killed 2014 (from Mapping Police Violence) 100 1.34 2.379 0 16 Black citizens killed 2015 (from Washington Post ) 100 1.18 1.641 0 8 Percent sworn black offi cers (2013) 100 14.438 14.24 .23 62.589 White citizens killed (2014) 100 1.07 1.519 0 8 White citizens killed (2015) 100 .9 1.381 0 7 Sworn offi cers per capita 100 0.0135 0.007 0.005 0.051 Black population 100 128,240.9 237,377.5 1,180.455 2,011,083 Violent crime rate (per 100,000) 100 709.3323 403.5674 47.91587 2072.329 Murder rate (per 100,000) 100 9.837 9.318 0.445 45.150 Percent population below poverty 100 20.787 6.610 6.6 40.7

Percent black population 100 20.774 17.883 0.5 80.3

Table 2 Correlation Matrix

Police Involved Homicides

Black Officers

Black Officers Squared

Pol . Homicides of Whites

Sworn Officers

Violent Crime Rate

Murder Rate

Below Poverty

Proportion Black Pop .

Black Pop .

Police-involved homicides 1 Black offi cers 0.2707 1 Black offi cers squared 0.1792 0.9464 1 Pol. homicides of whites 0.2057 −0.1823 −0.1518 1 Sworn offi cers 0.1905 0.6053 0.5816 −0.1388 1 Violent crime rate 0.2392 0.6751 0.5904 −0.087 0.5098 1 Murder rate 0.227 0.7794 0.7215 −0.1242 0.4906 0.7561 1 Below poverty 0.175 0.4899 0.428 −0.05 0.3738 0.5873 0.6225 1 Proportion black pop. 0.2583 0.8894 0.7893 −0.2321 0.5303 0.6923 0.8035 0.5336 1 Black pop. 0.6977 0.3758 0.2951 0.1022 0.2164 0.2821 0.2453 0.2071 0.3855 1

212 Public Administration Review • March | April 2017

Table 3 Relationship between Percent Black Officers and Police-Involved Homicides of Black Citizens

Variables 2014 Data 2015 Data

Black offi cers 0.085 0.118** (0.055) .044

Black offi cer squared –0.0012^ –0.0013** (0.0007) (0.0005)

White citizens killed 0.313** 0.151^ (0.129) (0.095)

Sworn offi cers per capita 18.326 –1.787 (33.639) (21.896)

Violent crime rate 0.0003 0.0005 (0.0004) (0.0004)

Murder rate 0.009 0.017 (0.025) (0.022)

Percent below poverty 0.005 –0.052 (0.027) (0.037)

Percent black population 0.001 –0.012 (0.019) (0.188)

Black population 2e–06 1.05e–05 (1.56e–06) (8.3e–07)

Intercept 2.204*** –0.792 (0.604) (0.541)

Log alpha –0.481 –1.519 (4.504) (24.607)

N 100 100

Bootstrapped standard errors clustered at the state level in parentheses. ̂ p < .1 (one-tailed); * p < .1; ** p < .05; *** p < .01.

– 2

0 2

4 6

8

P re

d ic

te d H

o m

ic id

e s

o f B

la ck

C iti

ze n s,

2 0 1 4

0 20 40 60 Percent Black Officers

Figure 1 Predicted Count of Police-Involved Homicides of Black Citizens across Percent Black Officers, 2014 Data

the predicted counts are significantly higher or lower at different percentages of black officers.

This is not the case for the model using 2015 data. In figure 2 , we see that the percentage of black officers and police-involved homicides of black citizens are positively correlated until a high density, at which point the relationship turns negative. The relationship is statistically distinguishable until the percent black officers reaches 26%, which means that the relationship is positive and significant in 85 of the police departments in the largest 100 U.S. cities.11 While the association appears to turn negative at 42%, the confidence interval

around the inflection point is wide enough that we cannot be sure if the relationship truly changes. It is important to remember that the very limited number of departments above 30% black officers (12) means that the significance of the quadratic relationship is highly sensitive to specification, treatment of errors, and even the way we examine the effects.

Before moving on, it is important to remember that our finding that increasing black officers actually correlate with an increase in black deaths in many departments is not inconsistent with previous research. Generally speaking, research on behavior in organizations suggests that minorities may be particularly zealous adherents to organizational or majority group norms when they are few in number (see Chrobot-Mason and Thomas 2002 ). Research specific to policing reinforces that notion, with Wilkins and Williams ( 2008 ) finding that African American officers are more likely than their white counterparts to engage in racial profiling of black drivers. The result is inconsistent with recent research on race and policing in England and Wales (Hong 2017 ). In our view, however, differences in the salience of race and the threat of lethal encounters for police in the United States versus the United Kingdom, as well as the very distinct dependent variables modeled in the two studies, easily explain the difference in findings.

Two-Stage Residual Inclusion Model Before drawing conclusions from the results just presented, it is important to determine whether endogeneity between police- involved homicides of blacks and the percentage of black officers is biasing the findings. While the key independent variables are measured prior to our dependent variable, endogeneity may still be present. The problem arises if the number of police-involved homicides in 2014 is correlated with the number in 2013 and 2012 and those homicides influenced the hiring of black officers.

In a linear specification, the best way to deal with endogeneity is by using an instrumental variables (IV) approach, but IV is

-5 0

5 1 0

P re

d ic

te d H

o m

ic id

e s

o f B

la ck

C iti

ze n s,

2 0 1 5

0 20 40 60

Percent Black Officers

Figure 2 Predicted Count of Police-Involved Homicides of Black Citizens across Percent Black Officers, 2015 Data

Research on behavior in organi- zations suggests that minorities

may be particularly zealous adherents to organizational or majority group norms when

they are few in number.

Will More Black Cops Matter? Offi cer Race and Police-Involved Homicides of Black Citizens 213

difficult if the second-stage model is a nonlinear specification. The preferred approach in that case is a residual inclusion (2RI) model. The approach regresses the endogenous independent variable on an exogenous instrument and all other predictors. The residuals from that equation are then included as a predictor in the model in which endogeneity is suspected. The method has been shown to produce consistent and unbiased estimates in nonlinear models in the presence of endogeneity (Terza, Basu, and Rathouz 2008 ). In our case, this means that the method should address any bias arising from endogeneity between police-involved homicides of black citizens and the percentage of black police officers on a force.

Like IV, 2RI still requires the identification of an exogenous instrument that predicts the percentage of black officers, but is not (and preferably cannot be) influenced by the number of black citizens killed by police. We use the percentage of black firefighters within a jurisdiction as the instrument. That measure accurately proxies the potential labor pool from which the police department might recruit black officers, as well as any municipal policies that might increase the likelihood of a more diverse force, but it cannot influence the number of police homicides of black citizens. The F -statistic for the model predicting the percentage of black police officers with the percentage of black firefighters is 50.28, well above the cutoff of 10 used to identify suitable instruments.

The residuals from this first-stage model, presented in table 4 , first column, are then included in a model that is otherwise identical to our first primary models. The results of this second stage of the 2RI are presented in table 4 , second and third columns. The residual term is not significant, suggesting that endogeneity is not biasing the main findings.

Conclusion The hiring of more black police officers has been a frequently offered policy solution in the face of increased public awareness and concern about the death of black citizens in encounters with police. Some of the literature on policing actually suggests that black police officers may be more likely to discriminate against black citizens because of increased pressure to adopt an organizational role that prescribes such behavior. Other work has suggested that there is no relationship between officer race and racially disproportion in policing outcomes. Finally, representative bureaucracy theory suggests that there may be a relationship between force composition and decreased police violence, but only once there are enough black officers that they feel safe representing the interests of clients of the same race.

Our analyses provide some support for all of these expectations. We see no significant relationship in the 2014 data. In 2015 we see a positive and significant relationship between the percentage of black officers and fatal encounters between police and black citizens in the vast majority of the largest U.S. cities. The results also suggest that there is an inflection point at which black officers may become less likely to discriminate against black citizens and more inclined to assume a minority advocacy role or to become neutral enforcers of the law, though the limited number of observations makes it impossible to know for sure. Nonetheless, the potential quadratic shape of the relationship could have significant implications for work in representative bureaucracy. It suggests that work looking for a simple linear relationship between minority bureaucratic representation and outcomes for minority citizens, particularly in organizations with strong socialization norms, may miss the role played by critical mass in changing the behavior of bureaucrats.

Table 4 Residual Inclusion Model of Relationship between Percent Black Officers and Police-Involved Homicides

Variables Coef. (s.e.) Variables 2014 Data 2015 Data

Percent black fi refi ghters 0.504*** Percent black offi cers 0.067 0.119^ (0.122) (0.088) (0.076)

White citizens killed 0.196 Percent black offi cers squared 0.0012* −0.0013** (0.371) (0.0006) (0.0005)

Sworn offi cers per capita 339.1** Residual 0.023 –0.002 (146.317) (0.069) (0.055)

Violent crime rate 0.143 White citizens killed 0.321** 0.15 (0.248) (0.123) (0.093)

Murder rate 0.339** Sworn offi cers per capita 26.007 –2.249 (0.147) (37.767) 44.870)

Percent population below poverty −0.0378 Violent crime rate 0.0003 0.001 (0.127) (0.0004) (0.001)

Black population 1.8e-06 Murder rate 0.015 0.017 (1.7e-06) (0.026) (0.027) 0.313*** Percent population below poverty 0.001 –0.051

Percent black population (0.0997) (0.024) (0.03) Intercept −4.421** Percent black population 0.011 –0.012

(2.080) (0.038) (0.034) N = 100 Black population 2.9e–06 1.04e–06 F = 50.27, p < .0000 (1.95e–06) (8.4e–07) R 2 = .85 Intercept –2.248*** –0.789 Standard errors in parentheses. (0.659) (0.576) * p < .1; ** p < .05; *** p < .01. Log alpha –4.793 –1.518

(6.433) (7.615) N = 100 100 100

Bootstrapped Standard errors clustered at the state level in parentheses. * p < .1; ** p < .05; *** p < .01.

214 Public Administration Review • March | April 2017

Returning to our initial question regarding policy recommendations in the wake of Ferguson and other high-profile police-involved homicides of black citizens, our results suggest that, for the vast majority of cities, simply increasing the percentage of black officers is not an effective policy solution. The analysis of 2015 Washington Post data suggests that in most of the largest cities, increasing minority officers may actually be associated with more of these incidents until, potentially, the percentage of black officers grows quite large. The analysis of 2014 Mapping Police Violence data shows no significant relationship between black officers and fatal encounters with black citizens.

Taken together, at best the results suggest that an increase in African American officers is only a solution for the small handful of large departments around the country that already have a high percentage of black officers. In places like Ferguson, where blacks are only 11 percent of the police force, efforts to double or even triple the share of the police force that is black would not reach this threshold and may actually be associated with an increase in violent interactions between police and black citizens until the critical mass is achieved.

A final challenge raised by our findings is that, in most cities, the point at which the association between black officers and fatal encounters becomes statistically insignificant or turns from positive to negative can only be achieved by overrepresenting blacks and making bureaucracy even less representative of the community it serves. There may be reasons to pursue such overrepresentation in policing, but our results suggest that reducing police-involved homicides of black citizens should not be one of them. This conundrum suggests that understanding the benefits and costs of representation in light of the multiple goals that police organizations pursue, and the appropriate balance to be struck among these, is an important topic for future research. Reading recent research on race and policing, one could arrive at varying and at times inconsistent conclusions, including that increasing the share of black police officers will result in fewer complaints against police by black citizens; that it will increase racial profiling in traffic stops; that it will increase the perceived legitimacy of those stops; and that it will lead to more police-involved homicides of black citizens. Future research should be conducted in a way that helps police organizations understand and predict the costs, benefits, and trade-offs from increased representation of minority groups within their jurisdictions.

Notes 1 . Despite those circumstances, it remains somewhat of a mystery as to why this

particular event was so incendiary. The NAACP Legal Defense Fund suggests that, since 2009, 45 unarmed black Americans have been killed by white police officers. In the vast majority of these cases, the officer was not charged with a crime. The number of these deaths does appear to have accelerated in 2013 and 2014, and some of them, such as the strangling death of Eric Garner, received national media coverage. These events, along with advocacy efforts such as #BlackLivesMatter started after the death of Trayvon Martin, and the increasing ease with which everyday citizens can video police–citizen interactions, may help explain the national attention that has become focused on police-caused homicides in the wake of Ferguson.

2 . In an ABC/ Washington Post poll taken a year after the killing of Michael Brown, a majority of white Americans agreed that racism is a “big problem” in this country and that the country “needs to continue making changes to give blacks equal rights with whites.” Similar questions asked in the years before Ferguson

had consistently solicited positive responses from less than 40 percent of white respondents (see Lauter and Pearce 2015 ). 3 . Statement made during an interview on 60 Minutes, CBS, January 23, 2015, http://www.cbsnews. com/news/more-black-police-not-solution-says- clevelands-top-cop/ (accessed January 9, 2017). 4 . Data are available at http:// mappingpoliceviolence.org/cities/ .

5 . Interview by e-mail with one of the creators of the database on June 7, 2016. 6 . Recent work by the Bureau of Justice Statistics and the Human Rights Data

Analysis Group (not broken down by race) estimates that the total number of police shootings in the United States in 2014 was between 1,200 and 1,300. This suggests that the Mapping Police Violence data capture between 90 percent and 98 percent of incidents.

7 . Data are available at https://www.washingtonpost.com/graphics/national/ police-shootings/ .

8 . Five departments have greater than 55 percent black officers. In order to ensure that these are not driving the results, we conduct the primary analysis without them. Our results remain substantively unchanged, although we do see the inflection point move back to 38 percent and a larger negative effect on homicides above that threshold in the 2015 data when we remove these observations.

9 . Not surprisingly in a model that includes black officers, black officers squared and measures of black population and population density, some of the correlations are high. In order to ensure that these relationships were not biasing the estimates, we conducted stepwise regressions, adding one control variable at a time to our base model. These analyses show that the inclusion of control variables does not change the results for our independent variables of interest, black officers and black officers squared. Indeed, in the absence of controls that are endogenous to the IVs, changes in predicted police involved homicides in 2014 become significant across the range of black police officers. These models are available from the authors.

10 . The maximum difference in predicted versus actual counts is –.05 for the negative binomial regression (NBRM) and .06 for the zero-inflated negative binomial (ZINB). Thus, prediction accuracy very slightly favors the NBRM. Fit statistics, including the AIC (−.404) and BIC (−8.21) confirm this, strongly favoring the NBRM over the ZINB.

11 . All significance tests are at the .1 level.

References Anbarci , Nejat , and Jungmin Lee . 2008 . Speed Discounting and Racial Disparities:

Evidence from Speeding Tickets in Boston. Discussion Paper no. 3903, Institute for the Study of Labor.

Antonovics , Kate , and Brian G. Knight . 2009 . A New Look at Racial Profiling: Evidence from the Boston Police Department . Review of Economics and Statistics 91 ( 1 ): 163 – 77 .

Ashkenas , Jeremy , and Haeyoun Park . 2015 . The Race Gap in America ’ s Police Departments . New York Times, updated April 8 , 2015 . http://www.nytimes. com/interactive/2014/09/03/us/the-race-gap-in-americas-police-departments. html?_r=1 [accessed January 9, 2017].

Bouie , Jamelle . 2014 . Black and Blue: Why More Diverse Police Departments Won ’ t Put an End to Police Misconduct. Slate, October 13. http://www.slate.com/ articles/news_and_politics/politics/2014/10/diversity_won_t_solve_police_ misconduct_black_cops_don_t_reduce_violence.html

Our results suggest that, for the vast majority of cities, simply increasing the percentage of

black offi cers is not an eff ective policy solution.

Will More Black Cops Matter? Offi cer Race and Police-Involved Homicides of Black Citizens 215

Bradbury , Mark , and J. Edward Kellough . 2011 . Representative Bureaucracy: Assessing the Evidence on Active Representation . American Review of Public Administration 41 ( 2 ): 157 – 67 .

Brown , Michael K. 1988 . Working the Street: Police Discretion and the Dilemmas of Reform . New York : Russell Sage Foundation .

Brown , Robert A. , and James Frank . 2006 . Race and Officer Decision Making: Examining Differences in Arrest Outcomes between Black and White Officers . Justice Quarterly 23 ( 1 ): 96 – 126 .

Chrobot-Mason , Donna , and Kecia Thomas . 2002 . Minority Employees in Majority Organizations: The Intersection of Individual and Organizational Racial Identity in the Workplace . Human Resource Development Review 1 ( 3 ): 323 – 44 .

Close , Billy R. , and Patrick L. Mason . 2006 . After the Traffic Stops: Officer Characteristics and Enforcement Actions . Topics in Economic Analysis and Policy 6 ( 1 ): 1 – 41 .

_____. 2007 . Searching for Efficient Enforcement: Officer Characteristics and Racially Biased Policing . Unpublished manuscript .

Cordner , Gary W. , and Kathryn E. Scarborough . 2010 . Police Administration . New York : Routledge .

Feldman , Daniel C. 1984 . The Development and Enforcement of Group Norms . Academy of Management Review 9 ( 1 ): 47 – 53 .

Fernandez , Sergio , Deanna Malatesta , and Craig R. Smith . 2013 . Race, Gender, and Government Contracting: Different Explanations or New Prospects for Theory? Public Administration Review 73 ( 1 ): 109 – 20 .

Fischer-Baum , Rueben , and Al Johri . 2014 . Another (Much Higher) Count of Homicides by Police . FiveThirtyEight, August 25 , 2014 . http://fivethirtyeight. com/datalab/another-much-higher-count-of-police-homicides/ [accessed January 9, 2017] .

Fyfe , James J. 1980 . Geographic Correlates of Police Shootings: A Microanalysis . Journal of Research in Crime and Delinquency 17 ( 1 ): 101 – 13 .

Giegerich , Steve . 2014 . Ferguson Mayor Announces Plan to Increase Minority Hiring ; No Severance for Wilson. St. Louis Post-Dispatch , November 30. http:// www.stltoday.com/news/local/crime-and-courts/ferguson-mayor-announces- plan-to-increase-minority-hiring-no-severance/article_7c27e9b4-fdb5-5f8d- 9ee9-6d571d82fdfb.html [accessed January 9, 2017].

Gilliard-Matthews , Stacia , Brian R. Kowalski , and Richard J. Lundman . 2008 . Office Race and Citizen-Reported Traffic Ticket Decisions by Police in 1999 and 2002 . Police Quarterly 11 ( 2 ): 202 – 19 .

Grissom , Jason A. , Jill Nicholson-Crotty , and Sean Nicholson-Crotty . 2009 . Race, Region, and Representative Bureaucracy . Public Administration Review 69 ( 5 ): 911 – 19 .

Hackman , J. Richard . 1992 . Group Influences on Individuals in Organizations . In Handbook of Industrial and Organizational Psychology , 2nd ed. , Vol 3 , edited by Marvin D. Dunnette , and Leaetta M. Hough , 199 – 267 . Palo Alto, CA : Consulting Psychologists Press .

Henderson , Lenneal J. 1979 . Administrative Advocacy: Black Administrators in Urban Bureaucracy . Palo Alto, CA : R&E Research Associates .

Hindera , John J. 1993 . Representative Bureaucracy: Imprimis Evidence of Active Representation in the EEOC District Office . Social Science Quarterly 74 ( 2 ): 95 – 108 .

Holmes , Malcolm D. , and Brad W. Smith . 2008 . Race and Police Brutality: Roots of an Urban Dilemma . Albany : State University of New York Press .

Hong , Sounman . 2017 . Does Increasing Ethnic Representativeness Reduce Police Misconduct? Public Administration Review 77 ( 2 ): 195 – 205 .

Jackson , Gavin . 2014 . Diversity Discussion Dominates Florence City Council Meeting. SCNow.com, September 8. http://www.scnow.com/news/local/article_832eaad4- 37b1-11e4-8d2f-0017a43b2370.html [accessed January 9, 2017].

Jacobs , David . 1998 . The Determinants of Deadly Force: A Structural Analysis of Police Violence . American Journal of Sociology 103 ( 4 ): 837 – 62 .

Jacobs , David , and David Britt . 1979 . Inequality and Police Use of Deadly Force: An Empirical Assessment of a Conflict Hypothesis . Social Problems 26 ( 3 ): 403 – 12 .

Kania , Richard R. E. , and Wade C. Mackey . 1977 . Police Violence as a Function of Community Characteristics . Criminology 15 ( 1 ): 27 – 48 .

Kanter , Rosabeth Moss . 1977a . Men and Women of the Corporation . New York : Basic Books .

_____ . 1977b . Some Effects of Proportions on Group Life . American Journal of Sociology 82 ( 5 ): 965 – 90 .

Keiser , Lael R. , Vicky M. Wilkins , Kenneth J. Meier , and Catherine A. Holland . 2002 . Lipstick and Logarithms: Gender, Institutional Context, and Representative Bureaucracy . American Political Science Review 96 ( 3 ): 553 – 64 .

Langworthy , Robert H. 1986 . The Structure of Police Organizations . New York : Praeger .

Lauter , David , and Matt Pearce . 2015 . After a Year of High-Profile Killings by Police, Americans’ Views on Race Have Shifted. Los Angeles Times, August 5. http:// www.latimes.com/nation/la-na-race-poll-20150805-story.html#page=1 [accessed January 9, 2017].

Liska , Allen E. , and Yu. Jiang . 1992 . Specifying and Testing the Threat Hypothesis: Police Use of Deadly Force . In Social Threat and Social Control , edited by Allen E. Liska , 53 – 68 . Albany : State University of New York Press .

McElvain , James P. , and Augustine J. Kposowa . 2008 . Police Officer Characteristics and the Likelihood of Using Deadly Force . Criminal Justice and Behavior 35 ( 4 ): 505 – 21 .

Meier , Kenneth J. 1993 . Representative Bureaucracy: A Theoretical and Empirical Exposition . In Research in Public Administration , Vol 2 , edited by James L. Perry , 1 – 35 . Greenwich, CT : JAI Press .

Meier , Kenneth J. , and John Bohte . 2001 . Structure and Discretion: Missing Links in Representative Bureaucracy . Journal of Public Administration Research and Theory 11 ( 3 ): 455 – 70 .

Meier , Kenneth J. , and Jill Nicholson-Crotty . 2006 . Gender, Representative Bureaucracy, and Law Enforcement: The Case of Sexual Assault . Public Administration Review 66 ( 4 ): 850 – 60 .

Meier , Kenneth J. , and Joseph Stewart . 1992 . The Impact of Representative Bureaucracies: Educational Systems and Public Policies . American Review of Public Administration 22 ( 3 ): 157 – 71 .

Meier , Kenneth J. , Joseph Stewart , Jr., and Robert E . England . 1989 . Race, Class, and Education: The Politics of Second-Generation Discrimination . Madison : University of Wisconsin Press .

Meier , Kenneth J. , Robert D. Wrinkle , and J. L. Polinard . 1999 . Representative Bureaucracy and Distributional Equity: Addressing the Hard Question . Journal of Politics 61 ( 4 ): 1025 – 39 .

Mosher , Frederick C. 1968 . Democracy and the Public Service . New York : Oxford University Press .

Nicholson-Crotty , Jill , Jason A. Grissom , and Sean Nicholson-Crotty . 2011 . Bureaucratic Representation, Distributional Equity, and Democratic Values in the Administration of Public Programs . Journal of Politics 73 ( 2 ): 582 – 96 .

Ochs , Holona Leanne . 2011 . The Politics of Inclusion: Black Political Incorporation and the Use of Lethal Force . Journal of Ethnicity in Criminal Justice 9 ( 3 ): 238 – 65 .

Ouriel , Andy . 2014 . Blacks Call for More Diverse Police Force . Sandusky (OH) Register ( October 8 ).

Riccucci , Norma M. , and Gregg G. Van Ryzin . 2017 . Representative Bureaucracy: A Lever to Enhance Social Equity, Coproduction, and Democracy . Public Administration Review 77 ( 1 ): 21 – 30 .

Scarry , Eddie . 2015 . N.Y. Times’ Charles Blow Said Nothing about Cop Who Arrested His Son Being Black. Washington Examiner, January 27. http:// www.washingtonexaminer.com/n.y.-times-charles-blow-said-nothing-about- cop-who-arrested-his-son-being-black/article/2559374 [accessed January 9, 2017].

Selden , Sally Coleman . 1997 . The Promise of Representative Bureaucracy: Diversity and Responsiveness in a Government Agency . Armonk, NY : ME Sharpe .

216 Public Administration Review • March | April 2017

Sharp , Elaine B. 2014 . Minority Representation and Order Maintenance Policing: Toward a Contingent View . Social Science Quarterly 95 ( 4 ): 1155 – 71 .

Simon , David , and Edward Burns . 1998 . The Corner: A Year in the Life of an Inner- City Neighborhood . New York : Broadway Books .

Smith , Brad W. 2003 . The Impact of Police Officer Diversity on Police-Caused Homicides . Policy Studies Journal 31 ( 2 ): 147 – 62 .

Smith , Brad W. , and Malcolm D. Holmes . 2014 . Police Use of Excessive Force in Minority Communities: A Test of the Minority Threat, Place, and Community Accountability Hypotheses . Social Problems 61 ( 1 ): 83 – 104 .

Sorensen , Jonathan R. , James W. Marquart , and Deon E. Brock . 1993 . Factors Related to Killings of Felons by Police Officers: A Test of the Community Violence and Conflict Hypotheses . Justice Quarterly 10 ( 3 ): 417 – 40 .

Terza , Joseph V. , Anirban Basu , and Paul J. Rathouz . 2008 . Two-Stage Residual Inclusion Estimation: Addressing Endogeneity in Health Econometric Modeling . Journal of Health Economics 27 ( 3 ): 531 – 43 .

Theobald , Nick A. , and Donald P. Haider-Markel . 2009 . Race, Bureaucracy, and Symbolic Representation: Interactions between Citizens and Police . Journal of Public Administration Research and Theory 19 ( 2 ): 409 – 26 .

Thompson , Frank J. 1976 . Minority Groups in Public Bureaucracies: Are Passive and Active Representation Linked? Administration & Society 8 ( 2 ): 201 – 26 .

Wilkins , Vicky M. , and Brian N. Williams . 2008 . Black or Blue: Racial Profiling and Representative Bureaucracy . Public Administration Review 68 ( 4 ): 654 – 64 .

_____. 2009 . Representing Blue: Representative Bureaucracy and Racial Profiling in the Latino Community . Administration & Society 40 ( 8 ): 775 – 98 .

If you are interested in submitting a manuscript to

Public Administration Review,

please review the guidelines under submission process at publicadministrationreview.org

and follow the submission instructions.