McCleskey v. Kemp Petitioners Reply Brief

    J

No. 84-6811 St<4dr
IN THE

SUPREME COURT OF THE UNITED STATES 
October Term, 1985

WARREN McCLESKEY,
Petitioner, 
-  v . -

RALPH M. KEMP, Superintendent, \
Georgia Diagnostic & Classification Center,.

On Writ of Certiorari to the United States 
Court of Appeals for the Eleventh Circuit

PETITIONER'S REPLY BRIEF

JULIUS L. CHAMBERS 
JAMES M. NABRIT, III 
*JOHN CHARLES 30GER 
DEVAL L. PATRICK 
VIVIAN 3ERGER

99 Hudson Street
New York, New York 10013
(212) 219-1900

ROBERT H. STROUP
141 Walton Street 
Atlanta, Georgia 30303

TIMOTHY K. FORD
600 Pioneer 3uilding 
Seattle, Washington 98104

G. AMSTERDAM 
York University 
ool of Law 
Washington Square South 
York, New York 1Q01L «.

f NEY?ri!CnP=FETTT~E ONER
cMd *

0*4



TABLE OF CONTENTS

Page
ARGUMENT

I. There Is No "Death Penalty 
Exception" To The Equal 
Protection Clause, And 
The Court Should Firmly 
Decline Respondent's 
Invitation To Insulate 
Capital Sentencing 
Systems From Claims Of 
Racial Discrimination ..... 3

II. None Of The Purported
Deficiencies Identified 
By Respondent Or Found 
By The District Court 
Impeach The Basic Show­
ing Of Racial Discrim­
ination Made By
Petitioner ................  14
A. The Reliability of the

Data Base ..............  16
(i) Data Sources ...... 16
(ii) Questionnaire

Design ........  19
(iii) Purported Coding

Errors ........  23
(iv) Purported Mis­

treatment of
"Unknowns" .... 2 6

i



B. The Effect of Multi-
collinearity ...........  30

C. The "Direct Rebuttal
Evidence" ...............  32

D. The Suggestion of
Contrary Findings....... 35

CONCLUSION .......................  42

ii



TABLE OF AUTHORITIES
Cases Page

Batson v. Kentucky,_U.S.__,
90 L . Ed . 2d 69 (1986) .........  3,4,8

Bazemore v. Friday,_U.S.__,
92 L . Ed . 2d 315 (1986)...... 10,16,33

Detroit Police Officers Ass'n 
v. Young, 608 F.2d 671 
( 6th Cir . 1979 )...............  11

Gregg v. Georgia, 428 U.S. 153
(1976)  5,6,13

Keyes v. School District No. 1,
413 U.S. 189 (1973)   8

McGautha v. California, 402 U.S.
183 ( 1971)....................  5

Segar v. Smith, 738 F.2d 1249
(D.C. Cir. 1984)..............  11

State v. Andrews, 574 P.2d 709
(Utah 1971)   23

Trout v. Lehman, 702 F.2d 1094
(D.C. Cir. 1983)..............  11,33

Village of Arlington Heights v.
Metropolitan Housing Development 
Corp., 429 U.S. 252.(1977).... 4,7

Vuyanich v. Republic Nat11 Bank,
505 F. Supp. 224 (N.D. Texas
1980).........................  9,11

iii



Washington v. Davis, 426 U.S.229
(1976)........................  4

Statutes
Ga. Code Ann. §27-2537 (Supp. 1975) 6
Ga. Code Ann. §27-2537(e)

(Supp. 1975)   6

Other Authorities:
Fisher, Multiple Regression in Legal 

Proceedings, 80 Colum. L. Rev.
702- (1980)....................  9,31

Gross, Race and Death: The Judicial 
Evaluation of Evidence of 
Discrimination in Capital 
Sentencing, 18 U.C. Davis L.
Rev. 1275 ( 1985 )...............  31

Kleck, Life Support for Ailing
Hypotheses: Modes of Summariz­
ing the Evidence for Racial 
Discrimination in Sentencing,
9 Law and Human Behavior 271 
(1985)........................  41

Note, 33 Stan. L.Rev. 75
(1980)  38,39,40

United States Dep1t of Justice,
National Prisoner Statistics,
Capital Punishment 1930-1970
(Aug. 1971)...................  42

iv



No. 84-6811
IN THE

SUPREME COURT OF THE UNITED STATES 
October Term, 1985

WARREN McCLESKEY,
Petitioner,

-  v . -

RALPH M. KEMP, Superintendent, 
Georgia Diagnostic & 
Classification Center.

On Writ of Certiorari to the 
United States Court of Appeals for 

the Eleventh Circuit

PETITIONER'S REPLY BRIEF

The State of Georgia stakes its case 
largely on two propositions: first that 
capital cases are so unique that any 
statistical analysis of capital 
sentencing patterns is impossible as a



matter of law (Resp. Br. 7-16)1. an^ 
second, that certain methodological 
flaws in Professor Baldus's analysis of 
the Georgia data undermine his findings. 
(Resp. Br. 16-23; 31-36). We deal with 
these points in Parts I and II 
respectively. Neither impugns 
petitioner's strong evidence of racial 
discrimination, which continues to stand 
unmet and unrebutted by respondent, as 
it did during the 1983 federal hearing.* 2

*Each reference to the brief for 
Respondent filed in the Court will be 
indicated by the abbreviation "Resp. 
Br.," followed by the number of the page 
on which the reference may be found. 
References to the Brief of Amici Curiae 
State of California, etc.,et al., will 
be indicated by the abbreviation "Cal. 
Br." and to the Brief Amicus Curiae of 
the Washington Legal Foundation et al. 
by the abbreviation "WLF Br."

2We emphasize that petitioner has 
presented his case on alternative Eighth 
and Fourteenth Amendment grounds. We do 
not discuss our Eighth Amendment claim 
in this brief, since respondent's Eighth 
Amendment arguments were anticipated and 
dealt with in our opening brief . (See 
Pet. Br. 41-44; 97-104.)

2



THERE IS NO "DEATH PENALTY EXCEPTION"
TO THE EQUAL PROTECTION CLAUSE, AND 
THE COURT SHOULD FIRMLY DECLINE 
RESPONDENT'S INVITATION TO INSULATE 
CAPITAL SENTENCING SYSTEMS FROM CLAIMS 
OF RACIAL DISCRIMINATION_______________

Respondent first contends that, in a 
capital case, "[s]tatistical analyses 
are inadequate as a matter of . .
law." (Resp. Br. 7). While this
argument is couched in terms of 
"statistics," its actual purport is to 
exclude capital sentencing decisions 
entirely from the normal procedures and 
standards for proof of discrimination 
outlined in Batson v. Kentucky,
__U.S.__, 90 L.Ed.2d 69 (1986) and its
precursors. Those standards permit a 
defendant to "make out a prima facie

I

case of purposeful discrimination by
showing that the totality of the
relevant facts gives rise to an
inference of discriminatory purpose,"

3



Batson v. Kentucky, 90 L.Ed.2d at 86.
Under those standards, statistical 
evidence may sometimes "'for all 
practical purposes demonstrate 
unconstitutionally because in various 
circumstances the discrimination is very 
difficult to explain on nonracial 
grounds.1" Batson v. Kentucky, 90 
L.Ed.2d at 85, quoting Washington v. 
Davis, 426 U.S. 229, 242 (1976). See 
also Village of Arlington Heights v. 
Metropolitan Housing Development Coro., 
429 U.S. 252, 265-66 (1977).

Respondent offers three interrelated 
reasons why normal principles of Equal 
Protection jurisprudence should not 
apply in capital cases: (i) "[e]ach 
death penalty case is unique," (Resp. 
Br. 8), involving "too many unique 
factors relevant to each individual 
case," to permit any meaningful 
comparisons (id. at 6); (ii) the "myriad

4



of factors" that influence capital 
sentencing decisions makes statistical 
analysis impossible (id. at 14); and 
(iii) many of the influential factors in 
capital cases are so subjective that 
they can neither be identified nor 
quantified (id.).

Respondent's first argument echoes 
the skepticism expressed in McGautha v. 
California, 402 U.S. 183, 204-07 (1971), 
that any objective standards can be 
found or formulated to regulate the 
choice of sentence in capital cases. 
Since every case is unique, respondent 
reasons, there can be no way to compare 
the dispositions in different cases, and 
thus there can be no proof that race has 
played an impermissible role. Yet 
McGautha1s viewpoint has not stood the 
test of time: as this Court recognized 
in Gregg v. Georgia, 428 U.S. 153, 193- 
94 (1976), "the fact is that [capital

5



sentencing] standards have been 
developed . . . [and that they] provide 
guidance to the sentencing authority."3 
Respondent elsewhere argues that 
Georgia's capital system is 
constitutional precisely because its 
statutory sentencing standards guide 
jury discretion among more and less 
death-worthy cases (see Resp. Br. 23-31; 
see also, Cal. Br. 18-23; WLF Br. 5-7); 
he cannot simultaneously maintain that 
capital cases are so unique that it is 
impossible to identify meaningful

Georgia's statute requires 
appellate review of death sentences to 
determine whether each is 
"disproportionate to the penalty imposed 
in similar cases, considering both the 
crime and the defendant." Ga. Code Ann. 
§27-2537. "If the [Georgia Supreme 
Court] . . . affirms a death sentence, 
it is required to include in its 
decision reference to similar cases that 
it has taken into consideration. §27- 
2537(e) (Supp. 1975)." Gregg v. 
Georgia, 428 U.S. at 167. All of this 
is, of course, fundamentally 
inconsistent with respondent's position 
that every capital case is so unique as 
to be incomparable with others.

6



sentencing standards when cases are 
actually compared.

Respondent1s second argument
transforms a quantitative distinction 
into a qualitative one. He suggests that 
since a capital sentencing decision 
involves many more considerations than 
an employment decision, evidentiary 
tools long held to be appropriate in 
employment discrimination cases simply 
have no place in capital cases. (Resp. 
Br. 8-10; see also Cal. Br. 23-28).

This argument is wrong in law and 
fact. Governmental decisions in which a 
large variety of legitimate variables 
must be considered are subject to proof 
of discrimination by the usual 
"sensitive inquiry into such
circumstantial and direct evidence of 
intent as may be available." Village of
Arlington Heights____v. Metropolitan
Housing Development Corp., 429 U.S. 252,

7



266 (1977)(rezoning); Keyes v. School
District____No ■____ 1, 413 U.S. 189
(1973)(school location and zoning). For 
example, in no area of the law-- 
including capital sentencing —  has a 
wider or more subtle array of factors 
traditionally been deemed legitimate 
than in the exercise of peremptory 
challenges to prospective jurors. Yet 
in Batson the Court specifically held 
that traditional Equal Protection 
principles should guide lower courts in 
determining whether race has affected 
the prosecutor's decision to strike 
jurors peremptorily. Batson v. Kentucky, 
90 L.Ed.2d at 87-88.

Furthermore, perhaps the greatest 
virtue of multiple regression analysis 
is precisely its ability to sort out and 
determine the influence of one factor, 
despite the presence of a large number 
of other factors which may also exert

8



independent influence. See, e . q . ,
Fisher, Multiple Regression in Legal 
Proceedings, 80 Colum. L. Rev. 702, 704- 
06 (1980); Vuyanicn v. Republic Nat 11 
Bank, 505 F. Supp. 224, 267-69 (N.D.
Tex. 1980). Multiple regression
techniques have regularly been relied 
upon determining medical, scientific, 
and agricultural issues many of which, 
like capital cases, involve scores of 
relevant considerations. As Professors 
Baldus and Woodworth testified below, 
for example, the statistical methods 
employed in petitioner's case were also 
central to the National Halothane Study, 
through which the national medical 
community made its life-or-death 
decision on the best anaesthetic for 
general medical use. (Fed. Tr. 155-58; 
1200-02).

As his third justification for 
jettisoning normal Equal Protection

9



principles, respondent invokes an 
unnamed and apparently unknowable host 
of subjective factors that, he surmises, 
might affect capital decisionmaking. 
Respondent concludes that petitioner's 
failure —  indeed, the probable 
inability of any researcher —  to 
include all such factors in an empirical 
study necessarily renders proof of 
racial discrimination "impossible" in 
capital cases. (Resp. 3r. 14).

Again, the argument is both legally 
and scientifically unsound. Bazemore v. 
Friday specifically held that "[w]hile 
the omission of variables from a 
regression analysis may render the 
analysis less probative than it might 
otherwise be, it can hardly be said, 
absent some other infirmity, that an 
analysis which accounts for the major 
factors 'must be considered unacceptable 
as evidence of discrimination.'" 92

10



L.Ed.2d 315, 331 (1986). Accord: Seqar
v. Smith, 738 F .2d 1249, 1287 (D.C. Cir. 
1984); Trout v. Lehman, 702 F.2d 1094, 
1101-02 (D.C. Cir. 1983); Detroit Police 
Officers Ass'n v. Young, 608 F.2d 671, 
687 (6th Cir. 1979); Vuyanich v. 
Republic Nat11 Bank, 505 Supp. at 255- 
56 .

This legal holding accurately 
reflects the scientific reality. 
Information on a missing variable does 
not make a significant difference unless 
the omitted variable (i) has an 
influence on death-sentencing which is 
not "captured by" other variables 
included in the analysis, and (ii) is 
itself systematically associated with 
race. (Fed. Tr.1690-94;1265-73). Thus, 
omission of a variable such as the 
"credibility of the witnesses" (Resp. 
Br. 12) would be important only if such 
"credibility" was not correlated with

11



detailedthe detailed "strength-of-evidence" 
variables which Professor Baldus did 
include in his questionnaire (see S.E. 
23-26, questions 61-62D; .id. at 29-33),4 
and if it were plausible to believe that 
jurors regularly (for non-racial 
reasons) find witnesses more credible in 
white-victim rather than black-victim 
cases. The crucial point is that 
unidentified factors which are not 
associated with race will occur randomly 
in white-victim cases and black-victim 
cases. They will not produce, and cannot 
explain away, a systematic pattern of 
differential capital sentencing between 
racial groups. Professor Baldus, who 
controlled for over 230 relevant 
variables, undoubtedly accounted for the 
"major sentencing determinants" in the 
Georgia system. Respondent's expert

4Each reference to the Supplemental 
Exhibits will be indicated by the 
abbreviation "S.E."

12



admitted that he had no idea whether any
additional variables might make a 
difference (Fed. Tr. 1569; 1591-92); and
the eminent social scientists who have 
submitted a brief amici curiae in 
support of petitioner find it "extremely 
unlikely" that any such factor remains 
lurking in the shadows. (Brief Amici 
Curiae for Dr. Franklin M. Fisher, et
al., at 22).

* * * * 
Respondent has, in sum, suggested no 

valid reason why constitutionally-based 
standards of proof that are applicable 
in all other areas of the law should be 
abandoned when a death-sentenced inmate 
seeks to prove that "in fact the death 
penalty [is] . . . being administered
for any given class of crime in a dis­
criminatory, standardless, or rare 
fashion," Gregg v. Georgia, 428 U.S. at

13



223 (White, J., concurring in the 
judgment).

II
NONE OF THE PURPORTED DEFICIENCIES 
IDENTIFIED BY RESPONDENT OR FOUND BY 
THE DISTRICT COURT IMPEACH THE BASIC 
SHOWING OF RACIAL DISCRIMINATION 
MADE BY PETITIONER_________________

Respondent alternatively invokes 
findings of the District Court to 
contend factually: (i) that the "data
base" upon which Professor Baldus and 
his colleagues drew for their 
information was so deficient —  and
their questionnaire and data-collection 
methods so flawed —  that no valid 
statistical analyses can be based on 
their work (Resp. Br. 2; 17-19); (ii)
that "multicollinearity" distorts
Professor Baldus1s research findings 
(Resp. Br. 2-3; 20-21); (iii) that the 
State's hypothesis that "white victim

14



cases are simply more aggravated and 
less mitigated than black victim cases" 
(Resp.Br. 6) provides "direct rebuttal 
evidence to Baldus1 theory" (Resp. Br. 
4); and (iv) that contrary findings on 
the race-of-victim issue were reported 
by Professor Baldus himself (Resp. Br. 
4), and, amici contend, by other sources 
as well. (See Cal. Br. 12 n.5; WLF Br. 
18-19 & n.4.)

None of these arguments is correct; 
the record plainly refutes each one. It 
is hardly accidental that not a single 
judge of the Court of Appeals rested his 
or her opinion, even in the alternative, 
on the factual findings of the District 
Court. To the limited extent that those 
findings were not marred by legal 
error,5 they were clearly erroneous.

5 The 
incorrect 
reliance 
fact. For 
held that

District Court's use of 
legal standards precludes 
upon its findings of 
example, the District Court 
"[a]n important limitation

15



We will examine below a number of the 
more prominent findings relied upon by 
respondent.5

A. The Reliability of the Data Base
(i) Data Sources —  Respondent 

contends that the sources of data 
available to Professor Baldus and his 
colleagues were "'very summary in many

placed on the data base was the fact 
that the questionnaire could not capture 
every nuance of every case," (J.A. 136), 
and that "[mjultiple regression requires 
complete correct data to be utilized." 
(J.A. 144). The District Court also
concluded, after an extensive, somewhat 
confused discussion, entitled "What a 
Multivariate Regression Can Prove," 
(J.A. 162), that "multivariate analysis
is ill suited to provide the court with 
circumstantial evidence of the presence 
of discrimination, and it is incapable 
of providing the court with measures of 
qualitative difference in treatment 
which are necessary to a finding that a 
prima facie case has been established 
with statistical evidence." (J.A. 168- 
69)(italics omitted). All of these 
holdings are contrary to this Court's 
subsequent opinion in Bazemore v. 
Friday, 92 L.Ed.2d at 331.

®A more thorough examination of the 
District Court's findings appears in 
Appendix E to the Petition for 
Certiorari. McCleskey v. Kemp, No. 84-6811.

16



respects.1" (Resp. 3r. 18). The Parole 
Board records, respondent claims, "were 
brief and the police reports from which 
the Parole Board summaries were prepared 
were usually only two or three pages 
long." Id. The Parole Board files, the 
District Court found, provided "'no 
information about what a prosecutor felt 
about the credibility of any witnesses.
R. 1117. ' " (Resp. Br. 18) .

This characterization of the data 
sources is incorrect almost from start 
to finish. Professor Baldus drew his 
information from an extraordinary range 
of official Georgia sources on each 
case, including the full trial 
transcripts, all appellate briefs, the 
files of the Georgia Department of 
Offender Rehabilitation, the files of 
the Georgia Bureau of Vital Statistics, 
and the files of the Parole Board. (See
S. E. 43, DB 39). L.G. Ware testified

17



that the parole officials like himself 
who compile each of its files (all
college graduates) prepare a special 
report: "We check local criminal 
records, we go to the clerk of [1331] 
court, get sentence information, 
indictments, jail time affidavits, we 
get police reports from the agency that 
handled the case." (J.A. 52). In 
homicide cases "if we didn't think the 
report had all the information we 
thought we needed, we may interview the 
officers that were involved in the case
. . . [or] the district attorney." (J.A.
54). A Parole Board Manual guides
officials at every step:

[The report] should be obtained 
in narrative form. It should be 
taken from the indictment, the 
district attorney's office, the 
arresting officers, witnesses 
and the victim. A word picture 
telling what happened, when, 
why, where, how and to whom 
should be prepared. . . . Parole
officers should be as thorough 
as possible when conducting 
post-sentences on persons who

18



have received life sentences or 
sentences in excess of fifteen 
years.

(J.A. 56-57). Although the District
Court faulted the Parole Board files for 
lacking photostatic copies of the actual 
police reports in 75% of the cases (J.A. 
137), Officer Ware testified without 
contradiction that nothing "contained in 
the police reports . .. would [be]
routinely omit[ted]" from the Parole 
Board files. (J.A. 53). These files, in 
fact, were often superior to the written 
police reports since they contained the 
results of direct interviews with police 
officers and prosecutors about each 
case. (J.A. 54).

(ii) Questionnaire Design —  
Respondent also faults Professor 
Baldus's questionnaire design, alleging 
that there are "problems with the format 
of critical items on the questionnaires, 
such that there was an insufficient way

19



to account for all factors in a given 
case." (Resp. Br. 17). Respondent 
quotes the District Court that "'[a]n 
important limitation placed on the data 
base was the fact that the questionnaire 
could not capture every nuance of every 
case.1" (See J.A. 136). Once again, 
these findings and contentions are 
themselves riddled with error. Professor 
Baldus developed three questionnaires, 
two for his first study, the Procedural 
Reform Study ("PRS") and one for the 
Charging and Sentencing Study. ("CSS"). 
Any restrictions on data entry under 
the "foil entry" method criticized by 
respondent were largely limited, as 
respondent's expert conceded (Fed. Tr. 
1392), to the two PRS questionnaires. 
(Fed. Tr. 274). Professor Baldus 
testified that the "foil entry" format 
was virtually abandoned in the CSS 
questionnaire. (Fed. Tr. 1098-1101).

20



Since CSS data were used to conduct 
virtually all of the major analyses 
relied upon during the federal hearing, 
(Fed. Tr. 1437), respondent's criticism 
is essentially irrelevant.7

The District Court's complaint that 
the Baldus studies "could not capture 
every nuance of every case" not only 
reflects a legally erroneous standard of 
proof but a straightforward 
misunderstanding of the record. 
Professor Baldus testified that because 
no questionnaire could be devised to

Nevertheless, as a check on the 
impact of the use of the foil method, 
Baldus identified some 50 PRS cases in 
which there had been "overflow
information that wouldn't fit into the 
original foils." He created new foils to 
handle the overflow information and 
reran each of the analyses. The race 
effects became "somewhat intensified 
when this additional information was 
included." (Fed. Tr. 1099-1100). Baldus 
also recoded and reran analyses
involving the only two CSS items
retaining the foil method of entry that 
contained overflow information. His CSS 
findings remained identical. (Fed. Tr. 1101).

21



anticipate every possible factor, he 
constructed a special "narrative 
summary" section precisely so that 
coders could "capture every nuance of 
each case." (Fed. Tr. 239; see S.S. 
36) .8

8The brief of amici Washington 
Legal Foundation et al. displays a 
similar misunderstanding of Professor 
Baldus* 1s powerful data collection 
instrument. Amici assert that one 
gruesome Utah murder they describe 
"would have been listed by the coders as 
a shooting." (WLF Br. at 24, citing the 
District Court, see J.A. 136-37). Amici 
are mistaken. 3oth the PRS and the CSS 
questionnaires contain numerous entries 
through which the special features of 
this case could have been coded. For 
example, the CSS questionnaire provides:
(i) three entry foils, plus an "other" 
blank space, to code the method of 
killing (see S.E. 14); (ii) an entry to 
reflect a contemporaneous rape (S.E. 
7, Q. 29); (iii) 34 separate options 
(including torture, mental torture, 
unnecessary pain, victim bound or 
gagged) in a section on special 
aggravating features of the offense 
(S.E. 15, Q. 47 A & B); as well as (iv) 
extensive entries for the respective 
roles of co-perpetrators. (S.E. 16-18, 
Q. 47-48). Finally, the "Narrative 
Summary" section instructs the coder to 
include "all facts" and "any special 
circumstances that are not covered in 
the preceding questions." (S.E. 36).

22



(iii) Purported Coding Errors —  
Respondent faults the Baldus data 
because of ostensible "'errors In coding 
the questionnaire'" (Resp. Br. 2), which 
it says are reflected in "numerous 
inconsistencies between the coding for 
the Procedural Reform Study and the 
Charging and Sentencing Study." (Resp. 
Br. 18). The District Court, noting 
these "mismatches" between the PRS and 
the CSS questionnaires, concluded that 
coding errors pervaded the study. (J.A. 
137-39).

Respondent's challenge to the 
accuracy of the data entries, however,

In sum, the charge that "the 
collective horrors of [the Utah case] .
. . cannot be reduced to neatly coded
variables," (WLF Br. 24), is wide of the 
mark. Professor Baldus's questionnaires 
would have captured all relevant factors 
in the Utah case, including one factor 
amici have misstated —  the substan­
tially different involvement of the two 
Utah defendants. See State v. Andrews, 
574 F.2d 709 (Utah 1977).

23



did not entail any actual comparison of 
Baldus's questionnaires with the 
underlying files from which the data 
were drawn. Instead, respondent's 
expert simply ran a computer check on 
items from cases that were included in 
both the PRS and CSS studies. The 
expert admitted that he made no attempt 
to compare the coding "protocols" from 
the PRS and CSS studies, to see whether 
the ostensibly "mismatched" items had 
been coded according to different 
instructions. (Fed. Tr. 1447).9

9 All CSS coders were law students 
who were carefully selected (Fed. Tr. 
301-03) and extensively trained. (Fed. 
Tr. 309-11). Coding decisions were 
guided by a comprehensive written 
"protocol" developed by Professor Baldus 
(see DB 43) which contained hundreds of 
instructions on general coding issues 
and on rules for coding specific items. 
(Fed. Tr. 310-11). During the data 
collection period, a coding supervisor 
reviewed a large proportion of all 
completed questionnaires on a daily 
basis. (Fed. Tr. 401-03). Entries were
ultimately checked by computer for 
internal inconsistencies. (Fed. Tr. 
595-99) .

24



In fact, Professor Baldus testified 
that the instructions often varied 
significantly. For example, the PRS 
coders were required to draw inferences, 
if reasonable, from the file data; in 
the CSS study, coders were instructed 
not to draw inferences if information 
was not present in the file. (Fed. Tr. 
367). Respondent was eventually forced 
to concede, "I don't believe [our 
expert] is indicating either one is 
necessarily right or wrong in his 
judgment. He's just indicating he's 
done a computer count and found these 
inconsistencies." (Fed. Tr. 1444).

Professor Baldus, however, did 
conduct a broad reanalysis of the 
alleged mismatches, and reported that 
approximately one percent were 
attributable to data entry, coding, or 
key punch error:

[T]hat translates into an error 
rate of approximately one-half

25



of one percent in each of the 
two studies. However, we found 
on further examination that . .
. the error rate in the 
Procedural Reform Study was 
higher than it was in the 
Charging and Sentencing Study.

(Fed. Tr. 1719-20). Since the findings
presented to the District Court came
largely from the CSS study, the relevant
error rate was very low.
( iv) Purported____Mistreatment____ of
"Unknowns" —  Respondent poses one 
final data collection issue —  the 
number of items that were coded
"unknown" in the studies, and Professor 
Baldus's treatment of those unknowns in 
his analyses. (Resp. Br.19). Throughout 
the CSS study, Professor Baldus's coders 
were instructed to enter a "1" if a fact 
were "expressly stated in the file", a 
"2" if the fact was "suggested by the 
file but not specifically indicated," a 
blank if the fact were not present in 
the case, and a "U" if the coder could

26



not classify the item based on the file. 
(Fed. Tr. 444-45). Once statistical 
analysis began, the "U" was recoded as 
"not present."

In his testimony, Professor Balaus 
examined one aggravating variable 
that the "victim pled for his life"—  
to clarify the logic behind this 
standard coding procedure. If there had 
been witnesses present during the crime, 
he explained, a coder would code the 
variable either present or absent, 
depending on the witnesses' accounts. 
But in the absence of witnesses or other 
evidence, Balaus reasoned that one could 
draw no inference either way, and the 
item would be coded "U." (Fed. Tr. 
1685-86; see also id. 1155-58).

This explanation casts in a 
radically different light the District 
Court's ominous-looking list of 
variables coded "U" in more than ten

27



percent of the cases. (J.A. 139-41). 
Many Involve either state-of-mlnd or 
relational variables that are often 
unknown to any outside investigator. 
For example, "Defendant's Motive was 
Sex" would be important if known to a 
prosecutor or jury. If the fact could be 
neither eliminated nor confirmed from 
the evidence, however, Baldus's rule 
would be to code it "unknown" and 
ultimately discount its impact either 
way by treating it as non-existent.

The District Court challenged the 
basic logic of this coding treatment: 
"the decision to treat the 'U' factors 
as not being present in a given case 
seems highly questionable . . .  it would 
seem that the mere rational decision 
would be to treat the 'U' factors as 
being present." (J.A. 139). Yet neither 
petitioner's experts (Fed. Tr. 1684-90) 
(Baldus); Fed. Tr. 1761-63 (Berk)), nor

28



respondent's experts (Fed. Tr. 1502-04; 
(Katz); Fed. Tr. 1656-58 (Burford)) 
suggested that a "U" should be coded as 
"1" or "present" for purposes of 
analysis. Indeed, Dr. Berk, 
petitioner's rebuttal expert, testified 
that the National Academy of Science 
expressly considered this issue during 
its two-year study of sentencing 
research and endorsed the very approach 
Baldus adopted. (Fed. Tr. 1761-63). The 
District Court's conclusion that a 
contrary code should have been used is 
entirely baseless.1(̂

i°Moreover, Baldus testified that 
he conducted a series of alternative 
analyses to test the District Court's 
assumptions. (See generally Fed. Hab. 
Tr. 1693-1705 and S.E. 64-66). He 
recoded unknowns as "1" or "present" 
just as the Court had recommended. The 
effects on racial disparities "were 
within a percentage point of one another 
and all the coefficients that were 
statistically significant in the one 
analysis were in the other." (Fed. Tr. 
1701). Another alternative analysis, 
employing "list-wise deletion" of all 
cases with "U" codes —  a procedure

29



B. The Effect of Multicollinearity
Both respondent (Resp. Br. 2; 20-21) 

and the District Court (J.A. 150-53), 
decry "[m]ajor problems" presented by 
the phenomenon of "multicollinearity," 
which, they assert, invariably 
"distort[s] the regression coefficients 
in an analysis." (Resp. Br. 20).

Their concern is misguided. 
Professor Gross, in his thorough 
examination of the district and circuit
court opinions in this case, has
directly addressed the point:

Multicollinearity occurs, in the 
court's view, whenever 'there is any 
degree of interrelationship among 
the variables,' and it distorts the 
regression coefficients. . . This is
false. There is nothing in the 
assumptions of multiple regression

recommended by the State's principal 
expert, (Fed. Tr. 1501-02) —  also had 
no adverse effect upon Baldus' original 
findings. (Fed. Tr. 1695-96; see S.E. 
64). Indeed it increased the race-of- 
victim coefficient by two percentage 
points.

30



analysis that requires uncorreiated 
regressors; indeed, multiple 
regression analysis is primarily 
useful in analyzing data in which 
there are correlations among the 
predictor variables."

Gross, Race and Death:____ The Judicial
Evaluation of Evidence of Discrimination 
in Capital Sentencing, 18 U.C. Davis L. 
Rev. 1275, 1292 n.83 (1985). The social 
scientists who appear in this Court as 
amici strongly concur in this judgment. 
(Brief Amici Curiae for Dr. Franklin M. 
Fisher, et al., 25-26). See also Fisher,
Multiple_____Regression_____in_____Legal
Proceedings, 80 Colum. L. Rev. at 713. 
Moreover, the issue is not one on which 
the record is silent. Petitioner's 
experts testified without contradiction 
that the effects of multicollinearity, 
far from increasing the Baldus findings 
of racial influences in the Georgia 
system, would, if anything, tend to

31



dampen their appearance by decreasing 
their reported statistical significance. 
(Fed. Tr. 1281-82; 1782). Since 
Professor Baldus found racial 
disparities that were highly 
statistically significant despite any 
multicollinearity, the entire issue is a 
spurious concern.

C. The "Direct Rebuttal Evidence"
Apart from his attacks on Professor 

Baldus1s data sources, discussed above, 
respondent offered virtually no rebuttal 
evidence to undermine either the stark 
racial disparities found or their 
significance. His reference to "direct 
rebuttal evidence . . . that 
contradicted any prima facie case of 
system-wide discrimination, if one had 
been established" (Resp. Br. 4) is 
misleading. This reference is to an 
hypothesis, put forward by his expert at 
the federal hearing, that Georgia's

32



racial disparities might be explained by 
the fact that white-victim cases are, on 
the whole, more aggravated than black- 
victim cases, and thus that they receive 
deservedly harsher penal treatment. 
(Resp. Br. 6; see J.A. 169-70).

Respondent's hypothesis, like any 
other, might easily have been tested by 
determining whether white- and black- 
victim cases at the same levels of 
aggravation are, in fact, similarly 
treated. (Fed. Tr. 1664). Although 
respondent's expert admitted on cross- 
examination that such critical testing 
"would be desirable" (Fed. Tr. 1613), he 
chose not to undertake it. Instead, 
respondent rested his case on untested 
assumptions of precisely the sort 
condemned by the Court as inadequate in 
Bazemore v. Friday, 92 L.Ed.2d at 333 
n.14. See, e.g., Trout v. Lehman, 702 
F .2d at 1102.

33



Petitioner, however, did not permit
these assumptions to go unexamined. 
Instead, his experts addressed this 
hypothesis directly (Fed. Tr. 1297; 
1729-32;1759-61), tested it thoroughly 
(Fed. Tr. 1291-96; see GW 5, 6, 7, 8; 
see also DB92), and conclusively proved 
that racial disparities in Georgia are 
not the result of any differences in 
average aggravation levels between 
white- and black-victim cases. (Fed. Tr. 
1732). One powerful indicator of this 
finding appears in the Supplemental 
Exhibits at page 72. The different bands 
for white- and black-victim cases reveal 
that as aggravation levels rise, a 
substantial gap in the death-sentencing 
rate opens between cases at the same 
level of aggravation. Nothing in 
respondent's hypothesis addresses, much

34



less refutes, this central truth.11
D. The Suggestion of Contrary 

Findings
Respondent quotes the Court of 

Appeals for the proposition that 
Professor Baldus1s first study, the PRS, 
"revealed no race of defendant effects

^One brief submitted by amici 
speculates, despite the evidence, that 
race-of-victim disparities in Georgia 
surely could not reflect decisionmaker 
bias, since "the victim is perforce 
absent from the trial and the victim's 
race is rarely a matter of relevant 
concern at trial." (WLF Br. 4). The 
remark betrays lack of familiarity with 
the record and with the normal course of 
capital trials. As a matter of record, 
Professor Baldus1s data demonstrate that 
much of the reported racial 
discrimination occurs through the 
pretrial and presentencing decisions of 
Georgia prosecutors, who invariably know 
the race of the victims involved. As a 
matter of trial practice, moreover, it 
is the rare Georgia case where the jury 
is not exposed, during the trial itself, 
to photographs of the victim, to 
testimony from the victim's family, or 
to other clear indicators of the 
victim's race. In addition, pretrial 
exposure to newspaper accounts of 
homicides, as well as local knowledge of 
the victim among jurors in rural areas 
and small towns, often gives most jurors 
knowledge of the victim's race well 
before trial.

35



whatsoever and revealed unclear race of
victim effects." (Resp. Br. 4, citing 
J.A . 247). Several amici also suggest 
that contrary findings on the race-of- 
victim issue have been reported by other 
researchers. (See WLF Br. 4; 18- 
20)(Bureau of Justice Statistics); WLF 
Br. 18-19 n.4 (Note, 33 Stan. L. Rev. 75 
(1980); Cal. Br. 12 n.5 (Kleck, 9 Law & 
Human Behavior 271 (1985).) None of 
these assertions is accurate. Every 
researcher who has ever studied 
Georgia's post-Furman sentencing 
patterns has found a significant race- 
of-victim effect. (See the articles 
cited in petitioner's principal brief at 
page 51 n.16.)

The Court of Appeals' adverse remark 
about the race-of-victim findings in the 
Procedural Reform Study is unsupported 
by any citation and is wrong. (J.A. 
247). The record reveals that many of

36



Professor Baldus1s PRS analyses did find 
strong racial effects. For example, DB 
98, included in the Supplemental 
Exhibits at 58, reports highly 
statistically significant race-of-victim 
effects, using PRS data, for a wide 
range of statistical models, including 
5-variable, 9-variable, 61-variable, and 
164-variable models. Baldus reported and 
commented upon many other strong race- 
of-victim effects disclosed by his 
analysis of the PRS data. (See Fed. Tr. 
905-914; 917-919; 939-40; DB95; DB96.)

One amici brief suggests that 
statistics compiled by the Bureau of 
Justice Statistics of the United States 
Department of Justice, which report a 
higher death-sentencing rate for white 
defendants than for black defendants, 
"discredit petitioner's sweeping 
contention that anti-black 
discrimination permeates the capital

37



sentencing process. (WLF Br. 18). In
fact, these BJS statistics are
consistent with Professor Baldus's own
findings. For example, Baldus found
that 7 of every 100 white defendants,
but only 4 of every 100 black
defendants,, received a death sentence in
Georgia during the 1973-1979 period.
(See S.E. 46) . Upon further analysis,
however, he concluded that the
differences are not explained by any
"anti-white" bias in Georgia, but rather 
by the fact that most white defendants 
in Georgia murder other whites, while 
most black defendants murder other 
blacks. (See S.E. 47) The powerful 
influence of the victim1s race in 
Georgia death-sentencing decisions 
simply overwhelms the less powerful 
race-of-defendant effects.

Amici Washington Legal Foundation et 
al. also mention "other reputable

38



studies [that] undercut the claims of 
victim-anchored racial discrimination in 
capital sentencing." (WLF Br. 18 & 
n.4). They cite a single work, a 
student note reporting a limited 
analysis of data from a four-year period 
collected in another state. Note, 33 
Stan. L. Rev. 75 (1980). Even this 
study, however, largely replicates 
Baldus1s principal findings. The student 
found that "black offenders who killed 
whites were convicted of first degree 
murder about four times more often than 
black offenders who killed blacks," id. 
at 87, and that such defendants received 
death sentences nearly seven times as 
often. Id. While the student asserted 
that "the inference of discrimination 
collapses" when the analysis is 
restricted to felony-related murder 
cases, id.at 88, his data actually 
reveal the following death-sentencing

39



5 of 61 8%

4 of 52 8%

1 of 26 4%

felony-related
homicides:

Black kills white 
White kills white 
Black kills black 
White kills black 0 of 3 0%

Id. at 89, Table 4. Although the small 
number of felony-related murder cases 
involved precludes a statistically 
significant finding, the pattern of 
results supports Professor Baldus's
claims.

Other amici refer the Court to 
Professor Kleck's article for "a recent, 
objective review of some of these
studies and conclusions." (Id.) 
Professor Kleck's article is indeed 
instructive; it concludes that while 
most hypotheses of racial discrimination 
in the criminal justice system are 
overstated, prior research does support 
the following conclusions:

rates among all

40



(1) The death penalty has not 
generally been Imposed for 
murder in a fashion
discriminatory toward blacks, 
except in the South, (emphasis 
added) . . .

* * *

(5) There appears to be a 
general pattern of less severe 
punishments of crimes with black 
victims than those with white 
victims, especially in
connection with the imposition 
of the death penalty.

Kleck, Life____Support for Ailing
Hypotheses: Modes of Summarizing the
Evidence for Racial Discrimination in
Sentencing, 9 Law & Human Behavior 271, 
272 (1985). Professor Kleck thus 
directly ratifies the principal results 
reported in this case, and reaffirms 
Professor Baldus1s observation that the 
"triangulation" of research findings 
provides one fundamental reason for 
believing "that there are real race 
effects operating in the charging and 
sentencing system in this state." (J.A. 
48) .

41



CONCLUSION
The history of the adminis­

tration of the death penalty in Georgia 
is a history marred by racial discri­
mination. Over 81% of all those 
executed beween 1930 and 1970 were 
black, (see United States Department of 
Justice, Capital Punishment 1930-1970 
at 13), just as 6 out of 7 executed in 
the post-Furman period —  under 
Georgia's revised capital statutes-- 
have been blacks whose victims were 
white. (Brief for the Congressional 
Black Caucus, et al., as Amici Curiae, 
at 5.) Although respondent continues to 
insist that Georgia's post-Furman system 
is "functioning as it was intended to 
function," (Resp. Br. 5), Professor 
Baldus has amply demonstrated the

42



existence of strong race-of-victim 
disparities, as well as race-of- 
defendant disparities against blacks 
whose victims are white. This 
discrimination occurs exactly where it 
might have been predicted —  among the 
"midrange" of moderately aggravated 
cases, where petitioner McCleskey's own 
case is found.12

The Eighth and Fourteenth Amendments 
surely require no more of petitioner 
than this evidence, which renders it 
"more likely than not" that racial

12Respondent quarrels with this 
mid-range analysis —  hypothesizing that 
"different rankings" could be given to 
the cases "depending on what variables 
might be included in a particular 
regression." (Resp. 3r. 22). Yet 
predictably, respondent has offered 
no analysis in which Georgia racial 
results are different. Respondent's 
expert, who spent over 1000 hours prior 
to the federal hearing reanalyzing the 
Baldus data (Fed. Tr. 1576) never 
uncovered any defensible model or any 
set of variables that could explain, 
or even diminish significantly, the role 
played by race as a determinant in the 
Georgia capital system.

43



discrimination has been at work in
Georgia's capital sentencing system 
during the 1973-1979 period. The State's 
demand for still further proof is 
certainly not, at this juncture, a 
legitimate plea for more careful 
examination. It is instead a heedless 
request that Georgia be permitted to 
continue its age-old capital sentencing 
practices —  despite the facts, despite 
the law, despite the Constitution.

Amici have contended that it would 
be "repugnant to any decent sense of law 
and justice" for a capital inmate to 
"escape an otherwise valid death 
sentence by invoking the race of his 
victim." (WLF Br. 2). That's not what 
this case is about. The real issue is 
whether petitioner and other Georgia 
inmates have received their death 
sentences in part because of the race of 
their victims. Decency, law, and

44



* t ♦

justice are properly invoiced to guard 
against such a possibility, not to 
condone it.

The Court should reverse the 
judgment of the Court of Appeals.
Dated: October 3, 1986

Respectfully submitted,

JULIUS L. CHAMBERS 
JAMES M. NABRIT, II 
*JOHN CHARLES BOGER 
DEVAL L. PATRICK 
VIVIAN BERGER

99 Hudson Street
New York, New York 10013
(212) 219-1900

ROBERT H. STROUP
141 Walton Street 
Atlanta, Georgia 30303

TIMOTHY K. FORD
600 Pioneer Building 
Seattle, Washington 98104

ANTHONY G. AMSTERDAM
New York University 
School of Law
40 Washington Square South 
New York, New York 10012

♦Attorney of Record
ATTORNEY^ FOR PETITIONER/ll 0 \BY: l̂ .v ‘M  V.

45