Racial profiling and searches: Did the politics of racial profiling change police behavior?

\\server05\productn\C\CPP\8-2\CPP204.txt unknown Seq: 1 5-JUN-09 8:30

Racial profiling and searches: Did the politics ofracial profiling change police behavior?*

Patricia Y. WarrenFlorida State University

Donald Tomaskovic-DeveyUniversity Massachusetts, Amherst

Research SummaryScholarly research has documented repeatedly that minority citizensare disproportionately stopped, searched, and arrested relative to theirbaseline populations. In recent years, policymakers have broughtincreased attention to this issue as law-enforcement agencies across theUnited States have faced allegations of racial profiling. In the 1990s, thepolitics generated by accounts of racially biased policing placed height-ened pressure on law-enforcement agencies. However, to date, fewstudies have explored whether the increased social and political scru-tiny placed on police organizations influenced or changed their generalpattern of enforcement among black and white citizens. Using data inthe search and citation file from the North Carolina Highway TrafficStudy, this research specifically examined whether the politics gener-ated by the media coverage of racial profiling and racial profilinglegislation in North Carolina influenced the search practices of officersof the North Carolina State Highway Patrol’s drug interdiction team.The findings suggest that media accounts and the passage of new legis-lation were particularly powerful influences, which thereby reducedracial disparity in searches. Declines in the use of consent searches andan increased probability of finding contraband also were influenced bythe politics of racial profiling.

* This article is part of a larger project on racial profiling funded by the NationalInstitute of Justice (#99-MU-CX-0022). The points of view expressed in this article arethose of the authors and do not represent the official position of the National Instituteof Justice. The authors would like to thank Eric Stewart, Robert Kaufman, and theanonymous reviewers for their helpful comments. Direct correspondence to Patricia Y.Warren at 634 West Call Street, College of Criminology and Criminal Justice, FloridaState University, Tallahassee, FL 32301 (e-mail: [email protected]).

CRIMINOLOGY & Public PolicyVolume 8 Issue 2 Copyright 2009 American Society of Criminology

343


344 Warren and Tomaskovic-Devey

Policy ImplicationsThe results of this study suggest several important policy recommend-ations. First, law-enforcement agencies must monitor the policingpractices of their officers because such supervision can influence officercompliance with fair and unbiased policing policies. Next, supervisorsneed to familiarize themselves with enforcement data to identify poten-tial problem officers and organizational practices. By doing so, policeorganizations will not only increase officer accountability but also willpotentially improve communication with their local communities. Suchcommunication can empower community members to file chargesagainst officers who violate their civil rights. Next, the media and politi-cal effects documented in this study suggest that external oversight canbe particularly influential on police practices. Therefore, police agen-cies should consider developing a citizen complaint board, which is anexternal oversight board that would be responsible for investigatingallegations against officers who engage in racial profiling or otherracially biased enforcement practices. Such a board could assist policeadministrators in identifying problem officers as well as in making rec-ommendations for corrective action. Finally, external social andpolitical pressure must continually be placed on police forces toimprove equity in police behavior. These external forces can assist increating opportunities for police to increase the overall quality and effi-cacy of policing.

Keywords: racial profiling, police searches, hit rates, racial inequality

Racial profiling is a term used to describe the practice of targeting orstopping an individual based primarily on race or ethnicity, rather than onindividualized suspicion or probable cause. It became particularly politi-cized in the late 1990s as police departments came under media and politi-cal scrutiny for disproportionately targeting minority drivers (Farrell,McDevitt, Cronin, and Pierce, 2003; Harris, 2002; Weitzer and Brunson, inpress). Civil rights organizations such as the American Civil LibertiesUnion (ACLU) and the National Association for the Advancement ofColored People (NAACP) launched national media campaigns calling fora ban on racially biased policing, with particular emphasis on the use ofracial profiles in drug interdiction efforts. Shortly thereafter, the U.S.Department of Justice entered into consent decrees with several jurisdic-tions, explicitly prohibiting discriminatory vehicle stop and searchpractices. The consent decrees required data collection within these juris-dictions to monitor institutional compliance (Ramirez, McDevitt, andFarrell, 2000). By 2000, more than 4,000 law-enforcement agencies across


Racial Profiling, Politics, and Media 345

the United States were involved in data collection efforts. As a result ofthe national scrutiny, police departments became acutely aware that theirlaw-enforcement practices were being monitored. Few studies, however,have systematically examined the extent to which this external scrutinyinfluenced police behavior. Therefore, the purpose of this research is toexamine changes in police behavior that might have resulted from thelocal and national politics generated by racial profiling. This researchfocuses in particular on changes in vehicle search practices, as they are thekey behaviors implicated by the charge of racial profiling.

We highlight the relevance of this issue because scholarly research hasrepeatedly demonstrated that organizations, particularly those that areresource dependent, are responsive to the political pressures that originatein their social and political environments (Staw, Sandelands, and Dutton,1981). Thompson (1967) noted, “organizations cope with potential sourcesof adversity that arise from external sources by taking the necessary actionto rebuild a positive position in the social environment.” Police organiza-tions should be particularly vulnerable to external pressures because theyare monitored closely by the media and are resource dependent on electedofficials. These sources of normative and coercive pressure (DiMaggio andPowell, 1982) seem to be present in the politics surrounding racial profil-ing. Evidence exists of such responsiveness by police to media and politicalpressure in past research. For example, Prendergast (2001) found that themedia’s focus on the Los Angeles Police Department’s corruption scandalin the 1990s forced them to create a citizens’ complaint division to investi-gate police misconduct. Within just 3 years, approximately 10% of officerswere disciplined for violating citizens’ civil rights.

In this research, we investigate changes in the pattern of searches ofblack and white citizens by the North Carolina State Highway Patrol’s(NCSHP) Criminal Interdiction Team (CIT) after the politics surroundingracial profiling heightened locally and nationally. We question whether themedia and legislative pressures generated by the politics of racial profilinginfluenced police discretion in the decision to conduct a search. Thisresearch builds on the work of Prendergast (2001) by modeling how thelocalized external pressures generated by the media and local legislativechanges might have influenced CIT officers’ pattern of enforcement.

In the next section, we recount the rise and fall of racial profiling as apolicy issue, with particular emphasis on the role the media played inbringing this phenomenon to national attention. In addition, this researchhighlights the importance of legislative mandates and their influence onpolice behavior. The empirical analyses that follow begin by first establish-ing the degree of racial disparity in searches, then by assessing howdiscretionary searches and racial disparities declined across the studyperiod (1997–2000).

https://www.researchgate.net/publication/5196636_Selection_and_Oversight_in_the_Public_Sector_With_the_Los_Angeles_Police_Department_as_an_Example?el=1_x_8&enrichId=rgreq-621708ee-26e1-4a57-bd02-d4a4d022c301&enrichSource=Y292ZXJQYWdlOzIyNzg5MTQyODtBUzoyMzExNjU4MjMwMjUxNTRAMTQzMjEyNTYzNTUzNg==



History of Racial Profiling

The use of racial profiles dates back to the late 1970s when federalagents created drug courier profiles for the purposes of apprehending drugtraffickers in American airports. By the mid-1980s, lawsuits were filed inlarge numbers challenging their use because anyone who fit the profilecould be stopped and searched by law enforcement. In U.S. v. Sokolow(1989), the U.S. Supreme Court upheld the use of drug courier profiles asan acceptable basis for the temporary detention of potential drug couriers.

This practice was later extended to highways and became a widespreadpolicing strategy in the early 1990s after the U.S. Drug EnforcementAgency (DEA) offered drug interdiction training to local and state patrolofficers. During this time, race/ethnicity was introduced as a legitimate andnormal characteristic in drug courier profiles, and police departments usedthese profiles to make search decisions as they proactively attempted tofind drug couriers (Harris, 2002). This highway drug interdiction program,known as Operation Pipeline, trained more than 27,000 police officersfrom 48 states to use these profiles (Allen-Bell, 1997; ACLU, 1999; Harris,2002). Examples of explicit racial profiles are as follows:

• The Eagle County Sheriff’s Office used the following drug courier profiles onInterstate 70 in Colorado: presence of fast-food wrappers strewn in the car,out-of-state license plates, and dark skin.

• Delaware’s drug courier profile commonly targeted young minority men driv-ing newer model cars and carrying pagers or wearing gold jewelry. Theprofiles also considered the age of the car and whether it was rebuilt withcompartments; similarly a profile was used to stop suspects because they wereyoung minority male occupants in an old vehicle (Allen-Bell, 1997: 5).

In August 1988, under DEA encouragement, the NCSHP created theCIT with the goal of apprehending drug traffickers on North Carolina’smajor interstates and highways (Smith et al., 2003). Because the CIT wascreated under DEA guidance and officers were trained to use racialprofiles in their search for drug couriers, it is an ideal case to study—bothto establish whether racial profiling seems to have been operating and tofollow its progression after the politicization of racial profiling delegiti-mated this policing practice.

Although in this study we highlight the importance of racial profiling inproducing racial disparities in police outcomes, other mechanisms can pro-duce racial disparities in police searches (Tomaskovic-Devey, Mason, andZingraff, 2004). The first is the practice of intensively patrolling high-crimeareas. If these areas have higher minority populations, then patrol prac-tices may produce racially disproportionate vehicular stops and searches.In this case, the observed racial disparity may not result from systematic



racial bias in police search decisions but from concentrated police deploy-ment (Sampson and Bartusch, 1998; Sherman, Gartin, and Buerger, 1989).Racial disparities also can be produced by officer-level cognitive bias and/or racial prejudice. Cognitive bias refers to the types of errors in informa-tion processing that result from unconscious attention to stereotypes (e.g.,black dangerousness) or favoritism to high-status groups (e.g., leniencytoward well-spoken, well-dressed middle-class drivers). Racial prejudicerefers to more active or self-conscious racial animus in specific policeofficers. Such bias might be manifested empirically if we observed a few“bad apple” officers with high rates of stops or searches of minority driv-ers. If a police force disproportionately recruits officers who have highlevels of bias or are prejudiced, then racial disparities in searches mightreflect individual racial bias rather than the organizational practice ofracial profiling.

The distinctions between sources of racial disparity have implicationsfor the current study. First, to adjust for deployment patterns, we muststatistically control for the racial composition of drivers who police officersare most likely to search before we infer that officers are targeting minor-ity drivers for searches. In this study, we use the racial composition ofcitations to capture drivers most vulnerable to the probability of a search.1Second, even if we observe racial disparity in searches, we should distin-guish between officer bias and racial profiling as an organizationalpractice. We accomplish this by officer-level fixed-effects models that sta-tistically control for the stable influence of officer prejudice and cognitivebias in the search decision.

Media Accounts of Local and National Politics

In the late 1990s, racial profiling became a major race relations issue inU.S. politics, and the media played a prominent role in bringing it tonational attention. The televised beating of Rodney King in Los Angelesas well as the Amadou Diallo shooting in New York raised grave questionsabout the role of race in law-enforcement outcomes. The ACLU spon-sored direct investigations of the degree of racial disparity in vehicularstops in several states. For example, in their study of the Philadelphia Nar-cotics Unit, they found that although blacks make up 42% of thepopulation, they comprise 79% of all vehicular stops. Working with the

1. The use of citations is a conservative baseline, as it does not capture all stops.Instead, it captures only stops that resulted in a formal citation, which potentiallyunderestimates the size of racial disparity in searches. However, data on stops are notavailable in North Carolina during the study period.



ACLU, Lamberth’s (1997) study was among the first empirical investiga-tions to substantiate claims of Operation Pipeline linked to racial profilingamong the New Jersey State Police. Similar studies were conducted in Del-aware, Maryland, and California, and similar race disparities wereobserved (Dominitz, 2003).

The media’s consistent focus on racial profiling helped bring the discus-sion about the role of race in police outcomes to center stage in local andnational politics. Figure 1 presents data on media accounts of racial profil-ing both nationally and in North Carolina. The data are presented inquarters and represent trends in reporting that occurred in the New YorkTimes and the two major North Carolina papers, The News and Observerand the Charlotte Observer, from 1997 to 2000.2 The general pattern showsthat racial profiling was not on the media’s radar until the second quarterof 1998. By the fourth quarter of 1998 and through the end of 2000, storiesabout racial profiling routinely appeared in both the New York Times andthe two North Carolina newspapers. As this graph demonstrates, themedia was an important mechanism in the politicization of racial profiling,and we think this politicization also led to the delegitimation of racial pro-filing as a viable policing strategy.3

Ultimately, media accounts of gratuitous treatment of minority citizensled many state governments to enact legislation that would ban racial pro-filing and hold police organizations accountable for systematicallytargeting minority citizens (Dominitz, 2003; Miller, in press). In addition tothe legislation, in 1999, Attorney General Janet Reno convened theStrengthening Police-Community Relations Conference held in Washing-ton, DC, with the purpose of assisting police organizations in collectingdata on this issue. By 2000, more than 4,000 law-enforcement agenciesacross the United States were involved in data collection efforts. As earlyas 1999, several states—including North Carolina—adopted legislationthat required data collection on every police-initiated stop. In North Caro-lina, Senate Bill 76 required all state law-enforcement agencies to trackthe race and ethnic backgrounds of motorists stopped to monitor officers

2. The graph represents a count of the number of times the newspapers featuredarticles about racial profiling or racial bias in police stops and outcomes from 1997 to2000. The two media sources highlighted here from North Carolina represent the larg-est and most widely read newspapers in the state. We examined the media stories in theNew York Times as it is a widely read national newspaper; in focus groups that weconducted with the NCSHP, the officers referenced the national stories circulating inthis news source as problematic and unfair to police organizations.

3. Ironically, soon after the racial profiling was delegitimated as a policing strat-egy, it became legitimated as a practice in customs screening of Arab and Arab-resembling individuals at U.S. airports.



Figure 1. Cumulative media references to racial profiling

North CarolinaNewspapers

New York Times

300

1997

-1

1997

-4

1998

-1

1998

-2

1998

-3

1998

-4

1999

-1

1999

-2

1999

-3

1999

-4

2000

-1

2000

-2

2000

-3

2000

-4

2001

-1

2001

-2

250

200

150

100

0

50

who might be targeting minority drivers. Prior to Senate Bill 76, officersonly recorded the race/ethnicity of motorists who were formally cited for aviolation (e.g., citations and written warnings). By 2005, more than 25states had passed similar legislation.

Prior to the passage of Senate Bill 76 and concurrent with the rise ofracial profiling to national attention, the NCSHP settled a lawsuit allegingthat officers of the CIT unlawfully apprehended and searched an African-American motorist. Although the highway patrol did not admit guilt, theydid settle out of court with the plaintiff (Neff and Stith, 1996). We suspectthat, in addition to the national politics, this local scrutiny influenced theNCSHP enforcement strategies to avoid additional allegations of racialprofiling. However, this lawsuit could be interpreted as a local manifesta-tion of the national politics about racial profiling. Whereas in earlier years,officers received organizational rewards for drug seizures (as we willdemonstrate in this study), this process was extinguished by 2000 when theNCSHP was threatened with local and federal sanctions for targetingminority citizens.

Police Responses to Allegations of Racial Profiling

As media attention and legislative mandates were adopted across theUnited States, significant administrative- and officer-level changes wereimplemented. For instance, the New Jersey Highway Patrol equipped theirpolice vehicles with video cameras to provide evidence of the conduct ofofficers and citizens during motor vehicle stops. Verniero and Zoubek



(1999) noted, “Video cameras, coupled with other data-collection systems,will provide the basis for a reliable and trustworthy system to detectproblems, to prevent abuses, and to protect officers and citizens alike.” Insome police jurisdictions, law enforcement reduced their enforcementactivities after heightened scrutiny. For example, Mas (2006) found thatthe New Jersey State Patrol significantly reduced their enforcement effortswhen they came under scrutiny for arbitrary and unfair policing tactics. Asimilar weakened enforcement effort was observed in Los Angeles. Evenif these observed changes are temporary, the results suggest that social andpolitical scrutiny directed toward police organizations can influence pat-terns of police enforcement. If drug interdiction strategies encouragepretextual stops in “fishing expeditions” for drugs, then the politicizationof racial profiling can be expected to lead to fewer searches, particularly ofblack drivers.

In focus groups conducted in 2000 (after racial profiling was delegiti-mated as a police practice) with the CIT and regular troopers of theNCSHP, both groups of officers vehemently denied that they profiledblack drivers (Smith et al., 2003). They did, however, describe numerouscharacteristics other than traffic violations that potentially made one carmore suspicious than another, which leaves the possibility of individual-officer racial bias in stops and searches open. These factors include loudmusic, bumper stickers connected to drug cultures, Hispanic drivers whothey characterized as unlikely to have valid driver licenses, and cars theyfelt “were out of place.” The use of stereotypes potentially correlated withrace or ethnicity is broadly consistent with Kennedy’s (1997) definition ofracial profiling as the use of race as a symbol of criminal activity.

Empirically Investigating Racially Biased Policing

Despite consistently finding racial disparities in law-enforcement out-comes (Farrell and McDevitt, 2006; Gaines, 2002; Smith et al., 2003; Smithand Petrocelli, 2001), there is little agreement about when the observeddisparities rise to the level of racial profiling (see also Engel, Calnon, andBernard, 2002). Because at least three mechanisms other than racial profil-ing can produce racially biased outcomes, the observation of racialdisparities is not an automatic indication of racial profiling. Rather, dispar-ities indicate that, relative to some baseline population (e.g., racialcomposition of drivers encountered by officers), the enforcement of lawsis unequal. Other mechanisms—such as deployment patterns, cognitivebias, and prejudice—also can produce race disparities.



Scholars have consistently emphasized the importance of baselines andbenchmarks because they help identify the relative risk that groups haveof being subjected to a traffic stop or search. However, there is little con-sensus over which particular baseline is the best benchmark for identifyingthe normal risk of police stops. Studies have used the racial composition ofresident drivers, accidents, and even census population counts to estimatethe relative size of the driving population (Alpert, Smith, and Dunham,2004; Smith and Alpert, 2002; Smith et al., 2003; Walker, 2001). Eachbenchmark has limitations in its ability to identify adequately those mostat risk of being stopped or searched. In addition, the diversity inbenchmarks reflects different conceptualizations of at-risk populations aswell as limitations in available data.

In this study, we use the racial composition of the officers’ own citationsto control for the racial composition of drivers most at risk to be searched.This strategy adjusts for deployment patterns as well as for racial bias onthe part of the officer in the decision to cite drivers. However, we are lessconfident that this baseline controls for officer cognitive bias or prejudiceas sources of racial disparity in searches. For this reason, we use officer-level fixed effects in multivariate models of change in search behavior andoutcomes. These fixed-effects models control for stable officer disposi-tions, which presumably include racial prejudice and cognitive biaseslinked to race.

In addition, some scholars have advocated the use of hit rates in deter-mining the presence of racial bias in policing (Knowles, Persico, and Todd,2001). Hit rates refer to the likelihood of seizing contraband in a search. Ifcomparisons across racial/ethnic groups show lower hit rates for minoritiesthan for whites, this suggests that more pretextual searches are being con-ducted (Persico and Todd, 2004). To the extent that police are not raciallybiased in their search decisions, search and hit rates should be statisticallyequivalent.4 If police organizations are racially biased, it is expected thatracial disparity in searches will be significantly higher than the rate of find-ing contraband (see Knowles et al., 2001).

Numerous empirical studies have found that, although minorities aresearched at significantly higher rates than whites, their hit rates are typi-cally equivalent or lower (Harris, 2002). One of the most compellingnational studies was conducted by the Bureau of Justice Statistics (2005)and demonstrates clear evidence of racial disparity in searches. Black andHispanic drivers were significantly more likely than whites to be searched

4. It is possible that the decision to stop and search drivers varies across racialgroups because the rate of criminal offending varies across racial/ethnic groups. To theextent that this information is based on true criminal offending and not stereotypes, thelikelihood of a search should still be equivalent to the likelihood of finding contraband.



after a police stop; however, their hit rates did not significantly vary fromwhite drivers. Farrell and McDevitt (2006) found that minority driverswere four times as likely as whites to be subjected to a discretionarysearch, but they were less likely than whites to be found with contraband.

In this study, we use hit rates as another search activity that potentiallychanged as a result of the heightened racial politics surrounding racial pro-filing. Our expectation is that, in earlier years of the study period, officerswere more likely to rely on extralegal (i.e., race, age, and vehicle type)factors when making the decision to conduct a search. After the height-ened politics, hit rates are expected to increase as officers were less likelyto rely on “gut feelings” when making enforcement decisions. Instead,they came to rely more heavily on legalistic reasoning, therefore searchingfewer drivers but increasing the likelihood of seizing contraband. Weexpect these effects to be particularly strong among black drivers, as theywere most susceptible to a pretextual search in the earlier part of ourstudy period.

The remainder of this article empirically examines the search behaviorof the NCSHP between 1997 and 2000. We highlight this time periodbecause racial profiling and pretextual stops became politicized anddelegitimated as a policing strategy in North Carolina during these peakyears. We demonstrate in the multivariate analyses that changes inobserved NCSHP enforcement practices were in part influenced by thepolitics surrounding racial profiling. Because the CIT was created specifi-cally to conduct drug interdiction—if racial profiling is the mechanismproducing search disparity—then it is likely that the most dramatic changein search behavior will be observed among these officers. The results ofthis study will provide some baseline evidence that increasing policeaccountability in the social environment can bring changes in the way lawenforcement interacts with citizens.

Data and Methods

The data are taken from the 1997–2000 Search and Citation file fromthe North Carolina Highway Traffic Study. The analyses begin with firstestablishing the size of racial disparity in searches and hit rates in 1997. Wethen follow with an examination of the racial composition of search activ-ity. If racial profiling is the source of disparity, then it should be highest in1997 before the heightened politicization of police practices and resultingdecline due to media and legislative scrutiny. If officers are using racialprofiles in determining who to search, then it is expected that they willproduce inefficient searches of African-American drivers; this result is



observable in lower hit rates in comparison with white drivers. Conver-gence of black and white search and hit rates across the study periodsuggests the erosion of racial profiling as an organizational practice fordetermining who to search.

The analyses that follow are limited to consent and probable-causesearches. Consent searches require verbal or written permission from thedriver. Probable-cause searches are conducted on the grounds that suffi-cient evidence exists to warrant a search. Such evidence includes the visibilityof drug paraphernalia or a weapon.5 Prior research identified consentsearches as the search type most likely to exhibit racial bias (Harris, 2002;Persico and Todd, 2008; Smith et al., 2003). Probable cause, however,leaves little room for racial bias because it requires officers to have verifia-ble suspicion that the driver is engaged in more severe criminal activity.

Dependent Variables

We evaluate the influence of the political environment on the searchpractices of officers by estimating the effects of local and national mediaaccounts as well as racial profiling legislation on three dependent vari-ables. The first dependent variable is a measure of racial disparity in policesearches. It is calculated as a ratio of the percentage of black drivers whoare searched out of those who are cited relative to the percentage of whitedrivers who are searched out of those who are cited. This disparity mea-sure is calculated for each month from January 1997 to December 2000,for each officer.6 The second dependent variable is the probability of aconsent search. We expect the probability of officers who conduct a con-sent search to decline across the study period, as they are highlydiscretionary searches that are not based on any evidence of wrong doing.The variable is coded as 0 if the search was not based on consent and 1 if itwas. The final dependent variable is the probability of finding contraband

5. Searches incident to lawful arrest are excluded because, if sufficient evidenceexists to arrest, the search is considered routine practice.

6. This measure is calculated using the following formula: ([% black driverssearched/% of black drivers cited]/[% of white drivers searched/% of white driverscited]).



(hit rates).7 Hit rates are expected to improve across the study period, asofficers are less likely to subject drivers to discretionary searches. Thisdichotomous measure is coded as 0 if no contraband was found and 1 ifcontraband was found.8

Independent Variables

The following two political mechanisms have been hypothesized toinfluence the search activity of officers across the study period: (1) mediastories published about racial profiling and racially biased policing inNorth Carolina and nationally, and (2) the introduction of Senate Bill 76in North Carolina. The media politics variables are intended to capturewhen stories about driving while black and racial profiling were printed inthree news sources: the New York Times and two North Carolina newspa-pers—The News and Observer and The Charlotte Observer. We highlightthese media sources because they represent the largest and most widelyread newspapers nationally and in North Carolina. In addition, weexamine the influence of stories that circulated in the New York Timesbecause officers in focus groups conducted with the NCSHP referencedthe stories about racial profiling that were being chronicled in nationalnews sources. They were particularly aware of the problems that wereoccurring in New Jersey. Therefore, we use the stories circulating in theNew York Times as a control for the national stories about racial profilingthat were chronicled in national news sources across the United States. Toobtain a count of these news stories, a LexisNexis search was conductedusing the following search terms: “racial profiling,” “racially biased polic-ing,” and “driving while black.” Each article was reviewed for relevance toracial profiling. This variable is coded as 0 if no article was printed and 1 ifat least one article was printed within each month-year interval.

Next, we underscore the importance of Senate Bill 76 because itrequired all state law-enforcement agencies, which included the NCSHP,to record the race, age, and gender of every police-initiated stop. The billwas introduced in the North Carolina legislature in 1998 and passed in1999. Prior to the passage of Senate Bill 76, only stops that resulted in a

7. To test the models for robustness, we also estimated models using changescores (y(t) – y[t – 1]). The results did not change substantively from those reportedhere.

8. The results presented here do not examine when changes in police behavioroccurred. Instead, the analyses demonstrate general changes in police practices thatmight have been influenced by the politics surrounding racial profiling.



formal written citation were recorded. This measure is coded as 0 prior tothe introduction of the legislation and 1 thereafter. Because the legislationwas not introduced until March 1998, every month-year time periodbefore this time is coded as 0 and 1 thereafter.9

Results

Table 1 presents the racial composition of drivers cited and searched foreach of the nine troops of the NCSHP in 1997. Most troops recorded fewsearches. Officers of the CIT were responsible for approximately 70% ofall searches. Clear differences are evident across troops in both the rate ofsearch activity and the size of racial disparity in searches. The last columnof Table 1 displays the race-specific relative ratios of searches to stops. Theratios are calculated using the following formula: ([%African Americansearched/% African American cited]/[% white searched/% white cited]).In 1997 African Americans had higher rates of being searched than whitedrivers across all troops, although the magnitude of this increased risk var-ies across troops. At the low end, African Americans were 1.27 times morelikely than whites to be searched after a stop in Troop 8. In the CIT andTroop 1, African Americans were four times as likely to be searched aswhite drivers in 1997. Although black drivers had higher odds of beingsubjected to a search, the probability of finding contraband was smallerthan that for whites. Table 2 displays the hit rates for all troops across thestudy period. Except for Troop 3, hit rates were lower among black driversdespite the fact they were subjected to searches at rates that were two tofour times higher than whites.10

One of the most striking findings in Tables 1 and 2 is that, with theexception of the CIT, regular troopers of the NCSHP did not routinelysearch vehicles. At the organizational level, if racial profiling as an institu-tional practice is producing the observed racial disparity in searches, thenit seems that regular troopers were not routinely involved. In fact, 90% oftroopers did not conduct a single search in 1997. In contrast, the officers ofthe CIT conducted 70% of searches in 1997, despite comprising only 2%of all NCSHP officers. Therefore, the remaining analyses will focus exclu-sively on the search activity of the CIT because that is the only unit wherewe conclude that it is reasonable to believe that racial profiling in searchactivity was a common organizational practice. Evidence in Tables 1 and 2consistently points to the CIT as both search specialists and producers of

9. The study period occurred 15 months before the introduction of Senate Bill 76and 33 months after the introduction.

10. Although the hit rates for whites are higher across all troops, chi-square valuesonly yield significant race differences in the CIT and Troop 4.



Table 1. Race differences in the probability of a search by troop, 1997

Number Total Ratio of aof Officers White/ Black to

in Each Black % Black % Black % White % White WhiteTroop Troop Searches Cited Searched Cited Searched Search

CIT 25 630 23.7 56.3 76.3 43.6 4.151 154 43 5.5 18.8 94.5 79.1 4.082 147 5 16.3 40 83.7 60.0 3.423 139 23 30 52 70 47.8 2.544 127 59 8 18.6 92 81.3 2.635 144 17 26.8 47.1 73.2 52.9 2.436 139 52 37.8 48 62.2 51.9 1.537 182 9 37.4 44.4 62.6 55.5 1.338 170 20 29.8 35 70.2 65 1.27

Total 1227 858

Note. CIT = Criminal Interdiction Team.

substantial racial disparities in searches and hit rates. Although othertroops also have high levels of search disparity, they conducted fewsearches. Only Troop 4 had both high levels of racial disparity in searchesand enough searches to produce a significant racial disparity in hit rates.But even in this troop, only 10 of the 127 officers conducted a search in1997. However, searches were a normal policing behavior in the CIT,where evidence of racial profiling in 1997 is most compelling.

Table 2. Race differences in hit rates, 1997

% Black Black Hit % White White Hit Chi-SquareTroop Searched Rate Searched Rate ValueCIT 56.3 23.6 43.6 41.6 24.74*

1 18.8 22.2 79.1 47.0 2.732 40 0 60.0 33.3 .8333 52 50 47.8 45.5 2.544 18.6 27.2 81.3 54.1 5.370*

5 47.1 87.5 52.9 88.8 3.086 48 24.0 51.9 37.0 1.047 44.4 25.0 55.5 60.0 1.238 35 0 65 69.2 9.10*

Note. CIT = Criminal Interdiction Team.*p = .05.



Search Activity of the CIT

Table 3 examines the stability of the racial disparities observed in Tables1 and 2 for officers of the CIT. The last column of Table 3 displays therelative ratios of a black-to-white search. The rate of black searchesdeclined dramatically. In 1997, the odds of a black driver being searchedby a member of the CIT were four times that of white drivers. By 2000, theodds had been cut in half. The general trend in Table 3 suggests that, in1997, officers of the CIT were disproportionately searching black drivers;however, this behavior eroded. Meanwhile, searches of white drivers wererelatively stable in 1997, rose in 1998, and then dropped in 1999. Thesedata suggest a dramatic shift in the selection of who officers chose tosearch, with the targeting of black drivers declining precipitously. It is stillthe case, however, that black drivers remained twice as likely as whitedrivers to be searched in 2000, given their incidence of being cited fordriving violations.

Table 3. Race differences in search activity of the CIT, 1997–2000

Ratio ofNumber Total Total Black to

of Total Black % Black White % White WhiteYear Officers Searched Searched Searched Searched Searched Searched

1997 25 630 355 56.3 275 43.6 4.151998 16 568 237 58.3 331 41.7 3.591999 14 377 147 38.9 230 61 1.722000 13 347 147 42.4 200 57.6 2.21

Note. CIT = Criminal Interdiction Team.

Race Differences in Consent Searches and Hit Rates

This section highlights race differences in probable cause and consentsearches. We emphasize this distinction because consent searches are dis-cretionary and do not require any evidence of wrong doing. Instead, aslong as the driver provides the officer with verbal or written consent, theofficer is justified in conducting the search (Harris, 2002). The SupremeCourt has ruled that consent searches are an important part of policeinvestigatory work, reasoning that officers must have some discretionarypower to control crime effectively. It has been noted that African Ameri-cans are more frequently subjected to consent searches (Farrell andMcDevitt, 2006; Smith et al., 2003), because they are often perceived to becriminally dangerous (see also Chiricos, Welch, and Gertz, 2004;Kennedy, 1997).



Table 4 displays black and white differences in the proportion of consentand probable-cause searches. In 1997, among black and white drivers, 90%of all searches resulted from verbal or written consent. By 2000, this num-ber had dramatically declined to 53%. Interestingly, blacks and whites in1997 were almost equally likely to be subjected to a consent search. By2000, this trend changed substantially with 77% of whites and only 53% ofblacks subjected to a consent search. We suspect that this shift resulted inpart from the external scrutiny generated by the problem of racial profil-ing, and officers consequently relied less on discretion in the decision tosearch black drivers.11

Table 4. Proportion of consent and probable cause searches conductedby the CIT by race, 1997–2000

% White % BlackProbable % White Probable % Black

Total Cause Consent Cause ConsentYear Searched Searches Searches Searches Search1997 630 .09 .91 .07 .931998 568 .15 .85 .25 .751999 377 .28 .72 .36 .642000 347 .33 .77 .47 .53Note. CIT = Criminal Interdiction Team.

Table 5 reports race differences in hit rates. In 1997, the hit rates forwhites are significantly higher than for blacks—despite the fact that blackdrivers are subjected to a search more frequently (see Table 1). Assearches of black drivers declined in 1998, hit rates converged. In 2000, thehit rates for black drivers was marginally but not significantly higher thanthat of whites, which is consistent with their lower incidence of consentsearches. This convergence in hit rates is a product of officers reducing thetotal number of searches, particularly consent searches, among black driv-ers. In 1997, 355 black searches occurred; by 2000, this number haddeclined to only 147. Although officers of the CIT eventually reduced thenumber of searches of white drivers, that decline is not as dramatic as thedecline observed among black drivers. This notable shift provides what weinterpret as evidence that officers abandoned enforcement practices thatcould potentially be considered racially biased.

11. It is the case that consent searches result from officers asking drivers for con-sent to search their vehicle. Therefore, it is possible that reduction in consent searchesamong black and white drivers resulted from drivers becoming more informed abouttheir rights and who were therefore less compliant with these kinds of police requests.This study cannot empirically model this process because of data limitations.



Table 5. Race differences in hit rates for CIT, 1997–2000

Number Total Black Black Total White White Chi-of Black Contraband Hit White Contraband Hit Square

Year Officers Searches Found Rate Searched Found Rate Value

1997 25 355 84 24% 275 99 36% 11.45*

1998 16 237 62 26% 331 90 27% .081999 14 147 50 34% 230 72 31% .302000 13 147 53 36% 200 62 31% .98Note. CIT = Criminal Interdiction Team.*p = .001.

It is possible that the decline in observed racial disparities might resultfrom some other process that we have not yet accounted for. More impor-tantly, the number of CIT officers dramatically declined across the studyperiod. In 1997, there were approximately 25 officers; by 2000, this num-ber had declined to 13. The decline in the size of the CIT may have alsobeen a response to the politics of racial profiling. It is also likely that thedecline in the total number of searches is associated with the decliningnumber of officers. Importantly, the change in racial disparity mightreflect a management decision by the NCSHP to weed out “bad appleofficers” to avoid allegations of racial profiling.

Multivariate Analyses

We hypothesized that the politics generated by the media scrutiny ofracial profiling and the introduction of Senate Bill 76 in North Carolinainfluenced officers’ pattern of enforcement, particularly toward black driv-ers. Although some of this influence could be direct—that is, officers areresponding to political discussions in the media and the legislature—wesuspect that much of it is organizational, as the NCSHP began to monitorthe racial composition of searches more closely and senior officers cau-tioned troopers against racially biased policing. To assess systematicallywhether the politics surrounding racial profiling influenced law-enforcement practices, we use time-series analysis to estimate their effectson three dependent variables: (1) racial disparity in searches, (2) theprobability of a consent search, and (3) the probability of finding contra-band (hit rates).12

12. Because we use time series as our modeling strategy, we must test and, if nec-essary, correct for the autocorrelation that often exists when data are ordered by time.If substantial autocorrelation is present, then the standard errors are biased, whichaffects significance tests. To test whether autocorrelation is problem in our data, theDurbin Watson statistic was examined and significant positive autocorrelation wasfound; therefore, the standard errors are reweighted, which corrects for autocorrelation.



The models presented here are intended to measure general changes inofficer behavior that might have been influenced by the racial profilingpolitics. These models assess overall changes in police enforcement thatare associated with the politics of racial profiling. The fixed-effects modelsallow us to examine whether the changes in search behavior are associatedwith changes in the composition of officers or whether they represent trueorganizational change. In addition, because it is unlikely that the explana-tory variables have contemporaneous effects on the dependent variables,all three mechanisms are measured with a lag of 2 months. We examinedthese models with longer and shorter lags. The maximum effects and thebest model fit happen with lags of 2 months. To that end, the results of thisstudy provide baseline evidence that the influence of media stories andSenate Bill 76 were mediated through changes in managerial or organiza-tional practices, rather than being simply the result of officers reading thenewspaper or listening to the evening news.

Table 6 analyzes racial disparities in searches between black and whitedrivers from 1997 to 2000. In these analyses, the unit of the analysis isofficer months and the dependent variable is the relative odds of a blackto a white search. Model 1 begins with a simple time-trend model andassesses the influence of time (as measured in months) on racial disparitiesin searches. The second and third models gauge the influence of mediastories and the enactment of Senate Bill 76 net of this time trend. Model 4includes officer-level fixed effects to control for any variance in the depen-dent variable that might be caused by unobserved heterogeneity associ-ated with officer characteristics (Piquero and Rengert, 1999). In thesemodels, a separate constant term is estimated for each officer in the sam-ple. The fixed-effects estimator allows for the examination of shifts in therelative proportion of blacks and whites searched that are associated withunobserved attributes of individual officers. Most importantly, the officerfixed effects control for the declining number of officers in the CIT. If thesize of the disparity shifts in the fixed-effects model, then it most likelyreflects the removal of the most racially prejudiced or aggressive officerswho were targeting black drivers in 1997.

Tables 7 and 8 present time-series estimates of the probability of con-sent searches and contraband seizures. The coefficients are presented inboth tables as odds ratios because the dependent variables are dichoto-mous. For both, the unit of analysis is the search itself. We pay particularattention to consent searches because they, unlike probable-causesearches, allow officers to go on “fishing expeditions” in the discovery ofcontraband (Harris, 2002; Heuman and Cassak, 2003). For both analyses,



the first model displays estimates of race-specific time trends. We also con-trol for the age and sex of the person searched because both measures arepotentially correlated with the search decision. Model 2 adds the two mea-sures of media stories, whereas Model 3 includes the indicator variable forSenate Bill 76. Model 4 includes officer-level fixed effects.

Racial Disparity in Searches

Model 1 of Table 6 demonstrates that racial disparities in searchesdeclined across the study period. Model 2 incorporates the media mea-sures for the North Carolina newspapers and the New York Times. Mediastories generated by the New York Times do not significantly influenceracial disparities in searches. However, the two North Carolina newssources do produce a reduction in race disparity in searches (–.098)approximately 2 months after the stories appeared. That is, racial disparityin searches becomes significantly smaller after periods of local media cov-erage of racial profiling. Model 3 incorporates the influence of the passageof Senate Bill 76, with the disparity in searches between blacks and whitesagain decreasing (–.079).

Table 6. Time-series models of racial politics and its effect on racialdisparities in black and white searches, 1997–2000

Model 4:Model 1 Model 2 Model 3 Fixed Effects

Month-Year −.050 (.001)** −.020 (.006)** −.015 (.006)** −.011 (.001)*

Driving-while-Black PoliticsNC media stories (t–2) −.098 (.116)** −.082 (.107)** −.901 (.113)*

New York Times (t–2) −.060 (.142) −.045 (.140) −.018 (.139)Senate Bill 76 (t–2) −.079 (.176)** −.092 (.164)**

R2 .035 .073 .124 .155

Notes. Standard errors are reported in parentheses. NC = North Carolina*p = .05. **p = .01.

Model 4 includes officer fixed effects that control for unobserved heter-ogeneity associated with officer-level characteristics. The effect sizes forboth the local media measure and the passage of Senate Bill 76 are slightlylarger than the one observed in Model 3, which suggests that the externalpolitical environment influenced not only a shift in officer-level practicesbut also a shift in practices at the organizational level. We suspect that, in



periods after greater external scrutiny, police supervisors might have moreclosely examined officers and thereby modified their pattern of enforce-ment. Ultimately, organizational changes as well as those associated withofficer behavior reduced the size of racial disparity in searches.

Racial Politics and Consent Searches

Table 7 presents the times-series analyses for the probability of a con-sent search. Model 1 explores the influence of time and driver character-istics. In general, officers are less likely to subject all drivers, particularlyblack drivers, to a consent search across the study period. Controlling forthese dual time trends, black drivers end up with fewer consent searchesthan white drivers (we observed this pattern in Table 4). We hypothesizedthat these observed changes were influenced by the heightened mediascrutiny placed on police departments and the passage of Senate Bill 76.We examine these hypotheses in Models 2 and 3. With a 2-month lag, con-sent searches decline after the publication of racial profiling stories in theNorth Carolina newspapers. The national stories have no observed localeffects. The temporal decline in searches actually strengthens marginallyafter media stories are entered into the model, which suggests that,although new media stories do depress consent searches, they do notexplain race-specific time trends in consent searches.

Table 7. Time-series models of racial politics and its effect on consentsearches, 1997–2000


Control VariablesMonth-Year .994 (.001)** .990 (.008)** .987 (.008)* .992 (.004)*

Time*Black .997 (.001)* .992 (.001)* .998 (.138)* .999 (.001)*

Black .889 (.005)** .881 (.005)* .871 (.002)* .879 (.001)*

Male .690 (.163) .684 (.162) .687 (.162) .685 (.168)Age 1.052 (.008) 1.052 (.008) 1.049 (.008) 1.048 (.008)

Driving-while-Black PoliticsNC media stories(t–2) .905 (.145)** .886 (.014)* .892 (.157)*

New York Times(t–2) .816 (.158) .802 (.181) 1.016 (.204)Senate Bill 76(t–2) .867 (.106)* .881 (.109)*

Chi-Square 26.55** 30.02** 43.21** 47.20**

Notes. Standard errors are reported in parentheses. Coefficients are reported as odds ratios.NC = North Carolina.*p = .05. **p = .01.



Model 3 explores the relative influence of the passage of Senate Bill 76.The passage of this legislation significantly reduces the likelihood thatdrivers will be subjected to a consent search across the study period.Model 4 incorporates officer-level fixed effects, and the size of the effectsfor the NC media and the passage of Senate Bill 76 is slightly larger thanthat observed in Model 3. Because these are within-officer change models,we can conclude that both the NC media stories and Senate Bill 76reduced officers’ use of consent searches. Race differences in thedecreased use of consent searches remain significant and even strength-ened in these final models. Thus, officers’ use of consent searchesdecreased partly in response to the external scrutiny. In other words,officers simply searched fewer drivers, particularly African Americans,and it was consent searches that were curtailed.

Racial Politics and Hit Rates

The final search activity of interest is the probability of finding contra-band. We expected the probability of finding contraband to improve afterthe politics surrounding racial profiling heightened and as officers sub-jected all drivers to fewer discretionary searches. Both the reduction in theoverall number of searches as well as the heavier reliance on probablecause should increase the likelihood of finding contraband.

The results from Model 1 suggest that, across time, hit rates increasedand they did so more dramatically for searches of cars driven by blackdrivers. Model 2 incorporates the influence of the media on hit rates, andwe observe that hit rates increase after periods with greater local mediacoverage about racial profiling. This finding explains some of the observedincrease in hit rates and race differences in hit rate trajectories. Model 3 incor-porates the influence of the passage of Senate Bill 76, and it also increasesthe likelihood of finding contraband across the study period. At this point,all time trends and race differences in hit rates are no longer significant,and the coefficients are close to one. Therefore, the increased efficiency in offi-cer searches seems to have been influenced by the political environment.

The introduction of officer-level fixed effects again marginally increasesestimates of political influence, which suggests that real changes occurredin officer behavior—not simply attrition in officers with low hit rates.Again, we have fairly consistent evidence of real organizational change,rather than simple changes in the array of officers with stable dispositions.



Table 8. Time-series models predicting racial politics and its effect onhit rates, 1997–2000


Control VariablesMonth-Year 1.006 (.006)* 1.002 (.007)* .998 (.007) .999 (.007)Black*Time 1.001 (.001)* 1.000 (.009)* .999 (.001) .998 (.001)Black 1.015 (.002)* 1.009 (.004)* 1.000 (.004) .999 (.003)Male .910 (.168) .916 (.171) .917 (.170) .945 (.179)Age .964 (.005) .963 (.005) .964 (.006) .964 (.006)

Driving-while-Black PoliticsNC media stories(t–2) 1.520 (.113)** 1.478 (.115)** 1.484 (.112)*

New York Times(t–2) .811 (.246) .818 (.244) .778 (.146)Senate Bill 76(t–2) 1.146 (.196)** 1.161 (.205)*

Chi-Square 54.19** 61.02** 64.67** 68.54**

Notes. Standard error is reported in parentheses. Coefficients are reported as odds ratios.NC = North Carolina.*p = .05. ** p = .01.

In summary, the results presented here suggest that the external envi-ronment can significantly influence officer and organizational practices.The support studies by Prendergast (2001) and Mas (2006) by demonstrat-ing that social and political pressures generated by the externalenvironment can influence significant organizational accountability amongpolice forces. The new accountability standards encourage new policingprocedures with the goal of reducing racial disparity and bias in police-citizen encounters.

Discussion

Scholarly research has well documented that blacks are stopped,searched, and arrested disproportionately relative to their baseline popu-lations. In recent years, the enforcement practices that led to suchdisparities were brought into question amidst numerous accounts of gratu-itous treatment of minority citizens, such as the Rodney King beating andthe Amadou Diallo shooting. The heightened awareness of incidencessuch as these directed increased social and political attention to the prob-lem of racial profiling. In the 1990s, police organizations across the UnitedStates were placed under investigation for unfairly targeting black drivers.To date, few studies have explored whether the national scrutiny ofracially biased policing influenced changes in officers’ pattern of enforce-ment. This research examined the issue by assessing whether and how theracial politics surrounding racial profiling influenced changes in officers’search practices of black and white citizens. Racial politics were theorized



to be particularly consequential for black drivers, as they were most vul-nerable to racially biased policing practices. Although this hypothesis wasconfirmed, we also observed that the NCSHP reduced its search activity ofwhite and black drivers in addition to reducing the size of the CIT (whowas primarily responsible for drug interdiction) during the study period.

Using time-series analyses, this research examined changes in the fol-lowing three separate search practices: (1) racial disparity in searches ofblack and white drivers, (2) the probability of a consent search, and (3) hitrates. This research was particularly interested in identifying whether andhow the external environment (i.e., media and racial profiling legislation)influences police practices. In general, we found that the size of racial dis-parity in searches declined across the study period. It is also the case thatthe decline was closely associated with the politics generated by racial pro-filing. The media played a major role in weakening police enforcementefforts, presumably because individual troopers were made more aware ofracial bias and senior officers were encouraged to monitor officer behaviorclosely. However, Senate Bill 76—which increased the monitoring of theNCSHP by the state legislature—led to more due process in searches,which increased the use of probable cause to justify searches anddecreased the number of discretionary searches. Not surprisingly, as dis-cretionary searches of black and white drivers fell, hit rates increased.

In sum, this research provides some baseline evidence that the socialand political environment can be particularly consequential for policeorganizations. It also provides insight for policymakers and police adminis-trators. First, police organizations should consistently monitor the policingstrategies of their officers because, with proper oversight, officers willscreen their behavior to avoid allegations of racial profiling (Alpert, Dun-ham, and Smith, 2007; Bureau of Justice Statistics, 2005; Farrell andMcDevitt, 2006). An example of this strategy is to encourage police orga-nizations to collect data regularly and encourage supervisors to becomefamiliar with the data to monitor properly the enforcement practices oftheir officers. Senior officers can monitor troopers’ behavior both throughreports on the racial composition of stops, citations, and searches andthrough stressing the value of race neutrality in training and squad meet-ings. Miller (in press) reports that more organizations have policies againstracial profiling than those that collect data to make monitoring possible.We have no estimates of how frequent monitoring is conducted across theUnited States. However, given the findings in this study, we suspect it is notcommon, except perhaps during periods of intense political or media scrutiny.

Despite police organizations’ notorious resistance to change—particu-larly change that is generated by the external environment—someresearch has shown that police organizations are sensitive to public per-ceptions, which thereby increases the likelihood that they will implement



organizational policies that discourage discriminatory and biased behavior(Miller, in press). Recent work on race and gender employment bias inprivate sector organizations suggests that holding management accounta-ble for equal employment opportunity is particularly efficacious (Kaley,Dobbin, and Kelly, 2006). Senate Bill 76 may have worked in exactly thisway, holding senior NCSHP officers accountable and thereby shifting rou-tine policy and managerial behavior. We do not have more recent data onracial disparity in searches by the NCSHP. However, we do wonder whathappens to police organizations in the absence of political pressure. Dothey revert to earlier practices and biases? Or does the new organizationalsystem become institutionalized? We suspect the answer will hinge onmanagement accountability and routine monitoring of police behavior.

The results presented in this study provide some evidence that externaloversight can be particularly influential on police behavior, especiallygiven the fact that there are relatively few officers responsible for most ofthe disparate search practices. To that end, police organizations should notonly regularly supervise officers, but they should also use their data-collection efforts to empowering their officers and local community mem-bers to work together to build better relations. Because the externalenvironment is particularly influential on police behavior, the creation ofexternal oversight boards comprising civilians is potentially another way ofevaluating police behavior. This oversight board would be charged withevaluating complaints against officers who violate citizens’ civil rights. Themandate of these citizen boards would need to be expanded to react notmerely to complaints but to evaluate senior officers routinely for the racialneutrality of their organizations’ police practices. External accountabilityboards might be more efficient than internal boards based upon the argu-ment that law enforcement is incapable of being impartial wheninvestigating “one of their own.” The creation of these boards also willsend a message to police organizations that they are not free from externaloversight. Although there has been some criticism launched against civil-ian review boards (see Goldsmith and Lewis, 2000), in general, they canhelp identify potential problems in police organizations while forcingpolice chiefs to remedy problems that might otherwise have been ignored.

References

Allen-Bell, Angela A. 1997. The birth of the crime: Driving while black (DWB). South-ern University Law, 25: 195–225.

Alpert, Geoffrey P., Roger G. Dunham, and Michael R. Smith. 2007. Investigatingracial profiling by the Miami-Dade Police Department: A multi-method approach.Justice Quarterly, 6: 25–55.



Alpert, Geoffrey P., Michael R. Smith, and Roger G. Dunham. 2004. Toward a betterbenchmark: Assessing the utility of not-at-fault traffic crash data in racial profilingresearch. Journal of Justice Research and Statistics Association, 6: 43–69.

American Civil Liberties Union (ACLU). 1999. Driving while black: Racial profiling onour nation’s highways. New York: Author.

Bureau of Justice Statistics. 2005. Contacts between police and the public: Findings fromthe 2002 national survey. Washington, DC: U.S. Department of Justice.

Chiricos, Ted, Kelly Welch, and Marc Gertz. 2004. Racial typification of crime and sup-port for punitive measures. Criminology, 35: 107–131.

DiMaggio, Paul J. and Walter W. Powell. 1982. The Iron Cage revisited: Institutionalisomorphism and collective rationality in organizational fields. American SociologicalReview, 48: 147–160.

Dominitz, Jeff. 2003. How do the laws of probability constrain legislative and judicialefforts to stop racial profiling? American Law and Economics Review, 5: 412–432.

Engel, Robin Shepard, Jennifer M. Calnon, and Thomas J. Bernard. 2002. Theory andracial profiling: Shortcomings and future directions in research. Justice Quarterly,19: 249–273.

Farrell, Amy and Jack McDevitt. 2006. Rhode Island traffic stop statistics: 2004–2005final report. Boston, MA: Northeastern University.

Farrell, Amy, Jack McDevitt, Shea Cronin, and Erica Pierce. 2003. Rhode Island TrafficStop Statistics Act final report. Boston, MA: Northeastern University.

Gaines, Larry K. 2002. An analysis of traffic stop data in the city of Riverside. Riverside,CA: City of Riverside.

Goldsmith, Andrew and Colleen Lewis. 2000. Civilian oversight of policing. Portland,OR: Hart Publishing.

Harris, David. 2002. Profiles in injustice: Why racial profiling cannot work. New York:New Press.

Heumann, Milton and Lance Cassak. 2003. Good cop, bad cop: Racial profiling andcompeting views of justice. New York: Peter Lang Publishing.

Kaley, Alexandra, Frank Dobbin, and Erin Kelly. 2006. Best practices or best guesses?Assessing the efficacy of corporate affirmative action and diversity policies. AmericanSociological Review, 7: 589–617.

Kennedy, Randall. 1997. Race, crime, and the law. New York: Pantheon.

Knowles, John, Nicola Persico, and Petra Todd. 2001. Racial bias in motor vehiclesearchers: Theory and evidence. Journal of Political Economy, 109: 203–230.

Lamberth, John. 1997. A report to the American Civil Liberties Union. New York:ACLU.

Mas, Alexandre. 2006. Pay, reference points, and police performance. Quarterly Journalof Economics, 121: 783–821.

Miller, Kirk. In press. The institutionalization of racial profiling policy: An examinationof antiprofiling policy adoption among large law enforcement agencies. Crime &Delinquency.

Neff, Joseph and Pat Stith. 1996. Highway drug unit focuses on blacks. News andObserver, July 28: A1.



Persico, Nicola and Petra Todd. 2004. Generalising the hit rates test for racial bias inlaw enforcement, with an application to vehicle searches in Wichita. The EconomicJournal, 116: 351–367.

Persico, Nicola and Petra Todd. 2008. The hit rates for racial bias in motor vehiclesearches. Justice Quarterly, 25: 37–53.

Piquero, Alex R. and George Rengert. 1999. Studying deterrence with active residentialburglars: A research note. Justice Quarterly, 16: 451–471.

Prendergast, Candice. 2001. Selection and oversight in the public section, with the LosAngeles Police Department as an example. University of Chicago, National Bureau ofEconomic Research Working Paper #8664.

Ramirez, Deborah, Jack McDevitt, and Amy Farrell. 2000. A resource guide on racialprofiling data collection systems: Promising practices and lessons learned. Washington,DC: U.S. Department of Justice.

Sampson, Robert and Dawn Bartusch. 1998. Legal cynicism and subcultural toleranceof deviance. Law & Society Review, 32: 777–804.

Sherman, Lawrence W., Patrick R. Gartin, and Michael E. Buerger. 1989. Hot spots ofpredatory crime: Routine activities and the criminology of place. Criminology, 27:27–55.

Smith, Michael and Geoffrey P. Alpert. 2002. Searching for direction: Courts, socialscience, and the adjudication of racial profiling claims. Justice Quarterly, 19: 673–703.

Smith, Michael and Matthew Petrocelli. 2001. Racial profiling? A multivariate analysisof police traffic stop data. Police Quarterly, 4: 4–27.

Smith, William R., Donald Tomaskovic-Devey, Matthew Zingraff, H. Marcinda Mason,Patricia Y. Warren, Cynthia Pfaff Wright et al. 2003. The North Carolina highwaytraffic study. Final report to the National Institute of Justice. Washington, DC: U.S.Department of Justice.

Staw, Barry M., Lance E. Sandelands, and Jane E. Dutton. 1981. Threat rigidity effectsin organizational behavior: A multilevel analysis. Administrative Science Quarterly,26: 501–524.

Thompson, James D. 1967. Organizations in action. New York: McGraw-Hill.

Tomaskovic-Devey, Donald, Marcinda Mason, and Matthew Zingraff. 2004. Lookingfor the driving while black phenomenon: Conceptualizing racial bias processes andtheir associated distributions. Police Quarterly, 7: 3–29.

Verniero, Peter and Paul H. Zoubek. 1999. Interim report of the State Police ReviewTeam regarding allegations of racial profiling. Newark: New Jersey Office of theAttorney General.

Walker, Samuel. 2001. Searching for the denominator: Problems with police traffic stopdata and early warning system solutions. Justice Research and Policy, 3: 63–95.

Weitzer, Ronald and Rod K. Brunson. (In press). Strategic responses to the policeamong inner-city youth. Sociological Quarterly.

Cases Cited

U.S. v. Sokolow, 490 U.S. 1 (1989).



Patricia Y. Warren is an assistant professor in the College of Criminology and Crimi-nal Justice at Florida State University. Her research interests include racial profilingand racial bias in crime and justice outcomes. In addition to her work on racial profil-ing, she is exploring how the intersection of race, class, and gender influences judicialdecision making.

Donald Tomaskovic-Devey is a professor and the chair of sociology at the Universityof Massachusetts, Amherst. In addition to racial profiling research, he is interested inthe processes of organizational inequality and change, particularly in regard to equalemployment opportunity. He is also exploring interorganizational market relationships,including exploring the contrasts among market, hierarchy, and embeddednessapproaches.

Racial profiling and searches: Did the politics of racial profiling change police behavior?

Documents