Who Survives on Death Row? An Individual and Contextual ... · Who Survives on Death Row? An Individual and Contextual Analysis David Jacobs Zhenchao Qian The Ohio State University

What individual and contextual factorsexplain what happens to offenders who

have been sentenced to death? An executionclearly is the most severe legal punishment,but no systematic research on the combinedpolitical and individual factors that determine

which death row inmates will be executedapparently exists. A few studies describe whathappens to these offenders after sentencing(Aarons 1998; Liebman, Fagan, and West2000), yet there are almost no systematicinvestigations on the most important deter-minants of executions. Studies by Spurr (2002)and Blume and Eisenberg (1999) are partialexceptions, but these studies focus only onoffender characteristics and ignore environ-mental conditions. Although studies repeatedlyshow that victim race is the most importantdeterminant of death sentences, we do notknow if this factor influences the fate ofoffenders on death row because no investiga-tions have determined whether this accountexplains executions.

The factors that influence execution prob-abilities are of interest partly because there aresuch substantial disparities in this outcome.Largely as a result of the appeals process thatoccurs after a death sentence, less than 10percent of all offenders on death row ulti-

Who Survives on Death Row? An Individual and Contextual Analysis

David Jacobs Zhenchao QianThe Ohio State University The Ohio State University

Jason T. Carmichael Stephanie L. KentMcGill University Cleveland State University

What are the relationships between death row offender attributes, social arrangements,

and executions? Partly because public officials control executions, theorists view this

sanction as intrinsically political. Although the literature has focused on offender

attributes that lead to death sentences, the post-sentencing stage is at least as important.

States differ sharply in their willingness to execute and less than 10 percent of those

given a death sentence are executed. To correct the resulting problems with censored

data, this study uses a discrete-time event history analysis to detect the individual and

state-level contextual factors that shape execution probabilities. The findings show that

minority death row inmates convicted of killing whites face higher execution

probabilities than other capital offenders. Theoretically relevant contextual factors with

explanatory power include minority presence in nonlinear form, political ideology, and

votes for Republican presidential candidates. Inasmuch as there is little or no systematic

research on the individual and contextual factors that influence execution probabilities,

these findings fill important gaps in the literature.

AMERICAN SOCIOLOGICAL REVIEW, 2007, VOL. 72 (August:610–632)

#3154-ASR 72:4 filename:72406-jacobs page 610

Direct correspondence to David Jacobs,Department of Sociology, 300 Bricker Hall, 190North Oval Mall, The Ohio State University,Columbus, OH 43210 ([email protected]). Wethank Douglas Berman for his valuable advice oncriminal procedure law applied to the death penaltyand Ruth Peterson for her comments. We are indebt-ed to Dan Tope for his research assistance. We alsothank Ohio State colleagues for their comments inpresentations at the law school, the political sciencedepartment, the Criminal Justice Research Center,colleagues at the University of Cincinnati School ofCriminal Justice, the editors, and the referees. All dataused in this study and in the analyses discussed in thetext but not shown are available on request. Thisresearch was supported by NSF grant #0417736.

mately are executed (Liebman et al. 2000).1

There is great variation in the duration of thisprocess as well, apparently because death penal-ty states differ sharply in their willingness toexecute. In many reluctant jurisdictions capitaloffenders can spend well over two decades ondeath row, but other states execute in far lesstime. To assess the determinants of these legaldecisions about who will live and who will die,we test theoretically based explanations usingan event history approach to discover the fac-tors that influence post-death sentence executionlikelihoods.

Trial court studies on the offender attributesthat lead to death sentences show that offend-ers who kill whites are far more likely to receivethis sentence (Baldus and Woodworth 2003;Dodge et al. 1990; Paternoster 1991). Yetwhether victim race continues to explain thefate of condemned prisoners after they havebeen sentenced remains a complete mystery.There are good reasons to think that this factorwill continue to matter. Yet it is equally plausi-ble that the appellate court decisions that large-ly determine death row outcomes are unaffectedby this consideration that, of course, should notbe relevant. In any event, both theoretical con-siderations and concerns about equity makethis relationship between victim race and exe-cution probabilities a critical issue.

This article will provide important evidenceabout whether the death penalty is adminis-tered impartially. But our primary goal is torefine theories of punishment by using a com-prehensive approach to gauge the explanatorypower of both individual and contextual effects.It is unlikely that judges and the political offi-cials who decide which death row inmates will

be executed are unaffected by their politicalenvironment, especially because this punish-ment is such an intensely moral issue. In partbecause some death row offenders face far lowerexecution probabilities than those in less lenientjurisdictions (Liebman et al. 2000), we gauge theeffects of the sociopolitical environment as wellas offender and victim attributes.

There are strong reasons for such a com-bined approach. Two isolated traditions havecoexisted in the literature. Many studies useindividual data to explain trial court sentencing,but others rely on aggregate data to study addi-tional criminal justice outcomes. State or nation-al attributes have been used to explain shifts inincarceration rates (Jacobs and Carmichael2001; Jacobs and Helms 1996; Stucky, Heimer,and Lang 2005; Sutton 2000; Western 2006).The urban conditions that explain police depart-ment size (Jacobs 1979; Kent and Jacobs 2005),arrest rates (Brown and Warner 1992), or the useof deadly force by the police (Jacobs andO’Brien 1998) have been researched as well. Yetexcept for a few studies that assess how com-munity and individual determinants combineto affect the sentencing of non-capital offend-ers (Helms and Jacobs 2002; Myers and Talarico1987), there is little research on the combinedeffects of individual and contextual determi-nants. But the many results based on aggregatedata showing that context is a strong determi-nant of multiple criminal justice outcomes makeit difficult to believe that such environmentalfactors do not influence post-sentencing deci-sions about executions.

We therefore use an integrated theoreticalapproach that emphasizes political explanationsand the racial accounts in earlier conflict stud-ies (Turk 1969). An execution is an intrinsical-ly political act. Foucault (1977) views executionsas rituals designed to enhance political power byreminding potential miscreants of the state’svast coercive resources. But there are more con-crete reasons for studying political effects. Mostresearchers who first tested conflict explanationshypothesized that larger and therefore morethreatening minority populations would increasesupport for repressive law and order measures.Citizens threatened by expansions in a minori-ty presence often react by demanding harshcriminal justice policies. Because criminal jus-tice agencies are operated by the state, this pres-sure has to be directed at political officials. By

WHO SURVIVES ON DEATH ROW?—–611


1 Local trial courts sentence, but appeals are han-dled by higher state and federal courts. The first twocapital appeals typically are decided by state appel-late courts. Condemned offenders then can petitionthe federal courts. About 41 percent of all death sen-tences are reversed on first state appeal and about 9.5percent are reversed in the second. About 40 percentof those who then seek federal relief are successful(Liebman et al. 2000). Most of the remainder are exe-cuted. But a few receive executive clemency, somedie before execution, and a few are removed fromdeath row for miscellaneous reasons. Almost all ofthe condemned who obtain appellate relief are resen-tenced to long prison terms.

assessing the political factors that result fromadded demands for this severe punishment, weseek to broaden the conflict approach to pun-ishment.

This article therefore offers the promise offilling many important gaps in the sparse liter-ature on post-death sentence execution proba-bilities by using a survival analysis that adjustsfor censoring and assesses both offender andpolitical characteristics. The multiple advan-tages that result from the inclusion of both indi-vidual and contextual factors suggest that thisanalysis will provide an accurate picture of thepost-sentencing death penalty process. Resultsbased on models that assess many explanationsare most accurate (Johnston 1984, see note 11),but such an inclusive approach means that thetheoretical section cannot focus on only a fewexplanations.

THEORY

Inasmuch as scholars claim that race continuesto have powerful effects on U.S. politics(Goldfield 1997; Jacobs and Tope 2007; Key1949), and because the administration of thedeath penalty is such an intense political issue,racial politics provide the primary conceptualbasis for this analysis. Most research on thedeath sentence focuses on the race of individ-ual offenders and their victims. We assess suchmicro minority accounts by gauging theexplanatory power of various offender-victimminority-majority combinations, but we fillcontextual gaps in the literature by analyzing theinfluence of minority threat in the political envi-ronments in which these decisions are made.Ideology also should matter as this penalty issuch an intensely moral issue and since westudy it in the most direct of all large democra-cies. In contrast to other nations, U.S. politiciansroutinely encourage citizens to vote on the basisof their views about capital punishment. Anotherclosely related account suggests that theparochial interests of politicians influence exe-cutions. Tactical rhetoric that stresses the deprav-ity of a minority underclass and the need forharsh measures should affect support for thispenalty (Beckett 1997) and its probability.

We present the theoretical foundation forfour interrelated proposition sets. First, we dis-cuss the conceptual basis for micro-level racialhypotheses derived from the literature on sen-

tencing. Second, we present theoretically basedhypotheses about the contextual effects ofminority threat, political ideology, and parti-san politics. At the end of this section, we pre-sent justifications for additional explanations forexecution probabilities.

INDIVIDUAL EXPLANATIONS: EXTRA-LEGAL

MICRO RACIAL ACCOUNTS AND LEGAL

FACTORS

If penal measures are best explained by theirinterrelationships with other social arrange-ments rather than by their alleged legal pur-poses (Garland 1990:91), it would be surprisingif the most important U.S. social division did notinfluence the administration of the most severelegal punishment. A conceptual grounding inracial politics therefore should provide the besttheoretical foundation for a study that analyzesthe administration of the death penalty. In the-oretical essays Wacquant (2000, 2001) usesarguments about social impurity and taboo toshow how the Jim Crow racial caste systempersists in socially altered but less conspicuousforms in the contemporary U.S. criminal justicesystem. To discover if such racial considera-tions still affect decisions about legal punish-ments, the question that motivates almost all ofthe many sentencing studies is whether the trialcourts treat minorities the same as nonminori-ties (Chiricos and Crawford 1995; Walker,Spohn, and DeLone 1996; Zatz 1987).

Findings about non-capital sentences, andthose from the less ample literature on the fac-tors that account for death sentences, shouldprovide the best available empirical basis forselecting the individual factors that explainwhich condemned prisoners will be executed.Despite their number, only a bare majority of thenon-capital sentencing studies show that thetrial courts give harsher sentences to minorities,but almost as many careful investigations do notfind such biases (Chiricos and Crawford 1995;Walker et al. 1996). Studies on the links betweenoffender race and death sentences are even lesslikely to suggest that white offenders are treat-ed with greater leniency than minorities (Baldusand Woodworth 2003; Dodge et al. 1990;Paternoster 1991). In light of the systematicdiscrimination faced by African Americans,however, we follow precedent and hypothesizethat: The likelihood of executions should be

612—–AMERICAN SOCIOLOGICAL REVIEW


greater for African Americans on death row. Insome jurisdictions Hispanics are the most threat-ening minority that faces severe discrimina-tion. It is equally plausible that: Hispanics ondeath row should face higher execution proba-bilities than whites.

Offenders convicted of killing whites are par-ticularly likely to be sentenced to death (Baldusand Woodworth 2003; Dodge et al. 1990;Paternoster 1991), especially if the offender isblack (Baldus and Woodworth 2003). Wacquant(2000) provides one theoretical foundation forthis persistent association when he claims thatharsh legal punishments continue to be used tomaintain the “symbolic distance needed to pre-vent the odium of ‘amalgamation’with [minori-ties] considered inferior, rootless, and vile” (p.380). The ultimate symbolic assault on such acaste system occurs when an underclass minor-ity kills a white. In this research we discover ifthis account that explains death sentences sowell also explains execution probabilities.

Two causal paths seem most plausible. Incontrast to the great majority of homicides,media outlets focus on interracial murders, par-ticularly if the victim is white (Bandes 2004,Lipschultz and Hilt 2002). This increased cov-erage is critical as victories in well-publicizedcapital trials often help politically ambitiousprosecutors reach higher office. But such tri-umphs must be protected. Prosecutors there-fore have strong reasons to vigorously opposecapital appeals that may jeopardize legal victo-ries that are likely to further their politicalcareers.2 Even if other facilitative conditionsare absent, the media’s focus on those murdersin which a white was killed by a minority putsadded pressure on state appellate court justicesto rule against these death row offenders whenthey appeal (Liebman, Fagan, and West 2002).

And disregarding these pressures can be cost-ly. State judges who ignored intense public sup-port for particular executions and grantedappellate relief to such petitioners have losttheir seats in retention elections, even thoughonly the incumbent appears on retention elec-tion ballots (Brace and Hall 1997; Bright andKeenan 1995). Hence: African American orHispanic offenders convicted of murdering awhite should be less likely than other offendersto avoid the death chamber.

Most studies of the determinants of non-cap-ital sentencing that assess the effects of genderfind that in comparison to females, trial courtsgive less lenient sentences to males (Bickle andPeterson 1991). This finding may be partiallybased on a failure to control for the ways offend-ers participated in their crimes. Women con-victed of robbery, for instance, often do notengage in the violence associated with thisfelony. Inasmuch as their involvement is lesspernicious, such offenders receive lighter sen-tences than their male associates. It neverthelessis reasonable to expect that chivalrous inclina-tions should reduce female execution probabil-ities, so: Women on death row should be lesslikely to be executed than males. Finally, find-ings show that offenders with prior convictionsare sentenced more severely by trial courts. Thispractice is reasonable as repeat offenders arelikely to pose a greater threat to the communi-ty after their release or to guards and otherinmates while they are imprisoned. Hence:Execution probabilities will be greater for thosecondemned prisoners with previous convictions.

CONTEXTUAL EXPLANATIONS: RACIAL

THREAT, POLITICAL IDEOLOGY, AND

PARTISANSHIP

The multiple decisions that ultimately producean execution do not occur in a social vacuum.A few studies of sentencing decisions for non-capital crimes productively treat the trial courtsas complex organizations. Yet analyses of con-textual forces external to organizations havesharply increased our understanding of organi-zational behavior (Perrow 1986; Scott 1987).The multiple findings based on aggregate datathat provide such robust explanations for crim-inal justice outcomes and the strong organiza-tion-environment relationships so oftenuncovered in the organizational literature point



2 After a death sentence, local prosecutors canhave important effects on the subsequent appeals. Forexample, their delays in filing for an execution dateslow appeals. In fact, any failure to act promptlyfavors condemned offenders as delayed appeals arenot as likely to be vigorously contested. For this andother reasons, the local prosecutor who won an ini-tial death sentence verdict often plays an importantrole in resisting subsequent appeals, although this rolemay be informal. Hence, prosecutor commitment toefforts to resist death row appeals should influenceexecution probabilities.

in the same direction. Both research streamssuggest that legal decision makers who areembedded in political environments do notignore such conditions when they decide if anexecution will occur.

RACIAL MIX. The fierce U.S. disputes aboutrace in the past probably make this fissure themost resilient and influential division in con-temporary U.S. politics (Goldfield 1997; Jacobsand Tope 2007; Key 1949). A majority’s eth-nocentric views and that group’s inclination toview minorities as trespassers enhance such agroup’s presumption that they should retainexclusive claims over important rights and priv-ileges (Blalock 1967; Blumer 1958; Bobo andHutchings 1996). Hostility and entrenchedbeliefs about a majority’s “rightful” positionare solidified by the political struggles thatoccur when minority groups seek to alter thesearrangements (Blumer 1958). According tothreat theorists, when large minority popula-tions endanger their dominance, whites oftenreact by supporting law and order measures thatat least indirectly target these minorities.

Findings are supportive. Racist views aremore widespread in cities with more black res-idents (Fosset and Kiecolt 1989; Quillian 1996;Taylor 1998). An enhanced minority presenceproduces added votes for anti-minority candi-dates (Giles and Buckner 1993; Giles and Hertz1994; Heer 1959) who are likely to endorseharsh criminal punishments. With crime ratesheld constant, Liska, Lawrence, and Sanchirico(1982) and Quillian and Pager (2001) find thatfear of crime is greater in cities or neighbor-hoods with more black residents. Larger minor-ity populations lead to additional police officers(Jacobs 1979; Kent and Jacobs 2005). Otherfindings show that the death penalty is likely tobe legal in states with the highest percentagesof African American residents (Jacobs andCarmichael 2002), while the number of deathsentences is greater in states with the largestAfrican American populations (Jacobs,Carmichael, and Kent 2005).

These results suggest that severe punishmentswill increase in areas after a growth in minori-ty presence. Yet if minority proportions expandand their political influence becomes sufficient,the positive relationship between this threat andpunitiveness may reverse. All governors, almostall local prosecutors, and most state appellate

justices are elected. Expansions in AfricanAmerican or Hispanic proportions past a thresh-old should give these minorities enough votesto influence decisions about executions. Hence:The relationship between the percentage ofblacks and execution probabilities should bepositive if minority presence is modest, but afterthis percentage reaches a threshold, this asso-ciation should become negative and executionsshould diminish. For this sign reversal to occur,a minority need not outnumber other groups.Minority size need only reach the point wheretheir votes may help decide elections, and thisproportion can be modest if other voting blocsare evenly matched. This logic and nonlinearfindings about death sentence frequency (Jacobset al. 2005) suggest that an inverted U-shapedrelationship between African American pres-ence and executions will be present.

The relationship between Hispanic presenceand executions may take a different nonlinearform. In many non-southwestern states, the pro-portion of Hispanic residents is minute. Themedian percentage of Hispanics was 2.4 percentin the sampled states, yet the same statistic forblacks was 6.8 percent or 2.9 times greater. Asthe Hispanic population was so modest in somany states, it is plausible that this populationmust reach a threshold size before whites seeHispanics as sufficiently threatening. Hence:The nonlinear relationship between the per-centage of Hispanic residents and executionprobabilities should become increasingly pos-itive only after the comparative size of this eth-nic minority reaches a level sufficient to threatenmajority Anglos.

POLITICAL IDEOLOGY. Claims that ideologyhelps shape legal penalties are especially com-pelling when U.S. sanctions are at issue, as thisnation is such an exceptionally direct democracy(Savelsberg 1994; Whitman 2003). The result-ing close voter control over criminal punish-ments, which are decided by bureaucraticexperts in the less direct European democracies,ought to make mass ideologies crucial whensuch intensely moral decisions about who willdie must be made. Images of evil and the result-ing beliefs about the most appropriate punish-ments that stem from these assumptions abouthuman nature are a foundational component ofpolitical ideologies. Conservatives often seecrime as resulting from freely made but amoral



choices (Lacey 1988). If such presumptions arecorrect, increases in expected costs should beeffective. Many conservatives therefore stressthe irreversibility and the deterrent effects ofexecutions. Such views about the efficacy ofincapacitation and deterrence provide the basisfor empirically dubious conservative claimsthat a few executions will protect many innocentvictims from criminal brutality.3

Liberals instead see criminal acts as imposedby circumstances (Garland 2001; Thorne 1990).These acts result from noxious environmentalconditions such as poverty or discrimination.Liberals view policies that alleviate these inju-rious conditions (Taylor, Walton, and Young1973) and therapeutic efforts to resocializeoffenders (Garland 2001) as the best remedies.In contrast to conservatives, liberal surveyrespondents are far less likely to support thedeath penalty (Lakoff 1996; Langworthy andWhitehead 1986). Findings corroborate theseclaims as they show that the most liberal statesare unlikely to legalize this punishment (Jacobsand Carmichael 2002). It follows that:Executions should not be as probable whereliberal views predominate, as prosecutors andappellate justices should be less likely toendorse this penalty in these jurisdictions.

POLITICAL PARTISANSHIP. Because they seekoutcomes that help the prosperous, politicalparties closer to the right face election obstacles.These parties, for example, usually choose taxpolicies that benefit their affluent core sup-porters at the expense of the less affluent (Allenand Campbell 1994). Yet prosperous voters arein the minority. Because such parties have asmaller voter base than their rivals, conservativesoften use law and order claims to appeal to less

affluent voters who are more likely to be crimevictims and who often live where such risksare greater. Officials in the Nixon and firstBush campaigns for the presidency admit thatthey emphasized this issue to attract anti-minor-ity voters.4 By focusing on street crime andother social problems readily blamed on under-class minorities, Republicans won elections byusing this “wedge” issue to gain sufficient votesfrom less prosperous citizens. Multiple findingsshow that Republican (Jacobs and Carmichael2001; Stucky et al. 2005; Western 2006) or con-servative political strength (Sutton 2000) ledto severe criminal justice outcomes. Becausecapital punishment has been an important issuein many state political campaigns (Constanzo1997) and because Jacobs and Carmichael(2002) find that this punishment is likely to belegal in states with the strongest Republicanparties, we expect that greater Republican polit-ical strength in a state should increase executionprobabilities.

Republican campaigns for the presidencyhave relied on and probably accentuated(Beckett 1997) mass perceptions about the linksbetween purportedly venal underclass life stylesand lawlessness. Findings show that votes for thef irst of these presidents who successfullyexploited public views about the linkagesbetween race and crime (Nixon in 1968) helpexplain how quickly states relegalized capitalpunishment after the 1976 court decisions that



3 Careful reviews of the multiple empirical stud-ies on this issue conducted by legal scholars (Zimringand Hawkins 1986), criminologists (Hood 1998;Paternoster 1991), sociologists (Bailey and Peterson1999), and economists (Donohue and Wolfers 2005;Levitt 2002) conclude that the death penalty has nodiscernable general deterrent effects beyond thoseimposed by long prison terms. This list of skepticsabout the deterrent effects of executions includes ascholar (Levitt 2002) who has published multiplefindings showing that imprisonment and other poli-cies designed to control crime are effective deterrents.

4 A participant described Nixon’s 1968 campaign:“We’ll go after the racists. That subliminal appeal tothe anti-black voter was always present in Nixon’sstatements and speeches” (Ehrlichman 1982:233).Other vivid examples occurred in the 1988 Bushcampaign against Dukakis. Republicans ran an adver-tisement declaring, “‘Dukakis not only opposed thedeath penalty, he allowed first-degree murderers tohave weekend passes from prison.’.|.|. [as the] clear-ly black [offender]—Willie Horton stared dully intothe camera.” They next released an advertisement fea-turing a victim. “‘Mike Dukakis and Willie Hortonchanged our lives forever .|.|. Horton broke into ourhome. For twelve hours, I was beaten, slashed, andterrorized. My wife Angie was brutally raped’”(Carter 1996:76–77). This emphasis did not abate. Ina House debate in 1994 “29 Republican[s] .|.|. spokederisively about midnight basketball .|.|. character-izing the program as ‘hugs for thugs’” (Hurwitz andPeffley 2005:99–100).

forced changes in these statutes (Jacobs andCarmichael 2002). Incarceration rates also werehigher after more voters supported a Republicanlaw and order presidential candidate (Weidnerand Frase 2003). Hence: Death row inmatesshould be less likely to avoid the death cham-ber in states in which more voters supportRepublican presidential candidates becauseprosecutors, appellate justices, and governorswill face stronger pressures to allow executionsin these jurisdictions.

ADDITIONAL CONTROLS

Murder is the most threatening crime. Executionprobabilities therefore should be greater in stateswith higher murder rates. Following Durkheim’semphasis on the determinants of restitutive law,states with a larger and more diverse populationand an enhanced division of labor should not beas likely to use this most severe punishment.This perspective also suggests that solidarityshould matter. Migration interferes with socialcohesion. Outsiders inspire hostility and fear, buttheir absence strengthens bonds and intragroupempathetic feelings (Hale 1996). Citizens instates with few outsiders therefore should not beas willing to support executions. This factoraccounts for the legality of the death penalty(Jacobs and Carmichael 2002), so executionsshould be less likely when most residents wereborn in the states where they now live.

Research on imprisonment has focused on theMarxist view that punishment is used to controlthe supply of labor (Rusche and Kirchheimer1939). Many researchers have assessed the linkbetween unemployment and incarceration, but theresults are mixed. A review (Chiricos and Delone1992) shows that about 60 percent of the 147associations between unemployment and impris-onments are significant. Despite such mixedresults, we expect that high joblessness willenhance execution probabilities because the pros-perous may view the unemployed as a threat orbecause high unemployment enhances resent-ments against criminals and accentuates demandsfor harsh sanctions. Yet the unemployment ratemay have to reach a threshold before it matters,so we test a nonlinear relationship between thisvariable and execution probabilities. Primarily asa result of the multiple state and federal appeals,expensive expert testimony, and the other costsrequired for due process in decisions that may end

a life, executions are far more expensive thanalternatives such as life imprisonment. Stateswith a superior tax base should be more likely touse this expensive punishment. Finally, to see ifthe unique arrangements in the South influencedeath row outcomes, we also control for thisregion.

METHODS

ESTIMATION

Our aim is to discover how offender attributesand the political and social characteristics of thestates affect post-sentencing execution likeli-hoods. To test these hypotheses, we use an eventhistory approach. This procedure provides aremedy for the censoring that occurs becauseoffenders face different execution risks. Someoffenders remained on death row after 2001, sothis right-censored group was in danger of beingexecuted after the end of the observation peri-od. The other right-censoring occurs whenoffenders were removed from death row large-ly as a result of successful appeals. Event his-tory analysis uses the information from caseswith incomplete duration who were not exe-cuted to avoid the bias that would occur if thesecensored cases were not included.5 We employdiscrete-time logit models to predict executionprobabilities. Our first model will assess theinfluence of individual factors. This model takesthe form:

Pijtlog ( 1 – Pijt) = �td +

M

�m=1

�mXmi (1)

where Pijt is the conditional probability of exe-cution for death row offender i in state j at time



5 Because event history analysis takes advantageof all available information, we include death rowoffenders near the end of the analysis period althoughsuch offenders are not as likely to be executed.Offenders still on death row are right-censored regard-less of how long they have been on death row, but ourestimation approach will not be biased by this orother forms of censoring. Note that selection biasesshould not be problematic because we only explainwhat happens to offenders sentenced to death. Wemake no claims about whether those who received aless severe punishment than death would have dif-ferent likelihoods of being executed if they had beensentenced to death.

t given that the execution has not alreadyoccurred to that individual prior to time t. t isyears from death sentence. The log odds of exe-cution at time t is a function of a set of time-con-stant covariates as well as a set of temporaldummies (td) to account for time dependence(�td).6 The individual level covariates entered inthis model include offender race, ethnicity, gen-der, prior convictions, and victim race. Themodels also include interactions betweenoffender and victim race. We assess the explana-tory power of state-level contextual effects byusing variables such as the percentage of astate’s vote for Republican candidates in time-varying form. The subsequent and more exhaus-tive models therefore take the following form:

Pijtlog ( 1 – Pijt) =

�td + M

�m=1

�mXmi +N

�n=1

�nXnj(t–1) (2)

where the conditional probability of executionat time t is a function of a set of individual con-demned prisoner factors as well as state char-acteristics at time t-1. Death row offenders maybe treated similarly within a state with the samedeath penalty provisions and the same officialsdeciding appeals. To adjust for this possibledeparture from statistical independence, wereport z-values corrected for within-state cor-related errors with a cluster procedure (Rogers1993).

SAMPLE

Most offender information was taken from

(ICPSR study 3958) “Capital Punishment inthe United States” (CPUS) from 1973 to 2002(this source is limited to those years). Thissource gives state, date of removal from deathrow, and reasons for removal, but it does notcontain data on victim race. For all offenderswho were executed, race of the victim, offend-er race, state, and execution date are availablefrom the Death Penalty Information Center.7

To obtain victim race we merged this data intothe CPUS file by matching execution date andrace of executed offenders.8

To obtain race of victim for former deathrow offenders removed from death row beforeexecution or for offenders still on death row, wehad to use another source. The SupplementalHomicide Reports (SHR) contain informationon offense date, victim race, offender race, andage. We obtain victim race by matching SHRdata with offender data from state correctiondepartment records by f irst using date ofoffense, and then offender race and age, butonly 16 states could or would provide offensedate. We merge the offender data with the CPUSdata using date of sentence, offender race, andage. This matching process yields victim racefor 1,012 current and former death row offend-ers who were not executed in 16 states. To cap-ture the effects of this explanatory variable, theanalysis must be restricted to 16 states, but thesestates are selected by a presumably randomprocess based on whether a state provides dateof offense—an administrative decision inde-



6 Because execution risk is small in the first 12years after a death sentence, we combine these yearsinto one period and use it as the reference group. Wethen create three additional dummy variables: thefirst is coded 1 for periods since death sentencebetween 13 and 20 years, the second between 21 and23 years, and the third from year 24 on death row andlater. This categorization captures the distribution ofexecution likelihoods as this distribution has a longflat tail to the left resulting from low initial risks andan inverted-U shaped distribuion thereafter. To dis-cover if this categorization is unrepresentative, we cre-ated alternative codes for death row years beyond 12.The significance test results and the theoretical impli-cations persist when we reestimate the models withthese alternatives.

7 This Web site is www.deathpenaltyinfo.org.Although the Supreme Court’s 1972 Furman decisionmade executions unconstitutional, Furman left openthe possibility of a constitutional death penalty. Manydeath penalty states therefore maintained their deathrow populations and altered their statutes. In 1976 inGregg and other cases the Court declared some mod-ified death penalty statutes constitutional (Paternoster1991; Zimring and Hawkins 1986). The data for aplausible study of death row outcomes before 1973apparently do not exist.

8 Since 1973, 768 of the 820 offenders executedwere matched. Data inconsistencies in these twosources explain the modest unmatched remainder.States analyzed are: Arizona, California, Delaware,Florida, Georgia, Illinois, Kansas, Kentucky,Maryland, Missouri, New Jersey, Ohio, Tennessee,Texas, Virginia, and Washington.

�A singleline was

permittedhere to

avoidexcessive

verticaljustifica-

tion; other-wise,

everythingfrom the A-

head onwould have

beencarried

over.

pendent of the state or offender characteristicsat issue.9

We therefore have information on victim racefor all executed offenders and for about 28 per-cent of current or former death row offendersin these 16 states who were not executed.Although the execution rate from our sample ismuch higher than the equivalent rate in theCPUS data, such disparate rates should not biasour multivariate results. Our goal is to discov-er which explanatory variables affect executionrates. We are interested in how relative execu-tion risks shift based on, for example, the mur-der of a white instead of a nonwhite. Becausewe include all executed offenders in these states,but only a fraction of death row offenders whowere not executed, our sample design is equiv-alent to a response-based sample design. Sucha design uses all events that occur in a specifiedtime period, but samples individuals who are atrisk of experiencing the event but did not(Prentice and Pyke 1979). To analyze aresponse-based sample, Xie and Manski (1989)propose a weighted maximum likelihood esti-mator for a logit analysis. Xie and Manskidemonstrate that estimates from a response-based sample behave asymptotically with nobias and with little loss in efficiency. We applythis weighted maximum likelihood estimatorin our logit model. Following Xie and Manski(1989), we define a weight variable that equals

qpwt =

fp(3)

where if p = 1, qp is the proportion executedfrom the CPUS sample and fp is the propor-

tion executed from our sample; if p = 0, qp isthe proportion not executed in the CPUS sam-ple and fp is the proportion not executed in oursample.10

The selection procedures we use are like-ly to produce a random sample because ourselection method depends on between datasource matches. This assumption that our esti-mates of explanatory effects will be unbiasedis corroborated by the results in Table 1 show-ing between sample contrasts. Most betweensample offender attributes are extremely sim-ilar. Although Hispanics are underrepresent-ed and whites are overrepresented in oursample, this bias is eliminated by includingoffender race and ethnicity in all analyses.Table 1 shows a total of 4,145 (3,597 + 548)death row offenders between 1973 and 2002in 16 states from the CPUS data, and 13.2 per-cent were executed (548/4,145). In compari-son to their proportion in the population,disproportionately more African Americansand Hispanics were on death row. Of thoseexecuted, slightly more were white. An over-whelming majority of death row offenderswere male (98 percent); and even higher pro-portions of males were executed (99.3 per-cent). Absent controls, death row offenderswith prior conviction records faced a higherchance of execution than those with no priorconviction. Among current or former deathrow offenders, two-thirds killed a white, butamong the executed offenders, four-f ifthskilled a white. The multivariate analyses willshow if victim race interacts with race ofoffender to alter execution likelihoods withother relevant factors held constant.

Finally, because the proportion of Asianand Native American death row offenders istoo modest to produce statistically signifi-cant results, these offenders are excluded



9 Included states exhibit considerable variation inexecution probabilities. Most but not all of the miss-ing death penalty states are small with few at risk ofexecution. Two exceptions are North Carolina andPennsylvania, but the corrections departments inthese and the other missing states with death row pop-ulations would or could not provide offense dates, sothe effects of victim race are not analyzed in thesestates. The funds from a generous grant were exhaust-ed by the matching process we had to use to obtaindata on victim race. We therefore cannot employcostly supplemental data sources such as newspaperaccounts, local prosecutor records (if any exist), ormake detailed appraisal of this process in particulardeath penalty states.

10 Weighted maximum likelihood estimates of alogit model make response-based samples behaveasymptotically as if they were random (Manski andLerman 1977). This method does not require thatresearchers select the best estimator. Even if the bestestimator is probit rather than logit, Xie and Manski(1989) use Monte Carlo simulations to show thatthe weighted maximum likelihood logit estimatorgives good results as long as response-based samplesare larger than 1,000 as ours is.

from our analyses. We also exclude offenderswho died on death row for reasons other thanan execution, but these decisions have no dis-cernable effects on the results.11 We thereforeanalyze outcomes experienced by 1,560 post-Furman death row offenders in 16 states from1973 to 2002. The corrections we use—thatare grounded in the econometric literature onhow such samples can be analyzed to pro-vide accurate findings—should give unbi-ased and consistent est imates of thedeterminants of executions. This is so eventhough our sample includes all offenders whowere executed, but only about 28 percent ofthose who were not executed in these 16states.

MODEL SPECIFICATION AND EXPLANATORY

VARIABLE MEASUREMENT

One general specification of the discrete timelogit model that predicts the log of executionodds for death row offender i in state j at timet is:

Log EXECUTION ODDSij(t) = (td DURATION DUMMIES) + b1BLACKij + b2HISPANICij + b3VWHITEij + b4(BLACK ij �

VWHITEij) + b5(HISPANICij �VWHITEij) + b6MALEij +

b7PRIORij + b8%BLKj(t–1) +b8%BLK2j(t–1) + (4)

b10%HISP3j(t–1) + b11POPj(t–1) +

b12MURDRTj(t–1) + b13BORNINSTj(t–1) +b14IDEOLOGYj(t–1) + b15%PPRSVOTEj(t–1) +b16 RLMEDINCj(t–1) + b17 %UNEMPj(t–1) + b18 %UNEMP2j(t–1) + b19YEAR92PLUSj +

b20STHj

where all explanatory variables are definedbelow.12



Table 1. Percentage Distributions of Offender Characteristics in 16 States by Samplea

Sample in Which RaceTotal CPUS Sample of Victim is Available

Not Executed Executed Not Executed Executed

3,597 548 1,012 548Race (Percent)—White 48.1 55.5 53.6 55.5—Black 39.8 34.9 40.1 34.9—Hispanic 12.1 09.7 06.2 09.7Sex—Percent Maleb 98.3 99.3 98.0 99.3Marital Status at Offense—Percent Marriedb 24.0 30.8 21.2 30.8College Education at Offense—Percent Yesb 07.6 07.7 10.0 07.7Prior Convictions—Percent Yesb 59.4 68.2 59.5 68.2Race of the Victim (Percent)—White 66.5 80.3—Nonwhite 33.5 19.7

a Marital status and college education at offense cannot be used in the logit regressions because too many missingvalues are present; we provide their distributions to better compare samples that do not include and include raceof victim.b These percentages plus the omitted category for each variable sum to 100 percent.

11 We eliminated 179 offenders from the samplebecause they died on death row before execution; 118Native Americans and 34 Asians or Pacific Islandersalso were eliminated leaving a total of 1,560 offend-ers. In analyses (not shown) we find that these exclu-sions do not affect the reported results. Deficienciesin the data made efforts to code race of victim in thefew instances when there were multiple victims toospeculative, so most of these cases were omitted.

12 We use exhaustive models. According toJohnston (1984), “It is more serious to omit relevant

INDIVIDUAL EFFECTS. We create two dummyvariables for minority offenders: the first equals1 if an offender is African American (BLACK),the second equals 1 if an offender is Hispanic(HISPANIC), and both dummies are set equalto 0 if the condition in question is not true.Another dummy variable is coded 1 only if thevictim is white (VWHITE). We next create twointeraction terms: the f irst (BLACK �VWHITE) is equal to the dummy variable coded1 for an African American offender times thedummy variable coded 1 for a white victim; thesecond (HISPANIC � VWHITE) is equal to thedummy variable coded 1 for an Hispanic offend-er times the dummy variable coded 1 for a whitevictim. Additional individual effects are heldconstant with a dummy variable coded 1 if adeath row inmate is male (MALE) and anoth-er coded 1 if a death row inmate has prior con-victions (PRIOR).13

CONTEXTUAL EFFECTS. We measure racialthreat effects with the percentage of AfricanAmericans (%BLK) and the percentage of

Hispanic residents (%HISP) in each state, butin contrast to all other contextual variables, thepercentage of Hispanics in the states is unavail-able in the non-census years before 1990. Wetherefore use decennial census figures in the fiveyears after a census and figures from the nextcensus in the next five years before 1990; after1990 this variable is time varying by year (theresults persist if we use yearly linear interpola-tions to estimate these missing values).14 Tocapture nonlinear quadratic relationships, bothminority threat variables are entered in untrans-formed and squared form. All state-level vari-ables, save the percentage of Republican votesfor president and Hispanics (which time vary byfour or five years), are time-varying by year.Following convention in studies of public pol-icy, the yearly time-varying explanatory vari-ables are lagged by a year because their effectsshould not be instantaneous.

Theory, however, suggests that Hispanic pres-ence should have a nonlinear relationship withexecution probabilities that becomes increas-ingly stronger as this indicator reaches extremevalues, yet there is no reason to think this asso-ciation should shift direction. Higher percent-ages of Hispanics, for example, may wellproduce increasingly greater execution proba-bilities, but it is difficult to see how growthfrom a low to a somewhat higher percentage ofHispanics could lead to reductions in execu-tion probabilities. When a priori considerationssuggest that such nonlinear relationships thatshould not change direction are present, powertransformations are most appropriate (Cohen etal. 2003:225–54). Threat theory thus suggeststhat a single exponential transformation will bebest. To avoid iterative searches for the bestexponent and overfitting, we cube the percent-age of Hispanics. This power transformation issomewhat unconventional in sociology (but notin psychology). To provide comparisons andreassure the reader, we present otherwise iden-tical models with and without this transforma-tion.

We assess the influence of the division oflabor with population (POP). Following Jacobsand Carmichael (2002), who find that this meas-



variables than to include irrelevant variables since inthe former case the coefficients will be biased, the dis-turbance variance overestimated, and conventionalinference procedures rendered invalid, while in thelatter case the coefficients will be unbiased, the dis-turbance variance properly estimated, and the infer-ence procedures properly estimated. This constitutesa fairly strong case for including rather than exclud-ing relevant variables in equations. There is, howev-er, a qualification. Adding extra variables, be theyrelevant or irrelevant, will lower the precision of esti-mation of the relevant coefficients” (p. 262). Henceour comprehensive models should produce moreaccurate point estimates but the significance tests willbe relatively conservative.

13 Although we cannot introduce a control foroffense severity because information is not avail-able, a legal requirement probably makes such a con-trol unnecessary. In the Gregg case that relegalizedcapital punishment, the Court held that jurors can sen-tence offenders to death only if they deem the crimeto be sufficiently horrific. Aggravating criteria stip-ulated by legislatures that jurors must consider whenmaking this judgment include the killing of multiplevictims, murdering a child, or homicides that involvedeliberate torture (Paternoster 1991). Such prereq-uisites probably create a more precise control foroffense severity than the alternatives used in studiesthat assess sentencing for non-capital crimes.

14 The great majority of the death row outcomeswe study occurred after 1990 when the percentage ofHispanics in the states is calculated with yearly data.

ure of outsider presence explains whether astate has a legal death penalty, outsiders aremeasured with a dummy coded 1 if over 75percent of a state’s population was born in-state(BORNINST). We assess partisanship with thepercentage of a state’s vote for the Republicancandidate in the last presidential election(PPRSVOTE). The threat from higher murderrates is gauged by murder rates provided by theUniform Crime Reports.

Berry and colleagues (1998) view mass ide-ologies (IDEOLOGY) as the mean on a liber-al-conservative continuum. They identify theideological position of each member ofCongress using interest-group ratings(Americans for Democratic Action, Committeeon Political Education) of a representative’s vot-ing record and then estimate mass ideology ineach congressional district with this score for thedistrict’s incumbent and with an estimated scorefor the incumbent’s challenger in the last elec-tion. Incumbent ideology scores are combinedwith estimated challenger ideology scoresweighted by district election results to capturedistrict ideologies. Berry and colleagues cal-culate state-level scores on liberalism-conser-vatism with the mean of these within-statecongressional district scores. We analyze themost recent version of this index, which gaugescongressional votes until 2003 and adds a fewcorrections to the values first published in 1998.The most liberal states receive the highestscores, so the coefficients on this variable shouldbe negative.15

We measure the tax base and a state’s abili-ty to support this costly punishment with realmedian household income (RLMEDINC). We

capture joblessness with the percentage ofunemployed (%UNEMP). Because this factormay have a nonlinear relationship, we includeits square. We also include a dummy variablecoded 1 (YEAR92PLUS) for years after 1992when execution frequencies increased and adummy variable coded 1 for southern states(STH). We use one-tailed tests because theo-retically based predictions about sign have beenstipulated.16

ANALYSES

DESCRIPTIVE STATISTICS AND

MULTIVARIATE MODELS OF EXECUTION

PROBABILITIES

INITIAL ANALYSES. Figure 1 shows the distribu-tion of death row outcomes during the years inquestion and Table 2 shows the predicted signs,means, and standard deviations. In Table 3 webegin the multivariate analyses by restricting thefirst model to individual explanatory factors toshow contrasts with subsequent models thatinclude contextual effects. Recall that twodummy variables are set equal to 1 if offendersare African American or Hispanic. A third is setequal to 1 if the victim is white, and this victimvariable is interacted with the two minorityoffender variables to create two interactionterms. The coefficients on these two productterms capture what happens to either AfricanAmerican or Hispanic offenders who killedwhites (all necessary main effects are included).A sixth dummy variable is set equal to 1 if a con-demned offender is male, while the last is equalto 1 if an offender has prior convictions. InModel 2 we begin to add contextual variablesby entering the percentage of African Americanand Hispanic residents in quadratic form. Toassess division of labor effects, population isincluded as well. We add the percentage born in-



15 This index has face validity. From 1972 to 2002Rhode Island and Massachusetts tied for second asthe most liberal states. But the sole Vermont repre-sentative—who probably was the only member ofCongress who claimed to be a socialist in this peri-od—earned the highest liberalism score. Mississippiin 1972 had the most conservative score followed byVirginia in 1974. Specialists have accepted the useof roll-call, vote-based indexes to identify ideology(Fowler 1982; Poole and Rosenthal 2000). Rankingsby Americans for Democratic Action (ADA) and theAFL-CIO’s Committee on Political Education(COPE) are the most widely used and have withstoodconsiderable scrutiny (Herrera, Epperlein, and Smith1995; Shaffer 1989).

16 Despite repeated attempts we have not locatedinformation on the partisanship of state appellatecourt justices. In almost all states, one appellate courthandles each level of state death row appeals, sovariation in offender outcomes probably cannot beattributed to within state appellate court differences.Data on factors such as offender demeanor, theirbehavior in prison, offense characteristics other thanthose measured, and additional information onoffender appeals unfortunately are not available.



Figure 1. Three Death Row Outcomes in 16 States by Years After Death Sentence

Table 2. Predicted Signs, Means, and Standard Deviations

Variable Predicted Sign Mean SD

1 if Executed .010 .0991 if Black Offender + .390 .4881 if Hispanic Offender + .068 .2511 if White Victim + .710 .4541 if Black � 1 if White Victim + .161 .3671 if Hispanic � 1 if White Victim + .036 .1871 if Male + .986 .1161 if Prior Conviction + .650 .477Percent Black + 12.772 5.331Percent Black2 – 191.548 171.170Percent Hispanic 0 15.212 11.189Percent Hispanic3/10,000 + .943 1.169Population/1,000,000 – 13.751 8.0211 if Percent Born in State > 75 Percent – .225 .417Murder Rate per 100,000 + 9.332 2.906Citizen Ideology – 43.608 9.149Percent Republican Vote for President + 48.695 9.330Real Median Household Income/100 + 366.480 47.638Percent Unemployed + 6.051 1.5321 if Southern State + .528 .499

Note: Weighted by 16,361 offender-years from 16 sampled states.

state and a quadratic test of the effects of unem-ployment in Model 3.

The results in Model 1 show that several indi-vidual offender attributes contribute to execu-

tion probabilities. The coefficient on the firstmain effect shows that African American offend-ers whose victims are not white are less likelyto be executed. But the coefficients on the two

interaction terms suggest that African Americansor Hispanics found guilty of killing members ofthe majority race face a greater likelihood of thispunishment (we defer discussing significancetests on contrasts between the coefficients onindividual variables until discussion of the bestmodel). If they persist, these last findings areparticularly important as they suggest that theextra-legal attribute with the most powerfuleffects on death sentences accounts for post-death sentence execution probabilities as well.

In Model 2 we add five contextual variablesto the individual determinants in the first model.Although these additions increase modelexplanatory power as the BIC statistic fallssharply, the coefficients on the interaction termsthat assess execution probabilities for minoritieswho kill whites remain significant. We againfind that African Americans convicted of mur-dering nonwhites are less likely to be executed,but execution probabilities remain higher forAfrican Americans who kill a white. The con-textual results show that execution probabilitiesare greater in states with more AfricanAmericans until a threshold in this proportion

is reached (we also defer reporting this inflec-tion point until the presentation of the bestmodel). Although the percentage of Hispanicresidents has no effect, this ethnic threat vari-able in squared form enhances execution prob-abilities. Additional findings show that stateswith larger populations and a greater division oflabor are not as likely to execute. After we addthe unemployment rate and its square togetherwith the percentage born in-state, these initialresults suggest that unemployment rates influ-ence execution probabilities, but this penalty isless likely in states with few outsiders.

COMPREHENSIVE MODELS. Yet other accountsmay matter. In Model 4 of Table 4, we add statemurder rates, citizen ideology, votes forRepublican presidential candidates, real medi-an income, and a dummy variable for yearsafter 1992 when the number of executionsexpanded. Model 5 is identical to Model 4, butwe replace the quadratic Hispanic threat spec-ification with the cubic power transformation.Model 6 only differs from Model 5 because a



Table 3. Post-Capital Sentencing Determinants of Executions Estimated with Discrete-Time Logit

Model 1 Model 2 Model 3

Explanatory Variables Coef. SE Coef. SE Coef. SE

Individual Effects—1 if Black Offender –1.015*** .349 –.761*** .220 –.758*** .215—1 if Hispanic Offender –.375 .465 .131 .228 .102 .238—1 if White Victim –.348 .258 –.067 .230 –.157 .219—1 if Black Off. � 1 if White Victim 1.394*** .371 .898** .322 1.042*** .337—1 if Hispanic Off. � 1 if White Victim 1.130** .422 .332 .255 .449* .257—1 if Male .458 .499 .907* .545 .870 .539—1 if Prior Conviction .082 .130 .134 .103 .123 .086Contextual Effects—Percent Black .— .— .443*** .106 .423*** .077—Percent Black2 .— .— –.011*** .003 –.012*** .002—Percent Hispanic .— .— –.052 .058 –.036 .054—Percent Hispanic2 .— .— –.005*** .001 .003** .001—Population/1,000,000 .— .— –.149*** .040 –.100*** .029—1 if Percent Born in State > 75 Percent .— .— .— .— –1.813*** .322—Percent Unemployed .— .— .— .— –1.144** .392—Percent Unemployed2 .— .— .— .— .069** .027Constant –5.101*** .633 –8.556*** 1.118 4.093*** 1.258Log Likelihood –873.3*** –816.6*** –793.7***BIC Statistic 1853.4 1778.8 1732.8

Notes: N = 1,560 offenders and 16,361 offender-years from 16 states; state-level cluster corrected standarderrors; coefficients on offender-year dummies not shown.* p ≤ .05; ** p ≤ .01; *** p ≤ .001 (one-tailed tests except for intercept).

control for state location in the South is addedto the explanatory variables entered in Model 5.

The results in Model 4 show that the additionof political variables and the tax base measureincrease model explanatory power as the BICstatistic again declines from its value in Model3. The findings confirm other contextual expec-tations: liberal states are less likely to execute,but capital punishment likelihoods are higher instates where Republican presidential candidatesreceived the most votes and where the tax baseis substantial. These findings show that the mur-der rates have the expected positive relationshipwith executions, but the unemployment rates areno longer significant after the addition of con-textual variables that assess state political envi-ronments. When we replace the quadraticHispanic specification with the cubic poweralternative in Model 5, all results persist, but the

coefficient on Hispanic threat becomes signif-icant.

A test of the significance of the differencebetween the absolute value of the coefficient onthe black offender main effect and the coeffi-cient on the interaction term that gauges exe-cution likelihoods for blacks who kill whitesshows that the interaction term coefficient isgreater than its counterpart on the main effect(two-tailed p = .038). Other tests that gauge thestatistical significance of contrasts reveal thatthe coefficient on the blacks who kill whitesterm is most substantial. For example, whenwe replace the condemned blacks main effectwith the same variable for whites, we find iden-tical significance test results again indicatingthat the coefficient on the black killings ofwhites interaction term is more substantial.Similar tests show that the coefficient on the



Table 4. Post-Capital Sentencing Determinants of Executions Estimated with Discrete-Time Logit

Model 4 Model 5 Model 6

Explanatory Variables Coef. SE Coef. SE Coef. SE

Individual Effects—1 if Black Offender –.737*** .233 –.744*** .234 –.764*** .231—1 if Hispanic Offender .108 .248 .107 .248 .099 .253—1 if White Victim –.141 .202 –.144 .202 –.158 .200—1 if Black Off. � 1 if White Victim 1.007** .339 1.011** .342 1.033*** .334—1 if Hispanic Off. � 1 if White Victim .429* .256 .437* .263 .454* .269—1 if Male .845 .527 .852 .531 .848 .530—1 if Prior Conviction .118* .069 .119* .073 .121 .075Contextual Effects—Percent Black .450*** .066 .439*** .065 .479*** .113—Percent Black2 –.014*** .002 –.014*** .002 –.014*** .003—Percent Hispanic –.019 .068 .— .— .— .——Percent Hispanic2 .002 .002 .— .— .— .——Percent Hispanic3/10,000 .— .— .533*** .109 .569*** .134—Population/1,000,000 –.099*** .025 –.099*** .027 –.099*** .026—1 if Percent Born in State > 75 Percent –2.000*** .265 –1.998*** .284 –2.089*** .397—Murder Rate .088* .045 .102** .043 .102* .045—Citizen Ideology –.030** .011 –.029** .011 –.033*** .009—Percent Republican Vote for President .048*** .011 .048*** .012 .052*** .010—Real Median Household Income /100 .010*** .003 .010*** .003 .009** .003—Percent Unemployed –.749 .463 –.778 .493 –.783 .500—Percent Unemployed2 .050 .033 .052 .035 .052 .036—1 if after 1992 1.597*** .307 1.602*** .316 1.635*** .302—1 if Southern State .— .— .— .— –.291 .611Constant –12.123*** 2.268 –12.016*** 1.951 –12.004*** 1.966Log Likelihood –776.1*** –775.6*** –775.5***BIC Statistic 1697.7 1696.8 1696.5

Notes: N = 1,560 offenders and 16,361 offender-years from 16 states; state-level cluster corrected standarderrors; coefficients on offender-year dummies not shown.* p ≤ .05; ** p ≤ .01; *** p ≤ .001 (one-tailed tests except for intercept).

black killings of whites interaction term isgreater than the coefficient on the Hispanic-white victim interaction term (two-tailed p =.023). Finally, when we add the South in the lastmodel, we find no noteworthy changes in theresults.

We now can calculate the inflection pointwhere the relationship between the percentageof African Americans and execution probabili-ties shifts from positive to negative. Again usingcoefficients from the best model (Model 5), wefind that this change occurs after the percent-age of African American residents reaches 16.2or about the 86th percentile in the percentage ofAfrican American residents in these 16 states inthis period. This result suggests that executionprobabilities start to fall only after the potentialAfrican American vote reaches a rather sub-stantial threshold. And odds ratios suggest thatthese effects are not weak. The exponentiatedcoefficient from Model 5 that gauges executionprobabilities for African Americans convictedof killing a white is 2.75. The odds ratio forAfrican Americans who kill nonwhites is .476.The odds ratio for Hispanics convicted of killinga white is 1.55. The size of some contextualeffects are substantial as well: for example, theodds ratio for the percentage of blacks in thesestates is 1.55 and its counterpart on the cube ofHispanic presence is 1.70.

ADDITIONAL CONSIDERATIONS. Some deathrow inmates accept their sentences and try tostop all appeals. These “volunteers” are dis-proportionately white (Lofquist 2002). In statesthat rarely execute, a greater proportion of thefew executed offenders are volunteers (Lofquist2002), so a failure to control for this effect maybias the estimates. The data at hand do not letus identify such offenders, but we can discov-er if this effect matters because volunteers areexecuted with less delay. If those executed earlydiffer sharply from those executed later, analy-ses limited to inmates executed later—or mod-els restricted to offenders executed from 5 to 10years after sentencing—should produce findingsthat differ, but they do not. The same variablesare significant in each of these five restrictedanalyses (not shown but available on request).Such theoretically identical findings suggestthat our inability to directly control for volun-teers has not distorted the findings.

We find no evidence for additional interac-tion effects and these negative results persistwhen we assess cross-level interactions betweenindividual and jurisdictional characteristics.This result is important for methodological rea-sons. Cross-level interaction significance testsare too optimistic if such tests are not estimat-ed with an Hierarchical Linear Model (HLM)approach. Yet none of these interactions are sig-nificant when they are assessed with the non-HLM method we employ. Such cross-leveleffects therefore will not be uncovered if theyare estimated with HLM because this estimatorwill produce larger estimates of the standarderrors. The cluster adjustment we use to correctthe standard errors for within state interdepen-dencies therefore should suffice, as HLM willproduce the same nonsignificant and redun-dant findings about cross-level interactions.17

We entered religious fundamentalism alongwith various measures of Republican gover-nors and this party’s strength in state legislatures,but these determinants had no effects (modelsnot shown). We find no evidence that federaljudge partisanship influences executions.18 Toisolate the effects of otherwise omitted cross-state national political, social, or macroeco-nomic changes, we included dummies coded for



17 For readers who prefer two-tailed significancetests even if theoretical justification for direction isprovided, we list coefficients that reached the one- butnot the two-tailed .05 level: in Table 3, the maleoffender variable in Model 2 is significant at the .05one- but not the two-tailed threshold and the same istrue for the Hispanics who killed whites interactionterm in Model 3. In Table 4, each of the coefficientson the Hispanics convicted of killing a white inter-action term are significant at the one- but not the two-tailed level and the same generality holds for thecoeff icient on the prior conviction variable inModel 4.

18 We gauged the degree to which the estimates arerobust by dropping each of the 16 sampled states inanalyses not shown. The implications persist but thisdependent variable already was skewed. To avoid alogit model that could not be estimated after weremoved a state with many executions and increaseddependent variable skewness, we had to remove a fewexplanatory variables from some of these 16 equa-tions. Only one of the 16 sampled death penaltystates (Kansas, which had a tiny death row popula-tion) had no executions. The elimination of states withfew executions had only tiny effects on the estimates.

each year (models not shown). All explanatoryvariables that matter in Table 4 were significantin these models. Such results suggest that shiftsin the national political or macroeconomic cli-mate cannot account for these results. The mod-els in Table 4 pass the link test formisspecification. These considerations provideadded reasons to think that the most compre-hensive models capture the primary determi-nants of death row outcomes.19

We nevertheless must acknowledge that wewere forced to use advanced statistical tech-niques to overcome data limitations. Althoughthere are good reasons to believe that the report-ed estimates are unbiased and consistent, supe-rior data always are preferable to such statisticalalternatives (for a forceful illustration, seeDonohue and Wolfors 2005). In particular, wehope that subsequent researchers can obtainmore exhaustive information on victim raceand other considerations from additional deathpenalty states. Perhaps researchers with theresources and the time to achieve a better rap-port with state correction agency officials whowould not cooperate with our requests for datacan produce superior findings about this issueby analyzing additional states.

DISCUSSION

THE FINDINGS

Possibly because offender prior convictionscontribute to prosecutors’ decisions to seek thedeath sentence and lead trial courts to supportthis request, we find weak and inconsistent evi-dence that this factor affects post-death sen-tence execution probabilities. The findings showthat gender is nonsignificant, but the proportion

of condemned women in our sample is tiny. Yetvictim race clearly is the most important indi-vidual finding. The coefficients and the signif-icance tests that gauge contrasts in the size ofthese point estimates show that blacks convict-ed of killing whites are more likely to be exe-cuted than other death row offenders. Suchresults do not contradict hypotheses that localprosecutors with an interest in protecting well-publicized legal victories that should furthertheir political careers make successful efforts toresist death row appeals. These findings alsosupport a hypothesis that state justices do notignore pressures to deny appellate relief tominority death row offenders who kill whites.But the findings suggest that African Americanson death row for killing nonwhites are less like-ly to be executed than other condemned pris-oners.

The evidence corroborates hypotheses thatHispanics also face higher execution probabil-ities if their victims are white. Yet the oddsratios that gauge the strength of this Hispaniceffect and the untransformed coefficients aresignificantly weaker than their counterparts thatassess the strength of the black offender, whitevictim effect. Such contrasting results aboutthe explanatory power of ethnic rather thanracial effects should not be surprising in lightof the fiercely divisive and violent conflictsabout race throughout U.S. history (Myrdal1944; Tocqueville 1948). Only a few states hadsubstantial Hispanic populations before the1990s. This ethnic group may have been toosmall in most death penalty states to be suffi-ciently threatening. Subsequent studies, how-ever, may provide greater support for this ethnicthreat account because the Hispanic populationrecently has expanded so rapidly.

The contextual results always show thatminority proportions matter. Larger percent-ages of African American residents producehigher execution probabilities, but this effectbecomes negative only after the proportion ofAfrican Americans reaches a relatively sub-stantial threshold. The positive sign on theunsquared coefficient provides evidence for aracial threat account, but the negative sign onthe squared percentage of African Americanresidents suggests that execution probabilitiesrespond to election pressures. The form of thenonlinear relationship between Hispanic pro-portions in a state and execution probabilities



19 Attempts to use other diagnostic tests failedbecause these estimates are weighted. Tests con-ducted without the weights suggest that estimationproblems are not present. No observation has a highleverage score, and there are only a few modest out-liers. Estimate stability and a VIF analysis conduct-ed on the state variables (without the squared terms)yields a maximum score 4.33. Both considerationssuggest that collinearity is not present. Although weanalyze 16 states, there are 480 state-years as all buttwo of our state-level variables are time-varying byyear. Fewer case-years are common in pooled time-series analysis.

differs from the nonlinear association betweenAfrican American presence and these proba-bilities. Probably because the proportion ofHispanics in all but a few states was so modest,the findings indicate that a growth in the per-centage of Hispanics yields increasinglystronger execution probabilities without a rever-sal in the sign of this relationship.

Consistent with prior findings on the factorsthat produce a legal death penalty in the states(Jacobs and Carmichael 2002), the findings inthis article suggest that jurisdictions with themost residents born in-state are less likely to usethis ultimate punishment. This association prob-ably is based on latent hostility to strangers(Hale 1996) and to a corresponding reluctanceto invoke the ultimate penalty against peoplewho are regarded as one’s neighbors. A differ-ent aggregate finding supports another threataccount, as the results show that execution prob-abilities are greater in states that have the high-est murder rates.

Votes for Republican presidential candidates,who often run on law and order platforms, helpexplain execution likelihoods as well. Beckett’s(1997) findings suggest that law and order cam-paign rhetoric magnifies the public salience ofthe crime issue. Successful law and order polit-ical campaigns therefore should produce greatersupport for the harshest punishment. The sen-sitizing effects of this rhetoric and the relativesuccess of Republican law and order candidatesin some jurisdictions—which almost certainlyindicates considerable preexisting support forcapital punishment—help explain why votesfor Republican candidates provide such a robustexplanation for executions. It is interesting thatthe presence of a Republican governor does notincrease execution probabilities (results notshown), but this finding should not be surpris-ing as this factor does not explain the legaliza-tion of capital punishment (Jacobs andCarmichael 2002). Governors must decide thefinal appeal before an execution. This awesomemoral responsibility gives these political offi-cials good reason to be ambivalent about thispunishment (Jacobs and Carmichael 2002;Zimring and Hawkins 1986).

As theorists such as Garland (1990, 2001) andSavelsberg (1994) would expect, the explanatorypower of political ideology suggests that polit-ical values contribute to the propensity to exe-cute. When or where there is greater political

support for liberal values, execution probabili-ties diminish. This result supports claims(Lakoff 1996) that support for the death penal-ty is one of the most important differencesbetween liberals and conservatives. Finally, theresults suggest that the affluent states that canbetter afford this costly punishment use it moreoften.

WIDER IMPLICATIONS

This analysis fills critical gaps in the sparse lit-erature on the application of capital punish-ment. First, we gauged the effects ofoffender-victim racial contrasts and found thatvictim race is a strong determinant of execu-tions. Until this study, and despite the empiri-cal strength of victim race in trial-court studiesabout the death sentence process, apparentlyno research has gauged the explanatory powerof this factor. In addition to the findings abouttheoretically important individual and legal fac-tors, our data set lets us test environmentaldeterminants that have been ignored in the lit-erature. The findings suggest that these con-textual omissions are unfortunate, asenvironmental accounts have substantialexplanatory power in this and in other analysesthat seek to explain criminal justice outcomes.Compared to the studies of the determinants ofdeath sentences that are restricted to individualdata from one or a few jurisdictions, this study’scoverage is far greater as we assessed the effectsof both individual and contextual accounts overa 30-year period in a diverse set of multiplestates.

Reviews of the related research on race andfelony sentencing (Chiricos and Crawford 1995;Walker et al. 1996) suggest that substantial dis-agreement exists in this literature. These reviewsboth base their conclusions on results that appearin only a bare majority of the many availableinvestigations. Almost all of the many studiesthat analyze trial court sentencing are based onoffender samples from just one or only a fewtrial courts in a specific region. Yet since trialcourts exist in diverse environments, inconsis-tent results can be expected. If each court studyinvestigates death penalty sentences in just oneor a few localities, and their political climatesdiffer, it will be difficult to uncover generalrelationships, particularly because court envi-ronments have so much explanatory power



(Helms and Jacobs 2002). The same reasoningholds for death penalty investigations. It followsthat a study of executions in many states thatassesses both contextual and individual factorsshould provide more accurate results than wouldanalyses restricted to individual explanationsin just one or only a few jurisdictions.

Another plausible explanation for these dis-agreements concerns the neglect of politicaleffects in almost all of the studies of criminaljustice system outcomes. As one might expectin light of the recent theoretical emphasis on thepolitical nature of legal punishments (Chambliss1994; Foucault 1977; Garland 1990, 2001) andthe Republican party’s tactical use of law andorder appeals to covertly enhance voters’ racialfears (see the quotes in note 4), these resultsshow that executions increased after statesawarded additional votes to Republican candi-dates. Yet increased electoral support for law andorder Republicans who enthusiastically embracethe death penalty may be based on preexistingconservative values. But this alternative hypoth-esis is unlikely as votes for Republican presi-dential candidates account for executions aftercitizen ideology has been held constant. Suchfindings, of course, show that these supposed-ly objective and purely legal decisions aboutwho will die are shaped by factors that shouldnot be relevant.

These findings also confirm the racial poli-tics perspective that provided the primary con-ceptual impetus for this research. In thehomogeneous European democracies, predom-inant penal emphasis is on the reintegration ofoffenders into a solidaristic society adminis-tered by experts who are only distantly account-able to the voters (Savelsberg 1994; Whitman2003). In the United States, however, the pub-lic’s Manichean image of human nature, com-bined with a history of bitter racial conflict,has produced an exclusionary penal system.According to Wacquant (2000, 2001), theresilient divisions produced by slavery and bythe later virulent measures used to maintainwhite supremacy created a segregative penalsolution consistent with the dominant publicimage of criminals as members of a vile racialunderclass. This racial history and the resultingcultural premises helped to produce a criminaljustice system that primarily uses incapacitationto control street crime. The most reliable way toincapacitate incorrigible offenders is to kill

them. This incapacitative solution fits with theprevailing public view—often forcefullyexpressed politically in this most direct of alllarge democracies—that the primary goal ofthe U.S. criminal justice system should bevengeance.

Executions therefore should be most likely injurisdictions in which covertly racist law andorder political appeals have been most suc-cessful and where the presence of a large AfricanAmerican population—that has not becomequite large enough to enforce its political val-ues—threatens dominant whites. Prior researchsupports such political and racial results.Findings show that the same aggregate politi-cal and racial factors account for the likelihoodthat a state will legalize the death penalty(Jacobs and Carmichael 2002). In another study,Jacobs and colleagues (2005) report that simi-lar political and racial factors explain the num-ber of death sentences. But compared to thesetwo studies and other aggregate-level research,this investigation is unique as it analyzes bothstate and individual characteristics. Probablythe most important finding that emerges fromthis multilevel approach concerns victim race.African Americans who breach the racial castesystem by killing whites can more often expectthe harshest of all penalties, and this findingholds after many contextual level political andracial effects have been held constant. Suchresults do not contradict Wacquant’s (2000,2001) theoretical claims that a primary goal inthe current U.S. criminal justice system is to sus-tain the prior racial caste system even though thecurrent methods are less transparent than thelawless brutality used in the past.20

The most important legal implication con-cerns racial equity. Many claims have beenmade that the U.S. criminal justice system is notcolorblind, yet after multiple studies, definitiveevidence showing that the trial courts sentenceAfrican Americans less leniently than whites hasremained elusive. This study instead analyzedracial effects on the post-sentencing executionprocess. The findings show that despite efforts



20 For additional findings on how past racial, eth-nic, or religiously based brutality continues to affectcurrent practices, see Archer and Gartner (1984),Messner, Baller, and Zevenbergen (2005), Savelsbergand King (2005), and Jacobs and colleagues (2005).

to transcend an unfortunate racial past, residuesof this fierce discrimination evidently still linger,at least when the most morally critical decisionabout punishment is decided. Findings indicat-ing that African Americans who kill nonwhitesare less likely to be executed than their coun-terparts who kill whites show that the post-sen-tencing capital punishment process continues toplace greater value on white lives. Evidently,inclinations to devalue African Americans inconstitutional compromises in the distant pastremain today, although they are expressed in lessconspicuous ways. Associate Justice WilliamBrennan helps us understand the importanceof such discriminatory findings:

Those whom we would banish from society orfrom the human community itself often speak intoo faint a voice to be heard above society’sdemand for punishment. It is the particular role ofthe courts to hear these voices, for the Constitutiondeclares that the majoritarian chorus may not alonedictate the conditions of social life. (Brennan 1987)

This evidence shows that the state and federalappellate courts—who should pay special atten-tion to those accused of the most horrif iccrimes—evidently continue to listen to somevoices more than others.

David Jacobs is Professor of Sociology and (by cour-tesy) Political Science at The Ohio State University.He uses a political economy approach to study out-comes in the criminal justice system and other issuesin political sociology. A study of racial politics recent-ly appeared in the American Journal of Sociologywhile another publication on the determinants ofyearly execution frequencies will soon appear inSocial Problems. In addition to these interests, he con-tinues to investigate the politics of labor relations.

Zhenchao Qian is Professor in the Department ofSociology and research associate in the Initiative inPopulation Research at The Ohio State University. Hisresearch focuses on family demography, race andethnicity, and immigration. He studies changes inmate selection by taking into account marriage mar-ket conditions. His work with Daniel T. Lichter, recent-ly published in American Sociological Review,centers on changes in racial/ethnic intermarriage. Hisother research examines changing racial identifica-tion among children born to interracial couples.

Jason T. Carmichael is an Assistant Professor in theDepartment of Sociology at McGill University. Hisinterests include criminology, criminal justice, and,more broadly, political sociology. Current projectsinclude, among other things: an analysis of the polit-ical and social determinants of the number of juve-

nile delinquents who are adjudicated in the adultcriminal justice system, and an examination of theability of foundation funding to shape the size and/orsuccess of social movement organizations.

Stephanie L. Kent is an Assistant Professor ofSociology at Cleveland State University. Her researchfocuses on the politics of crime control and usesmacrosocial explanations to predict social controloutcomes. Recent publications include a pooled-time-series analysis of police strength in U.S. citiespublished in Criminology. Additional work on capi-tal punishment has or will be published in theAmerican Sociological Review and Social Problems.She is currently exploring the social and politicaldeterminants of homicide and the use of lethal forceby and against the police in U.S. cities.

REFERENCES

Aarons, Dwight. 1998. “Getting Out of this Mess:Steps Toward Addressing and Avoiding InordinateDelay in Capital Cases.” Journal of Criminal Lawand Criminology 89:1–58.

Allen, Michael P. and John L. Campbell. 1994. “StateRevenue Extraction from Different IncomeGroups.” American Sociological Review59:169–86.

Archer, Dane and Rosemary Gartner. 1984. Violenceand Crime in Cross-National Perspective. NewHaven, CT: Yale University Press.

Bailey, William C. and Ruth Peterson. 1999. “CapitalPunishment, Homicide, and Deterrence.” Pp.257–77 in Homicide: A Sourcebook, edited by M.D. Smith and M. A. Zahn. Thousand Oaks, CA:Sage.

Baldus, David C. and George Woodworth. 2003.“Race Discrimination in the Administration of theDeath Penalty: An Overview of the EmpiricalEvidence with Special Emphasis on the Post-1990Research.” Criminal Law Bulletin 39:194–226.

Bandes, Susan. 2004. “Fear Factor: The Role ofMedia in Covering and Shaping the Death Penalty.”Ohio State Journal of Criminal Law 1:585–97.

Beckett, Katherine. 1997. Making Crime Pay. NewYork: Oxford.

Berry, William D., Evan J. Ringquist, Richard C.Fording, and Russell L. Hanson. 1998. “MeasuringCitizen and Government Ideology in the States.”American Journal of Political Science 42:327–48.

Bickle, Gayle and Ruth Peterson. 1991. “The Impactof Gender Based Family Roles on CriminalSentencing.” Social Problems 38:372–94.

Blalock, Hubert M. 1967. Towards a Theory ofMinority Group Relations. New York: CapricornBooks.

Blume, J. and Theodore Eisenberg. 1999. “JudicialPolitics, Death Penalty Appeals, and Case



Selection.” Southern California Law Review72:465–503.

Blumer, Herbert. 1958. “Race Prejudice as a Senseof Group Position.” Pacific Sociological Review1:3–7.

Bobo, Lawrence and Vincent Hutchings. 1996.“Perceptions of Racial Group Competition:Extending Blumer’s Theory of Group Position ina Multiracial Social Context.” AmericanSociological Review 61:951–72.

Brace, Paul and Melinda G. Hall. 1997. “TheInterplay of Preferences, Case Facts, Context, andRules in the Politics of Judicial Choice.” Journalof Politics 59:1206–23.

Brennan, William J. 1987. Dissent in McClesky v.Kemp (481 U.S. 279, 343).

Bright, Stephen B. and Patrick J. Keenan. 1995.“Judges and the Politics of Death: DecidingBetween the Bill of Rights and the Next Electionin Capital Cases.” Boston University Law Review75:759–835.

Brown, Craig and Barbara Warner. 1992.“Immigrants, Urban Politics, and Policing in 1900.”American Sociological Review 57:293–305.

Carter, Dan T. 1996. From George Wallace to NewtGingrich: Race in the ConservativeCounterrevolution, 1963–1994. Baton Rouge, LA:Louisiana State University Press.

Chambliss, William J. 1994. “Policing the GhettoUnderclass: The Politics of Law and LawEnforcement.” Social Problems 41:177–94.

Chiricos, Theodore G. and Charles Crawford. 1995.“Race and Imprisonment.” Pp. 281–309 inEthnicity, Race, and Crime, edited by D. Hawkins.Albany, NY: SUNY Press.

Chiricos, Theodore G. and Miriam Delone. 1992.“Labor Surplus and Punishment: A Review andAssessment of Theory and Evidence.” SocialProblems 39:421–46.

Cohen, Jacob, Patricia Cohen, Stephen D. West, andLeona S. Aiken. 2003. Applied MultipleRegression/Correlation Analysis for the BehavioralSciences. 3d ed. Hillsdale, NJ: Lawrence ErlbaumAssociates.

Constanzo, Mark. 1997. Just Revenge: The Costsand Consequences of the Death Penalty. NewYork: St. Martins Press.

Dodge, Lowell, Laurie E. Eckland, Harriet C. Ganson,Lisa Cassady, James L. Flemming, and Douglas M.Sloan. 1990. “Death Penalty Sentencing.” Pp.268–74 in Death Penalty in America, edited by H.A. Bedau. New York: Oxford.

Donohue, John J. and Justin Wolfers. 2005. “Uses andAbuses of Empirical Evidence in the Death PenaltyDebate.” Stanford Law Review 58:791–845.

Ehrlichman, John D. 1982. Witness to Power: TheNixon Years. New York: Simon & Schuster.

Fosset, Mark A. and K. Jill Kiecolt. 1989. “TheRelative Size of Minority Populations and White

Racial Attitudes.” Social Science Quarterly70:820–35.

Foucault, Michel. 1977. Discipline and Punish.London: Vintage.

Fowler, Linda L. 1982. “How Interest Groups SelectIssues for Rating Voting Records of Members ofthe U.S. Congress.” Legislative Studies Quarterly7:401–13.

Furman v. Georgia. 1972. 408 U.S. 238.Garland, David. 1990. Punishment and Modern

Society. Chicago, IL: University of Chicago Press.———. 2001. The Culture of Control. Chicago,

IL: University of Chicago Press.Giles, Michael W. and Melanie A. Buckner. 1993.

“David Duke and Black Threat.” Journal of Politics55:702–13.

Giles, Michael W. and Kaenan Hertz. 1994. “RacialThreat and Partisan Identification.” AmericanPolitical Science Review 88:317–26.

Goldfield, Michael. 1997. The Color of Politics. NewYork: New Press.

Gregg v. Georgia. 1976. 96 Sup. Ct. 2902.Hale, Chris. 1996. “Fear of Crime: A Review of the

Literature.” International Review of Victimology4:79–150.

Heer, David M. 1959. “The Sentiment of WhiteSupremacy: An Ecological Study.” AmericanJournal of Sociology 64:592–98.

Helms, Ronald and David Jacobs. 2002. “ThePolitical Context of Sentencing: An Analysis ofCommunity and Individual Determinants.” SocialForces 81:577–604.

Herrera, Richard, Thomas Epperlein, and Eric R. A.N. Smith. 1995. “The Stability of CongressionalRoll-Call Indexes.” Political Research Quarterly48:403–16.

Hood, Roger. 1998. “Capital Punishment.” Pp.739–76 in The Handbook of Crime andPunishment, edited by M. Tonry. New York: OxfordUniversity Press.

Hurwitz, Jon and Mark Peffley. 2005. “Playing theRace Card in the Post-Willie Horton Era.” PublicOpinion Quarterly 69:99–112.

Jacobs, David. 1979. “Inequality and Police Strength:Conflict Theory and Coercive Control inMetropolitan Areas.” American SociologicalReview 44:913–25.

Jacobs, David and Jason T. Carmichael. 2001. “ThePolitics of Punishment Across Time and Space.”Social Forces 80:61–89.

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

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. “Towards aPolitical Model of Incarceration: A Time-Series



Examination of Multiple Explanations for PrisonAdmission Rates.” American Journal of Sociology102:323–57.

Jacobs, David and Robert M. O’Brien. 1998. “TheDeterminants of Deadly Force: A StructuralAnalysis of Police Violence.” American Journal ofSociology 103:837–62.

Jacobs, David and Daniel B. Tope. 2007. “The Politicsof Resentment in the Post Civil-Rights Era:Minority Threat, Homicide, and Ideological Votingin Congress.” American Journal of Sociology113:1458–94.

Johnston, Jack. 1984. Econometric Methods. NewYork: McGraw-Hill.

Kent, Stephanie L. and David Jacobs. 2005.“Minority Threat and Police Strength from 1980to 2000: A Fixed-Effects Analysis of Nonlinear andInteractive Effects in Large U.S. Cities.”Criminology 43:731–60.

Key, V. O. 1949. Southern Politics. New York: VintageBooks.

Lacey, Nicola. 1988. State Punishment. London:Routledge.

Lakoff, George. 1996. Moral Politics. Chicago, IL:University of Chicago Press.

Langworthy, Robert H. and John T. Whitehead. 1986.“Liberalism and Fear as Explanations forPunitiveness.” Criminology 24:575–91.

Levitt, Steven D. 2002. “Deterrence.” Pp. 435–50 inCrime: Public Policies for Crime Control, editedby J. Q. Wilson and J. Petersillia. Oakland, CA: ICSPress.

Liebman, James S., Jeffrey Fagan, and Valerie West.2000. “A Broken System: Error Rates in CapitalCases, 1973–1995.” Report, Columbia UniversityLaw School.

———. 2002. “A Broken System, Part II: WhyThere is so Much Error in Capital Cases, and WhatCan be Done About It.” Report, ColumbiaUniversity Law School.

Lipschultz, Jeremy H. and Michael L. Hilt. 2002.Crime and Local Television News: Dramatic,Breaking, and Live from the Scene. Mahwah, NJ:Lawrence Erlbaum Associates.

Liska, Allen E., J. J. Lawrence, and A. Sanchirico.1982. “Fear of Crime as a Social Fact.” SocialForces 60:760–71.

Lofquist, William S. 2002. “Putting Them There,Keeping Them There, and Killing Them: AnAnalysis of State-Level Variations in Death PenaltyIntensity.” Iowa Law Review 87:1505–57.

Manski, Charles F. and Steven R. Lerman. 1977.“Estimation of Choice Probabilities from ChoiceBased Samples.” Econometrica 45:1977–89.

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

Myers, Martha A. and Suzette M. Talarico. 1987.

The Social Contexts of Criminal Sentencing. NewYork: Springer-Verlag New York Inc.

Myrdal, Gunnar. 1944. An American Dilemma. NewYork: Harper and Brothers.

Paternoster, Raymond. 1991. Capital Punishment inAmerica. New York: Lexington Books.

Perrow, Charles. 1986. Complex Organizations: ACritical Essay. New York: Random House.

Poole, Keith T. and Howard Rosenthal. 2000.Congress: A Political-Economic History of RollCall Voting. New York: Oxford University Press.

Prentice, R. L. and R. Pyke. 1979. “Logistic DiseaseIncidence Models and Case-Control Studies.”Biometrika 66:403–11.

Quillian, Lincoln. 1996. “Group Threat and Changesin Regional Change in Attitudes Toward African-Americans.” American Journal of Sociology102:816–60.

Quillian, Lincoln and Devah Pager. 2001. “BlackNeighbors, Higher Crime? The Role of RacialStereotypes in Evaluations of NeighborhoodCrime.” American Journal of Sociology107:717–67.

Rogers, William H. 1993. “Regression StandardErrors in Clustered Samples.” Stata TechnicalBulletin 13:1923–28.

Rusche, Georg and Otto Kirchheimer. 1939.Punishment and Social Structure. New York:Russell and Russell.

Savelsberg, Joachim J. 1994. “Knowledge,Domination, and Criminal Punishment.” AmericanJournal of Sociology 99:911–43.

Savelsberg, Joachim J. and Ryan D. King. 2005.“Institutionalizing Collective Memories of Hate:Law and Law Enforcement in Germany and theUnited States.” American Journal of Sociology111:579–616.

Scott, W. Richard. 1987. Organizations: Rational,Natural, and Open Systems. Englewood Cliffs,NJ: Prentice Hall.

Shaffer, William R. 1989. “Rating the Performanceof the ADA in the U.S. Congress.” Western PoliticalQuarterly 42:33–51.

Spurr, Stephen J. 2002. “The Future of CapitalPunishment: Determinants of the Time from DeathSentence to Execution.” International Review ofLaw and Economics 22:1–23.

Stucky, Thomas D., Karen Heimer, and Joseph B.Lang. 2005. “Partisan Politics, ElectoralCompetition, and Imprisonment.” Criminology43:211–48.

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

Taylor, Ian, P. Walton, and Jock Young. 1973. The NewCriminology. London: Routledge and Kegan Paul.

Taylor, Marylee C. 1998. “How White Attitudes Vary



with the Racial Composition of Local Populations.”American Sociological Review 63:512–35.

Thorne, Melvin J. 1990. American ConservativeThought Since World War II. New York: GreenwoodPress.

Tocqueville, Alexis de. 1948. Democracy in America.New York: Knopf.

Turk, Austen. 1969. Criminality and the Legal Order.Chicago, IL: Rand McNally.

Uniform Crime Reports [United States].“Supplementary Homicide Reports, Various Years.”Ann Arbor, MI: Inter-University Consortium forPolitical and Social Research.

United States Department of Justice, Bureau ofJustice Statistics. 2003. “Capital Punishment in theUnited States, 1973–2002.” Ann Arbor, MI: Inter-University Consortium for Political and SocialResearch.

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

———. 2001. “Deadly Symbiosis: When Ghetto

and Prison Meet and Mesh.” Punishment & Society3:95–134.

Walker, Samuel, Cassia Spohn, and Miriam DeLone.1996. The Color of Justice: Race, Ethnicity, andCrime in America. Belmont, CA: Wadsworth.

Weidner, Robert R. and Richard S. Frase. 2003.“Legal and Extralegal Determinants of IntercountyDifferences in Prison Use.” Criminal Justice PolicyReview 14:377–400.

Western, Bruce. 2006. Punishment and Inequality inAmerica. New York: Russell Sage Foundation.

Whitman, James Q. 2003. Harsh Justice: CriminalPunishment and the Widening Divide BetweenAmerica and Europe. New York: Oxford.

Xie, Yu and Charles P. Manski. 1989. “The LogitModel and Response-Based Samples.”Sociological Methods and Research 17:283–302.

Zatz, Marjorie S. 1987. “The Changing Forms ofRacial/Ethnic Biases in Sentencing.” Journal ofResearch in Crime and Delinquency 24:69–92.

Zimring, Franklin E. and Gordon Hawkins. 1986.Capital Punishment and the American Agenda.New York: Cambridge.



Documents

Who Survives on Death Row? An Individual and Contextual ... · Who Survives on Death Row? An Individual and Contextual Analysis David Jacobs Zhenchao Qian The Ohio State University