Towards an understanding of modern policing norms: social identity, organization identity, and efficient policing

Transcription

1 Florida State University From the SelectedWorks of Patrick L. Mason Winter February 17, 2014 Towards an understanding of modern policing norms: social identity, organization identity, and efficient policing Billy R. Close, Florida State University Patrick Leon Mason, Florida State University Available at:

2 Towards an understanding of modern policing norms: social identity, organization identity, and efficient policing Billy R. Close College of Criminology & Criminal Justice Florida State University Patrick L. Mason Department of Economics & African American Studies Program Florida State University February 17, 2014

3 Abstract This study examines the relationship between bureaucratic identity, social identity and policing outcomes. We utilize alternative outcomes tests to examine traffic stop data collected by the Florida Highway Patrol during This study finds that representation of African American and Hispanic troopers improves outcomes for all groups of drivers, by increasing efficiency in searches. All troopers, regardless of race, engage in fewer searches when they are assigned to racially diverse or minority troops. Importantly, we show that this decrease in search activity simultaneously yields higher hit rates, thereby increasing efficiency. Finally, the data reveal that the greatest change in search behavior is by majority officers within minority troops. This suggests that majority officers modify their behavior to align with the search activities of the minority troop they serve in. This increase in efficiency not only reduces unnecessary and often discriminatory searches faced by minority drivers, but it also increases public safety.

4 Modern policing is frequently confronted with the complicated challenge of continuously seeking innovative practices, policies and procedures to increase the overall production of public safety among increasingly diverse populations. The ability to meet lofty public safety goals involves the continuous development and evaluation of a complex array of both technical and social processes. The technical elements of policing consist of finding the optimal combination of officers, capital equipment (computers, cars, telephones, etc.), training, and deployment. As a social process, policing often focuses on managerial strategy, prioritization of accountability measures, interactions between officers and citizens, and interactions among officers. Successful implementation of both the technical and social processes involved in policing should maximize public safety and convince alternative social groups that the policing process yields equal treatment. Despite the lofty goals of modern policing, racial and ethnic minorities frequently claim that the policing process is racially biased; thereby, group advocates sometime seek to alter the social process of policing by demanding an increase in the racial and ethnic representation of police officers. The assumption is that racial and ethnic minority officers will more actively represent racially and ethnic minority concerns by constructively engaging the issue of racially biased policing without harming policing efficiency. Such active representation by racial and ethnic minority police can change the social process of policing; hence, representation can also change the provision of public safety. This paper explores the following empirical questions: 1) is racial and ethnic minority representation actively expressed as a change in police organizational norms; and, 2) does African American and Hispanic active representation improve policing efficiency and, thereby, the provision of public safety? Without efficiency gains, it may be the case that representation 1

5 creates tradeoffs; some social groups are made worse off as the policing bureaucracy becomes more representative. We investigate how representation can promote equality of treatment while also increasing efficiency in the provision of public safety. 1 Among other factors, the decisions of institutional agents are shaped by their professional identity, their bureaucratic identity, and their social identity (Stewart, 2009). Specifically for bureaucrats, institutional structures help to determine the purpose and scope of their work and how much discretion they have to carry out their duties. In addition, institutions shape bureaucrats cognition by conferring identity that is, by selecting the factors that are to be considered relevant in the decision-making process and performance of professional duties and by not sanctioning behavior that is discordant with bureaucratic norms. This study focuses on the interaction between police officer s racial identity and identification with the policing bureaucracy. Specifically, this study examines the search outcomes associated with African American, Hispanic, and white police officers during traffic stops. We use data collected by the Florida Highway Patrol (FHP) to explore the links between organizational policing norms (bureaucratic identity), officer characteristics (race and ethnicity), and the efficiency of driver searches by the race and ethnicity of drivers. Law enforcement organizations rely heavily on organization socialization to modify the behavior and attitudes of their employees to be consistent with organizational goals and missions. This socialization often creates identities tied to both the profession of law enforcement and the particular department in which the police officer works. Given this, officers have both social group and institutional identities that guide their work, making this the perfect setting to test the influence of social identities (race and ethnicity), as well as institutional identities (profession and organization) on the provision of active representation. The statistical results of this paper suggest that race and 2

6 ethnicity influence an officer s decision to search, however this behavior is moderated by the policing norms of an officer s troop. This study controls for professional identity by focusing on a single (but very large state) police agency; namely, the FHP. In addition to common recruitment training, the FHP seeks to develop a common professional identity via its recruitment selection process. Recruits are US citizens, at least 19 years of age, possess a valid driver s license, must be willing to accept an initial assignment anywhere in the state, and possess at least one of the following: one year of law enforcement experience, either sworn or non-sworn; or, two years public contact experience; or, two years active duty continuous US military service with a honorable discharge; or, completed 30 semester hours (45 quarter hours) of college credit from an accredited college or university. The recruit selection process may take 4 to 8 months to complete and, among other things, includes a written examination, a polygraph examination, a psychological examination, and a background examination. Professional training establishes the desired norms of all police officers within the agency. New recruits to the FHP who are without law enforcement experience and/or Florida Law Enforcement Certification must complete a 28-week training academy and pass the Criminal Justice Standards and Training Commission Law Enforcement Exam. (See Florida Department of Highway Safety and Motor Vehicles, 2013). Florida certified law enforcement officers recruited to the FHP must undergo an 8-week transitional academy. Out-of-state law enforcement officers must obtain Florida Law Enforcement Certification prior to entering the Transitional Academy. For all recruits, the training is physically demanding and regimented in a para-military format. Each recruit receives common study materials and must reside at the academy during training, though they may leave on weekends. All recruits are trained at a single 3

7 site, the Pat Thomas Law Enforcement Academy, Havana, Florida. Since the FHP provides selective recruitment and common professional training for all officers in the state, variations in policing strategy to increase public safety are derived from the social identity of officers, bureaucratic norms of the troop (organizational identity), and individual characteristics of officers. Law enforcement and other legal institutions have competing accountability measures (Ward and Kupchik, 2009). Common professional training bestows importance on each of these accountability measures, but the prioritization of accountability measures can differ according to organizational norms and social identities of officers. Changes in the racial and ethnic composition of the troop induce changes in organizational norms. Hence, whether and to what extent own-group driver-officer identity matches have a probability of yielding different outcomes than other-group driver-identity matches may depend on whether the officer is supervised within a majority troop or a minority troop. The search norms of a substantively minority troop, that is, a troop where racial and ethnic minorities are a large fraction of the law enforcement officers, may systematically differ from the search norms of substantively majority troop, that is, a troop where race and ethnic minorities are too few in representation to have a discrete effect on changing policing norms. Watkins-Hayes (2011) has emphasized that racially representative bureaucracy theory predicts that the demographic composition of a police troop directly and indirect shapes the policing process due to the greater or lesser presence of racial and ethnic minorities within policing agencies or because police are making discretionary decisions informed by their social identity status, that is, membership in a particular racial or ethnic group, and these decisions provide more favorable outcomes for own-group citizens. Similarly, drawing on the economics of racial identity, scholars have demonstrated that the racial identity match of drivers and officers 4

8 has a substantive effect on the probability a stopped driver will be searched and the nature and extent of police enforcement actions (Close and Mason 2006, 2007; Darity, Mason, and Stewart, 2006). Hence, both racially representative bureaucracy theory and the economics of identity predict that the identity match of drivers and officers may produce differential patterns of outcomes during a traffic search according to whether the traffic search involves own-group racial and ethnic matches or other-group driver-officer racial identity matches. Variance in traffic searches according to the racial and ethnic match of the drivers and officers invites the suspicion that the police search process is inefficient, that is, it may be possible to increase public safety by altering search decisions rather than by purchasing new policing inputs into the production process. However, street-level bureaucracy theory reminds us that institutions and professions may constrain the discretionary decision-making of organizational actors. Professional identity is a barrier to entry. Specifically, one becomes a police officer via a training process that does not vary according to the racial or ethnic identity of the individual being trained. All police are professionally trained to be blue, regardless of a social identity as black, brown (Hispanic), white, or other. Organizational identity is imposed on organizational actors through rules, procedures, institutional oversight, and cultural norms (Stewart, 2009). Professional and organizational identity limit variation in policing outcomes that is associated with differences in the social identity matches of police and drivers. Police searches and, hence, the efficient provision of public safety are governed by an individual law enforcement officer s use of bureaucratic discretion in combination with the possibly contending social norms of social identity, professional identity, and organizational identity. Darity, Mason, and Stewart (2006) show that agents construct racialized social identity 5

9 norms through repeated social interactions governed by a process of own-group altruism and other-group antagonism. Whether or not an agent becomes an individualist, that is, able to engage in social interactions without conforming to a racial identity and without being seen by others as a member of the same or a different racial group, he is shaped by the fraction of other agents who are either individualists or persons with a racialized identity. In a society governed by strong and persistent racial identity norms, police are more likely to be racialized police than individualist police. Watkins-Hayes (2009) suggests that there is a process of racialized professionalism among street-level bureaucrats, viz., law enforcement officers, such that bureaucrats (in our case, police) seek to integrate their racial identity into their understanding and operationalization of their work and their goals for what it should accomplish (Watkins- Hayes, 2011:237). There are multiple avenues for racialized professionalism to affect organizational processes. Through formal and informal interactions among colleagues, troops will develop differential standards for which drivers to stop, and which factors constitute suspicious behavior sufficient to request or impose a search of a stopped driver. Using traffic stop data collected by the Florida Highway Patrol during , this study expands our understanding of the relationship between bureaucratic representation and organizational performance by empirically investigating this relationship in the context of policing. Specifically, this study examines the decision to search a vehicle and the efficiency of searches. Previous models have failed to sufficiently capture the strategic interactions among the race of the driver, race of the officer, and the racial composition of the officer s work unit. Hence, we extend the traffic search game offered by and Close and Mason (2006, 2007) to capture variation in policing strategy related to the racial composition of the police troop. We contend that if representation improves efficiency, then more representative police units will 6

10 have higher hit rates (i.e. the proportion of searches yielding evidence of illegal activity or contraband will increase). We find that representation of African American and Hispanic troopers improves outcomes for all groups of drivers, by increasing efficiency in searches. More representative troops strike the proper balance between the level of searches the troopers undertake and the hits they achieve. The strongest effect seems to be on the behavior of white officers serving in units with greater numbers of minority troopers; members of the dominant group appear to adjust their searching patterns to align with the norms of the minority troopers they serve alongside. This increase in efficiency not only reduces unnecessary and often discriminatory searches faced by minority drivers, but it also increases public safety. I. The decision to search Effective law enforcement requires discretionary decision-making by police. However, police discretion may lead some officers to indulge prejudicial biases or the belief that members of minority racial groups represent a social threat (Blalock, 1967). Most non-traffic illegal activities do not receive law enforcement attention until the victims or witnesses of the crimes notify law enforcement. During traffic enforcement however, officers use their own discretion to determine if a driver has committed an infraction, if the infraction is sufficiently serious to warrant intervention, and finally the appropriate nature of the law enforcement remedy. Accordingly, even if executive policy and managerial oversight does not encourage or condone discriminatory behavior, officer biases, stereotypes, and prejudices combined with the high level of officer discretion required for effective law enforcement may produce biased policing. Scholars from multiple academic disciplines have extensively examined whether the race or ethnicity of an officer influences their behavior on the job, especially during a traffic stop. For example, Antonovics and Knight (2004) find that Boston police are more likely to search a 7

11 stopped driver if the race of the officer differs from the race of the driver. Similarly, Donohue and Levitt (2001) show that increases in the minority composition of a city s police force increases arrests of whites, but has little impact on arrests of nonwhites. They also find that increases in the presences of white officers lead to increases in the number of arrests of nonwhites, but has no effect on the number of white arrests. Close and Mason (2007) provide evidence that white officers police differently than their African American and Hispanic colleagues. White officers are 73 percent of FHP s sworn police force, conduct 88 percent of the searches, and have a hit rate of 20 percent. Hispanic officers are 11 percent of the sworn labor force, conduct 8 percent of the searches, and have a hit rate of 24 percent. African American officers are 15 percent of the sworn labor force, conduct 4 percent of the searches and have a hit rate of 26 percent. White female driver status reduces the probability that any evidence will cause police to be sufficiently suspicious to initiate a consent search request. African American male driver status increases the probability that any sort of evidence will cause police to be sufficiently suspicious to initiate a consent search request. The probability that an African American male driver is searched is 0.43 percent when the stopping officer is African American, but it is 2.09 and 1.34 percent when the stopping officers are white male and Hispanic, respectively. The probability that a white driver is searched is 1.10 percent when the stopping officer is white, but it is 0.29 and 0.65 percent when the stopping officers are African American and Hispanic, respectively. The probability that a Hispanic driver is searched is 0.82 percent when the stopping officer is Hispanic, but it is 0.30 and 2.14 percent when the stopping officers are African American and white, respectively. Relative to white officers, African American officers reduce the relative odds that African American, Hispanic, and white drivers are searched by 82 percent, 82 percent, and 77 8

12 percent, respectively. Hispanic officers reduce the relative odds that African American, Hispanic, and white drivers are searched by 58 percent, 64 percent, and 60 percent, respectively. White drivers are recipients of neutral or favorable treatment by African American, Hispanic, and white officers, but this preferential treatment is attenuated with increases in the fraction of racial and ethnic minority residents in the county where the stopped occurred. This study assumes that police executives seek to optimize an organizational objective, for example, to maximize arrests or to minimize crime. Nevertheless, line officers use discretion in determining the appropriateness of driver stops and searches. African American, Hispanic, and white officers differ in their utilization of discretionary authority; thereby, the enforcement norms of a given police unit may vary according to the racial composition of the troop. There is a discrete change in policing norms when the racial minority composition of the police unit, that is, the troop, rises above a certain level of the force. As such, there will be a distinct change in the probability of driver searches when the percent of racial minority officers rises above or falls below a particular percentage of the police force. In other words, officers will adjust to the policing norms of their administrative units. This study uses 33 percent as the point of delineation of minority and majority troops. This break point is also roughly equal to the proportion of African American and Hispanic officers within the Florida Highway Patrol and the fraction of African American and Hispanic citizens within the state of Florida. We posit that majority troops (those with 67 percent or greater white officers) do not have a critical mass of racial minority officers. Minority troops have a critical mass of African American and Hispanic officers, sufficiently so such that these officers are able to alter the policing norms of the entire troop. Hence, an agency s norms of policing will vary according to the racial composition of the troop, 9

13 specifically whether the unit is a majority or minority troop. In their model of racial identity formation, Darity, Mason, and Stewart (2006) show that antagonistic intergroup relations combined with intragroup altruism are elements of individual identity construction in a racialized society. This perspective suggests that traffic stop outcomes may vary according to whether the racial identity of the police officer matches the racial identity of stopped drivers. This racial identity effect is consistent with the theory of representative bureaucracy, which argues that given their shared values and attitudes minority police officers will be less likely to demonstrate bias in the treatment of minority drivers, such as subjecting them to unnecessary searches. The traffic search game presented in Close and Mason (2006, 2007) is used and extended to derive empirical predictions from the theories of racially representative bureaucracy, streetlevel bureaucracy, and the economic theory of identity. Figure 1 captures the strategic interactions of drivers and police in the traffic search game. The police search rate is measured along the Y-axis and may range from 0 (no driver is searched) to 1 (all drivers are searched). Driver criminal behavior is captured along the X-axis and ranges from 0 (no driver engages in crime) to 1 (all drivers engage in crime). For drivers, as the probability of police search increases (decreases), the probability that drivers will engage in criminal activity decreases (increases). For officers, as the probability that drivers will engage in criminal activity increases police increase the probability of searching stopped drivers. [Insert Figure 1] The equilibrium search and crime rates are σ e and χ e, respectively. With homogeneous police and homogeneous drivers, as represented in Figure 1, the equilibrium search and hit rates represent both unbiased policing and efficient enforcement, that is, the minimal level of crime (or 10

14 maximal level of public safety) and the optimal hit rate associated with a given level of policing resources, criminal propensities, policing strategy, etc.. Consider drivers from distinct social groups D {African American, Hispanic, white}, officers from distinct social groups O {African American, Hispanic, white}, and heterogeneous troops T {majority, minority}. Previous research demonstrates that search rates vary by the race of the officer: African American officers have the lowest search rate, white officers have the highest search rate, and Hispanic officers have an intermediate search rate. Additionally, search rates and hit rates vary by the race of the driver. Given the nature of policing and high-levels of organizational socialization, individual officers will face strong incentives to adjust their policing practices to the norms of their troop. All other things equal, majority troops will have a high search rate norm, that is, all officers within these units will have a greater than average probability of searching stopped drivers, regardless of the social group membership of the driver. Similarly, all other things equal, minority troops will have a low search rate norm, that is, all officers within these troops have a lower than average probability of searching stopped drivers, regardless of the social group membership of the driver. Assume that police seek to minimize crime. If so, drivers with the highest search rates also have the highest criminal propensities. Figure 2 presents the empirical implications of active representation and identity theory in police-driver interactions. D m and D w are the driver reaction functions for minority (African American and Hispanic) and white drivers, respectively. These reaction functions capture the assumption that minority drivers have a greater criminal propensity, that is, for a given search rate minority drivers are more likely to engage in crime than white drivers; hence, D m is further from the origin than D w. [Insert Figure 2] 11

15 O w,ma and O m,ma are the reaction functions of white officers in majority troops and minority officers in majority troops, respectively, while O w,mi and O m,mi are the reaction functions of white officers in minority troops and minority officers in minority troops, respectively. These reaction functions capture two assumptions. First, policing practice varies according to the social identity of officers; in particular, white officers search all drivers more than minority officers. Hence, for a given level of criminal intensity, white officers have higher search rates than minority officers, that is, O w,ma is higher than O m,ma and O w,mi is higher than O m,mi. Second, policing practice varies by organizational identity; specifically, majority troops have a higher search norm than minority troops; hence, for a given level of criminal intensity, majority troops have higher search rates than minority troops, that is, O w,ma is higher than O w,mi and O m,ma is higher than O m,mi. Given heterogeneous drivers, officers, and troops, Figure 2 may be employed to establish alternative rank-order tests, assuming either police seek to maximize safety (minimize crime) or that police seek to maximize the hit rate. Each test has equilibrium search conditions and noarbitrage conditions. The equilibrium search conditions are derived from the assumptions we make regarding officer and driver diversity. The no-arbitrage condition establishes the expected pattern of hit rates. Let σ(w,mi m) represent the search rate of minority drivers (either African American or Hispanic) by white officers within a minority troop, that is, point B in Figure 2. Similarly, χ(w,mi w) represents the hit rate of white drivers by white officers within a minority troop, that is, point E in Figure 2. At Point A we have σ(m,mi w), the search rate of white drivers by minority officers (either African American or Hispanic) within a minority troop and at Point F we have χ(m,mi m), the hit rate of minority drivers by minority officers within a minority troop. 12

16 Maximizing hit rate Standard economic analysis of biased policing focuses on police treatment of the marginal driver, that is, the last individuals deemed sufficiently suspicious to stop and search. Regression analysis focuses on the mean driver. For most datasets, including ours, there is no way to identify the marginal driver. Thus, we cannot determine if the marginal driver is dissimilar to the mean driver. Additionally, empirical analysis is often plagued by the difficulty of identifying an appropriate benchmark for assessing the extent of biased policing: the racial and ethnic composition of the population may differ from the racial and ethnic composition of drivers, who may also differ from the racial and ethnic composition of criminals. It is argued that the so-call infra-marginality problem and the benchmark problem undermine successful inference from regression analysis. Outcomes tests provide a way around these problems and are far simpler to implement than regression analysis. Outcomes tests seek to ascertain only whether the data are consistent with the qualitative pattern of outcomes suggested by the theoretical model and its assumptions regarding the optimization objectives and specific strategies of police and the criminal strategies of drivers. Suppose, for example, police define efficient enforcement as maximizing the hit rate. Knowles, Persico, and Todd (2001) show that search rates will differ by the social identity of drivers, but there will be a uniform hit rate (χ*). To the extent that hit rates differ for drivers of alternative social groups, then police can arbitrage this difference, that is, search high hit rate groups more and search low hit rate groups less. As the search rate for a group goes down (up), its criminal activity will increase (decrease). Police search arbitrage will continue until drivers of alternative social groups have equal hit rates. Therefore, for a Knowles, Persico, and Todd (KPT) equilibrium, the equilibrium search conditions (i) and no-arbitrage conditions (ii) are: 13

17 (i) σ(w,ma m) = σ(m,ma m) = σ(w,mi m) = σ(m,mi m) > σ(w,ma w) = σ(m,ma w) = σ(w,mi w) = σ(m,mi m) and (ii) χ(w,ma m) = χ(m,ma m) = χ(w,mi m) = χ(m,mi m) = χ(w,ma w) = χ(m,ma w) = χ(w,mi w) = χ(m,mi m) = χ*, where σ(w,ma m) is the search rate of white officers in majority troops given that a minority driver has been stopped, χ(w,ma m) ) is the hit rate of white officers in majority troops given that a minority driver has been stopped, and the other terms are defined in a similar manner. If a KPT equilibrium exists, racial representation among officers has no effect on either search rates or hit rates. Regardless of troop composition or the racial identity of the officer, minority drivers will have higher search rates than white drivers because minority drivers have tend to engage in more criminal activity. Regardless of troop composition, race of officer, or race of driver, there should be a uniform hit rate. Maximizing public safety (minimizing crime) Maximizing the hit rate is not the same thing as maximizing public safety. Reconciling the search process to driver heterogeneity is the core requirement for maximizing the hit rate. Minimizing crime, or what is the same thing, maximizing public safety, requires a search strategy that accounts for driver heterogeneity, officer heterogeneity, and troop heterogeneity. Just as drivers may differ in their criminal propensity, troops may be charged with providing safety for districts with differing criminal intensity, and officers may not be deployed to areas of uniform criminal activity within a particular troop s district (Close and Mason; 2006, 2007). Consider Figure 2, which has heterogeneous officers, troops, and drivers. For given agency policing norms T {Majority, Minority}, if police define efficient enforcement as minimizing crime, efficient enforcement exists if and only if both of the following conditions 14

18 hold (i) A white officer in a majority or minority troop searching minority drivers, B* or B, respectively, has the maximum search rate and minority officers in a majority or minority troop searching a white driver, A* or A, respectively, have the minimum search rate: maximum search rate = σ(w,t m) > σ(m,t w) = minimum search rate, and (ii) The hit rate for minority officers in a majority or minority troop searching a minority driver, F* or F, respectively, is greater than the hit rate for white officers in a majority or minority troop searching a white driver, E* or E, respectively: maximum hit rate = χ(m,t m) > χ(w,t w) = minimum hit rate. These efficiency conditions are consistent with a specific pattern of heterogeneity among officers and drivers: white officers have a high search rate norm, because they are assigned to high crime area; minority officers have a low search rate norm, because they are assigned to a low crime area; white drivers are a low crime group; and, minority drivers are a high crime group. Given this assumed pattern of heterogenity, when police seek to minimize crime, the search rate of the high crime group (in this example, African Americans or Hispanics) will exceed the search rate of the low crime group (in this example, whites); the hit rate for the high crime group will exceed the hit rate for the low crime group. Accordingly, if the search and hit rate conditions for efficiency do not hold, then the organizational identity of troops or the social identity of officers are an impediment to the efficient provisioning of public safety; police can search differently and increase public safety, that is, have more hits. Consider first the interaction between drivers and officers in minority troops. Crime minimization require that the maximum search rate (shown at point B on Figure 2) for the 15

19 combination of officers with the most intense search practices and drivers with the most criminal activity, that is, white officers with minority drivers. The minimum search rate (point A) will occur for the combination of officers with the least intense search practices and drivers with the least criminal activity, that is, minority officers and white drivers. For public safety maximization, we should observe the maximum hit rate for the combination of officers with the least intense search practices and drivers with the most criminal activity (point F), that is observed when, minority drivers are searched by minority officers. The minimum hit rate (point E) will occur for the combination of officers with the most intense search practices and drivers with the least criminal activity, that is, point E which captures the interaction between white drivers and white officers. Consider now the interaction between drivers and officers in majority troops. Points A, B, E, and F are analogous to the outcomes for minority troops. Note however that all racedriver interactions in majority troops should have higher search rates and lower hit rates than the same race-driver interaction in minority troops. For example, the minority driver minority officer interaction in a minority troop (F) should have a lower search rate and a higher hit rate than the minority driver minority officer interaction in a majority troop (F ). II. The Florida Highway Patrol The data for this study are taken from the Florida Highway Patrol s Traffic Stop Data Report: and the Florida Highway Patrol s Characteristics of Troops dataset: The Traffic Stop Data Report Instructions specify that: 1. All Florida Highway Patrol sworn personnel are to submit a Traffic Stop Data Report form on a weekly basis. 2. The report should include all traffic stops made during the week except those related 16

20 to traffic crash investigations, disabled vehicles, and checkpoints. 3. If a stop is initiated based on aircraft and radar observation, the report is to be made by the trooper actually conducting the stop. 4. Traffic Data Stop forms are to be turned in or mailed to the reporting Florida Highway Patrol station no later than the third day following the end of workweek. Information from the Traffic Stop Data Report includes the county, date, and time of the traffic stop, as well as the trooper s identification number, the assigned troop identifier, and troop s headquarters county. The stopped vehicle is identified by state of registration. Drivers are identified by race, ethnicity, sex, and driver s age. The driver s ethnicity includes whether or not the individual is Hispanic. Racial categories include black, white, Asian, and Native American (American Indian or Alaskan). Hispanics may belong to any racial category. Additional information in the dataset includes the number of passengers in the vehicle, the reason for the traffic stop, the enforcement action, search type, rationale for consent search, the type of contraband seized, and the items seized for forfeiture (other than contraband). Florida Highway Patrol can undertake two types of searches consent and non-consent. A consent search is one where the troopers seek the permission of the drivers to search their vehicle or property. Consent searches can go forward with no evidence of criminality. For example, a trooper s reason for requesting a search can be motivated by racial bias or a desire to practice requesting consent from a driver. On the other hand, a non-consent search is based upon probable cause, negating the need to seek the permission of the driver. Consent searches offer the trooper the most discretion and avoids the administrative hassles, time, and risks associated with obtaining and executing a warrant. In our analysis we present findings for both types of searches, however we focus on consent searches because of their relevance to officer discretion and their 17

21 potential to reflect the strongest relationship between troop representation and search behavior. The Characteristics of Troops dataset includes officer s identification number, date of birth, race, and sex. Unlike Hispanic drivers, Hispanic officers are separate for all other racial groups. The combined dataset will allow us to determine whether and to what extent there are racial and/or ethnic differences in the treatment of drivers after a traffic stop has occurred. The FHP made 6,392,439 stops of black and white drivers for the years Speeding was the major reason for traffic stops (69 percent). Most traffic stops do not result in a search and must searches do not yield contraband: about 1 percent of drivers are searched and less than 20 percent of searches uncover illegal activity. The racial distribution of stopped drivers is black (16.5 percent), Hispanic (19.7 percent), and white (71.4 percent). Asian and Native American represent less than 1.5 percent of all stopped drivers and are omitted from the sample. African Americans and Hispanics represent 15.0 and 16.8 percent, respectively, of Florida s citizens (Florida Legislature, 2003). Men are two-thirds of the stopped drivers. The Florida Highway Patrol (FHP) is responsible for policing 80,967 miles of federal, state, and county roads within the state of Florida (Florida Highway Patrol, 2006). The FHP has 1,770 sworn officers. Non-Hispanic white males represent 66 percent of the sworn officers of the FHP. African American and Hispanic males represent 13 and 10 percent, respectively. Non- Hispanic white females are 7 percent of the sworn force, while African American and Hispanic females represent 2 and 1 percent, respectively. Other minorities, principally Asians and Native Americans, represent about 1 percent of the Florida Highway Patrol. Of course, these percentages vary across the troops. III. Results Tables 1a and 1b present search rates and hit rates, respectively, by race of driver, race of 18

22 officer, and racial composition of troop. Troops are separated into minority and majority troops according to the percent of African American and Hispanic officers assigned to each troop. 2 The data show that regardless of race of driver or racial composition of the administrative unit, white officers have a higher search rate than Hispanic officers and Hispanic officers have a higher search rate than African American officers. For example, the African American male driver white officer search rate is 3.02 percent in majority troops and 1.62 percent in minority troops. The African American male driver African American officer search rate is 0.88 percent in majority troops and 0.30 percent in minority troops. The African American male driver Hispanic officer search rates is 2.03 percent in majority troops and 1.11 percent in minority troops. These search differences are quite large, given that only 1 percent of stopped drivers are searched. (See Table 1). [Insert Tables 1a and 1b] A similar pattern is repeated for all other matches of offers, troops, and drivers. Accordingly, the majority troop is the high search organizational identity, while the minority troop is the low search organizational identity. Controlling for organizational identity and the social identity of officers, hit rates vary by race of drivers. Consider, for example, white officers of majority troops. For all searches, the hit rates for white, African American, and Hispanic drivers are percent, percent, and percent, respectively (Table 1b). But, consent searches are the result of greater officer discretion than non-consent searches. When we examine non-consent searches, which are based on probable cause of criminal behavior, the hit rates for white, African American, and Hispanic drivers are percent, percent, and percent, respectively. When we examine consent searches, which are based on officer suspicion but there is no probable cause to search, the hit rates for white, African American, and Hispanic 19

23 drivers are percent, percent, and 9.26 percent, respectively Indeed, Table 2 shows that in all but one case (Hispanic officers searching African American drivers), the hit rates for consent searches significantly increase in minority troops and in every case the search rates decrease, often times by half. These findings offer the initial and most basic confirmation that troops with increased minority representation search less and find more. However, to understand the impact of minority representation on the efficiency of searching, we must employ a more complex analysis that considers the objectives of law enforcement executives (Knowles, Persico, and Todd 2001; Close and Mason 2006, 2007). [Insert Table 2] Inefficient enforcement: maximizing hit rates If the enforcement objective of the highway patrol is to maximize the organization s hit rates, then our findings demonstrate that the highway patrol is not engaged in efficient policing. A Knowles, Persico, and Todd (KPT) hit rate maximization equilibrium is consistent with unequal search activity. Crime may vary by district of troop coverage, by area of officer assignment, and by the criminal propensity of drivers. Hence, Table 1a is consistent with a KPT equilibrium. However, Table 1b is not consistent with a KPT equilibrium. Hit rate maximization for the highway patrol requires equality of hit rates for all officer-troop-driver identity combinations. Across and within both majority and minority troops, Table 1b shows that hit rates vary according to the identity match of drivers and officers. We focus on consent searches since they are completely discretionary. Consider first majority troops. The consent search hit rates for white officer searches of white, African American, and Hispanic drivers are percent, percent, and 9.26 percent, respectively. White officers can increase their hit rate by searching 20

24 white drivers more and African American and Hispanic drivers less. The consent search hit rates for African Americans officer searches of white, African American, and Hispanic drivers are percent, 8.16 percent, and 2.11 percent, respectively. African American officers can increase the hit rate by searching white drivers more, African American less, and Hispanic drivers much less. The consent search hit rates for Hispanic officer searches of white, African American, and Hispanic drivers are percent, percent, and 7.81 percent, respectively. Hispanic officers can increase the hit rate by searching white drivers more, African American drivers much more, and Hispanic drivers less. Within minority troops, the consent search hit rates for white officer searches of white, African American, and Hispanic drivers are percent, percent, and percent, respectively. White officers can increase the hit rate by searching white drivers more and African American and Hispanic drivers less. The consent search hit rates for African Americans officer searches of white, African American, and Hispanic drivers are percent, percent, and percent, respectively. African American officers can increase the hit rate by searching white and Hispanic drivers more and African American drivers less. The consent search hit rates for Hispanic officer searches of white, African American, and Hispanic drivers are percent, percent, and percent, respectively. Hispanic officers can increase their hit rate by searching white drivers more and African American and Hispanic drivers less. Additionally, as we have previously indicated, hit rates can be increased if all officers in majority troops were to follow a search strategy similar to their same social group in minority troops. Table 2 presents intertroop differences in policing outcomes. In particular, for each officer-driver combination, Table 2 presents the net difference in search rates and hits between majority and minority troops. For example, the white officer consent search rate of white drivers 21

25 is 0.27 percent higher in majority troops than in minority troops, while the white officer hit rate for consent searches of white drivers is 0.92 percent lower in majority troops than in minority troops. Given downward-sloping driver reaction functions, a higher search should result in a lower hit rate. Hence, the positive intergroup net difference in search rates should be matched by a negative intergroup net difference in hit rates. All the search differences are positive, confirming that officers in majority troops search more regardless of the race of driver-race of officer interaction. However, not all of the hit rate differences are negative. In 14 out of 27 combinations, we find a positive hit rate difference by majority troops. A positive hit rate difference implies relatively too much and inefficient searching by majority troops. For non-consent searches, Table 2 shows that white, African American, and Hispanic officers within majority troops over-search African Americans with respect to non-consent searches. Also, for non-consent searches, white and African American officers in majority institutions over-search white and Hispanic drivers. But, for consent searches, Hispanic officers over-search African American drivers. Inefficient enforcement: minimizing crime If the efficient enforcement objective of the highway patrol is to maximize public safety, then the highway patrol is not engaged in efficient enforcement. Table 3 presents Close-Mason outcomes tests. The binary comparisons satisfy the search criterion for maximizing public safety. However, none of the binary comparisons satisfy the hit rate criterion for minimizing crime. Consider for example white male and African American male drivers searched by white and African Americans officers, in majority troops. If efficient enforcement exists, we should observe the following outcomes for majority troops: the maximum search rate (African 22

26 American drivers searched by white officers, from Table 1a) should be greater than the minimum search rate (white drivers searched by African American officers, from Table 1a) and the maximum hit rate (African American drivers searched by African American officers, from Table 1b) should be greater than the minimum hit rate (white drivers searched by white officers, from Table 1b). These results are evidence of inefficient enforcement. For both majority and minority troops, the data are not consistent with efficient enforcement: the required maximum hit rates are less than the required minimum hit rates. In light of this finding, we now examine how search rates and hit rates vary across majority and minority troops. Focusing on consent searches, for majority troops, Table 3 shows (i) maximum search rate = σ W,Ma A = 1.04% > 0.19% = σ A,Ma W = minimum search rate & (ii) maximum hit rate χ A,Ma A = 8.16% < 18.25% = χ W,Ma W minimum hit rate. Similarly, for minority troops, Table 3 shows (i) maximum search rate = σ W,Mi A = 0.42% > 0.03% = σ A,Mi W = minimum search rate & (ii) maximum hit rate χ A,Mi A = 15.22% < 19.17% = χ W,Mi W minimum hit rate. For both majority and minority troops, none of the driver-officer matches are consistent with minimizing crime (maximizing public safety). [Insert Table 3] The bottom panel of Table 3 presents outcomes tests under the assumption that crime is higher in areas patrolled by majority troops; white officers patrol higher crime areas than Hispanic officers who patrol higher crime areas than African American officers; white drivers have a lower criminal propensity than Hispanic drivers and Hispanic drivers have a lower criminal propensity than African American drivers. Given these heterogeneity assumptions, none of the binary comparisons are consistent with minimizing crime as the goal of efficient 23

27 enforcement. For example, for the Hispanic-white identity match the required maximum search rate consistent with efficient enforcement occurs for white officers assigned to majority troops who have stopped Hispanic drivers = 1.13 percent. The required minimum search rate consistent with efficient enforcement is Hispanic officers assigned to minority troops who have stopped white drivers = 0.17 percent. Hence, the search rate criterion holds. The required maximum hit rate consistent with efficient enforcement is Hispanic officers assigned to minority troops who have stopped Hispanic drivers = percent. The required minimum hit rate consistent with efficient enforcement is white officers assigned to majority troops who have stopped white drivers = percent. Hence, the hit rate criterion does not hold. Regardless of the racial composition of the troop or type of search, the search and hit rate data are not consistent with the outcomes test of efficient enforcement. The results suggest that there are efficiency gains, that is, a higher average hit, that can derived from changes in the current pattern of searches. Within majority troops, the following changes raise public safety: fewer searches of African American drivers by African American (consent searches should decrease from 0.30 percent toward 0.07 percent); more searches of white drivers by white officers (consent searches should increase from 0.35 percent toward 0.84 percent); fewer searches of Hispanic drivers by Hispanic officers (consent search should decrease from 0.21 percent toward 0.16 percent). Within minority troops, the following changes increase public safety: fewer searches of African American drivers by African American (consent searches should decrease from 0.05 percent toward 0.01 percent); more searches of white drivers by white officers (consent searches should increase from 0.22 percent toward 0.46 percent); fewer searches of Hispanic drivers by Hispanic officers (consent searches should decrease from 0.11 percent toward 0.00 percent). Examining all troops collectively; the following changes will 24