The online version of this article can be found at: http...

http://asr.sagepub.com/American Sociological Review

http://asr.sagepub.com/content/74/2/291The online version of this article can be found at:

DOI: 10.1177/000312240907400207

2009 74: 291American Sociological ReviewRyan D. King, Steven F. Messner and Robert D. Baller

Contemporary Hate Crimes, Law Enforcement, and the Legacy of Racial Violence

Published by:

http://www.sagepublications.com

On behalf of:

American Sociological Association

can be found at:American Sociological ReviewAdditional services and information for

http://asr.sagepub.com/cgi/alertsEmail Alerts:

http://asr.sagepub.com/subscriptionsSubscriptions:

http://www.sagepub.com/journalsReprints.navReprints:

http://www.sagepub.com/journalsPermissions.navPermissions:

http://asr.sagepub.com/content/74/2/291.refs.htmlCitations:

at Serials Records, University of Minnesota Libraries on January 10, 2011asr.sagepub.comDownloaded from

http://asr.sagepub.com/

http://asr.sagepub.com/content/74/2/291

http://www.sagepublications.com

http://www.asanet.org

http://asr.sagepub.com/cgi/alerts

http://asr.sagepub.com/subscriptions

http://www.sagepub.com/journalsReprints.nav

http://www.sagepub.com/journalsPermissions.nav

http://asr.sagepub.com/content/74/2/291.refs.html


Conflict theories of crime and criminal lawposit that the state largely serves the inter-

ests of dominant groups in society (Quinney1974; Turk 1969; Vold 1958) and this functioncan be expressed in two distinct ways. On theone hand, the legal apparatuses of the state—law and law enforcement—can be used to sub-

ject members of subordinate groups to puni-tive control. Research in the group threat vari-ant of the conflict tradition, for instance,suggests that state social control increases inresponse to perceived threats from subordi-nate groups (Behrens, Uggen, and Manza2003; Jacobs and Carmichael 2001; Jacobsand O’Brien 1998; Liska, Chamlin, and Reed1985). On the other hand, the state can fail toserve the interests of subordinate groups byoffering limited protection from harmful andunlawful behaviors. The state can serve dom-inant group interests not only by administer-ing punitive sanctions against a subordinategroup, but also by “looking the other way”when civilians discriminate against subordi-nate group members.

One of the more egregious examples of such“malign neglect” on the part of the state in U.S.history is the phenomenon of lynching in the late

Contemporary Hate Crimes, LawEnforcement, and the Legacy of RacialViolence

Ryan D. King Steven F. MessnerUniversity at Albany-SUNY University at Albany-SUNY

Robert D. BallerUniversity of Iowa

This article investigates the association between past lynchings (1882 to 1930) and

contemporary law enforcement responses to hate crimes in the United States. While prior

research indicates a positive correlation between past levels of lynching and current

social control practices against minority groups, we posit an inverse relationship for

facets of social control that are protective of minorities. Specifically, we hypothesize that

contemporary hate crime policing and prosecution will be less vigorous where lynching

was more prevalent prior to 1930. Analyses show that levels of past lynching are

associated with three outcome variables germane to hate crime policing and

prosecution, but the effect of lynching is partly contingent on the presence of a minority

group threat. That is, past lynching combined with a sizeable black population largely

suppresses (1) police compliance with federal hate crime law, (2) police reports of hate

crimes that target blacks, and in some analyses (3) the likelihood of prosecuting a hate

crime case. Our findings have implications for research on law and intergroup conflict,

historical continuity in the exercise of state social control, and theories that emphasize

minority group threat.

AMERICAN SOCIOLOGICAL REVIEW, 2009, VOL. 74 (April:291–315)

Direct correspondence to Ryan King, Universityat Albany, SUNY, 351 Arts and Sciences, 1400Washington Avenue, Albany, NY 12222([email protected]). An earlier version of this arti-cle was presented at the 69th Annual Meeting of theSouthern Sociological Society, New Orleans, LA,March 22 to 25, 2006. The authors thank RichardFelson, Chris Galvan, Kelly McGeever, and the edi-tors and anonymous reviewers of the AmericanSociological Review for comments and assistancewith this work.



nineteenth and early twentieth centuries.1

Lynching “was a powerful tool of intimidation”for the white population to use against the sub-ordinate black population (Brundage 1997:2).Although lynching was not overtly sponsored bythe state, lynching incidents were infrequentlyprosecuted and lynching “was tolerated (andoften applauded) by local politicians and lawofficers” (Garland 2005:810).2

Today, there is a growing interest in hatecrime as an extra-legal, indeed illegal, behaviorthat can be conceptualized as a means of socialcontrol. Scholars suggest that, much like lynch-ing, hate crimes do not merely victimize par-ticular individuals. Rather, they constitute ameans of controlling the behavior of an entiregroup through intimidation and often violence(Craig 2002; Perry 2001). While such behavioris hardly a new phenomenon, legislation thatmandates additional police attention to hatecrimes, and criminal statutes that provideenhanced penalties for crimes motivated by big-otry, are relatively recent concepts in U.S. law.Legal protections “on the books,” however, arenot always implemented in practice (Jennessand Grattet 2005).

We suggest that racial antagonism is deeplyingrained in some pockets of the country, andthere is continuity in how the state, via its lawenforcement agencies, reacts to offenses moti-vated by bigotry. Our research investigates theassociation between legacies of past lynchingand contemporary law enforcement responsesto hate crimes. We regard the lynching of blacksnearly a century ago as indicative of two con-ditions pertinent to this research. First, lynch-ing was perhaps the most egregious expressionof overt prejudice and demands for whitesupremacy during the Jim Crow era. Second,

lynching dramatically depicts the state’s failureto protect a racial minority group from violent,extra-legal social control.

Mindful of the categorical differences in racerelations when comparing the Jim Crow erawith the present, we suggest that racial antag-onism dies hard and may take on new forms indifferent historical periods. In line with Boboand colleagues’ (Bobo, Kluegel, and Smith1997; Bobo and Smith 1998) notion of a shiftfrom Jim Crow to laissez faire racism, we positthat an undercurrent of racial antagonism per-sists in some locales but is now manifested inless overt actions. The spectacle of lynchinghas given way to less violent expressions ofinterracial conflict. The state’s past tendency to“turn a blind eye” to anti-black lynching is nowexpressed through resistance to policies per-ceived as giving special treatment to racialminorities. A logical extension of this notion isthat in places where bigotry was culturallyaccepted and institutionalized in the Jim Crowera, law enforcement is now apt to resist “affir-mative action” policies that give special atten-tion to discrimination against minorities.

Hate crime statutes represent one type of lawthat is likely seen as giving special protectionto racial minority groups. Although the writtenlaws must apply equally to majority and minor-ity group victims alike to remain constitution-al (Wisconsin v. Mitchell 1993), we argue thathate crime laws are largely protective of racialminorities for two reasons. First, non-whiteracial groups (particularly blacks) are statisti-cally most likely to be victims of hate crimes(Messner, McHugh, and Felson 2004). Second,the congressional history of hate crime laws inthe United States largely entails testimony byaggrieved minority groups and makes refer-ence to heinous acts ranging from the Holocaustto modern anti-gay violence (see Jenness andGrattet 2001, e.g., pp. 35–37, 55).

We argue that past lynching is likely predic-tive of current policing and prosecution of hatecrime laws. Our reasoning is based on threepremises: (1) racial antagonism was a powerfulsocial force underlying the phenomenon oflynching, particularly, but not exclusively, inthe South; (2) racial antagonism tends to bedeeply ingrained in culture (i.e., it dies hard);and (3) the manifestations of racial antagonismevolve over time; specifically, an importantmanifestation of racial antagonism in the con-

292—–AMERICAN SOCIOLOGICAL REVIEW

1 Malign Neglect is the title of Michael Tonry’s(1995) book on racial discrimination in criminal jus-tice administration.

2 We are careful to point out that local politiciansand law enforcement were not always supportive ofthis practice, and volumes of theAmerican Negro YearBook from the 1920s and 1930s describe severalefforts to prevent lynching. It appears well acceptedby scholars of lynching, however, that a significantproportion of the population did not regard the pre-vention of lynching during this era as a high priori-ty.



temporary era is resistance to special legal pro-tections for racial minorities. We further arguethat the impact of lynching is likely to dependon the relative size of the racial minority, con-sistent with insights from conflict and threatperspectives on social control.

PAST RESEARCH, THEORY, ANDHYPOTHESES

CONTINUITY IN SOCIAL CONTROL—LYNCHING AND ITS LEGACY

Prior research shows considerable continuitybetween past and present levels of social con-trol, and such path dependence is prominentlycaptured in scholarship on past incidents oflynching and contemporary punitive sanctions.Zimring (2003), for instance, finds a statisticalassociation between past lynching and currentexecutions. He suggests that the death penaltyis largely an extension of a vigilante tradition,and the racial overtones associated with lynch-ing continue to motivate the legalization and useof capital punishment. To that end, the racialimbalance in capital punishment (Baldus,Pulaski, and Woodworth 1983; Paternoster1984) is analogous to disproportionate blackvictimization in past lynchings (Tolnay, Deane,and Beck 1996).

Temporal continuity in social control is alsothe focus of related work on other forms of lawand punishment. Wacquant (2000) views thehistory of social control in the United States asa succession of institutions that disproportion-ately subject minority populations to discrimi-natory punishment. Chattel slavery gave way toJim Crow, which was then replaced by the “sur-rogate ghetto” of the modern penal apparatus(Wacquant 2000). Such continuity entails instru-mental and emotive characteristics. Punitivecontrol can serve instrumental ends by main-taining existing power arrangements and strate-gically limiting minority group power (Behrenset al. 2003). Punitive acts may also entail sym-bolic and emotional attributes. Lynching servedpartly as a means of moral enforcement (Wyatt-Brown 1982), consistent with sociological argu-ments that criminal punishment is invested withmoral power (Durkheim 1973; Garland 1990).Lynching was not analogous to other homi-cides, as its purpose was to incite terror and senda message beyond the immediate victims(Tolnay and Beck 1995). This practice was also

laden with the politics of racial domination(Garland 2005), again akin to contemporaryideas on punishment (Jacobs and Carmichael2002).

The symbolic and cultural framework sur-rounding lynching is central to recent socio-logical research that links past lynching withcurrent violence and sanctioning. For instance,Messner, Baller, and Zevenbergen (2005) pro-pose that lynching was indicative of culturalsupport for violence. They show that lynchingexhibits staying power over several decades,evidenced by the correlation between past lynch-ing and current homicide levels in the AmericanSouth. Jacobs, Carmichael, and Kent (2005)reach a similar conclusion with respect to puni-tive sanctions. Accounting for other knowncovariates of death sentences, their researchindicates that past lynching is predictive of cur-rent death sentences administered againstblacks. Similar to Zimring’s arguments, theyconclude that the “tradition of legal vigilan-tism” (Jacobs et al. 2005:672) continues toinform the sanctioning process.

While prior research links past lynching withcontemporary social control, this body of workfocuses almost entirely on violence (e.g., inter-racial homicide) or state sanctions (e.g., exe-cutions) that disproportionately and adverselyaffect racial minorities. As an extension to thisline of research, we suggest that lynching isalso predictive of law enforcement actions thatare protective of minorities, such as the polic-ing and prosecution of hate crimes, but thisassociation should work in the opposite direc-tion. It is tenable, we suggest, that the legacy oflynching has two simultaneous effects. First,past lynching may predict current racial antag-onism that subsequently manifests in the com-mission of hate-inspired crime. Second, lawenforcement agencies may respond to hate-motivated attacks less vigorously under thesame conditions, thereby reducing the likeli-hood that such incidents will be reported.3 Weare concerned primarily with the latter propo-sition, which raises two additional questions:What mechanisms connect past lynching and

PAST LYNCHING AND PRESENT HATE CRIME LAW ENFORCEMENT—–293

3 While these two outcomes are conceptually dis-tinct, distinguishing between them becomes com-plicated in empirical assessments. We address thisissue in greater detail below.



current law enforcement practices? And whatare the specific implications for hate crimelaws?

FROM EMBRACING OVERT DISCRIMINATION

TO RESISTING SPECIAL PROTECTION

Lynching was a complex social phenomenon,and the literature advances numerous explana-tions for its prevalence. These include psycho-logical and psychoanalytic accounts, argumentsabout the reliance on popular or vigilante jus-tice to compensate for the perceived ineffec-tiveness of legal institutions, culturalinterpretations highlighting southern notionsof chivalry and honor, and “social threat”accounts emphasizing the political and eco-nomic competition between blacks and whites(Brundage 1997; Pfeifer 2004; Tolnay and Beck1995). Despite disagreements about the relativeimportance of the various alleged causal factors,the distinctive racial character of lynching ishardly disputed. Beck and Tolnay (1997) esti-mate that of the more than 2,700 southernerslynched between 1882 and 1930, about 8 out of10 were blacks brutalized by white mobs.Lynching was more than an effort to enforcesocial conformity in general on behalf of thecommunity. These acts were extreme expres-sions of racial antagonism designed to intimi-date and control the black minority (Brundage1993, 1997; Tolnay and Beck 1995).

A second feature of lynching was the “malignneglect” by the state. The rampant lynching ofblacks signified the state’s conspicuous failureto protect a racial minority. Many lynching inci-dents were perpetrated with impunity, as lawenforcement either tolerated such behavior oracquiesced to it (Garland 2005). For example,some local law enforcement officials appearedto be complicit in Mack Charles Parker’s abduc-tion from his jail cell immediately before hislynching in 1959 (Smead 1986). Althoughlynchings were sometimes prevented, and a feworganizations overtly sought to end the practice(Tuskegee Institute 1932), lynching was rarelypursued as a criminal offense. Statutes for arrest-ing and prosecuting lynchers were readily avail-able (e.g., homicide statutes), but as Garland(2005:809) notes,

laws were rarely enforced against the lynchers.Prosecutions were not brought for lack of politi-cal will at the state level, or for lack of coopera-

tion in the local community. In practice, lynchersenjoyed immunity from the state or local prose-cution. Anti-lynching campaigners .|.|. appealed forintervention by the federal government. .|.|. But thisappeal to national law also failed.

The cultural norms that motivated and per-mitted lynching have clearly changed. Today,lynching rarely occurs in the United States, andwhen such violent crimes of bigotry are perpe-trated, they are likely to be dealt with prompt-ly and severely by law enforcement. The 1998lynching of James Byrd, for instance, resultedin prosecutions, convictions, and death sen-tences. The absence of overt Jim Crow racism,however, has not been replaced by ardent sup-port for policies designed to prevent discrimi-nation or compensate for past inequities. Boboand Smith (1998) refer to this categorical shiftin race relations as the movement from “JimCrow” to “laissez faire” racism. The formerrefers to “overt bigotry” (Bobo and Smith1998:185), namely, beliefs that blacks are cat-egorically inferior to whites and the concurrentinstitutionalization of such sentiments (e.g.,segregation). Laissez faire racism rejects theinherent inferiority argument; under this para-digm, whites are likely to attribute inequality toindividual shortcomings and to “actively resistmeaningful efforts to ameliorate America’s racistsocial conditions and institutions [e.g., affir-mative action policies]” (Bobo and Smith1998:186).

The relevance of Bobo and colleagues’ ideasin the realm of criminal law is nicely illustrat-ed in Behrens and colleagues’ (2003) work onthe changing justifications for felon disenfran-chisement laws during and after Jim Crow. Inthe course of debates about felon disenfran-chisement in the Jim Crow era, judges and leg-islators made reference to the enrichment ofAnglo-Saxon civilization, and they openlylabeled blacks as “menacing” and inherentlydifferent from whites (Behrens et al. 2003:570).Significant support still exists today for felondisenfranchisement laws, but they are now jus-tified by race-neutral arguments against repeal-ing laws that disproportionately affect blacks(pp. 570–72).

We advance a similar argument with respectto law enforcement’s role in responding tocrimes of bigotry, or what are now commonlyreferred to as hate crimes. Prosecutions fol-lowing the lynching of blacks were spotty and




convictions “nearly nonexistent” (Klarman2000:55) during the Jim Crow era. We expectthat in counties with a history of lynching acentury ago, there is now a greater proclivity toresist policies that might give special treatmentto blacks. In this regard, we draw an analogywith prior research on police use of violence. AsJacobs and O’Brien (1998:842) explain withrespect to why police would use more violenceagainst blacks under conditions of minoritythreat, “it is not necessary to claim that privi-leged groups make direct demands that thepolice use greater amounts of deadly force.Without restraints, police violence is probable.”When privileged groups are threatened, thepowerful are less likely to interfere with policetactics, opening the way for more police vio-lence.

With respect to hate crime law, the situationis in some sense reversed. Compliance withand enforcement of hate crime laws place addedburdens on police officers and prosecutors.Recording offenses requires complicated judg-ments about motives, which makes provingoffenses more difficult. To the extent that racialantagonism is manifested in resistance to spe-cial protections for racial minorities, skepti-cism about the legitimacy of hate crime laws(i.e., Why should there be special statutesdesigned largely to protect minorities?) andindifference to enforcement of such laws, orcompliance with them, is increasingly likely.Furthermore, without support from the gener-al population, police and prosecutors are like-ly to follow the path of least resistance andinfrequently put these laws into action.

We investigate the following general hypoth-esis: the contemporary enforcement of hatecrime laws and compliance with hate crimemandates are inversely associated with priorlevels of lynching. Another possibility, which weexplicate below, is that the legacy of lynchingmatters only in the presence of a minority groupthreat.

RACIAL THREAT AND HATE CRIME LAW

Theories that emphasize racial threat viewlaw as an instrument for subverting econom-ic and political challenges posed by racialminorities. In line with classic explanations ofintergroup relations (Blalock 1967), the racialthreat thesis postulates that majority popula-

tions and elites perceive a large or growingracial minority group as threatening. Suchthreats incite a variety of reactions from major-ity groups, such as right-wing voting (Gilesand Buckner 1993), prejudicial attitudes(Quillian 1996), and expansive state socialcontrol (Jackson 1989).

Racial threat is implicated in our assessmentof lynching and hate crime law for two reasons.First, black population size as an indicator ofracial threat is likely to be predictive of hate-crime law enforcement. Research is mostly sup-portive of racial threat tenets in the realm ofcriminal law, as evidenced in work on arrestrates (Liska et al. 1985), policing expenditures(Jackson 1992), and the legalization of capitalpunishment (Jacobs and Carmichael 2002).These findings are generally congruent withresearch that connects black population size toprejudice toward blacks in the contemporaryUnited States (Quillian 1996) and to lynchingsa century ago (Tolnay et al. 1996).

We propose that if black population sizeincreases both prejudice and social control thatadversely affects racial minorities, it maydecrease social control perceived as protectingminorities. This argument is partly supported byresearch that finds less compliance with feder-al hate crime laws in southern cities and coun-ties with large black populations (King 2007).In line with that work, and in concert withresearch on the state’s neglect of blacks’ legiti-mate crime control needs (Anderson 1997,1999; Hawkins 1987), we hypothesize that blackpopulation size is inversely associated with hatecrime law enforcement and compliance withfederal hate crime laws.

Beyond this direct link between racial threatand hate crime laws, threat may also conditionthe effect of lynching. In their work on capitalpunishment, Jacobs and colleagues (2005:660)suggest that “in the absence of a black popula-tion sufficiently large to pose a current threat,a prior tradition of vigilantism [as indicated bylynching] may not be enough to lead to addi-tional death sentences.” To wit, past lynchingonly increases the contemporary use of capitalpunishment in areas with sizeable black popu-lations. It is reasonable to expect that racialcomposition might also modify the link betweenlynching and the state’s protective functions forracial minorities, which implies a statisticalinteraction. Specifically, we hypothesize that




the association between hate crime law enforce-ment (including compliance with hate crimelaw mandates) and lynching becomes increas-ingly negative as the size of the black popula-tion increases.

POLITICS, INTERGROUP CRIME, AND HATE

CRIME LAW ENFORCEMENT

Extant theory and research dictate that ouranalysis account for political partisanship.Politics is an increasingly salient factor in thestudy of punishment (Beckett 1997; Sutton2000). In particular, support for conservativepoliticians is linked with elevated use of crim-inal punishment (Jacobs and Carmichael 2002;Jacobs and Helms 1996, 1999). Yet conservativepolitics may decrease hate crime reporting(McVeigh, Welch, and Bjarnason 2003). Allour models thus include a measure of politicalconservatism.

Explanations of state responses to hate crimesmust also confront the notion that hate crime lawenforcement simply reflects the level of inter-group crime in a jurisdiction. This “reactiveexplanation,” commonly put forth as an alter-native to the threat perspective (Jacobs andO’Brien 1998), suggests that hate crime polic-ing and prosecution are direct functions of thelevel of intergroup crime motivated by animus.We thus test our hypotheses concerning lynch-ing, racial threat, and law enforcement respons-es to hate crimes while accounting for the levelof intergroup violence (to the extent possiblewith the available data).

DATA, MEASURES, ANDMETHODOLOGY

SAMPLE, UNITS OF ANALYSIS, AND THE

LYNCHING VARIABLE

Our samples and units of analysis are deter-mined by the strategic independent variable—lynching. We assess the effect of past lynchingon current law enforcement responses to hatecrimes using two different samples and threeoutcome variables. First, we use rich lynchingdata for a sample of counties and county clus-ters in 10 southern states: Alabama, Arkansas,Florida, Georgia, Kentucky, Louisiana,Mississippi, North Carolina, South Carolina,and Tennessee. County-level lynching data forthese states come from Tolnay and Beck’s inven-

tory of southern lynching.4 Tolnay and Beck(1995:259) developed their inventory by firstcreating an “unconfirmed master list” of alllynchings that occurred within these states, asreported in previously used and publicly avail-able inventories. They then validated these inci-dents through meticulous cross-checking withnewspaper stories “to determine whether theevent truly was a lynching and to correct for fac-tual errors that may have been included in theoriginal inventories” (p. 260). Their final prod-uct is “a confirmed inventory” of lynching in theSouth from 1882 to 1930 (p. 260). Tolnay andBeck’s register is the “gold standard” forresearch on lynching. Our primary analyses arethus based on the 10 southern states for whichthese data are available.

As noted, counties serve as the focal geo-graphic units. However, because county bor-ders have changed over the past century and thuscurrent boundaries do not always mirror theperiod from which our lynching data are gen-erated, we aggregate adjacent counties into clus-ters in some cases. We created clusters inaccordance with the Horan and Hargis (1995)county template, which includes an aggregationkey indicating which counties should be com-bined to account for boundary changes thatoccurred between 1880 and 1990. To retain asmuch geographic detail as possible, we do notreplace individual counties with county clustersin two situations: (1) when a county cluster hadno lynching activity and (2) when such activi-ty occurred only after boundary changes werecompleted. In these cases we can determinewhether a modern county experienced a lynch-ing between 1882 and 1930; where no lynchingsoccurred, there is no reason to create countyclusters. This “southern sample” consists of726 counties and county clusters.5


4 Stewart E. Tolnay and E. M. Beck generously pro-vided the lynching data. For a description of data col-lection methods, see Tolnay and Beck (1995).

5 In supplementary analyses, we created addition-al clusters and reestimated the models. We did thisbecause some law enforcement agencies are coveredby other agencies in a different county. For example,a few agencies report no crime data to the FBI, butinstead send their data to an agency in another coun-ty that submits those data, plus their own, to the FBI.We deal with this in the current analysis by control-ling for general crime reporting as a predictor vari-



Because our theoretical argument concernsthe legacy of lynching, which constitutes a vio-lent act laden with symbolic power of domina-tion and suppression (Garland 2005), wemeasure the lynching count rather than thelynching rate. The symbolic significance oflynching is unlikely to be diluted in citizens’minds by the proportionate representation oflynching in the immediate area (cf. Tolnay et al.1996). Moreover, we include only those lynch-ing incidents that had one or more black victims.This maintains continuity with our theoreticalargument, which concentrates on the racialantagonism associated with lynching. Our mainindependent variable thus depicts the frequen-cy of lynchings with black victims in countyclusters from 1882 to 1930.

Although we focus primarily on the 10 south-ern states identified above (because of the highquality of the lynching data), we supplement thisset of regional analyses with an assessment oflynching and law enforcement responses to hatecrimes for all counties in the United States.Replicating the analysis of 10 southern statesspeaks to the robustness and generalizabilityof our findings, with the caveat that the nation-al lynching data are of lesser quality outside thestates in the Tolnay–Beck sample.6 For this“national sample,” we catalogued all lynchingincidents with black victims for counties outsidethe aforesaid 10 southern states using data fromthe NAACP’s 1919 publication, Thirty Years ofLynching in the United States, 1889–1918. Thispublication’s goal was to compile detailed infor-mation on lynching incidents that had beenreported in various sources, such as the ChicagoTribune and the Tuskegee Institute, and to pub-lish these “facts” about lynching without edi-torial commentary. Appendix II of thepublication is a chronological listing of personslynched in the United States between 1889 and1918. Nearly all the cases include informationon the race of the victims and the city (and typ-ically county) of the lynching.7 We aggregated

the NAACP data to the county level and mergedthese data with the counts based on the Tolnayand Beck inventory to create a national sample.As with the 10-state sample, we formed clustersin other states using the Horan and Hargis tem-plate. The Appendix lists counties that were notformed into clusters because all their lynchingactivity occurred after modern boundaries werein place. Some clustering was also required tolink data for 1990 and 2000 (e.g., in Alaska).These counties are also listed in the Appendix.

Because the NAACP data are available for aperiod of 30 years, whereas Tolnay and Beck’sinventory encompasses 49 years, we annual-ized the lynching counts for the supplementaryanalyses of all U.S. counties. That is, we divid-ed the NAACP lynching count by 30 and theTolnay and Beck count by 49 to standardize forthe number of years for which data were record-ed. We are thus able to examine whether anyeffect of lynching observed in the southern sam-ple can be replicated in a national sample ofcounties and county clusters.

DEPENDENT VARIABLES

Crime statistics can reflect the actual amount ofcriminal offending or the reporting practices oflaw enforcement agencies (Black 1971; Kitsuseand Cicourel 1963); more likely, it will be somecombination of the two. This duality withrespect to crime statistics complicates ourinquiry because we cannot know whether hatecrime data furnished by the government reflectthe true number of offenses motivated by big-otry (as used by Medoff 1999) or law enforce-ment’s willingness to execute hate crime statutes(the approach adopted by McVeigh et al. 2003).Mindful of this “realist versus constructivist”debate, we use three separate dependent vari-ables that cover two primary institutions of for-


able, but there is no substantive difference in thefindings when dealing with the issue via additionalclustering. The results are consistent for the 10-stateand national samples (described below).

6 See Appendix B in Tolnay and Beck (1995) foran extended discussion of the types of errors oftenfound in publicly available inventories on lynching.

7 We recorded 494 lynching incidents with blackvictims (outside the 10 southern states described

above) from the NAACP (1919) publication. In eightcases, no city or county information was provided,and thus we could not code the geographic locationof those incidents. In 18 cases, the NAACP listed acity but not a county. With the assistance of second-ary sources (namely Abate 1974; Forstall 1996;Gannett 1902, 1975), we found the correspondingcounty in two of those cases. We dropped nine otherincidents because the county identifier in the NAACPdata appears to be wrong. In all, we have countyidentifiers for 461 incidents. The Appendix providesa full list of changes we made to the NAACP data.



mal social control—police departments and dis-trict attorneys’ offices.

The police represent the gateway to the crim-inal justice system, and crimes not identifiedby the police are not likely to receive attentionfrom other law enforcement agencies. Bell(2002), for instance, reports that police inves-tigations of hate crimes have notable implica-tions for charging decisions made later in theadjudication process. We thus measure hatecrime policing via two measures. Our firstpolicing variable measures compliance withthe Hate Crime Statistics Act (HCSA). TheHCSA requires the Justice Department toacquire and publish data about crimes thatmanifest prejudice based on certain groupdefining characteristics, including race, eth-nicity, sexual orientation, and disability (seePublic Law 101-275, section b(1); also PublicLaw 103-322 and Public Law 104-155).Compliance with the HCSA is recognized asa f irst step toward policing hate crimes(Jenness and Grattet 2001), and it is viewed as“critical” for hate crime law enforcement (U.S.Department of Justice 2002). Such compli-ance indicates a basic recognition of the law,even if a law enforcement agency reports nohate crime offenses. To that end, investigatingcompliance with the HCSA has the advantageof making no assumptions about the preva-lence of offending motivated by bigotry in anarea, because policing agencies can complywith the law even if they report no hate crimeincidents. We suggest that failure to complywith the HCSA data collection mandate, par-ticularly when compliance with other federalcrime reporting initiatives is in evidence (gen-eral Uniform Crime Report reporting), signi-fies a resistance to or ambivalence about thehate crime law. In addition, this variable isuseful because the HCSA applies to all polic-ing agencies in the United States, regardless ofstate statutes, although compliance is both vol-untary and highly variable (McDevitt et al.2000).

We focus primarily on sheriffs’ departmentsin this analysis because the lynching data areorganized at the county level. Sheriffs largelycorrespond to counties and, with some excep-tions, their responsibilities include law enforce-ment and crime reporting. Using smaller unitsof analysis, such as municipal policing agencieswithin counties, would be problematic because

our focal independent variable, lynching, ismeasured at the county (or county-cluster) level.

In singular counties not grouped into clusters,we measure the compliance variable by thenumber of quarters that the sheriff’s office sub-mitted hate crime reports in compliance with theHCSA between 1992 and 2003.8 Where coun-ties are clustered because of border changessince 1880, we take the average number of com-pliant quarters for the cluster.9 That period, 1992to 2003, corresponds to the first year that datawere generated as a result of the HCSA and endswith the most recent available year when thisanalysis began. There is noticeable variation inthe degree of compliance during our period ofstudy (see Table 1 for descriptive statistics),with several departments failing to comply forthe duration of the study. We suspect that com-pliance with the federal hate crime law is notrandomly distributed; rather, we expect it tocovary with historical lynching activity andracial demographics.

While the compliance variable has theadvantage of making no assumptions aboutthe actual prevalence of hate crimes, it fails tocapture the degree of actual hate crime report-ing. That is, policing agencies could ceremo-nially “comply” with the letter of the lawwithout recording a single hate crime offense.We thus include a second outcome variablethat measures the number of anti-black moti-vated hate crimes reported by police in thecounty or county cluster. We include hatecrimes reported by sheriffs’ departments andlocal policing agencies within the county clus-ter because the data can be meaningfully aggre-gated to the county (or county-cluster) level.10

Since we focus partly on the concept of racialthreat, and because our lynching variable isused to indicate historical tolerance for extra-legal violence against blacks, our “hate crime


08 The Department of Justice requests that lawenforcement agencies submit quarterly hate crimereports. Our data cover 12 years; thus there are 48quarters in which law enforcement agencies were“at risk” of compliance (12 years � 4 quarters peryear).

09 See the Appendix for exceptions and addition-al notes about the law enforcement agencies.

10 We exclude hate crimes reported by state patrolsor regional police.



reporting” outcome variable focuses specifi-cally on the number of hate crimes that targetblacks, as identified by police. This dependentvariable represents a count of the crimes moti-vated by animus toward blacks as reported bypolicing agencies between 1992 and 2003. Asdepicted in Table 1, this variable includes atremendous range of hate crime reportingactivity. During the 12 years under investiga-tion, the number of anti-black hate crimes

reported in the 10 southern states varies fromzero in some counties to a high of 480.11


Table 1. Descriptive Statistics for Southern Sample (N = 726)

Mean SD Min. Max.

Dependent Variables—Compliance with federal hate crime law 20.2 13.41 0 45.00—Racially-motivated hate crimes that target blacks, as reported by policea 3.53 20.22 0 480—Hate crime prosecution in 2000 (N = 674) .18 .38 0 1.00Independent Variables—Lynching 3.37 5.53 0 61.00—Percent black (1990) 20.17 18.86 0 86.12—Percent black (2000) 20.31 19.32 0 86.13—Population logged (1990) 10.32 1.12 7.55 15.62—Population logged (2000) 10.45 1.14 7.64 15.85—Percent urban (1990) 32.04 26.65 0 99.96—Percent urban (2000) 35.54 28.13 0 99.90—Percent divorced (1990) 9.93 1.62 4.80 19.08—Percent divorced (2000) 12.46 1.50 7.23 19.79—Percent ages 15 to 29 (1990) 22.44 3.35 14.03 47.41—Percent ages 15 to 29 (2000) 20.20 3.36 12.94 44.46—General crime reporting to UCR – dummy measure .97 .16 0 1.00—General crime reporting to UCR – ordinal measure 1.34 .53 0 2.00—Percent voting for Bush in 1992 40.88 9.34 12.94 74.96—Percent voting for Bush in 2000 54.98 10.72 12.35 84.02—Black county commissioners per 100,000 (logged) 1.58 1.97 0 6.79—Any elected black sheriff .06 .23 0 1.00—White on black homicides .56 3.42 0 83.00—Percent born in state (1990) 75.78 13.45 17.22 97.24—Percent born in state (2000) 72.05 13.45 19.42 96.51—Percent black officers in county cluster (1990) 16.78 21.26 0 100—White unemployment (1990) 5.80 2.43 1.37 18.39—White unemployment (2000) 4.80 1.71 .39 12.93—White poverty (1990) 14.86 6.55 2.29 52.20—White poverty (2000) 12.63 5.18 2.40 45.31—Black–white unemployment ratio (1990) 2.7 1.13 0 11.75—Black–white unemployment ratio (2000) 2.91 1.89 0 26.84—Black–white poverty ratio (1990) 3.11 1.05 .78 8.96—Black–white poverty ratio (2000) 2.85 1.00 .56 9.38—Years under NIBRS 2.29 4.29 0 14—Police per thousand (agency – Table 2 only) .73 .62 .02 7.45—Police per thousand (county cluster – Table 3 only) 1.55 .75 .2 7.45—Hate crimes reported in 2000 .9 5.25 0 121

Note: Missing data on the hate crime prosecution variable decreases the valid N to 674. The N is also smaller forthe following variables that were included only in analyses of counties with 500 or more black residents: percentblack officers in the county, black–white unemployment ratio, and the black–white poverty ratio.a The maximum value is an outlier, which is omitted from analysis, although the results are consistent when it isincluded.

11 The high value represents an outlier and corre-sponds to a cluster of counties. We omit that casewhen estimating the effects of our predictor vari-ables on the outcome of hate crime reporting,although the results are consistent when including thecase. The same is true for the supplementary analy-



Our assumption is that higher levels of hatecrime reporting, in part, indicate greater atten-tion by police to identifying and formally report-ing potential anti-black hate crime incidents.12

Still, we acknowledge that the interpretation ofthis variable is somewhat problematic becausepolice identification of hate crimes to someextent reflects the actual volume of hate crimeoffenses. This is, in part, why we examine mul-tiple dependent variables, although we offertwo additional arguments in support of using thismeasure. First, our models include a measure ofinterracial violence perpetrated by the dominantracial group against a racial minority—the num-ber of white-on-black homicides. To the extentthat this measure is correlated with the level ofracially-motivated hate crimes, the confoundingof offending with police reporting practicesshould be reduced. Second, we note that themain effect of lynching on anti-black hate crimesmay be positive if the variable measuring report-ed anti-black hate crimes captures some actualcrimes of bigotry, although it would be consis-tent with our account if the interaction term forpercent black and lynching is negative.

Because policing captures only one institu-tional facet of law enforcement, we include anadditional variable germane to prosecution.Prosecution is arguably a more proximate indi-cator of law enforcement than is compliancewith federal reporting statutes or overall hatecrime reporting. We estimate the likelihood ofa hate crime prosecution, which we measureby the presence (coded 1) or absence (coded 0)of a single hate crime prosecution during a one-year period beginning in 2000.13 An additionalanalysis of the number of hate crime prosecu-

tions would be desirable, but data limitationsconfine our analysis of this dependent variableto the presence or absence of a prosecution.The only source containing data on prosecutionsfor all U.S. counties is the 2001 NationalProsecutors Survey (NPS; conducted by theU.S. Department of Justice.) The NPS is a bian-nual survey that monitors trends in prosecu-tion. The 2001 survey is a full census of alldistrict attorneys’ offices that operate in statecourts with felony jurisdiction in the UnitedStates. Among the information included in thissurvey is whether a district attorney’s officeprosecuted special categories of crime, includ-ing hate crime. For the year in question, 18 per-cent of the offices in our southern sampleprosecuted one or more hate crime cases (20percent in the national sample). We are unableto disaggregate this measure into types of hatecrime prosecution (e.g., racial motivation, reli-gious motivation), so this variable includes someadditional error compared with the previoustwo measures. It would, however, be consistentwith our theory if past lynching, particularly incombination with current racial threat, decreas-es the likelihood of hate crime prosecution.14

ADDITIONAL INDEPENDENT VARIABLES

In addition to lynching, a second focal inde-pendent variable concerns racial threat.Consistent with prior work on prejudice andpunishment, we measure racial threat as thepercentage black in a county or county cluster.Because our hate crime policing data begin in1992, we use 1990 U.S. Census data on racialcomposition for the policing dependent variablesand 2000 data for the prosecution outcome vari-able. We also include a product term for lynch-ing and percent black to test for an interactioneffect. We note, however, that the statisticalinteraction may have different implications forthe respective dependent variables. When


sis of all counties in the nation, where one caseexceeds a value of 2,000. We removed that case fromthe analysis, but there is no meaningful change in thecoefficients and standard errors when it is included.

12 McVeigh and colleagues (2003:846), forinstance, suggest that hate crime reports “reflect dif-ferent incentives to call acts of bias to the attentionof local authorities, as well as different incentives thatinfluence law enforcement agents to respond to, andreport, hate crimes.”

13 The exact one-year period could vary slightlydepending on the date that the National ProsecutorsSurvey was actually completed, but this period cov-ers 12 months in late 2000 and early 2001 for alloffices.

14 We coded clustered counties 1 if any of thecounties were under the jurisdiction of a districtattorney’s office that prosecuted a hate crime. TheNPS data include some missing cases on the hatecrime prosecution measure, thus reducing our sam-ple size. We dropped counties in some states from thesupplementary analysis of prosecution in the nation-al sample; those cases and the justifications for drop-ping them are discussed in the Appendix.



assessing compliance, we expect the effect oflynching to become increasingly negative asthe percent black increases. For our measure ofanti-black reported hate crimes, where the maineffect of lynching is likely positive, we expectthis effect to be tempered in the presence of alarge black population because of suppressedreporting practices. We thus expect consisten-cy in the direction of the interaction coeffi-cients even if specific tipping points may differ.A negative interaction coefficient would be par-ticularly revealing compared with recent workon other types of social control. Whereas theinteraction between lynching and percent blackwas found to be positive for types of social con-trol that are punitive against blacks (Jacobs etal. 2005), we expect that the same interaction isnegative for social control perceived as protec-tive of blacks.

Our analysis also accounts for three indica-tors of political control. First, all models con-trol for the general level of politicalconservatism. For the policing dependent vari-ables, we measure political conservatism as thepercentage of a county or county cluster thatvoted for President George H. W. Bush in the1992 presidential election. This indicator pro-vides a standardized measure of conservatismsince all eligible citizens chose from the sameslate of presidential candidates. Because ourprosecution measure pertains to the years 2000to 2001, this set of analyses uses the percentagevoting for George W. Bush in the 2000 electionas a comparable indicator of conservatism.Second, we use the number of black electedcounty commissioners per 100,000 residentsas an indicator of black political representa-tion.15 Third, all models include a dummy vari-able indicating that the county, or at least onecounty in a cluster, had a black elected sheriffat some point between 1994 and 2002. These lat-ter two measures are reasonable indicators ofblack political control, or at least of an under-

lying propensity to elect blacks to lead localpolitical institutions, which could potentiallymediate the purported effects of black popula-tion size and lynching.

It is also imperative to control for character-istics of law enforcement agencies in the respec-tive counties. We control for a sheriff ’sdepartment’s propensity to comply with anydata collection mandates when modeling thecompliance dependent variable. Using county-level Uniform Crime Report (UCR) data fur-nished by the FBI, we employ two controlvariables to account for that propensity. First, wecoded a general crime reporting dummy variableas 1 if a sheriff ’s department submitted anycrime data in accordance with the UCR for1992 to 2003; we coded complete abstainers 0.Where counties are clustered, we coded thecluster as 1 if any of the sheriffs’offices report-ed crime information. Second, we include anordinal variable consisting of three categories:agencies that reported no data for the UCR(“abstainers”), those that reported some of thetime (“partial reporters”), and finally depart-ments that always reported general crime infor-mation to the UCR (“consistent reporters”).Our assumption is that this measure will accountfor a general proclivity to report crime infor-mation and should be strongly and positivelycorrelated with the compliance dependent vari-able. We can then assess the effects of lynchingand black population size on hate crime lawcompliance net of the propensity to report gen-eral crime data.16

Beyond these agency indicators, we alsomeasure the number of years that an agencyhas submitted crime data as part of the NationalIncident Based Reporting System (NIBRS).Because hate crime reporting is structurally apart of that system, there should be a positivecorrelation between participation in NIBRS andhate crime reporting. We calculate this variableby subtracting the year an agency began theNIBRS program from 2004 (agencies not par-ticipating are coded 0). In addition, we use datafrom the U.S. Department of Justice to measure


15 Specifically, we counted the number of blackelected commissioners in a county or county clusterfor each year between 1994 and 2002, standardizedthat measure by population size (in 1990), and thenlogged the variable to reduce extreme skew (we addeda constant of 1 before logging). Data for black com-missioners and black elected sheriffs were providedby the Joint Center for Political and EconomicStudies. Data are not available for 1992 and 1993.

16 As described below, we use the dichotomous“general crime reporting” variable only to predict“zero-counts” in our zero-inflated negative binomi-al model. We use the ordinal measure to predict theactual number of compliant quarters.



the number of officers per capita in the polic-ing agency (for the compliance outcome vari-able) or the number of police officers per capitain the county (for hate crime reporting) as con-trols for law enforcement capacity.17 It is plau-sible that larger policing agencies are betterable to absorb additional data collection respon-sibilities. Finally, in some models restricted tocounties (or clusters) with 500 or more blackresidents, we control for the percentage of blackpolice officers using data from the 1990 U.S.Census Equal Employment Opportunity File(ICPSR 9929). It seems reasonable to expectthat the policing of hate crimes against a racialminority group will positively correlate withthe representation of this group in law enforce-ment.

We also control for several other potentialcorrelates of hate crime policing and prosecu-tion.18 All analyses control for the logged pop-ulation of the county or county cluster, alongwith the percent urban. In addition, we accountfor the percent divorced and the percentage ofthe population age 15 to 29 because, in theory,these proxy measures of social disorganizationcould influence police investigation practices(Borg and Parker 2001). We also control forthe percentage of whites in the civilian laborforce who are unemployed and the percentageof whites below the poverty line because it isplausible that any effect of racial threat could beattributed to economic competition and adverseeconomic conditions for the majority group(data are from the 1990 and 2000 U.S.Censuses).

In some models restricted to counties withmore than 500 black residents, we further testwhether black/white unemployment and pover-ty ratios are consequential. We calculate theseratios as the percentage of the black populationunemployed divided by the percentage white

unemployed and the percentage of blacks inpoverty divided by the percentage of whites inpoverty, respectively. For each indicator, high-er values indicate that blacks do worse relativeto whites. An additional control variable meas-ures the percentage in a county (or cluster) thatwas born in the state of residence. This meas-ure is potentially important because culturaltraditions may change appreciably where thereis significant population turnover.

As mentioned above, we also include aproxy measure of interracial criminal violence,as indicated by the number of white-on-blackhomicides using data from Fox (2005).Following standard practices, this measurerefers to murders and non-negligentmanslaughters, as recorded by police in theFBI’s Supplementary Homicide Reports(SHR), that occurred in 1986, 1987, 1992,1993, 1994, and 1995. We use these years toaccount for the fact that police in Florida pro-vided no information to the FBI from 1988 to1991.19 We exclude homicides recorded bystate police departments because the county ofincident is often missing or may be incorrect.Furthermore, to create an unambiguous meas-ure of interracial homicide, we exclude inci-dents involving more than one victim oroffender (Williams and Flewelling 1988).20

Since both ethnic heterogeneity and the lega-cy of lynching predict interracial violence (onlynching, see Messner et al. 2005; on hetero-geneity, see South and Messner 1986) and for-mal social control (on lynching, see Zimring2003; on race, see Jackson 1989), a viablealternative hypothesis is that racial threat and


17 Data are missing for a small number of cases (sixin the southern sample and 68 in the national sam-ple). We used mean substitution for these cases.Additional models that omit these cases from analy-sis yield the same substantive results and are avail-able from the authors upon request.

18 Given the different time frames for our policingand prosecution dependent variables, we measurethe control variables at 1990 for policing and at 2000for prosecution, unless otherwise noted.

19 Agencies in Kentucky contributed no data to theSupplementary Homicide Reports in 1988. From1996 to 2003, agencies in Florida provided no datato the FBI, so we use the same 1986 to 1995 meas-ure of white-on-black homicides for the analysis ofprosecution. Finally, results presented below holdwhen white-on-black argument-related homicides,or those that occurred in the context of lovers’ trian-gles, alcohol-induced brawls, narcotics-inducedbrawls, arguments over money or property, or otherarguments, are controlled instead.

20 The homicide measure does not reflect Fox’sadjustment for missing data. In a personal commu-nication with one of the authors, Fox (January 2006)advised that his weighting scheme not be used incounty-level analyses.



the legacy of lynching increase interracial vio-lence, which in turn influences state respons-es to interracial crimes. We test the impact oflynching and race on our outcome variables netof our indicator of interracial violence.

When using hate crime prosecution as thedependent variable, we also control for the num-ber of hate crimes reported by police in a coun-ty (or cluster) for the year 2000. This controlvariable is important because police reportingof hate crimes is likely to be consequential forprosecution.

The policing and prosecution of hate crimescould also be influenced by state laws or direc-tives from state agencies, such as the respectiveattorneys general. To account for potential state-level influences, we include dummy variablesfor all states, with one state omitted as a refer-ence category. By fixing the state effect, weessentially investigate within-state variation inour outcome variables of interest.

METHODS

We use binary logistic regression to assess theeffects of our predictor variables on our dichoto-mous outcome variable—hate crime prosecu-tion. The respective measures of hate crime lawcompliance and police reports of racially-moti-vated hate crimes are continuous, but also high-ly skewed, and they each include multiple zeros.Ordinary Least Squares regression may be prob-lematic when the distribution is positivelyskewed and heteroskedasticity is present.However, Poisson-based estimators can providenonbiased estimates for positively skewed eventcounts (Osgood 2000). We use negative bino-mial models because overdispersion is presentin both variable distributions.

We estimate the models for hate crime lawcompliance using zero-inflated negative bino-mial regression because the distribution includesmany zeros and several of these cases are sys-tematically related to an agency’s general crimereporting practices (i.e., non–hate crime report-ing to the FBI). As Cameron and Trivedi (1998)spell out in detail, and Jacobs and Carmichael(2004) illustrate for the case of death sentences,the zero-inflated procedure is suitable for a dis-tribution with many zero-counts and wheresome cases are at reduced risk, or no risk at all,of experiencing an event, such as law enforce-ment agencies that generally do not report crime

data.21 The zero-inflated routine simultaneous-ly estimates two models: one on the likelihoodof a zero-count and another on the expectedcount of the outcome variable. When modelingzero-counts, we rely on a single control variablepresumed to be strongly correlated with theabsence of hate crime reporting—an agency’sgeneral crime reporting record. Agencies thatgenerally fail to report crime data are notexpected to report hate crime data, but for rea-sons that may have little to do with racial threatand the legacy of lynching.

We use both the negative binomial and thezero-inflated models for the analysis of anti-black hate crimes reported. The models yieldnearly identical substantive results, but the zero-inflated model is more appropriate for oneanalysis, as explained below.

RESULTS

We begin with an analysis of compliance withthe HCSA, measured by the number of quartersa department reported hate crime data for 1992to 2003 (Table 2). Model 1 reports estimates foradditive effects based on our zero-inflated neg-ative binomial model; Model 2 includes theinteraction term for lynching and racial com-position; and Model 3 adds race-specific meas-ures such as the black–white unemploymentand poverty ratios. Intuitively, the results inModel 1 indicate that the degree of hate crimelaw compliance increases with the degree ofgeneral crime reporting to federal authorities (b= .191), and the likelihood of reporting zero hatecrimes decreases substantially if an agencyreports general crime data (b = –4.441). Thenumber of compliant quarters also increaseswith the length of participation in the NIBRSprogram (b = .082). Our measure of law enforce-ment capacity (police officers per capita) is sig-nificantly and negatively correlated with theoutcome variable, which may appear counter-intuitive. Notably, the correlation is positiveand significant in models that omit the statedummy variables. One plausible interpretationis that such counties are more invested in puni-


21 These agencies would technically still be at risk,but prior research suggests the likelihood of report-ing hate crimes is highly dependent on general crimereporting (King 2007).




Table 2. Zero-Inflated Negative Binomial Regression Coefficients: Number of Quarters thatCounty Sheriffs’ Departments Submitted Hate Crime Reports Regressed on PredictorVariables

Model 1 Model 2 Model 3

Lynching of blacks –.005 .004 .004(.003) (.005) (.006)

Percent black –.0001 .0006 –.0007(.001) (.001) (.002)

White unemployment –.006 –.006(.008) (.008)

White poverty –.010* –.010*(.003) (.003)

Black–white unemployment ratio –.008(.019)

Black–white poverty ratio .042*(.019)

Logged population –.004 –.004 –.014(.022) (.022) (.026)

Divorce rate .004 .005 –.0007(.009) (.009) (.011)

Population ages 15 to 29 –.002 –.001 –.001(.004) (.004) (.005)

Percent urban .0001 .00004 .0003(.0008) (.0008) (.001)

Crime reporting (ordinal) .191* .189* .237*(.028) (.028) (.033)

Voting for Bush in 1992 –.002 –.002 .0005(.002) (.002) (.002)

White-on-black homicides –.001 –.002 –.0006(.003) (.003) (.003)

Percent born in state –.001 –.0008 –.0002(.001) (.001) (.001)

Years reporting NIBRS .082* .081* .062*(.016) (.016) (.020)

Percent black officers .003*(.001)

Police officers per capita –.073* –.071* –.099*(.026) (.026) (.036)

Elected black county commissioners –.004 –.004 –.0009(.010) (.010) (.011)

Black county sheriffs .025 .045 .036(.058) (.058) (.062)

Lynching of blacks � percent black –.0003* –.0003(.0001) (.0002)

Constant 3.909* 3.835* 3.572*(.261) (.263) (.331)

Zero-valueAny UCR reporting –4.441* –4.440* –4.145*

(.602) (.602) (.612)N 726 726 562�2 / df 892.41 / 25 896.95 / 26 699.05 / 27

Notes: The state dummy variables, which are included in the model but not shown in the table, are statisticallysignificant, indicating that state effects account for part of the variation in compliance. Many predictor variables,such as political conservatism (–), percent black (–), percent born in state (–), and population size (+) are signifi-cant when the state dummies are omitted from the model. Police force size is positive and significant when statedummies are omitted. Standard errors are in parentheses.* p ≤ .05 (all tests two-tailed).



tive social control and concomitantly less apt tocomply with civil rights–related laws.

The results for the lynching measure and itsinteraction with racial composition are of par-ticular theoretical interest. Net of the controlvariables, Model 1 shows that past lynching isnegatively associated with hate crime law com-pliance, albeit at a more permissive level of sta-tistical signif icance (b = –.005; p < .10,two-tailed). Counties and county clusters withfrequent lynchings in the past are less likely tocomply with the HCSA. This coefficient is mod-est with respect to magnitude; an increase of 10past lynchings yields an expected 5 percentdecrease in the number of compliant quarters(1 – e–.005*10).

Consistent with our interaction hypothesis,the coefficient for the product term (Model 2)suggests that the legacy of lynching is contin-gent on black population size. In that model, thelynching coefficient is not significantly differ-ent from zero when the black population is zero.However, the interaction coefficient shows thatthe effect of lynching on compliance becomessmaller and ultimately turns negative after theblack population surpasses about 13 percent,and it becomes increasingly stronger in the neg-ative direction thereafter. This contingent asso-ciation aligns with recent arguments that thelegacy of lynching is associated with lawenforcement outcomes only in the presence ofa sizeable black population (Jacobs et al. 2005).

Model 3 in Table 2 provides a test of robust-ness by incorporating additional race-specificmeasures. The interaction coefficient in Model3 remains significant, albeit at a more modestlevel of statistical significance (p < .10, two-tailed). This is based on a smaller N because thesample is limited to counties and clusters with500 or more black residents to permit the inclu-sion of control variables for black officers andmeasures of black–white economic inequality.The results in this model also indicate that com-pliance increases with the proportion of blackpolice officers in a county (b = .003). Net ofsuch race-specific measures, a main effectsmodel (without the interaction term) shows thatlynching is negatively and significantly corre-lated with compliance (model not shown).

Table 3 shows the negative binomial regres-sion coefficients for our second outcome vari-able, the number of anti-black hate crimesreported by police. Looking first at the main

effects model (Model 1), a few of the controlvariables are significantly associated withreported anti-black hate crimes. Police reportsuch crimes with greater frequency in morepopulous areas (b = 1.003), where the numberof police officers per capita is higher (b = .609),and where there is a sizeable young population(b = .06). Net of the control variables, and unlikethe previous analyses of compliance, the maineffects model (Model 1) in Table 3 indicates apositive association between past lynchings andreported hate crimes that target blacks (b =.028). This result is consistent with Messnerand colleagues’ (2005) finding that lynching ispositively associated with white-on-black argu-ment-related homicides, although at first glancethis finding appears to contradict our hypothe-sis.

Yet, Models 2 and 3 in Table 3 again point toan interaction between past lynchings and cur-rent black population size. The models suggestthat lynching has a positive effect on policereports of anti-black hate crimes where noblacks reside, but this effect diminishes as thepercent black increases, ultimately reversingdirection as the black population surpasses about40 percent. In Models 1, 2, and 3 in Table 3, wefind little support for our hypothesis that anti-black hate crimes are inversely related to pastlynching of blacks, but the direction and sig-nificance of the coefficient for the interactionterm is consistent with our hypothesis con-cerning racial threat and the legacy of lynching.

Table 3 raises a question that warrants addi-tional scrutiny: Why is the main effect for lynch-ing positive for anti-black hate crimes reportedby police, in apparent contradiction to one of ourpredictions? One possibility is that policingagencies in counties with a history of racialantagonism encounter offenses motivated byracial animus with some frequency. While theinclination in such counties may be to refrainfrom reporting these as hate crimes, some agen-cies may report them either on their own voli-tion or due to incentives such as state directivesor participation in NIBRS. If true, we wouldexpect a legacy of lynching to decrease the like-lihood of reporting any anti-black hate crimes,but to increase reports among counties report-ing one or more anti-black hate crimes.

To shed some light on this possible scenario,we turn to Model 4 in Table 3. This model pre-sents a zero-inflated negative binomial model





Table 3. Negative Binomial Regression Coefficients: Number of Racially-Motivated Hate Crimesthat Target Blacks Reported by Police, Regressed on Predictor Variables

Model 4Model 1 Model 2 Model 3 (zero-inflated)

Lynching of blacks .028* .079* .070* .038*(.012) (.025) (.025) (.012)

Percent black –.005 –.0006 –.002 –.002(.008) (.008) (.010) (.009)

White unemployment –.024 –.016 –.046(.051) (.051) (.053)

White poverty –.029 –.029 –.021(.020) (.020) (.021)

Black–white unemployment ratio .039(.101)


Logged population 1.003* 1.010* 1.006* .801*(.123) (.123) (.134) (.135)

Divorce rate –.067 –.059 –.050 –.062(.055) (.055) (.060) (.057)

Population ages 15 to 29 .060* .058* .066* .052*(.020) (.020) (.022) (.020)

Percent urban .005 .004 .0004 .006(.005) (.005) (.005) (.005)

Voting for Bush in 1992 .002 .004 .006 .002(.009) (.009) (.011) (.009)

White-on-black homicides –.063 –.064 –.045 –.046(.033) (.034) (.036) (.033)

Police officers per capita .609* .642* .736* .660*(.139) (.139) (.159) (.154)

Percent born in state –.002 –.0004 .0007 –.004(.007) (.007) (.008) (.007)

Percent black officers –.009(.007)

Elected black county commissioners .109 .113 .086 .010(.059) (.059) (.060) (.061)

Black county sheriffs .461 .572 .695* .553(.312) (.311) (.315) (.316)

Lynching of blacks � percent black –.002* –.002*(.0008) (.0008)

Constant –11.205* –11.778* –13.244* –8.809*(1.496) (1.513) (1.755) (1.656)

Zero-count—Logged population –3.354*

(1.146)—Lynching of blacks .158*

(.081)—Percent black .044

(.026)—Constant 29.560*

(10.463)N 725 725 561 725�2 / df 568.27 / 23 574.59 / 24 487.40 / 25 439.42 / 23

Notes: Dummy variables for states are included in the analysis but not shown in the table. Standard errors are inparentheses.* p ≤ .05 (all tests two-tailed).



with lynching and percent black predicting thelikelihood of a zero-count for hate crime report-ing (net of population size; bottom part of thetable), as well as the number of anti-black hatecrimes reported (top part of the table). Theresults are consistent with the previous account.The odds of reporting zero anti-black hatecrimes increase with each past lynching (b =.158). At the same time, when looking at polic-ing agencies that report one or more hate crimes,the frequency of reported hate crimes increas-es with each past lynching (b = .038).

We also acknowledge a plausible alternativeinterpretation of our f indings in Table 3.Research convincingly shows that hate crimesperpetrated against blacks are more frequentwhere the black population is small, presumablybecause whites are emboldened and seek toprotect traditionally “white turf ” (Green,Strolovitch, and Wong 1998; Lyons 2007). It fol-lows that a legacy of lynching could be associ-ated with current hate crime behavior againstblacks, and that this tendency is exacerbatedwhere the black population is small. The inter-action term in Model 2 is consistent with thisexplanation if one assumes that official statis-tics reliably measure the prevalence of hatecrime offending, and we suggest that the coef-ficients may partly reflect such an account. Still,three sets of findings support our interpretationof the coefficients. First, lynching is negative-ly correlated with compliance (Model 1 of Table2), a dependent variable that is not contingenton the level of offending. Second, a legacy oflynching increases zero-counts of anti-blackhate crimes (Model 4 in Table 3), which is con-gruent with our assumptions. And third, if thereporting data entirely reflect actual offending,and if blacks are more often hate crime victimswhere they lack numbers and power, then wewould expect significant and negative coeffi-cients for the respective main effects of per-cent black, black sheriffs, and blackcommissioners. Yet Model 1 in Table 3 showsthat percent black is not statistically significantand the latter two coefficients are actually pos-itive in direction.22

Table 4 shows logistic regression coeffi-cients for the last dependent variable, hatecrime prosecution. Several variables are asso-ciated with prosecution. As expected, a hatecrime prosecution is more likely where morehate crimes are reported by police (b = .103).We also find a negative association betweenpolitical conservatism (voting for Bush) andhate crime prosecution (b = –.036). Withrespect to the focal independent variables, theevidence is consistent with the racial threatperspective as articulated above; the odds of ahate crime prosecution decrease by about 4percent for each percentage increase in theblack population (100*[1 – e–.039]). Models 2and 3 in Table 4, however, show that the inter-action coefficient is in the predicted directionbut it is not statistically signif icant.23 Animportant limitation of the hate crime prose-cution variable is that it does not distinguishbetween racially-motivated hate crime casesand cases not entailing racial bias.

The preceding analyses are limited to theSouth, the region for which the lynching data areof the highest quality. As an additional test toincorporate a wider geographic area, we usedata on all counties from the NAACP’s catalogof lynching incidents from 1889 to 1918 to testour interaction models for each dependent vari-able. These results, shown in Table 5, are con-sistent with the analysis of the 10 southern


exhibit an interactive effect on crimes based on racialbias. Given the logic of this argument, there is no rea-son to expect an interaction between lynching andblack population size when modeling non–racially-motivated hate crimes. Indeed, in additional analysesnot shown here but available from the authors uponrequest, neither the main effect of lynching nor itsinteraction with percent black is statistically signif-icant when counts of non–racially-motivated hatecrimes are regressed on the same set of predictorvariables.

23 The interaction coefficient for percent blackand lynching is statistically significant and negativein alternative model specifications in which thedummy variables for states are omitted and the stan-dard errors are adjusted for clustering within states.In these models, the interaction effect is very simi-lar to that found in the analysis of hate crime law com-pliance in Table 2. It is likely that variation in statelaws and other state directives covary with both thelegacy of lynching and prosecution.

22 One additional point concerning the analysis ofanti-black hate crimes is noteworthy. Our argumentemphasizes racial antagonism and racial threat, andthus we predict lynching and racial composition to



states. In Models 1 through 5, which includesan analysis of hate crime prosecution, the inter-action coefficient is negative and statisticallysignificant (p value for the interaction term inModel 6 is .064). Bearing in mind our earliercautions about the quality of this national dataset, these results suggest that the joint impact oflynching and racial composition on law enforce-ment responses to hate crimes hold whenextending our sample beyond the South.

SUMMARY AND DISCUSSION

The results from our analyses of hate crimepolicing and prosecution are largely in line withour theoretical premise that the legacy of lynch-ing, which we use as an indicator of historicalracial antagonism and the state’s failure to pro-tect a minority group, is predictive of contem-porary law enforcement responses tohate-motivated crimes. Past lynching is nega-tively correlated with hate crime law compliance


Table 4. Logistic Regression Coefficients: Prosecution of a Hate Crime Case on Predictor Variables

Model 1 Model 2 Model 3

Lynching of blacks –.008 .045 .021(.029) (.058) (.061)

Percent black –.039* –.034* –.029(.015) (.015) (.018)

White unemployment .043 .043(.104) (.104)

White poverty –.074 –.073(.045) (.045)

Black–white unemployment ratio –.203(.147)


Logged population .343 .313 .393(.231) (.232) (.252)

Divorce rate –.087 –.085 –.126(.094) (.094) (.105)

Population ages 15 to 29 .037 .037 .035(.040) (.040) (.040)

Percent urban –.017* –.017 –.018(.009) (.009) (.009)

Voting for Bush in 2000 –.036* –.035* –.022(.016) (.016) (.020)

White-on-black homicides .011 .029 .043(.086) (.091) (.094)

Hate crimes reported in 2000 .103* .105* .103*(.051) (.051) (.051)

Percent born in state –.028* –.026* –.029*(.012) (.012) (.013)

Elected black county commissioners .065 .064 .066(.095) (.094) (.097)

Black county sheriffs .481 .538 .562(.545) (.549) (.556)

Lynching of blacks � percent black –.002 –.002(.002) (.002)

Constant 1.613 1.554 –.519(2.845) (2.859) (3.390)

N 674 674 538�2 / df 133.11 / 23 134.31 / 24 103.55 / 24

Notes: Dummy variables for states are included in the analysis but not shown in the table. Standard errors are inparentheses.* p ≤ .05 (all tests two-tailed).



by policing agencies (Table 2, Model 1) andincreases the likelihood of reporting no hatecrimes (Table 3, Model 4). Moreover, fortwo of the three dependent variables in analy-ses restricted to 10 southern states, and forfive of six analyses of the national sample ofU.S. counties, the interaction coefficient forpast lynching and our measure of currentracial threat is negative and significant. Inother words, a history of lynching in com-bination with a relatively large racial minor-ity is associated with lesser compliance with,and enforcement of, hate crime legislation.

We readily acknowledge several importantlimitations in this research. First, as dis-cussed earlier, a question that affects virtu-ally all work on hate crime offending and lawenforcement responses is whether the num-ber of reported hate crimes and the legalprocessing of these crimes actually reflect theprevalence of this type of behavior. Or, doreports and prosecutions of hate crimesreflect a willingness to enforce hate crimelaws, consistent with the constructivist per-spective? It is difficult to adjudicate betweenthese competing ideas without independentmeasures of offending and enforcement.Mindful of this difficulty, we examined hatecrime law enforcement via multiple outcomevariables and multiple institutions (policeand prosecutors), and by introducing proxymeasures of plausible intervening mecha-nisms to the extent possible. Still, while ourmeasures of reporting and compliancebehave in a manner consistent with a con-structivist interpretation of the outcome vari-ables, future work might attempt to replicateour findings with data on related outcomes,such as arrests, judges’ sentencing enhance-ments, and convictions. In addition, qualita-tive inquiry into policing agencies andprosecutors’ offices could shed direct lighton the activities and orientations of thosewho enforce hate crime laws (cf. Bell 2002),thereby enhancing the veracity of our inter-pretations.

Second, given the aggregate nature of thedata, we are unable to identify the specificactors involved in the enforcement of hatecrime legislation, the characteristics of theseactors, and the nature of the decision-mak-ing processes associated with the recordingand prosecution of hate crimes. A particularly


Tabl

e 5.

Coe

ffic

ien

ts f

or L

ynch

ing,

Per

cen

t B

lack

, an

d T

hei

r P

rod

uct

Ter

m f

or A

nal

ysis

of

all

U.S

. Cou

nti

es

(Mod

el 4

)(M

odel

2)

(Mod

el 3

)A

nti

-bla

ck

(Mod

el 6

) (M

odel

1)

Com

pli

ance

An

ti-b

lack

h

ate

crim

es(M

odel

5)

Pro

secu

tion

Com

pli

ance

>50

0 bl

acks

hat

e cr

imes

>50

0 bl

acks

Pro

secu

tion

>50

0 bl

acks

Cor

resp

ondi

ng m

odel

from

Tabl

e 2,

Tabl

e 2,

Tabl

e 3,

Tabl

e 3,

Tabl

e 4,

Tabl

e 4,

sout

hern

sam

ple

Mod

el 2

Mod

el 3

Mod

el 2

Mod

el 3

Mod

el 2

Mod

el 3

Coe

ffic

ient

s—

Lync

hing

of

blac

ks

.247

.242

3.49

9*2.

790*

5.18

2*3.

928

——

(ann

uali

zed)

(.16

1)(.

189)

(.86

5)(.

853)

(2.1

56)

(2.2

32)

—Pe

rcen

t bla

ck.0

008

–.00

02–.

001

.000

2–.

021*

–.01

5(.

0008

)(.

001)

(.00

5)(.

006)

(.01

0)(.

011)

—Ly

nchi

ng *

per

cent

bla

ck–.

013*

–.01

3*–.

093*

–.07

2*–.

184*

–.14

4(.

005)

(.00

6)(.

031)

(.02

9)(.

077)

(.07

8)N

2,80

11,

312

2,82

31,

326

2,56

11,

318

Not

es:A

naly

sis

incl

udes

dum

my

vari

able

s fo

r st

ates

and

con

trol

s fo

r al

l var

iabl

es in

the

corr

espo

ndin

g m

odel

s fr

om T

able

s 2,

3, a

nd 4

. Add

itio

nal d

ata

note

s ar

e in

clud

ed in

the

App

endi

x. S

tand

ard

erro

rs a

re in

par

enth

eses

.*

p≤

.05

(all

test

s tw

o-ta

iled

).



important issue for future research to investigateis how the individual attribute of race interactswith the features of the larger social context. Forexample, what political pressures and organi-zational constraints impinge upon black enforce-ment agents (police, sheriffs, and prosecutors)who work in settings characterized by deeplyingrained traditions of racial antagonism? Howdo such actors adapt to these pressures and, insome cases, overcome organizational constraintsthat might discourage vigorous enforcement ofhate crimes?

Finally, our indicator of lynching is ultimatelyan indirect proxy for an intervening mecha-nism—cultural traditions. We argue that lynchmobs’ actions in the past reflected the racialantagonism of the time and place and the state’sfailure to protect the predominant racial minor-ity, and this general cultural orientation has dis-sipated but not completely evaporated over time.Deeply ingrained traditions “die hard,” and wepropose that traces of the cultural sentimentthat permitted lynching linger into the presentand are manifest today via lax enforcement oflaws that deal with hate crimes. However, directmeasures of such latent cultural traditions andcontemporary cultural orientations are unavail-able. We also cannot tally the number of lynch-ings prevented by whites who viewed thepractice as contemptible, which could furtherelaborate our findings.24 Nevertheless, ourresults demonstrate a coherent series of rela-tionships involving lynching, racial composition,and multiple indicators of hate crime lawenforcement that align with our theoreticallygrounded account.

With these caveats in mind, the analyses fur-ther underscore other researchers’ claims abouthistorical continuity in the exercise of socialcontrol (Jacobs et al. 2005; Wacquant 2000;Zimring 2003). They also largely support Bobo

and Smith’s (1998) suggestion that race relationshave morphed from overt discrimination intoskepticism about laws that assume a protectiverole for racial minorities, in this case evidencedby our focus on blacks. Jim Crow racism haslargely ceased, but race and the history of racialantagonism remain vital for understanding vari-ation in state responses to crimes of bigotry.

Our finding that the effect of lynching onlaw enforcement responses to hate crimes iscontingent on current racial threat is also con-sistent with recent research on other facets ofsocial control, such as death sentences (Jacobset al. 2005). It appears, however, that this inter-play increases punitive actions that dispropor-tionately fall on minorities (see, e.g., Jacobs etal. 2005) while decreasing law enforcementprotective of minorities. In light of this, ourconclusion that the social control of intergroupconflict is in part a function of both currentracial threat and a cultural tradition of apathytoward the protection of minorities suggestspromising topics for future research on relatedlaws and policies. For example, our modelwould predict that reports of employment dis-crimination are less likely to be investigatedwhere past lynching episodes are numerous andthe current black population is relatively large.The results are also germane to other institutionsthat often assume a protective function forlower-class minority groups, such as indigentdefense systems. It would be consistent with ouraccount if the same counties that scarcelyenforce hate crime laws also provide less accessto effective defense counsel. These ideas rep-resent a sampling of research topics that mightbe informed by our model, and empirical workon such issues could further test the credibilityof our interpretations.

We close by emphasizing that this researchpartly corroborates prior work on state respons-es to hate crimes by further illustrating the socialconstruction of hate crimes (McVeigh et al.2003) and explaining the gap between writtenand practiced law (Jenness and Grattet 2005).We go beyond extant work by drawing theoret-ical and empirical attention to the cultural under-pinnings of state responses to crimes entailingbigotry. Although all states are “at risk” ofreporting hate crimes and complying with fed-eral law, substantial intraregional and intrastatevariation exists. Laws that remain dormant insome places are enforced, or complied with, in


24 While we cannot measure prevented lynchings,we did measure the continuation of lynching after theGreat Migration began and elites increasingly chal-lenged the practice (Tolnay and Beck 1995). We rees-timated the models using the number of lynchingswith black victims from 1916 to 1930 (as opposed tothe full 1882 to 1930 period), and the substantiveresults are consistent with those reported in the text.These results are available from the authors uponrequest.



APPENDIX

ADDITIONAL NOTES ON DATA

We made the following changes to the NAACP data, using SAS. The following two lines of codeassign county identifiers based on city identifiers:

if state = ‘Texas’ and city = ‘Hedsville’ then county = ‘ROBERTSON’;if state = ‘Texas’ and city = ‘Lang’ then county = ‘GILLESPIE’;

The following lines of code edit the NAACP county identifier to match that found in the Horanand Hargis (1995) template:

if state = ‘Illinois’ and county = ‘VERMILLION’ then county = ‘VERMILION’;if state = ‘Illinois’ and county = ‘SAINTCLAIR’ then county = ‘STCLAIR’;if state = ‘Indiana’ and county = ‘WARWICK’ then county = ‘WARRICK’;if state = ‘Maryland’ and county = ‘ALLEGHANY’ then county = ‘ALLEGANY’;if state = ‘Maryland’ and county = ‘ARUNDEL’ then county = ‘ANNEARUNDEL’;if state = ‘Maryland’and county = “PRINCEGEORGE’S” then county = ‘PRINCEGEORGES’;if state = ‘Missouri’ and county = ‘GREEN’ then county = ‘GREENE’;if state = ‘Missouri’ and county = ‘GREENLEE’ then county = ‘GREENE’;if state = ‘Michigan’ and county = ‘SAINTCLAIR’ then county = ‘STCLAIR’;if state = ‘Missouri’ and county = ‘ST.CHARLES’ then county = ‘STCHARLES’;if state = ‘Missouri’ and county = ‘SAINTLOUIS’ then county = ‘STLOUISCOUNTY’;if state = ‘Oklahoma’ and county = ‘MANNFORDCREEK’ then county = ‘CREEK’;if state = ‘Texas’ and county = ‘FT.BEND’ then county = ‘FORTBEND’;if state = ‘Texas’ and county = ‘BEAVER’ then county = ‘POLK’;if state = ‘Texas’ and county = ‘GARFIELD’ then county = ‘RUSK’;if state = ‘Virginia’ and county = ‘ACCOMAC’ then county = ‘ACCOMACK’;if state = ‘Virginia’ and county = ‘ALBEMARIE’ then county = ‘CHARLOTTESVILLECITY’;if state = ‘Virginia’ and county = ‘ALEXANDRIA’ then county = ‘ALEXANDRIACITY’;if state = ‘Virginia’ and county = ‘ALEGHANY’ then county = ‘ALLEGHANY’;if state = ‘Virginia’ and county = ‘CULPEPPER’ then county = ‘CULPEPER’;if state = ‘Virginia’ and county = ‘FARQUHAR’ then county = ‘FAQUIER’;if state = ‘Virginia’ and county = ‘HOUSELOUISA’ then county = ‘LOUISA’;if state = ‘Virginia’ and county = ‘MEEKLENBURG’ then county = ‘MECKLENBURG’;if state = ‘Virginia’ and county = ‘MERCER’ then county = ‘TAZEWELL’;if state = ‘Virginia’ and county = ‘NOTTAWAY’ then county = ‘NOTTOWAY’;if state = ‘Virginia’ and county = ‘PITTSYLVANIA’ then county = ‘DANVILLECITY’;if state = ‘Virginia’ and county = ‘ROANOKE’ then county = ‘ROANOKECITY’;if state = ‘Virginia’ and county = ‘WARWICK’ then county = ‘NEWPORTNEWSCITY’;if state = ‘Texas’ and county = ‘WENO’ then county = ‘COLORADO’;


others. We investigated such variation to shedlight on the social control of intergroup conflictand to illustrate the current consequences ofconflict-laden histories. History reveals clues tocontemporary criminal justice behavior, but afull understanding of such behavior requiresconsideration of current social structures incombination with cultural legacies.

Ryan D. King is an Assistant Professor of Sociologyat the University at Albany, State University of NewYork. His research focuses on crime, criminal law, andintergroup conflict. Current work in this vein exam-ines hate crime law, criminal deportations in U. S.history, and sentencing.

Steven F. Messner is Distinguished TeachingProfessor of Sociology, University at Albany, StateUniversity of New York. His research focuses onsocial organization and crime, the spatial patterningof crime, and crime and social control in China. Inaddition to his publications in professional journals,he is coauthor of Crime and the American Dream andPerspectives on Crime and Deviance, as well as co-editor of Theoretical Integration in the Study ofDeviance and Crime and Crime and Social Controlin a Changing China.

Robert D. Baller is an Associate Professor ofSociology at the University of Iowa. His researchexamines violence, features of social structure, andculture.



We did not form the following counties withlynching activity into clusters because the activ-ity occurred entirely after modern countyboundaries were in place. In Texas: Reeves,Valverde, Crockett, Kinney, Pecos, Schleicher,and Sutton. In Virginia: Charlottesville City,Albemarle, Pittsylvania, Danville City, Roanoke,Roanoke City, Salem City, Buchanan,Dickenson, Russell, Wise, Norton, ElizabethCity, Newport News City, Warwick, WarwickCity, and Hampton City. In Wyoming: Albany,Converse, Goshen, Laramie, Niobrara, andPlatte. In Arkansas: Cleburne, Independence,Van Buren, and White. In Florida: Monroe. InKentucky: Ballard and Carlisle. In NorthCarolina: Cumberland, Durham, Hoke, Orange,Robeson, and Wake. In South Carolina:Charleston. In Tennessee: Fentress, Overton,and Pickett.

We clustered the following counties to linkvarious data from 1990 and 2000. In Alaska, weformed the Yakutat borough, the Skagway-Hoonah-Angoon census area, and the Skagway-Yakutat-Angoon census area into one cluster.Also in Alaska, we clustered Denali, Yukon-Koyukuk, and Southeast Fairbanks. In Virginia,we clustered South Boston and Halifax, as wellas Alleghany and Clifton Forge City. InMontana, we clustered Yellowstone NationalPark (part), Gallatin, and Park. In Hawaii, weclustered Kalawao and Maui. Finally, in NewYork, we clustered Bronx, Kings, New York,Queens, and Richmond counties.

DATA CODING NOTES ON POLICING AND

PROSECUTION

We used sheriffs’ departments for the compli-ance outcome variable (Table 2). If there was no

sheriff’s department, but a county police depart-ment was available, we used the latter. If nocounty or reasonable substitute (with county-level jurisdiction) was available, we coded thecase as missing. Three states in our supple-mentary analysis of the national sample wereparticularly problematic: Connecticut, Alaska,and Virginia. In the former two states, sheriffs’offices were systematically missing because oflaw enforcement organization in those states,and we could identify no county policing agencyor a reasonable proxy. (None of these states arein our focal sample of 10 southern states.) ForVirginia, we used the municipal policingagency’s degree of compliance for county FIPScodes greater than 500. The issues describedabove are relevant only for Models 1 and 2 ofTable 5.

Table A1 lists additional special cases where,for reasons detailed in the table, we used anon–sheriff’s department policing agency. Withthree exceptions, these apply only to the sup-plementary analysis of compliance in thenational sample (Table 5, Models 1 and 2).

We aggregated counties coterminous withNew York City boroughs into one “New York”cluster.

Given the unique organization of districtattorneys’ offices in Connecticut (correspond-ing to county subunits) and no county-leveldata in the National Survey of Prosecutors forAlaska, we dropped these states from the analy-sis of prosecution in Models 5 and 6 in Table 5.

We had to drop several county clusters forModel 6 in Table 5 because the model is restrict-ed to counties with more than 500 blacks; hencethe much smaller N for that analysis.


We dropped NAACP lynchings with the following county identifiers because the identifiers appearto be wrong:

if state = ‘Missouri’ and county = ‘LIBERTY’ then drop = 1;if state = ‘Texas’ and county = ‘AVALON’ then drop = 1;if state = ‘Texas’ and county = ‘PASO’ then drop = 1;if state = ‘Texas’ and county = ‘SANPETE’ then drop = 1; *6 lynchings dropped;if drop ne 1;



REFERENCES

Abate, Frank R. 1974. American Places Dictionary.Detroit, MI: Omnigraphics, Inc.

Anderson, Elijah. 1997. “Violence and the Inner-City Street Code.” Pp. 1–30 in Violence andChildhood in the Inner City, edited by J. McCord.New York: Cambridge University Press.

———. 1999. Code of the Street: Decency,Violence, and the Moral Life of the Inner City.New York: Norton.

Baldus, David C., Charles Pulaski, and GeorgeWoodworth. 1983. “Comparative Review of DeathSentences: An Empirical Study of the GeorgiaExperience.” Journal of Criminal Law andCriminology 74:661–753.

Beck, E. M. and Stewart E. Tolnay. 1997. “WhenRace Didn’t Matter: Black and White MobViolence against Their Own Color.” Pp. 132–54 inUnder Sentence of Death: Lynching in the South,edited by W. Fitzhugh Brundage. Chapel Hill, NC:University of North Carolina Press.

Beckett, Katherine. 1997. Making Crime Pay: Lawand Order in Contemporary American Politics.New York: Oxford University Press.

Behrens, Angela, Christopher Uggen, and Jeff Manza.2003. “Ballot Manipulation and the ‘Menace ofNegro Domination’: Racial Threat and FelonDisenfranchisement in the United States,1850–2002.” American Journal of Sociology109:559–605.

Bell, Jeannine. 2002. Policing Hatred: LawEnforcement, Civil Rights, and Hate Crime. NewYork: New York University Press.

Black, Donald. 1971. “The Social Organization ofArrest.” Stanford Law Review 23:1087–1111.

Blalock, Hubert M. 1967. Toward a Theory ofMinority-Group Relations. New York: Jon Wileyand Sons.

Bobo, Lawrence D., James R. Kluegel, and Ryan A.Smith. 1997. “Laissez Faire Racism: TheCrystallization of a Kinder, Gentler Anti-Black

Ideology.” Pp. 15–44 in Racial Attitudes in the1990s: Continuity and Change, edited by S. A.Tuch and J. Martin. Greenwood, CT: Praeger.

Bobo, Lawrence D. and Ryan A. Smith. 1998. “FromJim Crow Racism to Laissez-Faire Racism: TheTransformation of Racial Attitudes.” Pp. 182–220in Beyond Pluralism: The Conception of Groupsand Group Identities in America, edited by W. F.Katkin, N. Landsman, and A. Tyree. Urbana, IL:University of Illinois Press.

Borg, Marian J. and Karen F. Parker. 2001.“Mobilizing Law in Urban Areas: The SocialStructure of Homicide Clearance Rates.” Law andSociety Review 35:435–66.

Brundage, W. Fitzhugh. 1993. Lynching in the NewSouth: Georgia and Virginia, 1880–1930. Urbana,IL: University of Illinois Press.

———. 1997. “Introduction.” Pp. 1–20 in UnderSentence of Death: Lynching in the South, editedby W. Fitzhugh Brundage. Chapel Hill, NC:University of North Carolina Press.

Cameron, A. Colin and Previn K. Trivedi. 1998.Regression Analysis of Count Data. New York:Cambridge.

Craig, Kellina M. 2002. “Examining Hate-MotivatedAggression: A Review of the Social PsychologicalLiterature on Hate Crimes as a Distinct Form ofAggression.” Aggression and Violent Behavior7:85–101.

Durkheim, Emile. 1973. Moral Education. New York:Free Press.

Forstall, Richard L. 1996. Population of States andCounties of the United States: 1790 to 1990 fromthe Twenty-one Decennial Censuses. Prepared bythe U.S. Department of Commerce, Bureau of theCensus, Population Division. Washington, DC:U.S. Government Printing Office; distributed bythe National Technical Information Service.

Fox, James Alan. 2005. Uniform Crime Reports[United States]: Supplementary Homicide Reports,1976–2003 [Computer f ile]. Compiled by


Table A1.—Non-Sheriff’s Department Policing Agencies

FIPS Location

12025 Miami-Dade, FL13189 Mcduffie, GA47037 Davidson, TN8031 Denver, CO

1101 Washington, DC

15003 Honolulu, HI29189 St. Louis Co., MO29510 St. Louis City, MO30023 Deer Lodge, MT

Note: FIPS refers to the Federal Information Processing Standards Code for the county.

Description

Miami-Dade County Police Department is used.County Policing Agency is used (ORI #GA09400).Nashville-Davidson Metro Police Department is used.Denver Municipal Police is used because county and city borders are essentially the

same.Omitted because state dummy variables are included, and thus more than one

agency within a unit (e.g., state) is required.Honolulu Police Department is used because no county-level agency is available.St. Louis County Police Department is used.St. Louis City Police Department is used.Deer Lodge County Law Enforcement is used.



Northeastern University, College of CriminalJustice. ICPSR04351-v1. Ann Arbor, MI: Inter-university Consortium for Political and SocialResearch [producer and distributor].

———. January 2006. Personal communication.Gannett, Henry. 1902. A Gazetteer of Texas.

Washington, DC: Government Printing Office.———. 1975. A Gazetteer of Virginia and West

Virginia. Baltimore, MD: Genealogical PublishingCompany.

Garland, David. 1990. Punishment in ModernSociety: A Study in Social Theory. Chicago, IL:University of Chicago Press.

———. 2005. “Penal Excess and Surplus Meaning:Public Torture Lynching in Twentieth-CenturyAmerica.” Law and Society Review 39:793–834.

Giles, Michael W. and Melanie A. Buckner. 1993.“David Duke and Black Threat: An Old HypothesisRevisited.” Journal of Politics 55:702–13.

Green, Donald P., Dara Z. Strolovitch, and Janelle S.Wong. 1998. “Defended Neighborhoods,Integration, and Racially Motivated Crime.”American Journal of Sociology 104:372–403.

Hawkins, Darnell F. 1987. “Beyond Anomalies:Rethinking the Conflict Perspective on Race andCriminal Punishment.” Social Forces 65:719–45.

Horan, Patrick M. and Peggy G. Hargis. 1995. CountyLongitudinal Template, 1840–1990. Ann Arbor,MI: Inter-university Consortium for Political andSocial Research.

Jackson, Pamela Irving. 1989. Minority Group Threat,Crime, and Policing: Social Context and SocialControl. New York: Praeger Press.

———. 1992. “Minority Group Threat, SocialContext, and Policing.” Pp. 89–101 in Social Threatand Social Control, edited by A. E. Liska. Albany,NY: State University of New York Press.

Jacobs, David and Jason Carmichael. 2001. “ThePolitics of Punishment across Time and Space: APooled Time-Series Analysis of ImprisonmentRates.” Social Forces 80:61–91.

———. 2002. “The Political Sociology of the DeathPenalty: A Pooled Time-Series Analysis.”AmericanSociological Review 67:109–31.

———. 2004. “Ideology, Social Threat, and theDeath Sentence: Capital Sentences across Time andSpace.” Social Forces 83:249–78.

Jacobs, David, Jason T. Carmichael, and StephanieL. Kent. 2005. “Vigilantism, Current Racial Threat,and Death Sentences.” American SociologicalReview 70:656–77.

Jacobs, David and Ronald Helms. 1996. “Toward aPolitical Sociology of Incarceration: A Time-SeriesExamination of Multiple Explanations for PrisonAdmission Rates.” American Journal of Sociology102:323–57.

———. 1999. “Collective Outbursts, Politics, andPunitive Resources.” Social Forces 77:1497–1524.

Jacobs, David and Robert M. O’Brien. 1998. “The

Determinants of Deadly Force: A StructuralAnalysis of Police Violence.” American Journal ofSociology 103:837–62.

Jenness, Valerie and Ryken Grattet. 2001. MakingHate a Crime: From Social Movement to LawEnforcement. New York: Russell Sage Foundation.

———. 2005. “The Law-in-Between: The Effectsof Organizational Perviousness on the Policing ofHate Crime.” Social Problems 52:337–59.

King, Ryan D. 2007. “The Context of Minority GroupThreat: Race, Institutions, and Complying withHate Crime Law.” Law and Society Review41:189–224.

Kitsuse, John I. and Aaron V. Cicourel. 1963. “ANote on the Uses of Official Statistics.” SocialProblems 11:131–39.

Klarman, Michael. 2000. “The Racial Origins ofCriminal Procedure.” Michigan Law Review99:48–97.

Liska, Allen E., Mitchell B. Chamlin, and Mark D.Reed. 1985. “Testing the Economic Productionand Conflict Models of Crime Control.” SocialForces 64:119–38.

Lyons, Christopher J. 2007. “Community(Dis)Organization and Racially Motivated Crime.”American Journal of Sociology 113(3):815–63.

McDevitt, Jack, Jennifer M. Balboni, Susan Bennett,Joan C. Weis, Stan Orchowsky, and Lisa Walbolt.2000. Improving the Quality and Accuracy of BiasCrime Statistics Nationally: An Assessment of theFirst Ten Years of Bias Crime Data Collection.Final Report to the Bureau of Justice Statistics.

McVeigh, Rory, Michael R. Welch, and ThoroddurBjarnason. 2003. “Hate Crime Reporting as aSuccessful Social Movement Outcome.” AmericanSociological Review 68:843–67.

Medoff, Marshall H. 1999. “Allocation of Time andHateful Behavior: A Theoretical and PositiveAnalysis of Hate and Hate Crimes.” AmericanJournal of Economics and Sociology 58:959–73.

Messner, Steven F., Robert D. Baller, and MatthewP. Zevenbergen. 2005. “The Legacy of Lynchingand Southern Homicide.” American SociologicalReview 70:633–55.

Messner, Steven F., Suzanne McHugh, and RichardB. Felson. 2004. “The Distinctive Characteristicsof Assaults Motivated by Bias.” Criminology42:585–618.

National Association for the Advancement of ColoredPeople. 1919. Thirty Years of Lynching in theUnited States, 1889–1918. New York: NationalAssociation for the Advancement of ColoredPeople.

Osgood, D. Wayne. 2000. “Poisson-Based RegressionAnalysis of Aggregate Crime Rates.” Journal ofQuantitative Criminology 16:21–43.

Paternoster, Raymond. 1984. “ProsecutorialDiscretion in Requesting the Death Penalty: A




Case of Victim-Based Racial Discrimination.”Law and Society Review 18:437–78.

Perry, Barbara. 2001. In the Name of Hate:Understanding Hate Crimes. New York:Routledge.

Pfeifer, Michael J. 2004. Rough Justice: Lynching andAmerican Society, 1874–1947. Urbana, IL:University of Illinois Press.

Quillian, Lincoln. 1996. “Group Threat and RegionalChange in Attitudes toward African Americans.”American Journal of Sociology 102:816–60.

Quinney, Richard. 1974. A Critique of Legal Order.Boston, MA: Little, Brown and Co.

Smead, Howard. 1986. Blood Justice: The Lynchingof Mack Charles Parker. New York: OxfordUniversity Press.

South, Scott J. and Steven F. Messner. 1986.“Structural Determinants of IntergroupAssociation: Interracial Marriage and Crime.”American Journal of Sociology 91:1409–30.

Sutton, John R. 2000. “Imprisonment and SocialClassification in Five Common-Law Democracies,1955–1985.” American Journal of Sociology106:350–86.

Tolnay, Stewart E. and E. M. Beck. 1995. A Festivalof Violence: An Analysis of Southern Lynching,1882–1930. Champaign, IL: University of IllinoisPress.

Tolnay, Stewart E., Glenn Deane, and E. M. Beck.1996. “Vicarious Violence: Spatial Effects inSouthern Lynching, 1890–1919.” AmericanJournal of Sociology 102:788–815.

Tonry, Michael. 1995. Malign Neglect: Race, Crime,

and Punishment in America. New York: OxfordUniversity Press.

Turk, Austin T. 1969. Criminality and the LegalOrder. Chicago, IL: Rand McNally.

Tuskegee Institute. 1932. The Negro Year Book.Tuskegee Institute, AL: Negro Year BookPublishing Company.

United States Department of Justice. 2002. HateCrime Statistics, 2002. Washington, DC:Government Printing Office.

United States Department of Justice, Bureau ofJustice Statistics. 2002. National ProsecutorsSurvey, 2001. [Computer file]. Conducted by theNational Opinion Research Center. ICPSR 03418-v1. Ann Arbor, MI: Inter-university Consortium forPolitical and Social Research.

Vold, George. 1958. Theoretical Criminology. NewYork: Oxford University Press.

Wacquant, Loic. 2000. “The New ‘PeculiarInstitution’: On the Prison as Surrogate Ghetto.”Theoretical Criminology 4:377–89.

Williams, Kirk R. and Robert L. Flewelling. 1988.“The Social Production of Criminal Homicide: AComparative Study of Disaggregated Rates inAmerican Cities.” American Sociological Review53:421–31.

Wisconsin v. Mitchell. 1993. 508 U.S. 47.Wyatt-Brown, Bertram. 1982. Southern Honor:

Ethics and Behavior in the Old South. Oxford,UK: Oxford University Press.

Zimring, Franklin E. 2003. The Contradictions ofAmerican Capital Punishment. New York: OxfordUniversity Press.




The online version of this article can be found at: http...

Documents

Transcript of The online version of this article can be found at: http...