USF Libraries
USF Digital Collections

Assessing the issue of arbitrariness in capital sentencing in North Carolina


Material Information

Assessing the issue of arbitrariness in capital sentencing in North Carolina are the effects of legally relevant variables racially invariant?
Physical Description:
Earl, Judith Kavanaugh
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:


Subjects / Keywords:
Death penalty
Aggravating circumstances
Mitigating circumstances
Race discrimination
Dissertations, Academic -- Criminology -- Masters -- USF   ( lcsh )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )


ABSTRACT: This study analyzed case and sentencing data from 632 capital cases involving Black and White defendants and victims, processed in North Carolina from May 1990 through December 2002. Logistic regression analysis of all cases and race-specific data allowed assessment of the variable effects of jury acceptance of statutory aggravating and mitigating factors on capital sentencing outcomes (death versus life). The purpose was to evaluate the role race plays in shaping jury use of legally defined factors in capital sentencing. Significant variance in the effect of jury acceptance of aggravators was observed between Black and White defendants. Black defendants pay a higher premium in terms of the risk of a death sentence than do White defendants whose crimes are comparably aggravated.There was no overall disparity in the effect of jury acceptance of mitigatory factors observed, although certain mitigators reduced the risk of a death sentence significantly more for Black or White. Overall, the aggravators had a statistically significantly stronger effect on sentencing outcomes than did the mitigators, regardless of race, and on cases involving Black defendants, regardless of victim race. Racial invariance was not shown.
Thesis (M.A.)--University of South Florida, 2005.
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Judith Kavanaugh Earl.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 111 pages.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001709509
oclc - 68623224
usfldc doi - E14-SFE0001370
usfldc handle - e14.1370
System ID:

This item is only available as the following downloads:

Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001709509
003 fts
005 20060614112136.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 060510s2005 flu sbm s000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001370
HV6025 (Online)
1 100
Earl, Judith Kavanaugh.
0 245
Assessing the issue of arbitrariness in capital sentencing in North Carolina
h [electronic resource] :
b are the effects of legally relevant variables racially invariant? /
by Judith Kavanaugh Earl.
[Tampa, Fla.] :
University of South Florida,
Thesis (M.A.)--University of South Florida, 2005.
Includes bibliographical references.
Text (Electronic thesis) in PDF format.
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
Title from PDF of title page.
Document formatted into pages; contains 111 pages.
ABSTRACT: This study analyzed case and sentencing data from 632 capital cases involving Black and White defendants and victims, processed in North Carolina from May 1990 through December 2002. Logistic regression analysis of all cases and race-specific data allowed assessment of the variable effects of jury acceptance of statutory aggravating and mitigating factors on capital sentencing outcomes (death versus life). The purpose was to evaluate the role race plays in shaping jury use of legally defined factors in capital sentencing. Significant variance in the effect of jury acceptance of aggravators was observed between Black and White defendants. Black defendants pay a higher premium in terms of the risk of a death sentence than do White defendants whose crimes are comparably aggravated.There was no overall disparity in the effect of jury acceptance of mitigatory factors observed, although certain mitigators reduced the risk of a death sentence significantly more for Black or White. Overall, the aggravators had a statistically significantly stronger effect on sentencing outcomes than did the mitigators, regardless of race, and on cases involving Black defendants, regardless of victim race. Racial invariance was not shown.
Adviser: John K. Cochran.
Death penalty.
Aggravating circumstances.
Mitigating circumstances.
Race discrimination.
Dissertations, Academic
x Criminology
t USF Electronic Theses and Dissertations.
4 856


Assessing the Issue of Arbitrariness in Capital Sentencing in North Carolina: Are the Effects of Legally Relevant Variables Racially Invariant? by Judith Kavanaugh Earl A thesis submitted in partial fulfillment of the requirements for the degree of Masters of Arts Department of Criminology College of Arts and Sciences University of South Florida Major Professor: John K. Cochran, Ph.D. M. Dwayne Smith, Ph.D. Sondra J. Fogel, Ph.D. Date of Approval: November 9, 2005 Keywords: death penalty, aggravating circumstances, mitigating circumstances, race discrimination, jury 2005, Judith Kavanaugh Earl


Table of Contents List of Tables iii Abstract iv Chapter One: Introduction 1 Furman v. Georgia (1972): The Eighth Amendment Argument 2 Furman Implemented: Gregg/Proffitt/Jurek 4 Race Neutrality Under Furman: McCleskey v. Kemp (1987) 6 Denial of McCleskeys equal protection claim. 7 McCleskeys Eighth Amendment claim. 12 Latent Racism versus Blatant Racism 15 Chapter Two: Legal Overview 20 Alternative Constitutional Principles 22 Race-based Sentencing Disparity as De nial of Substantive Due Process 23 Remedies 26 Section 15A-2000, NC General Statutes 30 Chapter Three: Empirical Research and Statement of the Problem 32 PreFurman Research 32 Post-Furman Studies to 1990 33 Post-1990 Studies 37 Research Problem 45 Hypotheses 49 Chapter Four: Methodology 51 Data 51 Variables 56 Variable Recoding 57 Statistical Analysis 60 Chapter Five: Results 63 Models of Total Aggravating and Mitigating Circumstances Accepted 63 All-cases model. 63 Defendant race-specific models. 64 Victim race-specific models. 68 All defendant-victim racial combination models. 69 Accepted Individual Aggravating and Mitigating Circumstances 73 All cases. 74 Defendant race-specific models. 75 Victim race-specific models. 78 Racial combination models. 84 i


Chapter Six: Conclusions 88 Study Purpose and Results 88 Key Findings 90 Implications of Findings, Li mitations of Study, and Need for Additional Research 95 References 98 Articles 98 Statutes 104 Caselaw 105 ii


List of Tables Table 1 Frequencies, Variables a nd Coding: North Carolina Capital Penalty Trials, May 1990 December 2002. 58 Table 2 Logistic Regression Results (Odds Ratios): All-cases and Defendant and Victim Race-Specific Results Showing Effect of Total Jury-Accepted Aggravating and Mitigating Circumstances on Capital Sentencing Outcomes, North Carolina (May, 1990 December, 2002). 65 Table 3 Logistic Regression Results (Odds Ratios): Race-of-Defendant and Race-of-Victim Combinations Showi ng Effect of Total Jury-Accepted Aggravating and Mitigating Circum stances on Capital Sentencing Outcomes, North Carolina (May, 1990 -December, 2002). 72 Table 4 Logistic Regression Results (O dds Ratios): All-Cases, Defendant and Victim Race-Specific Models Showing Effect of Jury-Accepted Individual Aggravating and Mitigating Circumstances on Capital Sentencing Outcomes, North Carolina (May, 1990 December, 2002). 79 Table 5 Logistic Regression Results (Odds Ratios): Defendant-Victim Racial Combination Models Show ing Effect of Jury-Accepted Individual Aggravating and Mitigating Circumstances on Capital Sentencing Outcomes, North Carolina (May, 1990 December, 2002). 86 iii


Assessing the Issue of Arbitrariness in Capital Sentencing in North Carolina: Are the Effects of Legally Relevant Variables Racially Invariant? Judith Kavanaugh Earl ABSTRACT This study analyzed case and sentencing data from 632 capital cases involving Black and White defendants and victims, processed in North Carolina from May 1990 through December 2002. Logistic regression analysis of all cases and race-specific data allowed assessment of the variable effects of jury acceptance of statutory aggravating and mitigating factors on capital sentencing outcomes (death versus life). The purpose was to evaluate the role race plays in shaping jury use of legally defined factors in capital sentencing. Significant variance in the eff ect of jury acceptance of aggravators was observed between Black and White defendants. Black defendants pay a higher premium in terms of the risk of a death sentence than do White defendants whose crimes are comparably aggravated. There was no overall di sparity in the effect of jury acceptance of mitigatory factors observed, although certain mitigators reduced the risk of a death sentence significantly more for Black or White. Overall, the aggravators had a statistically significantly stronger effect on sentencing outcomes than did the mitigators, regardless of race, and on cases involving Black defendants, regardless of victim race. Racial invariance was not shown. iv


1 Chapter One Introduction It would seem to be incontestable th at the death penalty inflicted on one defendant is "unusual" if it discrimina tes against him by reason of his race, religion, wealth, social position, or cl ass, or if it is imposed under a procedure that gives room for the play of such prejudices. Furman v. Georgia, 408 U.S. 238 (1972), DOUGLAS, J., concurring. The primary purpose of this paper is to explore the extent to which the guided discretion approach to death sentencing under the seminal holding in Furman v. Georgia, 408 U.S. 238(1972) has achieved racial neut rality in the life-death decision. This is accomplished through stat istical analysis assessing the presence or lack of racial invariance in capital jury application of the North Carolina capital sentencing statute from 1990 to 2002. The secondary goal of the paper is to explore the viab ility of statistical analyses as proof of nonpurpos eful racial discrimination, which is potentially actionable as violative of constitutional principl es requiring evenhanded punishment, jury impartiality and substantive due process. To frame the empirical issues, it is n ecessary first to understand three pivotal United States Supreme Court decisions: 1) Furman v. Georgia, 1972 (prohibited standardless death sentencing; commuted all pending sentences to life and placed moratorium on further death sentenci ng); 2) the combined holdings in Gregg v. Georgia (1976), Proffitt v. Florida (1976) and Jurek v. Texas(1976) (established the two-pronged


2 guided discretion paradigm for death sentencing); and McCleskey v. Kemp (1987) (rejected statistical showing of group-based racial disparity as evidence of purposeful discrimination against an individual defendant; established undefined constitutional significance standard for st atistical showing of Eighth Amendment arbitrariness in operation of Furmancompliant sentencing procedures). Furman v. Georgia (1972): The Eighth Amendment Argument Ground zero for current capital punishment jurisprudence is the landmark case of Furman v. Georgia, 408 U.S. 238 (1972). In Furman the Supreme Court rejected the unlimited discretion previously afforded capita l juries in making th e life-death decision as violative of the Eighth Amendments prohibition of excessive, cruel and unusual punishment. This approach to the life-de ath sentencing choice was based on a new interpretation of what ma y constitute cruel and unus ual punishment under the Eighth Amendment. Only a year before Furman, in McGautha v. California (1971) and its consolidated companion case, Crampton v. Ohio (1971) the Court had rejected Fifth and Fourteenth Amendment depriva tion of procedural due proc ess arguments against capital sentencing procedures which employed a single jury for guilt and sentencing and allowed standardless jury discretion under the Calif ornia and Ohio capital sentencing statutes, finding the procedures adequate and fair with in the meaning of those two constitutional provisions and that the exercise of discretion is essential to the sentencing process. In Furman, the Court was presented with the different question of the constitutionality of standardless deat h sentencing discretion under the Eighth Amendments prohibition of excessive, cruel or unusual punishment, regardless of the procedural fairness surrounding the decision. This time a majority of the Court found


3 that, although unbridled jury discretion may be exercised in a way which meets procedural due process requirements, the lack of meaningful standards for deciding when a death sentence is appropriate amounts to unacceptable arbitrariness under the Eighth Amendment. The lack of standards capable of producing consistent and predictable outcomes for the same crime was found to be excessive, cruel, and unusual punishment, freakishly imposed, in the words of Justice Stewart ( Furman, Stewart, J. concurring). Among the reasons cited by the concurring plurality, and the dissenters as well, for distinguishing death senten cing was the acknowledgement that death is different in its irrevocable finality; and a lthough traditional in our system, death is unique as the ultimate penalty and requires maximum care to ensure equitable and correct decisions ( Furman, 1972) The potential for racial discrimi nation in death sentencing was not an express basis for the Furman holding, but was specifically cited as one of the risks, if not certain result of absolute jury sentencing discretion by Justices Douglas, Brennan, Marshall, and acknowledged by dissenting Justices Burger and Powell. Unlike other potentially legal and illegal bases for discriminatory governmental action, protection from racial discrimination finds its roots in our hi story and the Constitution itself; the Amendments were all enacted to end slavery a nd cure its ills by afford Black persons full citizenship and equal protection of our laws. If death is different, so is race, and where racial disparity is demonstrated in operati on of the law, it demands scrutiny by the courts for constitutional acceptability. It should also be noted that the dissenter s did not express a belief that the death penalty as then administered was fair or color-blind, but stressed reluctance under principles of judicial restraint to interfere with what was seen as a state legislative


4 function, or to expand so broa dly the interpretation of th e Eighth Amendment. In a foreshadowing of the controversy as to what constitutes proof of racial inequity in capital sentencing to be addressed fifteen years later in McCleskey, Justices Burger and Powell also questioned the degree to wh ich the Court should rely on the statisti cal evidence cited by members of the plurality in support of their positions ( Furman, 1972, BURGER, J. dissenting, at note 12; POWELL, J. dissenting ) Then there was the question of remedy. Among the concurring plurality, only Justices Brennan and Marshall pressed for ab solute rejection of death as a penalty, a position both maintained until their deaths. Justices Douglas, White and Stewart opted for rejection of the death penalty under then ex tant state statutes but did not rule out the possibility that a statute coul d be drawn which would limit ju ry sentencing discretion to impose death in constitutionally acceptable ways while pres erving the jurys ability to grant leniency in specific cases ( Furman, 1972 ) The result of Furman, then, was invalidation of existing state capital sentencing statutes a moratorium on further imposition of the death penalty under such st atutes, and commutation of all pending death sentences to life in prison ( Furman, 1972). Many believed, or at least hoped, that Furman effectively abolished death as a penalty in this country; but that turned out not to be the case (Radelet, 2001). Furman Implemented: Gregg/Proffitt/Jurek The state legislative response to Furman was immediate. Within four years, thirty-five (35) states had re -enacted capital sentencing statutes intended to incorporate the Furman principles, and five cases had ma de their way to the Supreme Court: Gregg v. Georgia, 428 U.S. 153 (1976), Proffitt v. Florida, 428 U.S. 242 ( 1976 ) ; Jurek v. Texas,


5 428 U.S. 262 (1976); Woodson v. North Carolina, 428 U.S. 280 (1976); and Roberts v. Louisiana, 428 U.S. 325 (1976). Three of the states, Georgia, Florida and Texas, whose capital sentencing systems were brought before the Court in Gregg, Proffitt, and Jurek, adopted procedures whereby the capital sentencing discretion was appropriate ly guided in a guided discretion model as impliedly acceptable under Furman. The Gregg/Proffitt/Jurek holdings collectively represent the second stage implementation of the new death sentencing approach announced in Furman. Each state statute provided two subs tantive, interacting criteria to be used by the sentencing entity to decide when the death penalty would be appropriate: 1) standards defining and narrowing the cla ss of crimes for which death could be considered -the aggravating criteria ; and 2) mandatory procedures allowing comprehensive consideration of individual mitigating factors which might justify mercy in a particular defendants case ( Gregg/Proffitt/Jurek) Each state also provided an acceptable pr ocedural framework for the application of these standards, in the form of bifur cated guilt-penalty proceedings, automatic and expedited appeal of all death sentences to the respective state supreme courts, and mandatory proportionality review. 1 Although different in certai n procedural respects, the three statutory schemes approved in Gregg/Proffitt/Jurek have provided the template for all state capital sentencing procedures today (Bowers & Foglia, 2003; Holowinski, 2002). 1 This last procedure was made optional some ten years later in Pulley v. Harris, 465 U.S. 37 (1983), based on the Courts conclusion that efficient aggravation standards made proportionality review redundant. Some states, notably New Jersey and Florida, have retained proportionality review as an extra procedural safeguard.


6 The two remaining states, North Carolina and Louisiana, whose statutes were reviewed in Woodson and Roberts, had also enacted procedures which significantly limited and clearly defined the crimes for whic h death was an available penalty, but made death mandatory for those few crimes, wit hout allowing consideration of mitigatory factors. The Court emphatic ally rejected this approach as inconsistent with Furman, holding that while the jurys discretion to impose death must be limited, its discretion to grant life to individual defendants base d on their unique situation could not ( Woodson, 1976) The Furman approach thus dictates the narrowe st of discretion to impose death, while at the same time affords the broadest disc retion to grant life. It is in the application of these principles that nonpurposeful racial bias may have the most play when capital sentencing entities decide objectively whethe r a defendants crime deserves death, but subjectively whether the defenda nt himself deserves life. Fifteen years after Furman and a decade after its new take on the Eighth Amendment notion of punishment was implemented per Gregg/Proffitt/Jurek, the Supreme Court was presented squarely with the question of whether Furman and its progeny had produced a sentencing procedure whic h was producing equitable death sentencing not just among cap ital defendants generally but between Black and White defendants, in McCleskey v. Kemp, 481 U.S. 279 (1987). Race Neutrality Under Furman: McCleskey v. Kemp (1987) McCleskey v. Kemp (1987) is the third of the triumvir ate of holdings which have shaped capital sentencing law for the thirty-odd years since Furman, and it continues to be the decision most affecting efforts to redr ess documented racial bias in the capital justice system. Defendant McCleskey was an African-American convicted of the murder


7 of a White police officer and sentenced to deat h in Georgia. He br ought a habeas corpus proceeding seeking the overturn of his death sentence based on an assertion, that, despite meeting the Furman standards per Gregg, distortions in the app lication of the Georgia statute to Black citizens, reflected a racial bias infecting the Georgia capital justice system which 1) violated his rights to equal protection unde r the Fourteenth Amendment, and 2) resulted in sentencing arbitrarin ess as to him which violated the Eighth Amendment under Furman. No direct evidence of bias against McCleskey personally was claimed or shown. Denial of McCleskeys equal protection claim. In asserting that his deat h sentence was a denial of equal protection, McCleskey relied on a theory of disparate impact be tween Black and White defendants and victims from application of Georgias capital ju stice procedures, which was supported by a comprehensive statistical study of the Georgi a system (the Georgia Study) by Professor David Baldus and his colleagues (Baldus, W oodworth & Pulaski, 1990). The study used modern logistic regression me thodologies like those previously accepted by the Court to prove implied purposeful discrimination based on disparate impact in other nonsentencing situations, such as Title VII employment and ho using discrimination cases and jury venire selection ( McCleskey, 1987). McCleskey asserted that the Georgia Study established both an increased risk of a death sentence for Black defendants over White defendants who had committed comparable crimes, and for those who kill ed White victims, indicating an overall devaluation of Black citizens. The study, he argued, showed that the statute as applied placed Black capital defendants such as McCleskey at an unfair and arbitrary, higher risk


8 of receiving a death penalty than White defendants committing comparable crimes which denied him equal protection under the Four teenth Amendment and subjected him to punishment violative of the Eighth Amendment. The real issue was what level and type of proof would sustain a denial of equal protection claim base d on disparate impact, or an Eighth Amendment arbitrariness claim in a capit al sentencing case: This case presents the question whether a complex statistical study that indicates a risk that racial considerations enter into capital senten cing determinations proves that petitioner McCleskeys capital sentence is unconstitutional under the Eighth or Fourteenth Amendment. ( McCleskey, 1987) It must be stressed that McCleskey was not challenging the Georgia statute directly; his claim sought to ove rturn his individual sentence. While the overturn of his sentence on a disparate impact basis may have set precedent applicable to other members of the affected group Black defendants sent enced to death the relief he sought was specific to him and this distinction between individual and general remedies is important to understanding the holding in McCleskey (Graines, 2000). There is an intent element in an equa l protection claim which requires proof of purposeful discrimination against an individual or a group to tr igger the strictest judicial scrutiny racial classi fication requires ( Washington v. Davis (1976)) The purpose to discriminate may be either express dispar ate treatment or inferred from informed tolerance of differences in the impact of f acially neutral governmental action on a group. The inference requires a disparity in effect so significant as to imply an intent of adverse effect on the group and its indi vidual members that is, di sparate impact discrimination is actionable only where it is so significant as to support an inference that the


9 decisionmaker selected or reaffirmed a part icular course of acti on because of, not merely in spite of, its adverse effects upon the identifiable group. ( Personnel Administrator of Massachusetts v. Feeney, 1979). Unconscious, but palpable, discrimination is not actionable as a denial of eq ual protection ( McCleskey, 1987 ; Lawrence (1987)). The McCleskey court denied his equal protectio n claim, holding that the broad discretion inherent in capita l sentencing and the myriad of factors which inform that discretion in a jury capital sentencing deci sion in an individual case prevents use of statistical proof of genera l racial disparities as sole proof of actionable purposeful racial discrimination, within the meaning of the E qual Protection Clause, in imposing the death sentence in an individual case. The Court relied on it prior decisions holding that statistical or other evidence of disproportionate negative impact of governmental action on one group over another is not enough, standing alone, to support a findi ng of discriminatory intent ( Arlington Heights v. Metropolitan Housing Dev. Corp. (1977) ) except where the pattern of disparity is stark, citing Gomillion v. Lightfoot (1960) [redistricting to exclude 395 of 400 Black voters a 99% disparity] and Yick Wo v. Hopkins (1886) [requiring permits for 310 laundries in wooden houses, but declining to issue permits for any of the 200 Chineseowned laundries 100% disparity]. In rejecting the use of the proffered st atistical proof in McCleskeys case, the Court also noted flexibility in its will ingness to consider st atistical proof of discriminatory intent in Title VII cases a nd particularly distinguished its acceptance of statistical proof in jury venire cases (Castaneda v. Partida (1977) [2-1 disparity between


10 Hispanic population and grand jury composition]; Whitus v. Georgia (1967) [3-1 disparity between Black population and grand jury composition]; ( Turner v. Fouche (1970) [1.6 disparity between Black population and grand jury composition], but based this distinction on the level of discretion and number of decisionmakers involved in the allegedly discriminatory action. The Court held that the more decisionmakers, the more factors affecting the decision, and the larger the number of decisions affecting the group from which the plaintiff seeks to infer discrimination in his individual case, the further his claimed harm is causally from the challenged action and the less acceptable statistics become as sole proof of intent to discriminate (McCleskey, 1987). The Court held that an application of an inference drawn from general statistics to a specific decision in a trial and sentencing [context] simply is not comparable to the ap plication of an inference drawn from these other contexts where the discretion was more defined and exercised by fewer people ( McCleskey, 1987) The McCleskey Court acknowledged the potential fo r racial bias skewing capital trial and sentencing decisions, but saw it as th e inevitable consequence of the jury system and the broad discretion it requires. The Court noted the framework of procedural safeguards it had constructed by then to k eep racial bias below as yet undefined constitutionally acceptable le vels at other discretionary decision points in the process ( Batson v. Kentucky (1986) [systematic prosecutorial strikes of Black jurors, or demonstration of venire disproportionate to minority population is prima facie proof of discriminatory intent for equal protection purposes, shifting burden to prosecutor to provide race-neutral explanation for exclusion]; Irvin v. Dowd (1961) [widespread racial


11 bias can require change in venue]; Ristaino v. Ross (1976) [significant likelihood of racial biasing influence requires questioning of jurors about ra cial attitudes]; Turner v. Murray (1986) [high potential for racism biasing jurors in Black-on-White murder requires voir dire re racial opinions under Sixth Amendment jury impartiality guarantee] and that effort to perfect procedural fairness continues today ( Miller-El v. Cockrell (2003)) [prosecutorial intent to discriminate racially in jury selection within meaning of Equal Protection Clause in jury selection may be inferred from practices such as venire shuffling and disparate questioning of White and Black jury candidates]. It is reasonable to conclude that McCleskey absolutely forecloses the use of statistical evidence of group disp arities as sole, competent pr oof of purposeful denial of equal protection in individual capital sentences; at a mini mum it expressly held that something more than the mere correlation between race and deat h sentencing the Court found the Georgia Study produced is necessary (McCleskey, 1987). No case has attempted to rely on an equal protection argum ent proffering solely statistical proof in a sentencing context postMcCleskey (Pillai, 2001), and it is questionable whether such a claim could be sustained as to any criminal sentence (Kennedy, 1988). Perhaps McCleskey would allow an equal protection cl aim in a sentencing case, were the statistically demonstrated disparity were su fficiently stark, or a di fferent individual or class-based remedy for denial of equal pr otection pursued which would allow use of statistical to show racial disc rimination in capital sentencing (Graines, 2000). In light of the empirical studies to date, such a stark di sparity is unlikely to be found, and continued focus on equal protection theories begs the que stion of whether the cau se of the disparity is actually the result of purposeful discrimination a required element of an equal


12 protection claim or may be caused by nonpurpos eful racial bias pe rhaps violative of other constitutional provisions such as the Eighth Amen dment, also addressed in McCleskey. McCleskeys Eighth Amendment claim. Having rejected McCleskeys equal prot ection claim as unproven, the Court next looked at the sufficiency of the Georgi a Study to demonstrate arbitrariness in McCleskeys sentencing result, violative of the Eighth Amendment under Furman. The Court accepted the Georgia Study as valid for purposes of the Eighth Amendment claim as it had for the equal protection claim, but where it may rejected the Georgia Study more as being incompetent to sustain an inferen ce of discriminatory intent in a sentencing context, it rejected it in th e Eighth Amendment context as being insufficient to document a racial bias strong enough to rise beyond constitutionally acceptable levels, holding the disparity demonstrated in th e study lacked the constitutional significance to be arbitrary or capricious within the meani ng of the Eighth Amendment. The Court noted that McCleskey coul d not say his death sentence was disproportional to his crime, as it involved a ggravating circumstances the Georgia statute had defined in sufficiently precise terms to qualify it for the death penalty. Nor could he say that the Georgia statute was defective in allowing the jury broad discretion to considering mitigatory factors when deciding lif e or death in his or any other case, since it had been upheld already in Gregg, no racial bias had been asse rted in its application to him personally, and Woodson required jury discretion in sentencing. The only possible basis for an Eighth Amendment claim was show ing constitutionally significant disparity


13 and this is the puzzle the McCleskey decision poses: What level of racial prejudice must be shown to trigger c onstitutional significance? Without explaining what level it might ha ve found constitutionally unacceptable, the McCleskey court took more of a well-know-i t-when-we-see-it approach, and did not see it in McCleskeys case. The Court noted that the jury is the criminal defendants fundamental protection of life and liberty against race or color prejudice ( McCleskey (1987), citing Strauder v. West Virginia (1880)) and that the purpose of requiring that the composition of a capital sentencing jury reflec t the criminals community is to assure diffused impartiality (citing Taylor v. Louisiana (1975)). The Court acknowledged again that individual jurors bring to the sentencing decision their personal attitudes, experience and other qualities which cannot be reduced to range of factors to which legislation can apply. Still, the Court rejected any argument that the broad discretion to be lenient is per se constitutionally unacceptable, and confirmed that the broad discretion for mercy is absolutely essential to the capit al sentencing decision pr ovided it is exercised in the facially fairest possible setting. Having accepted that McCleskeys sentence was proportional and the result of scrupulously fair procedures fo r the exercise of the senten cing discretion, but agreeing that racial prejudice is alwa ys a risk in sentencing situ ation, the Court examined the Georgia Study to determine exactly what it sh owed and whether it rose to the level of constitutional signifi cance requiring further i nquiry. The Court relied heavily on the trial courts analysis of the study, which had resulted in a rejection of the findings as flawed by problems with the methodology, such as sample size and selection, coding problems, variable selection and missing da ta: It is a major premise of a statistical case that the


14 data mirrors reality. If it does not in substantial degree mirror reality, any inferences empirically arrived at are untrustworthy ( McCleskey, at footnotes 5-7). The Supreme Court elected to accept the study as valid, as ha d the appellate court, but held that the results did not rise to constitutional significance within the meaning of the Eighth Amendment. The Court focused on two points: 1) logi stic regression analysis quantifies only the degree to which the risk of the outcom e occurring is increase d or decreased by the factor analyzed and the study failed to produce statistically significant ra cial disparities in the risks of a death sentence; and 2) the instability and lack of predictive efficiency in the models used, as evidenced by low R 2 values, weaken the validity of any inference of constitutionally unacceptable race-b ased sentencing disparities. The unadjusted data showed a notable race-o f-victim effect (11% of White victim cases resulting in a death sentence versus 1% of Black victim cases) and a slight, reversed race-of-defendant effect (4% Black defendants versus 7% White defendants with death sentence). When the data was subjected to logistic regression an alysis, the only racial effect demonstrated was a significant increas e [a multiplier of 4.3] in the risk of any defendant receiving the death penalty if his victim was White but a Black defendant was almost as likely to receive a death sentence as a White defendant, his risk being only 1.1 times higher. The highest fit was for a 230-variable model which attempted to incorporate all possible influences on a capital sentencing decision, with an R 2 of between .46 and .48, meaning it predicted th e sentencing outcome in less than 50% of the cases ( McCleskey at footnote 6).


15 Thus it can be concluded that 4.3 multiplie r of the risk of death sentencing, at least based on the race of a victim, was not significant enough in the eyes of the McClesky court to be constitutionally significant for Eighth Amendment purposes. It can also be concluded that logistic regression m odels with a predictive value of less than 50% are not sufficiently predictive to relay on their results for a finding of constitutionally unacceptable levels of race bias in a capital sentencing context. What higher disparities and predictive efficiencies might trigger th e Eighth Amendment or the Sixth Amendment protections against constitutionally significan t racial bias are unknown, but the levels shown in McCleskey were not enough. No case has since reached the Court rely ing solely on statistical evidence for a challenge to the death penalty. Unfortunatel y, despite three decades of fine-tuning these models and the construction of a complex system of super due process safeguards around them, the empirical research still finds significant unexplained disparities between Black and White citizens in the administra tion of capital justice. (Baldus & Woodworth, 2004; Bowers & Foglia, 2003; Howe, 2004). What remains in dispute is whether this research is sufficient to raise constitutional re d flags as to the con tinued legality of the death penalty. It is also like ly that any significant disparity found is not in tentional, but the result of unconscious racial attitudes, in light of th e continued integration of society. Latent Racism versus Blatant Racism Open racism is no longer politically co rrect, and it can fairly be said that Furman and its progeny have largely eliminated purposeful racism from death sentencing (Howe, 2004), although it may still exist in cert ain, defined geographic areas (Baldus & Woodworth, 2003; Paternoster & Brame, 2003). Many argue that latent racism still


16 infects capital justice, not in the form of ope n prejudice, but as the cu ltural relic of three centuries of slavery and Jim Crow (Howe, 2004; Lawrence, 1987). The McCleskey court acknowledged that race-based sentencing disparities exist, citing various studies and noting Congress development of sentencing guidelines for criminal cases to reduce unguided discretion arising out of individual predispositions not just in capital cases but in all criminal cases ( McCleskey at footnotes 35, 38), while at the same time accepting some level of racism as an unavoida ble by-product of the jury system. As we have moved further from the 1970s, and the civil rights movement has cooled from confrontation to implementati on of the laws intended to cure racial discrimination, the nature of racism has cha nged. Critical race theorists maintain that although open racism has dwindled, latent ra cism is still a pern icious and pervasive negative force in society (Lynch & Patterson, 19 96). The source of latent racism is a topic for theoretical debate with the so-called idealists, who have most affected the legal system in this country, viewing racism as a ttitudinal and thus capable of change over time as the law forces social change a nd acceptance of disparate groups (Lawrence, 1987). The realists among postmodern theorists view raci sm as the product of the continuing struggle between the classes in a capi talist society, with those in power always consciously and unconsciously acting to mainta in their status and suppress those below by keeping them in their place (Delgado, 2001). Both sides agree that racism, both latent and blatant, is still an ac tive force in our culture. Intentional racism is relatively easy to eliminate because it is the result of conscious personal choices and thus can be a ddressed through educati on and the threat of legal consequences. Latent racism is unint entional and nonpurposeful; it just happens as


17 part of the social, cultural, geographical, re ligious and even political conditioning which occurs in development of individual pers onalities and attitudes (Lawrence, 1988; Lenhardt, 2004). It thus is not subject to traditional legal remedies applicable to intentional, purposeful conduct, such as equal protection cl aims, and can only be cured by time and proximity to the object of th e unconscious negative attitudes. To fear those of a different race and t hus different from ones self is human, apparently a learned survival reaction whic h can be unlearned through proximity (hme, 2005). The unconscious attribution of negative traits to Black persons may occur among White capital jurors who still make up the ma jority of juries, despite even the most conscientious effort at intellectual racial neutrality because the trial and sentencing experience is filtered through each individua l jurors viewpoint, life experience and attitudes, including racial attitudes of both Black and White jurors (King, 1993; Eisenberg, Garvey & Wells, 2001). Other factors which affect jury decision ma king are internal to the process itself, a process in which many capital jurors misunders tand their role so that their decision may be premature, arise out of incomprehension of the instructions given, or an erroneous belief that a guilty decision requires the death penalty or that a life decision may result in release of the defendant, and so on (Bowers & Foglia, 2002; Foglia, 2003 ). The effect of these factors will also interact with racial attitudes, although none of these factors can be attributed to a conscious desire to discriminate against a defe ndant based on his race. All can be attributed to individua l juror characteristics, incl uding unconscious racial bias, which cannot be standardized by additiona l due process but may still significantly undermine the fairness of the sentencing process.


18 The difficulty lies in detecting a nd quantifying unconscious and thus nonpurposeful racism in order to determine the degree to which it may affect capital sentencing decisions under Furman. To date the issue of nonpurposeful discrimination resulting in racially variance in capital senten cing has not been raised before the Supreme Court. If such bias exists th ere are no directly applicable legal standards to cure it. The primary empirical objective of this study was to detect and measure racial variance in application of the Furman standards as embodied in a Furmancompliant state capital sentencing law, in this case, that of North Ca rolina. Evidence that the apparently neutral application of the Furman-based, legally relevant sentencing standards the aggravating and mitigating circumstances accepted by a capital jury to similarly situated defendants -is not racially i nvariant casts doubt on the legality and the morality -of continuing to use the Furman model to send defendants to their death more because of their race than th eir crime. A finding of racial variance may indicate, as Justice Blackmon concluded to wards the end of his life 2 that the Furman experiment has failed, and that the Constitution demands a different approach to punishment of the most serious crimes. It may also be indicative of latent nonpurposeful racism still skewing the process which cannot be addressed by additional procedural fix-its. The results provide a basis for alternative constitu tional arguments under which the Supreme Court may be willing to consider supporting statistical evidence in evaluating the racial disparity in 2 As one of his last acts on the Court, Justice Blackmun receded completely from his position in Furman and voted to abolish the death penalty: Rather than continue to coddle the Courts delusion that the desired level of fairness has been reached I feel mo rally and intellectually obligated to concede that the death penalty experiment has failed. ( Ca llins v. Collins (1994), BLACKMUN, J. dissenting).


19 capital sentencing procedures and find such disparities constitutionally unacceptable and repugnant to fundamental concep ts of racial equity. This introductory chapter has outlined in the broadest terms the legal setting for the discussion. Chapter Two presents a more detailed summary of the constitutional and legal issues at play in death sentencing and the effect of the McCleskey decision on the use of statistical analyses to prove racial disparities. Chap ter Three provides an overview of the extant empirical studies on the influen ce of race on capital justice in this country and defines the research problem presented and the hypotheses tested. Chapter Four describes the North Carolina data and the lo gistic regression methodology used. Chapter Five presents the results of the logistic re gression analyses performed. Finally, Chapter Six discusses the results and draws conclusions as to the meaning and possible use of the findings in the continuing effort to ma ximize race-blind decision-making in capital sentencing.


20 Chapter Two Legal Overview Chapter One explained the import of Furman, Gregg/Proffitt/Jurek and McCleskey on the continued use of the death penalty in this country, despite the potential for race-based disparity. It is important to note several things about these cases. First, they are all capital sentencing cases and do not apply directly to the other discretionary decisions which lead a defendant to the lifedeath decision and might be racially-biased. Thus studies showing racial bias at the prosecutorial decisions may inform these sentencing holdings, and vice-ve rsa, but are not directly on point insofar as their legal reasoning. Secondly, although McCleskey held that only purposeful discrimination may be redressed under equal protection principles and rejected the use of statistical proof to imply intent, it did not require intent to disc riminate in an Eighth Amendment context, nor did Furman or Gregg. Rather the question posed and left unanswered -was what level of racial disparity would be constitutionally significant. Thirdly, no case has yet decided what level of statisti cally demonstrated race-based disparity in capital sentencing would trigger constitutional protections under doctrines other than equal protection, although McCleskey provided guidance as to what wont. The final major issue raised by these cases is the choice of remedy. For example, McCleskey was a habeas corpus proceeding seeki ng to overturn a specific defendants


21 sentence as a denial of equal protection under the Fourteenth Amendment, where Furman and Gregg/Proffitt/Jurek sought to answer the broader le gal question as to whether the death penalty statutes then administered in Texas and Georgia and echoed in other capital states resulted in cruel and unusual punishment under the Eighth Amendment. While all of these cases sought specific relief from death sent ences, relief in McCleskeys case would have been specific to him that his particular senten ce was unconstitutional and involved a very indivi dual burden of proof of purposeful harm to him as an individual notwithstanding the fact that his success would have effec tively invalidated the Georgia statute. Furman and Gregg were appeals of state supreme court decisions, petitions for writ of certiorari seeking the answer to broad lega l questions of general applicability, and thus affected directly all death sentences re ached prior to their holdings. The choice of remedy affects the burden of proof and the weight accorded the statistical evidence. A habeas proceeding is an original proceeding, and the Court is not limited to the record and legal questions as in certiorari proceedi ngs, but the burden of proof is heavier. McCleskey applied habeas corpus eviden tiary standards [exceptional, clear and convincing proof] to an indivi dual equal protection and Eighth Amendment claim seeking individual relief. Because of th e uncertainty in any st atistical analysis and the fact this McCleskeys claim was suppor ted only by statistical proof implying the necessary discriminatory purpose and arbitrarin ess, he simply did not meet his burden of showing purposeful discrimination. Had he ra ised claims in a proceeding with a lesser evidentiary requirement, or whic h did not have an intent el ement, or had the statistics been stronger, he may have fared better and certainly would have a lower bar to jump.


22 Racial disparity, be it intentional or not, which skews so irrevocable a decision as the life-death decision in a capital case, rais es a Constitutional alarm to which the court must at some point respond. Since the onl y way to prove systemic bias is through statistical studies which are by definition only estimates, not absolute fact -McCleskey seems to have eliminated equal protection as a basis for challenging racially disparate capital sentencing by requiring a showing of r acially discriminatory intent. Whether any statistically proven level of subs tantive latent racism in the capital justice system is or is not acceptable under our Constitution is beyond the scope of this analysis, except to frame the issues and suggest possible alternative legal vehicles to resolve them in the courts but there are arguments to be made. Alternative Constitutional Principles There are four constitutional provisions pertinen t to the issue of significant racial disparity in application of the death pena lty. The Fifth and F ourteenth Amendments guarantee substantive and pro cedural due process of la w before any governmental deprivation of life, lib erty, or property. It has been ar gued that substantive due process requires racial equity in capital sentenci ng (Bird, 2003) and it w ould seem fundamental that the Constitution would abhor as a matter of substance over form -a demonstrably racially inequitable result of application of a law, regardless of how procedurally perfect it may be. As to Sixth and Eighth Amendment concerns the Court has already indicated that some level of racial disparity in capital sentencing outcomes is unavoidable, but at an as yet undefined point such disparity may reach levels unacceptable under Sixth Amendment guarantees of trial and sentenci ng impartiality or amount to arbitrary and


23 capricious action violative of the Eighth Ame ndment. The Court has already expressly acknowledged that Black-on-White crime poses the greatest risk for play of racial prejudice ( Turner (1986); McCleskey (1987)) Statistical analysis is the only means to measure racial disparity inte ror intra-racial -when looki ng at the overall, long term operation of a statute As di scussed in Chapter One, there is no guideline beyond the Courts consideration of the Eighth Amendment argument in McCleskey as to what level of disparity would be unacceptable other than it should reach statis tical significance and be the product of a better-t han-50% predictive model whic h incorporates appropriate variables to allow reasonable comparison of similar levels of criminal culpability among the defendants. Once a statistically signifi cant disparity is demonstrated, these two constitutional protections are arguably tr iggered and thus may provide a basis to challenge the race-neutrality of a statute. Wh at the Court would make of such a disparity, or what degree of disparity might offend c onstitutional sensibilitie s is unknown. As to the possibility of a substantive due process cl aim, several legal and proof obstacles would have to be surmounted, but such a claim may also be viable. Race-based Sentencing Disparity as De nial of Substantive Due Process Legal scholars have argued that substan tive due process principles might sustain for a challenge to the death penalty because of the higher, strict scrutiny the Court must afford to claimed deprivations of fundame ntal rights (Bedau, 1996; Bird, 2003) and the heavier burden placed on the state to show a compelling interest which cannot be protected through other means ot her than the discriminatory action which also must be shown to be tailored as narrowly as possible to do so. Once the claimant demonstrates the existence of the fundamental right a lega l question -and the fact that it is being in


24 some way infringed upon by the state action in question a mixed question of law and fact -the primary burden of proof and persuasion shifts to the state. The viability of a substantive due process claim based on racial disparity has yet to be explored by the Court in a death penalty case, primarily because other cons titutional protections have been available to redress raci al discrimination in the past. If, however, none of the other constitutional provisions are adequate to support a claim, because of proof or procedural issues peculiar to a capital sentencing situ ation, a substantive due process claim might suffice. Substantive due process is a judicially created doctrine which holds that in addition to the rights specifically enumerat ed in the Constitution, certain rights are so fundamental to life and libert y that they are inherently protected by the Constitution despite the lack of guidance from the mo re specific provisions of the Constitution (Crump, 1996). No fundamental right to life has yet been fully recognized by the Court (Bird, 2003). Most of the cases touching on this issue have been based in fundamental liberty interests, such as the right to privacy ( Griswold v. Connecticut (1967) [establishing a womans privacy right to birth control]; Roe v. Wade (1973) [womans privacy right to chose whethe r to be pregnant or not]; Washington v. Glucksburg (1997) [no privacy right to die, even for terminally ill patients Other liberty interests recognized by the court are freedom of unw arranted governmental infringement of personal movement (County of Sacrament o v. Lewis (1998)) and travel (Kent v. Dulles(1958) ), and bodily integrity ( Rochin v. California (1952) ). The right to life has been discussed in the context of excessive force cases which resulted in death, from a police chase, and the Court has said it requir es something more than negligence and less


25 than intent to harm ( County of Sacramento v. Lewis, 1998). There is also a murky question of what, if any, substantive due pro cess rights exist in constitutionally fair procedures. In Albright v. Olivier, 510 U.S. 266 (1994), the Court held that there was no substantive due process liberty interest in the prosecutorial charging decision being based on probable cause in a case where a defenda nt was arrested and charged, but later released by the court upon a finding of no pr obable cause for either. A distinguishing factor, however, was that there had not been a palpable consequence to the defendant, who had turned himself in voluntarily ( Albright 1998). Here the harm is palpable, finite and irreparable. Although there is a certain optimism reflect ed in the legal literature about the viability of a fundamental right to life in a challenge to the death penalty (Bird, 2003), it seems unrealistic in todays political climat e, and may be unnecessary. An alternative and perhaps more politically palatable ar gument could be made that there is a fundamental liberty interest in a non-arbitrar y, racially neutral cap ital sentencing, which is not protected by application of the Furman standards that is, that liberty within the meaning of our law requires racial neutrality or at least statistically significant equity -in punishment. What is the point of perfect procedural due process if the results are still markedly unfair, regardless of the lack of purposeful discrimination because the bias is culturally based, not consciously applied? Th is would seem especially important when dealing with a protected clas sification such as race and a permanent and irrevocable loss of liberty and body. There is also a level of proof issue: the Supreme Court has held that the governmental action asserted to violate due process must be egregious or shock the


26 conscience ( County of Sacramento, 1997; Rochin, 1952). Would continued execution of citizens in the face of documented and statis tically significant racial disparity in the imposition of the death penalty shock the consci ence? The level of proof in a substantive due process claim would possibly ha ve to be as stark as the M cCleskey court implied for Eighth Amendment arbitrariness and if s o, could be resolved under that provision. Another obstacle to pursuit of a substantive due process analysis is that the Court will not consider it if any other protect ion specifically afforded under the Constitution also applies ( Graham v. Connor (1989)). Substantive due process violati ons can be pled, but will not be addressed by the Court if any other const itutional principle is directly raised and applies. To the extent, then, that the Sixt h or Eighth Amendment arguments are raised and found to apply regardless of whether th e Court finds the action in question violates them the substantive due process argument may never be reached. The Graham court suggested, although it has not yet been conc lusively held, that all post-conviction complaints about the death penalty as exces sive or unjustified must be brought under the Eighth Amendment. Since it is unknown whether either the Sixth or the Eighth Amendments can support a challenge to racial inequity in death sentencing caused by nonpurposeful racial bias proven with statis tical studies, it is also unknown whether a substantive due process argumen t is available at all, but Graham seems to make it the argument of last resort. Remedies Finally, even where a constitutional viola tion can be shown, the choice of remedy will affect how the Supreme Court will interpre t the Constitution in particular cases, as the McCleskey case shows. This is particularly important when the only basis for the


27 claim of a constitutional violation is a statistical analysis producing general results as to a group which is being asserted as evidence of ha rm to an individual. That is one reason why remedies such as habeas corpus and ot her appellate remedies seeking individual relief from a death sentence may not be the be st vehicle to challeng e an individual death sentence, as McCleskey demonstrated (Graines, 2000; Vetter, 2004). Habeas corpus [literally let you have the body] is an attack by a person in custody on the legality of his confinement, and is the appropriate remedy where a prisoner is attacki ng the fact or duration of his custody (Preiser v. Rodriguez (1973)). Habeas is a procedure particularly unsuited for a challenge to a death sentence based on statistical evidence of a pattern of racial disparity in sentencing decisions. The level of proof is high requiring exceptionally clea r proof of an abuse of the sentencing discretion, and may not be too speculative ( Bracy v. Gramley, (1997) ). Discovery rights are limited and individua l state and federal post-conviction remedies must first be exhausted (Graines, 2000). It has been suggested that Title 42, United States Code, Section 1983 (2004) [Section 1983], which affords a remedy for constitutional injuries arising under state statutes (color of state law), might prove a better vehicle to present race-based discrimination claims as to death sentencing and that a class action might be more appropriate where statistical analyses are the only proof of such disparity (Graines, 2000). A Section 1983 action is a civi l action a kind of constitutional tort action -brought in federal or state court, and is thus subject to the more relaxed procedural and evidentiary standards afforded in civil cases which particularly lend credence and weight to valid statistical evidence of the type found wanting in McCleskey (Graines, 2000). It also


28 offers systemic relief in the form of injuncti on, declarations as to the constitutionality of the state action or law, recovery of atto rneys fees on success, and possibly money damages. Unlike habeas or an individual appeal, S ection 1983 claims may be brought either individually or on behalf of a class of similarly harmed claimants for example, all death row defendants sentenced under a sentencing stat ute asserted to be unconstitutional. No one has tried a class action to raise racial discrimination in death sentencing and there are limits as to the kind of claim a convicted prisoner may assert in a Section 1983 action (Graines, 2000) and a question of when such a claim might ripen in the procedural path a death sentence follows post-convic tion. However, recently, in Nelson v. Campbell, 541 U.S. 637 (2004) the Supreme Court reaffirmed the vi ability of an individual death row inmate using Section 1983 to challenge the method of his impending execution. In Nelson the prisoner sought Secti on 1983 relief from the use of the painful open-cut method used by Alabama for lethal injection, where the prisoners arm is slit open to reveal a vein, insuring proper insertion of the needle. Nelson sought injunctive relief against the use of that procedure in his execution, and preliminary injunctive relief from execu tion by lethal injection so long as that procedure was used by Alabama, claiming it was cruel and unus ual under the Eighth Amendment. A deciding factor in the S upreme Courts approving the use of a Section 1983 action in this case was the defendants pr ior unsuccessful completion of the habeas corpus process and other state remedies. Section 1983 thus may not be available to an individual before he has sought habeas cor pus relief and any stat e remedies, including administrative relief, and has failed, unless he can show that such remedies would have


29 no chance of success (Nelson, 2005). This may apply to class actions as well the members of the class may be limited to t hose death row inmates who have exhausted other remedies. Section 1983 cannot be used to chal lenge a death sentence directly ( Nelson 2005). A key factor noted by the Nelson court for the allowing a Section 1983 claim in a death sentence context was the fact he was not attacking his conviction or sentence, only the method of execution. Thus the claim could not assert that the defendant or class was innocent, or that the death penalty was not proportional to or even appropriate for the crime. A claimant, or claimants, would ha ve to concede guilt a nd the death-worthiness of their crime, and attack not their sent ence but the statute and how it was applied, seeking not a reversal bu t alternative relief from applicati on of the statute challenge in the form of life sentence without parole -instead of death, as a punishment It is unlikely that the current Court woul d totally abolish the death penalty, even should it find the Furman approach to be lacking. An action raising questions of racial bias skewing app lication of the Furman standards and thus the imposition of the death sentence would more likely be successful seeking preliminary injunctive relief from executions until factual questions as to the existence and degree of racial disparity produced by application of a states death sentencing stat ute was settled. Should the claimant demonstrate constitutionally unacceptabl e levels of racial disparity in who has received the death penalty and who has not, all that should be sought or granted is invalidation of the state statute in ques tion, commutation of all death sentences it produced to life-without-parole, and at a minimum a moratorium on further death


30 sentencing in the state until a way is found to prevent constitutionally significant disparity i.e., the Furman remedy. Alternatively, a preliminary injunction could seek a nationwide comprehensive exploration of the issue of raci sm to determine if latent bias can be sufficiently reduced or eliminated in capital sentencing under Furman, and request a moratorium on further sentencing or executions pending the outcome. A litigant would have to be prepared not only to present strong evidence to support any preliminary relief, but also a plan to complete such a study in a defined and not overly long time frame. The focus would not be the Furman standards themselves, but the question of whether any legal standard is sufficient, at this point in time, to remove si gnificant but unconscious racial bias from the capital sentencing decision. The details pro and con for this remedy are beyond the scope of this discussion, but such an action could provide a forum for full study of the efficacy of the Furman, unhindered by the more stringent, narrow ev identiary and proce dural constraints on individual appeals or remedies such as ha beas corpus. The research presented here provides a good starting point because the North Carolina statute is an excellent embodiment of the Furman principles, and there is a t horough and reliable database available for capital penalty proceedings postFurman. Section 15A-2000, NC General Statutes Under the North Carolina sentencing statute (Sec tion 15A-2000, NC General Statutes (2003)) capital juries have only tw o options, a death senten ce or a sentence of life in prison, currently on e without the possibility of parole except by executive clemency. North Carolina uses a bifurcated guilt-penalty procedure and is a weighing


31 state. It is a three step process: the jury first determines whether based on the evidence sufficient aggravators, limited to the eleven enumerated in the statute, exist unanimously and beyond a reasonable doubt whic h are sufficiently subs tantial to justify a death sentence. Once one or more aggrav ators have been found by the group, the jury next decides individually but based on the evidence if there are mitigating circumstances or circumstances applicable to the defendant. If so, each juror then decides if they are sufficient to outweigh the aggravating circ umstance or circumstances found, and they vote (Section 15A-2000(e), NC General Statutes (2003)). A single juror thus can hang a jury and ensure a life sentence -by finding a single mitigator justifies at least to him or her mercy for that defendant. There are eight mitigators specified in the statute, which also has a ninth, catch-all provision allowing jurors to c onsider any other circumstance arising from the evidence which the jury deems to have mitigating value (Section 15A-2000, NC General Statutes (2003)(f)). The final decision the jury must reach, unanimously, is whether, based on the first two findings the defendant should be se ntenced to death or to imprisonment in the State's prison for life (Section 15A-2000, NC Ge neral Statutes (2003)). The statute also requires, per Gregg/Proffitt/Jurek, automatic and immediate app eal of death sentences to the North Carolina Supreme Court, which re views for errors and also for proportionality to other similar crimes and defendants (Section 15A-2000(d), NC General Statutes (2003)). The North Carolina statute thus pr ovides an excellent framework in which to employ empirical analysis of the degree to which Furman has produced fair and racially equitable death sentencing.


32 Chapter Three Empirical Research and St atement of the Problem The intuitive question of whether there is ra cial disparity in the application of the death penalty in America has b een the subject of numerous sc ientific studies, of varying degrees of sophistication and validity. The research can be best be understood organized in three time frames: Studies of cases from the years preFurman, studies of cases after Furman went into effect through 1990 when a co mprehensive federal study of racial disparity in capital sentencing was released (GAO Report, 1990), and the research since 1990. Pre-Furman Research Studies of capital sentencing in cases before Furman showed a significant bias against Black defendants, particularly in the South. Guy Johnsons 1941 study, The Negro and Crime looked at the comparative freque ncy of death penalty sentences between Black and White defendants betw een 1940-1940, and found that 32% of Black murder defendants received the death pe nalty, compared to only 13% of White defendants. He also found a race-of-victim effect, with the death penalty being imposed in 17.5% of White victim cases, but in only .4% of the Black victim cases (Johnson, 1941). Garfunkel (1949) found similarly disparat e racial effects in his study of death sentencing patterns in ten North Carolina counties for the years 1930-1941. In 1969, however, a study of first-degree murder trials preFurman in California produced no


33 evidence of either a race-of-defendant or race-of-victim effect (Stanford Study, 1969). Wolfgang and Reidels 1973 study of racial patterns in de ath sentencing between 1945 and 1965 in several Southern states, including North Carolina, discerned patterns similar to those found earlier by Johnson and Garfunkel, which were also consistent among the other states examined (Wolfgang & Reidel, 1973). As noted earlier, by 1972, the possibility of racial bias in sentencing pract ices resulting in unfair death sentences had already been considered by the Supreme Court in McGautha/Crampton in which the process was upheld. In Furman, however, the Court decided that death, as the punishment produced from the procedurally fair process was not substantively fair or acceptable under the Eighth Amendment, and the law of death sentencing changed. Post-Furman Studies to 1990 The thrust of the next stage in the empirical research was to see if the Furman principles, as implemented in procedures established under Gregg/Proffitt/Jurek changed the racial distribution of death se ntencing. In 1981, five years after Furman-compliant processes began to be implemented, Gary Kleck reviewed th e published literature on racial effects on capital se ntencing, and reached the conc lusion that there no general racial bias was shown against Black defendants, except in Southern states; but that the death penalty was imposed less when the vic tim was Black, than when the victim was White (Kleck, 1981). Six years later, Nakell and Hardy published a st udy of the initial post-Furman years, 1977-1978, and application of the Furman-based sentencing models. Unlike earlier studies, they controlled for case seriousness and found a race-of-defendant effect at the prosecutorial charging decision, but not at the sentencing stage where a race-of-victim effect was noted (Nakell & Hardy, 1987). In 1989, Gross and Mauro


34 published a study of death sentencing patterns from 1976-1980, using FBI and law enforcement data from eight states. Like Nakell and Hardy, they controlled broadly for case characteristics and found a statistically significant in crease in the likelihood of a defendant receiving the death sentence when the victim was White, which remained consistent, although reduced when race-of-defen dant was added to the analysis (Gross & Mauro, 1989). The debate as to the existence of raci al effects on capital sentencing outcomes continued in the first two decades after Furman, to the point that in 1990, the United States Senate authorized the General Acc ounting Office (GAO) to examine the issue. Rather than conducting a time-consuming a nd expensive new analysis, the GAO elected to perform an evaluative synthesis of the twenty-eight studies of postFurman sentencing patterns, six of which were also post-McCleskey (GAO, 1990). Included were the studies of Gross a nd Mauro, Kleck, Nakell and Hardy, and Wolfgang and Riedel (GAO,1990), as well as studies of death sentencing patterns generally (Berk & Lowery, unpublished 1985; Bowers, 1983; Foley & Powell, 1982; Keil & Vito, 1989; Klein, Abrahamse & Rolph, unpublished, 1989; Radelet & Pierce, 1985; Radelet, 1981; Riedel, 1976) and studies stat e specific to Georgia (Baldus, Woodworth & Pulaski, 1990; Barnett, 1985); Texas (E kland-Olson, 1988); Louisiana (Klemm, 1986); Smith, 1987); Florida (Arkin, 1980; Foley, 1987; Lewis, Mannie & Vetter, 1979; Radelet & Vandiver, 1983; Zeisel, 1981), California (K lein, 1989), South Carolina (Paternoster & Kazyaka, 1988), Kentucky (Gennaro & Ke il, 1988; Keil & Vito, 1990), New Jersey (Benin et al., 1988), Illinois (Murphy 1984), and Mississippi (Berk & Lowery, unpublished 1985) (GAO, 1990).


35 The GAO assessed and rated the studies fo r quality in terms of design, selection of variables and sta tistical methodology (GAO, 1990). The evaluation was not limited to sentencing, where Furman applies, but encompassed the en tire capital justice process, looking at the potential for racial bias at each discretionary d ecision point (GAO, 1990). The GAO found a consistent race-of-victim effect, although it acknowledged that some studies, such as the Kiel and Vito (1990) study found race-ofdefendant effects, and that Paternoster and Kazyaka (1988) found regional raceof-defendant effects (GAO, 1990). As a result, the GAO concluded that while there was evidence to support a finding of a main race-of-victim effect, any race-of-defendant effect was equivocal and varied across a number of dimensions (GAO, 1990). The South Carolina study by Paternoste r and Kazyaka looked at South Carolina prosecutorial charging decisions begi nning in1977, when South Carolinas Furmanbased statute went into effect (Pat ernoster and Kazyaka (1988) Th e study was important in that it demonstrated an urban-rural effect at the prosecutorial level which had a racial component all defendants were at more risk of being charged with a crime in rural areas than in urban areas, Black defendants were at most risk, and thos e Black defendants who killed White victims were at the highest risk of all (Paternoster & Kazyaka, 1988). This intrusion of racial bias at the charging stag e, it was argued, had little possibility of being corrected at the sentencing or appellate stages if the defendant was found guilty. One of the shortcomings of the study for purposes of examining the effect of race on the sentencing decision, however, was that neit her defendant characteristics beyond race and gender, nor case-seriousness f actors were controlled for in the analysis. Since Furman applies only at the sentencing stage, and requires consider ation both of case seriousness


36 as defined by the aggravators and defenda nts individual situa tion for purposes of mitigation, this study is not as valuable for a ssessment of bias in the sentencing process alone. Gross and Mauros examination of racial di sparities in capital sentencing is more pertinent to the limited issue of whether Furman as implemented through Gregg/Proffitt/Jurek produces color-blind results. They looked at homicide data from eight states, most in the Sout h: Arkansas, Florida, Georgia, Illinois, Mississippi, North Carolina, Oklahoma, and Virginia, between 1976 and 1980. Like Paternoster and Kazyaka, they were looking at urban-rural differences in treatment of Black and White crime. Unlike the South Carolina study, howev er, Gross and Mauro attempted to control for crime seriousness as well, incorporating variables felony homicide, stranger murder, multiple victims, victim female gender, and the method of killing {shot versus bludgeoning, strangling, stabbi ng), which they use to scale the cases roughly by aggravation level. They also controlled for urban-rural, defendant and victim race, and interactive effects of race (Gross & Mauro, 1989). Their study showed that the ruralurban effect was strongest in the Southern states, particular ly Georgia and Florida; and a similar geographic bias occurred for race-of-v ictim effects. As noted earlier, no main race-of-defendant effect was shown except where the race of the victim was White. In 1988, Bienan et al. examined all homicide cases committed in New Jersey from August 1982 -when New Jerseys Furmancompliant sentencing statute went into effect -through the end of 1986, to asse ss which factors affected the sentencing outcome. The study was not limited to the penalty stage fe w penalty cases were used and its main focus was on the probability that a case pr ogressed from the char ging decision to capital


37 trial (Bienen et al., 1988). Unlike the two ear lier studies, this New Jersey study included not only case characteristic s related to aggravation and defendant and victim characteristics such as gender, age and r ace (Black, White, Hispanic) but also socioeconomic factors such as income class, empl oyment and relative physic al disability. The study produced what was characterized as substantial differen ces based on racial differences, across geographic jurisdictions, including a signif icant race-of-victim effect (Bienen et al., 1988). Here again, however, the primary stage in the process examined was the pre-sentencing stage, which may indicat e a possibly pervasive racial bias in the capital justice system as a whole, but cannot be used to assess the race neutrality of the Furman sentencing process itself. Post-1990 Studies The post-1990 studies also generally doc ument a race-of-victim effect on capital sentencing and less so a race-of-defendant effect (Baldus & Woodworth, 2003). In a 1991 study, Klein and Rolph found r ace-of-victim disparity in California capital penalty trials, with an unadjusted result showing a 9 percentage point race -of-victim disparity. The authors acknowledge that the effect show n was limited to the penalty stage and did not take into account earlier discretionary deci sion point where race could have an effect, such as the prosecutorial charging decisi on (Klein & Rolph, 1991). The subsequent CART analysis 3 including multiple case characteristics, showed race-of-victim to be a significant predictor of sentencing outcomes (Klein & Rolph, 1991). 3 Baldus and Woodworth have reserva tions about the use of the CART methodology (Baldus & Woodworth, 2003), and the advantage of a CART analysis over a logistic regression analysis was the subject of a recent study by Berk, Li and Hickman (2004) in which they re-analyzed the Maryland cap ital sentencing data used by Paternoster and Brame in their 2003 empirical study. Berk and his colleagues argue that the CART methodology produces more valid results than the conventional logistic regression model most often used in assessing racial effects and capita l sentencing (Berk,


38 In 1991, Radelet and Pierce published an an alysis of Florida homicides between 1976 and 1987, including 415 death cases. The unadjusted results showed a race-ofvictim disparity of nearly 13 percentage points, notwithstanding similar aggravating circumstances. A logistic regression an alysis which included pertinent case characteristics such as victim sex and the location of the crime still produced a significant race-of-victim effect, with White victim crimes being 3.42 times more likely to end in a death sentence (Radelet & Pierce, 1991). In 1995, two studies were publis hed, one from Connecticut ( State v. C obb, 1995) and one from Kentucky (Keil & Vito, 1995) The Connecticut study was small, examining 66 cases of capital murder gu ilt trials in 1973-1994. The study was submitted in a direct appeal of a death sentence as support for a defendant-comparative [as opposed to a case-comparative] proportionality review ( State v. Cobb, 1995). The unadjusted data indicated a higher capital murd er conviction rate for Black defendants than for White and lower conviction rate where the victim was Black. In a lengthy opinion, the Connecticut Supreme Court rejected the study as insufficient overall, and al so rejected the concept of defendant-comparative proportionality review ( State v. Cobb, 1995 ). The second study by Keil and Vito in Kentuc ky, examined death eligible cases in 197691, and assessed the combined effect of racial influences at the prosecutorial charging decision and jury sentencing decision in a logi stic regression analysis that controlled for case and crime characteristics. They found th at Kentucky capital justice system was the most punitive of Black defendants who killed White victims, whose odds of receiving the Li & Hickman, 2004). Logistic regressi on, however, remains the methodology most often used in racial studies of capital sentencing, and is the method used here.


39 death penalty increased by 1.41 over other defendant-victim combinations, and who suffered a similar increase in the odds of being charged with a capital crime in the first place (Keil & Vito, 1995). By 1999, the New Jersey Supreme Cour t became about the validity of the methodology it employed for comparative propo rtionality review which had been developed by Professor Baldus and his colleagues in 1988. They had been retained by the state of New Jersey to develop a state-of-the-art me thod of assessing proportionality in capital sentencing, and developed a comp lex, multivariate model using what came to be known as an index-of-outcomes approach. This was adopted by New Jersey in 1991 and applied initially in 1992 ( Marshall v. Loftin 1992). The predictive value of the Baldus model came under scrutiny by the Ne w Jersey Supreme Court beginning in 1996, when it initiated a review of the model, and lead in 1999 to the Proportionality Review Project ( In re Proportionality Review Project, 1980 ), under the supervis ion of a Special Master, Judge David S. Baime. The purpose of the study was to assess the continued viability and utility of the system of proportionality review developed for the New Jersey by Professor Baldus, a system which had proven unwieldy and unreliabl e (Baime, 2003). With the assistance of Professors Weisburd and Naus, Judge Ba ime examined the Baldus methodology for proportionality review and in his report to the Court in 2001 found the Baldus model to be unreliable because of the small sample of cases on which it was originally based a defect which Professor Baldus had noted hi mself and because of the unwieldy number of variables used in the logistic regre ssion analysis (Baime, 2001; Weisburd & Naus, 2001).


40 In the same report, the first of continuing annual reports monitoring proportionality review in New Jersey, Baim e and his colleagues recommended a three strategy analysis: bivariate analysis using a re vised salient factors selection, regression analysis, and case-sort ing techniques, looking at three decision points in the capital sentencing process penalty stage, life-death outcomes over all death eligible cases, and prosecutorial election to seek the death penalty. When the re sults of the three analyses converge, it can be safely assumed that the results are valid (Baime, 2003). The 2001 report utilizing these three me thods showed no direct statis tically significant race-ofdefendant or race-of-victim effect at any stage. The study did find a geographic effect, with counties with a higher rate of Black victims the more urban counties -advancing fewer cases to a penalty trial than less urban counties with more White victims (Baime, 2001; Baime, 2003). In 1998, Professor Baldus and his collea gues completed a comprehensive study of capital cases in Philadelphia, Pennsylvania. The most important influence on our [Baldus and colleagues] methodology in de veloping the Philadelphia study was the ongoing New Jersey project (Baldus, Woodworth, Zuckerman, Weiner & Broffitt, 1998). The Philadelphia study acknowledge s the two primary shortcomings of the original New Jersey work: 1) the small sample size on wh ich it relied and 2) the unwieldy number of variables used to try to captur e case characteristics (Baldus et al., 1998). The sample size issue was resolved in the Philadelphia study by the collection of da ta on 672 cases, 384 of which went to the penalty stage, including cases where the jury hung and an automatic life sentence resulted.


41 The number of factors used in the core analysis of the Philadelphia data was limited to on the number of statutory aggravators and mitigators as found by the jury, or present in the case if not found by the jury (Baldus et al. 1998). Other logistic regression models were run incorporating additional va riables such as socio-economic status and other case seriousness factors, and examining other decision points, such as the charging decision. At the aggravator-mitigator weighing stage they found a statistically significant race-of-defendant effect based on the total aggravators and mitigators accepted an odds multiplier of 29.0, significant at the .01 leve l, for Black defendants receiving the death penalty. This effect was reduced, but s till significant, when hung jury cases were included in the core model (Baldus et al ., 1998). Although more complex and evaluating many more dimensions of the sentencing deci sion, this study of Philadelphia cases has provided the approach for the study presented here. In 2000, Brock and his colleagues published an ex post facto analysis of charging and sentencing decisions in homicides in sp ecified jurisd ictions within Texas in 19801996. Using ratios they compared arrest to sentence results overall to selected jurisdictions which had the highest numb ers of both homicide arrests and death sentences, all urban counties. The initial ratio s analysis showed a rural-urban effect, with more homicides and fewer death sentences in all urban areas ex cept Houston/Harris County. They also conducted ratio analyses controlling for defendant and victim race legitimate case characteristics, using an additive scale of case seriousness from 0 to 4. A White-victim effect was found to be strongly associated with defendant culpability as measured by the aggravation sc ale (Brock et al., 2001).


42 During the same time frame, the United States Department of Justice, in 2000 and 2001, released reports presenting unadjusted race-of-victim a nd race-of-defendant effects in charging rates by federal prosecutors. Th ese reports present essentially descriptive analyses, but Baldus and Woodworth re-ana lyzed the data using logistic regression techniques and confirmed a statistically significant race-of-victim effect (Baldus & Woodworth, 2003). In 2001, Professors Unah and Boger rele ased their preliminary findings of an analysis of racial disparities in capital sentencing in North Carolina (Unah & Boger, 2001). Using Baldus-type multiple regression an alyses and logistic regression methods, they examined all potentia lly capital cases in North Carolina between 1993 and 1997 with the stated purpose of taking a systematic look for patterns of racial discrimination in capital sentencing in the South, employi ng data more recent than 1984 (Unah & Boger, 2001). Because they were looking at a ll stages of the capital justice process, included in their sample were all death potential cases, incl uding homicides that resulted in a charge of murder, first de gree murder or second degree murder, for the period from January 1, 1993 to December 31, 1997. The sample included 402 first degree hom icides which went to a penalty phase and resulted in a life or death sentence. A multistage sample technique was used to create an additional group of cases of second degree murder resulting in life or term of year sentences, for a total of 502 cases in the core dataset (Una h & Boger, 2001). The independent variables include d were not only the North Ca rolina statutory aggravating and enumerated mitigating factors, but also factors reflecting other case, crime and


43 demographic factors, for a total of thirty-six variables. The model did not include nonstatutory catch-all mitigating factors. They ran models using all cases in the sa mple, only cases that went to trial, all death eligible cases, cases where the pros ecutor sought the deat h penalty, and cases where there was a penalty trial. As the litera ture would predict, they found a statistically significant race-of-victim effect those mu rdering White victims were 3.4 times more likely to receive the death penalty. They also found notable but not statistically significant -differences in the death sent encing rate for Black defendants murdering White victims, versus White defendants at 6.4% versus to 2.6% a nd overall that Blackon-Black homicides resulted in the lowest death sentenci ng rate of any combination (Unah & Boger, 2001). Also in 2001, the Joint Le gislative Audit and Review Commission of Virginia released an analysis of the charging decisions in a sub-sample selected from a larger sample of death eligible cases to reflect various geographic regions in Virginia. The unadjusted data reflected a race-of-victim e ffect at the charging stage, but no race-ofdefendant effect. In a logistic regression model of charging decisions which controlled for case characteristics reflecting crime seri ousness, no statistically significant main racial effect was observed, a lthough an increased effect of two case characteristics defendant-victim relationship and crime jurisdiction was seen when the victim was White (Joint Legislative Audit and Review Committee, 2001). Five studies were published in 2002, fr om Arizona (Bortner & Hall, 2002), Illinois (Pierce & Radelet, 2002), Indian a (Ziemba-Davis & Myers, 2002), South Carolina (MCord, 2002), and Nebraska (Baldus Woodworth, Grosso, & Christ, 2000).


44 Baldus and his colleagues looke d at the rates of death sent encing in 189 death eligible prosecutions in Nebraska, only 89 of which ha d progressed to the sentencing stage. The logical regression analysis of these cases produced no racial eff ect, neither race-ofdefendant nor race-of-victim (Baldus, Woodwor th, Grosso, & Christ, 2000). The original Arizona study did not analyze for race eff ects (Bortner & Hall, 2002). Baldus and Woodworth performed a separate analysis of the data, taking into account crime death worthiness to the extent allowed by the data, and found a race-of-vict im effect (Baldus & Woodworth, 2003). Unadjusted race-of-victim effects were also found in the Indiana study of 224 murder convictions. Only a rates analysis was performed, although a multivariate regression analysis is proposed for the future (Ziemba-Davis & Myers, 2002). Pierce and Radelet looked at over 4000 mu rder convictions in Illinois between 1988 and 1997, during which time only 76 death sentences were imposed. They looked at unadjusted data, and performed a logist ic regression analysis controlling for 28 variables in addition to race. No statistical ly significant race-of-defendant effect was seen, there was a significant race-of-victim ef fect and a geographic effect seen in the models (Pierce & Radelet, 2002). A limited analysis of South Ca rolina sentencing rates in eleven homicide cases in 1998, selected by McCord for their extreme depravity (McCord, 2002) also produced no race-of-d efendant effect, but demonstrated a substantial race-of-victim effect. The most recent studies, in 2003 and 2004 continue the pattern. In 2003, Raymond Paternoster and Robert Brame publis hed an analysis of 1,311 Maryland death cases death noticed and tried between Ju ly1, 978 and December 31, 1999, including 180


45 cases which progressed to penalty trial (Paternoster & Brame, 2003). The unadjusted results showed race-of-victim disparities wh ich reached significance, but no significant racial differences in senten cing, nor significant geographic di sparities in sentencing. The logistic regression results cont rolling for relevant case characteristics, however, showed no significant race effects at the sentencing stage (Paterno ster & Brame, 2003), although a race-of-victim and geographic effect were significant at the char ging and death-noticing stages. As previously noted, in 2004, Ber k, Li and Hartman re-analyzed the Maryland data used by Paternoster & Brame using a CART analysis instead of logistic regression, and maintain that it produces more reliable re sults in this case, no racial effects at all (Berk, Li, & Hartman, 2004). Finally, Baldus and Woodworth have ci ted a third, smaller but instructive analysis of race in capital sentencing by Pr ofessor James McAdams, a supporter of the death penalty (Baldus & Woodworth, 2003). Mc Adams concludes that, despite increased sophistication in the analyses performed to assess racial disparity in capital sentencing, the results remain the same: 1) race-of-defendant is not, in and of itself, a statistically significant main effect in death sentencing; 2) race-of-victim is a significant factor in death sentencing, with all defendants being more likely to receive the death penalty if their victim is White; 3) Black defendants who kill White victims have the highest rate of death sentence; and 4) Blacks who kill Black victims have the lowest rate of death sentence (McAdams, 1998). Research Problem All of the studies noted above focused on whether there are main racial effects showing racial disparity in capital sentencing outcomes, and most looked at sentence


46 outcome as the product of racial bias in the entire capital justice process. None examined the sentencing decision solely as a product of application of the Furman/Gregg standards. This may be because of an unw illingness to abandon the argument lost in McCleskey and a continued desire to prove a denial of equal protection and an inference of purposeful discrimination in the form of the disparate impact of the entire capital justice system on Black citizens (Baldus 2004; Howe, 2004) There may be no statistically significant disparity in the numbe r or distribution of d eath sentences between the races sufficient to demonstrate purpos eful discrimination in violation of the Fourteenth Amendment equal protection clause. Even so, significant race-based disparity in how the Furman sentencing standards the aggravating and mitigating circumstances are applied may be indicat ive of unconscious racial bias which offends other constitutional principles. This nonpurposeful form of bias may be incapable of cure through legal process yet still skew the sentencing process against Black citizens in a way repugnant to other constitutional principles requi ring jury impartiality, predictab ility in death sentencing and protection from irrational sentence disparity. Identical aggravators may be applied more harshly to one race or the other, or vary disproportionately based on the race of the victim. Similarly, the mitigators may work better in reducing the chances of a death sentence based on a defendants or the victims race they might be more mitigatory for White defendants than Black, despite comparability of the crime. Earlier studies were not designed to look at latent, nonpurposeful racial discrimination in the application of the Furman standards, but that is not to say that the methodology cannot be used for such research. The studies fall generally into three


47 approaches: 1) the use of unadjusted and case characteristic-adjusted data in rates analyses and logistic regression models per Baldus, Paternoster and others; 2) the use of CART methodology to analyze case data proposed by Berk et al. (2004) and used earlier by Klein and Rolph (1991), and 3) the thre e stage combination of methods used by Baime, Weisburd and Naus in New Jersey. The Baldus-type model is generally c onsidered the most reliable and costeffective, although the number of variables, the size of the case sample and the decision point modeled will greatly affect the outcome (Unah & Boger, 2001; Baldus & Woodworth, 2003). It is also the most flexible in turns of ease of use, and the ability to use available information and produce meaningf ul results. For that reason, it has been selected as the best methodology to assess r acial invariance in th e application of the North Carolina Furmanbased standards, as proposed here. Logistic regression analysis of racial disp arities in the facially neutral application of Furmanbased standards is a logical way to test for latent racism in jury decisionmaking under the Furman/Gregg design (Baldus & Woodwort h, 2004). The Philadelphia study suggests that the total number of aggravating circumstances as measure of crime seriousness and defendant culpability -and to a lesser degree, mitigating circumstances which a capital jury accepts may be the best predictors of capital sentence outcome (Baldus & Woodworth, 2004; Baldus et al. 1998). If there ar e statistically significant differences in the effect of jury acceptance and application of the same number of aggravating and/or mitigating circumstan ces on sentencing outcomes among similarly situated Black and White defendants, it arguably reflect s unconscious racial attitudes of jurors.


48 To assess whether the Furman capital sentencing model is working in a raceneutral way, it is appropriat e to look only at the sentenci ng stage; that is where the Furman/ Gregg factors operate. The question of race infecting other stages of the capital justice process is important overall to th e question of racial neutrality in the administration of capital justice, but irrelevant to the question of wh ether capital juries are color-blind. Indeed, studies have shown the largest racial bias effect appears at the prosecutorial charging phase, and that it can be a function of politics and geography, as well as race (Bienen, Weiner, Denno, Alison & Mills, 1988). As discussed earlier, other stud ies have demonstrated extraFurman influences, such as geography and socioeconomic differences, which may be important in explaining racial variance in capital sentencing results but may be irrelevant to whether the Furman standards are producing constitutionally accepta ble results. The same can be said for studies examining the jury panel itself, its racial composition and the lack of understanding of its role in the sentenci ng process (Bowers, Sandys & Steiner, 1998; Bowers, Steiner, & Sandys, 2001; Bowers & F oglia, 2003). While informative as to the dynamics of the jury decision-making process, these do not address the core question of whether the Furman model is working in a ra cially fair and equal way These studies are relevant to the questi on of juror unconscious predispositions to impose the death penalty, overall and based on race their own, the defendants or the victims -but are too limited to allow conclusi ons as to racial invariance in application of the Furman standards universally or under a specifi c states standar d. A meaningful test of the Furman model would encompass a long time frame postFurman, all or most of the capital cases which reached the pena lty stag during that time, under a specific


49 statutory scheme compliant with the Furman standards implemented through Gregg/Jurek/Proffitt statutory schemes. That is the st udy presented here, of most, if not all, of the capital penalty decisions made by juries under North Carolinas capital sentencing statute over a period exte nding of more than ten years. Hypotheses The starting point should be the bare application of Furman to the facts and the defendant, by the jury, at the sentencing stage using a Furman-compliant state capital statute, in this instance that of North Caro lina. The goal is to determine whether the North Carolina statute works to channel capita l jury discretion in imposing the death sentence in racially invariant and thus constitutionally acceptable ways, as Furman and Gregg intended The hypotheses to be tested are: 1. Each unit increase in the level of a ggravation a capital ju ry finds increases a defendants odds of receiving a death sentence. 2. Each unit increase in the level of m itigation a capital jury accepts decreases a defendants odds of receiving a death sentence. 3. Jury acceptance of aggravating circumstances increases the likelihood of a death sentence more for Black defendants th an similarly situated White defendants. 4. Jury acceptance of mitigating circ umstances decreases the likelihood of a death sentence more for White defendants th an similarly situated Black defendants. 5. Jury acceptance of aggravating ci rcumstances has a stronger effect on defendants whose victims are White than those whose victims are Black, but whose crimes are comparable, in increasing the odds of receiving a death sentence.


50 6. Jury acceptance of mitigating circumstances has a stronger effect on defendants whose victims are Black than those whose victims are White, but whose crimes are comparable, in decreasing the odds of receiving a death sentence. 7. Jury acceptance of aggravating circumstances has the strongest effect on increasing the odds of receiving a death senten ce where the defendant is Black and his victim is White. 8. Jury acceptance of mitigating circum stances has the strongest effect on decreasing the odds of receiving a death senten ce where the defendant is White and his victim is Black.


51 Chapter Four Methodology Data The analysis is based on information orig inally compiled from reviews of capital murder trials, including penalty phase pr oceedings, in North Carolina between 1978 and 2002. LexisNexis searches of North Carolina Suprem e Court and Court of Appeals cases allowed identification of cases in which the defendants were convicted of or pled guilty to first-degree murder, the state sought th e death penalty, the tr ial progressed to a sentencing phase whereby the jury heard evidence concerning aggravating and mitigating factors, and the jury issued a binding recommendation for a sentence. There are 1073 cases in the original data set from trials he ld during the period 1978-2003. North Carolina sentencing procedures with regard to jury consideration of mitigating factors changed in 1990, as a result of the holding in McKoy v. North Carolina 449 U.S. 433 (1990). Prior to McKoy the North Carolina sentencing statute disallowed jury consideration of mitigating factors unless the jury had accepted them as a whole. McKoy clarified the role of mitigators, holding that any individual juror could rely on any relevant mitigating circumstance when making the sentencing decision. To eliminate the possible effect of this change in how mitigators are used by North Carolina juries, cases using preMcKoy procedures as to mitigation and decided before May 1, 1990 were deleted. The May date was selected arbitrarily, being about 90 days from the


52 issuance of the McKoy holding, as sufficient time for the McKoy principles to have been adequately implemented by North Carolinas trial courts. The original data reflected the race-of-def endant and victim in three categories: White, Black and other. In order to focu s on the defined Black-White race distinction, those cases where either defendant or vic tim race was coded as other were deleted, leaving a total of 632 cases involving Black or White victims and Black or White defendants for use in the analysis. These cases include 374 Black defendants and 258 White defendants, and 244 Black victim a nd 388 White victim cases. The breakdown of defendant-victim racial composition is White defendant/White victim = 232 cases; White defendant/Black victim = 26 cases; Black defe ndant/Black victim = 218 cases, and Black defendant/White victim = 156. Most of the ca ses (71%, n=450) are intra-racial, fairly evenly distributed between Black-on-Black ca ses [218] and White-on-White cases [232]. Of the remaining 182 cases which are inter-r acial, the small number of White-on-Black cases [26] confounds any analysis of that se t of cases or comparison of inter-racial homicides. Because the focus of the analysis is jury application of the aggravating and mitigating factors in the sentencing decisi on, the dataset includes all cases where the sentencing phase was conducted, including hung ju ry cases where the jury declared that they could not reach the required unanimous decision regarding a sentence, resulting in the default sentence of life in prison; and retr ials, where two separate juries considered the facts independently and recommended a sent ence in the same case. The inclusion of hung jury cases has been found to decrease the effects of aggravators because a hung jury always results in a life sentence (Baldus et al., 1998); however, because that sub-sample


53 of cases represents application of the North Carolina sentencing statute, those cases have been kept in the sample used fo r the analysis. As to retrial cases, inclusion of these cases increases sample size of instances where separate juries applied accepted aggravators and mitigators to reach the life-death decision, albe it to the same case, they are thus relevant for purposes of examining racial invariance in this process. The units of analysis are individual capital cases from the North Carolina Capital Sentencing Project (NCCSP) database, limited to those cases which reached the penalty phase and in which aggravators and mitigators were considered by the jury in penalty trials in North Carolina between 1990 and 2003. The data was collected by M. D. Smith (USF), Beth Bjerregaard (UNC-Charlotte) and Sondra J. Fogel (USF), for the North Carolina Capital Sentencing Project. Because each homicide victim constitutes a separate capital offense requiring separate jury findings and sentencing, each victim case was coded separately, although the defendant was the same. The data collection instrument [DCI] used was a Baldus-type instrument designed to capture defendant, victim, jury and case char acteristics, incorpora ting over 125 items. The sources of the data were official state records, primarily thos e of the North Carolina Supreme Court and the North Carolina Courts of Appeal as maintained by the North Carolina Administrative Office of the Courts (AOC), trial courts in the counties, and other state criminal justice related agencies. To the extent possible, missing information was supplemented by media reports and individual interviews. Materials collected from state appellate re cords include defendant and state briefs, as well as the jury Issues and Recommenda tion Form which records jury acceptance or non-acceptance of aggravating and mitigating factors, and concludes with the jurys


54 sentencing recommendation. Historically, thes e materials have been published in hardcopy form and placed in two university law libraries in North Carolina, while other locations have microfilm copies. Beginning with decisions returned from cases tried in 1999, hard copies have not been made availa ble, but materials are accessible via an electronic data file ( ). This information was supplemented with newspaper accounts of th e trial where such coverage was available through LexisNexis or Newsbank another electronic databank th at includes varying years of stories from eight Nort h Carolina newspapers. Defendant and crime information was obtained online from the North Carolina Department of Corrections at ( ). Through 1996, victims age, race, and sex were take n from a commercially available CD-ROM, North Carolina Vita Records: Deaths 1968-1996 For 1997-2002, victim personal information was determined from some combination of cour t material (such as reference to the victim in the states or defendants appeals briefs), newspaper accounts, or obituaries obtained through World Wide Web search engines. Case s for which this information could not be obtained are not included in the dataset. The lack of a centralized source of inform ation regarding capital murder trials in North Carolina makes it difficult to determine the exact number of capital murder trials conducted during the period covered in the data Since appeals of death sentences are automatically referred to the state Supreme C ourt, all of those cases are included. A large proportion of defendants receiving a life sent ence appeal their first-degree murder convictions to the Courts of Appeal, so mo st of the life sentence cases included in the database were identified in the lower appellat e court records, as well as in those cases


55 subsequently appealed to the North Carolina Supreme Court, alt hough that court has the option of declining to hear the case. There are two instances where defendants are unlikely to appe al, and therefore their cases may not be included in the datase t. First, if a defendant pled guilty and received a life sentence, there was little basis for appeal. Second, some defendants convictions were upheld but their death sent ences vacated. If, upon retrial of the penalty phase only, the sentence was life, there was no basis for appeal and no means to identify these cases other than a count y-by-county search. However, in the former situation no penalty phase jury decision was involved, so such cases are irrelevant to the analysis. As to the latter situation, while these retrial cases are relevant to the analysis of the jury decision-making process, the number is likely to be small, based on the relatively small number [76] of guilt-and-penalty phase retrials, which are included in the database. A much smaller set of trials not included in the dataset are those that were identified, but whose case materials are not available because hard copies were missing from both libraries or not yet posted in elec tronic form. Given that the substantial majority of capital cases are appealed to at least one of these courts, we estimate that the available data contain reviews of more th an 90% of all senten cing recommendations made by North Carolina capital juries during the ten plus y ear period. This provides an excellent sample to test racial e ffects on the sentencing decision under the Furman model. Some data is missing from some cases. If the jury recommendation form is missing, the specific aggravators and mitigators used cannot be determined. Some cases were default life cases where either no aggravator was found or the aggravator was


56 insufficient to recommend death. However, fo r the variables used in the instant study total and individual aggravators and mitigators accepted only 7 cases were incomplete from the 632 cases available. Comparisons of these missing data cases with those remaining in the dataset revealed an overre presentation of life sentence cases, suggesting that the reduced dataset overstates the proportion of death sentence cases; however, comparisons of major demographic and le gal variables between life sentence cases included and not included in th e working dataset revealed no major sources of bias in the cases used for the analysis. Variables For purposes of the analysis, the dependent variable is the defendants sentence, a dichotomous variable coded 0 = life, 1 = d eath sentence. There are 302 life sentence cases and 330 death sentence cases for the tota l 632 cases. There are four (4) classes of independent variables: the two target variab les of defendant and victim race and the two Furman factors -the statutory aggr avating circumstances (a ggravators) and statutory and nonstatutory mitigating circumstances (m itigators) considered by capital juries at the sentencing phase. These factors are analy zed both as ranges of total factors accepted and as individual factors (Baldus et al. 1998, Baldus & Woodworth, 2004). The total number of aggravating and mitigating circumstances submitted to, rejected and/or accepted by each jury was c oded for each case. Because the number of mitigators which can be considered in a given case is limited only by relevance, as opposed to the statutorily limited number of aggravators which can be considered, the difference between the number of aggravators and mitigators considered can be very large. In order to achieve an approximati on of relative low, medi um and high levels of


57 jury-determined aggravation and mitigation, ra nges of total aggravators and mitigators accepted were used for comparison purposes (Baldus et al. 1998). Variable Recoding Table 1 sets out the vari ables and their coding and th e case frequencies. All independent variables have been dichotomized, with 1 representing the target category. The original race-of-defendant and race-of-v ictim variables, after deleting the cases where the race designation was o ther, were recoded from the original three categories of White=1, Black=2 and other=3, to White =0 and Black=1, in order to compare the effect of a defendant or victim being Black on the likelihood of receiving the death penalty under similar circumstances, to not being Black. As noted above, numeric counts for th e total number of aggravators and mitigators accepted in each case were re-organized into ranges of total factors accepted which could be compared. The ranges were chosen for rough approximation of levels of aggravation and mitigation, from lowest represented by 1 and 2, mid-range by 3, and 4 and 5 representing the higher levels of case seriousness/defendant mitigation the jury accepted (Baldus et. al, 1998). This also was in tended to achieve some standardization in comparative counts because of the very broad range of mitigators [0 to 50] considered compared to the much smaller range of aggravators accepted [0-50]. Counts of aggravating circumstances were recoded from actual counts to 0 aggravators accepted = 1, 1 2 aggravators accepted = 2, 3 4 a ggravators accepted = 3, 5 9 aggravators accepted = 4. Counts of mitigators accepte d, including both statut ory and non-statutory, were coded 0 mitigators accepted = 1, 1-2 mitigators accepted = 2, 3 5 mitigators accepted = 3, 6 10 mitigators accepted = 4, 11 50 mitigators accepted = 5. The ranges


Table 1: Variable Frequencies and Coding. All White Black D White V Black WD/ WD/ WV BD/ BD/ BV WV BV Cases D V N 632 258 374 388 244 26 232 156 218 Sentence (0)Life 302 118 184 180 122 19 99 81 103 (1)Death 330 140 190 208 122 7 133 75 115 Total Aggravators Accepted (1)=0 29 13 16 18 11 0 13 5 11 (2)=1-2 380 168 212 220 160 24 144 76 136 (3)=3-4 176 74 102 120 56 1 73 47 55 (4)=5-9 28 2 26 16 12 0 2 14 12 Total Mitigators Accepted (1)=0 23 11 12 17 6 0 11 6 6 (2)=1-2 42 19 24 32 11 0 19 13 11 (3)=3-5 95 37 58 55 40 1 36 19 39 (4)=6-10 174 68 106 99 75 10 58 41 65 (5)=11-50 289 123 166 180 109 15 108 72 94 Aggravators (0=not submitted/ submitted not accepted/missing, 1=accepted) Felony murder 241 99 142 168 79 5 94 68 74 Murder Pecuniary Gain 135 52 83 100 35 0 52 48 35 HAC 203 97 106 135 68 5 92 43 63 Viol. Conduct 273 105 168 158 115 11 94 64 104 >1 Person Risk 29 11 18 15 14 0 11 4 14 Prior Viol. Fely 179 61 118 104 75 3 58 46 72 Murder In Jail 5 3 2 2 3 1 2 0 2 Prior Capitol Off. 18 6 12 10 8 1 5 5 7 Escaping 75 33 42 50 25 3 30 20 22 Kill Law Enf.Officer 5 2 3 4 0 2 2 1 Kill Govt Officer 21 4 17 3 18 0 4 14 3 Mitigators (0=not submit/submit not accept/individual accept not required/ not accept/missing, 1=accepted) Insig.Prior Recd 217 104 113 138 79 15 89 49 64 Men/Em.Distress 299 144 155 180 119 15 129 51 104 D Age 78 26 52 45 33 4 22 23 29 D Capacity 207 102 105 118 89 18 84 34 89 D Minor Accomplice 25 7 18 16 9 0 7 9 9 D Under Duress 48 26 22 30 18 6 20 10 12 D Aid Prosecution 51 20 31 37 14 3 17 20 11 Vic. Consent 8 2 6 1 7 1 1 0 6 D Drink 120 80 40 77 43 13 67 10 30 D Drugs 137 64 73 86 51 10 54 32 41 D Physic. Abuse 103 53 50 64 39 6 47 17 33 D Sexual Abuse 36 17 19 19 17 0 17 2 17 D Broken Home 73 35 38 39 34 7 28 11 27 D Dad Gone 118 32 86 62 56 4 28 34 52 D Mom Gone 51 23 29 35 17 0 23 12 17 Foster Home 20 13 7 17 3 0 13 4 3 Parent Misconduct 127 66 61 80 47 5 61 19 42 D IQ 67 25 42 35 32 2 23 12 30 58 DSpecificMent.Cond. 101 52 48 67 35 7 45 21 28


59 were chosen for rough approximation of levels of aggravation and mitigation, from lowest represented by 1 and 2, mid-range by 3, and 4 and 5 representi ng the higher levels of case seriousness/defendant mitigation the jury accepted (Baldus et al., 1998). As for each individual aggravating and mitigating circumstance accepted, the coding instrument provided f our (4) categories for aggravating circumstances: 1=not submitted; 1=submitted but not accepted; 3=accepted and 9=missing/not found. These were collapsed into two categories: 0 = not accepted [including not submitted, submitted but not accepted, and missing cases] as the default category and 1 = accepted, as the effect being tested. The NC statutory aggravators are fel ony murder, murder for pecuniary gain, heinous/atrocious/cruel murder murder during violent condu ct, murder resulting from conduct threatening more than one pers on, prior violent felo ny, murder while incarcerated, prior capital offense [adult a nd certain juvenile], murder of a law enforcement officer, and murder of a court or governmental official in course of their duties (Section. 15A-2000(e), NC General Statutes2004). In the specifically enumerated mitigators in the North Carolina sentencing statute are defendant has no significant prior record, defendant acted under emotional or mental distress, defendants age, defendant lacked capacity to understand his act, defendant was a minor accomplice, defendant acted under dures s, defendant aided the prosecution, and the victim consented to the conduct result ing in his death (Section. 15A-2000(e), NC General Statutes, 2004). The non-statutory mitigators allowed under the catch-all provision at Section 15A-2000(3)(9) and recorded by the DCI are defendants alcohol abus e, defendants drug


60 abuse, defendant was physically abused as a child, the defendant was sexually abused as a child, defendant came from a broken home, defendants father was absent from the home, defendants mother was absent from the home, defendant was in foster care, defendant witnessed parental misconduct as a child, defendants low IQ, and defendant suffers from a specific mental disease. There were six (6) categories of possibl e responses in the coding instrument applicable to both the statutory and nonstatutory mitigators: 1=not submitted; 2=submitted but not accepted; 3=accepted; 4=acceptance of individual circumstances not required of jury; 5=aggravators not accepted; and 8=aggravating circumstances ruled by jury as not sufficient to justify death penalty. Since the analysis examines only the effect of jury accepted aggravating and mitigating circumstances on life-death outcomes, these variables were collapsed like the aggravators, and recoded to reflect acceptance as the targeted effect -0=not accepted, encompa ssing all responses except accepted, and 1=accepted. Statistical Analysis Logistic regression analysis was employed, using SPSS to assess 1) the effect, if any, of jury acceptance of aggravators or mitig ators -both as to total factors accepted and individual factor effects -on sentencing outcomes, and 2) whether there is racial invariance of these effects. The core analysis controlled for the effect aggravators and mitigators on sentencing outcomes by ranges of totals accepted, and defendant and victim race and is reported at Tables 1 and 2. The second set of models takes into account the possibility that some factors might have a greater effect on jury decision-making than other factors also accepted in a case, and looks at the individu al effect of each aggravator


61 and mitigator on sentencing outcomes in an effort to sort out possible dilution of effect from the initial grouping of f actors in the first model set The logistic regression analysis for both th e totaland individu al factor accepted included eight models: all cases, Black de fendant cases, White defendant cases, Black victim cases, White victim cases, and defendant-victim race combinations [WhiteD/WhiteV; BlackD/BlackV; BlackD/White V]. If a variable in one model showed a statistically significant effect, the z-test fo r the statistical significance of differences in effects across other models was used to de termine if the difference between the two models, and the subset of cases they repr esented, was actually si gnificant (Paternoster, Brame, Mazerole & Piquero, 1998). There were two few cases of White defendant-Black victim cases [26] to analyze. Logistic regression was employed because the dependent variable, life death sentencing outcomes, is dichotomous and linear regression was not a ppropriate (Baldus et al. 1998). Unlike linear regression, logistic regression does not produce coefficients which predict actual values; logistic regr ession estimates the effect an independent variable has on the likelihood the odds a particular outcome will occur, in this case, a death sentence. The regression coefficient (B ) indicates the direction and strength of a variables effect on the likelihood of rece iving a death penalty; the odds ratio, as expressed by the exp (B), indicates the size of the change in the odds of a death sentence a given variable has. Overall model fit wa s measured by the Wald Chi-square statistic, with variance in sentencing outcomes explai ned by the model indicated in the Pseudo-R 2 statistic. The unstandardized b coefficient statis tical significance for each test variable is reflected in the individual Wald statistic. Stat istical significance of an effect is measured


62 at alpha level .05 [p < .05], meaning that there is less than a 5% chan ce that the variable has no effect on the odds of the outcome o ccurring in the population given the observed effect for the sample analyzed. For purposes of the final analysis and discussion, only results at alpha levels less than .05 are discussed.


63 Chapter Five Results Models of Ranges of Total Aggravating and Mitigating Circumstances Accepted Table 2 sets out the results of the first five of the eight models run using the ranges of total aggravators accepted and total mitigators accepted, selected to approximate low, medium and high aggrava tion and mitigation levels for comparison purposes. These models examined all cases, only Black defendant cases, only White defendant cases, only Black victim and only Wh ite victim cases. Ta ble 2 represents the results of the defendant-victim race combinati on models. In each of the models, legally relevant factors the aggravating and mitigating circumstances are statistically significant at p < .05, and the effect is in the expect ed direction. All-cases model. The all-cases model shows each that each one unit increase in the range of aggravating circumstances accepted increases the odds of any defendant receiving a death sentence by a multiple of 5.4. Each increas e in the range of mitigators accepted decreases the odds of a death sentence by a multiple of .64, or 64%. Across all cases, however, neither defendant nor victim race we re significant predictors of a capital sentence. There was an unexpect ed negative direction in these effects, indicating a slight decrease in the likelihood of receiving a death sentence based on race of defendant or victim, but as noted, neither was statistically significant. The predominance of intra-


64 racial cases [71% of all cases] may, in part, account for this result. The adjusted R 2 statistic of .407 for the all-cases model indicat es substantial variation left unaccounted for by this model. Some of the remaining varia tion is likely due to the inability to measure differences in the weights a juror might afford different aggravators or mitigators, the weighing of aggravating and mitigating ci rcumstances, and the balancing of their respective net weights. It is also possible th at legally irrelevant factors omitted from the model may contribute to the unexplained variation. Nevertheless, looking only at the all-cases model, the North Carolina capital sentencing process appears at first glance to be operating as intende d -death sentences appear to be largely based on legally releva nt aggravating and mitigating circumstances and not on irrelevant or extra-legal factors su ch as defendant or victim race. However, the subsequent models demonstrate that the effects of the level of jury-accepted of aggravating and mitigating circumstances are not racially invariant, and that the dissimilar effects of the level of aggravat ors accepted produce sta tistically significant differences in the odds of receiving a death sentence based on defendant and/or victim race. Defendant race-specific models. The results of the defendant race-specific models, set out in Table 2, demonstrate an increased effect from jury acceptance of aggravat ors on sentencing for Black defendants the aggravators are substantiall y more aggravating for Black defendants, with the odds of receiving a death sentence 2.7 times more than similarly situated White defendants. The odds of a Black defendant receiving the death penalty increase by a multiple of 7.7 with each unit increase in the level of aggravation, compared to an


Table 2. Logistic Regression Results (Odds Ratios): All-cases and Defe ndant and Victim Race-Specific Re sults Showing Effect of Total Jury-Accepted Aggravating and Mitigating Circumstances on Capital Sentencing Outcomes, North Carolina (May, 1990 December, 2002). Variables All Cases White Defendants z-scores Black White Defendants Victims zscores Black Victims Exp (b) B Exp (b) B S.E. Exp(b) B Exp B S.E. Exp(b) B S.E. S.E. S.E. (b) Defendant Race -.348 .232 .706 NA NA NA NA NA NA .562 .267 .570* z=.31 .386 .506 1.472 (DRACENEW) Victim Race -.145 .232 .865 .731 .503 2.078 -.443 .276 .642 NA NA NA NA NA NA (VRACENEW) Total Aggravators 1.679 .198 5.362* 1.057 .304 2.878* z=2.4* 2.04 .272 7.693 1.56 2 .242 4.768* z=.73 1.877 .348 6.535* (ATOTNEW) Total Mitigators -1.018 .115 .361* -1.166 .189 .392* z=.95 -.937 .152 .392* 1.05 4 .145 .348* z=.52 -.929 .193 65 .395* (MTOTNEW) Constant .811 2.069 -.569 1.17 3 -.650 R 2 = .407 .391 .445 .421 .397 *p <. 05


increase of only 2.9 for White defendants. This is a statistically si gnificant difference [z = 2.4, p=.0164], demonstrating considerable vari ance in the effect of jury-accepted aggravating circumstances on sentencing outcome across defendants race. Some of this variance may be attributed to the weights afforded individual aggravators and the collective weighing process by the jury, but there is no data as to these factors. Even so, it appears that although for each unit shift in aggravation level the odds of a death sentence increase for all defendants, Black defendant may pay a higher premium for the level of aggravation found. As to the effect of mitigating circumstances, the slightly less mitigatory effect of each unit increase in the range of mitigators accepted for a Black defendant [61%], over the odds reduction from mitigation for a White defendant [69%] is essentially racially invariant; there is no statistically significant difference [z =.95, p=.177]. In these racespecific models, Black defendants benefit fr om nearly the same discount enjoyed by White defendants for each increase in the level of mitigation accepted. Moreover, since the number of mitigators which may be considered is limitless, the mitigatory effect of jury acceptance of mitigators is diluted w ith each additional mitigator accepted. Jury acceptance of aggravators has a statistically significant different effect from acceptance of mitigators [z=2.8, p=.0026]. This may be due to a dilution effect based on the relatively small number of aggravating circ umstances [11] available to North Carolina juries versus the unlimited number of mitigat ing circumstances which can be considered, but may also reflect the greater weight an aggravator which must be found unanimously and beyond a reasonable doubt -carries for a ju ry. In any case, it appears that jury 66


acceptance of aggravating circumstances driv es the capital decision-making process, and Black defendants are treated more harshly once aggravating circumstances are accepted. As in the all-cases model, the effect of vic tim race fails to attain statistical significance regardless of defendant race. Modeling total aggravating and mitigating circumstances is also more predictive for Black defendants than for White defendants. The R 2 values increased in the Black defendant case model to .445, compared to the .407 in the all-cases model. The White defendant model was slightly less efficien t than the all-cases model, with an R 2 of .391. This difference in predictive value -nearly 45% for Black defendants compared to 39% for White defendants -is another indicati on that jury acceptance of aggravating circumstances has an enhanced aggravati ng effect on sentencing outcomes for Black defendants over White. There is also a difference in the inte rcept values between the defendant racespecific models which should be noted. The constant in the White defendant model is positive; it is negative in the Black defendant model. This indicates that omitted variable bias and other sources of systematic error in these models enhance the odds of a capital sentence for White defendants, but decrease them for Black defendants. Here again it is likely that the absence of the weights applied to each factor and the balancing of accepted aggravating and mitigating circumstances account for a considerable portion of this modeling error. However, this may also be evidence of less arbi trariness operating in Black defendant cases than in White defendant cases -a lack of ar bitrariness which may work to the detriment of Black defendants in terms of consistently harsher treatment. 67


Victim race-specific models. The last two sets of models reported in Table 2 sets out the effects of different ranges of aggravators and mitigators accepted on capital sentencing outcomes in victim race-specific cases. Here again the effects of jury acceptance of additional aggravators and mitigators are statistically significant and in the expected directions for both Black and White victim cases, but the differences in the effects across models are not statistically significant and thus not demonstrat ive of race-of-victim variation. It should also be noted that the aggravator effect is stronger for Black victim cases [multiplier of 6.5] than for White victim cases [multiplie r of 4.7], an unexpected result, but the difference is not statistically significant [z = .73 (p=.4645) ]. This difference may also reflect the skewed distribution of intraversus inter-racial cases in the sample, in which there are more Black-on-Black cases [n= 218] than Black-on-White cases [n=156]. As to the effects of mitigator accepta nce by race of victim, the odds of a defendant whose victim was White are redu ced by 65% for each unit increase in the range of mitigators the jury accepts, compar ed to a reduction of 60% for those whose victims were Black, indicating an unexpected stronger mitigatory effect where the victim was White. This may be because the ma jority of White victim cases had White defendants [232/388], and thus this may reflect the overall advantage apparently enjoyed by White defendants. In any case, the differences in mitigatory effects between White and Black victim cases is not statistically significant [z = .52 (p=.6031) ]. The race-ofdefendant effect is also significant for White victim cases, but not for Black victim cases, but in an unexpected negative direction. Th e effect is small, decreasing the odds of a 68


defendant whose victim is White receiving the death penalty by .43, and the difference in the effect of race-of-defenda nt is not significant between Black and White victim cases [z=.73 (p=.4654) ]. As noted earlier, the negativ e constant for Black victim cases indicates that omitted variables and systematic errors decrease the odds of a death sentence when the victim is Black, and slightly increase the odds when the victim is White. Although the Black victim model is a sl ightly better fit, the White and Black victim models perform comparably in pred icting sentencing outcomes, with comparative R 2 values of .421 and .397, respectively and ne ither is as predictive as the Black defendant only case model. All defendant-victim racial combination models. The lack of a racially different effect fr om jury acceptance of aggravating factors on capital sentencing outcomes is more clearly seen in the defendant-victim racial combination models set out in Table 3. Anal yses for White defendant-Black victim cases could not be undertaken because there were t oo few cases [n=26], so it is not possible to compare inter-racial cases. However, the resu lts of the other three racial combinations show substantial racial varian ce in aggravation effects. As to intra-racial cases, the odds of a Black defendant who kills a Black victim receiving a death sentence is increased 6.4 tim es for each unit increase in the range of aggravators, compared to an increase by 2.8 times for a White defendant who kills a White defendant. This difference is statistically significant [z=1.8, p=.0359] Although the multiplier is higher for Black defendants whose victim was White, than for Black-onWhite cases, the difference was not statistically [z=.8 p=.4237) There is, however, racial 69


difference in effect between Black and White defendants when the victim is White. The death sentence odds for the Black defendants with White victims increase by 9.8 for each increase in the level of aggravation, comp ared to only a 2.8 multiplier for the White defendant whose victim is White and killed under similarly aggravated circumstances. This difference Black defenda nts being nearly five (5) times more likely to receive a death sentence than White defendants who commit comparable murders of White victims, is statistic ally significant [z=2.4 (p =.0164) ]. Consistent with the other models, the de fendant-victim racial combination models show significant effects of the ranges of mitigators accepted, regardless of defendant and victim race. The odds of receiving the death sentence are reduced 69% in White defendant-White victim cases, 60% in Black defendant-Black victim cases, and 63% in Black defendant-White victim cases for each on e unit increase in the range of mitigators accepted. Although there is a slightly higher benefit for White defendants whose victims are White, and the least effect is for Bl ack defendants whose victims are White, the differences are not statistically signifi cant between any defendant-victim racial combinations. Here again the intercept for Black defenda nts, regardless of victim race, is negative, indicating that omitted variables and systematic error account for a slight reduction in the odds of receiving a death sent ence; which may indicate less arbitrariness operating in Black defendant cases than in White defendant cases, a la ck of arbitrariness which works to the detriment of Black defendants in terms of consistently harsher treatment. Even so, the results indicate that Black defendants pay a premium for their 70


race with each increase in the level of aggr avation accepted by a capital jury, regardless of the race of their victim; the highest price may be paid when the victim is White. The R 2 values show a better f it for the models for Black defendants whose victims are White. The model explains 52% of th e variance for Black de fendantWhite victim cases, versus only 39% in White defendant-W hite victim cases, and 40% in Black defendant-Black victim cases as well as the al l-cases model. The fits for the intra-racial models are slightly lower than for th e all-cases model, but comparable. Finally, it should be noted that the aggravat or effect is consistently the stronger effect across all models controlling for comparable levels of aggravation and mitigation. Jury acceptance of aggravating circumstances appears to have the stronger effect on sentencing outcome for Black defendants, rega rdless of the race of their victim, but is strongest in Black defendant-White victim cases. This supports the conclusion that Black defendants are treated more harshly by juries overall, in terms of the risk of a death sentence, than White defendants whose crimes are comparable. This also supports the conclusion that despite application of the Furman principles significant racial invariance has not yet been achieved in capital se ntencing, and sentencing bias based on a defendants race is still a significan t factor in capital sentencing. 71


Table 3. Logistic Regression Results (Odds Ratios): Race-of-Defendant and Race-of-Victim Combinations Showing Effect of Total Jury-Accepted Aggravating and Mitigating Circumstances on Capital Sentencing Outcomes, North Carolina (May, 1990 -December, 2002). Defendant Race Victim Race Variables B S.E. Exp(B) White Defendant Black Victim 4 NA NA NA NA White Victim Total Aggravators Accepted 1.023 .306 2.782* Total Mitigators Accepted -1.182 .196 .307* Constant 2.946 R 2 =.385 Black Defendant Black Victim Total Aggravators Accepted 1.851 .349 6.367* Total Mitigators Accepted -.931 .199 .394* Constant -.197 R 2 =396 White Victim Total Aggravators Accepted 2.291 .431 9.883* Total Mitigators Accepted -.963 .238 .382* Constant -1.523 R 2 =.517 z-scores Defendant-Victim Combinations Total Aggravators Accepted Total Mitigators Accepted z= 2.4* z= .72 WhiteD-WhiteV BlackD-White V z=1.8* z= .9 WhiteD-WhiteV -BlackD-Black V BlackD-WhiteV BlackD-Black V z=.8 z= .1 *p < .05 4 The small number of White-on-Black cases [n=26] prevented analysis of these cases. 72


Results of Modeling Accepted Individual Aggravating and Mitigating Circumstances. All-cases models. The results of the individual factor models are consistent with the totals accepted models, and are set out in Tables 4 and 5. Table 4 sets out the all-cases and the defendant and victim race-specific models. Table 5 repo rts the racial combination models. In the all-cases models, there was a significant race -of-defendant effect and a non-significant effect for race-of-victim, both in an unexpected negative direction. The Black defendant race-specific model and both victim race-spe cific models echo thes e negative directions for the racial variables. Race-of-victim was negative and significant in the Black defendant model, but positive and not significant in the race-specific White defendant model. The difference in effect was not st atistically significant. In the victim racespecific models, the defendant race effect was negative in both the Black and White victim models, but neither wa s significant, nor was there any statistically significant difference between them. These unexpected directions are possibly a reflection of the Black defendant and intra-racial crime sample bias. There is a higher percentage of Black defendants in the sample overall, 59% [374/632] of the cases ha ve Black defendants, compared to 41% of the cases with White defendants [258/632]. As previously noted, the distribution based on defendant and victim race is somewhat sk ewed. The preponderance overall are intraracial cases [71% (450/632)], of which the majority -58% (218/450) -are Black-onBlack. As to the remaining 29% [182/632] of the cases which are inter-racial, 86% [156/182 are Black-on-White -only 26 cases involve White defendants and Black 73


victims. Thus analysis and comparison is onl y viable as to how juries treated Black defendants overall, and in comparison to White defendants who kill White victims. Thus the race-of-victim effects cannot be fully eval uated, except as to White victims, but the sample is sufficient to assess race-of-defendant effects. Returning to the all-cases model, shown in Table 4, five (5) aggravators showed strong, significant effects: felony murder [n=241], heinous/atrocious/cruel (HAC) murder [n=203], murder during violent c onduct [n=273], prior violent felony record [n=179], and prior capital record [n=18]. HAC showed the stronge st effect, with a multiplier for the odds of receiving a death sentence of 5.7, over all cases where accepted. These five aggravators had significant effect s where accepted, in most models, and HAC was significant in all models. The small numb er of cases in which a prior capital record was accepted make the results as to this a ggravator less reliable, although it is worth noting that two-thirds [12/18] of the cases wh ere this was a factor had Black defendants. As to the mitigators, over all cases five (5) statutory mitigators and three (3) nonstatutory mitigators had a significant effect, all in the expected ne gative direction: no significant prior record [n=217] age [n=78], capacity [n=207], minor participation in the crime [n=25], and duress [n=48]. The nonstatutory mitigators which showed a significant effect were alcohol abuse [n =120], father absent from childhood home [n=118], and defendant had a specific mental illness [n=101]. All were in the expected negative direction. Sample size was an issue as to the mitigation effect overall of defendant as minor accomplice, duress and ag e, and this issue becomes more pronounces as the models are parsed into racial-specifi cs and combinations. As would be expected 74


with the addition of variables, the model f it increased for the all-cases model from .407 using only ranges of aggravation and mitigati on accepted, to .516 for the individual factor model of all cases. That is, defendant and vi ctim race, and jury acceptance of individual aggravating and mitigating factors explains almost 52% of the variance in the life-death decision. Defendant race-specific models. Consistent with the all-cases model, there is a negative race-of-victim effect in the Black defendant model, which is significant and indicates a 58% decrease in the likelihood of a death sentence based on victim race but the difference with the White defendant effect is not st atistically significant [z=.31, p=.03783]. As to the aggravators, the same five aggravators signifi cant in the all-cases models ar e also significant in either the Black or White defendant model, and a si xth aggravator murder for pecuniary gain [For $$] shows a significant positive effect on the life-death outcome in both Black and White defendant models. There are race-based differences, some of which are significant and some of which are not. Felony murder and a prior capital felony record are not significant for White defendants but are for Black defendants -but the differences in effect are not statistically significant. The effects of a prior violent fe lony record and murder for pecuniary gain are also significant for Black defendants but not for White, but here the differences are statistically significant. A Bl ack defendant is almost 9 times more likely to receive the death sentence if he has a prior violent felony record than a White defendant with such a record. Where the murder is found by the jury to have been for pecuniary gain, the Black 75


defendants odds of receiving a death senten ce increase by a multiplier of 3.3, compared to only a multiplier of 1.2 for a White defendant. Two of the aggravators, HAC and murd er during violent conduct, have a significant enhancing effect on the likelihood of a death sentence for both Black and White defendants. There is no significant diffe rence in the effect of violent conduct, and the odds ratios are comparable a 3.91 increase for White defendants and a 3.95 increase as to Black defendants. HAC shows a much gr eater and statistically significant different [z=2.0, p =.0228] effect on Black defendants the multiplier for White defendants where HAC has been accepted in aggravation is 4.2, versus 15.8 for Black defendants. As discussed more particularly below, this difference increases dramatically in the racial combination models, showing the greatest e ffect where the defendant is Black and the victim is White, and the multiplier increases to 83.2. The significant mitigatory effects were in the expected negative direction and similarly mixed to some degree by race of de fendant. Only defendants age and capacity were significant for both Black and White de fendants with age being slightly more mitigatory for Black defendants [n=52] than White [n=26], and lack of capacity for White defendants [n=102] than for Black [n=105], but the differences were not significant [age z=.2, p=.4207]; capacity z=.5, p=.3805] A defendants minor role in the crime [n=25] significantly reduced a Black de fendants odds of a death sentence by 94% and showed no significant effect for White defendants, but th e difference was not statistically significant [z=.004, p=.4984] and the result undoubtedly affected by the small sub-sample size. 76


One interesting result was the effect of the lack of a significant prior record [n=217]. This factor was significant for Wh ite defendant cases [104] but not for Black defendant cases [113], reducing the odds of receiving a death sentence by 88%, compared to only 37% if the defendant was Black, and the difference in effect was statistically significant [z=2.6, p=.0047]. On the other hand, the eff ect of a jurys acceptance of a defendants duress [n=48] was significantly [z=1.96, p=.025] greater for Black defendant cases [22] than for White [25], reducing th e odds of a death sentence by 92% for Black and only 65% for White defendants. As in the level of aggravation and mitigation models, the effect of jury acceptance of the HAC aggravator is strongest for Black defendants, and the difference in effect between Black defendants [15.8 multiplier] and White defendants [4.2 multiplier] is significant [z = 2.0, p=.0228]. HAC has the strongest effect of all for Black defendants whose victims are White, with a multiplier of 83.2. The difference between the HAC aggravation effect on White defendant-White victim cases and Black defendant-White defendant cases shown on Table 5 was al so statistically significant [z=2.1, p =.0179], as was the difference in the HAC effect betw een White and Black intra-racial crime [z=1.97, p =.0244]. One interesting result was the l ack of statistical significance in the difference of the HAC effect between Black-B lack and Black-White cases; it seems that Black defendants are going to be penalized more than White defendants in terms of effect on sentencing from acceptance of the HAC a ggravator regardless of the race of their victim, but the greatest effect will be if the vi ctim is White. Again, it must be noted that more [58%] of the Black defenda nt cases had Black victims. 77


Felony murder was significant in the all-cases model but not statistically significant for White defendants in the de fendant-race specific model or the White-onWhite case model. As tables reflect, the effect of this aggravator is comparable across all models, and there was no stat istically significant difference among them. This may be in part because felony murder is not usually a stand-alone aggrav ator unless the defendant is the triggerman, since the holdings in Enmund v. Florida (1982) and Tison v. Arizona (1987), and because of the double-dipping issue which prevents counting both the felony itself and felony murder as two separate aggr avators. Felony murder had a significant aggravating effect on Black defendant cases ov erall, but only in the Black intra-racial model. It was significant for Black defendants, but the di fference in effect on Black and White defendant cases was not significant. Felony murder was also not significant. except where the victim was White. Murder for pecuniary gain [For $$] was Four others felony murder, murder during violen t conduct, prior violen t felony record and prior capital offense were also statistically significant in the all-cases model. As to the defendant race-specific models as would be expected, more individual aggravators were significant for Black defendants than for White. For Black defendants felony murder, murder for pecuniary gain, murder during vi olent conduct, prior violent felony record, and prior capital offense were all significant, in addition to HAC. For White defendants, only HAC, and murder during violent conduct had a statistically significant effect. Victim race-specific models. As to the victim race-specific cases, four of the aggravating factors were significant for both Black and White victim cases: felony murder, HAC, murder during 78


violent conduct, and prior vi olent felony, but the differen ces were not statistically significant. Prior capital offense was also sign ificant if the victim was Black, in the small sample of Black victim cases where this aggravator was accepted (n=8). Jury acceptance of the HAC aggravator proved to have a str ong increased effect on the odds of a Black defendant receiving the death penalty. 79


Table 4. Logistic Regression Results (Odds Ratios): All-Cases, Defendant and Victim Race-Specific Mode ls Showing Effect of Jury Accepted Individual Aggravating and Mitigating Circumstances on Capital Sentencing Outcomes, North Carolina (May, 1990 December, 2002). All Cases White Defendants z ** Black Defendants Black Victims z White Victims B S.E. Exp(b) B S.E. Exp(b) B S.E. Exp(b) B S.E. Exp(b) B S.E Exp(b) Defendant Race -.683 .307 .505* NA NA NA NA NA NA -.373 .811 .689 -.627 .367 .534 Victim Race -.410 .296 .664 .126 .693 1.134 .94 -.872 .394 .418* NA NA NA NA NA NA Felony murder 1.143 .268 3.136* .795 .477 2.214 1.2 1.520 .395 4.574* 1.305 .492 3.688* .6 80 .960 .365 2.611* For $$ .470 .300 1.599 .160 .530 1.174 1.46 1.194 .463 3.299* 1.065 .668 2.900 .541 .390 1.178 HAC 1.748 .271 5.744* 1.428 .447 4.172* 2.0* 2.760 .473 15.796* 2.659 .570 14.282* 1.1 1.934 .369 6.919* Viol. Conduct 1.045 .257 2.844* 1.364 .515 3.913* .02 1.375 .386 3.954* 2.021 .538 7.548* 1.7 .892 .360 2.441* >1 Person Risk .507 .547 1.661 -.332 1.044 .717 1.196 .808 3.307 1.092 .919 2/980 1.005 .786 2.731 Prior Vio. Fely .967 .297 2.630* -.171 .532 .843 2.9* 2.203 .453 9.055* 1.514 .570 .4.543* 1.007 .413 2.737 In jail -.009 1.449 .992 -.196 1.837 .822 17.359 25592. 34E+07 16.982 26723 2372278 -.813 1.696 .443 Prior Cap. Off. 1.654 .756 5.226* 1.655 1.639 5.232 .44 2.484 1.058 11.994* 3.901 1.477 49.464* 1.268 1.225 3.555 Escaping .598 .385 1.818 .830 .665 2.294 1.210 .717 3.355 1.507 .897 4.515 .200 .474 1.222 Kill Officer -.125 1.307 .882 19.914 27292 45E+08 -.821 1.699 .440 20.437 40192 7.5E+08 -1.084 1.624 .338 Kill Gov. Off. .337 .720 1.401 -.612 1.480 .542 .909 .861 2.482 -.089 1.529 .915 .834 .910 2.032 No SigRec. -1.163 .269 .313* -2.129 .493 .119* 2.6* -.464 .398 .629 -.788 .486 .455 1.56 -1.760 .386 .172* Mt/Em Dis. .221 .260 1.247 .144 .463 1.155 .323 .380 1.381 .449 .439 1.567 .010 .366 1.011 D Age -1.383 .384 .251* -1.563 .801 .209 .2 -1.751 .550 .174* -1.166 .711 .312 .9 -1.959 .540 .141* D Capacity -1.322 .260 .267* -1.614 .441 .199* .5 -1.336 .399 .263* -1.471 .488 .230* .02 -1.483 .357 .227* D Minor Acpl. -2.598 .768 .074* -20.30 14013 .000 --2.694 .957 .068* -.755 1.094 .470 22.796 7983 .000 D Duress -1.805 .506 .164* -1.051 .705 .349 2.0 -3.926 1.232 .020* -4.289 1.511 .014* -1.250 .618 .287 D Aid Prosec. -.094 .457 .910 -1.365 .811 .255 .6 2.335 .818 10.328* .939 1.250 2.559 .273 565 1.314 Vic. Consent -.745 1.186 .475 .617 2.226 1.853 -19.15 16329. .000 -19.03 14153 .000 21.183 40193 1.6E+09 *p <.05 ** 1 The z-score indicates whether the difference in effect between two models is statistically significant ((Paternoster, Brame, M azerole & Piquero, 1998).


Table 4, continued. Logistic Regression Results (Odds Ratios): All-Case, Defendant and Victim Race-Specific M odels Showing Effect of JuryAccepted Individual Aggravating and Mitigating Circumstances on Capital Sentencing Outcomes, North Carolina (May,1990 December, 2002) All Cases White Defendants z Black Defendants Black Victims z White Victims B S.E. Exp(b) B S.E. Exp(b) B S.E. Exp(b) B S.E. Exp(b) B S.E. Exp(b) D Alcohol -.805 .341 .447* -.654 .603 .520 -.693 .614 .500 -.062 .645 .940 -.873 .485 .418 D Drugs .115 .319 1.122 -.302 .603 .739 .302 .494 1.352 .060 .630 1.062 .132 .457 1.141 D Physical Abuse -.168 .353 .845 -.389 .531 .677 -.461 .630 .631 -.187 .684 .830 -.471 .478 .625 D Sexual Abuse .124 .543 1.131 -.523 .802 .592 1.345 1.014 3.840 1.350 1.180 3.858 -.482 .716 .618 D Broken Home -.139 .359 .871 -.283 .580 .754 .493 .563 1.638 -.279 .569 .756 -.235 .535 .790 D Father Absent -.756 .334 .470* -1.94 .795 .144* 1.8* -.395 .424 .674 -.327 .513 .721 -1.217 .525 .296 D Mother Absent -.144 .467 .866 -1.88 1.074 .153 1.283 .720 3.606 1.465 .934 4.329 -.834 .674 .434 D in Foster Home .611 .711 1.843 2.682 1.250 14.612* 1.1 .731 1.265 2.078 1.584 1.522 4.873 .872 .860 2.391 D Saw Parentl Misc. .067 .306 1.069 .695 .507 2.003 -.694 .510 .500 -.397 .535 .672 .558 .447 1.745 D Low IQ .231 .395 1.260 .397 .697 1.488 -.385 .595 .681 .279 .708 1.322 .96 81 .595 .571 1.813 D Spec. Mtl Ill -.70 .311 .495* -.917 .533 .400 -.598 .489 .550 -.624 .607 .536 -.551 .427` .576 Constant .478 1.301 -1.71 -1.52 .600 R 2 .516 .626 .888 .607 .568 *p < .05


The strongest effect is seen on Black defenda nts who commit crimes found by the jury to be heinous, atrocious or crue l (HAC). The odds multiplier for HAC in the all-cases model was 5.7, but in the Black defendant mode l rose to 15.8, compared to a decrease as to White defendants from the all-cases model to 4.2. The difference in effect of the HAC aggravator on the likelihood of receiving a death sentence, between White and Black defendants was statistica lly significant [z=2.1 (p<.05) ]. An unexpected result was the disparity in the effect of the HAC aggravator based on victim race. Jury acceptance of the HAC aggravator increased the odds of a death sentence more when the victim was Black by a multiplier of 14.3 -than when the victim was White, which increased the odds by a multiplier of 6.9, but the differen ce between them was not statistically significant [z=1.1, p=.2713) ] As to mitigating circumstances, only one factor was significant in all eight models using the individual aggravating and mitigating circumstances defendants capacity to understand the criminality of his act. The seven mitigators significant at p<.05 over all cases were defendants non-significant prior record, age, de fendant was a minor accomplice, under duress, abused alcohol, de fendants father was absent from the childhood home, and defendant suffered a specifi c mental illness. All of these effects were in the right direction, indicating a mitigatory effect albeit slight -on sentencing when accepted. When looking mitigation effects in the defendant race-specific models, in addition to defendants capacity, both sets of defendants showed other mitigatory effects significant at p<.05, although these ad ditional mitigators differed in some respects by race. 82


The White defendant model showed significant mitigatory effects as to no significant prior record, defenda nts age, lack of capacity, duress, defendant aided the prosecution. The Black defendant model showed age, lack of capacity, minor accomplice, duress, and aid to the prosecution as significant. All of the effects were in the right direction, indicating a mitigatory effect, except as to aiding the prosecution, which was significant for Black defendants but in the wrong direction, indicating an increase by a multiplier of 10, in th e odds of a death sentence for Black defendants who helped the prosecution. The difference between Black and White defendants was not statistically significant [z=1.1 (p=.2713) ]. There was a similar unexpected positive direction for the foster home mitigator as to White defendants, producing a statistically significant multiplier of 14.1, but the differe nce was similarly not significant [z=1.1 ( p=.2713) ]. In fact, there were no significant differences in the mitigatory effects of any of the mitigating factors, between Black and White defendants. The defendant victim race-specific m odels showed a reduced number of mitigatory factors where the victim was Black, with only defendants capacity and duress showing a significant effect. The White vict im model produced sign ificant effects from where the defendant had no significant prio r record, and based on his age, lack of capacity, duress, and the father being absent from the home. All effects were in the expected direction and none of the differences in mitigatory effects based on victim race were statistically significant. 83


Defendant-Victim racial combination models. The defendant-victim racial combination models are reported in Table 5 produced mitigatory effects consistent with the all-cases models. Defendant age and lack of capacity were significant for all racial co mbination, but the differences across models were not significant. The same mitigatory factors were significant for Black defendants in the racial combination models as had been in the defendant race-specific model, except defendant as minor accomplice was not significa nt in any racial combination model, and duress was only significant in the Black de fendant-Black victim cases. The foster home and witnessing parental misconduct mitigators showed as significant in White-on-White crime but not in the Black in traor inter-racial crimes. There was a statistically significant difference in the mitigatory effect of defendants duress on intra-racial cases, where the mitigatory effect was stronger for Black-on-Black crime, than for White-onWhite crime [z=2.4 (p=.0214) Except for the duress mitigator for Black defendants there were no significant differences in mitigat ory effect of other factors and there was no significant differences in the effect of any of the mitigators as between interracial cases where the defendant was Black, or any cases where the victim was White. As to model fit, the R 2 values increased for all of the individual factor models over the total factors accepted models. The all-cases value went from .407 to .516, explaining 52% of the variance. The White defendant case model R 2 value increased from .391 to .588, indicating that 59% of the variance is explaine d by the individual aggravating and mitigating factors. The Black defendant model value increased from an R 2 value of .445 to .626, explaining nearly 63% of the variance between sentencing 84


outcomes. The victim race-specific models also increased from an R 2 value of .421 to .568, increasing in fit by 15 percentage points. The model fit for Black victim cases also improved, the 40% [R 2 =.397] explained by the all-cases model growing to nearly 61% [R 2 =.607] of the variance explained in the individual factor Black victim case model. The use of individual factors improved the fit of the racial co mbination models. The White-on-White model using indivi dual factors explained about 60% [R 2 =.598] of the sentencing outcome variance, over only 39% of the variation explained by levels of aggravation and mitigation. The Black-onBlack model improved from 40% using the ranges of factors accepted, to 62% with individual factors accepted. Overall, the models appear to be more efficient in predicting sentencing outcomes for Black defendants, and when individual factor are measured. The larg est increase in fit between the aggravationmitigation level models and individual factor models is for Black-on-White crime, going from explaining 52% to 80% of the variation in sentencing outcome. The North Carolina statute overall seems to be working efficiently to produce consistent sentencing results, particularly for Black defendants who endure a consistent and predictable disadvantage as to how the aggravating factors affect their sentencing outcome. Most of the unexplained variance is in cases where the defendant is White. While some of the unexplained variance is at tributable to unknown factors such missing variables and systemic error, and to the inabil ity to factor in the i ndividual weights of the factors or the weighing of the aggravators and mitigators, it appears that race is playing a role in how the Furmancompliant standards are applie d, and raising the question of arbitrariness and possible jury bias. 85


Table 5. Logistic Regression Results (Odds Ratios) : Defendant-Victim Racial Combination Mo dels Showing Effect of Jury-Accepted Individual Aggravating and Mitigating Circumstances on Capital Sentencing Outcomes, North Carolina (May, 1990 December, 2002). White Defendant White Victim 1 Black Defendant Black Victim Black Defendant White Victim B S.E. Exp(b) B S.E. Exp(b) B S.E Exp(b) Felony murder .823 .512 2.277 1.385 .528 3.996* 1.380 .858 3.974 For $$ .014 .546 1.014 .871 .681 2.389 2.627 1.067 13.832* HAC 1.506 .484 4.510* 3.042 .625 20.952* 4.421 1.246 83.214* Viol. Conduct 1.287 .543 3.620* 1.999 .571 7.385* .860 .817 2.363 >1 Person Risk -.114 1.103 .892 1.450 .983 4.265 3.206 2.004 24.684 Prior Vio. Fely .024 .583 1.025 1.857 .639 6.403* 4.247 1.154 69.908* In jail -.281 2.046 .755 17.080 25916.037 26175712.907 NA NA NA Prior Cap. Off. .288 1.850 1.333 3.465 1.521 31.971* 19.992 6950.775 481179454.364 Escaping .292 .675 1.339 1.146 1.014 3.145 1.121 1.679 3.067 Kill Officer 19.775 27229.935 387506038.119 20.187 40192.970 584927857.974 -42.428 24716.259 .000 Kill Gov. Off. -.573 1.498 .564 -.001 1.628 .999 3.061 1.719 21.340 No SigRec. -2.205 .553 .110* -.604 .538 .547 -1.302 .933 .272 Mt/Em Dis. .160 .527 1.174 .348 .477 1.416 -.848 .941 .428 D Age -2.349 .898 .095* -1.654 .780 .191* -2.797 1.141 .061* D Capacity -1.615 .490 .199* -1.420 .535 .242* -2.389 .954 .092* D Minor Acpl. -21.001 13525.905 .000 -1.146 1.142 .318 -49.299 12440.513 .000 D Duress -.633 .726 .531 -4.656 1.527 .010* -17.048 9955.123 .000 D Aid Prosec. -.507 .906 .602 2.944 1.840 18.997 3.723 1.538 41.39* Vic. Consent 19.809 40192.970 400870714.217 -18.858 16422.489 .000 NA NA NA p .05 86


Table 5, continued .: Logistic Regression Results (Odds Ratios) : Defendant-Victim Racial Combination Mo dels Showing Effect of Jury-Accepted Individual Aggravating and Mitigating Circumstances on Capital Sent encing Outcomes, North Carolina (May, 1990 December, 2002). White Defendant White Victim 1 Black Defendant Black Victim Black Defendant White Victim B S.E. Exp(b) B S.E. Exp(b) B S.E Exp(b) D Alcohol -1.135 .669 .321 -.527 .722 .591 1.932 2.357 6.903 D Drugs -.107 .655 .898 .163 .731 1.176 -.396 1.051 .673 D Physical Abuse -.684 .591 .504 -.569 .831 .566 .822 1.503 2.275 D Sexual Abuse -.599 .810 .549 1.469 1.328 4.346 18.889 7295.652 159781410.692 D Broken Home -.188 .670 .828 .316 .649 1.372 87 .777 1.854 2.176 D Father Absent -2.334 .916 .097* -.295 .551 .745 -1.203 1.091 .300 D Mother Absent -1.595 1.164 .203 1.634 1.001 5.125 -.194 1.544 .823 D in Foster Home 2.541 1.298 12.689 1.387 1.528 4.004 3.054 10.908 21.193 D Saw Parentl Misc. 1.161 .601 3.192 -.497 .607 .608 -.982 1.667 .375 D Low IQ .054 .792 1.056 .239 .795 1.270 .214 1.367 1.239 D Specific Mental Illness -1.069 .588 .343 -.950 .686 .387 1.044 1.016 2.841 Constant 1.562 -2.026 -2.933 R 2 .598 .624 .795 *p < .05 1 There were two few cases (26) for SPSS to produce re sults for the White Defendant-Black Victim cases. zscores Z-Scores White Def/White Victim Black Def/White Victim White Def/White Victim Black Def/Black Victim Black Def/White Victim Black Def/Black Vic. Felony Murder z= .76 Felony Murder z= .025 Murder for Pecuniary Gain: z= 2.18* Murder for Pecuniary Gain z= .82 HAC z= 2.1* HAC z= 1.97* HAC z= 1 During Violent Conduct z= .44 During Vi olent Conduct z= .9 During Violent Conduct z= 1.14 Prior Violent Felony z= 3.27* Prior Vi olent Felony z= 2.1* Prior Violent Felony z= 1.8* Prior Capital Felony z= 1.32 Prior Capital Felony z= .0000+ No Sig. Prior Record z= .83 No Sig. Prior Record z= 2.07* No Sig. Prior Record z= .68 Defendant Age z= .3 Defendant Age z= .6 Defendant Age z= .8 Defendant Capacity z= .8 Defendant Capacity z= .4 Defendant Capacity z= .9 D Under Duress z= 2.4* Defendant Aid Prosecutn z= 1.8* Defendant Aid Prosectn z= .32 D Father Absent z= .8 D Father Absent z= 1.905* D Father Absent z= .7


Chapter Six Conclusions Study Purpose and Results The purpose of this study was to assess the degree to which capital sentencing under the guided discretion sentencing standards required by Furman v. Georgia (1972) has produced predictable, consistent and racial ly invariant death sent encing patterns. It was not intended to assess the degree to which racial bias might already have influenced the progress of a case to the life-death deci sion, but was limited to the penalty stage, where Furman applies, in order to look at possibl e racial influences on the sentencing process itself. Thus, to the extent race entered into the decisions which resulted in the racial composition of the sample reaching th e penalty stage, it is not measured or considered here. It is possible, however, that any cumulative racial bias injected into the process during earlier discreti onary decisions arguably makes racial variance discovered at the sentencing stage a conservative reflect ion of race in the capital justice system overall. To isolate the sentencing deci sion in order to look at raci al disparities or the lack of them, it was logical to compare sentenci ng outcomes among Black-White race specific cases, defendant-victim racial combination cases and all cases as the product of the application of the only legally-relevant sent encing criteria under Furman, the statutorily 88


expressed aggravating and mitigating circumstances. The aggravators presumably encompass all criteria which make a crime death worthy, and the mitigators allow broad consideration of factors which might offset the crime and justify a life sentence in his particular case. The goal was to determ ine whether jury sentencing discretion under Furman has been sufficiently channeled to avoid or minimize to insignificant levels the influence of extra-legal consideratio ns such as racial prejudice. The study looked at the sentencing decisi ons of 632 capital juries in North Carolina between May 1990 and December 2002 after they applied North Carolinas Furmancompliant capital sentencing statute (S ection 15A-2000, NC General Statutes). By controlling for defendant and victim race and the numeric range of aggravating and mitigating circumstances accepted and applied by the jury and also the effect of each factor individually -it was possible to observe the degree to which these factors affected the risk of defendants receiving a death se ntence for comparable crimes during the selected time-frame, and the degree to whic h these effects differed by race. To be conservative, only results at an alpha leve l of less than .05 are reported, indicating a confidence level of more than 95%. In other words, the study tried to determine if the North Carolina death sentencing statute has operated to produce the consistent and predictable results which Furman held the Eighth Amendment to require, and if it does so in a race-neutral way. A demonstrated, statistically significant lack of racial invariance in how a Furmancompliant sentencing statute works when operating in a presumptively race-neutral setting, should call into question the continued va lidity of the current, Furman-based death sentencing model. 89


Constitutional principles of racial equity, jury impartiality, arbitrariness of punishment, require a death sentencing process which if not completely even-handed is sufficiently consistent and predictable to l eave no room for doubt as to its fundamental fairness between the races. Statistically signifi cant racial disparities in how facially raceneutral sentencing standards are applied, regard less of how relatively even the unadjusted life-death sentencing rates may be distribute d between Black and White defendants, are indicative of the operation of racial bias in the system which should be just as unacceptable under the law as the stark Black-White disparities seen preFurman. However, significant bias observed at this po int in the societal integration curve is probably not the result of institutionally purposeful denial of equal protection under the law; it is more likely the result of the continued effect of li ngering unconscious, culturebased racial stereotypes. The issue is whether any statistica lly significant racial disparity, intended or not, is an acceptable outcome in the operation of our laws where death is the consequence. Key Findings The results of the study s upport a conclusion that the Furmanbased criteria North Carolina has adopted to guide the capital se ntencing decision have substantially reduced unpredictability in death sentencing in that state. The results do not support a conclusion that Furman has worked to eliminate or even reduce race-based arbitrariness to constitutionally insignificant levels in capital sentencing in North Carolina. These are the key findings of the study: 1. The North Carolina statute is moderately efficient in produc ing consistent and predictable capital sentenci ng results across all cases. 90


2. The North Carolina statute is highly efficient in producing consistent and predictable capital sentencing results for Black defendants, particularly as to death sentences. 3. The North Carolina statute is not e fficient in producing consistent and predictable capital sentencing results for White defendants. 4. Black defendants pay a significantly higher premium in terms of the risk of a death sentence for increases in the le vel of aggravation a jury accepts. 5. Black defendants pay a significantly highe r premium in terms of the risk of a death sentence where the murder for p ecuniary gain, HAC, and prior violent felony record aggravators are accepted. 6. Black defendants pay the highest premiu m in terms of the risk of a death sentence where their victim was White. 7. The race of the victim makes no statistic ally significant difference in the risk of receiving the death penalty, except where the defendant is Black and the victim is White. 8. Jury acceptance and application of a ggravating circumstances has a much stronger influence on sentencing outcomes regardless of defendant or victim race, than jury acceptance and application of mitigating circumstances. 9. There is no statistically significant racebased difference in the effect of an increase in the mitigation level of a case. 10. There is no statistically significant difference in the mitigatory effect of individual factors ba sed on victim race. 91


11. Only two mitigatory factors were significantly different based on defendant race: The lack of a significant prior record decreases the risk of a death sentence more for White defendants versus Black defendants; and jury acceptance that the defendant acted under duress favors Black defendants. These results indicate that while the No rth Carolina statute appears to working efficiently overall, there is a statistically significant bias against Black defendants in terms of how aggravating circumstances affect their risk of a deat h sentence, regardless of the race of their victim, over White defendants whose crimes are comparably aggravated. Where levels of aggravation and mitigation are controlled for, Black defendants are three (3) times more likely to receive the death penalty than White defendants. When looking at the individua l aggravators accepted, the odds multipliers increase significantly for Black defendants over White, with HAC the most subjective aggravator having a very significantly larg er effect for Black defendants, and an enormous effect if the victim was white, compared to White defendants who kill white victims in a heinous, atrocious or cruel way. These disparities are statistically significant at p < .05, but whether they are enough to demonstrate constitutional significance is unknown. The analysis shows little, if any, disparity in how the North Carolina statute was applied which can be attributed to the race of the victim. The only statistically significant difference possibly attributable to a race-of-victim effect was the enhanced aggravation effect on Black defendants whose victims were White over the effect on White defendants whose victims were White, but this appears more of a race-of-defendant effect than a race of victim effect because: 1) th ere was no statistically significant difference 92


between the way jury-accepted aggravators and mitigators affected a defendants sentencing outcome based on the victims race, except between White and Black defendants whose victims were White; and 2) there was no statistically significant difference in how the aggravators influenced sentencing outcomes for Black defendants, period regardless of the race of their victim. The comparison of the White defendant-White victim and Black defendantWhite victim models produced the la rgest disparities, thus conf irming the fears about Black-onWhite murders expressed in Turner and McClesky. The Black defendant who killed a White victim was at the highe st risk overall, with an odds multiplier 3.5 times higher [9.883] as for a White defendant who killed a White victim [2.782]. Where the victim was White and the HAC aggravator was accepte d, a Black defendants risk of a death sentence was 18.5 times higher than White defendants in the same s ituation, and if the prior violent felony aggravator was presen t the risk was 69 times higher for Black defendants than White. McCleskey did not find a 4 point dispar ity in the race-of-victim effect sufficient to raise Eighth Amendment concerns as to McCl eskeys individual sentence. Whether the disparities shown in the study presented he re are or are not constitutionally significant as showing un acceptable jury bias, arbitrary and capricious sentencing or a violation of s ubstantive due process requirement s of racial parity in death sentencing in either an individual or system ic context, is unknow n, but the statistical significance and the consistency of these disparities raise a question of whether the death sentence can ever be fair in a Black-on-White crime. The model fit values undersc ore the apparent race-of-de fendant disparity in how the Furman standards have been applied in North Ca rolina. The all-case model indicates 93


that the North Carolina sentencing statute is wo rking moderately well overall, with jury acceptance and application of the aggravating and mitigating factors to crimes within comparable aggravation and mitigation ranges explaining 41% of variance in sentencing outcomes, rising to 52% when parsed into indi vidual aggravator and mitigator effects. The defendant race-specific models increase the fit for Black defendants over the allcases models to 45% of the variance in outcome based on aggravation/mitigation levels and 63% in the individual factor models, compared to a slight decrease in fit from the allcases model for White defendants to 39% but an increase over the all-cases model where individual factors we re used to 59%. The victim race-specific aggravation/mitiga tion level models are not far apart in their fit the Black victim model explaine d about 40% of the va riance and the White victim model 42% of the variance. When i ndividual factors are us ed, however, the fit becomes slightly better for Black victim cases than White, explaining 61% of the variance in Black victim cases, and 57% in Wh ite victim cases. The racial combination models reported at Tables 2 and 5 also de monstrate higher model efficiency for Black defendants than White. The fit for the level of aggravation/mitigation models decreases from the all-cases model in the White-on-Wh ite sub-sample indicating only 38% of the variance in sentencing outcomes is explained; evidence that aggravation/mitigation levels operate less efficiently in predicting sentencing outcomes for White defendants. Where individual factors are used, however, the model fit increases to nearly 60% (R 2 =.598) The fit increases for Black defendant s, regardless of the race of the victim, across all models, reach ing the best fit in the Black defendant-White victim cases where the aggravation/mitigation level model explains 52% of the variance, and the individual 94

PAGE 100

factor model explains almost 80% (R 2 =.795). This is a much higher level of predictive efficiency than the 50% value rejected in McCleskey. Implications of Findings, Li mitations of Study, and Need for Additional Research The primary implication of these findings is that it does not appear that the Furman approach has eliminated race-based arbi trariness under the Eighth Amendment, at least as to the operation of death sentencing procedures in North Carolina. The statistically significant differences in how juri es have applied the North Carolina statute to Black and White defendants cannot reasonably be attributed completely to non-racial factors, and may reasonably be accepted as evidence of arbitrariness founded in unconscious juror racial bias which somehow works in fa vor of White defendants and against Black defendants. The statute is much less efficient in predicting death sentences for White defendants than for Black. Whether this is the product of unconscious juror bias in weighing the seriousness of th e aggravating factors when the defendant is Black, or unconscious bias as to the mitigatory e ffects favoring White defendants cant be determined. What can be said is that there appears to be inequity, pr obably race-based, in how the North Carolina juries apply the No rth Carolina sentencing statutes to Black defendants. Whether the racial disparities observed in this study reach the level of constitutional signif icance referenced in McCleskey remains to be seen, but seem sufficiently significant to warrant further research to determ ine whether there is a basis for an Eighth Amendment challenge to the No rth Carolina statute. Moreover, if it is assumed that these results are caused by th e unconscious racial attitudes of jurors 95

PAGE 101

impossible to detect or eliminate through pr ocedural safeguards and curable only by time, then the Sixth Amendment issue is raised: Wh at if it is impossible to select a racially impartial jury in North Carolina because of lingering cultural racial biases? How much, if any, demonstrated racial disparity in jury decisionmaking equals impartiality violative of the Sixth Amendment? The results here thus also support a Sixth Amendment challenge to the North Carolina sentencing statute. Finally, if it is impossible to eliminate racial disparity in the imposition of the death penalty -or at least reduce it to insigni ficant levels -because it is the product of unconscious social/cultura l attitudes, a substantive due proc ess issue is raised. Is it an infringement on liberty interests to allow the happenstance of a persons race to influence in any significant way whether he lives or di es, regardless of his crime? There would seem to be a substantive liberty interest, ar ising out of constitutional guarantees of due process, to be free from significant racial discrimination period, in the operation of law, as much as there is to be free from purpos eful racial discrimination, certainly where the result is death. Thus this study provides a possible basis for a claim of denial of substantive due process to Black defendant s sentenced to death in North Carolina. It must be stressed again, however, that a statistical analysis such as the one presented here is likely not to be found accep table to prove a claim requiring specific or implied intent to discriminate, as is necessa ry in an individual equal protection claim seeking individual relief like the McCleskey case. An analysis such as this could be used to seek blanket relief from further operation of a specific statute, including enjoining further executions based on sentences pr oduced from the challenged statute. 96

PAGE 102

Further research is needed. The study is limited, intentionally, to the FurmanGregg factors as applied in North Carolina and does not take into account the possible influence of other case characteristics such as urban-rural effects, type of attorney, specific victim vulnerability factors such as age, gender or disability which could also account for the unexplained variance. It is possible that an analysis of aggravator and mitigator submission-rejection patterns could bolster or call into question the conclusion that racial bias is influencing sentencing outcomes. Small sub-set sample sizes may also have confounded the effects. Finally, without another states sentencing outcome data to compare, it is not possible to conclude from this study that the Furman approach has not reduced racial inequity in capital sentenci ng anywhere, because the study is limited to North Carolinas statute and North Carolina se ntencing outcomes. The results raise a red flag, however, and the study offers a valid appr oach to legal arguments about the lack of racial invariance in death sentencing under Furman v. Georgia. 97

PAGE 103

References Articles Baime, D. (2001). Report to the New Jersey Suprem e Court Systemic Proportionality Review Project Trenton, NJ: New Jersey Supreme Court. Baldus, D. C., & Woodworth, G. (2003). Race di scrimination in the administration of the death penalty: An overview of the empi rical evidence with special emphasis on the post-1990 research. Criminal Law Bulletin 194-226. Baldus, D., & Woodworth, G. (2004). Srace di scrimination and the legitimacy of capital punishment: Reflections on the interaction of fact and perception. DePaul Law Review, 53(Summer), 1411-1487. Baldus, David C., Woodworth, George G., & Pulaski, Charles A., Jr. Equal Justice and the Death Penalty: A Legal and Empirical Analysis. Boston: University Press, 1990. Baldus, D., Woodworth, G., & Pulaski, C.A., Jr. (1994). Reflections on the inevitability of raci al discrimination in cap ital sentencing and the impossibility of its prev ention, detection, and correction, Washington & Lee Law Review, 51 (Spring), 359-428. Baldus, D.C., Woodworth, G., Grosso, C., & Christ, A. (2002). Arbitrariness and discrimination in the administration of the death penalty: A legal and empirical 98

PAGE 104

analysis of the Nebr aska experience (1973-1999). Nebraska Law Review, 81 486756. Baldus, D.C., Woodworth, G., Weiner, N., & Broffitt, B. (1998). Racial discrimination and the death penalty in the postFurman era: An empirical and legal overview, with recent findings from Philadelphia. Cornell Law Review, 83, 1683-1770. Bedau, H. (1996). Interpreting the Eighth Ame ndment: Principled vs. populist strategies. T.M. Cooley Law Review, 13, 811. Bienen, L.B., Weiner, N., Denno, D., Alison, P., & Mills, D. (1988). The reimposition of capital punishment in New Jersey: The role of prosecutorial discretion. Rutgers Law Journal, 41, 327-372. Bird, D.G. (2003). Life on the line: Pondering the fate of a substantive due process challenge to the death penalty. American Criminal Law Review, 40, 13291386. Birk, R., Li, A., & Hickman, L. (2004). Statis tical difficulties in determining the role of race in capital cases: A re-analysis of da ta from the State of Maryland. Retrieved March10, 2005, from University of California Los Angeles site: http://preprints.sta Blume, J., Eisenberg, T., & Wells, M. T. (2004). Explaining death ro ws population and racial composition. Journal of Empirical Legal Studies, 1(1)[March], 165-297. Blume, J., Johnson, S., & Eisenberg, T. (1998). PostMcCleskey racial claims in capital cases. Cornell Law Review 83 1771-1810. Bortner, P., & Hall, A. (2002). Arizona first-degree murder cases summary of 1995-1999 indictments: Dataset II research repo rt to Arizona Capital Case Commission Phoenix: Arizona Capital Case Commission. 99

PAGE 105

Bowers, W.J., & Foglia, W.D. (2003). Still singularly agonizi ng: Laws failure to purge arbitrariness from capital sentencing. Criminal Law Bulletin 51 -86 Bowers, W.J., Sandys, M., & Steiner, B. ( 1998). Foreclosed impar tiality in capital sentencing: Jurors predispositions, guilt -trial experience and premature decisionmaking. Cornell Law Review 83, 1476-1557. Bowers, W., Steiner, B., & Sandys, M. (2001) Death sentencing in Black and White: An empirical analysis of the role of jurors race a nd jury racial composition. University of Pennsylvania Journal of Constitutional Law, 3, 171-274. Brock, D., Cohen, N., & Sorensen, J. (2000). Arbitrariness in the imposition of death sentences in Texas: An analysis of f our counties by offense seriousness, race of victim and race of offender. American Journal of Criminal Law, 28, 43-71. Crump, D. (1996). How do the courts really discover unenumerated fundamental rights? Cataloguing the methods of judicial alchemy. Harvard Journal of Law & Public Policy, 19, 795). Delgado, R. (2001). Two ways to think about race: Reflections on the id, the ego, and other reformist theories of equal protection. Georgetown Law Journal, 89, 228285. Eckland-Olsen, S. (1988). Structured discreti on, racial bias and th e death penalty: The first decade after Furman in Texas. Social Science Quarterly 69, 853-873. Garfinkel, Harold (1949). Research no tes on inter-and intra-racial homicides. Social Forces, 27,369. Graines, S., & Wyatt, J. (2000). The Re hnquist Court, legal process theory, and McCleskey v. Kemp American Journal of Criminal Law, 28, 1-40. 100

PAGE 106

Gross, S. R., & Mauro, R. (1989). Death and discrimination: Racial disparities in capital sentencing Boston: Northeastern University Press. Holowinski, I.A. (2002). Note: Inherently arbitrary and capricious : An empirical analysis of variations among state death penalty statutes. Cornell Journal of Law and Public Policy, 12 231-268. Howe, S.W. (2004). The futile que st for racial neutrality in capital selection and the Eighth Amendment argument for ab olition based on unconscious racial discrimination. William and Mary Law Review, 45, 2083-2166. Johnson, Guy (1941). The negro and crime. Annals of American Academy of Political and Social Science, 217, 92-131. Joint Legislative Audit and Review Commission of the Virginia General Assembly (2002). Review of Virginias System of Capital Punishment. Retrieved March 11, 2005 from State of Virginia site: .us/reports/rptw274.pdf Lawrence, C.R., III (1987). The id, the e go, and equal protection: Reckoning with unconscious racism. Stanford Law Review, 39, 317 Keil, T.J., & Vito, G.F. (1995). Race and the death penalty in Kentucky murder trials: 1971-91. The Advocate 17 5-15. Kennedy, R.L. (1988). McCleskey v. Kemp : Race, capital punishment, and the Supreme Court. Harvard Law Review, 101, 1388. King, Nancy J. (1993). Postconviction review of jury discrimination: Measuring the effects of jury race on jury decisions. Michigan Law Review 92, 63 104. Klein, S.P., & Rolph, J. (1991). Relationship of offender and victim race to death penalty sentences in California. Jurimetrics, 32, 33-48. 101

PAGE 107

Lenhardt, R.A. (2004). Understanding the mark: Race, stigma, and equality in context. New York University Law Review, 79, 803. Lenza, M. Keys, D., & Guess, T. (2003). The prevailing injustices in the application of the death penalty in Missouri ( 1978-1996). Retrieved March 10, 2005, from University of Missouri-St. Louis site: nces/forlanglit/mbp/Lenza.html Leopold, A. (1998). Objective tests and subjective bias: Some problems of discriminatory intent in the criminal law. Chicago-Kent Law Review 73, 533-558. Lynch, M.J. & Patterson, E.B. (Eds.). Justice with prejudice: Race and criminal justice in America. Guilderland, NY: Harrow and Heston, 1996. McAdams, J.C. (1998). Racial di sparity and the death penalty. Law & Contemporary Problems, 61[Autumn], 152-170. McCord, D. (2002). A year in the life of d eath: Murders and capital sentences in South Carolina. South Carolina Law Review, 53, 250-359. Ohme, A. (2005). Psychology: Conditioned fear of a face: A prelude to ethnic enmity? Science 309, 711-713. Paternoster, R., & Brame, R. (2003). An Empirical analysis of Marylands death sentencing system with respect to the influence of race and legal jurisdiction. Retrieved March 9, 2004 from University of Maryland site: http://www.urhome/ Paternoster, R, Braymer, Mazerolle, P., & Pi quero, A. (1998). Using the correct statistical test for the equality of regression coefficients. Criminology 36, 859-866. 102

PAGE 108

Paternoster, R., & Kazyaka, A.M., (1988). Th e administration of the death penalty in South Carolina: Experiences over the first few years. South Carolina Law Review 39, 245-414. Pierce, G.L., & Radelet, M.L. (2004, Novembe r). The impact of legally inappropriate factors on death sentencing in Califo rnia: 1990-1999. Paper presented at 2004 Annual Meeting, American Society of Criminology, Nashville, TN. Pierce, M., & Radelet, G. (2002). Race, re gion, and death-sentencing in Illinois, 19881997. Oregon Law Review, 81, 39-85. Pillai, K.G. Jan (2001). Article: Shri nking Domain of Invidious Intent. William and Mary Bill of Rights Journal, 9, 525. Radelet, M. (2001) Gregg vs. Georgia: The beginning of the modern era of America's Death Penalty. Retrieved March 14, 2005 from Amnesty International site: Radelet, M., & Pierce, G. (1991). Choosi ng those who will die: Race and the death penalty in Florida. Florida Law Review, 43, 1-34. Rubin, P.J. (2003). Square pegs and round hol es: substantive due process, procedural due process, and the Bill of Rights. Columbia Law Review, 103, 833-894. Smith, M.D. ( 1987). Patterns of discriminati on in assessments of the death penalty: The case of Louisiana. Journal of Criminal Justice 15, 279-286. Sommers, S., & Ellsworth, P.C. (2003). The jury and race: how much do we really know about race and juries? A review of social science theory and research. 78 Chicago-Kent Law Review, 78, 997-1032. 103

PAGE 109

Sorensen, J., & Wallace, D. (1999). Prosecuto rial discretion in seeking death: An analysis of racial disparity in the pretri al stage of case processing in a midwestern city. Justice Quarterly, 16 559-579. Unah, I. & Boger, J.C. (April, 2001). Prelim inary report: Race and the death penalty in North CarolinaAn empirical anal ysis: 1993-1997. Retrieved October 2, 2004 from Commonsense Foundation of North Carolina site: U.S. General Accounting Office (1990). Death penalty sentenci ng: Resource indicates pattern of racial disparities. Washington, D.C.: U. S. General Accounting Office. Vetter, B. (2004). Habeas, Section 1983, and post-conviction access to DNA evidence. University of Chicago Law Review, 71, 587. Wolfgang, Marvin and Riedel, Mark (1973). Race, judicial disc retion, and the death penalty. Annals of American Academy of Political and Social Science. 407, 119133. Zeisel, H. (1981). Race bias in the admini stration of the death penalty: The Florida experience. Harvard Law Review 95 451-468. Ziemba-Davis, M. & Myers, B.L. (2002) The application of Indianas capital sentencing law: A report to Governor Frank OBannon and the Indiana General Assembly. (January 10, 2002). Retrieved March 12, 2004 from Indiana Criminal Justice Institute site: Statutes Title 43, United States Code, Section 1983 (2003). North Carolina General Statut es, Section 14A-2000 (2004). 104

PAGE 110

Caselaw Albright v. Olivier, 510 U.S. 266 (1994). Arlington Heights v. Metropolitan Housing Dev. Corp., 455 U.S. 252 (1977). Batson v. Kentucky, 476 U.S. 79 (1986). Bracy v. Gramley, 510 U.S. 899 (1997). Castaneda v. Partida, 430 U.S. 482 (1977). City of Sacramento v. Lewis, 523 U.S. 833 (1998). Enmund v. Florida 458 U.S. 782 (1982). Furman v. Georgia, 408 U.S. 238 (1972). Gomillion v. Lightfoot, 346 U.S. 399 (1960). Graham v. Connor 490 U.S. 386(1989). Gregg v. Georgia, 428 U.S. 153 (1976). Griswold v. Connecticut, 381 U.S. 479 (1967). Irvin v. Dowd, 366 U.S. 717 (1961). Jurek v. Texas, 428 U.S. 262 (1976). Kent v. Dulles, 357 U.S. 116 (1958). Marshall v. Loftin (1992). McCleskey v. Kemp, 481 U.S. 279 (1987) McKoy v. North Carolina 479 U.S. 433 (1990). Miller-El v. Cockrell, 537 U.S. 322 (2003). Nelson v. Campbell 541 U.S. 637 (2004). Personnel Administrator of Massachusetts v. Feeney 442 U.S. 256 (1979). Preiser v. Rodriguez, 411 U.S. 475 (1973). 105

PAGE 111

Proffitt v. Florida, 428 U.S. 242 (1976). Ristaino v. Ross, 424 U.S. 589 (1976). Roberts v. Louisiana, 428 U.S. 325 (1976). Rochin v. California, 342 U.S. 165 (1952). Roe v. Wade, 410 U.S. 113 (1973). State v. Cobb (1995). Taylor v. Louisiana, 419 U.S. 522 (1975). Tison v. Arizona, 482 U.S. 921 (1987). Turner v. Fouche, 396 U.S. 346 (1970). Turner v. Murray, 476 U.S. 28 (1986). Washington v. Davis 426 U.S. 229 1976). Washington v. Glucksburg, 521 U.S. 702 (1997). Whitus v. Georgia 385 U.S. 545 (1967). Woodson v. North Carolina, 428 U.S. 280 (1976). Yick Wo v. Hopkins 118 U.S. 356 (1886). 106