does a helping hand put others at risk - affirmative action, police departments, and crime (2000) by...

DOES A HELPING HAND PUT OTHERS AT RISK?:AFFIRMATIVE ACTION, POLICE DEPARTMENTS, AND CRIME

JOHN R. LOTT, JR.

Will increasing the number of minority and women police officers make lawenforcement more effectie by drawing on abilities that hae gone untapped andcreating better contact with communities and ictims? Or will standards hae to belowered too far before large numbers of minorities and women can be hired? Usingcross-sectional time-series data for U.S. cities for 1987, 1990, and 1993, I find thathiring more black and minority police officers increases crime rates, but thisapparently arises because lower hiring standards inoled in recruiting more minorityofficers reduces the quality of both new minority and new nonminority officers. Themost aderse effects of these hiring policies hae occurred in the areas most heailypopulated by blacks. There is no consistent eidence that crime rates rise when morewomen are hired, and this raises questions about whether norming tests or alteringtheir content to create equal pass rates is preferable. The article examines how thechanging composition of police departments affects such measures as the murder of

.and assaults against police officers. JEL J72, K14, H42

I. INTRODUCTION

Using preferential standards to aid minor-ity groups is frequently justified as rectifyingpast wrongs. Yet, since Richmond . Croson

1Co. 1989 , the U.S. Supreme Court hasheld that these preferences must pass the

*I would like to thank Stephen Bronars, TomCollingwood, Richard Epstein, Gertrud Fremling, EdGlaeser, Linda Gottfredson, Robert Hansen, Dan Ka-han, Larry Kenny, Dan Klerman, Bill Landes, StanLiebowitz, Scott Masten, Sam Peltzman, two very help-ful referees from this journal, and the participants inseminars at UCLA, the University of Chicago, CornellUniversity, George Mason University, Heritage Founda-tion, the NBER Law and Economics Summer Institute,University of Michigan, Michigan State University,SUNY Binghamton, University of Southern California,University of Washington, the American Law and Eco-nomics Association, the Western Economic AssociationMeetings, the Southern Economic Association meetings,and my students at the University of Chicago for theirhelpful comments. Stephen Bronars also deserves morethan normal thanks for the tremendous amount of workthat he has put in helping me put this data set together.John Whitley also provided valuable research assistance.Lott: Senior Research Scholar, Yale University School

of Law, New Haven, Conn., Phone 1-203-432-2366,Fax 1-203-432-8260, E-mail [email protected]

1. See also Epstein 1992, 42933 . Coate and Loury 1993 provide an important discussion on the costs andbenefits of affirmative action policies. They rigorouslylist out conditions under which these policies will breakdown negative stereotypes and those cases where theywill make them even worse.

2. Adarand overturned the decision in Metro Broad-casting, Inc. . Federal Communications Commission.

difficult strict scrutiny test and will be in-validated unless they promote a compellinggovernmental interest. Correcting societaldiscrimination was not viewed as a com-pelling interest. Remedial efforts to rectifypast discrimination will only be approved ifnarrowly tailored to correct specific instancesof discrimination. The question of what goalsconstitute a sufficiently compelling inter-est has never been clearly specified by theSupreme Court, though in a very closelydecided case it reversed its own past decisionthat Federal Communication Commission .FCC allocation of licenses by race is ac-ceptable to promote diversity in entertain-ment and news programming and appliedthese high standards of strict scrutiny andcompelling interest to federal building pro-jects.2

The standards set by the Supreme Courtin Richmond and Adarand were motivated

by the desire that The strict scrutiny testalso ensures that the means chosen fit this

ABBREVIATIONS

FBI: Federal Bureau of InvestigationFCC: Federal Communication CommissionLEMAS: Law Enforcement Management and

Administration Statistics

239Economic Inquiry .ISSN 0095-2583Vol. 38, No. 2, April 2000, 239277 Western Economic Association International

ECONOMIC INQUIRY240

compelling goal so closely that there is littleor no possibility that the motive for theclassification was illegitimate racial prejudiceor stereotype.3 One can hypothesize whatcompelling goals would meet these standardswhere there is little or no possibility thatan ulterior race-based motive might be thetrue motivation behind an affirmative actionrule, but the most obvious case would bewhen the racial preferences actually help tofurther the central purpose of the govern-mental agency.4 In the case of police, thismeans that minority police officers are beingemployed not because diversity is intrinsi-cally valued but because it is believed to helplower the crime rate.

The potential law enforcement advan-tages from multiracial or female officers seemobvious. Minority police officers may be moreeffective in minority areas simply becauseresidents could be more forthcoming aboutinformation that will lead to arrests and con-victions or because of the officers ability toserve as undercover agents. Trust is alsoimportant for other reasons, as reports ofriots erupting after white police officers haveshot a black man may attest.5 Officers from a

3. Adarand.4. A distinction must be drawn between two differ-

ent types of affirmative action programs: quotas andpreferential treatment. While preferential treatment al-ready must meet a very high threshold to be approved,the requirements are if anything even more difficult forquotas. It is doubtful that even a federal law establish-ing an affirmative action racial classification would be

upheld if the law used a racial quota system Nowak and Rotunda 1995, 695 .

5. For example, in 1996 riots errupted in St. Peters-burg Florida, after a white police officer shot and killedan 18-year-old black man driving a stolen car and inLeland, Mississippi, after a white police officer fatally

shot a black businessman named Aaron White Kis-simmee chief wants riot gear for police: The city shouldlearn from St. Petersburgs riots, John Sutphin said,Orlando Sentinel, Saturday, April 26, 1997, p. 1, andBartholomew Sullivan, Shooting death prior to Leland

riot ruled accidental, Commercial Appeal Memphis,. .Tenn , Friday, April 18, 1997, p. A15 . Further back In

1980, one of the worst recent U.S. race riots erupted inLiberty City and spread through Miami after an all-whitejury acquitted white police officers accused of killing a

black man Angus MacSwan, Drug gangs rule, chil-dren suffer in Miamis Liberty City, Reuters World

.Service, Friday, February 14, 1997 . Of course, probablythe worst recent riots occurred in 1992 after whitepolice officers were found not guilty in the Rodney Kingbeating. On the other hand, having a racially diversepolice department does not guarantee that these riotswill be prevented. The Los Angeles Police Departmentsshare of blacks very closely matched the citys.

community may also be better at understand-ing the behavior of criminals in those areasor even something as basic as understandingthe language of immigrants.6 In any event,police efforts to reduce crime are surely de-pendent on the help that they receive from

.the community Wilson 1983 .Rape victims or women abused by their

spouses plausibly find it easier to discuss thetraumatic events with women officers. With-out female officers, many attacks againstwomen may go undetectedthus loweringthe expected penalty from attacking womenand resulting in even more attacks. Policingis a rare case where the government outputis likely to be advanced by race- or sex-basedpreferences. Indeed, reducing reliance oncognitive tests for police entrance examina-tions has been justified with the motivationthat police departments cannot functioneffectively in minority neighborhoods whenvirtually all police officers are white males .Dunnette et. al. 1996 .

Another case might be education, where afrequently made claim is that a diverse stu-dent body better prepares students for adiverse world.7 These goals have also beenused to justify weighting applicants by raceor sex along with their test scores. By con-trast, how people use roads or machinesseems likely to be unrelated to the race ofthose who built them. Even the case of firedepartments, obtaining racial diversity seemstangential to the ultimate goal of extinguish-ing a fire.

Although the foregoing benefits are clear,there are countervailing factors that must betaken into account. Most important iswhether explicit race or sex preferences re-sult in less-capable individuals being hired.For women, this might result because of

6. Community leaders frequently claim that Wewant police who know the community. We want them to

spend time and become part of the community. Quotefrom Dennis L. Chinn, founder of the Asian Plaza

Youth Foundation, as reported by Phat X. Chiem 1995,B1 . The same article reports on the importance of

having bilingual officers. 7. For example, see Katyal 1995 and Keohane

1995 .

LOTT: DOES A HELPING HAND PUT OTHERS AT RISK? 241

less-stringent physical requirements.8 Slowerrunning speed might make it more difficultfor women to catch criminals.9 Weaker phys-ical strength might cause police departmentsto substitute away from single officer patrol

.units either foot or car and into units withtwo officers. If criminals believe that theyhave a greater chance of resisting arrest whenofficers are weaker, more assaults may becommitted by criminals against women of-ficers. In compensating for their weakerstrength, women may substitute into otherways of controlling criminalsthe most ob-vious method being guns. Although guns area great equalizer, they may not completely

8. Testing of the physical strengths of men andwomen public safety employees consistently finds largedifferences. These studies indicate that womens

strength rang es from 44 to 68% of mens in the upper body and 55 to 82% in the lower body Landy 1992,

.456 . The norming adopted by most police depart-ments for physical fitness tests creates equal probabili-

.ties for passing by men and women Flannery, 1995, 2 .The same types of rules are adopted by the militarywhere women recruits must run two miles in 18 min-utes, 54 seconds, which is three minutes slower than the

required time for men. Women must do 18 push-ups intwo minutes and 50 sit-ups in two minutes, while menmust do 42 push-ups and 52 sit-ups in the same time.Tom Collingwood, a consultant on physical testing stan-dards in Dallas, estimates that between 70% and 80% ofpolice departments explicitly use norming of physicalstandards in their hiring practices. However, he believesthat most of the departments that use objective stan-dards do not enforce these rules. Women who fail tomeet the absolute standards during academy trainingare unlikely to be failed out of the program. This beliefwas confirmed by conversations with other experts in

this area e.g., Mike Bahrke at Fitforce in Champaign,.Illinois . This creates a difficult problem for testing the

impact of norming physical standards because it impliesthat all cities really have the same standards whether

they explicitly claim so or not See also Bahrke and .Hoffman 1997 . Courts have also disallowed other types

of tests that produce differential pass rates betweenmen and women. For example, in a 1980 case involvingthe Philadelphia Police Department, the district courtruled that it was unlawful to discharge women whofailed to achieve a passing score on the firearms quali-

.fying test 499 F. Supp. 1196 .9. The New York City Police Department is said to

illustrate this point. The department abandoned allphysical screening of applicants in the 80s out fear oflawsuits by minority applicants and women. Some offi-cers hired under relaxed testing lack the strength to pullthe trigger on a gun, said Michael Julian, former NYPDchief of personnel. There are hundreds, if not thou-sands, of police officers on the streets today who, whena suspect runs from them, have no other option than tocall another cop, because they do not have the physical

ability to pursue them, Julian said Marzulli and Lewis, .1997, 7 .

offset differences in strength.10 Being lessable to fall back on their physical strength toprotect themselves when faced with a possi-ble attack, women may have to determinewhether they will fire their gun before thepossible attacker gets into physical contactwith them. If true, shorter reaction times riskresulting in more accidental shootings.

Although the U.S. Department of Justicestates that the appropriate testing proce-dures nearly eliminate disparate impact while

. 11improving merit hiring Gottfredson 1997 ,critics of affirmative action in policing arguethat these tests lower reliance on importantcognitive skills. According to a 1993 surveyof 23 large police and sheriff departmentsconducted for the Department of Justice

.and Nassau County, New York , the cogni-tive portion of police tests have been com-pletely removed in three cases, in an attemptto increase minority recruitment. Even theremaining 20 had reduced their emphasis oncognitive skills, with all the respondents indi-cating that adverse impact was consideredwhen determining the selection process .Dunnette et. al. 1993, 18 . Using this sur-vey to help justify its decision, Nassau Countyremoved all cognitive tests except for a read-ing comprehension test, which is gradedpass-fail and requires that applicants had toscore only as well as the bottom 1% ofcurrent police officers. The Louisiana StatePolice replaced a cognitive exam with a testthat initially contained six parts: three per-sonality, one biographical, and two cognitive,

10. A gun might not be as much of an equalizer forfemale officers as it is for women who use a gundefensively. Officers are frequently called on to havephysical contact with the criminals that they are pursu-ing, whereas women who use a gun defensively merelyuse the gun to keep a threatening person at bay.

11. Some testing consultants back up the Depart-ment of Justices position, and note the different waysthat questions can be worded which will hurt minorityapplicants. In particular, the use of double-negatives,homonyms, questions reflecting middle-class experi-ences, or complex sentence structures toward the end

of an exam all work to lower minority scores Wilson, .1996, A . President Clintons recent nominee as assis-

.tant attorney general for civil rights Bill Lann Leeargues that admission standards for schooling maynot disproportionately exclude members of any race,ethnicity, or gender unless justified by an educationalnecessity and no less discriminatory but equally effectivealternatives to the practice exist. Lee argued that University of California cannot demonstrate any edu-

cational necessity for standardized tests. Clint Bolick,A Vote for Lee Is a Vote for Preferences, Wall Street

.Journal October 27, 1997, p. A23 .

ECONOMIC INQUIRY242

but later threw out one of the cognitivesections to further reduce the impact on

. 12minorities Price 1997 . After spending$5.1 million to have consultants developunbiased exams, only to have minorities farepoorly again, Chicago moved to a heavilyweighted seniority system for promoting po-lice officers and a lottery system for hiring

. 13firefighters Spielman 1996, 16 . The De-partment of Justice has used legal action or

.so-called consent decrees to force policedepartments to adopt these rules.

Some academics have charged that thenew tests are consciously designed to worklittle better than simply picking applicants atrandom so that the pass rate is the same

across different racial groups Gottfredson .1997, 1996 . If minority applicants with lowcognitive skills are hired and if these skillspredict how good a police officer a candidatewould be, preferential treatment adverselyaffects the effectiveness of police depart-ments. Indeed, some shocking reports havebeen made about the importance of cogni-tive skills. Expressing concerns about thepoor English skills of new police recruits, a

Washington Post editorial 1993 claimed thatbetween 1986 and 1990, 311 of the 938murder cases the D.C. police brought to theU.S. attorneys officeroughly a thirdweredismissed. . . . One local prosecutor saysmany D.C cases were thrown out becauseprosecutors couldnt read or understand the

12. The Louisiana case provides a good example ofhow these cases work. As part of an agreement with theDepartment of Justice, the Louisiana State Police agreedto set aside $1 million to pay African Americans whofailed the test and hire new troopers from among quali-

fied African Americans who failed the test Shinkle,.1996, B1B2 . The test that was developed by the Coop-

erative Personnel Services, Inc., had been used in otherjurisdictions where it had been upheld as not discrimi-nating against minority applicants by a federal judge ina Torrance, California case. The Louisiana State Policedenied the allegations of discrimination, but agreed tosettle the case with the federal government to avoid theburdens of contested litigation. The Department ofJustice pointed to the disparate impact that the testwas having on blacks and that the test was not job-related. From August 1991 to May 1996, Of the 2,721white applicants who took the test, 66 percent passed; ofthe 1,293 African Americans who took the test, just 25percent passed.

13. The number of people participating in the lotteryis to be adjusted so as to ensure that enough minoritiesare found in the pool from which the new hires will be

.chosen Kass and OConnor, 1995, A1 . Other storieson the affirmative action process and its consequences

.in Chicago are provided by Martin 1997, A4 and .Oclander 1995, 22 .

14arrest reports written by the police . Still,some designers of the new tests defend thechanges: the validity of the cognitive ability

.test was not high Dunnette et. al. 1996 .The basic economics of these affirmative

action regulations is fairly straightforward.Voters value many objectives but face lim-ited resources. The question is whether vot-ers were previously discriminating againstcertain politically unfavored groups of poten-tial police officers at the cost of higher crimerates or whether affirmative action laws areforcing departments to accept higher crimerates as the cost of changing hiring policies.

This article examines the relationship be-tween the changing racial and gender com-position of police departments and the crimerate. As mentioned above, there are possiblyopposing forces, and the net effect is notobvious and it may not be the same for allcrime categories. For example, women policeofficers may deter rapists better than theydeter armed robbers.

Affirmative action can also affect crimerates in many different ways, for example,through changing the marginal quality ofnew officers or affecting which officers arepromoted and thus altering the incentives ofthe existing police force. If the critics of thenew rules are correct that the replacementsfor cognitive tests simply introduce random-ness into the hiring process, all new officers,and not simply the officers the new testswere designed to encourage, could be oflower quality. After first examining how courtorders altering the hiring and promotionprocess affect the crime rate, this articleseeks to provide a comprehensive picture forhow the changing demographic characteris-tics of police departments affect crime rates.The evidence will try to sort out the differ-ential impact of affirmative action on newhires and the existing police force as well astry to test whether the changes in effective-ness are due to the minority officers that arehired or the changing quality of all officers.

14. The Washington Post editorial went on to claimthat: Of the murder suspects who are indicted, manyend up being acquitted because of weak cases preparedby police. Washingtons Pretrial Services Agency reportsthat only 44 percent of the murder cases filed in 1990and closed by the first part of 1992 resulted in convic-tions.


Alternative explanations for the resultsare examined, such as whether any observedhigher crime rates merely reflect higher re-porting rates and whether police experiencelevels are affected by the altered hiring poli-cies. I also examine how changing genderand racial compositions alter how police de-partments operate and other measures ofeffectiveness such as arrest rates.

II. THE CHANGING COMPOSITION OFPOLICE DEPARTMENTS

During 1987, 1990, and 1993, the U.S.Department of Justice conducted a compre-hensive national survey of state and local lawenforcement agencies with 100 or more of-ficers, known as the Law EnforcementManagement and Administrative Statistics .LEMAS . My study focuses on city policedepartment data because they allow a moreprecise study of the relationship betweenhow police departments were organized andthe crime rate. By contrast, state and countydepartments are more difficult to investigate,because they have jurisdiction over largerbut overlapping areas.

.I separated the data into two sets: 1 the .entire Justice Department Survey and 2 a

subset in which demographic data are alsoavailable. The results that I report are con-siderably more significant statistically andimportant empirically when using the entireDepartment of Justice survey, yet I will focuson the subset with the demographic data,because changing demographics are relatedto both the changing hiring patterns by po-lice departments and crime.

Two characteristics stand out from thesurvey: city police departments vary greatlyin their racial and gender makeup, and therehave been large increases in the proportionof black and women officers. Tables I and IIillustrate these two points, with Table I illus-trating the distribution of the racial and gen-der composition of police departments andTable II examining the distribution of thechange in the composition. The first tableshows that although most departments haveno blacks, Hispanics, or Asians, the range islarge with the tenth and ninetieth percentiledepartments, respectively, employing 0% and18% blacks. The diversity for women officersis not quite as large, ranging from 0% at thetenth percentile to 14% at the ninetieth.

It is possible to subdivide these categorieseven further, but some of the racial and sexcategories have very small changes in the

TABLE 1The Race and Gender Composition of Police Departments

Distribution of Race and Gender Characteristics for Police Departments

10th Percentile Median 90th Percentile

Sample for Sample for Sample forWhich Yearly which Yearly which YearlyDemographic Demographic Demographic

Entire Estimates are Entire Estimates are Entire Estimates areSample available Sample available Sample available

% of the police 0% 0% 0% 0% .63% 1.5%force that isAsian Pacific% of the police 0% 1.3% 0% 7.8% 18.3% 26%force that is black% of the police 0% 0% 0% 2.1% 8% 14.5%force that isHispanic% of the police 72% 65% 98.5% 85.5% 100% 96.5%force that is white% of the police 86% 86% 97% 91.7% 100% 96.6%force that is male

Notes: The entire sample has 4,158 city year observations for 1987, 1990 and 1993. The sample for which yearlydemographic estimates are available from the Current Population Survey Contains 664 cityyear observations: 204Police Departments in 1987, 240 in 1990, 220 in 1993.

ECONOMIC INQUIRY244

TABLE IIThe Changing Racial Composition of Police Departments

Change at the Change at the10th Percentile Change at the Median 90th Percentile

No Consent Consent No Consent Consent No Consent ConsentDecree Decree Decree Decree Decree Decree

A. Changes in the Racial Composition of Police Departments With and Without Consent DecreesThat Occurred from 1987 to 1993.Percentage Point .6 .23 0 .2 1.0 1.8Change in the % ofthe Police Force that isAsian PacificPercentage Point 6.8 .2 .73 3.2 6.0 18.2Change in the % ofthe Police Force that isBlackPercentage Point .14 .12 .7 1.1 5.4 7.0Change in the % ofthe Police Force that isHispanicPercentage Point 11.5 21 2.3 5.9 .98 .7Change in the % ofthe Police Force that isWhiteB. Changes in the Sex Composition of Police Departments With and Without Consent Decrees From 1987 to 1993Percentage Point 5.3 6.3 1.1 2.8 6.4 0Change in the % of thePolice Force that isMale

Notes: Panel A again breaks down the sample on the basis of the complete LEMAS Survey and those cities forwhich information on changing city demographics are available. The table shows the change in the racial and gendercompositions of police departments. The entire sample contains 333 cities without consent decrees for whichinformation is available for the same city for all three years. Twenty one cities with consent decrees meet this criteria.By contrast, the restricted sample that is used for the regressions contains 163 and 19 cities in these two categories,though it provides very similar results.

Panel B again breaks down the sample on the basis of the complete LEMAS Survey and those cities for whichinformation on changing city demographics are available. The entire sample contains 343 cities without consentdecrees for which information is available for the same city for all three years. Fourteen cities with consent decreesmeet this criteria. By contrast, the restricted sample that is used for the regressions contains 163 and 19 cities in thesetwo categories, though it produces very similar results.

total number of officers. In my restrictedsample, 189 cities had detailed employmentdata within each race category by sex forboth 1987 and 1990. These cities employed

.155,071 or 40% of the 387,534 sworn full-time officers employed by local governmentsin 1990. As examples of the small number ofofficers in some of these subgroups, thenumber of male American Indian officersbetween 1987 and 1990 grew from 280 to 378officers; for female American Indians, thechange was from 47 to 91; and for femaleAsian Americans, 83 to 203. Even Hispanicfemales, the next-largest category, saw anincrease of only 378 officers. The number ofmale white officers, the only category to de-cline, fell by 6,912.

The second table illustrates the differentrates of changes over time as well as theimpact of the consent decrees which theDepartment of Justice entered into with citypolice departments regarding a citys hiringand promotion practices.15 Past work hasstudied the effect of these decrees on hiringof black men and found that indeed they do

.have an impact Lewis 1989 . The Depart-ment of Justices Civil Rights Division pro-vided information on both racial andorgender-based consent decrees over the pe-

15. These decrees are contracts that the Departmentof Justice and cities have signed that have been ap-proved by a court, which obligate the city to act incertain ways in the future.


riod from 1972 to 1994: 19 of these 189 citieswere covered by consent decrees during the198793 period, though only three of thesecities had consent decrees that were imposedas late as the end of 1987. The 19 cities wereBirmingham, Ala.; Montgomery, Ala.; LosAngles, Calif.; San Francisco, Calif.; Ft.Lauderdale, Fla.; Pompano Beach, Fla.; Mi-ami, Fla.; Tallahassee, Fla.; Macon, Ga.;Chicago, Ill.; Indianapolis, Ind.; Jackson,Miss.; Omaha, Neb.; Las Vegas, N.V.; Syra-cuse, N. Y. Cincinnati, Ohio.; Philadelphia,Penn.; Memphis, Tenn.; and Milwaukee,Wisc.

Many cities are adopting affirmative ac-tion rules on their own either because oftheir own support for such rules or becauseof the threat of Justice Department actions.Any examination of consent decrees is thuslikely to underestimate the impact of suchpolicies. Yet, consent decrees appear to haveclear impacts for both blacks and women.The median change in the percent of blackpolice officers was 2.5 percentage points morein cities with consent decrees than thosewithout them, and for women the medianincrease was 1.7 percentage points. Thesemay seem like small changes in the share ofpolice employment going to these groups,but compared to the median percent of blackand women officers over this seven-year pe-riod, these changes represent at least a 57%increase over past employment practices.

Finally, despite the large difference insample sizes between the entire sample andthe restricted one, both sets experienced re-markably similar changes in types of officersemployed during this seven-year period. Thissimilarity occurs despite the cities in thesmaller sample averaging about 40% morepeople.

III. EXPLAINING CHANGING CRIME RATESAS A FUNCTION OF THE RACIAL AND

GENDER COMPOSITION OF POLICEDEPARTMENTS

The Direct Impact of Consent Decrees

The FBIs Uniform Crime Report allowsus to study violent and property crimes, with

seven primary crime categories murder,rape, robbery, aggravated assault, burglary,

.larceny, and motor vehicle theft , and tenother subcategories manslaughter, forcible

rape, attempted rape, gun robbery, knife

robbery, other robbery, strong-arm robbery,assault with a gun, assault with a knife, and

.other assault . The results from most of thesesubcategories will not be reported, becausethey differ little from the results shown forthe primary categories. Data on arrest ratesfor these broader categories as well as thecity populations were obtained directly fromthe FBI.

The Current Population Survey was usedto determine the changing demographicmakeup of cities over the 198793 period.The percent of the population in differentdemographic categories was broken down by

age less than 30 years of age, 3054 years of. age, and 55 and older , race black, white,

. .and other , and sex male and female , thusyielding 18 demographic categories. This sur-vey also provided information on the averageweekly wage and the unemployment rate.The National Conference of Black Mayorsprovided me with copies of their entire na-tional roster by year so that the race of acitys mayor could be identified. Finally, theLEMAS survey provides information on theracial and gender composition of police de-partments, as well as on the per capita num-ber of sworn police officers, and other de-partmental characteristics. The means andstandard deviations for these variables areshown in the appendix.

Table III shows simple preliminary regres-sions that use a simple time trend for thenumber of years after a consent decree hasbeen imposed and a similar time trend forthe years before the decree to pick upchanges in before and after trends in crimerates. To do this, I used yearly violent andproperty crime data for 198594 for 495cities, a longer period than is available forthe LEMAS survey. Data prior to 1985 wasnot included because of severe problems withthe consistency between 1984 and 1985 inthe city-level crime data. Two sets of fixedeffects were used for these simple regres-sions: city and year fixed effects and cityfixed effects along with separate year fixedeffects for each state to control for any indi-vidual state trends. These regressions useordinary least squares weighted by city popu-lation. The results for both violent and prop-erty crime rates imply that crime rates weredeclining in cities before consent decreeswere imposed and were rising thereafter.Violent crimes were rising after the consent

ECONOMIC INQUIRY246

TABLE IIIChanges in Crime Rates for Cities with and without Consent Decrees for the Period 198594;

Using Only Fixed Effects

Crime Rates Per 100,000 People

Time Trend for Time Trend forYears before Consent Years after Consent

Decree Went Into Decree Went Into F-test( ( ( )Effect negative values Effect positive values Prob F

imply that crime imply that crime that beforewas falling until was rising after and afterthe decree went the decree went time trends Adjacent No. of

2) )into effect into effect are different R Observations

Controlling for City and Year Fixed EffectsViolent Crime Rate 138.6 126.1 36.35 .7939 4,947

.5.3% 4.8% .0000 . .4.204 11.433

Property Crime Rate 593.4 172.2 57.37 .7719 4,947 .9.4% 2.7% .0000

. .6.257 9.346Controlling for City Fixed Effects and Separate Year Fixed Effects for Each StateViolent Crime Rate 60.85 86.05 7.10 .8738 4,947

.2.3% 3.3% .0078 . .1.195 7.901

Property Crime Rate 464.0 133.76 50.94 .8845 4,947 .7.4% 2.1% .0000

. .5.998 8.085

Notes: The first number is the annual change in crimes per 100,000 people, while the second number is the changeas a percent of the mean crime rate. Absolute t-statistics are shown in parentheses. The regressions use weightedleast squares.

decrees by at least 3.3% per year, and forproperty crimes it was at least 2.1% per year.The differences in trends are all statisticallysignificant at the .01 level.

Given the significant declines in precon-sent decree crime rates, the results for threeof the four regressions raise the questionabout whether the decrees just happen to beimposed when the crime rates were at theirlow ebb and the post-decree increase is sim-ply a result of mean reversion. At least forthe cities with new consent decrees imposedduring 1987, the increases in violent crimeduring the period studied are 2.4 to 3.7 timeslarger than preceeding declines and thus ex-ceed any increase that could simply be at-tributed to mean reversion.16 The evidence

16. I tried a regression that predicted which citieswould have consent decrees imposed on them. The mostimportant factors were city size, whether the city wasthe largest in a state, and the type of administration.Republican presidential administrations tended to im-pose consent decrees on relatively Democratic states,whereas Democrat presidential administrations tendedto impose consent decrees on relatively Republicanstates.

is not clear cut for property crimes, wherethe declines and increases are of approxi-mately equal size.

Using the smaller sample that matchesthe LEMAS survey and just the time trendfor years after the imposition of the consentdecree produces similar, though smaller andless statistically significant increases in crime.Controlling for changing city-level demo-graphics as well as the average weekly wage,unemployment, per capita number of policeofficers, city population, and populationsquared, and city and year fixed effects im-

plies: violent crime rises by 1.9% t-statistic. 2.16 and property crime by 2.1% t-statis-

.tic 2.99 for each additional year the con-sent decree is in effect.

Figure 1 reports this same data slightlydifferently by including separate dummyvariables for the years before and after theimposition of the consent decree. The yearsincluded range from three years before thedecree to year 4 afterward, with anotherdummy variable that equals one for years 5or more after the decree. The estimates use


FIGURE 1Using Year Dummies for the Years

before and after the Implementation ofa Consent Decree

the full sample and correspond to thoseshown in Table III that account for city fixedeffects as well as separate year fixed effectsfor each state. The pattern is similar to thatimplied by the table with crime rates fallingimmediately before the consent decree is

imposed and rising after that. Recently, ex-tending this data set back to 1977 and ex-panding the number of years studied by seven

produced similar results Lott 2000, chapter. .9 .

An Initial Assessment of How Consent DecreesAffect Crime through the Typeof Officers Hired

Consent decrees change hiring policies,and changing hiring policies may affect thecrime rate. The consent decrees favor somegroups more than others, but the link be-tween these group-specific changes and the

crime rate are unclear. Even if the new hir-ing procedures introduce randomness, thedistribution of skills is not necessarily thesame across potential applicants in all groups.To make matters more complicated, the typesof officers hired may depend on the crimerate. For example, if departments hired mi-norities because of growing crime problemsin minority areas or hired women because ofgreater crimes against women, simple ordi-nary least squares estimates risk improperlyblaming some of the higher crime rates onthe new police who were hired to help solvethe problem.17 To explain this problem dif-ferently, crime rates may have risen eventhough a city hired black officers, and if theyhad hired white officers who were less capa-ble of policing minority areas, the crime ratecould have risen by even more. Unfortu-nately, an alternative explanation exists:higher crime rates may signal less concern bycity governments about crime and thus agreater willingness to indulge other objec-tives when hiring police officers.18

The opposite relationship between crimerates and hiring practices is also possible.Additional law enforcement efforts have agreater effect on crime in high-crime areas .Lott and Mustard 1997, 28, 29 . If affir-mative action actually increases crime, high-crime areas would find it more costly toengage in affirmative action, and thus, every-thing else equal, one suspects that they wouldengage in less such hiring. Failure to controlfor why the particular composition of policeofficers was chosen would underestimate thenegative impact from this policy.

To guard against this problem, I initiallyemploy two-stage least squares where thefirst equation attempts to explain the propor-

tion of black, minority black, Hispanic, and.American Indian , or male officers employed

17. While the hiring of minority officers is motivatedby the desire to assign these new officers to minorityneighborhoods, the legal prohibition against giving of-ficers assignments based on their race require that anynew minority officers be evenly distributed across dis-tricts. It is very easy for minority officers to bringdiscrimination suits if they feel that they are beingdisproportionately assigned to more dangerous neigh-borhoods. Black officers have no more desire than whiteofficers to be assigned to dangerous high crime areas.For another perspective with respect to New York City,

.see Fyfe 1981 .18. There are also questions about whether some

officers have stronger preferences for policing certaintypes of communities based on their level of crime.

ECONOMIC INQUIRY248

by a city. As discussed in Section II, I expectthat the imposition of a consent decree and,particularly, the length of time that the de-cree has been in effect to serve as the instru-ments and help explain the levels of minorityor female employment, depending on whattype of employment the consent decree dealswith. The number of years that a consentdecree has been in effect is an excellentinstrument since it is extremely unlikely thatthe causation runs from future crime rates tothe number of years since a consent decreehas been entered. I also account for thedemographic composition of the citys popu-lation; the average weekly wage and unem-ployment rate; whether its mayor was black;the citys population and population squared;and the per capita number of sworn policeofficers. The second equation that explainseach one of the individual crime rates in-cluded all the variables except for whetherthere was the consent decree and the mayorsrace. Weighted least squares, where the esti-mates were weighted by city population, wasused to deal with heteroskedasticity.19

The coefficients on the percent of thepolice force that is black, minority, or malein the second regression are thus adding theimpact on the crime rate of the consentdecree together with that particular groupbeing studied. I will disaggregate these twoeffects later when I report the reduced formregressions in Table VIII and Appendix 2.Because the instruments that I have for racialor gender hiring consent decrees are veryheavily correlated, the impacts of the racialand gender compositions of police depart-ments are initially estimated separately forthese two-stage least squares regressions.

Admittedly, there are many location-specific and year-specific differences in crimerates that are not captured by the variablescontrolling for demographic, income, andpopulation differences. One simple way ofdealing with this is the use of location andtime fixed effects, where a separate dummyvariable is used for each city and year. How-ever, this approach also has its drawbacks:although it may correctly measure left-outvariables, it may also cause us to falsely

19. Similar estimates are produced if unweightedestimates are employed, but these data exhibit definiteheteroskedasticity, with the smaller cities reporting amuch greater variation in crime rates over time.

attribute some of the impact of changes inour in our other variables for example, the

impact of changing racial or gender composi-.tion of police departments to these fixed

effects. Nevertheless, all the regressions re-port either city and year fixed effects orcounty fixed effects with separate year fixedeffects for each state.20

As an example, the two-stage least squaresestimates examining the percentage of thepolice force that is black with city and yearfixed effects take the following form:

.1 % Police Force That Is Black g Consent Decree in

Effect, Number of Years Decree inEffect, Dummy for Whether Mayor IsBlack, Per Capita Number ofSworn Officers, Citys DemographicComposition, Population andPopulation Squared, AverageWeekly Wage, Unemployment Rate,Fixed Year and City

.Effects . .2 In Crime Rate

f % Police Force That Is Black,Per Capita Number of Sworn Officers,Citys Demographic Composition,Population and Population Squared,Average Weekly Wage, Unemployment

.Rate, Fixed Year and City Effects

The results from the second equation arereported in Table IV separated out by thetype of fixed effects employed. All crime

20. I tried three different types of location fixedeffects: city, county, and state. Generally, using thebroader measures of location produced estimates thatagreed in sign with the city fixed effects, but the esti-mates were larger and more statistically significant. Todeal with possible state-level trends in laws, I also triedallowing a separate fixed effect for each state for eachyear, though when combined with county or city fixedeffects this dramatically reduces the degrees of freedomin each regression. Only the time-varying state fixedeffects are reported with the county fixed effects be-cause none of the estimates on any of the focus orcontrol coefficients was statistically significant with cityfixed effects.


rates are in natural logs, where .1 is added tozero values before the natural log is taken.With the exception of manslaughter, aggra-vated assault, and motor vehicle theft, anincrease in the percentage of a police forcethat is black is consistently associated withsignificant increases in crime. The effect is solarge that 18 of the specifications imply thata one standard deviation change in the per-cent of the police force that is black in-creases the corresponding crime rates by at

least 5% of its mean value see the percent-.ages listed next to the coefficients . The ef-

fects are dramatic no matter how one exam-ines these estimates. For example, increasingblack officers share by one percentage pointincreases property crimes by 4%, and thesame increase raises the murder rate by 1.9%and overall violent crime by 4.8%. As therelative median increase in black officersshare of police departments over this seven-year period because of consent decrees was2.5 pge points, I conclude that if nothing elsehad changed, the average citys murder ratewould have risen by 4.7%.21

One point should be made very clear atthis point. We are talking about the impacton crime of hiring additional blacks, manyof whom would not have been hired withoutthe consent decree. As mentioned in theintroduction, changes in testing that are usedto encourage hiring more minorities can ex-plain why these blacks are not of the samequality as previously hired blacks. It can stillbe true that qualified black officers are moreeffective but that the new less-qualified of-

21. One concern raised to me by Ed Glaeser iswhether the results are being driven solely by time-serieschanges in the data and whether these results are con-sistent across the years being studied. To test this, Ireran the regressions shown in Table IV with fixed stateeffects separately on the data for each of the threedifferent years. For blacks the coefficient signs are simi-lar to those already reported, though the results forthese smaller subsets of data are not always statisticallysignificant. The results for 1987 are as follows: for

.violent crimes the coefficient is 4.3 t-stat 1.822 ; .property crimes, 3.34 t-stat 1.811 ; and murder 5.52

.t-stat 1.254 . The results for 1990 are as follows: for .violent crimes the coefficient is 3.08 t-stat 1.071 ;

.property crimes, 5.46 t-stat 2.358 ; and murder, 3.45 .t-stat 0.872 . The results for 1993 are as follows: for

.violent crimes the coefficient is 3.025 t-stat 1.900 ; .property crimes, 2.509 t-stat 2.847 ; and murder,

.1.711 t-stat 0.815 . Similar results are also producedfor the percentage male and the percentage minorityspecifications.

ficers are associated with more crime. Thelarge impact suggests that more than just thequality of new minority recruits or new mi-nority promotions is affected. Changing teststo employ a greater percentage of blacks canmake it more difficult to screen out lower-quality candidates generally, including whitesand other racial groups. Independent of theconsent decree, the size of the change inblack employment may thus proxy forchanges in the level of standards used to hireemployees in general. Similarly, changingpromotion rules that favor seniority overachievement can affect morale and incen-tives across all categories of police officers.

For the next set of regressions, blacks,Hispanics, and American Indians were com-bined to represent the share of minorities ina department. The groups included in theminority classification was decided by using aseries of reduced-form equations where Itested to see whether the predicted impactof the different racial groups were statisti-cally different from each other. Generally,the coefficients for blacks, Hispanics, andAmerican Indians were not statistically dif-ferent from each other, and the whites andAsians usually fit together in a separategroup.22 More precisely, whites and Asians

22. More precisely, when the omitted group in the .reduced form regression represented by the intercept

is Hispanics, the probability that the coefficients forwhites and Asians are statistically significantly differentfrom each other at the following levels as: for violentcrimes is 34%; property crimes, 73%; murder, 39%;manslaughter, 41%; rape, 78%; forcible rape, 79%; at-tempted rape, 5.6%; robbery, 31%; gun robbery, 15%;knife robbery, 4.9%; other robbery, 0%; strong-armrobbery, 79%; assault, 66%; burglary, 77%; larceny,68%; and motor vehicle theft, 98%. The probability thatthe coefficients for blacks and Hispanics are statisticallysignificantly different from each other is: for violentcrimes it is 36%; property crimes, 51%; murder, 14%;manslaughter, 37%; rape, 77%; forcible rape, 56%; at-tempted rape, 73%; robbery, 1.6%; gun robbery, 1.5%;knife robbery, 46%; other robbery, 56%; strong-armrobbery, 74%; assault, 99%; burglary, 3%; larceny, 60%;and motor vehicle theft, 22%. The probability that thecoefficients for blacks and whites are statistically signif-icantly different from each other is: for violent crimes itis 5%; property crimes, 4.5%; murder, .12%; manslaugh-ter, .01%; rape, 43%; forcible rape, 39%; attemptedrape, 57%; robbery, .16%; gun robbery, .37%; kniferobbery, 58%; other robbery, 8%; strong-arm robbery,62%; assault, 3%; burglary, 6%; larceny, 34%; andmotor vehicle theft, .17%. State fixed effects were usedfor these estimates. A related set of regressions isreported in Section VII, though these regressions do nothave all these categories included at the same time.

ECONOMIC INQUIRY250

TA

BL

EIV

Usi

ngT

wo-

Stag

eL

east

Squa

res

toT

ake

Into

Acc

ount

the

Impa

ctT

hatC

onse

ntD

ecre

esan

dth

ePr

esen

ceof

aB

lack

May

orH

ave

onth

eC

ompo

sitio

nof

Polic

eD

epar

tmen

ts

%of

the

Pol

ice

For

ceth

atis

(%

ofth

eP

olic

eF

orce

Min

orit

yB

lack

,His

pani

c,%

ofth

eP

olic

eF

orce

)th

atis

Bla

ckan

dA

mer

ican

Indi

anth

atis

Mal

e

Cou

nty

Fix

edC

ount

yF

ixed

Cou

nty

Fix

edE

ffec

tsan

dE

ffec

tsan

dE

ffec

tsan

dSe

para

teY

ear

Sepa

rate

Yea

rSe

para

teY

ear

Fix

edE

ffec

tsC

ity

and

Yea

rF

ixed

Eff

ects

Cit

yan

dY

ear

Cit

yan

dY

ear

Fix

edE

ffec

tsfo

rea

chSt

ate

Fix

edE

ffec

tsfo

rea

chSt

ate

Fix

edE

ffec

tsF

ixed

Eff

ects

for

each

Stat

e(

)(

)(

)(

)(

)(

)1

23

45

6

coef

fici

ent

%of

coef

fici

ent

%of

coef

fici

ent

%of

coef

fici

ent

%of

Coe

ffici

ent

%of

Coe

ffici

ent

%of

Cri

me

and

t-m

ean

and

t-m

ean

and

t-m

ean

and

t-m

ean

and

t-m

ean

and

t-m

ean

Rat

esSt

atis

tic

expl

aine

dSt

atis

tic

expl

aine

dSt

atis

tic

expl

aine

dSt

atis

tic

expl

aine

dSt

atis

tic

expl

aine

dSt

atis

tic

expl

aine

d

Vio

lent

4.79

8%7.

1412

%3.

016%

6.06

813

%

8.19

9%

10.2

812

%

.

.

.

.

.

.cr

ime

2.38

3.57

2.52

3.85

1.97

1.67

Prop

erty

4.27

5%3.

965%

2.58

4%3.

205%

6.

425%

6.

576%

.

.

.

.

.

.

crim

e2.

513.

082.

593.

181.

181.

69M

urde

r1.

8711

%9.

4356

%1.

4611

%8.

0361

%

6.41

26%

8.

2633

%

.

.

.

.

.

.1.

642.

881.

823.

200.

961.

18M

an6.

3940

%6.

3339

%2.

5020

%5.

8446

%

13.0

655

%

3.69

15%

.

.

.

.

.

.

slau

ghte

r0.

781.

110.

511.

310.

860.

35R

ape

13.2

741

%4.

0712

%5.

9723

%3.

1412

%

8.98

19%

6.

6714

%

.

.

.

.

.

.1.

781.

981.

421.

960.

731.

37T

otal

7.86

16%

9.55

20%

4.57

12%

7.67

20%

11

.06

15%

13

.53

19%

.

.

.

.

.

.

robb

ery

3.07

4.08

3.18

4.25

1.22

1.78

Tot

al2.

104%

6.06

11%

1.83

4%5.

5513

%

6.64

8%

8.54

10%

.

.

.

.

.

.

assa

ult

0.90

2.71

1.30

3.13

1.07

1.44

Bur

glar

y7.

0511

%6.

199%

3.99

8%4.

999

10%

9.

239%

10

.45

11%

.

.

.

.

.

.

3.12

3.88

3.23

3.95

1.22

1.79

Lar

ceny

4.83

6%3.

535%

3.24

5%3.

055%

8.

688%

6.

386%

.

.

.

.

.

.

2.76

2.77

3.10

3.01

1.24

1.69

Mot

or

0.81

1%6.

3011

%

1.11

2%4.

6610

%3.

794%

7.

088%

.

.

.

.

.

.

vehi

cle

0.34

2.69

0.76

2.57

0.81

1.34

thef

tF

elon

ious

4.

2620

%5.

8228

%

4.39

27%

3.84

622

%11

.96

29%

10

.31

35%

.

.

.

.

.

.

killi

ngs

of1.

222.

231.

992.

141.

311.

58po

lice

offic

ers


TA

BL

EIV

Usi

ngT

wo-

Stag

eL

east

Squa

res

toT

ake

Into

Acc

ount

the

Impa

ctT

hatC

onse

ntD

ecre

esan

dth

ePr

esen

ceof

aB

lack

May

orH

ave

onth

eC

ompo

sitio

nof

Polic

eD

epar

tmen

ts

%of

the

Pol

ice

For

ceth

atis

(%

ofth

eP

olic

eF

orce

Min

orit

yB

lack

,His

pani

c,%

ofth

eP

olic

eF

orce

)th

atis

Bla

ckan

dA

mer

ican

Indi

anth

atis

Mal

e

Cou

nty

Fix

edC

ount

yF

ixed

Cou

nty

Fix

edE

ffec

tsan

dE

ffec

tsan

dE

ffec

tsan

dSe

para

teY

ear

Sepa

rate

Yea

rSe

para

teY

ear

Fix

edE

ffec

tsC

ity

and

Yea

rF

ixed

Eff

ects

Cit

yan

dY

ear

Cit

yan

dY

ear

Fix

edE

ffec

tsfo

rea

chSt

ate

Fix

edE

ffec

tsfo

rea

chSt

ate

Fix

edE

ffec

tsF

ixed

Eff

ects

for

each

Stat

e(

)(

)(

)(

)(

)(

)1

23

45

6

coef

fici

ent

%of

coef

fici

ent

%of

coef

fici

ent

%of

coef

fici

ent

%of

Coe

ffici

ent

%of

Coe

ffici

ent

%of

Cri

me

and

t-m

ean

and

t-m

ean

and

t-m

ean

and

t-m

ean

and

t-m

ean

and

t-m

ean

Rat

esSt

atis

tic

expl

aine

dSt

atis

tic

expl

aine

dSt

atis

tic

expl

aine

dSt

atis

tic

expl

aine

dSt

atis

tic

expl

aine

dSt

atis

tic

expl

aine

d

Ass

aults

on78

.14

128%

.823

1%61

.84

131%

6.35

13%

18

.716

%

15.3

118

%

.

.

.

.

.

.po

lice

3.49

.095

4.47

1.00

1.89

1.69

offic

ers

Acc

iden

tal

.5

433%

.315

1%

.097

1%

.179

1%

2.00

5%

4.80

16%

.

.

.

.

.

.

deat

hsof

0.29

.242

.083

.190

.742

1.60

polic

eof

ficer

s

.

Not

es:

The

regr

essi

ones

timat

esfo

req

uatio

n2

for

the

two-

stag

ele

ast

squa

res

that

are

repo

rted

abov

eac

coun

tfo

rth

esa

me

vari

able

sco

ntro

lled

for

inth

efir

st-s

tage

regr

essi

onex

cept

for

the

cons

ent

decr

ee,t

henu

mbe

rof

year

ssi

nce

the

cons

ent

decr

eew

ent

into

effe

ct,a

ndw

heth

erth

eci

tys

May

oris

blac

k,w

hich

wer

ein

stea

din

clud

edin

th

efir

st-s

tage

regr

essi

on.T

heot

her

vari

able

sco

ntro

lled

for

wer

eth

epe

rcen

tage

ofth

epo

pula

tion

indi

ffer

ent

dem

ogra

phic

cate

gori

esth

atw

ere

brok

endo

wn

byag

ele

ssth

an.

.

.

30ye

ars

ofag

e,30

54ye

ars

ofag

e,an

d55

and

olde

r,r

ace

blac

k,w

hite

,and

othe

r,a

ndse

xm

ale

and

fem

ale

soth

atth

isin

form

atio

nw

asav

aila

ble

for

18ca

tego

ries

,th

eav

erag

ew

eekl

yw

age,

the

unem

ploy

men

trat

e,an

dci

typo

pula

tion

and

popu

latio

nsq

uare

d.T

heab

solu

tet-s

tatis

tics

are

show

nin

the

pare

nthe

ses

belo

wth

eco

effic

ient

estim

ate,

with

the

perc

ent

ofth

een

doge

nous

vari

able

sm

ean

that

can

beex

plai

ned

bya

one-

stan

dard

-dev

iatio

nch

ange

inth

eex

ogen

ous

vari

able

show

nin

the

adja

cent

colu

mn.

All

regr

essi

ons

use

wei

ghte

dle

ast

squa

res

whe

reth

eva

riab

les

are

wei

ghte

dby

the

city

popu

latio

n.F

orth

epe

rcen

tage

ofof

ficer

sth

atar

ebl

ack

orm

inor

ity,t

hesa

mpl

esi

zes

is64

1an

dco

vers

the

year

s19

87,1

990,

and

1993

.For

the

perc

ent

ofof

ficer

sth

atar

em

ale

over

that

peri

od,t

hesa

mpl

esi

zeis

648.

ECONOMIC INQUIRY252

had different effects on crime in only threeof the 19 crime categories when the broad-

.est set of categories was used , whereasblacks and Hispanics were statistically dif-ferent in five cases. In the three cases wherewhites and Asians differed in their impact on

crime attempted rape, knife robbery, and.other robbery , Asians had a greater deter-

rent impact on crime. Hiring additional His-panics and American Indians did not tend toincrease crime by the magnitude shown byhiring additional black officers.

The two-stage least square estimates con-tinue to confirm this pattern. Putting to-gether blacks, Hispanics, and American Indi-ans continued to produce very similar, thoughsmaller, results compared to what I foundfor blacks alone. The minority portion of apolice force in column 3 explains about75%80% as much of the percent of themean violent and property crimes as did theregressions in column 1 for the percent ofthe police force that is black. Nineteen ofthe 20 crime regressions imply that increas-ing the percentage share of minorities in adepartment increase crime, and the relation-ship is statistically significant for three-quarters of the estimates.

The last two columns in Table IV implythat increasing the share of males in thepolice force decreases crime in 19 of the 20specifications shown, though the aggregateproperty crime category implies a statisticallysignificant relationship only for the timevarying state fixed effects that include thecounty fixed effects. The specifications formurder, manslaughter, and rape provide nosignificant evidence that increasing womensshare of the police force increases thesecrimes. Using either simple county or stateand year fixed effects produces a much moreconsistent negative relationship betweenhigher males shares of the police force andcrime.23

23. Limits on the number of variables that could behandled using two-stage least-squares with STATA re-stricted the regressions on the larger data set to the

state fixed effects specifications. This is the data setthat was not restricted to those cities for which demo-

.graphic data was available. Estimates using these dataremain similar to those already reported in Table IV.The sample size for this larger data set is 1,015 observa-tions for the regressions explaining the percentage ofthe police force that is black or minority and 1,026 forthe percentage of the police force that is male.

Table V reports some of the first equationresults from the two-stage least squares esti-mation of the percent of the police forcethat is black, minority, or male used in theregressions analyzing violent crime with cityfixed effects. The results imply that for theracial components, the number of years thata consent decree has been in effect dramati-cally increases the percentage of minoritiesin police forces. Every 10 years after theconsent decree goes into effect increases thenumber of blacks by another 4.1 percentagepoints and minorities by 4.8 percentagepoints. These results are comparable in mag-nitude with those shown in Table II. I alsotried these estimates with a squared term forthe number of years that the consent decreehad been in effect, but including this did notnoticeably alter the results. One city in thesample had consent decrees as long as 21

years with both the sample median and the.mean being about 10 years , and the esti-

mates indicate the percentage of the policeforce that is black is still rising at that time.

Interestingly, the election of a black mayordoes not appear to significantly change thenumber of minority police officers, with thecorresponding coefficient even being nega-tive is the minorities specification. There isalso surprisingly little relationship betweenpast crime rates and the composition of thepolice force, and the t-statistics are quitesmall. Only past violent crimes imply moreblacks on a police force with a t-statisticeven greater than 1. The Hausman endo-geneity test indicates that the numberof years a consent decree is in effect is avalid instrument for the black and minorityregressions.

A possible concern with these results isthat the consent decree not only directlyaffects the number of minorities or womenwho are hired but may also implicitly signalconcerns about future crime rates. If oneexpects that higher crime rates can be bestcombated with more minority police officers,there is also the concern that this motivatedthe adoption of the consent decree. Whilethis is possible, it is not clear why the De-partment of Justice has better informationon a particular citys future crime rate thanthe city itself. In any case, as a check, Ireestimated the regressions in Table IV by


TA

BL

EV

Exp

lain

ing

the

Perc

enta

geof

the

Polic

eF

orce

that

isB

lack

,Min

ority

,or

Mal

e:

.T

heR

egre

ssio

nsfo

rE

quat

ion

2fo

rth

eC

ityan

dY

ear

Fix

edE

ffec

tsR

egre

ssio

nsin

Tab

leIV

.

Con

sent

Dec

ree

Dum

my

( adi

ffer

ent

dum

my

Num

ber

ofY

ears

that

isus

edfo

rw

heth

erC

onse

ntD

ecre

eIs

inE

ffec

t(

the

cons

ent

decr

eein

volv

esco

nsen

tde

cree

sin

volv

ing

race

sex

orra

ceca

ses

and

sex

are

sepa

rate

dou

t(

and

ism

atch

edw

ith

and

are

mat

ched

wit

hth

eC

ity

Has

lnV

iole

ntC

rim

eth

eap

prop

riat

ere

gres

sion

appr

opri

ate

regr

essi

onli

sted

aB

lack

Rat

epe

r2

))

)li

sted

belo

wbe

low

May

or10

0,00

0pe

ople

F-S

tati

stic

Adj

acen

tR

%of

the

polic

e.0

14.0

041

.0

0028

.0

100

53.4

6.9

581

.

.

.

.

forc

eth

atis

1.00

33.

289

.049

1.73

1bl

ack

%of

the

polic

e.0

556

.004

8

.001

8

.008

940

.26

.944

6

.

.

.

.fo

rce

that

is2.

150

2.89

90.

237

1.14

5

min

ority

blac

k,hi

span

ic,a

ndo

ram

eric

anin

dian

%of

the

polic

e

.031

.0

015

.0

026

.0

061

2.15

.330

7

.

.

.

.fo

rce

that

is0.

708

0.44

50.

192

0.22

2m

ale N

otes

:The

regr

essi

ons

liste

dab

ove

repr

esen

tju

stth

ree

ofth

edi

ffer

ent

first

-sta

gere

gres

sion

s.In

addi

tion

toth

ela

gged

viol

ent

and

prop

erty

.

crim

era

tes,

the

regr

essi

ons

from

equa

tion

2al

soin

clud

edal

lth

eth

eot

her

vari

able

slis

ted

for

the

wei

ghte

dle

ast

squa

res

estim

ates

.The

fixed

effe

cts

used

inth

edi

ffer

ent

first

-sta

gere

gres

sion

sco

rres

pond

edw

ithth

ose

used

inth

ese

cond

stag

e.T

hees

timat

esre

port

edab

ove

use

city

and

year

fixed

effe

cts.

ECONOMIC INQUIRY254

including the consent decree dummy vari-able directly into the second equation thatestimates the crime rate. This change hasvirtually no effect on the results reportedpreviously. As a further test of the sensitivityof the results, I also tried reestimating theresults in Table IV by removing the crimerate variables from the first equation and thepattern of results remained similar to thosealready reported.24

The question of whether more black po-lice officers had a differential impact in moreheavily black areas can be examined by in-teracting the percent of the police force thatis black with the percent of the populationthat is black. The violent and property crimeestimates corresponding to the regressions inTable IV, and the estimates that did notinclude the crime rates in the first regressionall imply that the increase in crime from hiringblack officers is greatest in communities withthe most blacks. For example, the violent andproperty crime estimates that correspond tocity fixed effects estimates in Table IV arepositive and have t-statistics of 4.8 and 4.2,respectively.

Finally, data on whether a police depart-ment was unionized and the gross salary paidper sworn officer were available, though foronly 1987 and 1990. Using these two vari-ables and the smaller data set, I reestimatedthe results reported in Table IV and foundvery little change in results. For the mostpart neither of these variables was significantin explaining changes in the crime rate.

24. For example, after excluding the crime ratesfrom the first-stage regression, the city fixed effectsregressions produced estimates for the percentage of

.the police force that is black of 2.43 t-statistic 1.741 .for violent crimes and 2.25 t-statistic 1.864 for prop-

erty crimes. For the percentage of the police force thatis minority, the city fixed effect results were: 1.98 t-sta-

. tistic 1.810 for violent crimes and 1.86 t-statistic .2.055 for property crimes. For the percentage of the

police force that is male, the city fixed effect results .were: 7.73 t-statistic 1.012 for violent crimes and

.7.9 t-statistic 1.042 for property crimes. As wastrue in Table IV, the level of significance tended to behigher for county fixed effect regressions. The first-stageregression results also remain similar to those alreadyreported. For the regression estimating the percentageof the police force that is black, the consent decree

.coefficient is .017 t-statistics .899 and the number of .years that it is in effect is .0042 t-statistic 3.376 . For

the regression for minorities, the consent decree coef- .ficient is .059 t-statistics 2.274 and the number of

.years that it is in effect is .0049 t-statistic 2.962 . Forthe regression for males, the coefficients are again sta-tistically insignificant.

IV. ARE HIGHER CRIME RATES A RESULTOF LESS-EFFECTIVE POLICE OR GREATER

REPORTING RATES?

Unfortunately, the FBIs Uniform CrimeReport Data relies on reported, not actual,crimes. The problem is potentially critical forthis study, because the racial or gender char-acteristics of the police officers could eitherbe altering the behavior of criminals andorthe rate at which victims report crimes. Theproblem is made even worse by the fact thatboth sides of the debate can provide expla-nations for the preceding results. Those fa-voring affirmative action can argue that thehigher reported crime rates when more mi-norities are hired implies that the commu-nity feels more comfortable about report-ing crimes. In contrast, those who believethat lower standards mainly result in less-qualified officers can say that the resultsconfirm the poor performance of the less-qualified officers.

There are several ways of investigatingwhether the results are being driven byhigher reporting rates. The simplest ap-proach is to look at murder and manslaugh-ter, where underreporting is essentiallynonexistent. Thus, the race or sex of thepolice officer does not produce additionalreporting. For both murder and manslaugh-ter, the results are very consistent. Moreminority, black, or female officers are associ-ated with higher murder and manslaughterrates, while more white and male officersimply fewer deaths. These two crimes arealso the most accurately reported for an-other reason: if multiple offenses are perpe-trated at the same time, only the most seri-ous offense is reported. Thus, if an armedrobbery resulted in murder, only the murderand not the robbery is recorded.

Further, the importance of the reportingproblem should vary systematically acrosscrime categories as the loss from the crimevaries. For example, suppose that a blackperson is making a decision on whether toreport a theft to a predominantly white po-lice department. His decision to report thecrime depends on the value of the itemstolen, the probability that the item will berecovered, and the cost involved in going tothe police station, including whatever diffi-culties might arise in how the black manmight be treated by white police officers.The victim would only report crimes where


either the value of the item stolen or theprobability of recovery is relatively high.Lowering the cost of the black person report-ing the crime by introducing more black of-ficers would result in more reporting of rela-tively low-value, low-probability-of-detectioncrimes. Since the cost of making the com-plaint constitutes a much bigger percentageof the return to acting on relatively smallharms, actions that reduce those costs have amuch bigger effect on reporting minorcrimes.

For at least broad categories of propertycrimes it is possible to make this comparison.

.Miller, Cohen, and Wiersema 1996 claimedthat in 1992 the average larceny involvedproperty loss of $270, burglary $970, andauto theft $3,300. By comparison, the dif-ferences in the arrest rates are small: larceny30%, burglary 21%, and auto theft 25%.These figures would imply that the biggestincrease in reporting from changing the racialmix of police should occur for larceny, next

for burglary, and least for auto theft. Autotheft and burglary should also tend to haverelatively high reporting rates compared tolarceny simply because these crimes must bereported as a condition of obtaining reim-

.bursement from insurance companies. Yet,all of the two-stage least squares estimates inTable IV indicate that the racial or gendercompositions of the police department havealways smaller impacts on larceny than onburglary, and half the time the impact onlarceny is smaller than on auto theft. Noneof the estimates are consistent with the ear-lier results arising from increased reportingrates.

V. DISAGGREGATING FURTHER BY RACEAND SEX

For 1987 and 1990, the Department ofJustice survey determined the percent of eachracial group that was male or female. Thetwo-stage least squares regressions reportedearlier were therefore reestimated with twochanges: the previous racial or sex break-downs were replaced one at a time with theeight new race and sex categories and thefirst equation in the two-stage least squaresincluded a dummy variable that equals 1when the consent decree dealt with eitherrace or sex.

Table VI reports the county fixed effectswith separate year fixed effects for each state.

Despite the sample size being about one-third smaller, the results are similar to thosealready reported. Gender plays an evensmaller role than it did in the earlier results.The effectiveness of different types of policeofficers lies more along racial than genderlines, though there are notable exceptionsfor Asians, where males are associated withfewer crimes and females more. Murder di-vides along racial lines, with more whites .both males and females coinciding withlower death rates but the reverse being truefor blacks and Hispanics. In all but a few ofcases, more blacks and Hispanics are associ-ated with higher crime rates.

The variables explaining rape provide verylittle evidence that the gender of the policeofficer affects this crime differently. Forwhites and blacks, the different gender racialgroups have the same coefficient signs andare statistically indistinguishable. Althoughdifferences do exist for Hispanics and Asians,even here the effects do not suggest a consis-tent pattern with the relative impacts of maleand female officers having the opposite im-pacts in the two cases. The strength of theseresults make it very difficult to believe thatmale and female officers have much of adifferential impact on deterring rapes. Al-though it is still quite likely that male andfemale officers have different skills in deal-

ing with rape e.g., female officers may bebetter able at getting rape victims to reveal

.details , the tests do not allow us to differ-entiate what the skill differences are foreach gender. Victims or potential victimsmay also value more than simply deterrence.For example, they may value how they feelgoing through the process, and that is an-other dimension that we are unable to mea-sure. However, even if these other attributesare significantly valued, the results presentedhere allow us to discuss the trade-off be-tween the number of rapes and these otherpossible dimensions.

As another attempt to control for differ-ences in law enforcement across states, I alsoreestimated the regressions shown in TableVI with city and year fixed effects and in-cluding variables for both the per capita stateemployment in corrections and the judicialsystem.25 Including these variables had nodiscernible impact on the results reported.

25. See Lott and Mustard 1997 , for a discussion ofthese data.

ECONOMIC INQUIRY256

TA

BL

EV

IU

sing

Tw

o-st

age

Lea

stSq

uare

sto

Fur

ther

Exa

min

eth

eD

iffer

ence

sby

Rac

ean

dSe

x:C

ontr

ollin

gfo

rC

ount

yF

ixed

Eff

ects

and

Stat

eF

ixed

Eff

ects

that

Var

yby

Yea

r

Cat

egor

y%

ofP

olic

eof

Cri

me

%of

Pol

ice

%of

Pol

ice

For

ce%

ofP

olic

e%

ofP

olic

e%

ofP

olic

eF

orce

Asi

an%

ofP

olic

eF

orce

Bei

ng%

ofP

olic

eF

orce

Bla

ckH

ispa

nic

For

ceH

ispa

nic

For

ceW

hite

For

ceW

hite

Pac

ific

Asi

anP

acifi

cE

xpla

ined

For

ceB

lack

Mal

eF

emal

eM

ale

Fem

ale

Mal

eF

emal

eM

ale

Fem

ale

Vio

lent

10.4

314

%38

.62

14%

13.9

818

%23

4.02

27%

6.

2813

%4.

545

2%

56.1

231

%24

1.91

10%

.

.

.

.

.

.

.

.

crim

e2.

672

2.65

72.

065

0.89

02.

858

0.29

00.

735

2.16

0Pr

oper

ty3.

954%

15.0

24%

5.36

65%

93.9

48%

2.

434%

3.80

41%

28

.03

12%

97.2

53%

.

.

.

.

.

.

.

.

crim

e1.

723

1.82

71.

444

0.84

91.

883

0.38

90.

621

1.55

4M

urde

r19

.72

60%

66.9

155

%18

.46

56%

79.2

874

%

10.3

778

%

2.45

54%

54

.34

68%

340.

9450

%

.

.

.

.

.

.

.

.2.

742.

719

1.75

80.

911

2.73

80.

095

0.98

21.

904

Man

-26

.074

60%

101.

5763

%27

.21

63%

148.

7330

%

16.4

761

%33

.60

28%

95

.46

64%

682.

163

49%

.

.

.

.

.

.

.

.

slau

ghte

r2.

317

2.57

11.

976

0.87

42.

649

0.67

00.

839

2.10

1R

ape

4.26

910

%15

.50

10%

6.18

215

%90

.44

19%

2.

588

10%

1.

221

1%

6.33

36%

97.0

257%

.

.

.

.

.

.

.

.

1.34

51.

354

1.10

10.

732

1.40

40.

088

0.10

31.

090

Tot

al11

.40

18%

41.7

418

%13

.90

22%

252.

8236

%

6.66

517

%6.

354%

82

.098

56%

251.

9713

%

.

.

.

.

.

.

.

.ro

bber

y2.

583

2.57

81.

951

0.88

72.

728

0.36

00.

910

2.10

4T

otal

9.80

214

%37

.28

14%

15.4

922

%22

7.57

29%

6.

2614

%3.

015

2%

21.0

7113

%25

1.05

11%

.

.

.

.

.

.

.

.

assa

ult

2.38

72.

445

2.02

40.

881

2.64

20.

174

0.27

72.

047

Bur

glar

y7.

101

8%27

.87

9%9.

553

11%

181.

0619

%

4.44

68%

12.0

665%

70

.765

35%

185.

997%

.

.

.

.

.

.

.

.

2.44

42.

534

1.87

90.

896

2.70

30.

868

1.05

12.

284

Lar

ceny

2.31

92%

11.7

53%

6.49

07%

86.8

58%

2.

096

4%9.

061

3%

3.77

32%

110.

964%

.

.

.

.

.

.

.

.

1.10

11.

534

1.69

40.

843

1.69

50.

839

0.09

61.

803

Mot

or13

.30

18%

39.4

615

%3.

090

4%20

1.95

25%

5.

3795

12%

6.

265

3%

193.

2454

%11

2.87

5%

.

.

.

.

.

.

.

.ve

hicl

e2.

558

2.17

00.

457

0.85

32.

018

0.33

81.

289

0.93

2th

eft

Ass

ault

24

.16

36%

24.2

17%

38

.05

27%

133.

068%

3.

0512

%16

.68

14%

35

.32

3%33

4.5

2%

.

.

.

.

.

.

.

.on

polic

e1.

292

0.41

81.

280.

905

2.29

5.2

801.

671

0.90

5of

ficer

s

Not

es:T

hese

cond

-sta

gere

gres

sion

estim

ates

that

are

repo

rted

abov

eac

coun

tfo

rth

esa

me

vari

able

sco

ntro

lled

for

inth

efir

st-s

tage

regr

essi

onex

cept

for

the

cons

ent

decr

ee,t

henu

mbe

rof

year

ssi

nce

the

cons

ent

decr

eew

ent

into

effe

ct,a

ndw

heth

erth

eci

tys

may

oris

blac

k.C

ount

yfix

edef

fect

sar

eus

ed,w

ithad

ditio

nal

sepa

rate

fixed

effe

cts

for

each

stat

eby

year

topi

ckup

any

chan

ges

atth

est

ate

leve

lth

atm

ight

expl

ain

chan

ges

incr

ime

rate

sov

ertim

e.A

llre

gres

sion

sus

ew

eigh

ted

leas

tsq

uare

sw

here

the

vari

able

sar

ew

eigh

ted

byth

eci

typo

pula

tion.

The

abso

lute

t-sta

tistic

sar

esh

own

inth

epa

rent

hese

sbe

low

the

coef

ficie

ntes

timat

e,w

ithth

epe

rcen

tof

the

endo

geno

usva

riab

les

mea

nth

atca

nbe

expl

aine

dby

aon

e-st

anda

rd-d

evia

tion

chan

gein

the

exog

enou

sva

riab

lesh

own

inth

ead

jace

ntco

lum

n.Sa

mpl

esi

zeis

439

and

cove

rsth

eye

ars

1987

and

1990

.


While the impact of hiring more people incorrections and the judicial system usuallyreduced crime, the effect was never statisti-cally significant. Variables to account forconcealed handgun laws, waiting periods andthe length of those waiting periods in buyinga gun, penalties for using guns in commis-sions of crimes, and cocaine prices were alsoincluded,26 but only the variable for thepresence of concealed handgun laws reducedcrime and none of these variables apprecia-bly altered the other findings. Passage ofconcealed handgun laws reduced murderrates by about 10.5%.27 Controlling for theuse of Lojack automobile anti-theft devicestended to make the results for black andminority officers more positive and statisti-cally significant though the coefficient forLojack was not significant.28

VI. FELONIOUS KILLING OF POLICE,ACCIDENTAL POLICE DEATHS, ASSAULTS

ON POLICE, AND SHOOTING CIVILIANS

Many studies have focused on whetherblacks and other minorities civilians havebeen shot by police at disproportionately

.higher rates Matulia 1985, 7 . The stan-dard view is that the higher rates at which

26. See Lott and Mustard 1997 , for a discussion ofthese data.

27. Given the possible relationship between drugprices and crime, I reran the regressions in Table IV byincluding an additional variable for cocaine prices. Oneargument linking drug prices and crime is that if thedemand for drugs is inelastic and if people commitcrimes in order to finance their habits, higher drugprices might lead to increased levels of crime. Using theDrug Enforcement Administrations STRIDE data set

from 1977 to 1992 with the exceptions of 1988 and.1989 , Michael Grossman, Frank J. Chaloupka, and

Charles C. Brown 1996 estimate the price of cocaine asa function of its purity, weight, year dummies, yeardummies interacted with eight regional dummies, andindividual city dummies. However, these data are notperfect. Because of the lack of observations for 1993, Iused the drug prices for 1992. While the drug pricevariable was positive it was not statistically significantand its inclusion had very little impact on the relation-ship between the type of police

does a helping hand put others at risk - affirmative action, police departments, and crime (2000) by...

Documents