Folder
Deposition of Dr. Joseph L. Katz
Public Court Documents
June 30, 1983 - July 1, 1983
301 pages
Cite this item
-
Case Files, McCleskey Legal Records. Deposition of Dr. Joseph L. Katz, 1983. ba5eb077-5aa7-ef11-8a69-7c1e5266b018. LDF Archives, Thurgood Marshall Institute. https://ldfrecollection.org/archives/archives-search/archives-item/fe30b9c6-5b12-4a83-be35-2d2a576172fb/deposition-of-dr-joseph-l-katz. Accessed December 06, 2025.
Copied!
ase Biles: | le vl Y (03) ‘Legal :N.D (a. No Cvl- Z HZHA Depo nN: Ko tz, Dr. Jos Pr « Jun. 20-dul. | 13%
3}
N
IN THE UNITED STATES DISTRICT COURT
NORTHERN DISTRICT OF GEORGIA
ATLANTA DIVISION
WARREN McCLESKEY,
Petitioner,
CIVIL ACTION FILE
VS.
NUMBER: C81-2434A
WALTER D. ZANT, Superintendent,
Georgia Diagnostic & Classification
Center,
Respondent.
VOLUME TI
DEPOSITION OF DR. JOSEPH L. KATZ taken at the instance of
the Petitioner at 132 State Judicial Building, Atlanta, Georgia
before Earlene P. Stewart, Certified Court Reporter and Notary
Public, on the 30th ‘day of June, 1983, at 3:00 p.m.
APPEARANCE OF COUNSEL
For the Petitioner: For the Respondent:
JOHN CHARLES BOGER MARY BETH WESTMORELAND
Attorney at Law Assistant Attorney General
NAACP Legal Defense and 132 Judicial Building
Educational Fund Atlanta, Georgia 30334
10 Columbus Circle
New York, New York 10019
Also Present
Robert H. Stroup, Esq. Professor David C. Baldus
Co-Counsel Samuel P. Laufer
EARLENE P. STEWART
CERTIFIED COURT REPORTER
4334 GREENVALE DRIVE
DECATUR, GEORGIA 30034
(404) 981-2311
VOLUME I
CONTENTS
WITNESS DIRECT CROSS REDIRECT RECROSS
DR. JOSEPH 'L. KATZ
By Mr. Boger 6
EXHIB 1'T:S
Marked Identified Page
Petitioner's Exhibit 1 Three Page Notice of 5
Deposition
“Petitioner's Exhibit 2 Analysis of the Pro- 8
cedural Reform Study and
the Georgia Charging and
Sentencing Study
Petitioner's Exhibit 3 VITA of Dr. Katz 10
“Petitioner's Exhibit 4 Continuation of Prelimi- 97
nary Report "Analysis of
the Georgia Charging and
Sentencing Study"
*It was requested by Counsel that the above exhibits not be
attached to their copies of the deposition.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
1 DEPOSIT TON
A 2 MR. BOGER: We're here for the deposition of
3 Professor Joseph Katz this afternoon in the case of McCleskey
4 versus Zant, and before we get started, I would like to ask
5 Ms. Westmoreland if the usual stipulations that apply in
5 these kinds of depositions will be applied today, that all
o objections except to the form of the question are reserved
8 for the time of the hearing?
9 MS. WESTMORELAND: That's correct.
io MR. BOGER: One other thing I want to note for the
record -- Ms. Westmoreland, correct me if I'm mistaken -- I
believe that in the conduct of this deposition, we have agreed
that the following ground rules should apply: that Dr. Katz
13
® 14 has extensive documents connected with this case, that you've
5 indicated to me that it will be difficult, although not
16 impossible, to bring them from Dr. Katz's office to the room
17 where we're holding the deposition, but that Dr. Katz would
6 bring among the most important documents for this deposition
9 || 2S well as certain ones that I've indicated to you previously
z that I was confident we would want to see; that we would be
ot free to ask Dr. Katz about any other relevant documents, and
2 if those documents were unavailable in this room, we could
23 adjourn after this session this afternoon of the deposition
2d to Dr. Katz's office and review the documents there at his
N 28 office or any documents that we've identified that were useful
EARLENE P. STEWART
CERTIFIED COURT REPORTER
=
~
for us for the deposition or to be copied, we could either
have those documents brought here tomorrow, or if they were
too voluminous for some reason, maybe schedule the deposition
closer to Dr. Katz's office or make some arrangements so that
we could have the use of them during the deposition; and that
you would make available for copying either through your own
facilities or through giving us a short loan of any documents
we'd want from Dr. Katz, and that we'd have them copied some-
time in the near future, perhaps, after the July llth cut-off,
although we'd hope earlier, actually, to accommodate Dr. Katz
in his use of them, but also obviously to accommodate us in
preparing for the hearing. And when we get to Syracuse, the
same kind of courtesies would be extended with respect to
documents that Profesor Baldus has; is that our understanding
of the ground rules?
MS. WESTMORELAND: I think that's our understanding,
Just to go ahead and state it in regards to Professor Baldus,
Dr. Katz and I are going to give you at least a tentative
list of documents that we know that we would like to have at
the deposition. And as your list was not exclusive, our list
also is not meant to be exclusive. And we also understand tha
the questionnaires would be available for us to at least look
through on our own, perhaps, one night while we were in
Syracuse. And I believe that was our understanding on the
telephone as well.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
MR. BOGER: What I told you just to be sure, though
is that the record reflects Professor Baldus's questionnaires
are in storage, in effect, in the basement of the Law School
at Syracuse in boxes. And my understanding is that there is
construction going on around them, but they are accessible in
that kind of form. And I don't think the Dr. or Professor
Baldus has any control any more over the law school, but my
impression is there's access to it until some reasonable
hours in the evening. Okay.
MS. WESTMORELAND: One other thing while we're
discussing this as well. In regard to the list that you had
given me previously, Dr. Katz and I have been through it, and
to the best of our understanding of what you want, we have
managed to obtain, I think, everything that you have requested
and do have two boxes full of printouts and documents here
that are an attempt to comply with the request. And we'll
see if they do.
MR. BOGER: All right. Madam Reporter, I'd
like to have you mark as Petitioner's Exhibit 1 for identifi-
cation a three page notice of deposition in this case.
(Whereupon, Petitioner's
Exhibit 1 was marked
for identification.)
Whereupon,
DR. JOSEPH L. KATZ
EARLENE P. STEWART
CERTIFIED COURT REPORTER
was called as a witness and, having been duly
sworn, was examined and testified as follows:
DIRECT EXAMINATION
BY MR. BOGER:
Q Dr. Katz, would you state your name and address for
the record?
A I'm Dr. Joseph L. Katz, and I live at 1741 Smyrna
Roswell Road.in Smyrna, Georgia.
Q Dr. Katz, you're aware that you're appearing here
today in the deposition being taken in the Federal Case of
Warren McCleskey versus Walter Zant; is that correct?
A Yes.
Q And you've spoken with the Attorney General's
Office about the nature of these proceedings in general?
A Yes.
Q Let me ask you to take Petitioner's Exhibit 1 for
identification and look through it briefly.
MS. WESTMORELAND: While he's looking through that,
could we just note for the record the people present at the
deposition? I should have done that a minute ago. I don't
know if the Court Reporter got that or not, but I just wanted
to =~
MR. BOGER: Fine; I'll be glad to. I'm Jack Boger,
taking the deposition with N.A.A.C.P. Legal Defense Fund. Wit
me is Co-counsel Robert Stroup from Atlanta. We have with us
EARLENE P. STEWART
CERTIFIED COURT REPORTER
a summer law student from my offices -- actually a pre-law
student -- Samuel Laufer from New York, and Professor David
C. Baldus who is an expert witness for the Petitioner in this
case is here. For the Respondent, it appears to be Mary Beth
Westmoreland and Susan Boleyn from the Attorney General's
Office. And Dr. Katz has just given us his name and address.
MS. WESTMORELAND: Thank you.
BY MR. BOGER:
Q Dr. Katz, have you reviewed this Petitioner's
Exhibit 17
A Yes.
Q And have you previously seen this document?
A Previously seen this document?
Q Yes.
A No.
Q This document is a Notice of Deposition noticing
that your deposition was going to be taken today. Had the
Attorney General's Office informed you about that prior to this
time?
A Pardon me?
Q That there would be a deposition this afternoon?
A Yes, I'm aware of that.
Q Okay; that's why you're here, obviously.
A The Notice at the bottom of the first page has a
paragraph that begins, "PLEASE TAKE FURTHER NOTICE that,
EARLENE P. STEWART
CERTIFIED COURT REPORTER
pursuant to Rule 45(b) of the Federal Rules of Civil Pro-
cedure, said deponent is requested to bring with him all
documents upon which the deponent relies or refers to in
formulating, substantiating or explaining his analyses or
opinions, but not including those documents supplied by
petitioner to respondent.’
Had the Attorney General informed you of this
request for documents?
A Yes.
Q And if you would, tell us specifically what you
brought with you this afternoon in response to this request.
A I brought only two boxes of about ten or fifteen
boxes of computer printouts that I have used in analyzing
the data. Specifically, this first set of computer printouts
has to do with the specific request on the data that was
generated on a certain table. Let's see -- that would be
Tables XI and XII.
MR. BOGER: Let me ask the Court Reporter to mark
as a Petitioner's second exhibit a copy of the document
entitled "Analysis of the Procedural Reform Study and the
Georgia Charging and Sentencing Study by Dr. Joseph L. Katz,
Preliminary Report."
(Whereupon, Petitioner's
Exhibit 2 was marked
for identification.)
EARLENE P. STEWART
CERTIFIED COURT REPORTER
1 THE WITNESS: These documents have to do with the
es » (| multivariate testing that: 1 —-
3 MR. BOGER: Dr. Katz, I'm going to give you this
4 document -- let me explain what I'm doing. You were kind enough
5 to provide us with a copy of your report. It was unnumbered,
6 and we've taken the liberty of numbering the document and I
2 hope that we're going to refer to those numbers throughout
8 this deposition so that everybody is clear about the pages
o we're all talking about. So, if you would, I'll ask you to
30 identify that document and tell me if you recognize it?
THE WITNESS: Yes.
BY MR. BOGER:
Q What is that document?
The page number?
14
16 Q Just for the record, I want you to note what it is.
" A It's the preliminary report that I completed.
7 Q All right. Now, if you would, go back to that
8 first table that you mentioned and just give us a page
5 reference for that so --
os A Okay; Pages 32 and 33, 34, 35, and that's it.
21 Q In other words, those pages contain the table with
3 reference to which you brought these documents; computer
bo printouts today?
2 A I generated those tables based on the information
25 contained in these first documents.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
10
Q All right. And would you continue with your
description of what else you brought?
A Okay. This set of computer printouts is all the
multivariate testing that I've done in trying to analyze the
data for both studies.
Q In other words, you've done no additional runs on
the multivariate?
A I have not run any original regressions or other
multivariate. tests.
Q All right. TI think in terms of my proceeding most
clearly, it would be helpful, not at this point though, to
get into what other documents or studies you've generated or
runs you've generated, but simply to go through, if we can
quickly, what you brought so that we'll have on the record
what's here. And then I'll ask you some questions about what
you've done and what you've not done in addition to that.
A Okay. Now, these are, I believe, the regressions
that you requested where I included the unknowns for the
specific, two specific regressions that are mentioned in the
preliminary report. That would be Pages 50 -- starting 51,
52, 53,“and that ‘should do it.
Q All right.
A Okay; these are printouts of unknowns in the LDF
file pursuant to the table -- is this all the Report that you -
Q Let me tell you that we didn't have time and our
EARLENE P. STEWART
CERTIFIED COURT REPORTER
11
Xerox machine broke. The second part of your report, the
continuation, we didn't get numbered. So, you can simply
refer to the table numbers if there are tables in the second
part.
A All right. In the second part of the report that
is for the table of unknowns, which is Table XLII.
Q All right; ves.
A This contains printouts of all the variables for
these cases that are common to both studies plus some programs
comparing the data for common variables between the two studie
Q Okay.
A This is the computer printout that gave me the
information to generate the information on Table VIII, compari
the set of variables between the race of the victim, including
all the cases in the first study.
Q And that Table VII is on page what, for our reference
later?
A Page --
Q Is that 23?
A Yes.
Q Thank you, Dr. Katz.
A Okay. These are some additional printouts that
helped me generate this program. Okay; that takes care of
the first box.
Q What is this large pile that you've just put on
EARLENE P. STEWART
CERTIFIED COURT REPORTER
ng
12
the table?
A Well, I couldn't generate this information directly
Q When you say this, let's be sure on the record we
know what you're talking about.
A For Table VIII.
Q All right. ;
A So, I had to run this program and then Sods Sty
in a new file that I could gather from this program to help
generate the results for Table VIII.
Q For Table VIII; okay, thank you.
MS. WESTMORELAND: It might be simpler if I place
these back in the box. at this time.
MR. BOGER: Let me suggest, if you wouldn't mind
Ms. Westmoreland, if we put them down at the other end of the
table.
MS. WESTMORELAND: We can go off the record for
just a minute.
(Whereupon, there was a brief
interruption of the record.)
THE WITNESS: Okay. Now, starting with the second
box, this is my preliminary report. It's not obviously the
most recent copy but close to that. These are documents that
were sent to me by Mr. Baldus; similar documents sent by Mr.
Baldus.
Okay. This is some analysis I did on one of the
EARLENE P. STEWART
CERTIFIED COURT REPORTER
13
experiments done by Mr. Baldus in the first study, the Parole
Board questionnaires.
Q Yes. Then, if you would clarify, or do the
documents themselves clarify which one of the analyses you
performed -- which one of the parts of Dr. Baldus's Study you
performed that analysis on?
A This is an updated analysis of the fourth part of
his first study where he tries to or he categorizes the cases
according to seven factor levels.
Q Okay; thank you.
A There are some additional things that I've gene-
rated to help me understand the data in both studies. Do you
want specifics?
Q No. I'll just note for the record these look like
handwritten documents with numbers and some notations on them
on lined notebook paper.
A These are more of my analysis of a working draft
that I recently received from Mr. Baldus. This is a printout
of all the variables in the Parole Board Questionnaire Study
for all the cases.
Q Would you clarify for the record, when you say
Parole Board Questionnaire Study, let me make a suggestion
that I'm going to try to follow myself throughout the depositig
and these hearings. We can call the first study the Procedural]
Reform Study, as you've done in your preliminary report, and
EARLENE P. STEWART
CERTIFIED COURT REPORTER
DT]
14
call the second one the Charging and Sentencing Study. I'm
afraid that the Court might get confused if we start referring
to them by different names. So, this is a printout of variabl
in the --
A Procedural Reform Study.
Q Thank you.
A Okay. This is a printiout of the variables and
indexes and scales generated by Mr. Baldus from the second
study, the Georgia Charging and Sentencing Study, which
is ‘a task file.
This is a printout of all the variables for all the
cases in the Georgia Charging and Sentencing Study that is
from the master file of the original data off the question-
naires.
And here is a printout which generates most of the
tables for my work for the Georgia Charging and Sentencing
Study.
Q That's a printout of what, did you say?
A Of information that I used to generate most of
the tables that I used for my testing.
Q What kind of information 40 you mean?
A To generate, I'll give you an example of the table,
table numbers. I believe this starts with Table XLIV and runs
through most of the other tables succeeding Table XLIV.
Q In other words, it doesn't relate to --
EARLENE P. STEWART
CERTIFIED COURT REPORTER
eS
A To just one table.
Q And it doesn't relate to prior, tables prior to
A I don't believe so.
Q All right; thank you. That's some of the documents
you've brought with you this afternoon?
A Yes.
Q Now, you have other documents in your office; is
that correct?
A Yes.
Q I think we're going to ask you a little later in
this deposition to generally describe some of those documents,
and we're going to ask you specifically about certain cate-
gories of additional documents we may be interested in. My
understanding with the Attorney General's Office, and I'd
like to confirm it with you on the record, is that you'll
make those documents available to us at some point after
this deposition this afternoon so that we could use them if
we need them tomorrow morning; is that correct?
A Yes.
Q What I'd like to do now, however, is to go over
with you a little biographical information so that we can
have that for the record. Let me just note for the record,
Professor Baldus would like to proceed to the end of the
room and look through these documents that you brought for
EARLENE P. STEWART
CERTIFIED COURT REPORTER
16
us, Dr. Katz. That's what he's going to be doing for the next
few minutes while we talk about your background.
MS. WESTMORELAND: That's fine; I have no objection
to that. Can we go off the record just for a moment before
we start, please.
(Whereupon, there was a brief
interruption of the record.)
MR. BOGER: .Back on the record. Dr. Katz, how late
is your office open this evening in case we need to be there
after hours?
THE WITNESS: At least until 11:00 tonight.
MR. BOGER: Thank you. Madam Reporter, I'd like
to have you mark as Petitioner's Exhibit 3 for identification
a three page document entitled "VITA".
(Whereupon, Petitioner's
Exhibit 3 was marked
for identification.)
BY MR. BOGER:
Q Dr. Katz, I'd like to show you the document that's
just been marked Petitioner's 3 and ask you to examine it for
a moment. Can you identify that document for us?
A Yes, this is a copy of my current Vita.
Q Have you examined it enough to know whether this
is your current Vita at this point?
A There is one small change.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
17
Q What's that, sir?
A My marital status; I'm single.
Q All right. We will not get into questions in that
area this afternoon. Is your Vita accurate, to your best
knowledge, other than that change you've mentioned?
A Yes.
Q I'd like to proceed rather quickly to your :educatios
background, but let me ask, first, where were you born?
A I was born in Montreal, Canada.
Q And when did you come to the United States?
A When I was seventeen, 1969, to study at Louisiana
State University.
Q Where did you do your growing up; in Montreal?
A In Montreal, Canada.
Q And what kind of preparatory school did you attend?
A I attended a private elementary school and a public
high school.
Q Did you receive a high school diploma?
A Yes, I did.
Q And what year was that?
A 1969.
Q Was there any special track that you were going in
the schools in Montreal, or was it just a general high school
diploma?
A It was a general high school diploma.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
nal
18
Q You entered, you said, Louisiana State University
in 19697?
A Yes.
Q And what was your major?
A Mathematics.
Q Let me ask you -- and I know this is a difficult
thing to recall specifically -- but can you give us all of the
mathematical courses that you took as an undergraduate; if you
can remember. And let me, before you answer, ask you if you
can remember the professors as well?
A Okay. I took calculus, which was Math 50, under
Professor John Keisler, and I took Math 50 honors. And I took
Math 51 honors, which was the second part of the calculus unde
the same Dr. John Keisler. And then I took Math 57, which was
multi-dimensional calculus, under Al Horn. And then I took
linear algebra and matrices, which was Math 85, I believe.
The number originally was 57, but they changed it. I think
when I took it it was Math 85. And that was under a -- I
forget her first name -- Cs Then I took Math 131, which
was the first part of real analysis, and Math 132 which was
the second part of real analysis.
Q Do you remember who was the professor there?
A The Professor in the first and second part ‘was a Di.
David P. Roselle. Now, these other courses, I don't think
I'll remember them in the order that I took them but --
EARLENE P. STEWART
CERTIFIED COURT REPORTER
a
19
Q Were the courses you've just given us in sort of
sequential order in which you took them while at LSU?
A Yes. But these later courses, I took, I'm not
sure if I completely remember when. There was Math 153, which
dimensional vector spaces;
is finite dimension; a Math 136, complex variables; Math 122,
which was abstract algebra. Then, there were two or three
Math 199's which were directed readings courses. And those
courses --
Q Who were those under the direction of, do you recal
A They were under Dr. Brooks Reid. And in those
LombD | -—~
directed readings courses, I studied graph theory in cemmen
parts. I also took a Course 145, whichwas a numerical analysi
Q And do you remember who the professor was there?
“7
Lavin,
A oh EAE
Q Do you recall any additional courses as an under-
graduate in mathematics?
A As an undergraduate?
Q Let me clarify from your response, all of the
courses that you've mentioned thus far have been undergraduate
courses?
A Yes.
Q Do you recall any further mathematics courses?
A Well, when I was in computer science, there were
some mathematics courses that were taught under the computer
science numbers.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
1?
20
Q Well, now, I'm going to ask you in just a minute
about this feat you achieved to having a B.S. one year in
mathematics and the next year in computer science. But still
focusing on your mathematics B.S., does that complete the list
of courses you recall?
A That's all I remember.
Q All right. Do you remember any courses in
statistics?
A As an undergraduate?
Q As an undergraduate.
A Yes.
Q And what courses did you have?
A I had an engineering statistics course --
Q Under?
A Under Lawrence Mann; it was in industrial engi-
neering.
Q All right. Any other statistical courses?
A That was my introduction to statistics.
Q And did you have any courses as an undergraduate
in multivariate analysis or regression analysis?
A As a undergraduate, no.
Q Any course in research design, mathematical researc]
design, statistical research design?
A There was one other course as an undergraduate,
and that was an operations research course.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
21
Q And who was that taught by?
A William Thompson in the Quantitative Methods
Department.
Q Let me ask you just generally about other related
areas; that is, areas related to this litigation. Did you
take any courses in economics?
A Yes.
Q What courses did you take?
A Economics 55, which was a combination of macro and
micro economics, in one semester.
V ©
) Any other courses in economics?
No.
Did you take any courses in sociology?
Yes, one course.
What was that?
P
D
ri
dO
Sociology 51.
Which was what?
Beginning sociology.
Was there any quantitative component to that course?
No.
Did you take any courses in psychology?
No.
Any courses in political science?
A
>
OO
I
©
i
SE
H
©
No.
Po
) Any courses in law or legal studies?
EARLENE P. STEWART
CERTIFIED COURT REPORTER
22
A No.
Q Any courses in criminal justice?
A No.
Q Any courses in empirical research?
A No; other than those directed readings courses.
Q Let's clarify a little more about that independent
study, your directed reading. What were you doing in those
courses?
A Part of the time I was learning graph theory. Ther
was one course where it was formally taught as a graph theory
course. The other time I was working with Dr. Reid on his
graph theory problems, his research problems. And so, I was
using the computer to help him. What I would do is, he would
devise a problem and try to get some insight into what the
problem was dealing with. He would ask me to use the computer
to print out some numbers to give him some feel for what was
going on.
Q Well, I'm not familiar with graph theory. Let me
just ask you, is that some theoretical mathematical area or
is it an empirical mathematical area; what are you doing in
that area?
A It's a theoretical mathematical area.applying --
there's been a lot of applications, though -- but my interest
at that point wasn't in the applications.
Q Let me just summarize if I could or ask you, was
EARLENE P. STEWART
CERTIFIED COURT REPORTER
D
23
your undergraduate mathematical training more theoretical or
more empirically oriented?
A More theoretical.
Q Did you have any substantial empirical training at
that point?
A No.
Q Apart from the directed readings that you've
mentioned, did you do any other kind of independent study in
the area of mathematics or the area of statistics?
A No.
Q Did you do any other work apart from course-
work for employment in this area as an undergraduate?
A I did tutor undergraduate mathematics courses
around my junior year. So, it must have been my third year.
Q What kind of courses were: they?
A Well, for people taking the basic calculus courses.
Q Did you do any student papers or other kinds of
publications or written work that was not published but was
circulated in the professional community as an undergraduate?
A Well, as a result of my work with Dr. Reid, he
did publish papers on his results. But I didn't get any
formal credit for the work. All I did was give him an insight
into the problem.
Q You've answered my next question. You didn't have
any formal co-authorships, although you had worked with Dr.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
24
Reid on thisgraph theory work?
A Right. He acknowledged to me personally that I
had helped a great deal in the solution of one of his research
problems.
Q Did you have any: ‘other undergraduate exposure to
issues involving research design; questionnaire construction;
data collection in empirical work?
A No .
Q Okay. Your resume indicates you graduated with a
B.S. Degree in mathematics in '72 and got a B.S. Degree from
LSU the next year in computer science; is that right?
A That's correct.
Q What academic requirements does LSU impose on
someone to permit them to receive two bachelor's degrees a
year apart?
A Well, after they get the first degree, I believe
you need to complete at least thirty hours, semester hours,
A atta all the requirements for the second degree.
Q And that's what you did?
A Yes.
Q Tell me what courses you took to satisfy these
thirty hours of computer science requirement?
A I took some of those statistics courses that I
mentioned earlier, and I continued with the basic computer
courses that I had been taking; the undergraduate computer
EARLENE P. STEWART
CERTIFIED COURT REPORTER
courses.
Q Let me ask a few questions. Then I assume, perhaps,
your computer work all came during this --
A No.
Q What computer courses did you take as an under-
graduate? Now, let's move back. What computer courses did
you take as an undergraduate before your 1972 degree?
A I took computer Science 51, which was an intro-
duction to computer science; Computer Science 52, which was
assembler language.
Q Can you spell it for us just so we won't have it
wrong in the record?
A It's basic assembler language. The assembler is
a-s—s-e-m-b-l-e~-r.
Q Thank you. Do you recall any other computer
courses as an undergraduate?
A Well, one course was Computer Science 80, which
was computer hardware. Another was Computer Science -- I'm
not sure of this number -- my best guess is Computer Science
57, which was data structures, discrete structures. Then I
took -- I think that may be all I took within the first
three years before I graduated in 1972.
Q And then you mentioned you took at least thirty
hours of additional work to receive your Bachelor of Science
in Computer Science. What courses did you take during that peri
EARLENE P. STEWART
CERTIFIED COURT REPORTER od?
26
A I took Computer Science 101, which was programming
languages; Computer Science 103, which was software, software
development; Computer Science 151, which was compiler con-
struction. Let's see; the Computer Science 57 was discrete
structures; and then there was a Computer Science 104 -- the
numbers are getting foggy -- which was on data structures.
And I think I took my directed readings with Dr. Reid, one of
them during that time.
Q In other words, these are additional readings
beyond those that you took for your B.S. in mathematics
where you did the graph theory?
A Right; they were under the computer science number.
Q Dr. Reid sounds as if he was one of the principal
professors that took an interest in your work or in whose work
you took an interest; is that correct?
A Yes.
Q Were there any other professors who were instru-
mental in your education during this period?
A Yes, Dr. Roselle.
Q Now, you mentioned earlier that you took a basic
statistics course, an undergraduate. And I think I asked you
whether you took any further statistical courses and your
answer was no; is that my recollection?
A As an undergraduate, I think I took two courses
that might qualify. The first is the industrial engineering
EARLENE P. STEWART
CERTIFIED COURT REPORTER
10
1
13
14
15
16
17
18
19
20
21
22
23
24
25
statistics course. The second is the operations research
Se |g
course,which was taught by the Quantitative Methods Department
which has statistics involved with it.
0 Did you use a statistical text during that course;
do you recall?
A It's a text that used statistical techniques. A lot
of operations research is just to find statistical techniques
to different types of problems.
Q Do you remember that text?
A Yes, Levin and Kilpatrick were the authors, and the
title was "Quantitative Techniques in Management." They have
eA i LioNS
some further addits out; I took the third or fourth
edition.
Q Did you do any other work apart from your course
work in this ‘directed reading during that year that you
were involved in computer science in the area of mathematics
or statistics or computers?
A Other than what was required in course work?
Q That's right.
A No.
Q And during that year, did you take any subjects
outside the area of computer science?
A I don't remember.
0 Let me ask you more specifically, have you taken
any courses in economics or sociology or psychology or politic
EARLENE P. STEWART
CERTIFIED COURT REPORTER
28
science, law, criminal justice; areas I asked you about earlier
A During that period?
Q During that period.
A Yes, I took some Latin readings courses.
Q I'm sorry; Latin ~~
A I think I studied Tacitus as a summer --
Q 1 see; okay.
A I'm not very: good at it, but 1've tried to be.
0 As far as you can recollect, that's all outside
that area during this period of time?
A Yes. Now, my economics course, though, I'm not
sure when I took that.
Q Dr. Katz, did you attain any academic honors during
your undergraduate career?
A No.
Q Did you have any scholarships or fellowships?
A No.
Q Did you join any professional or other kind of --
A I joined a Professional Computer Science Fraternity,
which was just an honorary fraternity.
Q What's the name of that?
A I don't remember the Letters.
Q All right. What was the basis for selection for
that organization?
A The students had to have a certain grade point
EARLENE P. STEWART
CERTIFIED COURT REPORTER
Ps
29
average in computer science and overall.
Q Do you remember what the grade point averages were?
A I think it was a 3.0 average overall and a 3.0 or
3.5 in computer science major.
Q Did you graduate with honors in either of your B.S.
programs?
A No.
Q Dr. Katz, what did you do upon completion of your
B.S. in computer science in 1973?
A I went to graduate school and pursued a master's
degree in mathematics at Louisiana State University.
Q Let me ask you before we get into that masters
program work, during any of the summers after you began to
attend college, did you do any outside employment that related
to mathematics or computer science?
A No.
Q Or statistical work?
A No.
Q Let me before we begin the discussion of your master
program, how would you describe your areas of academic interes
as an undergraduate if you were asked, as I'm asking you, what
were your areas of academic interest, what's your answer? At
one level I suppose you'll say mathematics and computer
science; is that correct?
A Yes.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
30
Q Within the areas of mathematics and computer science,
did you have any specific interest or concerns?
A No.
Q All right. And you began your masters program at
LSU -- what is the nature of that program at LSU?
A It's a theoretical mathematics program.
Q All right. And how many people are in the Depart-
ment as professors?
A Well, at the time‘'l was there, it was over forty.
Q And how many students were enrolled in the graduate
program; if you remember?
A Approximately twenty or thirty in the graduate
program.
Q All right. You said there were forty professors?
A Right.
Q That's a good faculty-student ratio. Were there
specialties or sub-specialties within the Department?
A Yes.
Q And what were those specialtivs?
A One of the specialties was algebraic typology.
Another specialty was functional analysis. I would say those
are probably the two major areas that they were known for.
Then, there were people who were in combinatorics, graph
theory, who had reputations. There you had a big department
covering all kinds of different areas in mathematics.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
31
0 What specialties did you pursue in your master's
program?
A It would be combinatorics and the graph theory,
but that's where my interests were.
MS. WESTMORELAND: You might spell that for the
Court Reporter, I think.
THE WITNESS: Combinatorics; c-o-m-b-i-n-a-t-o-r-i-c:
BY MR. BOGER:
Q You've listed those as your two areas of interest?
A Of interest; yes.
Q What kind of course work requirements were there
for your master's degree?
A For the master's degree, algebra, analysis, and
typolo
gy ¥ t (&] Fe { O 9 y \'
Q Did you take courses in all of those?
A Yes,
Q Under what professors; do you know?
A Okay; my first semester I took real analysis, Math
165, under -- I don't remember the name.
Q That's all right if you can't remember it.
A I took algebra, which was Math 161, under -- this
is more recent, but I can't remember their names.
Q This was undergraduate --
Pree K Oo WiTZ . TOPOS Sol
A I remember; Bruce Breekoewitz. Then I took .
which was Math 164, under Dr. Madison.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
32
Q All right. Those were the only three courses that
you took?
A Yes; and I taught undergraduate mathematics at the
same time.
Q Okay. Well, let me just clarify for the record, the]
you didn't take any course in your master's work on multi-
variate analysis?
A In the first semester.
Q In the first semester; I see. I thought you had
given me all the courses for your --
A Three courses for a master's degree?
Q I don't know. Let's go onto -—-
A You want to go onto the next ones?
Q Why don't we do that.
A Okay. The next semester I took Math 162, which was
the second part of algebra under -- this was theoretical
Pe eK OWITZ.
algebra -- under Dr. Bxeeckowitz. And I took Math 264, which
t |
+opoiogy
was the second part of OLE under Dr. Madison. And then
took Math 265, which was measure theory, under the same
guy I couldn't remember for Math 165. And then for the summer
I took a course in combinatorics under Dr. Gordon Paul. And
the second course --
0) If you can't remember it, that's fine.
A Okay.
Q Let me ask you this: does the course, again, have
EARLENE P. STEWART
CERTIFIED COURT REPORTER
33
anything to do with multivariate analysis?
A It may.
Q I'll give you another moment, then, if your recoll-
ection can come back; if it doesn't, then, we'll go on.
A Well, I'll just give it to you in the next -- I
know what the course is but --
Q Well, if you can describe even what the course was
about, that's all I need.
A The course was mathematical statistical theory, well,
probability in statistical theory, Math 155.
Q Who was the Professor?
A Timothy Etem. Eaton.
Q And what did you cover in that course?
A We covered distributions, theory of distributions,
random variables, sums and products of random variables; some
other probability topics.
Q Since the two courses you mentioned in your under-
graduate career, industrial statistics and the other one,
is this, then, the next course that you've had that looked at
multivariate analysis and statistical theory; is that corrects:
A No.
0 What else did you have that I've missed?
A Before that I had a quantitative methods course
which was =--
Q Let me say before that, at what time did you have
EARLENE P. STEWART
CERTIFIED COURT REPORTER
34
the quantitative methods course in your career?
A It may have been the spring before the summer that
I took the probability and statistical theory course.
Q Okay. And who was that taught by, do you remember?
A Dr. Carolyn Hargrave; and it was research methods I,
Q Do you remember what text you used.
Fazer
A We used Pesser and Swanson's Statistical Analysis.
I'm not certain about the name of the text.
Q Just for my clarification, is that course in
empirical research methods?
A Yes.
Q All right. And then you've mentioned this course
in the summer -- I think you've given me the name of the
Professor -- did you have a text in that course?
A Yes.
Q What was that, if you remember?
DwAss
A The author's name is Gloss, and it was a title like
Probability Statistics.
Q After your summer work, did you have any additional
course work for completing your master's requirement?
A Well, TIT missed one year of math courses.
Q I'm sorry; I mean that first year you were taking
math courses?
A Well, I gave you the first year, and then the second
year we jumped to the second semester in the summer.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
35
1 Q Okay; let's go through the rest. As you can tell,
p “ I'm trying to find out is the full summary of all the courses
3 that you took in mathematics or computer science or in
4 statistical methods.
5 A Okay. The second year I took complex variables,
6 which was Math 266.
7 Q And who was the Professor; do you recall?
8 A I don't remember his name.
9 Q All right.
10 A I took algebraic typology, which was Dr. Weintraub.
11 Then I took Algebra 261, which was taught by -- I don't
12 remember his name.
13 Q Now, does that complete the description of your
®
14 course work? You've described, I think, two years and one
15 summer ?
16 A No. Then there's Algebra 262 which I took under
LAX
17 Robert laeks, who also taught me 261.
18 Q All right. Now, any additional courses you
19 remember?
20 A That's as much as I remember.
21 Q Were there any other requirements for completion
22 of the master's degree at LSU?
23 A Yes, there was a thesis.
24 Q What was your thesis?
% 25 A I've got it somewhere here.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
10
1
36
Q I didn't recall seeing your master's thesis. You
do have a dissertation reflecting --
A L've omitted it. I think the title is, "Basic
+3
| = 4 ge -Cri TIC
Results in Amtrieritieal Graphs."
Q This was a thesis in the graph theory where you
had done your work, undergraduate, in your directed study?
A Yes.
Q Who did you do that thesis under or with; did you
have just a thesis committee?
A Yes.
Q Who were they?
A My advisor was Dr. Linda Lesniak, and my --
Q Excuse me. Can we get that name spelled for the
Court Reporter, if you know?
A L-e-s-n-i-a-k.
Q Thank you.
CASTLER; .
A Dr. Kassler; Dr. Keisler; Dr. Luther Wade; and I
don't remember the last one.
Q All right. Your Resume or Vita indicates that you
were a Graduate Assistant in the Department of Mathematics
during this time; is that correct?
A Correct.
Q And what were your responsibilities as a Graduate
Assistant?
A Well, my responsibilities were teaching undergraduat
EARLENE P. STEWART
CERTIFIED COURT REPORTER
11
%
37
math courses.
Q What kind of courses did you teach?
A They were basic algebra courses; remedial algebra
courses; some trigonometry courses.
Q Anything else?
A Some math for teachers; some geometry courses.
Q You mean for teachers, you mean for like high school
teachers?
A Education majors, right.
Q Anything else that you recall?
A No.
Q Did you do this work under some professors, or
were you the instructor?
A I was the instructor.
Q Were there any other written papers that you authored
or co-authored during this period of time within your
professional discipline?
A No.
Q And did you have any fellowships or grants during
your masters?
A I had an assistantship.
Q In other words, the teaching position was itself
a remunerative position; nothing else?
A No.
Q Were there any kind of honors or other academic
EARLENE P. STEWART
CERTIFIED COURT REPORTER
38
conferments bestowed on you during this period of masters
work?
A No.
Q And after you obtained your masters degree, Dr.
Katz, what did you do then?
A I then pursued a doctorate degree in the Department
of Quantitative Methods.
Q How large is the Department of Quantitative Methods
at LSU?
A I believe at the time they had about nine or ten
faculty members.
Q And how many graduate students?
A About fifteen or twenty; something in that range.
Q Did this represent a shift of interest or a shift
in specialization for you?
A Yes.
Q In what respect?
A Well, my interests, which were originally combina-
torics and graph theory, find a lot of applications in statisti
and operations research. And there was no future in pursuing
mathematics.
Q As a pure discipline?
A As a pure, theoretical discipline. So, I always
had a great interest in statistics and operations research,
so I pursued the courses in quantitative methods. pP
EARLENE P. STEWART
CERTIFIED COURT REPORTER
CS
39
Q All right. Let's be clear, for me at least, for
N / the record. Before you entered the Department of Quantitative
Methods for your Ph.D. work, I believe I recall your testi-
mony was that you had had an industrial statistics course and
your other undergraduate statistics course. You talked about
6 your quantitative methods course --
7 A Research.
8 Q -- one spring, and then another course in probability
9 theory runs in the summer?
10 A Yes.
11 Q Anything else in this area prior to your Ph.D.
& 13 A I think that covers it.
14 Q Are there any areas of specialization within the
15 || quantitative methods Ph.D. program at LSU? Can you subdivide
16 the Quantitative Methods Department into several specialties
17 or course areas?
18 A You mean like what their expertise is?
19 Q Well, I know so little about it that I didn't know
20 || whether there were major branches of work within quantitative
21 || methods that are recognized as --
22 A Well, there are two main areas in quantitative methods.
23 || One is the statistical area; the other is the operations
24 || research area.
® 25 Q All right. And are both of those represented in the
EARLENE P. STEWART
CERTIFIED COURT REPORTER
40
1 Department at LSU?
) A Yes.
3 Q Did one as a graduate student find oneself more
A in one area than another, or did you take courses in both
5 areas?
6 A I took courses in both areas.
, Q Let's talk about your course work, then, in the
, P.hD. Program. What courses did you take?
9 A Okay. The first semester I was there, I took
10 Statistical Theory II, which was QM -- I don't remember the
number. But that was the second part of the probability
and statistics that was taught in the QM Department.
12
1 Q Let's just clarify for the record. By QM, it's a
ia shorthand for Quantitative Methods?
os A Quantitative methods, yes. And since I'd had a
vh comparable course in the Math Department, I took the advanced
47 statistical methods or statistical theory in probability
” theory course.
ie Q Who was the professor there?
20 A The professor was Dr. Roger Burford.
#0 Q And what was the text; do you recall?
Gray bi if Bo es
os A The text was Mood, Gzabile and Bewes-, ''Probability
3 Statistics." They all seem to have that title; something
like that.
24
ra os MS. WESTMORELAND: Would you spell Dr. Burford's
EARLENE P. STEWART
CERTIFIED COURT REPORTER
41
name for the Court Reporter, please.
THE WITNESS: Yes. It's B-u-r-f-o-r-d.
BY MR. BOGER:
Q And that first semester, what other courses did
you take?
A I took Introduction to Operations Research, which
was Quantitative Methods 7001. The first course was
Quantitative Methods 7000, and they renumbered all the courses.
Q And who was that course under?
A Dr. John.Pisa.
MR. BOGER: I think we'll need that spelled as well
I'm sorry.
THE WITNESS: P-i-s-a.
BY MR. BOGER:
Q Any other course work?
A And there was another course, Accounting 3001.
Q And that was under who?
A I don't remember.
Q Is that the extent of your course work for the
first semester?
A Yes.
Q And then your second semester, what were your
courses and your professors?
A The second semester I took QM 7002, which was
Operations Research II, which dealt mostly with dynamic
EARLENE P. STEWART
CERTIFIED COURT REPORTER
Db
42
inventory
1 programming and empleyee theory,
2 Q Who was that professor?
3 A Again, Dr, John Pisa. And TIT believe I had an
Le rEe™S
4 economics course under Dr. Lawrence Scott; micro-economics.
MAR X
5 And then TI think I had QM 7025 under Dr. Maxks, which was
6 the second part of the research design course. The first
: part I had in graduate school in the math department.
3 Q All right. And the first part had been under
9 whom?
10 A The first part had been under Dr. Carolyn Hargrave.
1 Q Right. And the second part under --
MARK
7 A Dr. Don Marks. And then we moved to the summer,
» and I believe I took a course in macro-economics and a
14 directed readings course.
ie Q And who gave you the macro-economics course?
near
16 A Dr. Thompson; and it concentrated on non-lineal
17 programming.
18 Q Okay. And your directed reading was with whom?
7 A Dr. Thompson.
9 Q I'm sorry; I asked you who gave you your economics cdurse?
A Dr. Beard.
21
2 Q Okay. That was your summer's work; any other
on courses at that time?
ws A I don't believe so; just two.
Q Now, you started a second year's worth of course 25 y
EARLENE P. STEWART
CERTIFIED COURT REPORTER
43
work; is that correct?
A Correct.
Q And what were the courses you had then?
A QM 7003, which was Theory of Linear Programming;
Finance 7001, which was Introduction of Finance.
Q Did you have any other quantitative methods course
that semester, do you recall, or operations research?
A I'm not sure.
Q And then I guess you had a spring semester that
year?
A Yes.
Q And what courses do you recall you took?
A One course was simulation.
Q Under whom?
A Under Dr. Burford. Another course was --
Q Was that a general introduction to simulations or
did it have some specific focus to it?
A It was just a general simulation. The second course
was Sampling Theory and Design.
Q And with whom was that?
A Dr.Burford:
Q Do you recall the text in that course?
A We used more than one text. One text that we used
was Hansen, Hurwitz and Madow; probably some sampling theory
Kish
course. And the second text was by Kiseh on Sampling Theory.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
44
Q Is this just, so I'm clear, that's the second course
you had with Dr. Burford; is that correct? You had one before
A I had an earlier course; I had the second part of
Statistical Theory with Dr. Burford. Then, another course,
the third course for the first semester, second year, was
Stochastic Processes, which I had with Dr. Burford.
MS. WESTMORELAND: You might spell that for the
Court Reporter, please.
THE WITNESS: Stochastic, s-t-o-c-h-g-s-t-i-c.
BY MR. BOGER:
Q And if you wouldn't mind just very briefly
explaining that one to me?
A Well, that's a course in which processes are
modeled, processes that vary over time, according to some
probability. And the course deals with different models
to model those processes.
Q And what kind of text did you use for that course;
do you recall?
Cin lar's Ed
A We used Semlaxs Stochastic Processes', I believe
was the title.
Q Now, we've advanced, I think, through two years of
your course work in the Ph.D. Did you have any additional
course work beyond that spring semester?
A Well, there was one other course in the spring which
was a course in Micro-economics Theory, another economics cours
EARLENE P. STEWART
CERTIFIED COURT REPORTER
18
%
45
Q And who was that with?
A That was with Alden Tades, loevs.
Q All right. Beyond that of the spring semester of
your second year, is there additional course work that you
pursued for your Ph.D.?
A I'm not sure. If there is any additional work,
it was during that summer after the second year.
Q All right. What else did you do to complete
requirements at LSU for a Ph.D.?
A I wrote a dissertation.
Q How long did that consume?
A That took one year after I had finished my course
work.
Q So, beginning sometime in the spring of that year
after your second year of course work you worked --
A Well, during the second year of my course work
while in the Stochastic Processes course taught by Dr.
Burford, he was doing some research on Insénsitivity of
Leontief Multipliers and --
MR. BOGER: Excuse me for a minute; I just want to
be sure that the Court Reporter has that word down, and I
think that's in your Vita. All right, I'm sorry; if you'll
continue,
THE WITNESS: And he presented some of his research
and what he was pursuing, and I became interested in it at
EARLENE P. STEWART
CERTIFIED COURT REPORTER
46
that time and at that point started working closely with
Dr. Burford on this area. So, by the time I did finish my
coursework, I had completed, I had most of my results for my
dissertation already completed based on the work that I had
done with Dr. Burford.
BY MR, BOGER:
q Who was your Dissertation Committee?
A Dr. Roger Burford was the Chairman; Dr. Carolyn
Hargraves; Dr, William Thompson, Dr. Don. Hecke and Dr.
LOENS .
Alden Taves-
Q Those are all professors at LSU?
A Yes.
Q Apart from your dissertation, which you listed
in your Vita, did you complete any other written work during
this period?
A Papers?
Q Papers; that sort of thing.
A Yes.
Q Are they all reflected in your Vita?
A Yes, ‘I think so.
Q Do take the time to look through it.
A Yes, I think they are.
Q Now, during your Ph.D. work, were you on any
fellowships or scholarships?
A I was on an assistantship.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
47
Q Again, ‘that's a teaching position?
A Yes.
Q Your Vita reflects -- well, perhaps, let me ask you:
what was the nature of your assistantship; what were your
responsibilities?
A Teaching,
Q Teaching undergraduates or graduates?
A Undergraduates,
Q And what were the courses?
A They were the basic statistics courses for business
aM
undergraduates. There were two courses, Haeang 2001 and 2002.
The first course is just basic statistics; the second course
is basic operations research as applied to business.
Q Do you recall the text you used in those courses?
A They varied.
Q What about the statistics course, would it vary
in that?
A Yes, they change those books pretty much every
year. One -- in the first year, the basic statistics book
was Harnett and Murphy, Basic Statistics. One year they
used Dr. Burford's book.
Q Which book do you mean?
A His Basic Statistics book.
Q All right. Were you responsible for a choice of
texts here, or was that something that the Department
EARLENE P. STEWART
CERTIFIED COURT REPORTER
dictated?
A That was & Department matter.
Q This was a two year instructorship; is that --
A Yes.
Q And I note that for two years on your Vita it
appears that you were a graduate assistant, and then you
became an instructor.
A Yes.
Q What's the difference between those titles?
A Well, as a graduate assistant, I got $4,000.00 for
two courses; as an instructor, I got $15,000.00 for four
courses.
Q Were there any differences in responsibilities;
were they all teaching courses?
A Simply teaching; yes.
Q Were the instructorship courses the same ones
that you've ‘reflected when you were a graduate assistant;
that's basic statistic courses for undergraduate and basic
quantitative --
A Yes.
Q I'm sorry, you didn't say quantitative methods, you
said --
A Operations research.
Q Yes. What were your operations research texts; did
they vary as well?
EARLENE P. STEWART
CERTIFIED COURT REPORTER
49
A Well,yes. The first text was Bierman, Bonini and
Hausman, "Application -- I'm not sure what the title is. The
second one was ''Management Science'' by Anderson.
Q Again, were those chosen by the Department. or by
you?
A By the Department.
Q I believe we're about to finish the graduate work,
but let me just ask one or two additional questions. Are
there any honors that you received in your Ph.D. work, awards
or other professional recognition?
A No.
Q During your graduate courses or studies, masters
or Ph.D., were there any courses that we've not discussed
or identified in the area of sociology?
A No.
Q Or political science?
A No.
Q Law?
A No.
Q Criminal justice?
A No.
Q Any other courses that gave you training and
research design beyond those we've mentioned?
A Beyond those formal courses?
Q Formal courses.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
50
A Yes, there is one course that I didn't take for
credit, which I eusht—te—have- dene,
Q What was that?
A Multivariate Analysis,
Q With whom?
A Dr. Burford.
Q In what year?
A I think it was in, like, while I was an instructor.
Q So, that would be during the '77, '78 year?
A Yes,
Q Do you recall what text Dr. Burford used?
A I think it's Multivariate Analysis by Loones; I'm
not sure.
MR. BOGER: Do you know how to spell that?
THE WITNESS: L-o0-o0o-n-e-s; I'm not sure of the
title.
BY MR. BOGER:
Q Are there any other courses that involved training
in data collection or data analysis beyond what we've just
discussed in the earlier --
A As formal courses; no.
Q Let's ask about informal courses. Are there any
other forms of instruction = that you got other than talking
with people that dealt with these areas?
A No.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
51
1 Q Let me ask you even more generally, are there any
& 2 other courses upon which you could say that you have drawn or
3 relied in conducting your analysis of Professor Baldus's Study?
4 Obviously, one level, that would include every course in
: English taken, but I mean, courses that are in the broad
5 professional discipline that you've had to employ to conduct
your analysis here.
3 A I can't think of any others right now.
o Q And one last question in this area: have you ever
10 done any course work that involved analysis of the issue of
racial discrimination?
A No.
12
i Q Any course work that ever involved analysis of the
" ia issue of capital punishment?
A No.
15
i Q Have you ever done any course work that dealt in
3 empirical research in criminal justice matters?
8 MR. BOGER: If there's no objection, I'd like to
‘9 go off the record and take a short break.
25 MS. WESTMORELAND: I don't think any of us will
24 have an objection.
(Whereupon, there was a brief
on interruption of the record.)
a MR. BOGER: We're back on the record. Dr. Katz,
® he after receipt of your Ph,D, degree, what course did you pursuef
EARLENE P. STEWART
CERTIFIED COURT REPORTER
52
THE WITNESS: I remained an additional year at
Louisiana State University,
MR. BOGER: In what capacity?
THE WITNESS: As an Assistant Professor for just a
one year appointment.
BY MR. BOGER:
Q And what were your responsibilities?
A Teaching and research.
Q What teaching assignments did you have?
A Both undergraduate and graduate courses.
Q Can you tell us what courses you taught that year?
A I taught Basic Statistics and Operations Research.
Q What text did you use?
A I think we still used the Anderson text for
Operations Research, and I don't remember what for the Basic
Statistics. One of the texts that I mentioned earlier, I
suspect.
Q Again, did you have any choice of texts on these,
or was this something that the Department stipulated or
recommended; do you recall?
A No, I don't think I had any choice.
Q Okay; and was that an undergraduate course?
A Yes.
Q What other course: did you teach that year?
A Then I taught the QM 7001, which was the first
EARLENE P. STEWART
CERTIFIED COURT REPORTER
53
Operations Research course.
Q Is that an undergraduate course?
A No, that's a graduate course. Then I taught QM 7003
5 al
§) Lk . ~ fC f BOY
TC rOGraim imine t DCO
which is the Linear Preogram-and Theory Course.
Q Any others besides those?
A I believe those were the only graduate courses at
that time.
Q And you mentioned you had done research during this
one year period. What was your research on?
A On Input-Output Models; a continuation of the
kind of work I was doing for my dissertation.
Q And did you work with anyone or collaborate with
anyone on that research?
A Yes, with Dr. Burford.
Q Did that research result in any written document
papers?
A I believe we made some presentations; I don't think
we got anything published in journals during that year.
Q All right. Would the presentations be reflected
in your Vita of the List of Presentations?
A Yes.
Q Let me ask you generally, I guess, before I turn
away from your LSU experience, could you describe or categori
what your specialty was --
MS. WESTMORELAND: Excuse me, please.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
~
Ze
54
(Whereupon, there was a brief
interruption of the record.)
BY MR. BOGER:
Q What I was asking you, Dr. Katz, back on the
record, is would you have any description of the area of
your concentration in graduate work, Ph.D. work?
A The degree in Quantitative Methods is generally a
degree to use quantitative techniques in applications of
business problems. And so, for that purpose, you get a strong
background in statistics and operations research, but it's
not strongly theoretical. They're not preparing you for work
in abstract statistical theory. It's more -- you do get
statistical theory courses, but it's not designed to go to
high levels of abstraction. It's not mathematical statistics.
Q So, your undergraduate work had been more abstract
than theoretical?
A Yes.
Q And I guess your early graduate work, your masters
work, but then you switched in your Ph.D. Program to a more
applied approach, and you said it was largely directed toward
business systems?
A Yes.
Q All right. Now, during this year you were at LSU
as an Assistant Professor, apart from your teaching, apart
from your research, did you have any consultant's jobs or
EARLENE P. STEWART
CERTIFIED COURT REPORTER
35
other outside professional employment?
A I didn't have any employment in which I was
compensated for directly,
Q Can you explain, did you have any uncompensated
employment?
A Well, I wouldn't formally call it employment, but in
the area of consulting, I've been very close with Dr. Burford
for my first year in graduate school. And he has --
Q Let me clarify that. That's your first year in
your master's program or your Ph.D. work?
A My Ph.D. work.
Q Okay.
A And we have talked at length over many areas, many
application areas, because he has a consulting firm, and many
problems arise based on his work. And so, I wouldn't want to
say that I had no outside activities, because formally I didn'
But I still talked with Dr. Burford who -- I wasn't com-
pensated for this -- but as part of our research, we had
talked about many different aspects of application in
statistics.
Q All right. So, in other words, he had a consulting
firm, and a lot of commercial clients, and that can rate a
great many theoretical and applied problems?
A Yes; and he likes to talk.
Q And so, you and he would discuss particular problems
EARLENE P. STEWART
CERTIFIED COURT REPORTER
56
he was having, individual cases if they call them that, or
consultations?
A Yes, There were times when I would work on,
formally on some of these projects, and be compensated for
that. But most of the time, he would seek my opinion on
certain things. Some of the things he was doing was a direct
application of our research, and so --
Q Would there be any easy way to indicate or recall
the kinds of consultant jobs you had where there was some
compensation?
A A particular case was a case where Greyhound
Bus Lines wanted to open a route from, I believe, Atlanta to
Dallas. And to do that, they had to go in front of some
regulatory board to get the permits to run their line. At
that point, Continental Trailways was the:only bus service
between Atlanta and Dallas along that route. And Dr. Burford
was hired by Greyhound to help develop a case that the public
would be better served by Greyhound having a line between
Atlanta and Dallas. And he hired me to collect data, run
some regressions, to help determine what the demand would be
for that line; those kinds of things.
Q What kind of data did you collect?
A Mostly census type data for the affected county
areas.
Q Did that require you to engage in research design,
EARLENE P. STEWART
CERTIFIED COURT REPORTER
57
or was there already design to the collection of this data
that followed?
A Well, it was more trying to determine what variablegd
seemed to be significant in determining sales. And what we
did was looked at cities and towns that supposedly were
comparable to cities and towns on the line and try to forecast
what -- and Greyhond had the sales numbers for those cities
and towns; I think it's in dollars or the number of passengersg.
And we were trying to determine what variables were signi-
ficant in explaining why certain towns would be more likely
to use the bus service than other towns.
Q Did that involve you in actual data collection?
A Yes.
Q How did you collect data?
A I went to the LSU Library and pulled out these huge
census books and looked up the appropriate towns and had a
list of variables and considered some other variables, and
recorded the data.
Q So, you did actual data recordation?
A Yes.
Q But you created some sort of an instrument for this
recordation?
A It was simply certain variables for the different
counties or I think the earliest. -- I guess at that point
the best census data we had was 1970.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
58
Q S50, your source of data was the 1970 census?
A Yes.
Q And you'd do something like look at Mobile, Alabama
and find out how many households there were with incomes over
certain amounts; that kind of thing?
A Those kinds of areas; right.
Q Now, what year was that, if you remember?
A I think that was 1977.
Q And was that your first compensated consultant work?
A Yes.
Q Did you have any other major consultant jobs during
the period you were in graduate school?
A You mean, other than just what I had worked out
with Burford on a casual basis?
Q That's right.
A No.
Q Or during that one year you were at LSU as an
instructor?
A No.
Q Now, after you left LSU, Dr. Katz, you went where?
A To the University of Arizona at Tucson.
Q And what was your title there?
A I was an Assistant Professor.
Q In what department?
A I started off in the Department of Management.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
59
Q And then where did you go?
A The next year I moved to the Department of Manage-
ment Information Systems.
Q Let's back up a minute and let me ask you what the
Department of Management is, how big a department is it, and
what's its focus?
A The Department of Management at that time had the
Statistics and Operations Research courses.
Q Was this for the undergraduate and graduate?
A For the undergraduate; and they had the graduate
statistics courses.
Q How long were you in that department?
A I was in that department for one year. And then
the courses were switched over to the Management Information
Systems Department. So, two of us who seemed to, according
to the Dean, be more applicable for the MIS Department were
transferred along with the courses.
Q Why don't you tell me the difference between Manage-
ment Information System and Management. What did the Dean
have in mind when he said you were more applicable to Manage-
ment Information Systems Department?
A Well, his view of the Management Department was that
it should be a strictly usual basic kind of management
department dealing with the areas of personnel, I think
business law fell in that category; organizational behavior;
EARLENE P. STEWART
CERTIFIED COURT REPORTER
60
| those kinds of topics, where Management Information Systems
5 already contained, computer courses plus some operations
3 research courses. So, we felt it would combine all the
a courses under one department, and the Management Information
3 Systems was a more applicable department.
5 Q Was that a newly created department?
; A No, it had been established. There was a newly
8 created dean who came in and reorganized.
i Q What courses did you teach your first year in
5 management or in the Management Department?
4 A My first year I taught basic statistics courses,
is a graduate statistics course for masters in business admini-
5 stration, and a graduate operations research course for masters
® ia in business administration,
» Q For the basic statistics course, what text did you
is use; do you recall?
he A Yes, it was by George Summers, ''Basic Statistics".
i Q Was that a text of your choice?
5 A No; George Summers was a member of the faculty in
= the Department of Management.
by Q Your graduate course, you said, was in Business
Statistics?
22
A Yes.
23
or Q What text did you use there?
w oy A Those texts also varied. One text that I remember
EARLENE P. STEWART
CERTIFIED COURT REPORTER
61
f i A YY) oS 7X
was ats some kind of business statistics.
Q Now, did you continue any research during this
first year?
A Yes.
Q What was your area of research interest?
A I continued my research on Input-Output Models.
Q Were you in touch with Dr. Burford?
A Yes.
Q Did you work or consult or collaborate, TI guess is
the word, with anyone else in your professional research at
this time?
A Yes.
Q Who?
A Dr. Bruce Billings?
Q And did your research with either Dr. Burford or
Dr. Billings result in any publications or papers or pre-
sentations during this year?
A Yes.
Q Are they reflected in your presentation portion
of your Vita or your publication portion?
A Yes.
Q Let me ask you, Dr. Katz, just to be sure the recorc
is complete on this, do you have any publications or pre-
sentations which you've completed but are not reflected in
your Vita?
EARLENE P. STEWART
CERTIFIED COURT REPORTER
62
A I have papers that are in the process of being
completed and being submitted for consideration to journals
which are not indicated here,
Q We may talk about those a little bit later. But as
far as prior publications or prior presentations, the Vita is
complete?
A More or less.
Q More or less means that you can't remember at this
point any omissions?
A Well, I made a lot of presentations, and there may
be one or two that hasn't been reflected here in this Vita.
Q All right. But this is your best recollection at
this point of what you've done?
A Yes.
Q Did you have any consultants jobs during this year,
this first year at the University of Arizona?
A No.
Q Any other outside activities beyond your course work
and teaching and your research?
A No.
Q You mentioned the second year you moved your courses
in the Department of Management Information Systems. What did
you teach that year?
A I taught Basic Fortran and I taught the Graduate
Operations Research course that I taught in the Management
EARLENE P. STEWART
CERTIFIED COURT REPORTER
63
Department, And TI also belieye I taught the graduate statistics
course that IT taught in the first year.
Q Was that using the same text, do you recallZ- that's
the graduate business statistics course?
A Yes.
Q And you indicated you're not sure what text -- they
varied; is that correct?
A Yes,
Q Now, you came to Georgia State in 1981; is that
right?
A Yes,
Q And in what position?
A As an Assistant Professor.
Q What were your responsibilities as Assistant
Professor?
A To teach and do research.
Q What teaching responsibilities did you have the firs
year -- let me broaden that question so that I won't have to
ask several sub-questions. Tell me the courses that you've
taught since you've come to Georgia State and when?
A My first year I taught Mathematics for Statistical
Applications, which is a graduate course.
Q What text did you use?
TRL aa
A I used Salas and Hittev's. Calculus and I used
Mathematical Methods in Statistics or Matrix Methods for
EARLENE P. STEWART
CERTIFIED COURT REPORTER
64
Statistics by Green, I'm not sure of the title. I also
taught BA 605 which is the masters MBA Graduate Course in
Statistics; that was the first quarter. The second quarter,
I taught Statistical Theory I, which is QM 805.
Q And let me ask you about the text in that course?
A The text was Pg "Mathematical Statistics.’
And the second course was DM 310, which is the first under-
graduate statistics course, And then the third quarter --
Q What was your basic text in that course?
A That text was Daniel and Terrell, "Business Statistici.
The third quarter I taught Statistical Theory II, which is
oM 806, the second half of the sequence.
Q And the text remained the same?
H 343 O A 049 Qs
A It was the second part Hegan;Craig. Aol T believe
DM 121, which is Basic Mathematics for Business Undergraduates H
which is a course that prepares them for the next course, which
is calculus. Then in the summer, I taught two DM 121's, and %
Q I'm sorry; DM 121 is what?
A Is the Basic Mathematics for Business Students, which
is the --
Q I'm sorry.
A And BA 652 and BA 653; BA 652 is a course for Ph.D.
students to improve their background in calculus, and BA 653
is a course for Ph.D. gtudents to improve their background in
matrix algebra. Then, in the fall of, this would be 1982.1
EARLENE P. STEWART
CERTIFIED COURT REPORTER
65
taught the OM 865, which is the mathematical techniques for
statistics.
Q What was your text in tongs course?
A Again, it RRR a wrong; that course was
cancelled. I was scheduled to teach it, but we didn't have
enough enrollment. Therefore, instead, I taught DM 312, which
is the second part of the statistics course -- DM 310 is the
first one. And I taught a sampling course.
Q Now, the sampling course was --
A Was an undergraduate course.
Q And what text did you use for that?
A We used Mendenhat and Scheoffer and Ott,' Sampling.
Q Let me ask you, that course is outside the range
of a lot of courses you taught previously, is that correct;
or am I wrong? Is it sampling theory or are you talking
about sampling -=
A No, sampling applications. We talked about
different sample designs; we talked about questions dealing
with how to design questionnaires; those kinds of things.
Q Had you previously done teaching in that area?
A No, that was my first time teaching in that area.
Q And what text did you use?
Mendenhall , Seheafjer
A Mendenhahl, Sehaeffer and Ott; it was called
1" "Sampling,
Q The previous courses that you had in that area
EARLENE P. STEWART
CERTIFIED COURT REPORTER
66
were courses you had in graduate school?
A Yes,
Q I guess one summer you talked about multivariate
analysis?
A I had a course in sampling theory.
Q And that was the fall of 1982 you mentioned, not
your course, but the one that you taught?
A Yes.
Q And was that to undergraduates you said or --
A Undergraduate, yes.
Q Were there any course materials that you've
generated since you've been at Georgia State, any of the
courses that you've mentioned here beyond the text that
you've used? In other words, did you generate any handouts
Or Xerox course materials?
A Well, starting in 1982, fall of 1982, I became
Coordinator of the DM 310 - 312 sequence. So, as part of my
duties, I have to design the syllabus, create a final exam,
and so forth.
Q I'm sorry; the 210 - 212 series is --
A 310 - 312.
Q All right; 310 - 312 is what?
A It's two courses in business statistics, Business 0
Statistics TIT and Business Statistics II.
Q Some professors I've seen in the past work not as
EARLENE P. STEWART
CERTIFIED COURT REPORTER
67
much from texts but from xerox copies of materials they
produce, That's not --
A That's not the way I do it; no.
Q This brings us all the way, I think, to the spring
of 1983, which is this spring. What courses did you teach?
A The winter of 1983,
Q I'm sorry; you're in'a quarter system.
A That would be the winter quarter; I taught QM 805,
which is the Statistical Theory, first part, and I used Hosdal!
Craig. And I taught, I believe it was DM 312, the second part
of the Business Statistics. I think the idea is now that I'm
Coordinator, I must teach at least DM 310 and 312 every
quarter. And then in the spring I taught QM 806, which is
the second part of the Statistical Theory, Probability Theory
course, and another DM 312. And the summer, I'm not teaching?
Q It will be a relief.
A Well, my time is going to be needed in other areas.
Q I may ask you a little about that later. During
this time you've been at Georgia State, have you engaged in
additional research?
A Yes, on Input-Output Models.
Q With Dr. Burford and Dr. Billings or others?
A With Dr. Burford. There's also other areas of
research that I'm involved in that originated in Arizona. One
is on incompatibility matrices, which is with Dr. Singual, who
EARLENE P. STEWART
CERTIFIED COURT REPORTER
68
is now at the University of Middle Tennessee. Then, at
Georgia State
Q Let me ask you -- and then you can come back -- in
compatibility matrices; is that a more theoretical --
A Yes, that's a theoretical -- I've developed proof of
some question that had come up in the literature about compati-
bility matrices, dealing with means and variances. So, it's
really a statistical application. Then, I'm also working with
two accountants who, they designed a random response question-
naire, and I'm helping them analyze the data.
Q Is this an outside consultant?
A No, this is just research.
Q Okay. Can you give me a little more information
about what area of accounting they were doing the work
in?
A Well, they wanted to determine whether State,
Federal, or Local accountants generally cheat in recording
information about audits. So, these are all auditors. So,
they wanted to collect information about sensitive data. And
so they used the random response design to try and convince
the respondents that their answers could not be traced to
them. And our trying to analyze ~--
MR. BOGER: Let's go off the record during the
telephone call.
MS. WESTMORELAND: I'm sorry.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
69
(Whereupon, there was a brief
interruption of the record,)
BY MR. BOGER:
Q Let's go back on the record. We were in the middle
of discussing this research project with the accountants.
Let me just ask a question or two. Are these accountants
just doing this on their own initiative; do they have some
funding?
A They had a grant.
Q From whom?
A I don't know.
Q Do you know how they approached you?
A Well, they needed a statistician to help them
analyze the data. And so, I'm a statistician, so they
approached me to help.
Q Okay; are they accountants here in Atlanta?
A Yes, they're with Georgia State -- they're in the
Accounting Department at Georgia State.
Q And what were their names?
MS. WESTMORELAND: Ts this something that you don't
THE WITNESS: TI know their names; it's just --
MR. BOGER: If you momentarily can't recall, I've
had that experience myself.
THE WITNESS: Gordon Harwood and -- this must be too
current for me to recall -- and Gene, I forget what his last
EARLENE P. STEWART
CERTIFIED COURT REPORTER
70
name is.
BY MR. BOGER:
Q All right, And your assistance to them has been in
the area of what, and what have you done for them?
A I've been analyzing their data.
Q Who collected the data?
A They collected the data.
Q Who designed the data questionnaire?
A They designed the questionnaire.
Q S50, your responsibility has been analysis?
A Yes.
Q Is that analysis complete at this. point?
A I believe so.
Q Is there a report?
A No, we're in the process doing the report.
Q When would you anticipate the report would be
available?
A Well, he said a rough draft would be ready in the
next few weeks or so, and that was a few weeks ago. But, you
know, he hasn't contacted me yet about giving me the rough
draft so I could look at it and correct any problems he had
statistically.
Q Now, I asked you about any outside activities other
than teaching and research, and you mentioned this accounting
project. Is there anything else that you've been engaged in
EARLENE P. STEWART
CERTIFIED COURT REPORTER
Z1
since you came to Georgia State?
A Such as doctoral students, dissertation advisors,
those kinds of areas?
Q Well, I didn't know. TI take it that is an area
that you've been engaged in as well?
A Yes,
Q Any other consultant's jobs?
A No, nothing formally.
Q Have you worked with Dr. Burford on some of his
projects?
A No. I occasionally do get calls from usually
lawyers who want to know statistical information over the
phone. And I give them statistical information over the
phone.
Q We may call you. My impression -- and I know this
is not something you're holding back on us -- the State of
Georgia has asked you to work on the project you're engaged
in now, this Report?
A Yes, I am working with the State of Georgia on this.
Q All right. TI may go tomorrow a little bit into the
background about that, not a great deal, but that's the only
outside consultant's job that you've --
A I thought that was known.
Q Yes, but I just want to be sure for the record that
we've got everything down, and that didn't suggest that you
EARLENE P. STEWART
CERTIFIED COURT REPORTER
72
were meaning to keep that one back, And I assumed that you
had assumed that that was known; yes,
You mentioned, Dr, Katz, the empirical research
that you were engaged in in a consultant role with Greyhound
Bus. And you've talked about the empirical research that
you've been engaged in and the analysis you've engaged in with
the accountants. Have you been involved in any other empiricdg
research since, really, the beginning of your graduate career?
A Yes.
Q What?
A I've been involved in empirical research ‘as a
dissertation advisor or a master's advisor: a thesis advisor.
Q And when has that been?
A At the University of Arizona I was on one Ph.D.
committee and one master's level committee.
Q What were the subjects of those masters thesis and
Ph.D. dissertations; if you recall?
A The masters thesis was testing high school seniors
attitudes between levels of job safety and what factors
seemed to be important in their selection of jobs; if safety
had any factor or any affect.
Q What was your particular role on the thesis committe
or what was your expertise?
A I worked with them in designing a questionnaire from
the statistical part,
EARLENE P. STEWART
CERTIFIED COURT REPORTER
1
Z3
Q And what do you mean by that?
A Well, his analysis -- if he had followed through,
whatever his analysis would have been,
Q Does that mean that you worked with this graduate
student before the questionnaire was designed, or was --
A Yes.
Q -- the questionnaire design and data already
collected and you --
A No, I worked with him to help him design the
questionnaire, and then he found out that he couldn't collect
any data.
Q Do you recall why that was?
A Well, he said that the Pima County School Board
had a regulation about letting people submit questionnaires
to their students, using up class time; that it had to be
submitted to the Board of Education six months before the
experiment was to be done. And of course, he wanted to
graduate at the end of the summer, and he was just getting
started about the beginning of the summer.
Q Okay; what happened?
A Well, he discussed it with his committee and his
committee seemed to think that, well, he's a hard luck person.
he ran into unforseen circumstances. So, they were willing
to accept some explanation of his experiment and the design
and how he would have analyzed the results and so forth.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
74
Now, at the time, this was just before I was going
to leave Arizona to come to Georgia State. At the time when
we had our first meeting, I found out that he hadn't really
learned any statistics, hadn't really figured out how he was
going to analyze the data. So, I made that point, and as
the committee agreed, he should be prevented from graduating
that point until he had indicated some of the statistical
information and techniques that would be required to analyze
his data.
But since 1 was going to leave in the next few
days, they felt the simplest thing for me to do was to sign
off and let them handle it from there; which is what I did.
Q You mean, you signed off of the committee. You
were no longer a committee member or you signed off of --
A I signed off approving it under the conditions that
they said they would insure that he would have the statistical
information in his thesis.
Q Do you have a copy of that work that you did?
A No.
Q Do you recall the name of the student?
A No,
Q Who were the other members of that committee?
A I don't remember their names either.
Q Was that in the Department of Management Information
Systems?
EARLENE P. STEWART
CERTIFIED COURT REPORTER
75
A No, it was the Department of Safety Education. or
Industrial Safety; something of that nature.
Q All right. You mentioned that as a masters degree
dissertation or thesis that you were involved in with your
involvement with empirical research, And then you mentioned
there was also, I think, a Ph,D dissertation committee that
you ‘were oOrll.
A Yes.
Q What was the nature of that research?
A That was research on a management topic, the
property of a person knowing whether he knows the answer to
a question. You can know the answer; you may not know the
answer, but knowing whether you know the answer or not is
another consideration.
Q That may be the most important subject I've ever
heard anybody --
A And she collected data and she was funded by, I
believe, the University of Arizona Grant Council or something.
And she investigated the property and did some statistical
analysis.
Q What was your role on that dissertation committee?
A Me, along with George Summers, our role was to
analyze the statistics.
Q Were you involved in the research design phase of th
work?
EARLENE P. STEWART
CERTIFIED COURT REPORTER
76
A No.
Q Or the instrument or questionnaire construction?
A No.
Q Was her dissertation approved?
A Yes.
Q Do you recall the name of that student?
A Jinoos Hosseini.
MR. BOGER: Would you mind spelling that for the
Court Reporter?
THE WITNESS: Jinoos, J-i-n-o0-0-s; Hosseini,
H-o-s-s-e-i-n-1i.
BY MR. BOGER:
Q Thank you. What department was that in?
A The Department of Management. But she was, her
major advisor was in the Department of Industrial Engineering.
Q All right. Was that possibly the department that
this other graduate student that you mentioned was in?
A No, he was in Safety Management; Industrial
Safety. I don't remember exactly the department. According
to what reports I had, it was 4a fly-by-night type department,
so, somehow they referred them to me. And now I know why
they referred them to me, because no one else wanted to get
involved in those kinds of theses. You see, I was new there.
Q You got them, the four instances that we have talked
about where you were involved in empirical research, your
EARLENE P. STEWART
CERTIFIED COURT REPORTER
77
Greyhound Bus research, the accountants research that you
have been engaged in recently, and the masters thesis and
the Ph.D dissertation, can you think of any other instances
where you've been involved?
A Yes. I'm now involved with a Ph.D. student at
Georgia State who is investigating what factors seem to
affect the decisions in labor arbitrations and wondering what
the factors are, whether there's any discrimination; those
kinds of issues.
Q What kind of discrimination do you mean?
A Based on sex; most sex discrimination and race
discrimination.
Q Have you had any previous either professional
training or research in the area of discrimination, race or
sex discrimination?
A No.
Q Is this Ph.D. dissertation one in which there are
other dissertation advisors?
A Yes.
Q What is your role on the dissertation committee?
A My role is the statistical aspects of the
dissertation, analysis of the data.
Q Is there anyone else involved who has experience
with discrimination?
A The other members of this committee have expertise
EARLENE P. STEWART
CERTIFIED COURT REPORTER
78
in arbitration.
Q What's the department?
A Management,
Q Who are the other members of the dissertation
committee?!
A I can give you the Chairman, Don Crane. We've only
had one meeting. I was sort of thrust into this committee,
and I read what he had done so far, and he hasn't done any-
thing yet. So, we had our meeting and he's been told to
pretty much redo it because he hasn't really come up with any
significant results,
So, I'm not all that familiar with all the names.
Q Okay. And your role is to oversee the analysis;
is that correct?
A Yes.
Q So, again, you're not involved in research design
or questionnaire construction?
A Well also research design and questionnaire con-
struetion.
Q Have either of those taken place yet?
A No, he has collected a lot of data, but he hasn't
really made any significant contributions in his analysis.
He probably will need to collect more data, and so I will
have to work with him on his collection of the data.
Q How long have you been involved, though, in this
EARLENE P. STEWART
CERTIFIED COURT REPORTER
79
particular dissertation?
A Since -- for the last year, But he's very slow;
it's not like there's been a lot going on for the past year.
I would get a chapter here, a chapter there.
Q All right. Any other empirical research that you've
been involved with?
A Other than my work with --
Q Other than with this.
A As far as I can recollect, no; that's as far as I
have gone.
Q Thank you Dr. Katz. What I'm going to do this
afternoon is ask you some questions about documents that I
hope will help us look through what you have both here and
in your office and then continue the deposition tomorrow
morning as we've discussed off the record beginning at 8:30
here. And I'll have a few more questions, then, tomorrow
morning about your professional activities and accomplishments
and then proceed more directly to the report that you've
written on Professor Baldus's Report.
I want to ask you, do you have a copy of the
Report on the highway project, the Greyhound Bus Report that
you mentioned?
A No.
Q Is that Report available anywhere?
A Dr. Burford may have a copy of that report. You
EARLENE P. STEWART
CERTIFIED COURT REPORTER
80
may be able to obtain that from him. It was an expert testi-
mony situation, so he may not have done a final report. He
may have just given his testimony.
Q In other words, he was involved in some sort of
litigation proceeding before a Federal Interstate Commerce
Commission or something of that sort?
A Yes, something like that,
Q You didn't have any role in that?
A No.
Q Do you have a copy of the dissertation of the
student that you did --
A Jinoos's?
Q Yes.
A Yes.
Q Can we have access to that?
A Yes.
Q I'm going to ask a series of questions about the
documents here and some other documents. Before I do,
Professor Baldus has indicated to me that if you want some
specific documents from him at his deposition on Tuesday and
you know about those now, if you can give him the list -- his
office is closed down Friday afternoon -- and we're talking
about a deposition that will begin next Wednesday. It would
be very advantageous to you, we can go off the record here
for about five minutes so that you can give that list. Or
EARLENE P. STEWART
CERTIFIED COURT REPORTER
81
maybe if it's not available now, give us that list tommorrow
morning.
MS. WESTMORELAND: Let's go off the record for a
minute, and we can discuss that.
(Whereupon, there was a brief
interruption of the record.)
BY MR. BOGER:
Q Dr. Katz, with Professor Baldus's assistance, I
have laid out on the far end of this table six piles of
documents, all of which you brought to the deposition today.
And those were the ones that were in your first box. I'm
going to go through your report and point tc tables seriatum
that you've got in your report and then ask you questions
about whether certain things, the underlying codes or the
computer instructions of the printouts are in, for those
particular tables, are in any of the piles that are before
us so that we can know what we're looking at in that sense.
And in that regard, I guess it will be best if
you'll look with me at our Petitioner's Exhibit 2 as I
go through, and then make yourself available to come to the
far end of the table as needed to answer my questions about
what piles the information we have is in or whether it's
not here this afternoon and something we'll need to get from
your office.
PROFESSOR BALDUS: Can I explain the piles to him?
EARLENE P. STEWART
CERTIFIED COURT REPORTER
82
I think it would simplify that.
MR. BOGER : Off the record.
(Whereupon, there was a brief
interruption of the record.)
MR. BOGER: Back on the record. To clarify what
we've got here, we've got the various stacks of data that you
brought this afternoon. And Professor Baldus, with Sam Laufer’
assistance, has put a written label on the different cate-
gories of these documents. We have all the documents that
were in your first box marked as A documents; A-1, A-2, A-3,
A-4, A-5, and A-6. And then all of the documents of interest
to us in your second box are marked in categories B; B-1
through B-6. And you had described to us seriatum the differen;
categories of documents that you brought, the numbers 1 through
6 following categories that you laid out for us at the outset
of the deposition.
So, now, what I'd like to do, and I hope this works
clearly, is ask you first with respect to Table Number I in
your preliminary report, which is at Page 5 of Petitioner's
Exhibit 2, is the code that consists of, are the computer
instructions generating that table anywhere within the
materials that you brought?
THE WITNESS: No.
BY MR. BOGER:
Q Is that in your office?
EARLENE P. STEWART
CERTIFIED COURT REPORTER
i
(&
1]
{
e
o
}
33
A Yes,
Q Can we have access to that this evening?
A Sure,
Q Is the printout of Table I, total printout, anywhere
within those materials before you?
A The printout of the information I used on the table
on here?
Q Yes.
A No.
Q Okay. Did you do any other runs of unknowns in the,
similar to the runs of the unknowns that you reflect in Table
17?
A Yes.
Q And are any of those other runs of unknowns present
here before us? A No.
Q Okay. Are they available in your office? |
A Yes.
Q Can we look at those?
A Yes.
Q And that, I guess, would include both the printouts
and the computer instructions to generate any additional
analysis?
A Yes.
Q Now, your TableII which you mention on Page 5, do
EARLENE P. STEWART
CERTIFIED COURT REPORTER
84
you have before us the computer instructions that generated
that Table II?
A Yes.
Q Is that in one of the piles; can you simply tell
us which pile letter that's in?
A Given that you've kept the piles together as I gave
them to you, it would be in Pile A-4,
Q Okay; thank you. And does that also include the
printout as well as the computer instructions?
A Yes.
Q Did you do any other runs of unknowns ‘similar to
the runs that you did on Table II for the Georgia Charging
and Sentencing Study?
A Yes.
Q Now, are those here with us this afternoon?
A I believe so; Pile 4, I believe.
Q Now, your Table III, which is mentioned on
Page 10 of the report, your preliminary report, Petitioner's
Exhibit 2, and which actually begins on Page 11 of Petitioner’
2, do you have anywhere on the table here the code of the
computer instructions to generate that table?
A I believe it's -- yes, on the table.
Q In what pile?
A A-5,
Q Thank you. Now, does that include the printout as
EARLENE P. STEWART
CERTIFIED COURT REPORTER
S
85
1 well as the code instructions?
2 A Yes, I believe it covers all the variables that
3 I have in the paper.
4 Q Do you believe, are you sure, or are you just not
5 positive on that?
6 A I'm not absolutely certain. There may be a variable
7 or two that is not contained in this file.
8 Q All right. Did you run other or do other runs on
9 mismatches between the two studies, the Procedural Reform
10 Study and the, Charging and Sentencing Study?
11 A I believe so, but I don't know where -- I'm not
12 sure. I think I had run a lot of runs on this particular
13 issue, but I can't be certain that I have all the printouts
® 14 that I got for those runs and I can't be certain that there
15 may not be some other variables that I didn't include in
16 || the printouts.
17 Q Would the printouts of those be in your office?
18 A Yes.
19 Q Could we have access to them if you can find them?
20 A If I can locate them; I think I put them in a
21 separate pile which is really my discard pile. You are free
29 to look through that file,
23 Q All right; thank you, Your Table IV which begins
24 on Page 12 of Petitioner's Exhibit 2 is a count or cross classi-
% 25 fication of the number of aggravating factors between the
EARLENE P. STEWART
CERTIFIED COURT REPORTER
86
Procedural Reform and the Charging and Sentencing Studies.
Is there on this table, or are there on this table any
documents that reflect the computer instructions that gene-
rated Table TV?
A Yes,
6 Q And will you tell us where those are?
7 A Again, that's in Pile A-5:
8 Q Okay; thank you. And that includes both the
9 computer instructions and the printout?
10 A Yes.
1 Q Let's look at your table -- before we turn to
Table V, let me ask you, did you compare these with respect
Ji 13 to other factors and aggravating factors, particularly,
14 for example, with mitigating factors between the two studies?
15 Did you do any of that kind of runs?
16 A I didn't perform that run yet,
17 | Q You --
18 A I may in the future do it, but I haven't done it
19 so far.
20 Q All right. Table:V, which is at Page 13 of
21 Petitioner's 2, "Difference Between Aggravating Factors by
22 Race of the Victim." Do you have anywhere on the table here
23 the computer instructions that generated that table?
24 A Yes; I believe it's in A-5 also.
% 25 Q Also in A-5., And that's both the instructions and
EARLENE P. STEWART
CERTIFIED COURT REPORTER
87
the printout?
A Yes.
Q Thank you. Now, let's look at Table VIII which is
on Page 22 of the master -- excuse me; Petitioner's Exhibit 2.
I believe, in fact, that it's not on Page 22, that it's
referred to at one point on Page 22. It's actually on Page
23, Table VIII.
Do you have anywhere on the table here documents
that reflect the computer instructions that generated Table
VIII?
A Yes, I believe it's in Pile A-6.
Q And that's both the computer instructions and the
printout?
A Yes.
Q Okay. Did you do any other runs apart from the ones
that are reflected in Table VIII where you compared certain
aggravating factors between black and white victim cases?
A Yes,
Q Are they present here this afternoon?
A Yes. These are later tables. Should I give you
the table numbers?
Q No, What I mean, I think, is are there any other
factors that you have, different variables that you've used
to make tables similar to the tables that you've made in
Table VIII?
EARLENE P. STEWART
CERTIFIED COURT REPORTER
9
88
A It's possible,
Q You don't recall?
A I believe I probably have, but this was the, I
believe this contains the comparisons for all of the wvariableg
that are indicated on the table, There may be other variables
that I have performed this kind of analysis on but for some
reason or other I have not used.
Q Would documents that contain those runs be present
your office?
A Yes, in my discard area.
Q All right. And we can have access to them?
A Yes.
Q Okay; fine. Now, let's look at Table XI, I believe
on Page 32 of Petitioner's Exhibit 2. Do you have here on the
table the documents that reflect the computer instructions
and the printout that generated Table XI?
A Yes.
Q And where are they located?
A A-1,
Q And that reflects both the instructions and the
printout?
A Yes.
Q Let me ask you about your Table XII which is found
on Page 34. Do you see on this table documents that reflect
the computer instructions and the printout involved in the
EARLENE P. STEWART
CERTIFIED COURT REPORTER
89
1 construction of Table XII?
2 A Yes.
3 Q Where is that?
4 A A-1,
5 Q Let me ask you: it looks as if there's a series of
6 tables that reflect the, Georgia's Statutory Aggravating
7 Factors that include Tables XI, XII, XIII, XIV. XV, XVI, and
8 XVII on Pages, respectively, 32, 34, 36, 38, 40, 43, and 45
9 of Petitioner's 2. Are all of those computer instructions
10 to generate those particular tables and printouts in a pile
11 that you've just given us two from, or do you know?
12 A Yes. It was my understanding that you only requested 13 || that I bring the programs for the first two tables.
14 | Q Okay. When you say first two tables; I'm sorry.
15 A For Statutory Aggravating Circumstance, B-1 and
16 B-2.
17 Q Okay; but as for the others, those are present
18 or not present here?
19 A They're not present here.
20 Q Okay. They're in your office?
21 A Yes.
29 Q We can have access to them if we need them?
23 A Yes.
4 Q Now, your table -- with respect to the other B
2% factors under the Georgia system Tables XIII through XVII
EARLENE P. STEWART
CERTIFIED COURT REPORTER
90
that we mentioned, in regard to coding, are they identical
to the coding that you've used to generate Tables XI and XII?
A Almost identical; yes,
Q What kind of variance is there?
A Well, I ‘had to change the code so it considers the
specific cases that are important to each table. The number
of cases and the actual cases change from table to table.
2
Any other changes in the coding that you can recall?
A No.
Q Okay. Do ‘the variables change?
A No.
Q Now, let's look at Table XIX on Page 48 of
Petitioner's Exhibit 2. Do you have on the table materials
which reflect the computer instructions that generated
Table XIX in the printout?
A No.
Q Are those available in your office?
A Yes.
Q And we can have acces to them?
A Yes.
Q Table XX on Page 50 of Petitioner's 2, is there
on the table information which reflects the computer in-
structions for generating that table?
A Not on the table,
Q That's available in your office?
EARLENE P. STEWART
CERTIFIED COURT REPORTER
A
Q
91
Yes, it's available in my office.
Are the codings for Tables XIX and XX identical,
apart from in terms of variables, to the earlier tables
we've talked about for the B factors, apart from the --
A
Q
A
Q
reflected
A
Q
The different selection cases --
Yes.
-- relevant to each situation; yes.
Did you generate any runs on other variables not
in the tables that you've got here in your report?
Be more specific.
With respect to your Tables XI through XX, did you
do any other runs on other variables?
A
Q
A
Q
A
Q
calculate
A
Q
Other than what there is; none.
Those that are reported?
Yes.
And do you have those with you here?
For Tables XI and XII, yes.
And for the others; they're in your office?
Yes.
And we could look at them if we need to this evening
Yes.
Thank you. Do you have a formula that you use to
the Z values that you reflect in some of these table
A formula?
Yes; do you have some sort of formula that's
EARLENE P. STEWART
CERTIFIED COURT REPORTER
oO
8
19
92
generated by computer otherwise to calculate the Z-Values?
A Not by computer,
Q How is that done?
A Okay; I had the P-Values from the printout. And
so from that, I can compute the Z-Values.
LO
You computed those by means of what formula?
A By the use of the tables.
Q What kinds of tables did you use?
A The normal tables,
Q Well, let me ask, because I may be beyond my own
technical abilities. Is a normal table a term of art in
your profession, or is it simply a standard table that one
uses to compute?
A It's a table of values for the standard normal
distribution, which is the Z-Distribution.
Q Thank you. Your Table XXI on Page 52 of Petitioner's
Exhibit 2 lists a number of unknowns for several variables.
Do you have any documents here on the table that reflect the
coding instructions that generated these unknowns?
A No.
Q Are those available in your office?
A I believe so; yes,
Q And we can have access to them?
A Yes,
Q And what about the underlying recodes for, involved
EARLENE P. STEWART
CERTIFIED COURT REPORTER
93
L with these tables? Is there any document in here that
2 reflects that?
3 A You mean the regressions?
4 Q No, I don't -- just one moment. It's the underlying
[8
] recode that enabled you to identify how many of these variables
6 had missing factors.
7 A I believe I have that in my office.
8 Q Do you have any additional runs which indicate
9 other unknowns for other variables?
10 A Yes.
1 Q And is that available here or is that in your office?
12 A It's in my office.
13 | Q Do you have any instructions here for the computer
® 14 on how to generate this list that you've got here in Table XXI?
5 A I'm not sure I'm understanding what you're asking
16 me.
17 MR. BOGER: Off the record.
18 (Whereupon, there was a brief
19 interruption of the record.)
o || BY MR. BOGER: |
21 Q Does Table XXI reflect the frequency distribution
22 for the unknowns with respect to these variables?
23 A Yes.
24 Q And do you have any others that are not reflected
% 95 here that you've done that way; frequency distributions?
EARLENE P. STEWART
CERTIFIED COURT REPORTER
94
: A Yes.
pA 7 Q And are they available here?
3 A No.
3 Q But they're in your office?
i A Yes.
6 Q All right. Now, you've got Table XXII on Page 55
/ of Petitioner's 2. Do you have anything on the table that
8 gives the computer instructions for generating Table XXII?
9 A No.
10 Q Is that available in your office?
11 A I'm not sure. I don't recall which pile it's in.
12 Q It might be in the discard pile?
13 | A It might be -- I'm not sure where it is. I will
® 14 | try and look for it. Now, I believe I have it.
15 | Q Okay. And with respect to XXIV--you have it in
16 your office, I take it?
17:04 A Yes.
18 Q With respect to XXIV on Page 60, would you have
19 the computer instructions which generate Table XXIV in your
ol office or here?
21 | A I believe they're here; let me check. Yes.
29 Q And where is that?
23 A Pile A-6.
24 Q Okay. Let me ask you with respect to comparison,
pe 2 | if you would, of Table XXII and Table XXIV -- those are on
EARLENE P. STEWART
CERTIFIED COURT REPORTER
No
6
24
25
95
Pages 35 and 60, respectively, of Exhibit 2, Is there any
difference in the instructions, coding instructions on how
you generated those two tables?
A There's a lot of difference.
Q Okay. We may, I think then, need to look at, for
certain in your office, Table XXII's coding instructions
in the printout. Did you do other runs of -- looking back
to Page 55 -- did you do other runs involving other variables
similar to the ones in Table XXII?
A I believe;it's possible.
Q Would those be in your office?
A In the discard area.
Q It sounds like you may have a big discard pile. Did
you do other runs similar to Table XXIV, comparing white and
black victim cases on some of these factors?
A Again, I believe so.
Q Would they be in your office?
A In the discard area.
Q All right. Now, let's look at Table XXV, which is o
Page 65 of the Petitioner's Exhibit 2. Do you have the docu-
ments here that would reflect the computer instructions which
generated this table?
A Yes,
Q And where are they found?
A A-6.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
No
96
Q A-6. Did you do any other runs similar to this
one involving other variables?
A I believe so,
Q Where are they; are they here or in your office?
A In my office in the discard ares.
Q Now, on Page 67, Table XXVI, once again, let me ask
you if there are any documents here that reflect coding
instructions for the generated Table XXVI?
A Yes,
Q And where are they?
A A-6.
Did you do any other runs similar to those? Q
A I believe so.
Q And they're in your office?
A Yes. Let me clarify; I'm not certain that there's
one or two other variables that I've done. It's Just thar
it's been a long time and I have been trying to get as many
variables as possible, and some variables don't cooperate
because there were very few observations and they weren't
included in the study. And that's my response.
Q Those are the kinds of variables that you're not
reflecting here; is that what you're saying?
A I think I remember I defined some other variables
that weren't defined by Mr. Baldus and ran these kinds of
experiments,
EARLENE P. STEWART
CERTIFIED COURT REPORTER
97
1 Q How did you define those? Can you give me an
“ 5 || example; if you can recall?
3 A I had a variable V tossed, which means was the
4 || victim tossed off of a high place as the method of death.
Q All right." “We're, 1 think, nearing the end of this
: inquiry and you've been very cooperative, Dr. Katz. Table
, || XVII on Page /1 of Petitioner's Exhibit 2, are there on the
8 table here documents which reflect the computer instructions
5 which generated Table XXVII?
io A Yes, A-6.
i“ Q All right. And did you do other runs similar to
5 this that are not reflected?
4s A Yes, I believe so.
“ a Q They're in your office?
+2 A I believe so.
in MR. BOGER:. Dr. Katz, I'm going to show you the
0 original of a document which we will mark as Petitioner's
a Exhibit 4 for identification.
2 (Whereupon, Petitioner's
” Exhibit 4 was marked
3 for identification.)
BY MR. BOGER:
22
> Q Let me just state for the record, in order to move
things along as quickly as we can, this is, I believe, and I
® hope you'll tomorrow identify this as the continuation of your
EARLENE P. STEWART
CERTIFIED COURT REPORTER
98
preliminary report simply with the numbering system on it for
all of our clarity, much as we put numbers on the first part
of your report. Let me ask you, if you would, to look at Page
82 of this Petitioner's Exhibit 4. And there begin a series
of tables, Table XXIX through Table XLI, all involving white
and black victim data. I'd like to ask you if the instructions
which generated those tables appear anywhere in the documents
on the table? Maybe the answer to that question is different
for different tables, but my information suggests it might be
similar for all the tables.
A I don't believe any of the computer programs that
generated these tables are on the table here.
Q Okay; are they in your office?
A Yes.
Q Are your instructions basically similar for each
of these tables except for material you separated out in
different cases that fell under the categories that you
indicated with earlier tables, or were there different
computer instructions for these different tables?
A Well, Tables XXX to XLI, they were similar as in
the first study when I compared each specific statutory
aggravating circumstance and totaled the number of aggravating
circumstances similar in the changes to generate these
tables. The first table, Table XXIX was generated from
a completely different program.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
99
1 Q Okay; that program is available to us in your
2 office?
3 A Yes.
4 Q And the program for XXX through XLI; is it also
5 available there?
6 A I believe so. I believe all these programs are
7 there.
8 Q Have you done any other runs including other
9 variables on the Charging and Sentencing Study that are not
10 reflected in your tables?
11 A Yes.
12 8) And are they charging -- are computer instructions
13 and printouts available in your office?
. 14 A In the discard area.
15 Q All right. On Page 107 of Petitioner's Exhibit 4,
16 there's Table XLII. Do you have available here the recodes
17 for all of the charging and sentencing variables that are
18 reflected there listing the unknowns?
19 A You mean do I have the programs used to generate
20 these unknowns on the table?
oh Q That's right.
29 A Yes; I believe they're in pile A-4.
23 Q All right. . Did you do any other runs with other
24 variables that are not included in those tables or that
® 25 table?
EARLENE P. STEWART
A CERTIFIED COURT REPORTER
100
A Yes.
Q Are they available in your office?
A I believe they're available in pile A-4.
Q Okay; so the other runs on XLII are in A-47?
A Yes.
Q Did you do any runs on the distribution of all
missing variables to reflect whether there was a random
pattern distribution or another pattern of distribution of
missing variables?
A Not formally, no.
Q What do you mean by that?
A I didn't run a program to try and determine the
pattern of missing variables. I take that back; I did do
one thing concerning that on the first study.
Q What was that?
A That's reflected in pile B-1, but not formally,
statistically analyzed, but just for my own purposes to get
an idea of the pattern.
Q Okay; why don't you tell us what that is just in
very general terms?
A For the 607 cases, it indicates unknown values for
responses between, deriving from the questions W-9, W-20 and
W-22, W-23, W-24; a whole host of the questions from the
Procedural Reform Study questionnaire. And in circles, those
questions were unknown responses in here.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
101
Q Okay; we can look at it further, I suppose. I just
wanted to get a general sense there for what we've got. There
a Table XLIII, which I regret to say looks, I believe it's
going to be on 112 or 111 on your Petitioner's Exhibit 4.
Do the computer instructions that generated that table appear
here on the documents you've got here with us?
A I don't believe so.
Q They're in your office?
A Yes, I believe so.
Q And did you do other runs similar to Table XLIII
involving other factors or other variables?
A I believe so.
Q Are they in your office?
A Yes.
Q Okay. On Page 123 of Petitioner's Exhibit 4, there!
a Table XLIV. Indeed, let me -- I'm not sure whether this
will speed things up or not, but I want to ask about Tables
XLIV, XLV, XLVI, XLVIL, XLVIII. With respect to all of
those, do you have with you the computer instructions which
generated those tables?
A I believe they're contained in Pile B-6.
Q That's for all of these tables we've mentioned?
A Yes.
Q Thank you. Does that include printouts as well as
the computer instructions?
EARLENE P. STEWART
CERTIFIED COURT REPORTER
h
102
! A Yes.
2 Q Okay. Now, let me ask, then, about another series
3 beginning with L, LI -- well, L, LI, LIV, and LV. We're
4 not interested in Table LIII. I'm sorry; I'm getting tired.
5 We're not interested in either LII or LIII. We just want
6 Tables L, LI, LIV, and LV; the computer instructions that
7 generated those tables.
8 A Yes, I believe they're also in Pile B-6.
9 Q Thank you. Let me ask you one or two more questions.
10 Did you do any runs to try to replicate Professor Baldus's
11 own regressions?
12 A Yes.
13 Q Are they reflected in the documents here?
*
14 A No.
15 Q Do you have those your office?
16 al Yes.
17 Q And those are available for us?
18 A Yes.
19 MR. BOGER: Just one moment off the record.
20 (Whereupon, there was a brief
21 interruption of the record.
29 BY MR. BOGER:
23 Q Dr. Katz, does Pile A-2 here before us represent
24 analysis that you have run, either regressions or cross
26 tabulations, with respect to the data?
EARLENE P. STEWART
CERTIFIED COURT REPORTER
103
! A Cross tabulations; no. This represents my original
® 2 multivariate testing in terms of trying to develop factor
3 classes for certain kinds of cases.
4 Now, if vou include cross classifications in what
: you want, then those are stacks and stacks of printouts that
6 are in my office.
7 Q In other words, you've done stacks and stacks of
8 cross classification or what we call cross tabulations?
9 A Yes.
10 Q Have you done any other multivariate regressions
11 or multiple regressions on --
12 A I haven't done any original regressions. I've
13 done regressions trying to obtain the same results that
*
14 Mr. Baldus obtained.
15 Q Other than, though, the attempted replications of
16 his work, have you done nothing else?
17 A Right. Plus, I also attempted to replicate the
18 factor classes as indicated in the Procedural Reform Study.
19 Q Good. Are those attempted replications before us
20 here, or are they in your office?
21 A They're in the discard file.
So Q All right. Do you have -- can you clarify that
23 last set of remarks? What were you trying to do there?
24 A Okay. In the fourth, in the preliminary report,
a or, the fourth method of analyzing the data for discrimination
EARLENE P. STEWART
CERTIFIED COURT REPORTER
104
was by taking the original 613 cases of the preliminary report
down to 607 in the amended report; trying to classify them
according to seven levels of factors. And I attempted, given th
information that I was given in the preliminary report and the
interrogatories, to determine which cases fell into each categd
Q Okay. And the results of that are now in your
discard file; did you say?
A Correct.
Q In terms of facilitating what might be a lot of looki
tonight, what were tie pediles of that set of analysis that you :
A I couldn't match the cases that he had indicated.
Q You said you could not?
A Right.
Q Okay. Let me ask you if you've done any other runs
involving race of victim categories other than the ones we've
talked about?
A Yes.
Q And where are they?
A They're in my office.
Q Any others involving race of defendant categories?
A Yes.
Q Are they also in your office?
A Yes.
Q And finally, have you done any runs with both race
of defendant and race of victim?
EARLENE P. STEWART
CERTIFIED COURT REPORTER
ne
ry.
ng
an?
105
A Yes.
Q © And they're in your office?
A Yes; and there's something on the table.
Q What's on the table?
A B-3.
Q And what's in B-37
A Defendant victim racial combination.
Q What distinguishes materials that are in B-3 from
those that are back in your office in this category?
A They're the compilation of the tables that I
obtained from the computer programs that are back in my
office.
0 So, the computer programs themselves, and I guess
the coding instructions are back in:your office, and this
is the compilation in B-67
A Yes.
Q Your compilation there in B-6, does that cover
and report upon all of your analyses that you found or some?
Excuse me; B-3.
A B-37
Q Yes.
A I believe it only reports on some.
Q What was the basis for the selection; do you recall}
A Well, a lot of the variables had very few
occurrences, and so, that and toting them around all the way
EARLENE P. STEWART
CERTIFIED COURT REPORTER
106
1 through the study, I discarded them.
2 Q Have you done any bivariate analysis with race
3 variables?
4 A Like correlation, correlation analysis?
5 Q Yes.
6 A Yes.
7 Q And do you have anything on the table here that
8 reflects those?
0 A Nothing that we haven't already referred to.
10 Q Okay. Apart from things that you've referred to,
11 do you have anything in your office?
12 A Yes.
13 Q Let me ask a general question -- you've been
14 || very gracious -- I take it all this material that we've
15 || talked about that's in your office is available to us until
16 the cleaning ladies close it up at 11:00 p.m. or whatever?
17 A Yes.
18 MR. BOGER: Let me reflect for the record,
19 || Professor Baldus has, while we've been conducting this depo-
20 || sition, inventoried all the documents that are included in
21 || the various groups that we've got, the A-1 through 6 and the
22 || B. We would like to take these documents with us this
23 || evening and keep them very carefully from harm, but have them
24 || so we can be ready tomorrow to ask you some questions, perhaps,
y 25 || about them.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
107
1 I believe we talked off the record about this, and
p 2 you asked for inventory. I think this is that inventory
3 if that's all right with Ms. Westmoreland and Dr. Katz.
4 MS. WESTMORELAND: You want to look through that?
5 ~ THE WITNESS: (Examines document).
6 MR. BOGER: Let me go back on the record for a
7 minute. Dr. Katz, with respect to all of these Tables that
8 involve these statistics, leaving the different cases that
9 I understand were involved with different tables and different
10 variables that may have come up, was the underlying statistical
11 methodology used to generate these tables the same throughout,
12 or did you use different methodology to generate different
13 tables and to generate the Z statistics?
® 14 THE WITNESS: Yes, I used different methologies.
15 MR. BOGER: Different methodologies; okay. We'll
16 explore that again tomorrow, maybe some of the rationales of
17 that.
18 MS. WESTMORELAND: Can we go off the record for a
19 few minutes?
20 MR. BOGER: TI think we'll complete the deposition
21 for this afternoon and continue again tomorrow at 8:30 a.m.
29 (Whereupon, the above proceedings were adjourned until
23 Friday, July 1,:.1983.)
EARLENE P. STEWART
CERTIFIED COURT REPORTER
VOLUME I 108
1 S. UBM I:8:8 TT ON
© 2 I, DR. JOSEPH L. KATZ, certify tlat I have read or have
3 || had read to me the foregoing deposition, and that I have sub-
4 || mitted “0 changes in form and substance to the officer
5 || before whom this deposition was taken, to be included by her
6 || in the deposition.
| ~M A
7 THIS (4. ‘day of J ula , 1983
~~ / j ' ; 4 / { & —f— —
" Ly pais Ki fA
I
4 T
Jo aed DR. JOSEPH L. KATZ by,
Signed, sealed and delivered
in the presence of:
16\ 24 Fé Wi 4 A ; nl
7
Woh PaV
" A iv Er’
No A L fd
17 | Notary Public Pak
/ &
EARLENE P. STEWART
CERTIFIED COURT REPORTER
® WARREN MCCLESKEY V. WALTER ZANT
No. C81-2434A
Corrections for Deposition
of Dr. Joseph L. Katz
volune JET
As 1n
Page Line Deposition Corrected
12 7 "date" data
18 19 "Zimant" Zemanek
19 5 "dimension" dimensional
vector spaces
19 114 12 "common parts" combinatorics
19 14 "Scholls” Scholz
® 24 18 "to satisfy" and satisfy
27 12 "additions" editions
31 14 "typology" topology
31 24 "Breckowtiz" Prekowitz
31 24 "typology" topology
32 17 "Breckowitz" Prekowitz
32 18 "typology" topology
33 12 nELon' Eaton
34 7 "Posser" Pazer
34 17 "Gloss" Dwass
35 17 "Lacks" Lax
36 “ "Antricritical’ Edge-critical
As in
Deposition
"Kassler"
"Grable and Bowes"
"employee theory"
"Lawrence Scott"
"Dr. Marks"
"Dr. Don Marks"
"non-lineal"
"Kisch"
"Semlar's"
"Alden Tades"
"Hargraves"
"Marks"
"Alden Taves"
"Huang 2001"
"ought to have done"
"Program and Theory"
"Hambert"
"Hilley's"
“Hogan, Craig"
Corrected
Castler
Graybill and
Boes
inventory
theory
Loren Scott
Dr. Marx
Dr. Don Marx
non-linear
Kish
Cinlar's
Alden Toevs
Hargrave
Marx
Alden Toevs
QM 2001
audited
Programming
Theory
Hamburg
Hille's
Hogg and Craig
"Hogan, Craig" Hogg and Craig
"Hogan" Hogg ang « « .
"Mendenhahl and
Schaeffer"
Mendenhall and
Scheaffer
Mendenhall and
Scheaffer
"Mendenhahl and
Schaeffer"
"Hogan-" Hogg and
109
C.BR'T LiFlInCA T E
STATE OF GEORGIA |
COUNTY OF DEKALB
I, Earlene P. Stewart, a Certified Court Reporter and
Notary Public in and for the State of Georgia at large, do
hereby certify as follows:
THAT I correctly reported by the stenomask verbatim
method the foregoing deposition of DR. JOSEPH L. KATZ (Volume I
at the date and place stated in the caption hereof;
THAT the witness was duly sworn and examined by counsel;
THAT my tape recorded notes were later reduced to type-
writing under my personal supervision and control, and that
the foregoing transcript contains the full, true and correct
testimony of said witness on said occasion;
THAT I am neither of kin nor counsel to any parties in-
volved in the matter, nor in any manner interested in the
result thereof;
THAT the reading and signing of the deposition was not
waived by counsel and the witness;
THAT all changes submitted by the witness have been
included in the deposition;
THIS 18th day of ‘July, 1983.
EARLENE P. STEWART
CERTIFIED COURT REPORTER
EARLENE P. STEWART
CERTIFIED COURT REPORTER
4334 GREENVALE DRIVE
DECATUR, GEORGIA 30034
(404) 981-231
IN THE UNITED STATES DISTRICT COURT
NORTHERN DISTRICT OF GEORGIA
ATLANTA DIVISION
WARREN McCLESKEY,
Petitioner,
CIVIL ACTION FILE
VS.
NUMBER: C81-2434A
WALTER D. ZANT, Superintendent,
Georgia Diagnostic & Classification
Center,
Respondent.
VOLUME II
DEPOSTION OF DR. JOSEPH L. KATZ taken at the instance
of the Petitioner at 132 State Judicial Building, Atlanta,
Georgia, before Earlene P. Stewart, Certified Court
Reporter and Notary Public, on the lst day of July, 1983,
at 8:30 a.m.
APPEARANCE 'OF COUNSEL
For the Petitioner: For the Respondent:
JOHN CHARLES BOGER MARY BETH WESTMORELAND
Attorney at Law Assistant Attorney General
NAACP Legal Defense and 132 Judicial Building
Educational Fund Atlanta, Georgia 30334
10 Columbus Circle
New York, New York 10019
Also Present
Robert H. Stroup, Esq. Professor David C. Baldus
Co-Counsel Samuel P. Laufer
VOLUME II
1 CONTENTS
& 2 | WITNESS DIRECT CROSS REDIRECT RECROSS
3 | DR. JOSEPH L. KATZ |
By Mr. Boger 3
9 EXHIBITS
10 | Marked Identified Page
11 Petitioner's Exhibit 5 Table of Cross Tabu- 17
lations
-
F
O
R
M
74
0
—
~
0
7
0
0
2
P
E
N
G
A
D
CO
.,
B
A
Y
O
N
N
E
,
N.
J.
N
ol
-
F
O
R
M
74
0
0
7
0
0
2
P
E
N
G
A
D
CO
.,
B
A
Y
O
N
N
E
,
N.
J.
24
25
VOLUME II
BD E<P 0 S-1.T ION
‘MR. BOGER: Back on the record, and Dr. Katz, you're
still under oath; good morning.
THE WITNESS: Good morning.
DIRECT EXAMINATION
BY MR. BOGER:
Q I have a few questions that I want to ask you to
continue from yesterday into the examination about your back-
ground and qualifications. Dr. Katz, are you a member of any
professional societies or organizations?
A Yes.
Q What?
A The American Institute for Decision Sciences; the
Institute of Management Science; Western Regional Science
Association; Southern Regional Science Association; American
Statistical Association.
Q All right. How long have you been a member of those
various organizations?
A For about five years since I graduated with my
doctorate.
Q From LSU?
A Yes.
Q Do you regularly read any professional journals?
A Yes.
Q What journals do you read?
-
F
O
R
M
74
0
B
A
Y
O
N
N
E
,
N.
J.
0
7
0
0
2
P
E
N
G
A
D
CO
.,
24
25
A The Annals of Regional Science, the Journal of
Regional Science, and the magazine that's put out by American
Statistical Association.
Q What is that?
A It's not the Journal: it's: the; I think it's called
the “'Statistician.™
Q Is it your testimony that you don't read the Journal
or that you do -- I'm just not clear what you mean.
A Well, I don't -- the Journal of the American
Statistical Association isn't a journal that you w wuld have
all that much interest in every single article that's in there
Most of those articles are highly theoretical in nature and
don't really apply to things that I'm doing. So, I do not
read it from cover to cover.
Q Okay. And the Journal of Regional Science, what
do you mean by Regional Science, and what does that field
encompass?
A The regional science is an economic area where it's
interested in the factors that affect regional growth; issues
like housing, migration, and input-output analysis, regional
structure, regional economic structure, issues of that sort.
Q From your Resume or Vita, is it fair to say
that's one of the major areas of interest?
A Yes.
Q Is that the major area of your interest, professionally
-
F
O
R
M
74
0
B
A
Y
O
N
N
E
,
N.
J.
0
7
0
0
2
P
E
N
G
A
D
Co
O.
,
24
25
at least, in the past?
A Well, it has been these past years.
Q Are you on the editorial board of any professional
journalg?
A No.
Q Since you've gotten your Ph.D., have you received
any grants from the National Science Foundation or other grant
organizations?
A No.
Q Are you on any grant review committees or anything of
that sort?
A No.
Q Have you gotten any academic awards of any sort?
A No.
Q Have you previously served as a consultant on any
occasions other than those we mentioned yesterday? We talked
about several ones. |
A Not that I -- I think I described them all yesterday.
Q Thank you. Have you previously given testimony or
expert opinion in any legal proceeding?
A No.
0 Dr. Katz, as you know, the State has identified you
as an expert witness in the upcoming hearing of McCleskey versus
Zant. You're aware of that?
A Yes.
-
F
O
R
M
74
0
B
A
Y
O
N
N
E
,
N.
J.
0
7
0
0
2
P
E
N
G
A
D
CO
.,
Q That's why you're here today, I suppose?
A Correct.
Q How would you describe the precise areas of
expertise that you bring to this case?
A I believe I have a general knowledge of the legal
system that has been given to me by the lawyers that I've
been working with. But my precise expertise is in the
analysis of Professor Baldus's Study; the statistical analysis,
analysis of the questionnaire design, analysis of the results
0 Okay. So that you focus that question, in effect, as
directly as I asked it. Let me back up a little and ask it
in a slightly more general way. If you were asked the
question by the Court or by myself here in what areas are you
an expert, professionally, what is your precise response?
Do you understand my question?
A Not completely.
Q For example, one could say "I'm an expert in
abnormal psychology, or I'm an expert in the field of law
with emphasis in constitutional law'; that kind of thing.
And that really is a title, I would suggest, or it's self
designation growing out of one's own background in research
and training. Given that kind of understanding of what it
means to designate oneself as an expert, in what area or
areas is your specialty?
A I'm a Statistical Expert.
-
F
O
R
M
74
0
B
A
Y
O
N
N
E
,
N.
J.
0
7
0
0
2
P
E
N
G
A
D
CO
.,
24
25
Q Do you -- let me ask several other related question
Do you claim expert status in the area of quantitative method
generally?
A Yes.
Q In the area of multivariate analysis?
A I wouldn't claim that I knew everything about
multivariate analysis.
Q What about the area of empirical research, empirica
quantitative research?
A Well, I'm an expert in the sense that I've been
trained to analyze statistical data properly. And my trainin
is both in the applications in some theoretical training, and
I'm an expert in analyzing the data.
Q Okay. Does your expertise extend to areas of
research design and data gathering, or is it more the end of
data analysis?
A Yes, it extends to research design and data
analysis.
Q Let me ask just one or two questions about how you
became involved in this case. When were you first contacted
by the State and asked to look at this case?
A I believe it was in early November, 1982.
Q Do you recall who contacted you?
A Yes; Mr. Dumich.
Q That's Nicholas Dumich from the Attorney General's
Ul
.
UJ
09
-
F
O
R
M
74
0
0
7
0
0
2
P
E
N
G
A
D
CO
.,
B
A
Y
O
N
N
E
,
N.
J.
24
25
office?
A Yes.
Q Did he explain to you what services he sought from yo?
A Yes.
Q And what did he ask you to do?
A He asked me to review some materials that had --
the Preliminary Report with Appendices -- that had been sent
by Professor Baldus to him. And he'asked me to review them
and comment on them.
Q Okay. And did he give you the materials, then, in
November?
A Yes.
Q Did you begin a review?
A Yes.
Q There was subsequently a time -- let me back up a
moment. Were those materials. Professor Baldus's Sreltningty
report of about 20 to 25 pages, I think, that had a date of
June 25, 1982; or do you recall?
A I don't recall exactly the date on -- are you
referring to Appendix D?
Q Well, let me ask you, do you have the materials
with you?
A I believe so.
Q Okay. Let's just be sure for the record what you've
gor,
-
F
O
R
M
74
0
P
E
N
G
A
D
CO
.,
B
A
Y
O
N
N
E
,
N.
J.
0
7
0
0
2
24
25
A It's not there. Where's the --
MR. BOGER: Off the record.
(Whereupon, there was a brief
interruption of the record.)
MR. STROUP: We're back on the record.
BY MR. BOGER:
Q Have you located the materials that were provided
for you by the State back in November of 198327
A Yes, I've located Appendix B, C, and D.
Q Would you just for the record read what's the
heading under those Appendices?
A Appendix B is "Questionnaire: Georgia Procedural
Reform Study, 1973 - 1978"; Appendix C, "Questionnaire, .
Georgia Charging and Sentencing Study, 1973 - 1979";
Appendix D, '"'Baldus, Woodworth and ulaoky, The Differential
Treatment of White and Black Victim Homicide Cases in
Georgia's Capital Charging and Sentencing i Preliminary
Findings."
Q All right. Then, I take it those Appendices
include those documents that are described, the question-
naires from the two studies. And I believe they're
Appendices to a pleading we filed in this case back in June
of:'72.
A Yes. wy
did :
Q At some subsequent time, -éid you receive any
-
F
O
R
M
74
0
B
A
Y
O
N
N
E
,
N.
J.
0
7
0
0
2
P
E
N
G
A
D
CO
.,
10
additional information or data that assisted you in your
conduct of a review?
A Yes.
Q What was that?
A We asked for in the discovery period all the data
for both studies that Professor Baldus used in compiling his
results for the preliminary report plus additional informatio
provided by Professor Baldus concerning how his variables wer;
defined, the names of the cases, case numbers, further
information about his conduct of the study.
He also provided me with a computer printout of
the raw data for the first study, the Procedural Reform Study
plus he sent me a condescriptive printout of the first study,
which is simply a printout of all the variables with some
statistical measure for each variable calculated.
He sent me copies of regressions he had run.
Q Did you receive at any point magnetic tapes,
underlying data for both studies?’
A Yes; cards for the first study and magnetic tape
for the second study.
Q Okay. Can you think of anything else that you
received?
A I think there was also something describing how
the data was stored on: the tape.
0 All right. To your knowledge, has all the
1§%
)
H
11
1 information the Attorney General has received from Professor
2 Baldus or the Petitioner generally Been "With respect to this
3 data, transmitted to you' --
4 A Yes.
5 Q -- and employed by you in your analysis?
3) A Yes.
7 Q What did you undertake to do by way of analysis
8 of this data that you received?
9 A When I first received the data, my first job was
10 to obtain computer facilities to analyze the data. And once
11 those had been secured and my files had been set up, the
12 first thing that I did =--
13 Q Let me ask you just one or two questions about
® 14 that before you go on. The computer facilities that you
15 obtained were where?
16 A They're the computer facilities for the Department
; 17 of Administrative Services for the State of Georgia.
: 18 Q Is that located in Atlanta?
19 A That's located in Atlanta.
20 Q Okay. And you said you set up a file. By that,
: 21 what do you mean?
: 29 A The tape and the cards were read in by employees
23 of the Department of Administrative Services' computer
24 systems specialistswho set up files for me on their computer,
on their IBM computer system. And then they taught me how
25
=
F
O
R
M
74
0
0
7
0
0
2
P
E
N
G
A
D
CO
.,
B
A
Y
O
N
N
E
,
N.
J.
24
25
12
to use their facilities, and then I proceeded to analyze the
data.
Q So, in other words, these were not manual files
that you're talking about setting up within the computer
itself a file for this data?
A Yes.
Q Who did you work with at the Department of Admini-
strative Services?
A I worked with Freeman Smedley and Dave Beck.
Q What were their titles, if you know?
A I wouldn't know precisely what their titles were.
Q Did they assist you with anything beyond filing
the information in the computer and teaching you how to use
the computer?
A They did not do any statistical analysis. Their
job was to help me in running whatever analysis I wanted to
work on; providing me with the job control language, making
it as easy as possible for me to run my programs. So, they
handled all the systems parameters, and I did all the
statistical analysis.
Q Okay. I didn't mean to interrupt you, but you
said after you had set up the file and you'd been taught how
to get access to it by these two gentlemen, what did you do
then?
A I believe the first thing that I did was I first
-
F
O
R
M
74
0
0
7
0
0
2
P
E
N
G
A
D
CO
.,
B
A
Y
O
N
N
E
,
N.
J.
24
25
13
wanted to get a feel for what the data was. And to do that,
I think I started with the Procedural Reform Study because
it was a smaller study and less variables. TI think I printed
out a cross tabulation of all the variables that had been
defined by Professor Baldus -- not the original questionnaire
variables but the subsequent variables.
0 Let me stop you a minute. You said all of them
had been defined by him. In what place or --
A As part of the data that he sent me. He sent me
all the variables as they were defined, and they were on the
cards.
Q Okay.
A So, I had all of his variables and I cross tabulate
that with race of the victim.
MR. BOGER: Did we receive -- perhaps this is a goo
moment to mention, Ms. Westmoreland, what we didn't do last
evening after we left the deposition. We did go over to
Professor Katz's office; we spent, perhaps, an hour in his
office. Professor Baldus was present, I was present, Ms.
Westmoreland, Dr. Katz, Co-counsel Robert Stroup, and Sam
Laufer.
Professor Baldus did most of the work reviewing the
computer runs and other information that is in your office.
We asked you for and you kindly gave us permission to take a
number of those computer runs, a number of your published and
-
F
O
R
M
74
0
0
7
0
0
2
P
E
N
G
A
D
CO
.,
B
A
Y
O
N
N
E
,
N.
J.
24
25
14
unpublished papers, and some other documents that were of
interest to us in the office for provincial copying and quick
return to you. I think we made an inventory of all the items
that we took and left the inventory with Ms. Westmoreland
or, perhaps, with you, Dr. Katz.
Is there anything else that you'd add to that, for
the record?
MS. WESTMORELAND: I think the only thing that I
would add is that I think that we attempted to provide
everything that had been requested. I don't think anybody is
making an absolute statement that everything has been
provided, but to the best of our knowledge, we have provided
everything that was requested.
MR. BOGER: You've been very cooperative on that.
I mean, there are problems that I trust you'll run into when
you see us next week in Syracuse when you've got this much
information and this much paper and finding the needles, if
there are needles, in the haystacks that are lying all over
the floor. But you were very helpful to us.
Let me just ask, though, Dr. Katz, do you know
whether you provided us with that cross tabulation between
the variables that Dr. Baldus had identified in the --
THE WITNESS: TI made it available to you, but I
don't believe Professor Baldus was interested in it.
MR. BOGER: Off the record just for a moment.
-
F
O
R
M
74
0
B
A
Y
O
N
N
E
,
N.
J.
0
7
0
0
2
P
E
N
G
A
D
CO
.,
24
25
15
(Whereupon, there was a brief
interruption of the record.)
BY MR, BOGER:
Q Dr. Katz, now you've indicated that you did these
cross tabulations between the variables identified by
Dr. Baldus in his Procedural Reform Study and race of victim.
What, then, did you do?
A Then, I think I ran some correlations correlating
these variables with race of the victim, ‘race of the defendant;
I think some outcome variables such as whether a penalty trial
was held or not; whether defendant was given a death sentence
or not.
Q Let me ask you: did you do any runs to: determine
whether race of victim was correlated with sentencing outcome?
A I'm sure that I didn't run that specifically, but
in the correlations that I ran I'm sure I did have. that
correlation many times in my printouts.
Q I'm sorry; how would that appear?
A Well, race of victim, the variable for race of
victim would be compared with the variable for a death
sentence outcome. And what we get on the printout is a
a correlation.
Q Between those two?
A Yes.
Q Okay. What else did you do in your -- you talked
-
z
ui
Zz
2
Oo
>
«
m
CO
0.
;
P
E
N
G
A
D
16
about other correlations, and then I interrupted you. Let
me ask you to go on.
A Okay. Than I proceeded to break down the cases
not just by race of victim and the variable, but by defendant
victim racial combination by variable, and then by sentencing
outcome where my alternatives were either a life sentence,
no penalty trial; a life sentence, penalty trial; a death
sentence; life sentence, unknown if there was a penalty
trial; and death sentence as a result of a multiple or a
second or third penalty trial.
Q If you would, explain to me a little further what
was this last set of runs that you did?
A What this run gave me were the counts. It first
took all the cases and classified them by their defendant
SAC END
victim racial combination -- that's a Rey. variable as
defined by Professor Baldus, which is either black killing
a white, a white killing a white, a black killing a black,
or a white killing a black -- by the sentencing outcome,
I'm sorry, by the variable, and then for each sentencing
outcome.
So, for example, for all the life sentence no penalt
trial cases, I would have a cross tab of the defendant victim
racial combination by the variable. So --
Q And that's by each variable identified in the
study?
o
5
-
z
wl
z
z
0
>
[4
a
P
E
N
G
A
D
C
O
.
,
17
A And I did that for each variable that I was given
by Professor Baldus, and then I did that for each different
set.
Q As you may have overheard as my expert and I
confarred on this matter, Dr. Katz, we're interested in
knowing not simply whether you did cross tabulations as
you've described, or. whether you did them controlling for
background factors, other legitimate factors, was any control
built in when you did those cross tabulations?
A No, I did those cross tabulations over all the
cases.
Q But not only over all the cases, did you do them
with other factors held constant as you looked at each of
the variables?
A I only had three levels. I didn't have any further
I didn't do four way cross tabulations or five or six or
seven way cross tabulations.
MR. BOGER: I'm going to have marked a blank piece
of paper as Petitioner's Exhibit 5, and I'd like to hand that
blank sheet to you, Dr. Katz, and ask you, if you would, to
sketch out on it something that would replicate the kind
of tables that you generated.
(Whereupon, Petitioner's
Exhibit 5 was marked
for identification.)
w
on
5
-
A
=
z
w
=
z
o
>
<C
m
P
E
N
G
A
D
C
O
.
,
18
MS. WESTMORELAND: Why don't we go off the record
for a minute or two while we do this.
(Whereupon, there was a brief
interruption of the record.)
BY MR. BOGER:
Q Dr. Katz, if you wouldn't mind showing us what
the rows and the columns stand for on that Petitioner's
Exhibit 5. 'Let me ask you several questions about Petitioner
Exhibit 5. At the top of the page when held horizontally,
you've got something that says, "XXX = 0 Life Sentence No
Penalty." Does this imply that you completed tabulations of
this sort for each of the sentencing outcomes?
A Yes.
Q So, that, for the category or the variable of
armed robbery, there would be four tabulations representing,
or three or four representing death sentence, life sentence,
no penalty, no life sentence?
A I think I had five. I had life sentence, no
penalty trial; life sentence, penalty trial; death sentence;
life sentence unknown with penalty trial; death sentence
for multiple penalty trial.
Q All right. And you did this for each of the
variables identified by Dr. Baldus?
A Yes.
Q This tabulation would not correlate sentencing
J
)
p
-
z
wl
z
=
Oo
>
<<
@
P
E
N
G
A
D
C
O
.
,
outcome with defendant victim racial categories; would it?
A The sentencing outcome is fixed in that table, so,
I'm not trying to Look at any correlation between sentencing
outcome or anything.
Q All right. ‘What additional kind of analyses did
you perform on these data beyond those that you've just
described to us?
A I went to the regressions that Professor Baldus
has provided for me and analyzed his regressions. I
attempted to try and obtain his regression results. I also -
Q Let me ask you about that. In other words, you
sought to replicate his regressions?
A Yes.
Q How did you do that?
A In the variable definition list that he provided
for me, he indicates the regression equation. So, from that,
I could determine the dependent and independent variables
that go along with the regression equation. And so, I ran
my own regression analysis.
Q Did those runs correspond with Professor Baldus's?
A For the Procedural Reform Study; yes.
Q Did you run similar comparisons or replications
of Professor Baldus's regressions in the Sentencing and
Charging Study?
A Yes.
i
<
-
z
I" :)
z
o
>
<
m
P
E
N
G
A
D
Co
O.
,
20
And did they also conform?
They didn't conform exactly.
Do you recall in what particulars?
+
J
©
EE
E
E
S
©
I believe the way he ran his regressions were
different from the way I ran the regressions.
Q In what particulars?
A I believe he weighted the cases.
Q And you ran unweighted cases?
A I ran unweighted cases.
Q So, did you even attempt to try the weighted
regressions?
A By that time, no. Up to this point, no, I haven't
tried the weighted regressions.
Q Which regressions in Professor Baldus's Charging
and Sentencing Study did you try to replicate; do you recal
A I need a copy of the variable list.
MR. BOGER: Let's go off the record.
(Whereupon, there was a brief
interruption of the record.)
THE WITNESS: Do you have copies of my regressiong
MR. BOGER: I'm not sure which regressions you're
talking about, Dr. Katz.
THE WITNESS: The regressions that you've been
looking at.
MR. BOGER: The material we've taken from you, we
wo
o
=
z
ui
z
z
o
>
<<
m
P
E
N
G
A
D
C
O
.
,
21
have available to us; yes.
THE WITNESS: Okay; I need to look at my regression
printouts.
MR. BOGER: Off the record again.
(Whereupon, there was a brief
interruption of the record.)
BY MR. BOGER:
Q You've located a set of printouts which has a
number on it, 1, on a sheet of notebook paper. And actually,
on your printout it has Regressions, III.
A I've located one regression, Post-Furman,Overall
Discrimination Index, C481IND4.
Q All right. And you ran that regression?
A Yes.
Q You'd indicated, I think, in your testimony --
let me just try to reorient me in time and space. You said
that you ran or attempted to replicate Professor Baldus's
regressions. in the Procedural Reform Study, and you were abld
do so; is that correct?
A Yes.
Q So, then we went to the Charging and Sentencing
Study, and there you said you ran some regressions and you
were not able to replicate Professor Baldus's --
A At that time.
Q What do you mean by at that time?
-
z
ul
z
z
Qo =
<
m
C
O
.
P
E
N
G
A
D
22
A Well, I did this analysis many months ago. This
whole process of analysis is a learning process. And you
try things, and as you do more and more you learn more and
more about the data.
And also, the information that was given to me by
Professor Baldus wasn't completely detailed in exactly what
everything meant. It was sort of in computer language, which
takes awhile for me to assimilate that great amount of
information. So, when I ran these regressions earlier, it
was very early, I was still in the process of trying to
duplicate what he had done to see what the numbers looked
like.
Q Right. And is the suggestion that since you have
learned more about this particular data set,that it's
possible that running the regression would produce different
results?
A Running the same regressions will give exactly
the same results.
Q Let me ask you this question. How did you make
your selection what variables to put in those regressions
that you ran?
A Professor Baldus indicated what variables he used
in his regressions, and I only tried to duplicate his experi-
ment.
Q This was in the Charging and Sentencing Study?
=
z
i"
z
z
o
=
Ld
a
P
E
N
G
A
D
C
O
.
,
23
A Yes.
Q The regressions that you've got that you've just iden-
tified for us, do you recall where Dr. Baldus or Professor Ba
had identified the variables that he used in that regression?
A Yes.
6) Where?
A In the variable that was provided to me.
Q Dr. Katz, are you confident that the regression tha
we're talking about that you've indicated, I think it's C48lI
is from the Charging and Sentencing Study, not the Procedural
A No, the Procedural Reform Study.
Q Let me clarify things again. I thought you had
testified just a few minutes ago that when you ran re-
gressions to replicate Professor Baldus's work in the Pro-
cedural Reform Study, you did replicate?
A Yes.
Q And your problems in replication occurred in the
Charging and Sentencing Study?
A Yes.
Q IT asked you if you could indicate or locate some of
the regressions that you ran where you could not replicate
from the Charging and Sentencing Study; is that what we've
got before us?
A No, we have the Procedural Reform Study.
Q I see. So, we've got one where you did have the
ND4,
24
% 1 | list of variables from Dr. Baldus. And this is a regression
2 | that you were able to replicate.
3 Do you have. any of the regressions from the
4 | "Charging and Sentencing Study available where you weren't
5 | able to .replicate the results?
6 A Yes.
7 +30 Can you locate some of those?
8 A I will try. Yes; here is an example, the first
og | regression on that printout.
10 MS. WESTMORELAND: Can we identify the regression
11 | by number or something for the record somehow?
12 MR. BOGER: You're referring to Regression DTSNIDXZ?
13 THE WITNESS: Yes.
14 | BY MR. BOGER:
15 Q And so, you employed the variables that Dr. Baldus
16 | had listed for that particular regression, and you were unable
17 | to replicate his --
18 A I was unable to exactly replicate his numbers.
3 19 Q What do you mean by exactly?
: 20 A It comes close.
: 21 Q How close; do you remember?
: 22 A Here's a copy of the printout.
23 Q That's part of the material you provided for us?
24 A Yes.
25 Q Do you have other examples -- I mean, how many
o~
-
z
ui
z
z
oO =
<
m
P
E
N
G
A
D
.C
O.
,
10
11
25
regressions did you run to try to replicate the Charging
and Sentencing Study?
A I believe they're contained in that printout.
Q The printout you have identified as Regressions I?
A yes.
Q So, ‘this comprises your efforts to replicate
Professor Baldus's Charging and Sentencing regressions?
A I think there is a little bit more for the
Charging and Sentencing Study.
Q You handed me a Printout Regression II. We've
talked a little about the replication efforts -- excuse me;
are there any other, documents that contain such regressions?
A Regressions concerning the Charging and Sentencing
Study?
Q And your reference to the replication.
A Just the Charging and Sentencing Study?
Q That's ‘correct.
A Not that I'm aware of; other than what may be in
my discard pile due to errors.
Q All right. Errors, by that you mean your-errors
or errors in your computer people running --
A Well, it wasn't always clear what the dependent
variable was supposed to be. I was given a list of about
thirty possibilities. And so, sometimes I would try a
dependent variable which I thought the index was trying to
“i
&
yo
ul
z
z
Oo
>
<
om
P
E
N
G
A
D
C
O
.
,
26
use, and it wouldn't come close.
Q Is this the part where you're talking about in the
sense of with large data set of learning what you've got and
learning how to deal with it?
A Yes. I was not given a three hundred page detailed
report as to exactly what everything meant. I was given
pretty much all this work computer short-cut formulations.
'Q Let me ask you; this is an area I'm not familiar
in. When one is all of a sudden presented with this large
kind of data set, is there a lot of trial and error in
figuring out how things relate to each other and how you
organized the dataset for analysis?
A Well, the way the data was given to me, once we
had received all the documents, Mr. Dumich said, "Here."
He gave me a stack of documents.
Q And these are the --
A Are documents that Professor Baldus made available
to us. He said he didn't know what they meant. And he said,
"You make .sense out of this." And so, I had to read all the
documents and try and figure out what was contained in there
and how they related to the information, the data that was
provided to me by Professor Baldus. And that isn't something
that you're going to completely appreciate in a short period
of time because I did not have the opportunity to ask
Professor Baldus specifically, what does this mean, why did
-
z=
wi
z
z
o
>
L-4
m
P
E
N
G
A
D
C
O
.
,
you do this; how come you have this instead of this. Those
kinds of things only come to you after spending a lot of time
playing with the data, trying to understand the data; reading
his documents; trying to figure out what all the variables
mean.
Q So, in a sense; it's not like ~-- and I'm trying to
get what you're saying -- it's not like reading a set of
instructions on how to put together a toy for your child or =~
I mean, in other words, you've got to work with the data to
figure out --
A Plus, there's no toll free number you can call to
ask for help. So, it takes time to fully analyze everything
in understanding the study.
Q Well, have you brsued any further analyses beyond
the ones that we've. now talked about?
A Yes.
Q Can you describe some of those?
A Well, In terms of the regressions -- but before I
get back to the regressions, I discovered after running all
of these cross tabs that it seemed that everything was, all
the data was there, until I recognized that there were some
variables, the race of the victim variable, the penalty
trial variable, where there were unknowns. And then I
realized there seemed to be too much data here for the number
of unknowns and some of what I thought were pretty, supposedly,
0
3
z
ui
z
z
oO
>
«
m
P
E
N
G
A
D
C
O
.
28
basic and obvious types of variables.
Q Let me stop you and just ask you, what do you mean
by too much data for the number of unknowns?
A Well, in running all the cross tabs with his
original data, there was not a hint of any indication that
any of the values were unknown. All the values were defined
either to be "0" or "1" values for the attribute variables
or total values for the total variables. For example, for
armed robbery, the variable was "ARMROB', and as the
Si variable, "0" is meant to indicate armed robbery didn't
occur at the time of the murder; "1" is meant to indicate
armed robbery did occur at the time of the murder.
There was no hint that anything, that there were
any cases where it may have been unknown whether armed
robbery occurred at the time of the murder.
Q So, at some point in your review of this data, you
discovered that there were possible ‘unknowns?
A I discovered as I .reviewed his questionnaire more
closely that in the items for the possible responses for a
particular question, there was a choice, "0", which indicates
that the answer to that question was unknown. And then I
started looking at the printout of the raw data, and I saw
lots of cases where the items were indicated to be unknown.
And they were coded to indicate unknowns.
Q And once you discovered this, what did you do?
=
3
z
ul
z
Z
[]
>
be
0
P
E
N
G
A
D
C
O
.
,
23
A Well, once I discovered this, then I realized that
all the cross tabulations that I had run up to that point
are seriously affected by these unknowns that are present
in the data.
Q In what sense: are seriously affected?
A Well, there's a difference between a variable
being unknown such as whether =-- there's a difference in
whether it's unknown whether an armed robbery occurred or
whether an armed robbery occurred or didn't occur. And what
had happened, is I discovered after I had looked at the data
and then looked at how he formulated his variables, he had
ignored the unknowns and he had coded them for the variables
in such a way that if unknowns were indicated as an item for
the questionnaire for a particular question, he coded that
variable to be automatically "0" regardless of the sense of
the variable.
Q What do you mean by the sense of the variable?
A Well, if it's a positive variable such as, "Did a
serious contemporaneous offense occur", okay, coding --
Q Let me ask what you mean by positive there?
A Positive there means that a "1" would indicate
that a serious contemporaneous offense had occurred; a "0"
would mean that, essentially, no serious contemporaneous
offense had occurred.
Q Okay.
~
3
z
ui
z
z
[e]
>
<C
[+
o
o
Q
<<
o
z
[}
o.
30
A A negative type variable would be a variable like
no contemporaneous offense; that's a negative variable.
Q Just the way the variable itself is structured?
A Right. And so, a "1" there would indicate there
was no contemporaneous offense, and a "0" would indicate a
contemporaneous offense.
Q I didn't mean to interrupt; I just wanted to
clarify that sense of the variable.
A All right. So, that changed the whole picture.
So, all the cross tabulations that I had run on all of those
variables were now subject to problems because of the
unknowns.
Well, then, I investigated how much unknowns are
there; what's the depth of the number of unknowns; does it
seriously affect the analysis that has been run to this
point. And I found that in the Procedural Reform Study,
there is a tremendous number of unknowns in very important
categories. For example --
Q Well, I hope, I mean, I don't want to get too
deeply into this because I do want to go over your report,
and I realize some of this is contained in your report.
This is very helpful and we do want to get back to it, but
I'm really interested now not as much in where your analysis
takes you, but what operations you've performed. So, what
you've told me is that having identified this as a problem,
AS
3
Zz
ul
z
=
Oo
>
<
m
[o]
Oo
Q
Pe
[L
Zz
uw
a
31
you have done some evaluation of --
A After I identified this as a problem, then I looked
at those regressions again and -- first I tabulated all the
unknowns and I found that there was a great, a tremendous
number of unknowns throughout the dataset.
Q You indicated that some distinction, it sounded
like, between the Procedural Reform Study and the Charging
and Sentencing Study. You said that certainly in the Pro-
cedural Reform Study there were a tremendous number. Did
you find fewer in the Charging and Sentencing Study?
A I found a great number in the Charging and Sen-
tencing Study, however, there are some advantages to the
Charging and Sentencing Study in the structure of the
questionnaire design which can allow you to do some legitimatq
analysis.
Q In other words, it reduced the number of instances
where there were unknowns?
A Where you can account for the number of unknowns
and,essentially,clean them up and not have them affect your
counts or your estimates, which is not really possible with
the first study, the Procedural Reform Study.
Q Having made this sort of distinction, then what
did you go on and do now?
A In the Procedural Reform Study?
Q Yes.
[1
3
z
ui
z
z
o
>
=
@
P
E
N
G
A
D
C
O
.
,
32
A Okay. Well, then, I also recognized, looking at
the questionnaire, that there were problems with the way
he designed his questions. And the problems, I couldn't
see, given the great number of unknowns how, really, I could
do any legitimate analysis or get any useful information from
the study.
The major problem is in this Foil Method of
designing the questions. For example, Question 26 on the
Procedural Reform Study asks for whether serious, to indicate
the serious contemporaneous offenses that occurred at the
time of the murder. And there is a list of -- I don't
remember the number of items -- a good many items. And the
data collector was instructed to fill in those numbers that
applied to this question, and he was given three foils to use
to fill in numbers to represent what serious contemporaneous
offenses had occurred at the time of the murder. If no
contemporaneous offense had occurred, there was a number
to indicate that. If three items were not needed totally,
then there was a filler number for Filler Code 9 for those
items that aren't needed. If, however, this area was
unknown, there was the Code "0". And so, the code that
would indicate that this was unknown: whether a serious
contemporaneous offense had occurred or the number of serious
contemporaneous offenses had occurred would be of the form
no99n,
o 5
"
z
ui
z
z
lo]
>
£4
m
[o]
[6]
a
<
0)
z
il
a
33
Okay. Now, the problem with that is, first of
all, with having only three items, if you have a case which
has more than three contemporaneous offenses, there is no
way the data collector can indicate that. He can only choose
three of .them, and maybe there was some convention as to how
he decided which three he would choose. But that in itself
adds additional error to the study, because not all the con-
temporaneous offenses are necessarily captured by that
question.
There are cases where more than three contemporaneo
offenses had occurred. And generally those are cases where
the death sentence wasgiven. And so, the effect of having
only three filler items or three items foils is to really
make death sentence case -- well, really the more aggravated
cases not look as aggravated as they would.
Q All right. I do, as I said, want to get back to
this analysis that you've given us into the flaws of this
procedure, but if we could, just come back a little and talk
about what operations you performed; describe these diffi-
culties --
A This is important because I need to describe this
to explain how, why I proceeded the way I proceeded.
Q Well, I'm definitely interested in finding why
you proceeded. If you can tell then, now, once you've
identified this kind of problem, what you did; what was your
34
rR A 1 | next series of analyses -- I don't mean analyses of the
2 | problems of the report-- but did you do any further runs or
3 | regressions yourself to deal with this?.
4 A Well, that's my next part.
5 Q All right; fine.
6 A And we again have to go back to the questionnaire
7 | design.
8 The second problem is the following: Given that
9 | there are so many unknowns and, for example, for the guestion
10 | that I indicated, whether a serious contemporaneous or indi-
11 | cating serious contemporaneous offenses had occurred at the
12 | time of the murder, I think there are 111 out of 607 cases
13 | where it was indicated that it was unknown. Now, given that
14 | there is such a high percentage of unknowns, it's very
15 | possible that the rest of the cases that are coded really
16 | represent incomplete data. For example, let's suppose we
17 | had a coder who found himself in this situation: He recognizes
E 18 | that armed robbery did occur, and so he codes the code for
19 | armed robbery. But he's not sure if kidnapping occurred.
20 | Well, there's no way he can get that information across,
given the flexibility he's given with those foils. You know, [if
B
A
Y
O
N
N
E
,
N.
J.
©2
76
2
-
N
m
h
22 | he codes "0", then that "0" can't be necessarily interpreted
P
E
N
G
A
D
C
O
.
,
23 | to represent kidnapping.
24 Q I understand what you're --
2b A So, the point is that these 111 unknowns for this
~
<r
=
z
wi
=z
z
o
>
4
@
[<}
(9)
o
<<
[4]
Zz
wl
a
35
question is possibly only the tip of the iceberg for the
amount. of information that's unknown. I can't look at these
items and be certain that if something is not coded,that it's
definitely known that it didn't occur. It may be coded, not
coded because it was unknown that it didn't occur and the
data collector was unable to tell me that that was unknown.
So, I couldn't wash those variables out.
Q All right. With those problems, then, what did you
do?
A With those problems, I realized that I couldn't
even get reliable counts on the outcome for all the variables
that Professor Baldus had defined because of these problems. An
it then dawned on me that any real concentrated analysis of
the study was really being wasted because if I don't have
reliable data, the rest of it, any analysis on garbage will
naturally be garbage.
Q Well, now, when you say don't have reliable data,
you mean you don't have complete data?
A Well, complete data; yes.
Q Let me ask you: have you done anything to suggest
that the data that is present is unreliable?
A Yes.
Q And what's that?
A Okay. Another test that I performed was, I compare
cases between the two studies that were common. I compared
d
2
3
Zz
ui
z
Zz
Oo
>
«
m
P
E
N
G
A
D
C
O
.
,
them by case number, and I found there were 360 cases that
had the same case number.
I then manually checked each case that had the
same case number to make sure that they had the same name
also. And I found there were --
Q When you say manually checked you --
A I read; I had a list of the case numbers and names
for the first study; case numbers and names for the second
study; and I cross checked them to make sure that the name
was the same... There were about five or six cases where they
weren't identical. But most of them were of the sort where
I could readily assume that there was some typing error that
had occurred and, so, they probably were the same case.
Q This was the five you mentioned; most of the five?
A Most of the five, yes. For example, there was
i eve n
one case where the name was indicated as &ureh, and the other
Leach
name was keeeh. So, I assumed it's one, it was the same guy|
whatever his name is.
There was another case where the name, one name
had III after it where the other name didn't. So, I assumed
that was probably the same case.
However, there were two cases where there was no
similarity in the names.
Q Do you recall what cases those were?
A l:have -- 1 can’ find it. ‘I'm trying to find it.
0)
a
3
z
wl
z
z
o
>
Led
m
o
0
a
<<
t)
2
[*%)
a
24
25
37
MS. WESTMORELAND: Let's go off the record for
a minute.
(Whereupon, there was a brief
interruption of the record.)
BY MR. BOGER:
Q Dr. Katz, have you found a document that indicates
what those mismatched names were?
A Yes, they're Case Number 635 and Case Number 964.
Q What were the names in those cases?
A For Case 635, the first study has James L. Graves,
and the Charging and Sentencing Study has Sam Wilson. And I
concluded that they were, that the names were different, even
though it may be that there's an error. I threw them out
from consideration of my matchings.
~The second one, Case 964, in the first Study that
wes aE outa Stephens, and the second study was James
Daniels.
Q All right. And you did the same thing; you
excluded them out --
A OI lianation. And so, I had 360 cases; I was
left with 358 cases. Then I merged the two, I merged the
Procedural Form Study data with the master file of the
Charging and Sentencing Study for ‘these common 360 cases.
And then I began comparing variables that were common to
both studies.
p 1 Q How did you conduct that comparison?
2 A I ran cross tabulations where I would have, I would
3 | compare, for example, long term hate, which is an item that
4 | is codeable from the first study, which is, I believe the
5 | variable is hate on the first study, with the item from the
6 | master file which indicates hate, and found the number of
7 | counts for each of the different cells.
8 Q All right; did you do those with every one of the
90 | common variables?
10 A I don't think so.
11 Q How did you select among the variables those on
12 | which you do this cross calculation, cross tabulation?
13 A I went through the mode of variables; the aggra-
14 | vating variables, which were item variables; I believe I
15 | went through some mitigating variables; I think I also did
16 | some precipitating event variables.
17 Q And how did you make your selection from among
£ 18 | those categories of variables?
19 A I tried to take most of them, but I didn't have
20 | any specific pattern. I would probably go back and do
some more testing because it's all set up,and it would be
BA
YO
NN
E,
N.
J.
@7
ud
2
No
—
22 | very easy for me to find more mismatches.
23 You have to understand that in running this sort
P
E
N
G
A
D
C
O
.
,
24 | of thing, Professor Baldus had three years to perform his
25 | study and do his analysis. I was only given his data, I
of
5
Zz
u
2
=
o
>
«
)
Oo
[s)
a
<<
©)
z
Ww
a
39
believe, in early February. I'm sorry; I think it was late
January, but it was only accessible to me in early February.
So, I only had three or four months to run all of these
different tests. : And what I really tried to do is really
get to know what all the data was, see what the problems
were, and not necessarily be completely comprehensive for
all the different problems.
Now, I hope for my final report I will be able to
indicate all the different mismatches and be a little bit
more complete. But I've really been just trying to keep my
head above water in getting all the things that need to be
done analyze within the time limits.
Q I was going to ask you a little later about that;
whether you had plans for a further report beyond the one
that we've received from the State, and I take it your answer
is yes?
A Yes.
Q Do you have any idea when such a final report will
be forthcoming?
A Well, a lot of it is going to depend on the
information that I receive after we take Professor Baldus's
deposition, on the information contained in his working
draft which I have recently received.
Q Let me just clarify for the record, the working draf
is the June 17, 1983 draft that we provided the State on the
<
a
3
z
v
z
z=
le]
>
<C
[ee]
P
E
N
G
A
D
C
O
.
,
40
17th of June as far. as you know?
A I don't know when it was provided.
129 1% eluctant to mark a copy at this point; I
suspect it may:be marked at the deposition of Professor
Baldus. There's more paper than we need in this deposition.
But let me show you a document that's on the desk here that's
titled "Working Draft: Death Sentencing Patterns" -- excuse
me. It's titled "Discrimination and Arbitrariness in
Georgia's Capital Charging and Sentencing System of Prelimin;
Report" with a date of June 15, 1983. Is that the document
to which you're referring?
A Yes, plus, some supplementary documents which are
also working drafts, "Death Sentencing Pattern of Fulton
County, Georgia‘ with special reference to the case of Warren
McCleskey"; and "Appendix D: Evidence of the Liberation
Hypothesis and Prosecution and Jury Decision-making."
He also provided at that time other documents
which were copies of what I had received earlier, or I
perceived were copies of what I received earlier. So, I
did not get any additional copies.
Q All right. Let's go back now, again. We talked
about your cross classifications of the two files and so
forth that had common cases. What other kinds of analyses
have you pursued of Professor Baldus's data?
A In what area, just --
Ay
41
% 1 Q '. At all. We've gone through a number of the steps
that you took to perfect your analysis. I can ask you --
3 | let me ‘ask you a few questions. Have you sought to measure
4 | the affect of the race of defendant variable in sentencing
outcome in anyway?
5
6 A Yes.
; Q How have you done this?
8 A What I have done is, I've done two experiments. And
g | these are recent experiments because only recently have I
noticed that race of the defendant has been raised as an
10
n issue. Professor Baldus previously, I believe, in his first
i report, he made a comment to the effect that he didn't think
13 there was discrimination based on race of the defendant;
$ 14 only discrimination, substantial discrimination, based on
5 | race of the victim.
16 And so, I concentrated my study on race of the
17 | victim. So, recently I have begun or I have been continuing t-
18 | I had looked a little bit at race of the defendant type
1g | information. And I'd run cross tabulations and comparisons
20 | ©f aggravating mitigating variables that compares cases
broken down by race of the defendant for the different
99 | sentences.
Q All right. Have you tried to not simply correlate
P
E
N
G
A
D
C
O
.
,
B
A
Y
O
N
N
E
,
N.
J.
@©
7e
uz
Z
the two, but to measure what the race of the defendant
o5 | effect is on the sentencing outcome?
L
42
A As in, say, regression type analysis?
2 Q That's right.
3 A No.
4 Q Let ‘me ask the same thing about race of victim
5 | variable; have you done any --
6 A In a regression type analysis?
7 Q Right.
8 A Trying to determine what are the =-- trying to explajin
9 | race of the victim by different factors, different variables,
10 | independent variables?
1 | Q Trying to measure the race of the victim effect,
12 | if any, on the sentencing outcome.
13 A The sentencing outcome is a dependent variable --
@® 14 0 Right.
15 A A bunch of other variable with race of the victim
16 | at the end?
17 Q That's right.
18 A No.
19 Q Have you attempted any other measures apart from
; 20 | regression of the racial effect, race of defendant effect
: 21 | or race of victim effect on sentencing outcome?
22 A Other than comparison within Sentences?
23 Q Right.
24 A No.
25 Q Have you sought any additional data beyond that
3
z
ui
2
z
[eo]
>
<
om
P
E
N
G
A
D
C
O
.
,
43
provided to you by Professor Baldus on patterns of charging
and sentencing in the State of Georgia on homicide cases ?
A For use in, for updating my files?
Q For any purpose relative to this hearing.
A I have not sought to obtain data to be used in my
analysis.
Q Whether you've sought it or not -- perhaps I asked
the question the wrong way. Have you received any additional
data beyond that provided to you by Professor Baldus on
charging and sentencing --
A Such as what?
Q For example, other datasets collected by the State
of Georgia officials) --
A No.
Q -- Or any other professors or researchers who are
studying charging and sentencing patterns?
A No.
Q So, the results of your analyses are confined to
the data that you received from Professor Baldus?
A correct.
Q And let me ask you: Are you seeking now for use at
the evidentiary hearing to be held in this case any additional
data from other sources?
A Such as what you described earlier, datasets?
Q That's right.
i
[5]
ot
3
z
ui
z
z
o
>
<<
@
P
E
N
G
A
D
C
O
.
,
Aa
A
Q
question.
some of the unknowns that you've identified in the case that
Professor Baldus has included in his Study?
A
the unknowns to then run analysis. I've left Professor
Baldus's Study intact with all the unknowns, and I've tried
to analyze as best I can what he has presented to me. I
haven't tried to update it or correct it in my analysis.
Q
gentlemen you've mentioned with the Department of Admini-
stration in the conduct of your analysis?
A
Q
A
Q
officials, employees with the State of Georgia in the Depart-
ment of Corrections, the Department of Parole Probation?
A
Q
A
Q
That can be used in analysis of this issue?
That's right.
No.
Have you -- well, let me ask a slightly different
Have you sought data that would fill in or complet
I have not sought data for the purpose of completing
Have you worked with anyone else beyond the two
Yes, I collaborated with Dr. Roger Burford.
Anyone besides Professor Burford?
No.
Have you consulted with others apart from legal
In aid to analyze Professor Baldus's Study?
Yes.
No.
For any other purposes relative to this hearing?
[»)
od
2
z
wi
z
z
oO
>
<
@
P
E
N
G
A
D
C
O
.
,
®
45
A No. Let me qualify: There is one statistician in
OfFender
the Department of BPefender Rehab. who is interested in
Professor Baldus's Study and has been given access to look at
the data. I don't believe he's done it, but I also don't
believe that he's going to be a factor in this case.
Q When you stated that he was given access to the
data, do you mean -- in what form has he been given access?
A Well, he was given a computer number that allowed
him to run programs and analyze data.
0 In other words, all the computer cards and the
computer tape which you have to set up your file in the De-
partment of Administration's computer is accessible to =--
A To him; yes. And I do not know if he's run anything
or what he's run, and do not, to the best of my knowledge, I
do not believe that he is going to be used in this case.
Q All right. But you've not consulted with him?
A I have not consulted with him.
MS. WESTMORELAND: Let me for the record -- to my
knowledge, our office has done nothing toward requesting his
assistance in this matter or anything of the sort.
MR. BOGER: His assistance; let's get this indi-
vidual's name out so --
THE WITNESS: His name is Tim Carr. He was the one
that gave you the data, I believe. And sO, it seemed only
appropriate that he be allowed to analyze and so forth. He rg WW
46
pe 1 quested it, and no one seemed to have any objection to it.
> | BY MR. BOGER:
3 Q When you say he's the one who gave us the data, what
4 | do you mean, exactly?
5 A From what I've been told, Professor Baldus requested
gs | a sizeable amount of data from his department, and he and the
vi research committee approved that request. So, he presented
Professor Baldus with the data, and later on the data was
8
g | sent back to us. Now, I don't believe he gave all the data
10 | on it; Professor Baldus collected additional data. But some
11 of the data, I believe, was given by him.
12 Q You've indicated you did have assistance in your
13 analysis from Dr. Burford. What was his role in the analysisp
. 14 A Well, I discussed with him different aspects of the
15 analysis. He had copies of the preliminary report and the
16 questionnaire and I tried to provide him copies of all the
17 | documents that I received. And as I did my analysis, we
18 | discussed the results and the implications, and I tried to obj
19 | tain suggestions’ from him about how to proceed further and would
20 | ask him questions, you know, to get some agreement that I was
Ul
doing a legitimate job in analyzing data. And I requested hi}
BA
YO
NN
E,
N.
J.
©
N
3
so | help for any problems he may see in my analysis so that I
could correct it. Plus, I provided him with two or three boxes
P
E
N
G
A
D
C
O
.
,
4] full ‘Of printouts to give him a chance to go through the data
05 | himself, look at some of the cross tabulations, look at some of the
©
3
Z
ul
z
z
Oo
>
<<
o
P
E
N
G
A
D
C
O
.
,
9
10
47
correlations, and give him an opportunity to get some idea
as to what's in the data. And then I would update him as
I did more and more experiments with the results.
He also read my preliminary report and we discussed
certain aspects of it. TI have been using him to help vali-
date all my procedures.
Q So, in a sense, you have been the principal
researcher or analyst on the matter?
A Yes.
Q He's stood in the role of a consultant to you or
a collaborator?
A A collaborator; yes.
Q You drafted the report, is what you indicated?
A Yes.
Q He reviewed the report?
A Yes.
Q And you discussed the report?
A Yes.
Q Do you know whether at this point the report in
the form that we're going to discuss it is a document that
he concurs in in his basic modes of analysis and conclusions?
A Yes.
Q Let me ask you just to make sure that I haven't
improperly narrowed the questions that we've been talking
about by improper use of language. I asked you if you had
i
z
wi
z
z
o
>
Ld
@
P
E
N
G
A
D
C
O
.
,
48
received data from any other sources; from professors or
from the Department of Offender Rehabilitation or elsewhere
in Georgia. Have you received information, as distinct from
data, relative to these cases?
A Yes.
Q what kind of information have you received?
A Well, because the cases are incomplete in all of
the information that was requested on the questionnaire and
because of other problems in terms of interpreting some of
the variables and degree of occurrence of variables and thing
I needed in several instances some information about what
really happened in the case, compared to what was indicated
by the codes on the variables.
Q And where did you obtain that information from?
A I obtained that information from Nicholas Dumich
and Ms. Westmoreland and William Hill on specific cases.
Q In what form did that information come to you?
A They described the case and the circumstances.
Q So, in other words, you would say what really happe
in the -- let's use an example -- Jack House case and they'd
say, "That was the case that occurred here in Atlanta where
two young boys were strangled and murdered." That's what
we're talking about by way of information?
A I think that information I got from a newspaper
article. But that's the kind of information that I would get
hed
G
bs
Q §
=
z
ul
z
=
Oo
>
<
m
oO
(S]
P
E
N
G
A
D
49
Q For the record, let me just clarify. The reason
I used that as an example is last night during our document
gathering you indfehted some familiarity with that case and
indeed said, "Oh, that's the case where two young boys were
strangled and raped."
A Well, I know I read a newspaper article on it, and
I probably received additional information from the Assistant
Attorney General and lawyers. I can't pinpoint exactly,
differentiate which information I received for that particular
case, but that is the kind of information that I was pursuing
to try to find out, you know, what was really indicated, what
really happened in the cases compared to what was coded in
the variables to indicate what happened.
Q Do you know what the source of information they
had?
A I believe they used trial transcripts, prosecutor
reports, whatever. I'm not completely sure as to what their
source of information is, but I believe it's reliable.
Q All right. In other words, you don't know what they
used. You didn't ask them does this come from a trial
transcript or a report. You relied on their knowledge of the
cases?
A Yes.
Q All right.
A For the most part, their usually experts on the case
&~
3
z
uw
z
z
o
>
<
om
P
E
N
G
A
D
C
O
.
,
50
having tried them for years and years and watched the
boxes fill up with brits on each individual case.
Q How often did this kind of communication of
information go on; is this an infrequent occurrence or does
this go on in a large number of cases?
A Well, I would say that I do not know the circum-
stances of all the cases. I would say I know the circumstanc
of maybe a very small proportion of the cases that I've
picked up in my association with the State Attorney General's
Office.
Q So, in other words, you're not talking about full
of information that would occur every day or two during the
last four months; this is more infrequent than that?
A Yes. Like, I do have some interest as to, for
example, the circumstances in. the McCleskey Case, or Mr.
Dumich would like to describe some of the gory circumstances
of some of the other cases. And so, I've learned about those
cases.
Or occasionally I had a question. There seemed to
be some problem with the information I was getting on a
particular case, and I asked him, Mr. Dumich, to explain how
can this occur. And he would look up and tell me what
happened. So, either it would substantiate the code that
was represented for that case, or it would indicate that ther
was some problem with the code for that case.
eS
51
pl 1 Q Did you have any other sources of information that
2 | you relied upon? You've talked about the Assistant Attorney
3 | General's, members of the Attorney General's Office; any
4 | other sources of information in your analysis?
5 A In terms of the facts of the case?
6 Q In terms of any other factors you were considering.
7 A Can you suggest some?
8 Q No, because I don't really know what I'm looking
og | for.
10 A I don't know what you're looking for either.
17. 1 Q All right; let me explain. You have been given
12 | information from Professor Baldus on the data. You've
13 | indicated you've received information from the Attorney
14 | General's Office on particular cases, all of which went into
15 | your evaluation of Professor Baldus's work and into your
16 | own analysis.
17 I'm trying to determine whether there were any
18 | other information sources -- I'm not talking about a casual
19 | look at newspapers -- but were you talking with people at
20 | the Department of Corrections; were you talking with people
in the Department of Offender Rehabilitation --
BA
YO
NN
E,
N.
J.
87
an
z
N
—
—
22 A No.
23 Q Were you talking with people in other Attorney
P
E
N
G
A
D
C
O
.
,
24 | Generals' Offices in other states?
25 A NO.
™
=
z
ul
z
z
o
>
«
om
P
E
N
G
A
D
C
O
.
,
52
Q Were you talking with prosecutors and other
representative of county government?
A No.
Q All right. Have you made any attempt to learn
whether there is other information available to the State of
Georgia that bears on the question of the charging and senten
patterns in the State?
A I don't believe I understand the question.
Q Have you made any attempt to determine whether
there are other data sources available to you, whether
computer banks or manual files or --
A I made no attempt to locate them or use them.
Q All right. Have you discussed the Baldus data or
the preliminary report at length with any professional
colleagues apart from Dr. Burford?
A Yes.
Q With whom?
A With fellow statisticians in the Quantitative Method
Department.
Q At Georgia State?
A Yes.
Any colleagues outside of Georgia State?
A I don't believe so.
MR. BOGER: Madam Reporter, do you have Petitioner’
2 from Yesterday. Off the record.
cing
53
& 1 (Whereupon, there was a brief
2 interruption of the record.)
3 | BY MR. BOGER:
4 Q I'll show a document we previously marked as
5 | Petitioner's Exhibit 2 and ask you, if you would, to identify
6 | that document?
7 A That's the first part of my preliminary report,
8 | "Analysis of the Procedural Form Study and the Georgia
9 | Charging and Sentencing Study."
10 Q And now I'll show you a document that's been
11 | marked Petitioner's Exhibit 4 for identification and ask
12 | you if you can identify that.
13 A This is the continuation of my preliminary report.
14 Q Does this report reflect the results of your
15 | analyses of Dr. Baldus's work?
16 A Yes.
: 17 Q As of what time?
! 18 A As of, I guess, the middle of May.
: 19 Q This is a time before or after you received
20 | Professor Baldus's working draft?
A Much before I received a copy of the working draft.
BA
YO
NN
E,
N.
J.
@7
a
No
—
22 Q Is your analysis of Dr. Baldus's material continuing
P
E
N
G
A
D
C
O
.
,
23 | in light of this subsequent draft?
24 A Yes.
25 Q So, as you testified earlier, we can anticipate
-
z
wi
z
z
Oo
>
«
«
P
E
N
G
A
D
C
O
.
,
54
some final report that will go beyond the conclusions of
this?
A Yes, you can. I expect to have a final report, yes
Q Let me, though, ask you -- we're going to spend
just a little time going through this preliminary draft today
If you would, turn to Page 1 of that document, the Introductiq
there. The introduction sets forth, I believe, your basic
conclusions; is that correct? Well, let me ask it a differen;
perhaps legally a better way. What does your introduction
attempt to do?
A My introduction outlines the two studies that I've
analyzed and I propose that the higher proportion of death
sentences that result in white victim cases as compared to
black victim cases is not due to a dual system based on the
race of the victim, but due to the substantially higher
level of aggravation and mitigation in white victim cases;
so, higher level of aggravation and lower level of mitigation
in white victim cases as compared to black victim cases.
Q And is that a conclusion that you presently agree
with?
A Given all the analysis that I've done to this
point, it seems clear to me that yes, that is the explanation
for the higher death sentence rate.
Q Okay. Let me just ask you a few questions about
the introduction, and we'll go into the more meaty part, which
55
% 1 | is your ‘explanation for your conclusions. The sentence
2 | midway down this paragraph in the introduction begins,
3 | "However, the key issue is whether the observed higher pro-
4 | portion of death sentences in white victim cases is the
5 | result of discriminatory practices as charged by Baldus."
6 | Do you see that sentence?
7 A Yes.
8 Q Do you agree that the data does show a higher
9 | proportion of death sentences in white victim cases?
10 A Yes.
11 Q That fact, at least, is not in contention?
12 A No.
13 Q You then say, "After analyzing the data in both
14 | studies, it is overwhelmingly clear.."; I'd like to break
15 | down this series of sentences into four or five different
16 | statements. "It is overwhelmingly clear that
17 | white victim cases are systematically more aggravated and
18 | less mitigated than black victim cases"; is that your
19 | conclusion?
! 20 A To this point?
21 Q Yes.
22 A Yes.
: 23 Q And second, again to this point, "It is over-
24 | whelmingly clear that this fact completely explains the
25 | difference in the death penalty rates."
‘
wa
=
=z
wi
=z
z
o
>
<<
m
P
E
N
G
A
D
Co
O.
,
56
A Yes.
Q It's the third, which is really the second sentence
now, it is overwhelmingly clear that "There is no dual
track in charging and sentencing as alleged by Baldus and
his coauthors."
Do you reject the idea of a dual track based on
the data you've analyzed?
A Since the sentencing patterns can be explained
completely by the higher levels of aggravation and lower
levels of mitigation in the white victim cases.
Q All right. And it seems to me the next sort of
conclusion that's contained in these sentences, that you can
prove that--""it will be proven using their own data that
prosecutors and juries base their prosecuting and sentencing
decisions on the aggravating and mitigating facts of the
case." Is that your conclusion, based on the data?
A Yes; based on his data, yes.
Q And you've testified that you don't have any other
data sources that you are relying on for that?
A No; simply his data.
Q Then, apart from the question about whether there
is in fact dual track discrimination, you also conclude that,
in this last sentence, "The conclusions reached by Baldus and
his coauthors are the result of faulty and incomplete analysis
of their own data"; is that still your opinion at the present}
L*J
4
z
ui
z
z
le]
>
<
[+
P
E
N
G
A
D
C
O
.
,
A Based on the information of the preliminary report.
Q Have you reviewed Professor Baldus's working draft?
A Yes,
Q Has your review to date altered any of your con-
clusions that we've just gone over?
A No; but I haven't spent enough time reviewing his
working draft. Again, it's a very substantial work. The
results contained in there for the most part are different
from the results contained in the preliminary report in terms
of what's been presented. So, I need more time to go through
that working draft before I can come to any real, definite
conclusions about the work contained in there.
Q All right. To date, you've seen nothing that
fundamentally alters your conclusions: of the preliminary
report?
A Correct.
Q We're going to talk a little later, I hope, early
afternoon, maybe later in the afternoon, about any work that
you that you may already have done on the preliminary draft.
Maybe it will help me in assessing the time we have to know
whether you've done any runs on the data that he's reported
in the preliminary draft.
A I have done no runs on the data; I expect to get
those runs when we visit Syracuse next week.
2 But you, yourself, have not made any effort to
~
[=]
=
z
w
z
=z
o
>
<<
m
P
E
N
G
A
D
C
O
.
,
58
independently analyze the tables or =--
A Or substantiate any of the information contained
init.
MS. WESTMORELAND: Excuse me. Could I let the
record reflect, I think we're referring to the working draft.
I think it got referred to as a preliminary draft. I think we
MR. BOGER: I'm sorry; we do need to keep that
terminology the same.
THE WITNESS: Yes.
BY MR. BOGER:
Q All right, Dr. Katz; let's move into the meat of
your preliminary report. The first category or heading is
something called "Questionnaire Design". Could you tell me,
briefly, what you attempted to say or you said in this sectior
on Questionnaire Design?
A Well, to start any statistical analysis, some
analyses work, the first thing that has to be considered is
how the data was collected; what were the tools, the instru-
ments that were used to collect the data, how much veri-
fication was there made to insure that the data is accurate,
these kinds of issues, because analysis of faulty data will
result in faulty conclusions.
In the Questionnaire Design section, I analyzed
the questionnaire designs of the two studies, the Procedural
Reform Study and the Georgia Charging and Sentencing Study.
59
. 1 The Procedural Reform Study Questionnaire design, I find, is
2 | faulty. It's not designed in such a way, given the actual
3 | codings that occurred, to allow an analyst to properly record
4 | the outcomes for each case of the relevant items.
5 To go to an example, maybe I'll try a different
6 | question. Question Number 30 --
7 MR. BOGER: Let us be clear, for the record.
g | Different question, you mean different than Number 26?
9 THE WITNESS: Right. The reason I choose 26 and
10 | 30 is I believe these are very, very important questions in
11 determining the sentence of a defendant.
12 In Question Number 30, the data collector is asked
13 to choose from a list of some number of items, approximately
14 | thirty-three or thirty-four, those aggravating and mitigating
15 | features of the effects. And he is given six foils in which
16 | to indicate the codes for the particular aggravating and
17 | mitigating features. For example, if the murder was bloody,
18 | then the data collector --
19 MS. WESTMORELAND: Excuse me; can we go off the
20 | record for a minute?
(Whereupon, there was a brief
B
A
Y
O
N
N
E
,
N.
J.
@r
ec
:z
No
s
r
22 interruption of the record.)
23 | BY MR. BOGER:
P
E
N
G
A
D
C
O
.
.
24 Q Dr. Katz, you were in the middle of explaining to
25 | us how Question Number 30 of the Procedural Reform Study was
“
4
oi
=
z
ul
z
z
Oo
>
£4
om
[e]
(&]
Q
[4
Q
z
wl
a
60
structured. Can you continue that description?
A Yes. If the murder was bloody, then the data
collector would be instructed to indicate 07 as one of the
six foils for Question 30. Or if the murder was luring
ambushing, lying in wait, where the defendant was luring,
ambushing or lying in wait, then the data collector would
be instructed to fill in Item Number 10. So, the data
collector would go through this list and for each item
decide whether --
MS. WESTMORELAND: Excuse me; we'll go off the reco
(Whereupon, there was a brief
interruption of the record.)
BY MR. BOGER:
Q All right; continue if you would, Dr. Katz.
A So, if the defendant committed the murder luring,
ambushing, lying in wait, then he would fill in item, the
Code Number 10 as one of the items in the foils.
So, after going through each of these items, he
would decide whether it was an actual aggravating or mitigatii
feature of the offense and if it was present in the offense,
he would be instructed to indicate its code as one of the
foils.
Now, there have only been six foils provided to
list all of the aggravating and mitigating features of the
offense. There are cases, as indicated by the Charging
rd.
hg
=
z
ul
2
z
o
>
<<
o
C
O
,
P
E
N
G
A
D
61
and Sentencing Study when I did some cross classifications,
where up to fourteen aggravating circumstances applied to
that case. A data collector on reading information of the
file for that case would have quite a dilemma. He only has
six foils in which to represent fourteen aggravating circum-
stances. The end result, then, is that some of the aggra-
vating circumstances ,and perhaps mitigating circumstances,
would be omitted from the list.
Since most of the circumstances represented in --
sorry —-- most of the features represented in this question
are aggravating circumstances, it's very likely that most
of those omitted aggravating features would never show up
anywhere in the analysis. There's just no way for the data
| collector to indicate the additional aggravating circum-
stances.
The result of this is that cases where, for
example this particular case where there was fourteen aggra-
vating circumstances --
Q Let me just stop you for a moment. I just want to
clarify -- we're talking about a case where you say there
are fourteen aggravating circumstances, and yet what you're
telling us is that there are foils only enough for six. How
do you obtain the information of this case with fourteen
aggravating circumstances?
A According to the Georgia Charging and Sentencing
=
z
ul
z
z
o
>
<C
om
P
E
N
G
A
D
C
O
.
;
62
Study, that case was given fourteen aggravating circumstances
Q I see. So, what you've done is taken that, the
number of aggravating that you find for that case and --
A Yes; we will get to that item later in the report.
Q Let me just clarify for the record, on Page 3 of
your report you are discussing this problem to some extent,
and you talk about there are in fact cases such as Numbers 26
266, and 267 where 13 or more distinct aggravating circumstan
listed as options for Question 30 can be identified.
Are those the cases you have in mind?
A Yes.
Q All right.
S fe
te
A Okay. The end result of only having six foils
that severely aggravated cases such as Case Numbers 265, 266,
and 267 are not given all the aggravating circumstances that
they deserve, Therefore, in any analysis of the data over
all the cases in the Procedural Reform Study, these three
cases do not look as aggravated as they really are.
Q Do you suggest that that introduces some sort of
bias?
A Yes. Furthermore, in later analysis I found that
in most cases when there was significant understatement of
the aggravating circumstances because of the restriction to
six foils, these were white victim cases, and generally they
were death sentence cases or life sentence penalty trial case
Ie
ces
[2
]
.
[%
B
A
Y
O
N
N
E
,
N.
J.
©
C
F
3
0
2
-
P
E
N
G
A
D
C
O
.
,
So, the problem with this question is that not enough room
has been created for a data collector to indicate all of the
aggravating circumstances, and this is probably one of the
most important features in determining whether or not the
defendant will be given a penalty trial or even a death
sentence.
Q That statement is based on what your analysis of
other --
A That statement is based on my own general knowledge
of =- I'll retract that statement. It's not ~~
Q I'm not trying to trick you at all; I'm just trying
to ~-- is that based on some data in Professor Baldus's Study?
A No, that can be substantiated that the higher the
sentence, the more are the aggravating circumstances ; the
higher the sentence meaning the one from life sentence, no
penalty trial; life sentence, penalty trial; and death senteng
But at this point, I haven't really created the foundation to
say that, so, we'll wait until later when I can give you all
the numbers.
Q Okay. I just wanted to know whether that's some-
thing you had already done. Okay; continue if you would. I
didn't mean to interrupt.
A Okay. So, in any analysis of this data, some of
the most aggravating cases, which I found to be death sentence
cases and found to be white victim cases ,are not represented
ee.
~
=
z
wi
z
z
(o]
>
<<
[4]
P
E
N
G
A
D
C
O
.
,
64
to be as aggravated as they truly are.
Then, there is the additional problem with the
general foil design in trying to determine whether an cutcome
actually occurred, or did not occur, -or is unknown to have occu
For example, if a data collector went through all of these
items and decided that some of the items applied, such as
bloody and luring, ambushing, ‘lying in wait, but was unsure
about other items such as multiple shots to body, defendant
actively resisted arrest, there is no way that he can
communicate that uncertainty about these last two items
given the foil method. Therefore, although he can indicate
yes, the murder was bloody, by indicating a 07 as one of the
foils and yes, there was luring, ambushing, lying in wait,
which would be code number 10 as, perhaps, the second foil,
there is no way he can give us any information about whether
these other items are known or unknown. They may be Known
not to have occurred, or it may be unknown whether they
occurred. So, it's impossible just given six items from
this questionnaire to determine that.
Q Does the foil method taint all of the questions in
the Procedural Reform Study?
A Yes.
Q So, all of the questions here are subject to the
same problems you've identified in 26 and 30?
A Yes.
red.
~
-
5
z
m
z
z
o
>
4
m
P
E
N
G
A
D
C
O
.
,
65
Q Now, does the Georgia Charging and Sentencing Study
have the same problems that you've just identified?
A Okay. The Georgia Charging and Sentencing Study,
which is a later study, improves on the design considerably
except --
Q Before we get to the except, in what way does it
improve on the design?
A It improves on the design in that specific items
are questioned rather than a group or a list of items being
given where the collector is asked to choose two or three
or six that apply.
For example, if we look at the comparable question
for aggravating circumstances in the Georgia Charging and
Sentencing Study questionnaire which is from Appendix C of
the preliminary report and look at Question 47, it's broken
down in terms of Part A and Part B. Now, Part A is again
using the foil method. So, it is possible that there are
some omissions, unknowns contained in Part A which I am
unable to determine because of the nature of the design of
the question. This is unfortunate because again this is an
important aspect of the offense, and it would be difficult to
separate out whether an outcome was known not to have occurred
or unknown whether it occurred.
Q In other words, your testimony is it's important
to determine not simply whether there was torture, but the
~N
0
3
z
Z
z
z
Oo
>
<
om
P
E
N
G
A
D
C
O
.
,
[@
3
aN
Aortvve
kind of touxrture that was involved.
A No, not whether it was a torture. If the data
collector does not indicate torture, then two things could
have occurred; either torture didn't occur or the data
collector doesn't know whether torture occurred or not. So,
it's the same criticism as the Procedural Reform Study
Questionnaire for those foil design questions. Okay. And thi
is another important aspect of the data.
However, in Part B, for what I think are termed
minor aggravating circumstances to some extent -- It's just
that Part B circumstances, the design is a significant
improvement over the foil questionnaire design. In Part B
each particular item is given a variable, and the data
collector is allowed to indicate not only whether that item
has occurred, but whether it has occurred, his level of
confidence in the occurrence of the item. And also, he is
allowed to indicate whether he is unable to classify it as
occurring or not occurring or whether that item is unknown.
For example, the first item which is 176 is the item
Homicide Planned for more than five minutes.
Now, after reviewing the file if the data collector
decides that the homicide indeed was planned for more than
five minutes and was expressly stated in the file, then he
would code a 01. If, however, the fact that the homicide
planned for more than five minutes was not expressly stated
on
Le)
of
-
z
w
z
Zz
[eo]
>
Ld
m
P
E
N
G
A
D
C
O
.
,
in the file but simply suggested, but not specifically indi-
cated, then he would code a 02. If the homicide was not
planned for more than five minutes, then he would leave a
blank. And if he was unsure whether the homicide was planned
for more than five minutes or not and had no information or
was unable to classify the item, then he would indicate a "U"|
Now, this is a superior method of collecting
information because for each item we can focus in on not
only whether it occurred or not, but whether it's unknown.
And so, any analysis with this variable allows us to know
the extent of information for each particular case.
0 Do you recall how many questions in the: Charging
and Sentencing Study reverted to the foil method?
A I believe there were three. The two major
questions, unfortunately, was 47, Part A, Special Aggravating
Features of the Offense and -- well maybe -- I guess there's
more than that. But for the significant ones, Question 29,
"Did the homicide occur while the defendant was engaged in
commission of another offense; whether or not the defendant
was charged and convicted of the offense; whether a con-
temporaneous offense occurred at the time of the murder, is
also used in the foil method.
Q Those are the only two major guestions that use
that method?
A A minor one, I believe, is in -- well, there are
PN
-
Zz
[7]
Zz
z
o
>=
<<
om
P
E
N
G
A
D
C
O
.
,
68
some minor questions on;Question 41,"Where did the homicide
occur, but =—
Q Let me read to you on Page 4 in the first paragraph
of your Report, Petitioner's Exhibit 2, what sounds like a
summary of what you've just told us at some length. "This
format overcomes!, this format meaning the format of the
Charging and Ssntencing Study, "..overcomes the major problem]
encountered with the foil method in the Procedural Reform
Study although the questionnaire still uses the earlier foil
design for two important questions namely 29 (contemporaneous
offense at the time of the murder) and Question 47 Part A
(special aggravating features of the offense).
That's, in effect, what you've told us here?
A Yes.
Q That's still your conclusion?
A Yes.
Q Now, below that discussion, you talk about a second
weakness of the questionnaire design. Would you elaborate
on: that for us a little?
A Okay. In many cases, murders are committed not by
a single individual, but by gangs of individuals or by co-
perpetrators. And the method of treating this situation was
that each copertetrator would obtain all the aggravating
circumstances of the offense and whatever contempraneous
offenses had occurred at the time of the murder and, I guess,
Ul
69
® 1 the victim mitigating circumstances and his individual
2 | defendant mitigating circumstances.
3 Now, there are some cases where the role of the
coperpetrator in the murder was minor. However, by looking
5 | at just the raw data for the variables concerning the aggra-
vating factors in the case, it can look like a heinous crime.
7’ | There are some questions, starting on Page C 22, 48B which
8 | tries to pinpoint to some extent what the role of the
9 | defendant was, giventhat there were copertetrators. However,
10 | IT do not believe that this is sufficient to pinpoint exactly
11 | what aggravating features of the offense were committed
12 | directly by the defendant, how much information and intent
13 | he had for each aggravating circumstance, et cetera.
®
14 My own general knowledge suggests to me that
15 | prosecutors in deciding whether to seek the death sentence
16 | would look at the actual role that the defendant played in
17 | the crime and although an accessory, perhaps, if he did not
kill the victim, the prosecutor would attempt to specifically
19 | know what his role was before making decisions about what
20 | sentence to prosecute for.
Q This is an extremely basic question, but I just
BA
YO
NN
E,
N.
J.
#7
00
2
N
o
h
22 | want to clarify this. The general assumption of what you're
P
E
N
G
A
D
C
O
.
,
23 | saying,I take it,is that aggravating circumstances do make a
24 | case more aggravated. Is that a plausible way of looking at
25 | reality?
0
a)
z
ul
z
z
Oo
>
<
0
P
E
N
G
A
D
C
O
.
,
70
A I would accept that, yes.
Q And your criticism of the question that you were
addressing just now about coperpetrators is that the certain
Charging and Sentencing Questionnaire insufficiently reflects
differences among coperpetrator aggravation levels?
A Yes. What is available, and I've researched this
aspect of it to try and see if I can‘somehow adjust the
information to account for coperpetrators, the first set of
questions deal with the acts of violence by the defendent.
And the possibilities are either the defendant committed or
participated in the act that caused the death of the victim;
the defendant did not commit the act but committed acts of
violence against the victim; or the defendant did not commit
the act that committed acts of violence in the same incident
as against a person or persons other than the victim whose
death is the basis of the defendant's prosecution; or the
defendant committed no acts of violence.
Now, in each of those circumstances or each of
those items, it is possible that a case would be aggravated
even though, perhaps, the defendant committed no acts of
violence or if the --
0 Excuse me. You mentioned you had researched that.
what do you mean by that?
A I had run some cross tabulations, and I'd looked at
role of coperpetrators in the different sentencing to try and
the
o
o
4
wl
Zz
=z
Oo
>
Ld
m
P
E
N
G
A
D
C
O
.
,
71
figure out what the effect is on the data as an attempt to
further clean up the data in my analysis.
Q And what have your results been of that research
of those runs?
A Specifically, I cannot see any way, given the
questionnaire and the way the questions are and given the
level of detail for each aggravating circumstance, to separatg
out plausibly the effect of coperpetrators.
Let me -- I'm trying to get to some of the problems
in trying to:do that:
0 Fine.
A Okay. For example, if the defendant committed the
murder, which is 193, that may be cause for him to obtain the
death penalty given the aggravating and mitigating circum-
stances. —-
0 Just for the Court Reporter's benefit, 193 refers
to what?
A The item.
Q All right; Item 193 in that questionnaire.
A Then Item 194, where the defendant did not commit
the act but committed acts of violence against the victim, it
is possible that the acts of violence committed against the
victim by the defendant were so heinous, that although he did
not actually cause the death of the victim, he played a major
role in the death of the victim.
3
z
ul
z
z
o
>
<
@
P
E
N
G
A
D
C
O
.
,
712
So, it wouldn't be sufficient, then, to just look
at cases where it was indicated that Item 194 applied because
there are relative degrees of aggravation that could apply.
And there are death penalty cases where the defendant did not
commit the act but committed acts of violence against the
victim.
Then 195, Defendant did not commit the act but
committed acts of violence in the same incident against a
person or persons other than the victim whose death is the
basis of the defendant!s prosecution, that is possible in the
sense that the defendant may be the leader of the gang of
coperpetrators instructing others to commit murder, committing
heinous crimes, and although he doesn't do any aggravated
crimes to the victim or commits acts of violence against the
victim, he does commit acts of violence against the others.
And the same kind of situation can occur for Item
Number 196 ,where the defendant committed no acts of violence.
So, the defendant may have been the agent to hire someone to
kill the victim or something.
So, I couldn't see any plausible way of trying to
separate out the different cases based on those four items.
0 I'm not sure at this point I follow: you. All I
think I've heard you do with 193 to 196 is sketch out in other
terms what kind of defendant activities these four items could
identify.
é 1 A Right.
2 Q Is there any problem with that?
3 A Well, my point is that I claim that there are not
4 | enough details given for each item as it applies to the
5 | defendant. All coperpetrators were given the same aggravating
6 | circumstances despite their involvement. It was this --
7 0 In other words, this sufficiently differentiates
g | among defendant involvement acts of violence, but when it
9g | came time in other sections of the questionnaire to talk about
10 | an aggravating circumstance, these distinctions were not
11 | honored; is that correct?
12 A Well, let me bring up a case in point, and I think
13 | I used this case in the study; it's the McCleskey Case. And
% 14 | my information is, from Mr. Dumich, that the circumstances
1s | in the McCleskey case were that McCleskey acted on his own
16 | to kill a police officer. That there were three coperpetrators
17 | in the back, one of them being David Burney. And they had no
18 | role in the murder, they were part of the armed robbery or
19 | the contemporaneous offense.
20 Now, in the first study, the Procedural Reform Study,
I went through those two cases and looked at the different
BA
YO
NN
E,
N.
J.
©7
00
2
N
n
h
oo | aggravating and mitigating features and so forth. And I
found that for the McCleskey Case, he was given five statutory
P
E
N
G
A
D
C
O
.
,
23
24 | aggravating circumstances. He was given B-1, B-2, B-4, B-8,
25 and B-10. However, David Burney who was in the back room,
74
1 | who had conceivably no involvement in the murder, was given
2 six statutory aggravating circumstances. He was given B-1,
3 B-2, B-3, B-4, B-8 and B-10.
4 Now, 1f we compared those two cases, it looks like
5 | there's discrimination of some sort. Here is someone with
6 | six aggravating circumstances in a similar pattern of aggra-
7 | vating features compared to McCleskey who had five who got
g | the death sentence. But the key is that these aggravating
9 | circumstances don't really apply in this case because he
10 | didn't commit the murder; he's only an accessory. So, that
11 | explains why his sentence is not as severe as McCleskey's
12 | sentence.
Ji
13 Q All right. Do you think, are you suggesting there':
14 | anything systematic about the biasing of the overall results
15 | of the studies because of the coperpetrator problem?
16 A Yes.
17 Q Can you explain?
18 A Okay. Let's suppose we run regressions. In
19 | regressions we're trying to explain,as Professor Baldus has
20 | done in some of the ‘earlier regressions, to sentence outcome
by these different variables. Now, what it is really trying
BA
YO
NN
E,
N.
J.
87
80
2
-
Fo
PM
20
04
N
a
m
d
22 | to do is give him a set of variables, is trying to weight
P
E
N
G
A
D
C
O
.
,
23 | each variable so that the predicted outcomes are equal to
24 | the actual outcomes. So, it's trying to weight these variables
25 | as best as we can so what you predict will occur is actually
3
z
In
Zz
z
Oo
>
<<
o
C
O
.
.
P
E
N
G
A
D
75
what occurred. And that's how the process is modeled.
Well, here you have what might be referred to as
an outlier in a sense.
Q Pardon?
A An outlier, which means a data point that doesn't
really apply, because David Burney has six statutory aggra-
vating circumstances, and all of these other aggravating
circumstances that McCleskey has which makes him really in
the regression look worse than McCleskey. But he only has a
life sentence, no penalty trial. McCleskey has a death
sentence. So, the regression has a very difficult time in
resolving this and trying to allocate the different weights
for all of the different aggravating factors.
Q All right. Let me ask you one or two follow-up
questions. Your suggestion there is that there may be some
difficulty in constructing the proper outcome or having a
regression that's appropriate to the underlying data.
But the question I asked you is whether you think there's
any systematic bias introduced by this kind of problem, co-
perpetrator?
A I believe that there is some systematic bias because
and this bias is in terms of the race of the victim, the white
victim cases, because those are the cases that generally in-
volve coperpetrators. Those are the cases that generally
involve contemporaneous offenses. So, those are the cases
Bh
A
PA
3
z
ul
z
z=
[e]
=
<
@
P
E
N
G
A
D
C
O
.
,
76
most likely to be affected by this kind of problem.
Q Have you done any analysis of whether in fact
coperpetrators are more frequent in white victim cases?
A Yes.
Q What were your results?
A They are more frequent.
Q We've talked a little about questionnaire design,
Dr. Katz. Are there any other major points apart from the
nes we've discussed that you have about the deficiencies in
questionnaire design in these two studies?
A I believe those are all the points that I've un-
covered to this point.
Q Well, is there anything in the nature of the
materials that you've been given that would bear on the
questionnaire design?
A I don't believe so. But remember, now, as I get
more and more into the study, more and more things occur to
me, I learn more about the study. After taking Professor
Baldus's deposition, I expect I'll know a lot more about the
study, and that might reveal other problems in the guestion-
naire design.
Q I didn't mean for that question to foreclose you
forever after from saying anything else.
A Fine. There may be some more problems; I just
haven't realized them at this point, given the level that
ay
z
Ww
z
z
o
>
<
)
P
E
N
G
A
D
C
O
.
,
77
I'm at in the analysis.
Q So, the longer you work with this questionnaire,
the more things you may recognize about it, changes?
A That's possible, yes.
Q Let's look at the third area that you have discusse(
in your report, which is the Analysis of the Unknowns in the
Procedural Reform Study and the Georgia Charging and Sentencit
Study.
What is the thrust of your criticism in this
section of the report?
A That both studies have a great deal of unknowns,
and they severely affect the analysis or the potential for
analysis of these studies.
@) Excuse me. You have two sections, I note, one
beginning on Page 4, that talks about the analysis, the
unknowns. In the second section that begins on Page 7 of
Petitioner's Exhibit 2, it talks about the effect of an
unknown value: on the creation of variables.
A Yes.
Q Why don't we take those step by step, even though
I understand they're related.
A Sure.
Q In the section on the Analysis of Unknowns, what
do you conclude?
A There are a great deal of unknowns in the Procedural
il
La
«
-
z
ui
2
z
[e]
>
<
0
P
E
N
G
A
D
C
O
.
,
718
Reform Study and the Georgia Charging and Sentencing Study.
And I give some evidence for the amount of unknowns. This
doesn't contain all the possible unknowns; this is just a
brief list of some of the more important items or guestions
where the unknowns occur.
Q We talked yesterday, I think, and you told us that
you had some other data on this, and I believe we asked if
we could see it.
A Yes.
Q On Page 5 your analysis and below, Table I begins,
let me quote, "In his analysis, Baldus recoded the five
cases where the race of the victim is unknown by giving white
defendants white victims and black defendants black victims.
Since the key issue in his study concerns the alleged dis-
crimination based on the race of the victim, arbitrarily
assigning races to victims where the actual race is unknown i
not a logical or defensible approach." Is that =--
A Yes, I still agree with that.
Q What should have been done in those cases?
A They should have been thrown out when considered
for the race of the victim because doing anything else is
just manufacturing data. And statistics isn't a tool in
where data is manufactured; it's the analysis of hard, correct
data in the search for relationships and information,
especially in a case of this kind where this is a critical
-
2
ul
z
z
le]
>
<<
m
P
E
N
G
A
D
C
O
.
,
72
issue. There are some situations where the analysis doesn't
have to be exact, where there's great room for error, where
that kind of assumption might be plausible because --
Q Can you give me an example?
A In a business situation where you're trying to
forecast sales and you're missing some unknown values. So,
you make a guess to fill them in to run your regressions.
0 How is that differentiated in the situation?
A Well, the impact of your decision will be relatively
minor if your results are off in that case, where the impact
of the decision here will be highly significant.
Q Are you talking about socially significant or ==
A Yes, socially significant here.
Q So, your criteria is that the more socially
significant a particular analysis is, the more hard the
data would have to be? I'm just trying to follow you.
A No. I'm a stickler for hard data as accurate as
possible and analysis of hard, good data for any statistical
analysis.
All that I'm saying is ‘that in‘ cases where the
results are not going to be used for important purposes, I
would not come down very hard on someone making rough esti-
mates and trying to get some information. But this is not
one of those situations. This is a situation where the
decisions and the results of the analysis will have a lot of
©
3
z
ul
z
z
Oo
>
<C
o
P
E
N
G
A
D
C
O
.
,
80
importance. And I think it's not appropriate in these cases
to start manufacturing data.
Q Let me just ask you at the bottom of Page 5 -- I
think I detected a typographical error, but I want to be
clear. "Again, the above .."; we're talking about something
above, "is an overstatement of the amount of data actually
collected. Table III counts the number of unknowns out of
a total number of 1082 cases." Is that a reference to Table
II that follows?
A Yes, that's an error.
Q Let's move on to the section beginning on Page 7
of your report concerning the Effect of Unknown Values on
the Creation of Variables. What is the thrust of your
criticism and your observations in this section?
A When Professor Baldus did his analysis in both
studies, he simply ignored the unknown values that had been
coded according to the variables. What he did was, he
systematically defined all unknown values for variables for
a case to be 0, regardless of the aspect of the variable,
whether it was positively defined or negatively defined.
So, he created some contradictions between cases where there
were unknown values.
And I indicate two instances. One is VMITCIR and
NOVMC, which are opposite variables. And so, the cases that
I indicate afterwards are cases where they were all coded to
81
2 And’ then the second one, which isn't a direct
3 | opposite but close, is SERCONOF and NOCONTOF. SERCONOF is
4 | a serious contemporaneous offense occurring, and NOCONTOF,
5 | no contemporaneous offense.
6 MR. BOGER: Off the record for just a moment.
7 (Whereupon, there was a brief
8 interruption of the record.)
9 MS. WESTMORELAND: I think we're ready.
10 | BY MR. BOGER:
11 Q Dr. Katz, you were explaining to us about these
12 | contradictions between outcomes and certain variables, and
13 | I asked the Court Reporter to go off the record while I
14 | explained to her what the variable titles were. Now, let's
15 | get back to your explanation, if we can.
16 A Will you ask the question, please, again?
17 0 All right. You've, I guess, described for us
18 | certain instances in which because of the coding pairs of
19 | variables appeared to wind up with contradictory results.
20 | How many such instances have you identified?
A I identified two in the report.
BA
YO
NN
E,
N.
J.
27
50
2
No
—
22 0 Were there any others that you've identified from
P
E
N
G
A
D
C
O
.
,
23 | your review of the questionnaire?
24 A Not that I remember at this point.
25 Q If you do, would you have a list of them somewhere?
Ie)
3
z
Ww
z
z
o
>
<C
I)
P
E
N
G
A
D
C
O
.
,
10
1
c
t
No
13
14
82
A No. I can advise you that you can look through the
variables set and look for -- define these things -- look for
variables that are defined opposites that have unknown values
Q But you've not performed that systematically?
A No, I've just chosen two for examples to show
partly the danger,if not obvious, of coding unknown values
to be 0.
Q Well now, let me go on. You state in the next to
the bottom paragraph on Page 7, "Perhaps Baldus is unaware
of the extent of the problem," the problem of unknown variab ls
Is that what you're speaking of?
A Yes,
0 "And believed, as he stated in the preliminary
brief..." ; what ‘are you referring to, in the preliminary
report there?
A Yes.
Q "That he had in fact collected over 250 aggravating
mitigating variables for each case. Nevertheless, this is
a critical data problem..". Is that still your testimony?
A Yes, it is 'a critical ‘data problem.
Q "And the manner in which Baldus has dealt with
the problem, even if he is unaware of its existence, raises
serious questions as to the validity of his analysis based
on this data." Do you confirm that?
A I agree with that.
a
11
0
@
R]
[w)
3
<<
“
z
ul
z
z
o
>
<C
m
P
E
N
G
A
D
C
O
.
,
83
Q What kind of questions does it raise; any ones
beyond the ones that you've mentioned here?
A In terms of his analysis?
Q Yes.
A Well, his analysis has unknowns coded as 0's. This
has a great impact on the regressions, where he is using
these variables to try and identify what variables are sig-
nificant in explaining the dependent variable, whether it
be usually sentencing outcome. Now, if he codes unknowns
to be 0, well there's no reason to believe in that case
the unknown actually occurred, that it was unknown. What he
doing is, essentially, he's trying to explain a dependent
outcome variable such as sentencing with variables that
aren't clean and won't necessarily reflect the information
in all ‘the cases.
And in regression, an error like that, especially
a wholesale error of that kind, really doesn't give any
validity to any of the regressions that are done.
Q Let me ask you this: If you were to delete all
the observations that had missing variables -- excuse me =--
yes, missing variables, would the resulting analysis, then,
be cured of its problem?
A I don't think so because given the great number of
unknowns, any kind of analysis of that sort out of 607
original observations, a great deal of these cases were throwi
is
o
-
Zz
ul
z
z
Oo
>
<<
om
P
E
N
G
A
D
C
O
.
,
84
out due to unknown values. It only takes an unknown in one
variable before the complete observation is deléted from
consideration.
@) You mean, that's the acceptable procedure?
A That's the generally acceptable procedure.
0 So, the objection to deleting all observations
with missing variables is not that there's anything wrong
with that procedure, but that when you're given the number of
observations in this study, I guess we're talking about both
studies, you'd wind up with too few observations to make
valid conclusions; is that your point?
A Yes.
Qf But if you had a thousand observations left when
you did that, then there's no problem of the sort that you're
describing; is that correct or not?
A 1,000 out of, say, the original 1,082?
Q Yes.
A I would concur with that; that would be true.
However, since part of your analysis is examining all
potential aggravating and mitigating circumstances, you're
extremely limited in the regressions you can run because of
the unknowns. For example, there's a variable -- let's go
with the first study -- for whether a serious contemporaneous
offense occurred or not. Okay. There are 1ll1 cases where
it's unknown, and other cases where I can't determine whether
od
-
z
ul
Zz
z
o
>
L- 4
m
P
E
N
G
A
D
C
O
.
,
85
it's known or not for particular offenses.
Okay. If we restrict ourselves to variables other
than those variables, then we're losing a very important set
of variables that could help explain sentencing outcomes. So
if you can end up with, say, 1,000 cases in running your
regression, then that would be meaningful, but you don't get
an opportunity to examine all the variables, 400 or so
variables, that Baldus claims to have collected and considered.
You can't consider all of them in a regression analysis
because of the level of unknowns.
Q What would the appropriate procedure be?
A At this point?
0 At this point.
A I cannot see how any multivariate techniques will
work; that all multivariate techniques involved specifically
having a big data base with good data and then trying to
get some information, some relationships between the variables.
Q So, it's just not acceptable, the standard, to play
with that many unknowns? In other words, multivariate
techniques are: just out when you've got --
A Yes, you don't have the data base to perform multi-
variate techniques.
Q And: that's -- this is Dr. Katz's view, or is this =--
A Another --
Q -—- a generally accepted view?
&
Q
o~
©
o
3
z
ul
z
z
Oo
>
<<
m
P
E
N
G
A
D
C
O
.
,
86
A This is a generally accepted view.
Q Okay. Well, would there have been an appropriate
procedure to do it from the beginning? In other words, if
you were designing a study of this sort, what procedures migh
you have employed in order to permit multivariate technigues?
A If I was designing the study from the
the beginning, what I would do is collect information from
the prosecutors, I guess from the prosecutors who tried the
cases, because conceptually they should have all the in-
formation contained or relevant to the case, even some in-
formation that probably didn't necessarily come out in the
trial. And they would be the ones who, even if they had
wrong information, if we're looking at the Georgia system
of sentencing and trying to second guess prosecutors and
how their deciding whether or not to send the case for
a ‘death sentence, we really have to do it given. the informati
they had available. So, my first thing would be, I would
collect data from the prosecutors.
0 I guess that would have to be contemporaneous data
following your observation; is that correct?
A Well --
0 In other words, if a prosecutor learns something
about a case at a point after he had made a decision to
charge it murder as opposed voluntary manslaughter --
A That information wouldn't be reflected; just the
on
87
p 1 information that he used to make his decisions about whether
2 or not to hold a penalty trial or to prosecute or to plea
3 | bargain or whatever.
4 Q So, in the system that you have in Georgia or any
5 | other state where the prosecutor is making a series of
6 decisions, then the. ideal data collection method would be,
7 | in effect, to go back to the prosecutor a number of times;
8 | go back when he's deciding what indictment to prefer; go back
9 | when he's deciding whether or not to plea bargain; go back
10 | again when he's deciding whether to go to trial on capital
11 charges or sentencing phase; then go back again when he's
12 deciding to appeal; is that --
13 A And then look at what information the jury has
14 | had in making their decision.
15 Q Okay. You really, in effect, need multiple data
16 | collection efforts for each case contemporaneous with the
17 | decision maker's knowledge at that point?
18 A Yes, we need to know exactly what information the
19 | decision maker had at that point if we're trying to model
20 | his decisions as a group.
Q Okay. And I take it from what you testified
B
A
Y
O
N
N
E
,
N.
J.
27
00
2
No
wa
ne
d
22 | earlier -- please correct me if I'm wrong -- that you would
P
E
N
G
A
D
C
O
.
,
23 | want a data entry form that would reflect both whether the
24 | prosecutor knew the information or whether he didn't know it
25 or whether it.was unknown. In other words, whether he knew
88
% 1 there was no armed robbery or he knew there was an armed
2 robbery or it was unknown to him. In other words, he would
3:1 need to. ——
4 A Well, the unknowns, I think, would then be none,
5 | 0, because his information, if his information is he doesn't
6 | know if an armed robbery occurred, then he should be making
7 | his decision based on the information that no armed robbery
8 | occurred.
9 Q I'm not sure I follow. In other words, unless he
10 | knows that an armed robbery occurred --
1: A He should be making his decision based on that
12 | :particular —-- Af he doesn't know that it occurred, then I
13 | don't believe he should be making decisions on whether it
14 | might have occurred. Part of his decision process should be
15 | no. Unless evidence is collected to indicate that an armed
16 | robbery occurred, you know, then he knows an armed robbery
17 | occurred.
18 For example, if the prosecutor doesn't know if the
. 19 | defendant had killed thirty seven other people in South
20 | Carolina, if he's unaware of whether that occurred or not,
he shouldn't make his decision based on that possibility.
BA
YO
NN
E,
N.
J.
07
No
—
22 His decision should be based on the information that he has
P
E
N
G
A
D
C
O
.
,
23 | about the particular case. If he doesn't know that he killed
24 | thirty-seven other people in South Carolina, then he can go
25 ahead and use that in his decisions.
89
8 1 Q So, the ideal information collection system would
2 include information in which the prosecutors, say, or the
3 | Juries had proper knowledge, but would exclude things about
4 | which they had no knowledge but maybe some suspicion?
5 A Well, I think they're directed or instructed to
6 | not include information that isn't presented in evidence.
7 9) You're talking about the juries now or the prose-
g |. cutorsias well?
9 A Well, I can't imagine, if I made a decision about
10 | something -- suppose I want to buy a car -- well, I'm going
11 | to have my information and based on that information make
12 | the.decision., Now, there's lots of unknowns, but I can't
13 | really use that unknown information to help my decision. I
a 14 | have to base it on what I have available. And in any decision-
15 | making process, there's all kinds of things that will be
16 | unknown, but you have to make your decision based on what you
17 | do know.
18 0 Absent the kind of data collection method we're
19 | talking about in which the, in effect, information is con-
20 | temporaneous with the knowledge of the decision maker and
unknowns really are not included in the analysis, is it your
BA
YO
NN
E,
N.
J.
27
20
2
N
—
29 | testimony that multivariate analysis of the overall pattern is
P
E
N
G
A
D
C
O
.
,
23 | inappropriate?
24 A Would you repeat that again; I didn't ‘get it.
25 2 I'm sorry. Absent the kind of data collection
-
z
w
z
z
Oo
>
«T
0
P
E
N
G
A
D
C
O
.
,
90
effort that you've described which includes collecting data
contemporaneous with the knowledge of the decision maker and
excluding unknown factors, would it be your testimony that
multivariate analysis of the overall charging and sentencing
system 1s inappropriate?
A I still don't understand exactly.
Q Well, let me --
A I think you're making it too broad.
0 Well, let me back up. You said that given the kind
of ‘data that Professor, ,Baldus collected in the way that he
had designed his questionnaire, that it was against standard
practice to. conduct multivariate analysis, and it was not
recommended or approved. But then we moved into a guestion
about how would you have done it, what method would you have
used. And you've described the method that you would use
and approve. If that method were used, multivariate analysis
would be appropriate; is that your testimony?
A Multivariate analysis could be used if the data
collected 1s accurate, correct data and you have complete
information about each case.
Q Absent that --
A If you don't have complete information, how much
information don't you have. Are you just missing a few
variables or ten percent of the variables or fifty percent
of the variables?
a
-
z
ul
z
z
o
>
<
I)
P
E
N
G
A
D
C
O
.
,
21
4) Do you have any kind of rough way in which to make
some judgments about how much has to be missing before multi-
variate analysis becomes inappropriate?
A I think you would have to look at the outcomes of
the regressions. I don't have any rough rule of thumbs to
what the proper percentages are. It would also depend on
which variables the information is missing on. Some of the
variables collected may have really no relevance to the sense}
So, whether you have complete information on that variable
may hot be important.
It's not a hard and fast rule that you can make
about if you have so much data, do multivariate analysis,
you take. out one more item for a case, then you can't use
multivariate analysis.
Q when you said you have to look at the outcomes,
I'm not sure what you meant by that.
A Well, the regressions, see how many observations
get thrown out; see what seems to be going on.
Q Based on all those factors --
A Yes.
Q -— considered, I guess, simultaneously, one makes
the decision about whether it's appropriate.
A Yes. Statistics is a science, but the analysis of
data is pretty much an art in trying to understand the data
in using these statistical techniques.
&
La]
z
uw
z=
z
0
>=
bad
m
P
E
N
G
A
D
C
O
.
,
92
You know, there are general rules about doing
certain things and guidelines as to what the appropriate
distributions are for the particular tests. And part of the
importance of an expert is identifying when to use certain
tests and when to recognize if problems can occur when you
use certain tests.
9) Let's turn to your next section which involves
Reliability and Consistency of Data Between the Procedural
Reform Study and the Georgia Charging and Sentencing Study,
beginning on Page:8 of Petitioner's Exhibit 2. What have
you tried to state in this section?
A Okay. Conceivably, the two studies are trying
to collect similar information on cases in the Georgia System.
TI found 360 cases where the case numbers matched, and 358 of
those in which the names also matched. So, that indicates
to me that those are similar cases that were considered by
both studies.
Then I compared items that were defined as
variables or items in the first and second study to see how
consistent they are over all the 358 cases. And I recoded
cases where the variable was unknown to be 0 to be consistent
with the way Baldus recoded it, and I used his pattern of
recoding the variables, and I defined a mismatch to be the
case when the eventual variable on the one hand in the first
study was a 0, and then in the second study it was 1, or
93
. 1 in the first study the variable was 1, and in the second
2 study the variable was 0.
3 So, when we're all finished, these are the off main
4 diagonal, the off main diagonal occurrences.
5 0 All right. And what did you find, having done thatp
6 A I found that there was a lot of inconsistency
7 | between the two studies.
8 Q Are all of the inconsistencies you found reflected
9 | in your report?
10 A No.
1" Q Where are any further inconsistencies reflected?
12 Fig ‘IT don't believe I tested for all possible incon=- ...
13]. sistencies. Buti I still may, you know, in the future, test
14 | for further inconsistencies between variables.
15 Q Well, ‘have you done further tests that are not
16 reflected here, though, in the cross.classification?
17 A There may be some other variables that I've already
18 | tested for that aren't reflected here, but I don't believe
19 “I've tested for all the variables. I:don't think.this is a
20 | list of all the variables that could be tested.on the table.
Q How did you make your choice of the variables to
B
A
Y
O
N
N
E
,
N.
J.
53
78
32
N
—
—
t
22 | cross classify?
motive
23 A I just started with .mede-af variables and went
P
E
N
G
A
D
C
O
.
,
24 | through some aggravating variables and some contemporaneous
25 | offense variables and some mitigating variables just to get
o~
-
z
w
z
z
o
>
Lod
m
P
E
N
G
A
D
C
O
.
,
24
25
94
an overall view of what the matches seemed to look like.
H) Let's look on Page 12 -- I'm sorry, it's going to
be Page 11 in your Table III in Petitioner's 2. At the very
top of the page, the Description, the first item is Number
of coperpetrators; the second column appears to be Number
of Mismatches, and your number there is 36; is that correct?
A That!s what I've indicated there.
Q And are the two variables that you have used to
determine whether mismatches occurred, W40 for the Procedural
Reform Study variable and LDF 192 for the Charging and
Sentencing variable?
A Yés, as far as’ I can recall.
Q And ‘so, by cross classifying the outcomes in the
different cases and those two variables, you reached that
number 367?
A Right.
0 Let's move to your next section beginning on
Page 13, which is the section on Evaluating the Data Analysis
by Baldus of the Procedural Reform Study.
Now, what is the substance of your evaluation
that you've set forth here?
A Okay. I first emphasized that given the
questionnaire design, there are tremendous problems with
the data. Furthermore, in the cross classifications I
indicate that the number of aggravating circumstances for
’
-
z
w
Zz
z
[eo]
>
<<
[+]
P
E
N
G
A
D
C
O
.
,
95
white victim cases is very likely highly understated in the
Procedural Reform Study. Still given that, I still tested tl]
data to see, or tested Baldus's preliminary report analyses
to see if they were valid. And that's what I review in
the section.
Q Okay. So, at the bottom. Of the page on 13, there's
a sentence that says, "Let me emphasize that these convention
do not follow generally accepted statistical"-- I guess it's
statistical” meant to be --"practices and the coding of
unknowns to be zero is in essence making up facts for the
different cases." You're not talking now about his analysis,
but about the matters you indicated above precede the
analysis, the way in which he's --
A Yes.
Q -- obtained the data and coded it.
A What I attempt to show in this section is even if
we accept all the conventions that Professor Baldus has used
in generating his variables --
2 And it's your testimony those do not follow the
accepted =--
A And they don't. But for now, I'm assuming that
we'll accept that and then just go on with looking at his
analysis.
0 I see; okay. Well, then, what's your conclusion,
accepting his conventions?
Le
Ul
©
«©
&
o
ia
z
wi
z
z
Oo
>
Ld
@
P
E
N
G
A
D
C
O
.
,
96
A Well, I analyzed the four methods he used to test
for discrimination for race of the victim.
Q Before you talk about the analysis, let me ask you
for the bottom line conclusion. The report appears to
contain one at the top of Page 14. Is it your conclusion
that there is "still no evidence based on this data that
any discrimination based on the race of the victim exists"?
A Just based on the Procedural Reform Study data?
Q Yes.
A Even using his conventions and so forth?
Q Yes,
A Yes.
Q It looks as if you go beyond that to say in the
second sentence, "In fact, this data even confirms the fair-
ness of the Georgia Charging and Sentencing system in assign-
ing death penalties based on the severity of the crime and
not arbitrarily or discriminatorily", I guess that's meant
to be -- a slight typo. Is that an additional conclusion
you've reached from this analysis?
A ¥rom this analysis, 1 think T'11l scratch that.
4, what do you mean?
A That given his analysis, what I can show is that
black victim cases are not discriminated against in sentencin
But I do believe that —--
Q Well, let me understand what you're saying here.
J «
~
3
5
z
[n)
z
z
Oo
>
<
m
P
E
N
G
A
D
C
O
.
,
97
In the preliminary report you say it confirms the fairness
of the charging and sentencing system. You no longer want
to stand behind that statement?
A Wait, no; the Charging and Sentencing Study is a
different study.
Q Your language says, "In fact this data even confirm
the fairness of the Georgia Charging and Sentencing system."
Does that refer to the actual real world system of charging
and sentencing?
A No; scratch that, that sentence.
Q Well, at the time you wrote that, what were you
saying?
A What I was trying to get across was that Professor
Baldus has made a claim that the system tolerates higher
levels of aggravation in black victim cases than white victim
cases before they are pushed along the sentencing system.
For example, he claims that black victim cases must have a
high level of aggravating circumstances before they will be
pushed onto a penalty trial or even the defendants pushed on
to a death sentence.
Okay. That is the .thing that his data in the
first study does not support.
Q So, that's your language, there's no evidence based
on this data that any discrimination based on the race of thi;
victim exists. That's what in effect you're saying?
Ul
Ul
of
3
z
ui
Zz
Zz
Oo
>
<
[+]
P
E
N
G
A
D
C
O
.
,
98
A Right. So, in terms of the fairness, more would
need to be done to completely establish that.
Q Professor Baldus's Procedural Reform Study, you
contend, does not provide evidence of discrimination; is that
correct?
A For race of the victim.
Q For race of the victim. But you now do not
continue the assertion on Page 14 that it provides affirmativ
evidence that there's no discrimination?
A Well, given all the assumptions that have been made
about the data, it would be meaningless even if it did becaus
the data.wouldn't be appropriate with all the errors in it.
Q But we've talked about the assumption at least .
on the bottom of 13, we said that we're going for the time
being assume that the conventions that he used are appro-
priate. And then on 14 you make a statement that the data
confirms the fairness. If we make those assumptions about
the convention, are you prepared to say that the data con-
firms the fairness of the system?
A I think that's too vague a statement.
0 It seems like the problem that causes you to
retract or to strike that statement is its vagueness?
A Yes.
Q Is there any more specific way that you can say
it which you would support?
U
29
% 1 A That there's no indication there's discrimination
2 | based on race of the victim where higher levels of aggra-
3 | vation is tolerated in black victim cases rather than white
4 | victim cases in determining sentences.
5 Q And beyond that, you're not prepared to go based
6 | on:the Procedural Reform Study?
7 A Not given the analysis that we have coming up.
8 Q Okay. 'Let me ask -- we're going to talk about
9 |' this: a little bit later, but do you have a different con-
10 | clusion when regarding the Charging and Sentencing Study?
11 A My conclusion for the Charging and Sentencing
12 | Study is basically there is no evidence of ‘any discrimination
13 | ‘based on race of the victim.
14 Q What about race of the defendant?
15 A I haven't found any evidence of that either.
16 Q SO =
17 A But there may be 77 other possible discrimination
18 | factors which I haven't considered. So, TI. .think din an
19 | instance like this, I need to qualify this as, in terms of
20 | the actual discrimination that's being looked at.
Q In other words, it's possible that there may be
B
A
Y
O
N
N
E
,
N.
J.
A
7
0
0
2
.
F
O
F
M
2
0
3
4
No
—
22 | other kinds of discrimination based on other factors in the
P
E
N
G
A
D
C
O
.
,
23 | system?
24 A I haven't tested for any of those, so I don't know
25 | what the answer is.
100
% 1 Q Okay .
2 A So, it's just the sentence is too vague, given
3 | the ‘analysis that I've done. And so, I retract it.
4 Q Let's go on, though, and ask you further -- I think
5 | I stopped you about asking for your conclusions first, and
6 | you were going to go through and tell us what analysis you
7 | had undertaken. Why don't we do that.
8 My conclusions, my recollection of the last five
©
minutes spent as we started on the section of evaluation of
10 | “the data of Dr. Baldus, you began to talk about the analysis
11 | you undertook, and I said before we get to that let's start
12 | with the recapitulation of your conclusions at the top of
13 | 14. .So,.1 in effect interrupted you. Let's talk now, if we
14 | can, about the analysis, your analysis of Baldus's analysis
15 | or your critigue cof Baldus's analysis.
16 MS. WESTMORELAND: Can I interrupt to ask if we're
17 | thinking about something that might take a little while to
18 | get into, we might want to consider breaking for lunch at
19 | this point.
20 MR. BOGER: Fine.
(Luncheon recess)
BA
YO
NN
E,
N.
J.
27
35
02
N
a
22 BY MR. BOGER:
P
E
N
G
A
D
C
O
.
,
23 Q Dr. Katz, why don't you tell us what evaluation
24 | you performed of Professor Baldus's data analysis.
25 A For the Procedural Reform Study?
[a]
Q
[5
™~N 5
3
z
ui
z
z
o
>=
Ld
m
P
E
N
G
A
D
C
O
.
,
101
Q Yes.
A Okay. Professor Baldus indicated four methods of
testing for discrimination by race of the victim. The first
method that he indicated was by comparing cases who had the
same statutory aggravating circumstance, breaking those cases
down in terms of white victim cases and black victim cases,
and then comparing the death sentencing rates between those
two groups. I believe I have an updated table of this on
Page. 31.
9) That's syour Table X?
A yes.
Q ‘And what did you do to evaluate that method?
A First of all, TI defined a list of variables that
encompassed most, if not all, of the important aggravating
and mitigating factors that could be applicable to whether,
to sentencing.
Q And how did you make the determination about which
factors could be important?
A I gave Mr. Dumich a list of all the variables
that Professor Baldus had defined, and I gave him a copy of
the list that I picked out of it and asked him if I had,
to his knowledge, obtained the important variables. And he
agreed that most, if not all, the important variables were
present.
0 Did you make any changes or deletions as a result
3
LJ
he
of
[+]
wo
-
z
ul
z
2
o
>
<<
m
P
E
N
G
A
D
C
O
.
,
102
of his consultation?
A No.
Q So, in essence, he approved the list that you had
made?
A Yes,
Q And what did you use to make the list; your own
sense of the data?
A I tried. to, for the first study I tried to pick
any conceivable variable that Professor Baldus has defined
that might apply. I think I even went overboard in the numbe]
of variables that I selected.
0 Your Pasi though, for selection was what?
A Variables that were defined by Professor Baldus
and variables that seemed to encompass most, if not all, the
information for the different categories that he had defined
in his questionnaire.
0 But within the different categories that Professor
Baldus had developed and using the variables that Professor
Baldus had as the universe, you selected the most important
aggravating variables based on your sense of what was most
aggravating; is that --
A No, I tried to select almost all of the aggravating
variables that he defined. I didn't make -- as I recall, I
did not try and make value judgments between the variables
except, specifically, there were some variables that had
LD
)
IX
Fo
-
z
ul
z
Zz
Oo
>
<
m
P
E
N
G
A
D
C
O
.
,
103
fewer occurrences; there were some variables that seemed to
indicate the same thing. And given the amount of analysis
I was going to do and the use that I was going to make of
these kinds of variables, I felt that if I did too much and
put every conceivable variable that Professor Baldus had
considered, this report may be much longer.
Q Let me ask you another question. I notice at the
bottom of Page 14 you said you "preserved the questionnaire
designation as to whether a particular variable was aggra-
vating or mitigating although there are a few cases where I
diagree with the designation.”
Can you identify for us those; were there a great
many or just a few?
A Well, there's two in particular. One is the
variable about the victim being drunk. It seems to me that
can be either an aggravating or a mitigating circumstance,
depending on whether the victim was an obnoxious drunk, just
partially drunk-- and that may serve as a victim mitigating
circumstance -- or if he was passed out and helpless, that
may be an aggravating circumstance. And so, I just went
with what was there. And I'm sure that I could go through
all these variables and question exactly what is being meant
by each variable, but I'm not an expert on that part. That
will be reserved for other people to question.
Q Okay. Well, you mentioned the victim's drunk.
-~
=
wi
z
z
o
>
«
mo
P
E
N
G
A
D
C
O
.
,
104
Were there any others, though, that --
A The defendant being an underling.
0) Being a -- pardon?
A An underling.
Q I. see; Okay.
A Because generally, you can have aggravated murder
cases where the defendant's an underling, usually as a result
of contemporaneous defense. And the defendant being an
underling is indicated as a mitigating factor, when that may
not be completely mitigating in terms of the circumstances.
Plus, there's also the fact that when you have a case where
there are perpetrators, those are generally white victim
cases. And so, there will be times when white victim cases
will have that variable defendant as an underling as a miti-
gating factor and be significant... And I think it may not
be completely reflective of the situation.
Q Thank you. Now, if you'll continue your descriptiol
of what you did once you began this evaluation?
A Okay. Once I had defined: these variables,and I
tried to select a broad number encompassing pretty much
everything that Professor Baldus had defined for his first
study, I then --
0 Let me just clarify for the record, that list is
the list contained in Table Vion Pages 15 to 21 of Petitioner
Exhibit 27?
-
z
ul
Zz
z
Oo
>
<
m
P
E
N
G
A
D
cC
oO
.,
105
A Yes.
Q Okay.
A I then went ahead and compared the white victim
and black victim cases for all cases in terms of each of
these variables to get a picture of the levels of aggravation
and mitigation in white and black victim cases.
Q Where did those comparisons appear?
A This is Table’ VIII on Page 23.
Q Before we get to 23, let me ask you guestion or two
about some of the text on Pages 21 and 22. .You may have
answered this yesterday and if so, I'm just slow. The 2
scores you mentioned will be used in these tables as standard+
ized measures to compute the probability. How did you
arrive at the Z scores?
A Okay; the Z scores were arrived at using the normal
approximation to the binomial by looking at the -- do you
want the formula?
0 Yes, if there's a readily given formula.
A Okay. Under the hypothesis that the proportions fox
each case, for race of the victims; under the hypothesis that
the proportion for the population for the white victim cases
and the black victim cases are the same, what we have are
two estimates of sample proportions for white victim cases
and black victim cases.
And conceivably, what we're doing is testing to see
4
3
z
u
z
z
o
>
£4
a
P
E
N
G
A
D
C
O
.
,
106
whether these proportions are the same. And so the assumptiol
then, is that these proportions are the same and we're test-
ing that hypothesis. Now, it's possible that the proportions
that we get are not completely identical due to some random
variation in the data. And the Z scores are a measure of how
plausible it is that differences observed in these two popu-
lations are due to the random variation.
9) And these were computed, you said, by =--
A By the standard formula for the normal approxi-
mation of the binomial comparing two populations where the
hypothesis, the’ proportions are the same.
Q Is that generated by the computer?
A No.
0 How 1s that done?
A I know the formula.
Q Did you apply the same formula throughout to all of
the Z scores?
A Yes.
Q Did you use any method to confirm the 7 scores
that you had computed?
A You mean by hand calculator to make sure that =--
Q Or any other methodology to confirm the scores that
appeared?
A You mean some other tests?
Q Tests or that kind of thing; yes.
-
z
ul
z
z
o
>
4
@
P
E
N
G
A
D
C
O
.
,
107
A Of that kind of nature?
Q Yes.
A No.
Q Okay; no chi squares or anything else?
A Well, chi sguares will be equivalent to the Z
test... 50 ~~
Q So, you didn't perform —-
A You mean to ‘check if the 7Z values in my table are
exactly right?
Q Right.
A No; but I did write a computer program that gene-
rated the Z scores, and I tested some of the Z values with
a calculator and presumed that since they worked for a few,
that all the Z scores were correct.
0 All right.
A However, I had to enter the data on the table,
and so it's conceivable that some of these might be off
by a little.
9) You've testified that the Z score assumes a random
variation, is that correct, in the distribution throughout
the population?
A No. The assumption is that there is no difference
between white victim case population in terms of an attri-
bute and black victim cases in the population in terms of
an attribute. Let me look at an example. On Table VIII,
oO
”
[=]
ni
-
z
ul
z
z
oO
>
«
om
c
o
.
,
P
E
N
G
A
D
108
let's look at the variable ARMROB for armed robbery.
2 What page are you on?
A Page 23. Now, 40% of the white victim cases in
armed robbery occurred at the time of the murder. 18.2%
of the black victim cases in armed robbery occurred at the
time of the murder. Now, the question is, are these pro-
portions really similar and the differences only due to
random variation. For example, if I had two coins and I
flipped one coin and got a heads -- so I had one heads -- I
flipped the other coin and got a tails =-- no heads -- then
there's 100% difference, but that's not all that signifi-
cant because of the sample sizes.
Q That's one of the few things in statistics I think
I do understand would be the point you're making here. The
Z value, though, is computed --
A The Z value encompasses the percent difference
along with the sample sizes used to help indicate the chances
of the observed differences due to random variation.
Q But the assumption that you use in the coin flip
is that the flips will on either side of your left hand or
your right will be distributed randomly heads and tails over
a period of time; is that correct?
A Sure; .ves.,
Q So, your assumptions similarly in using .Z values
here are that the variation or the variables will be dis-
be)
or
r
@
4
z
ul
z
z
o
>
<
m
P
E
N
G
A
D
C
O
.
,
109
tributed over the black and the white populations over a
period of time?
A Right.
9) And so, the Z value represents the prospects, the
differences that we see, are not random distribution, but
reflect real differences?
A Real differences.
Qe All right. Let's go ahead. You directed me on
over to your next table where you were going to do: some --
I guess it's Table VIII you directed me to.
A Right.
Q Now, what does this table reflect?
A After giving the proportions for white victim and
black victim cases and the percentages and percent differenceg
and the Z values, I then chose, saw or selected those variablg
from this table in which significant differences were indi-
cated at the .05 level level of significance. And this can
be obtained under a Z value of 1.645 or greater or a Z value
of -1.645 or less. And I think that's the usual kind of
significance level that's used.
And so —-
0 Those results are reflected where?
A Table IX:on. Page 27.
Q And what do you conclude after you've examined
Table IX?
p
ES
a
3
z
ui
z
z
Oo
>=
ES
m
C
O
w
P
E
N
G
A
D
110
A Table IX on Page 27 indicates that white victim
cases have significantly at the .05 level, at least, have
significant differences in the victim being -- I need to
refer to my dictionary here -- victim being defenseless;
either bedridden or handicapped; mental defected; defenseless
because of youth; defenseless by virtue of advanced age;
defenseless because of: physical condition or weakness; that
is frail woman or pregnant; have cases in which the victim
was on duty =-
Q That's fine. I don't think we need to go through
every single one of them; that's not what I really meant to
ask you.
A Nell, all these instances are aggravating circum-
stances, and they go down on Page 28, and then total varia-
bles of the aggravating circumstances. And those are the
variables which show significance in white victim cases
having more of those variables than black victim cases at
the significant level.
On the other side are those variables in which the
black victim cases showed that they had significantly more
of that factor than white victim cases. And as you notice,
that side of the Table is blank.
Then we move to mitigating factors. And again, I
take those Z values that showed significance at the .05 level,
So, if there was a Z value of 1.645 or greater, a -1.0645 or
o~
o>
-<
z
ui
z
z
Oo
>=
<
m
P
E
N
G
A
D
C
O
.
,
less, they're supposed to be reflected in this table. And
so, the left hand column indicates those variables in which
the white victim cases had more of that mitigating factor,
and the right hand side of the table indicates those miti-
gating factors that black victim cases had more of. And as
you see, black victim cases systematically have more miti-
gating factors. However, there are a few mitigating factors
to white victim cases.
Q Having done this analysis, how do you relate that
to Baldus's analysis in the methods that he's using to
answer questions?
A Well, the conclusion I came to is that white victim
cases are much worse than black victim cases.
Q By. worse you mean. ~—-
A They're systematically more aggravated and syste-
matically less mitigating.
Q From that conclusion, where do you proceed?
A Now we move to Baldus's first table. Now, in this
first Table X for his first experiment, he breaks down each
of the cases by whether or not each of the statutory aggra-
vating circumstances occurred or not. For example, if we
look: at the first line, "Prior. capital record (Bl)", there
were 26 white victim cases that received Bl, and thirteen of
those received death sentences. And there were 15 cases of
black victim cases that had Bl and only one of those received
112
y 1 the death sentence.
2 Then I've indicated the .50 as a proportion of
3 | white victim cases that received the death sentence, and .07
4 | as a proportion of black victim cases that received the death
5 | sentence. Then I computed the percent difference and computed
6 | the significant Z value to.determine if this percent
7 | difference is significant. And at the .05 level, you must
8 | accept, you cannot —==-‘at the .05 level you would reject that
9g | this difference is due to random variation.
10 So, when Baldus did this analysis he claimed that
11 this indicated that there was a difference here, a significant
12 | difference between sentencing patterns for white victim cases
13 | and black victim cases holding on statutory aggravating circunm-
14 | stance Bl. Then he does a similar kind of analysis for B2,
15 | B3, B4, down to B 10.
16 Now, after looking at the previous table, the
17 | question comes up, is this really a fair match in that in
18 | matching white victim cases and black victim cases on simply
19 | statutory aggravating circumstance Bl, will it really give
20 | us similar populations. Can we expect that the 26 white
victim cases are similar in aggravating and mitigating
BA
YO
NN
E,
N.
J.
©7
00
2
.
FO
RM
20
No
p
a
22 circumstances to the 15 black victim cases. That is the
P
E
N
G
A
D
C
O
.
,
23 | assumption behind taking this percent difference and computing
24 | the Z value and claiming this indicates some kind of discri-
25 | mination.
PS
&
3
z
ul
z
Zz
o
>
<
m
P
E
N
G
A
D
C
O
.
,
(8
)
o>
}
So, to test that, whether these two populations
that are compared are really similar, I went ahead and
developed Table XII. And what I did was I ran correlations
between --
MS. WESTMORELAND: Excuse me for a minute; can
go off «the record, please.
(Whereupon, there was a brief
interruption of the record.)
THE WITNESS: I ran correlations comparing only
on those cases where Bl occurred and comparing the variable
race of the victim using the recodes, recoding the five
unknown black defendant cases where it's unknown, the race
of the victim, giving them black victims, and computed the
correlations for all those variables defined earlier.
BY MR. BOGER:
Q That's what's reflected in Table XI?
A Table XII.
Q I'm sorry; Table XIT.
A Then from those correlations, I picked out those
variables that had values that indicated their significance
level was less than .05.
Now, correlations, when we're dealing with 01
variables, percent differences, the Chi Square Contingency
Tests are identical. The Correlations Test, I find, is
approximately equal to those tests.
“
ot
=,
z
ui
z
z
o
>=
Ld
m
P
E
N
G
A
D
C
O
.
,
114
0 What do you mean by approximately equal?
A Well, the distribution is T, but under the large
sample size it's approximately a normal distribution. So,
I would expect that it's testing the same kind of thing, and
I've checked a number of instances to make sure that you do
get essentially the same Z value. And I have gotten the
same Z value or something off by maybe, for example, 2.42
instead of 2.41. ‘So, nothing that would be important for
what I'm trying to demonstrate here in this table.
Now, what I find in Table XII is that white victim
cases seem to still be more aggravated than black victim
cases when we compare only those cases that had statutory
aggravating circumstance Bl. Then, when we moved to miti-
gating factors, I find that black victim cases have more of
those factors.
So, my conclusion is that it's very possible that
the difference in death sentencing rates is not necessarily
due to discrimination but due to the fact that white victim
cases still seem to be much more aggravated and less miti-
gated than the black victim cases.
Q I noticed your conclusion is that it's very
possible that that's the case. Let me direct you over to
Page 33 of Petitioner's Exhibit 2, and the first line of
text says "It is apparent from Tables XI that even when white
victim and black victim cases are matched on statutory aggra-
[ve]
3
=
ul
z
z
Oo
>
<
m
P
E
N
G
A
D
C
O
.
,
115
vating circumstance B2, white victim cases are still signi-
ficantly more aggravated and less mitigated than black victim
cases..". Is that the point you're making here?
A Yes.
Q Then you go on to say that "explains the discre-
pancy in death fated." Do you think this is an explanation
that --
A Let me say this is a legitimate explanation.
Q What do you mean by that?
A It's an alternative explanation to the assumption
that it's discrimination that's causing the difference in
death penalty rates.
0 Well, do you think that the greater aggravation
level in white victim cases precludes or makes it impossible
for there to be, nevertheless, discrimination?
A No. I'm just saying that given the higher level of
aggravation and lower level of mitigation in the white victim
cases as compared to the black victim cases, this is not a
meaningful match.
Q But there's a negative statement there and then
there's a positive statement. The negative statemen is the
one that given the significantly higher level of aggravation
in white victim cases, the match is not a fair match, if you
would. The second statement is this significantly higher
level of aggravation in white victim cases explains the
o
ho
&
©
o
=
z
ui
Z
z
1S)
bod
Ld
m
P
E
N
G
A
D
C
O
.
,
116
discrepancy. The latter statement --
A Okay; "explains" is too strong a word as a possible
explanation.
Q So it in effect offers something that could explain
the difference, but you've not from this data been able to
demonstrate?
A Not from this experiment to prove that's the
reason.
Q All right.
A My purpose in doing these tables is to show that
the comparisons in Table X in which Professor Baldus indi-
cates significant differences are not valid because the popu-
lations that he's comparing are not comparable in terms of
aggravating and mitigating factors.
And so, observing higher death sentencing rates
in white victim cases rather than black victim cases doesn't
necessarily point to discrimination.
Q It doesn't preclude discrimination either; does
it?
A It does -- however, I have provided an explanation,
a possible explanation for the death sentencing rates.
Q That's right. It's an hypothesis about the death
sentencing rates; is that correct?
A It's a possible explanation substantiated by the
data, by the table that I ran. It's a possible explanation
3
z
ul
z
Zz
o
>
<
@
P
E
N
G
A
D
C
O
.
,
117
for the difference in sentencing rates.
Q And your table substantiates that explanation?
A Yes.
Q Did you conduct any further analysis of Professor
Baldus's analysis of+his various methods of reviewing his
data?
A Okay. I then ran that same experiment for all
the: statutory aggravating circumstances that indicated a
significant’ Z+value at the’ .05 level. And those are the
tables that follow from Table IX all the way through Table --
Table XI, excuse me -- all the way through Table XVII; from
Pages 32 until and including 47.
0 And what did these tables reflect?
A In each instance I observed the same pattern of
white victim cases being more aggravated and less mitigated
than black victim cases.
Q All right. From this, what was your conclusion?
A My conclusion was that in testing for discriminatiol
the test that Professor Baldus has run in comparing cases
based on the given statutory aggravating circumstance does
not directly lead to any indication that that's caused by
discrimination. And a plausible explanation exists for the
difference in sentencing rates.
0 Plausible alternative theory; is that your point?
A Yes.
I
z
ui
=
z
o
>=
<
@
P
E
N
G
A
D
C
O
.
,
118
0 Now, did you examine any of Dr. Baldus's or
Professor Baldus's other means that he employed in analyzing
the data?
A Yes. I then went to his second experiment from
the preliminary report where he compared the cases on the
number of statutory aggravating circumstances and, again,
I updated the table. Originally, there were 615 cases indi-
cated ‘in the preliminary report. The data that I received
only had 607 cases, so I updated the table to reflect that
change. And I again looked at those situations where the
number of statutory aggravating circumstances yielded a sig-
nificant Z value at the .05 level.
Q Let's be; for the record, a little clearer. Dr.
Baldus's table that reflects the number of statutory aggra-
vating circumstances is what table in your report?
A Table XVIII.
Q And then the tables that you generated to show case
in which there was statistical significance at the .05 level?
A Tables XIX and XX. There were only two situations
with three and four aggravating circumstances where a signi-
ficant difference at the .05 level was indicated. And in
each of these cases, it still appeared that white victim
cases were more aggravated and less mitigated than black
victim cases.
0 All right. Did you do any further evaluation of
Ul
119
% 1 | his methods of analysis?
2 A The third method that he indicated in the preliminary
3 | report was the construction of a weighted index to try and
4 | account, using regression, for aggravating and mitigating cir¢um-
5 | stances. And I discussed what I know about how this weighted
6 | index was constructed. Originally, this weighted index was
7 | computed over 615 cases and I believe as I indicated on Page
8 | 51 that in an answer to an interrogatory, Professor Baldus
9 | indicated the following: "The weighted index was constructed
10 | from partial regression coefficients for legitimate factors
11 | that were found in statistically significant association with
12 | the death sentence outcome and whose coefficients had signs
13 | running in expected directions. A score for each case was cal-
14 | culated from these coefficients with each offender's overall
15 | score depending on the facts of his or her case. The cases wegre
16 | then ranked according to their scores. The factors found to be
17 | statistically significant in association with death sentencing
18 | outcomes and with the sign of their coefficients running in the
19 | expected direction were the following:". And I've listed the
20 | twenty variables. Now, I searched through the information
that Professor Baldus provided me for the Procedural Reform
B
A
Y
O
N
N
E
,
N.
J.
©1
83
2
No
m
—
22 | study, and I could not find a regression which had each of
P
E
N
G
A
D
C
O
.
,
23 | those twenty variables.
L
B
24 However, I realized that when the preliminary report
25 | had been done, there were 615 cases. Now, for whatever
~
©
Pe
<
B
A
Y
O
N
N
E
,
N.
J.
P
E
N
G
A
D
C
O
.
,
120
reasons, there were only 607 cases. So, I would presume that
Professor Baldus reran the experiment using only the 607 case
And under that assumption --
0 I think next Tuesday you'll probably find out
that's correct; next Wednesday.
A Well, under that assumption I found a regression
that was very close to the earlier regression, except armed
robbery isn't included, and there are four new variables
that are included -- 1, 2, 3, and 4 -- in the same victim
lower status.
Q That's all right. We've got the list before us.
So, having done that, what did you do?
A Okay. On running the multiple regression, I found
that each of those 23 variables were statistically signi-
ficant at the .05 level. And so that confirmed by belief
that this was indeed what was meant by this regression.
Q Did you perform or attempt to replicate his analysi:
employing this multiple regression weighted index?
A No, I simply ran the regression to check to see
if the coefficients I obtained were equivalent to the co-
efficients that he obtained or weights for the different
variables.
Q And what was your result there?
A And they matched.
Q Let me ask you, you appear to state an objection
Ul
°
3
z
uw
z
z
o
>
<C
m
P
E
N
G
A
D
C
O
.
,
121
on Page 52 to this methodology because of its use of multiple
regression; is that correct?
A Pardon?
0 This method that Professor Baldus had used to
attempt to measure discrimination, you object to on the groung
that multiple regression isn't appropriate here?
A I think I claimed the main problem in this case
is that multiple regression is a data intensive technique lik
all multivariate tests. To get accurate information from
the multiple regression, you must start off with good, hard
data for all the cases, for all the variables that you're
considering.
0) And if you don't have, as you put it, good, hard
data for all the variables, you can't perform multiple
regressions?
A You can do it, but the results aren't necessarily
meaningful.
Q When you say aren't necessarily meaningful, are
they sometimes meaningful?
A There is a random chance, occurrence, that errors
will cancel out and even though inappropriate means have
been used, the correct answer is arrived at.
3) But apart from that random chance --
A Apart from that random chance, whatever -- what
am I saying?
122
% 1 ®) Well, I want to ask you, apart from that random
2 | chance, what; there's simply no valid multiple regression
3:{>in that?
4 A There's no substitute for good, hard data in doing
5 |/your analysis.
6 Q Let me ask you, on Page 52 in your -- it appears
7 | tot be at the third or fourth paragraph; it's actually the
g | last paragraph of text. You have a sentence that says,
9 | "Multiple regression is a technigue that attempts to predict
10 | the outcome variable by assigning weight to the predicting
11 | variables in a manner which..", and this is what I want you
12 | to focus on, "minimizes the sum of the squared error between
13 | the actual outcome and the predicted outcome." What do you
14 | mean by that term, "minimizes the sum of the squared error?"
15 A Well, the purpose of regression is to assign
16 | weights to the different variables. And there must be some
17 | criterion that one uses to get these weights; otherwise,
18 | they could be arbitrarily assigned.
19 Well, the purpose of developing a model using multir
20 | ple regression is to try and develop weights so that after
B
A
Y
O
N
N
E
,
N.
J.
27
28
62
No
—
—
k
using these weights on the variables for all the different
} vy
af) Z
22 | cases that arent analyzed, that the predicted outcomes by
23 | using this index for these weights is very close to the
P
E
N
G
A
D
C
O
.
,
2a | actual outcomes, which in this case are either 0 or 1.
25 Now, the criterion that's used is for each actual
¢
~
fe]
~
-
z
ui
2
z
Oo
>
<<
[]
C
o
.
,
P
E
N
G
A
D
123
outcome, there will be a predicted outcome for a given set
of weights for each case. You take the difference of the
Minus ne. pred iced
actual minus—ef-predieting for a given set of weights over
all the cases, take the difference, square them, and sum
them, then find the weights that minimizes that sum.
So, that's a measure of closeness of fit of your
weights in your predicted outcomes to your actual outcomes.
Q I see.” So, in effect, the error when you're talking
a squared error is the differential between the predicted and
the outcome. So, you simply square that differential and
that's a measure of fit?
A Right. Squaring -- if you didn't. square
them, you'd have plus differences and minus differences, when ’
you're adding them together would cancel out. Squaring them
all positive differences.
0 I note on the next page that you say you did rerun
these regressions or this regression using a software package
which threw out all the observations if there were any
variables unknown, which is the standard convention; is that
correct?
A Yes.
Q And how many -- you've got ony 53 cases in which
there were 23 predictors?
A Yes.
0 And this, again, is in the Procedural Reform Study?
‘
‘
3
z
ui
z
z
o
>
<<
m
P
E
N
G
A
D
C
O
.
,
10
124
A Yes.
9) So, 1t's your conclusion, I .take it, it comes from
your earlier observations about the inability of multiple
regression to operate according to standard procedures if
there are so many unknowns that Professor Baldus's one test
that doesn't suffer from flaws, the earlier two that you des-
cribed simply cannot be performed. Do you know of any
alternative methods of attempting to evaluate discrimination
without using regression if the data is not as hard as you
described it?
A Yes, the methods I employed later on.
Q All right. These are the methods you were beginnin
to describe on Page 55; is that correct?
A Yes.
Q Now, on 54 you state that you have run most of .the
regressions sent to you by Baldus in both studies and have
not found any that explains the high percentage of the
variation of sentencing outcomes. So, this refers both to
the Procedural Reform Study and the Charging and Sentencing
Study?
A Yes.
Q How much variation did these regressions explain?
A I need a copy of the printouts. Now, these are.
approximations because I didn't run exactly the same
regression that Professor Baldus runs, but I didn't get
i
o
i
3
z
Ww
z
z
o
>=
<
oa
P
E
N
G
A
D
C
O
.
,
exactly the same outcome, the same weights, but I got close
weights. So, I would presume that his actual amount of
variation will be explained is not too much different from
what I have received.
Q Okay. So, you're proceeding with slightly different
weights in your regression, but close?
A Close; yes.
Q Let me just ask you while you're looking through
those regressions, are they close enough to be explained by
rounding error or not that close?
A No, they're not close enough to be explained by
rounding error.
0 All right. Go ahead, if you would, and look
through those and tell me what kind of explanation. or
explanatory values in variations you found.
A Well, let me add one other point. In analyzing
Professor Baldus's analysis, I find he has a very systematic
way of rounding and presenting his information. So, if
something deviated from that systematic procedure, then I
took note that it probably wasn't proper, that I hadn't
exactly duplicated his experiment.
0 Let me clarify it because I'm not sure what you
mean.
A For the first regressions, they were duplicated.
I saw exactly how he rounded and it was perfectly acceptable.
€
h
2
=
=
wl
z
z
0
>
<<
[eo]
P
E
N
G
A
D
C
O
.
,
126
But I didn't observe those same kinds of rounding conventions
in these regressions for the Charging and Sentencing Study.
Q Were there sort of random differences that you
couldn't explain?
A They were small differences but still distinct
differences.
Q Were there any systematic biases one way or another
that you detected?
A I didn't detect any, no. Let me refer back to
the first regression that we looked at for the Charging and
Sentencing Study. This is DTSNIDX2; I think it's a
regression that we discussed earlier. And the proportion
of variation explained by the regression that I received
running the regression my way was .28 --
0 That's significant --
A 28%. ~~
Q I'm sorry; +28.
A 28% of all the variation death sentence outcomes
where I believe looking at cases of "went to penalty trial."
Q Now, these are adjusted or unadjusted figures?
A I'm not sure whether this information is the
adjusted or unadjusted.
Q Do you recall whether you ran them both ways or
just one way?
A For this sample size, they're almost virtually
-
z
w
z
z
Oo
>
<
m
P
E
N
G
A
D
C
O
.
.
127
the same. There's varied difference between the unadjusted
and the adjusted R square.
Q But you don't recall which way you ran them?
A I'd have to calculate them to see which one the
regression did.
Q Let me ask you just once again for the record to
state what is the regression that we're talking about now
that you ran?
A This is the DTSNIDX2.
Q And what's the measure of the dependent variable
there?
A Death Sent; DEATHSNT. My coefficients that I've
received as compared to his coefficients, I have .07 for the
intercept compared to .08; .36 for oor copared to
.35; .33 for, I don't remember what that variable is, compare
to .32; .15 for scientific evidence compared to .13; =-.1l0
for clear ID compared to -.16; .03 for evidence index, I
believe, compared to .032; and .10 for IDEATHGUN compared
to 14.
So, they're close, but they're not exact.
Q Now, you said he explained 28% of the outcome,
sentencing outcome. What would you consider a highly
explanatory regression?
A Well, there's no clear-cut number that will, if
you get above that number you'll have a good model, if you
~
@ z
°
is
z
ul
Zz
z
o
>
«
a
P
E
N
G
A
D
C
O
.
,
128
get below that number, you don't have a good model.
0 So, ‘it's not a convention like the .05 level of
statistical significance or something that's accepted through;
out the statistical world?
A Right. I think it comes down to what the purpose
of this regression is.
Q The statistical purpose or the social purpose?
A No, what it will be used for. If this regression
is going to be used to then predict using other data whether
a death sentence would be given by judge or jury, given a
penalty trial, then 28% of the variation isn't enough.
Q What would be enough?
A 28% is not enough. A guess at what might be enough
certainly in the 90% would be good, but highly unlikely that
you're going obtain those kinds of results given the 01 Ol!
variables that you're dealing with.
So, 90% would be okay.
Q All right. What we've got now is a conclusion
by you, I believe, that the methods that Professor Baldus
has attempted to use to detérmine discrimination are in-
sufficient and inadequate and don't prove what he's --
A There was one other.
Q I'm sorry. that's ‘right; there is a fourth.
A His fourth experiment, which I believe is finally
getting to a proper method of analysis is Professor Baldus's
oe
od
a
z
ui
z
z
o
>
<C
m
P
E
N
G
A
D
C
O
.
,
129
attempt to break down the 607 cases into factor classes based
on seven different factor levels. And I don't know precisely
right now ‘what those different factor levels are.
0 But as far ‘as you're concerned, that's the best
start of any of the four methods, if you would?
A Yes.
Q What are the criticisms of that method as Professor
Baldus has worked it out?
A Okay. The problem with this kind of analysis is
that the number of different cells that'result increase at
a very rapid rate. And so, to explain, for example, seven
different factors, ‘it there are: two factors each, you're
talking about eventually at the bottom row where supposedly
all the factors have been washed out, and so you have similar
cases of two to the seventh --
Q Alot of cases.
A I CANS we
0 No, that's fine.
A And now, there's also a question as to how to facto]
the classes as to what are the variables that should be
factored. And that's something that takes some judgment
too. |
But the major problem that you run into is that
very quickly you ' have cells where you don't have a large
enough sample size to compare them. And so, given the sample
my
©
B
A
Y
O
N
N
E
,
N.
J.
C
o
.
,
P
E
N
G
A
D
130
size that we have in both studies, I don't believe that you
can analyze and break down the cases appropriately to eliminat
all the aggravating, more ‘aggravating, and less mitigating
factors that Vike vietim cases have until you get to a point
where you can actually compare similar cases.
Q How far, how deep would you have to go to get
comparisons that ‘had some meaning?
A Well, what would have to be done is some variables
selected that seem to be breaking down the cases appropriatelj
Then when you finally get --
Q Excuse me -- what do you mean by appropriately?
A That have meaningful correlation with sentencing
outcome; things that you would expect to affect whether a
defendant receives life sentence or death sentence. For examp
0 All right. Once you had gotten -- excuse me; I
didn't mean to interrupt your example.
A For example, Professor Baldus first breaks down
all cases in terms of the number of victims, which is a
plausible start. And he has three cells. Then he breaks
down the second level in terms of --
Q Let me show you a document that might refresh
your recollection.
A -- whether a serious contemporaneous offense
occurred or not. Then, there were two possibilities for
there, so we're now up to six cells.
le --
-
z
w
z
z
[o]
>
Eq
m
P
E
N
G
A
D
C
O
.
131
Then, the next level was whether there were serious
aggravating circumstances. But that in itself doesn't
necessarily explain all the aggravating circumstances. You'rg
going to include cases that have one aggravating circumstance
and cases that have perhaps four or five aggravating circum-
stances.
Then prior record is another factor level which
breaks everything down in another two ways. Then, whether
they were mitigating circumstances, and he defined certain
mitigating circumstances.
Q Okay; I don't think we need to go any further.
My real question was how deep do you have to go? There's
no one answer; is there?
A Pretty much, you have to go as deep until the thing;
at the bottom have completely washed out all the differences
between white victim and black victim cases. So that if you
look at each cell, you have a representative sample, some
size, plus you have similar cases in terms of aggravating
and mitigating factors. So, then, you have something that
truly can be compared in terms of death sentencing outcomes.
Q I'm not sure I understand. If in fact there are
differences between white victim and black victim cases,
will you get to a point in which the difference is washed
out?
A what you're trying to do is match those black
Av
7
132
% 1 victim cases that are similar to the white victim cases, and
2 | that's what all the factor levels are doing. They're matching
3 up, hopefully matching up, similar cases. So that when you
4 | get to the bottom, you will have in your cell cases that have
5 | the same number of victims -- will have -- will agree as to
6 | whether there was a serious contemporaneous offense or not ;
7 | will agree as to whether there were serious aggravating circum-
8 | stances or not; will agree as to whether there was prior
9 | record; will agree whether there are mitigating circumstancesj
10 | and will agree whether there are minor aggravating circumstanges.
11 | So, agreeing on those seven levels, if you can then look at
12 | those cases and see that in fact they are similar cases, then
13 | you can go ahead and test to see if there is any difference
in terms of death sentences.
15 Q My question is, when do you know that you've gotten
16 | deep enough that you've got cases that are similar?
17 A When you can get to a point where you can test
18 | each cell and see that they're similar in each cell. When you
19 | get to the bottom and you can look at each cell and test
20 | them and see that they're similar, more or less.
Q With respect to the categories that you've
BA
YO
NN
E,
N.
J.
21
00
2
-
Fo
Rr
M
28
32
No
—
22 | previously broken down?
23 A Well, with respect to all aggravating circumstances}
P
E
N
G
A
D
C
O
.
,
24 | If you get down to the bottom and you have cases that still
25 | are not similar, then it's still not a meaningful comparison
x
or
oF
-
=
wi
z
z
Q
>
[4
Qa
G
A
D
C
O
.
,
P
E
N
133
because you're still not comparing similar cases; you're
comparing different cases. And there's still the possibility
and I would say even the likelihood that the reason there
would be perhaps a difference in death sentencing rates 1is
that one population of cases is still more aggravated and
perhaps ‘less mitigatéd:than the other population.
Q But I guess what I hear you saying is that there's
some finite number of characteristics that determine whether
something is similar, and if we get one hundred factors
deep, can't someone say, you still haven't accounted for
this factor, this variable, which I contend is significant.
And you've got to go at least one more level deep because
these cases, although similar through the first hundred, may
be different on this 10lst factor; is that correct?
A Yes, someone could make that complaint.
Q So, my question really is, where do you know that
you've gone deep enough that there's some validity, statistic
validity, in saying we don't need to go further?
A I can't give you a'number. 1 can't say you.go
seven factors deep and that's going to do it. All I can
say is the way you can check when you're at that point is
by looking at the cases in all these different cells -- and
who knows how many cells you have to compare at this point --
and see that they're similar.
Now, hopefully by categorizing them properly, you
4
134
. 1 are forcing cases, similar cases, to go to the same cell,
2 | which is the whole object of this.
3 Q But if there's someone who can continue to generate
characteristics or variables that could separate out otherwis: WW
5 | similar cases, then you must continually proceed deeper in
6 | the cells and the analysis; is that correct?
7 A If the populations you're comparing at the bottom
8 | cell are not comparable, are not similar --
9 Q With respect to whatever you --
10 A With respect to a bunch of legitimate aggravating
11 | and mitigating circumstances that haven't already been con-
12 | sidered, then comparing them isn't going to give you any
13 | meaningful information.
Q All right.
15 A So, you just have to get to the point where you
16 | can get meaningful information.
17 Q How many white victim or black victim cases must
there be, in your opinion, in the bottom row when the case
19 | is the same for comparisons to be meaningful?
3 20 A As many as possible.
: 21 Q How few is acceptable?
22 A Well, the problem with having a small number --
P
E
N
G
A
D
C
O
.
,
23 | the problem with a small number of cases is just like the
24 | coin tossing experiment. If I have one white victim case
25 | and he received the death penalty and one black victim case
135
% 1 | and he didn't receive the death penalty, I have ‘a percent
2 | difference of 100%, And that doesn't mean anything because
3 | it's ‘not significantly different.
4 Q All right. . Are there any conventions that you
5 | know of?
3 A Yes. What you can do is you can test for signi-
7 | ficance the sample size -- the sample size is included in
g | the Z statistic in your test for significance, given that you
g | have a reasonable enough sample size to apply the approxi-
10 | mation theorem.
11 Q I remember several years ago having done a case,
12 | someone suggested 5 as a number that you needed in your cell
13 | at a minimum. Is that appropriate?
14 A That sounds like a Chi-square Rule. That's not
15 | appropriate in this situation.
16 Q Is that too few, it sounds like, or too many?
17 A If you had, say, 5 for 5 white victim cases and
18 | 5 black victim cases and all 5 white victim cases received
19 | the death penalty and all 5 black victim cases didn't
20 | receive the death penalty, then I believe that that is going
to be statistically significant ‘at the .05 level.
B
A
Y
O
N
N
E
,
N.
J.
07
80
2
No
—
22 Q Just employing --
P
E
N
G
A
D
C
O
.
,
23 A Right; but if you have something like 3 out of 5
24 | and 1 out 5, that's not going to be statistically signifi-
25 | cant. So, only very great discrepancies will turn out to
©
-
z
ui
z
z
o
>
<C
om
P
E
N
G
A
D
C
O
.
,
136
be statistically significant.
Q Well, now, the kind of calculation you've made
as ‘to statistical significance when you would get to this
bottom cells are not. significant within the cells, but some
of overall formula. I guess you said the Z scores --
A 7Z scores if the sample size is large enough. If
you have small sample sizes, then you'd have to do what I've
done in one of those stacks, is go back to the actual bi-
nomial distributions and look for the true probabilities to
see if it's .05 or smaller.
Q But given the sample size in the Procedural Reform
Study, if we had full data on all the variables, what would
you think would be the best way to measure the racial effects
controlling or non-racial background factors?
A I think given that white victim cases are system-
atically worse than black victim cases, 1 think that's ‘the
wrong approach.
Q Which is the wrong approach?
A Trying. to take your data and match similar cases
and then compare sentencing outcomes.
0 Well, that's the negative of the answer that I hopeq
for the positive. What's the right approach?
A The right approach is, what I started doing, which
I think should be the first kind of analysis done in any
test of sentencing discrimination.
4
i
~
=
z
ul
Zz
z
o
>
«
m
P
E
N
G
A
D
Co
O.
,
137
Q And is that the method that you began to describe
on Page 55 of Petitioner's Exhibit 2?
A Yes.
0 Why don't you tell me briefly what that method is?
A Okay. The charge that Professor Baldus made in the
preliminary report was that prosecutors and juries and the
sentencing system tolerates a higher level of aggravation,
and I guess he meant lower levels of mitigation, in black
victim cases compared to white victim cases in sentencing.
I concluded from that, that sentence, that he believes that
there are aggravated black victim cases which because of the
race of the victim aren't sent on in the sentencing system.
For example, there may be very bad black victim cases where
the defendants received life sentences, where comparable
white victim cases would be sent to penalty trial, whether
or not they received the death penalty. Furthermore, going
from penalty trial to death sentence, he charges that there
would be black victim cases who had penalty trials that were
very aggravated that didn't go and get a death sentence
compared to similar white victim cases that went and received
the death sentence.
Okay, now, this can be tested very, very simply.
If, in fact, black victim cases are being discriminated agains
in the sentencing system, and aggravating factors are being
ignored by prosecutors and prosecutors are tolerating a greats
a
o
-
z
ui
z
z
[o]
>
<<
m
P
E
N
G
A
D
C
O
.
,
number: of aggravated circumstances, then we would see that
effect in the.:sentence population. For example --
Q Before you:'go ahead, let me ask you -- I think I
may ask you to simply restate what you just said. What
exactly is the empirical question that your method sets out
to answer?
A Whether there's any evidence that there is race of
victim discrimination in sentencing; whether black victim --
Q In the State of Georgia?
A In the State of Georgia; whether black victim cases
are not prosecuted as vigorously or for some other reasons
are not sent ahead in the sentencing system as comparable
aggravated white victim cases.
0 All right; if you can proceed, then.
A Now, if in fact there was discrimination, then it
would show up in the following way: There should be a group
of black victim cases which are life sentence cases that
deserve to go to penalty trial under a fair system, but
because of discrimination were left behind as life sentence
cases. Then, there should be a group of black victim cases
who receive life sentences after penalty trials which are
very aggravated, that should have been pushed ahead to go to
death sentences but because of discrimination of race of the
victim were left behind as life sentence with a penalty trial.
Q These are your sort of logical assumptions based
~
2
Q
by
r,
z
ul
z
Zz
[o]
>
«<
@
P
E
N
G
A
D
C
O
.
,
139
on what we're talking about?
A Yes; whereas in terms of white victim cases, either
they're being prosecuted fairly or prosecutors are over-
reacting in pushing ahead defendants in white victim cases
through to higher potential sentences because of the race
of the victim.
So, that effect would tend to mean that there would
be cases, life sentence cases, for white victim cases that
weren't all that aggravated, but because they were white
victims, were sent onto penalty trial, and there would be
white victim cases that went onto penalty trial and really
did not deserve death sentences in a fair system, but were
pushed ahead because the race was white.
So, that would make life sentence black victim
cases highly aggravated compared to the white victim life
sentence cases --
2, Seen as a whole?
A -- if there was discrimination in the system.
0 Okay. So, what did you do with this assumption?
A So, I compared life sentence white victim cases
with life sentence black victim cases for all the variables
that I defined earlier.
0 Where is that reflected; where are those comparisons
A I believe that's Page 60.
MR. BOGER: Off the record just a moment.
52
3
Zz
ui
z=
z
le}
>
<
m
P
E
N
G
A
D
C
O
.
,
140
(Whereupon, there was a brief
interruption of the record.)
THE WITNESS: "Page 60, Table XXIV, Comparison of
White Victim and Black Victim Cases whose Defendants Received
Life Sentences."
Now, this includes life sentence no penalty trial
and life sentence penalty trial cases. And I've got all
these percentages —-- and I'll spare you the details -- but
in the next table, Table XXV, I pick out those variables from
Table XXIV that indicate significant differences similar to
the method that I used in comparing cases under the different
statutory aggravating circumstances. And what I find is
that it's not black victim cases that are more aggravated
for life sentence cases, it's white victim cases. They're
more aggravated and less mitigated.
So, I conclude that,just from this experiment,
there's no substance to charge that -- or partially from
this; there's more tables -- that the charge that a higher
level of aggravation is tolerated in black victim cases
isn't true.
Q All right. You mentioned some other tables. What
tables are those?
A The next table I compared, which is Table 26 on
Page 67, compared white victim and black victim cases whose
defendants received life sentences with no penalty trial.
3
=z
wl
z
z
o
>
£4
m
P
E
N
G
A
D
C
O
.
,
141
And the next table has those variables that were significant
on that table. I believe that's Page 71, Table 27. And
again, I see no evidence of black victim cases being worse
than white victim cases.
0 From these two, then, tables, what is your conclusion?
A That the system does not tolerate a higher level of
aggravation in black victim cases than in white victim cases
in sentencing, based on the data with all the coded variables
and so forth, and unknowns and assumptions for the Procedural
Reform Study.
0 Okay. Now, let me just go back for my own purpose.
I'm trying to stay with your words, but I'm also trying to
stay with the text because your words flow and then the text
still there a minute or two later to read. On 60 you engage
in what is in effect a process of deducting logic; is that
correct?
A Yes.
0) That if Professor Baldus's assumptions are true,
FY
or his hypothesis =--
A If his contention is true.
Q -- is true, let me quote your language, "If discri-
1S
mination based on race of the victim exists, then life sentence
black victim cases must show systematically more aggravated
and less mitigated circumstances when compared to the life
sentence white victim cases." That's your observation?
142
A YES.
2 Q And the tables refute that observation --
3 A YES,
4 0 Therefore, deductively, Professor Baldus's hypothesis
5 | or his contention is incorrect?
6 A Based on the data in the Procedural Reform Study.
7 0 When you say that, have you done a similar analysis
g | on the Charging and Sentencing Study?
9 A Yes.
10 Q And did you reach a similar conclusion?
11 A Yes.
12 Q So, based on all the data that's before us, this
13 | remains your contention and your conclusion?
% A Yes. 14
15 Q And then the second set of tables that you looked
16-( ati== you indicated starting on Page 71, I believe --
17 A Yes.
18 0 ~-- are the tables that are statistically significant,
19 | called from among all of those factors in the earlier tables?
20 A In the earlier table.
0 I'm sorry, earlier table; forgive me. Now, back
.,
BA
YO
NN
E,
N.
J.
©7
50
2
N
—
at the very outset of your report when you had your conclusions
co
No
No
23. | and in essence say that based on your review of the data that
P
E
N
G
A
D
24 | Baldus's contention that there is discrimination is unwarranted,
25 | is your conclusion predicated on this sort of analysis or
-
z
ul
z
z
o
=
<<
[eo]
P
E
N
G
A
D
C
O
.
,
143
method that you've used?
A On this method of analysis; yes.
9) You then throw in on Page 73 your qualifications,
even to the use of this data; is that correct?
A Yes.
Q You've got the problem due to coding of unknowns
in the core questionnaire design. Now, the core guestionnair
design there refers largely to the charging, excuse me, to
the Procedural Reform Study?
A It has some affect on the Charging and Sentencing
Study.
Q And you say, "For even if all the unknown values fo]
a given variable were thrown out, as is the generally accepte(
statistical practice, there is still the problem of deciding
whether nonindicated items or truly unknown or did not occur!
is that correct?
A That's correct.
MR. BOGER: Off the record for a moment.
(Whereupon, there was a brief
interruption of the record.)
MS. WESTMORELAND: Back on the record.
BY MR. BOGER:
0 What does your chart on Page 74 of Petitioner's
Exhibit 2 represent, Professor, or Dr. Katz?
A Well, the question I'm trying to answer is how like] Ly
od
3
7)
-
=
B
A
Y
O
N
N
E
,
P
E
N
G
A
D
C
O
.
,
144
is it that these same results would occur if the data was
properly coded and all unknowns would be accounted for. And
what I notice is that using one example which is the variable
armed robbery, that in fact I wouldn't expect the actual
percentages to be correct as indicated in the study. I would
expect that some of those unknowns that were indicated to be
unknown actually had armed robbery. And so, that would affect
those percentages.
But since what we're questioning here is whether in
fact black victim cases are worse than white victim cases
given the same sentence such as life sentence, no penalty
trial; life sentence, penalty trial, et cetera , what's
important is, how likely is it that the percent difference
would change. And what I've noted here is if there is really
no systematic error to the way the unknowns are compiled,
it's highly unlikely that if all the data had been collected
properly that the percent differences would radically change
for the variables that were indicated.
So, even though this data is unreliable, it's
highly unlikely probabilistically that if all the data had
been collected, the percent differences would all switch the
other way, showing that black victim cases were systematically
worse than white victim cases.
Q And what does that add to your analysis?
A So, that indicates that although the numbers that
145
% 1 | I have in these tables aren't accurate in terms of percentages,
2 | the gist or the sense of what the tables indicate, the white
3 | victim cases, that the black victim cases are not more aggra-
4 | vated than white victim cases when compared by sentence, will
5 | probably still hold true; very, very highly likely that it
6 | will still hold true.
7 So, this is some evidence, although not conclusive
8 | evidence, that in fact there is no bias based on race of the
9 | victim in the system as what was compared in that study.
10 0 In other words, you stop short here of conclusively |--
11 | well, your last paragraph, you seem to have, much as at the
12 | beginning, two conclusions. Let me read to you, "Therefore
13 | the main conclusion that can be reached from this study is
14 | that white victim cases are severely more aggravated and
15 | less mitigated than black victim cases overall, which explains
16 | the higher death penalty rate." That's the first conclusion.
17 A Change that to, which is an explanation, a possible
18 | explanation of the higher death penalty rate.
: 19 0 And your second conclusion, "Furthermore, there is
20 | not the slightest bit of evidence in the study to support the
charge by Baldus that the Georgia charging and sentencing
BA
YO
NN
E,
N.
J.
27
0
N
i
e
22 | system operates on the dual track tolerating higher levels of
P
E
N
G
A
D
C
O
.
,
23 | aggravation in black victim cases as compared to white victim
24 | cases." That conclusion remains?
25 A Starting at life sentence cases through -- given
®
aN]
-
z
ul
z
z
Oo
>
Ld
o
P
E
N
G
A
D
C
O
.
,
146
the sentences that were being considered by the study.
Q Clarify that for a minute.
A Well, the charging and sentencing system goes from
indictment all the way through death sentencing. And that
hasn't all been considered here in the study.
Q In the Procedural Reform Study?
A Right.
Q It is considered at some greater length in the
Charging and Sentencing Study; is that correct?
A Yes
0) Okay; that takes us, I think, throudh the first
part of your preliminary report. Now, what characterizes
the focus of your second portion of your preliminary report?
A In the second portion of my preliminary report,
I again investigate the methods used by Baldus to determine
discrimination based on race of the victim, and on the
preliminary report that he submitted to me there are only
two basic tests that he used. And so I updated those tests -;
Q Before you get to the tests, isn't it a fact that
the second portion of your report focuses on the Charging
and Sentencing Study as opposed to the Procedural Reform
Study?
A Yes.
Q All right; now, would you continue about the tests?
A Well, I defined another group of variables which
-
4
7
Zz
wl
Zz
z
Oo
>
<C
>)
P
E
N
G
A
D
C
O
.
,
147
are variables that Professor Baldus defined in the Charging
and Sentencing Study, and these are the variables I'm going
to use to compare race of victim populations as in the Pro-
cedural Reform Study.
Q You say on 76, "the improvement in the questionnaire
design of the second study will allow us, after cleaning up
the unknowns of the variables, to later analyze the data and
truly test whether the Georgia charging and sentencing system
is disparate or discriminatory." In other words, is it your
opinion that the Charging and Sentencing Study is a better
vehicle for testing the question of discrimination?
A Yes.
©) Once you've defined the variables in Table XXVIII
beginning on Page 77, then what do you do?
A Then I again establish that white victim cases are
worse than black victim cases in Table XXIX. And Table XXX
picks out those variables that are significant in the .05
level.
0 Then what do you do?
A I think there's one table here out of place, one
page out of place, I believe Page 88 is supposed to be Page 87
Q Page 88 has at the top a column headed "White Victim
Cases", with the first variable in the variable column being
A Yes.
o~
-
Zz
wl
z
z
Oo
>
<
o
P
E
N
G
A
D
C
O
.
,
148
0 And that should be Page 87, and I take it likewise
Page 87 should be Page 88?
A Yes.
Q What do you see once you've examined those tables?
A That white victim cases are systematically worse
than black victim cases.
0 This the same pattern we saw in the Procedural
Reform Study?
A Yes.
0 What did you do once you obtained those results?
A Then I performed the same experiment as before
on the data provided to me, matching all cases with the same
statutory aggravating factors on Page 89, and then I picked
those factors that turned out to be, have significant
differences, and I believe those were Bl, B2, B3, B4, BY.
And then I compared those populations, those sub-populations,
between white and black victim cases where all the aggravatin
and mitigating factors that I initially defined in the earlie
table would have the variables for this study.
Q And those are reflected in what tables?
A Tables XXXII on Page 90, through Table XXXVI, which
ends on Page 100.
Q And having made those comparisons --
A I found that white victim cases were still worse
than black victim cases, and that is a possible explanation
a
z
w
z
z
[e]
>
<
om
P
E
N
G
A
D
C
O
.
,
149
of the difference in death penalty rates.
0 All right. Did you engage in analysis of Professor
Baldus's other methods employed in the Charging and Sentencing
Study?
A I then looked at the number of statutory aggravating
circumstances in Table XXXVII, picked out those numbers that
were statistically significant, which turned out to be, to
occur when there was 1, 2, 3, or 4, and then generated Tables
XXXVIII through XLI, which ends on Page 107, comparing the
variables between white and black victim cases.
Q And what did you conclude?
A I concluded that white victim cases were still more
aggravated and less mitigated than black victim cases, which
is a possible explanation of the sentencing patterns.
MR. BOGER: Off the record.
(Whereupon, there was a brief
interruption of the record.)
BY MR. BOGER:
Q Dr. Katz, you've now discussed two of the methods
that Professor Baldus used in his Charging and Sentencing
Study to attempt to determine whether there was discriminatior
in the Georgia system. Did you analyze the other methods
that Baldus used?
A In the preliminary report, there were only those
first two methods in which he'd given some details and some
<
3
z
ul
z
z
[o]
>
<
m
P
E
N
G
A
D
C
O
.
,
150
indication that he viewed that as important.
He did indicate also that he had run a great deal
of regressions, and I felt and I feel at this time that
before I include that in my criticisms of those regressions,
I first need to know which regressions he believes are
important, and then I'll go ahead and analyze them and give
my results.
All I can further say at this point is that there
is a great deal of unknowns in the Charging and Sentencing
Study which will affect the regressions, and they have to be
accounted for. And any regression that he does run has to
note these unknowns. And once I find out what he believes
are important, those regressions, then I'll embellish his
part of the report.
Q And what were you doing in Table XLII, then?
A I was just establishing that there are a great deal
of unknowns that still exist in the Charging and Sentencing
study; so that his regressions will be affected by these
unknowns.
Q All right. If you eliminated those unknowns,
though, and somehow had enough cases left, or enough obser-
vations left, multiple regression techniques would not be
objectionable?
A If the multiple regression techniques could
analyze all the variables that he's considering, and if he
2,
z
ui
z=
z
oO =
aS
@
P
E
N
G
A
D
C
O
.
,
151
makes special emphasis as to those variables that couldn't
be considered because of incomplete data, then he could
introduce that and I can analyze it or consider it at that
time.
Q But I'm not talking about consideration of what
it shows; but whether it's appropriate. You've made some
serious methodological charges about the Procedural Reform
Study and about the Charging and Sentencing Study as it's
before you. What I'm asking is if we could somehow send
out people into the field for another summer and fill in all |
our unknowns, would you have the same objections method-
ologically to the use of regression?®
A I have no objections to the use of regression, but
I do need to look at the regression to see how valid it is,
based on statistical measures. and those kinds of things,
once the unknowns have been filled in.
Q But again, I'm not trying to ask for your look to
see whether the people have done the regressions right or
what they prove. But regressions as a method of analysis
is appropriate if they're not unknowns; is that correct?
A So, your question is can there be problems with
regression?
Q No. You've said that the usual statistical con-
vention is that you do not use regression Or multivariate
analysis if there are unknowns such as there have been in
bf
%
z
wl
z
z
Oo
be
[4
>)
P
E
N
G
A
D
C
O
.
,
the Charging and Sentencing Study as it's before you. My
question is, do those conventions approve of or at least not
disapprove of regression if there are no unknowns?
A If there are no unknowns, then regression as a
statistical technigue can be used. Now, whether it's meaningt
ful is another question.
Q What does the question depend upon?
A Well, it depends on the regression that's run,
an investigation of the parameters, measures of fit, these
kinds of things. So, just running a regression doesn't make
that regression valid; it has to conform to certain accepted
conventions about what a good regression is in terms of
measures of fit and validity of variables.
0 Did you propose any counter method of analyzing
the Charging and Sentencing Study data to determine whether
discrimination existed in the Georgia system?
A I proposed the same method that I used in the
Procedural Reform Study ,except in the Charging and Sentencing
Study I'm able to account for unknowns in many of the items,
although there are some cases where the foil method was used,
which possibly introduces some error into the proportions that
are calculated for those variables, since we again have the
problem where items from the foil method that are not marked
as a foil may not have occurred or be unknown whether it
occurred.
IN
[=]
-
z
ui
z
=
=)
>=
Ld
m
P
E
N
G
A
D
C
O
.
,
153
However, I noticed in the Charging and Sentencing
Study that there seemed to be a lot less unknowns for those
situations where the foil method was used. Now, I'm not
sure whether that's due to improved data sources or some
other possibility.
Q All right. But in essence, what you've done, as
I understand from your writing on 110, is to throw out cases
with unknown values; is that what you did?
A Yes, for each individual variable.
Q For each variable. And the assumption on which it
was based, and still use the smaller sample that you've got,
is that there was no systematic bias between the unknowns
and the race of the victim?
A Well, that won't be important because I'm still,
what I'm looking at now will be the proportion of, say, black
victim cases, the proportion of cases where it's known where i
had that attribute divided by the total number of cases where
it's known whether or not it had that attribute.
So, the percentages should be reflected if there's
no systematic bias between whether the thing was unknown and
the occurrence of that attribute.
Q Okay. Now, you indicated there was no reason to
believe otherwise?
A Correct; I have no reason to believe that.
Q Okay. So, what did you do having decided to under-
t
] take that kind of analysis of the Charging and Sentencing
2 | study?
3 A I first compared to essentially investigate whether
4 | the system is arbitrary in assigning death sentences =-- I'm
5 | sorry; excuse me -- sentences, voluntary manslaughter, life
6 | or death sentences. I compared the proportions of cases that
7 | have each of the variables defined in the earlier table for
8 | the Charging and Sentencing Study. I found the proportions
9 | for all the voluntary manslaughter cases, compared them with
10 | the proportions for all the life sentence cases and compared
11 | them with the proportions for all the death sentence cases.
12 | For example, if we look at the variable ARMRCB, which
13 | represents armed robbery on Page 111, out of 460 voluntary
manslaughter cases, 14 had armed robbery, or 3.0%.
15 | Out of the 410 life sentence cases, 116 had armed robbery or
16 | 28.3%. Out of the 138 death sentence cases, 2 of the cases
17 | recall that unknowns have been called from these counts, which
18 | will be reflected in the denominator. So, it's not the
19 | actual 140, which is the total number of death sentence cases]
20 | 55%, 55.8% committed armed robbery. And going through these
variables, we find there's a pattern. For an aggravated
BA
YO
NN
E,
N.
J.
£7
90
2
No
y
s
22 | variable, generally, as the sentence increases, the percentage
P
E
N
G
A
D
C
O
.
,
23 | of occurrence of that aggravating factor increases. Similarly,
24 | when we look at mitigating factors starting on Page 113,
25 | generally this pattern moves the other way. The more severe
oy
3
z
wi
z
z
Oo
=
«
ao
P
E
N
G
A
D
C
O
.
,
9
10
155
the sentence, going from voluntary manslaughter to life
sentence to death sentence, the lower the proportion of the
mitigating factor.
Now, this doesn't prove that the system necessarily
is not arbitrary, but it does strongly indicate that there's
no -- it doesn't disprove that it's -- let me think on that
just a minute. This is a pattern that would be expected from
a system that wasn't arbitrary.
Q Let me ask you, I guess, the same question about
this that I asked you before. Are you saying that it's
impossible for a system to be arbitrary if it has this kind
of pattern?
A It's probabilistically small for a system to be
arbitrary with this pattern,
Q Let's back up and ask one further question. What
do you mean by arbitrary?
A In which sentences are being given without regard
to aggravating and mitigating factors.
Q You talking about statutory aggravating and miti-
gating factors?
A No; all the aggravating and mitigating factors
that's listed in my variable set.
Q So, having reached this conclusion, did you make
any. further effort to determine whether Professor Baldus's
conclusions about discriminatory application of the death
% ! | sentence in Georgia were not supported by the data of the
2 | charging and Sentencing System Study?
3 MS. WESTMORELAND: Excuse me; can we go off the
4 | record for just a minute?
b (Whereupon, there was a brief
6 interruption of the record.)
7 THE WITNESS: In Table XLIV -- yes, I did -- in
8 | Table XLIV, I first compare all cases by race of victim to
9 | again establish that even with the unknowns from that, white
10 | victim cases are still more aggravated and less mitigated
11 | than black victim cases.
12 And Table XLV on Page 120 picks those variables
13.1 that are significant at the .05 level.
BY MR. BOGER:
15 Q What does that Table XLV show?
16 A Table XLV shows that white victim cases are still
17 | systematically more aggravated than black victim cases, and
18 | that black victim cases are still systematically more miti-
g 19 | gated than white victim cases.
20 Q All right. Then, what did you do?
A Then I, using the same rationale for the earlier
BA
YO
NN
E,
N.
J.
07
No
S
—
22 | tables by comparing white and black victim cases with
P
E
N
G
A
D
C
O
.
,
23 | common sentences, I then compared white and black victim
24 | cases with who received voluntary manslaughter sentences.
25 | Again, the =-
¢
¢
3
z
ul
z
z
o
>=
«
@m
P
E
N
G
A
D
C
O
.
,
0 Let's stop for just a moment; you say the same
rationale. The rationale, let's see if I can restate it:
is that if prosecutors are tolerating more aggravation in
black cases --
A Black victim cases.
Q ~- black victim cases -- before they will seek a
death sentence --
A or a higher sentence.
Q -— or a higher sentence -- and less aggravation in
white victim cases before they'll seek a higher sentence,
what one would expect to find in those cases that receive
lower sentences are more aggravated black victim cases as
a whole and less aggravated white victim cases as a whole,
comparatively.
A Compared by sentence, by the same sentence.
0 So, that's the deductive first or second step;
whatever it is?
A Yess.
Q And what did you find?
A Okay; Table XLVI compares race of victim for volun-
tary manslaughter cases, gives all the proportions. And then
on Table XLVII, I pick out those variables that are statisti-
cally significant at the .05 level. And based on these
variables, I cannot conclude that black victim cases are
worse than white victim cases.
158
y 1 Q I'm sorry; would you repeat that? You said you
2 cannot ‘conclude --
3 A I cannot agree with Professor Baldus's contention
4 | that black victim cases are worse than white victim cases.
5 0 I'm not sure that that's the conclusion that's set
6 | forth. You mean with any --
7 A What we're testing is whether, given voluntary
g | manslaughter sentences, whether black victim cases are worse
9 | than white victim cases. My results in Table XLVII do not
10 | substantiate that.
4 11 Q Did you conduct any further analysis of Dr. Baldus's
12 | or Professor Baldus's own study?
13 A Yes. I then established Table XLVIII which compared
— 14 | life sentence cases -- excuse me -- compared race of victim
15 | for life sentence cases. So, that includes life sentence
16 | no penalty trial and life sentence penalty trial cases. And
17 | Table XLVIII presents all the proportions. In Table IL,
18 | which is on Page 132, I then pick out those variables from
19 | Table XLVIII that are significant at the .05 level and again
20 | I cannot conclude based on Table IL that black victim cases
are worse than white victim cases, overall, given life sentences.
BA
YO
NN
E,
N.
J.
€7
60
z
No
—
22 I then compared in Table L white and black victim
23 | cases in terms of those life sentences where no penalty trial
P
E
N
G
A
D
C
O
.
,
24 | was held. ' And the proportions are set out in Table L.
25 Table LI picks those variables in Table L that
159
6 : showed significance at the .05 level. And what seems
interesting is that we're starting to see a little more
] parity between white and black victim cases in terms of aggraj
vating circumstances. Still I cannot conclude that black
® | victim cases are systematically worse than white victim cases)
6 | but it is starting to look like these populations are
comparable.
8 Q would that, then, support your rival hypothesis,
9 | or would that tend to support Dr., or Professor Baldus's?
10 A Whether they're comparable?
1" | Q Yes.
12 A That would support the hypothesis that the system's
4 130 Fair. What we would hope is that none of these variables
14 | would be significant, and that would indicate that the
15 | populations are comparable. And that would be the best of
16 | all possibilities to establish that the system is fair, to
17 | suggest that the system is fair.
Okay; one thing that you should note as you look
19 | at Table LI is also consider those variables that didn't turnout
20 | to be significantly different. And so, there is a process here
where white and black victim cases who received life sentences
B
A
Y
O
N
N
E
,
N.
J.
G
7
0
0
2
No
_
_
22 | with no penalty trial are comparable and similar.
P
E
N
G
A
D
C
O
.
,
23 Table LII, I compared life sentence penalty trial
24 | cases, life sentence cases —-- excuse me -- I compared white
25 victim and black victim cases or those cases that received
-
z
ul
z
Zz
Oo
>
<<
om
P
E
N
G
A
D
C
O
.
,
10
160
life sentences and a penalty trial was held. And the pro-
portions are set out in Table LII starting on Page 140.
On Page 143 in Table LIII I then pull out those
variables that are significant at the .05 level. And again,
we find there are some white victim cases variables that are
significant and aggravated and there are some black victim
cases variables that are significant and aggravated.
Q So, these two tables that we've just talked about
reflect the system moving more toward the ideal sort of
parity?
A More toward parity; exactly. And I cannot conclude
that -- or based on this table, there's no evidence that
black victim cases are systematically worse: than white victim
cases when we're dealing with a population of life sentences
and penalty trials.
Then finally on Table LIV on Page 145, I compare
death sentence cases by race of victim, and all those pro-
portions were set out.
And on Page 149, Table LV, I then pick out those
variables that. are significant at the .05 level, and I note
something very interesting. And that is, there seemed to
be -- and these are death sentence cases -- the white victim
cases are still, it seems if anything, worse than the black
victim cases when compared in terms of death sentences.
They have more aggravating factors and black victim cases
oN
3
z
uw
z
z
o
>
<
m
P
E
N
G
A
D
C
O
.
,
161
have more mitigating factors. And what's surprising is that
the white victim cases represent 119 out of the total number
of 509, I believe, white victim cases, or about 20% of the
cases have been pulled out of the system and given death
sentences. However, only 21 out of 504 black victim cases
have been pulled out and given death sentences.
Now, when you just compare those proportions
originally, it seems like if it can be assumed that white
and black victim cases are similar, that there's some
disparity. However, that's not what's indicated by this
table. White victim cases for some reason are systematically
worse than black victim cases,.even at the death sentence
level.
Q All right. You had some conclusions on 144 I want
& Ard] v
Ty CIV INT § ¥%
to point you toward. You reject, s vy reject, that the
charge of disparate treatment based on the race of victim
has been proven; is that correct?
A I can't find the sentence.
0 I'm sorry. "The charge levelled by Baldus that
there is disparate treatment based on the race of the victim
is absurd." And then you go on to say, "The Georgia charging
and sentencing system does not tolerate higher levels of aggrg
vation in black victim cases as alleged by Baldus." Is that
what your findings demonstrate?
A It probably should be expanded to indicate, does
«
a
%
@
od
3
z
w
z
Zz
lo]
=
<
om
P
E
N
G
A
D
C
O
.
,
162
not tolerate higher levels of aggravations in black victim
cases in terms of sentencing.
Q All right. As modified, you stand by that conclusit
A Well, I don't see how this is a new conclusion that
Tiida
Q I'm not suggesting that it's new; I just want to
get your answer to that and then move to the next sentence if
we can. Is your answer yes or. —-
A I'll retract the next sentence.
Q Let's get your answer to the first one. Is the
answer to the first part yes? That appears to be a conclusioj
you've been pushing toward for --
A The black victim cases are not --
0 The Georgia charging system --
A Does not tolerate more aggravation in black victim
cases in terms of sentencing; right.
Q The sentencing patterns are a direct result of
aggravating and mitigating factors?
A That should be modified somewhat.
0 In what direction?
A The sentencing patterns can be explained by aggra-
vating and mitigating factors.
Q Now, in‘'effect, what you're offering there and in
earlier portions through the use of this method that you've
adopted is a rival hypothesis, is that correct; a rival
bn?
[3]
[74
™
Pv
-
z
B
A
Y
O
N
N
E
,
P
E
N
G
A
D
C
O
.
,
163
hypothesis that explains the phenomena or the data that's
before you.
A Well, a rival procedure to analyze the data.
Q But you've done more than simply indicate a pro-
cedure. You've suggested fairly strongly in the preliminary
report and now a little less strong that this is another
possible explanation; isn't that correct?
A The aggravating and mitigating factors --
Q That's right.
A -- are a possible explanation of the death sentenciy
patterns; yes.
Q So that in effect is a rival hypothesis to
Professor Baldus's?
A To his hypothesis.
Q What methods, if you know, are generally employed
in empirical studies to test rival hypothesis? Do you know
whether there are any conventional methods that are employed?
In other words, when you have two competing hypotheses about
an explanation or hypotheses which explain phenomenon or data
do you know of any normal. accepted procedures to assess those
rival hypotheses?
A Yes, you develop a statistical test that will test
the hypothesis.
Q what kind of test?
A There are many kinds of tests, depending on what
ng
X
lo]
L
o
ay
=z
In
z
z
Oo
>
<<
m
c
o
.
,
P
E
N
G
A
D
those hypotheses are.
0 Then are cross calculations and tabulations one
appropriate test?
A You mean what was represented here?
Q Of the rival hypothesis; is it appropriate to cross
tabulate as a statistical measure of which hypothesis is corre
A I don't think I'm following what you're trying to
ask.
Q Well, a social scientist, a statistician, presented
with rival hypothesis to explain a set of data would like to
eliminate one of the other of them since they are incommen-
surable.
A Okay; what are the two hypotheses?
Q Well, Professor Baldus's hypothesis is, as I under-—
stand it, that there is in effect -- one of his hypotheses
set forth is that there is in effect a dual system in
operation in the Georgia capital sentencing and
charging system. And your hypothesis is no. In fact what is
the case is that for some reason it's beyond the scope of
our data, that white victim cases are systematically more
aggravated. And that explains the apparent discrepancies
or discrimination in outcome.
How does one resolve that clash of hypotheses?
A Well, I've already tested his hypothesis by this
analysis.
doe
ct
I
a
[+]
hy
z
ui
z
z
3)
>
<<
m
P
E
N
G
A
D
C
O
.
,
165
0 So, in other words, the way one resolves it is by
doing tests which can test the hypotheses one after another
to see?
A Yes.
9) And you've rejected his hypothesis based upon this
test which you've set forward both for the Procedural Reform
Study and for the Charging and Sentencing Study?
A Correct.
Q Are you -- having rejected his hypothesis based on
the method that you've employed, do you suggest any method
for testing your own hypothesis, or do you think your own
hypothesis has been validated by your method?
A Having rejected his hypothesis, my hypothesis is
still a possible explanation. Now, I'm still interested in
the reasons why white victim cases are worse than black
victim cases, and to try and understand why that phenomenon
occurs.
Q Are you doing currently any runs or other analysis
of the data before you in order to assist you in determining
the answer to that question?
A Yes.
Q Do you have any runs at this point which would
reflect this analysis that you're doing?
A Yes. |
0 Can we have access to those?
3
z
ui
z
z
o
-
£4
mo
P
E
N
G
A
D
C
O
.
,
166
A You have access already.
Q All right. Can you tell us in fact what you're
doing or what kinds of methods or procedures of analysis
you're employing?
A Well, as a statistician, given the data and given
this problem that's been set forth in many states as to
alleged discrimination based on race of the victim, and it's
been alleged in many states that discrimination is occurring,
I have found that there is no support for that charge given
the Georgia Charging and Sentencing Study, given my results
in anlayzing the Procedural Reform Study.
So, there seems like there must be some explanation
as to why we see those sentencing patterns, why are white
victim cases worse than black victim cases. Now, I'm not
a sociologist; I'm a statistician, but I am interested in
this question, and I've done some analysis. And what I did,
which I think fairly strongly indicates what's going on, is
I broke down the cases, not just by race of the victim, but
by defendant victim racial combination. So, I look at blacks
killing whites, which is the first one; whites killing whitesy 7
blacks killing blacks; and whites killing blacks. And I ran
all these variables broken down by those categories. And
I was trying to see if I could get any additional information
from that categorization. And I need my results to talk
further on it.
x
Zz
ul
z
z
lo]
>
<
[+
P
E
N
G
A
D
C
O
.
,
10
1
167
Q We will be happy to let you adjourn to the other
end of the room get those results.
A Okay. This is, I pulled out the stack labeled B3.
Now, I think what might be valuable here is if I give you
some of the variables I think are significant in explaining
why I believe each defendant victim racial combination really
represents a different set of crimes, a different set of
murders. I think the motive is usually different and the
circumstances are different
For example, for the variable armed robbery, when
blacks kill whites, armed robbery occurs 67.1% of the time.
When whites kill whites, it occurs 20.2% of the time. When
blacks kill blacks, it occurs 6.6% of the time, and when
whites kill blacks it occurs 22.2% of the time.
om
If we look at the variable CReue which 1s contem]
oraneous offense at the time of the murder pretty much, then ¥
blacks kill whites, that occurs 79.7% of the time; when
whites kill whites it occurs 32.7% of the time; when blacks
kill blacks it's 12.3% of the time; and when whites kill
blacks it occurs 29.6% of the time.
Now, that indicates to me just at the start that
these are different kinds. of crimes. When blacks kill
whites, it's generally as part of a contemporaneous offense.
And since there's a high proportion of instances of armed
robbery, ‘the motive is money.
D=
vhen
=
z
wi
z
z
Oo
=
<<
a
P
E
N
G
A
D
C
O
.
,
168
Now, why then white victim cases, I suggest that
generally in this society white victims -- excuse me -- white
people are more affluent than black people. So, if one is
going out to obtain money, they would pick on. a white person
rather than a black person. And I think that partially
explains why we have that high percentage.
Furthermore, when blacks kill whites, it's surprisil
to expect that to be the result of a precipitated event such
as a family dispute or a lover's quarrel, even though'I
believe there are instances of that. Generally, those kinds
of motives are not present when a black kills a white. .When
a white kills a white, as indicated, 20.2% of the time it was
due to armed robbery. Again, I believe this represents two
possibilities. One is a set where the motive is money, and
so they're picking a white person to rob.
However, when a white kills a white, I think
there's two possibilities. One is the family friend type of
dispute; the other is a money motive or a felony motive.
However, when blacks kill blacks, it's a different
crime; 6.6% armed robbery, 12.3% contemporaneous offense.
Those are very low. So, the event that preciditates these
murders ‘do not seem. to be exclusively from a contemporaneous
offense.
Ahd when whites. kill blacks, I have really no idea
what is being represented here, because I only have 27 or 28
Rg
~
i
=
ul
z
z
o
>
<<
m
P
E
N
G
A
D
C
O
.
,
24
25
cases of that occurring out of a total of 1,082 that were
presented to me. This is a rare occurrence, and I have no
real feel for what seems to be the overriding consideration.
Some other variables that I believe are of importancs
here in explaining the type of crimes and the sentencing
that would result, I believe that part of the thing that's
considered by prosecutors that is important is whether the
victim was killed unnecessarily during an armed robbery. And
: bin VWITNESS
there's a variable that represents this. This is a V-—witness
and that variable is the variable that the motive for the
killing was to silence a witness, whether for contemporary
Or contemporaneous Crime Or some previous crime.
Well, when blacks kill whites, 36.4% of the time
that is at least one of the motives. When whites kill whites
it's only 13.2% of the time. When blacks kill blacks, it's
only 3.1% of the time. And when whites kill blacks, it's
12.5% of the time. So again, when blacks kill whites, it's
more aggravated.
Now, if we look at the mitigated side, such as
precipitating events --
Q Let me interrupt you for just a second. I guess
I understood the assumption on which your analysis of the
armed robbery, aggravating circumstance; differences, to be
explainable was ‘that in our society whites are more affluent,
and therefore, during an armed robbery whites are more likely
UV
y
y
3
Zz
wi
z
z
Oo
bod
Ld
=)
P
E
N
G
A
D
C
O
.
,
170
targets, and therefore, armed robberies against whites are
more frequent percentage wise --
A Yes.
Q -—- or murders are more frequently accompanied by
armed robberies. Would that --
A When you have a black killing a white.
Q A black killing a white. Would that economic
assumption, as it were, translate into a lot of the other
categories or a lot of the other variables, aggravating
variables?
A Well, I don't believe that that's a complete
explanation of why white victim cases are more aggravated.
That's, I believe, part of it, but from my testing I haven't
found it to completely account for the additional aggravating
and less mitigating circumstances that we find in white
victim cases compared to black victim cases. However, what
one really has to remember is that when we talk about black
victim cases, we're talking about blacks killing blacks.
There are very few cases where whites kill blacks.
So, looking through that population, the reason tha
‘they look so mitigated is because their proportion of armed
robberies, felony murders, and I can read you some of the
precipitating events proportions which will be very, very
high. So, they result from different circumstances than when
blacks kill whites or whites kill whites.
\
o
-
z
ui
z
Z
Oo
>
Ld
m
P
E
N
G
A
D
C
O
.
,
171
So, I think, you know, you have to keep that in
mind when you try and investigate race of the defendant,
because when you have a black defendant, it depends, really,
whether he killed a white victim or a black victim in terms
of what the circumstances of the crime were.
And so what you have 1s you have very bad cases
where blacks kill whites and very mitigated cases where
blacks kill blacks and they sort of even out. And I think
overall if we look at race of the defendant and compare all
these variables, we find that white defendants are slightly
more aggravated -- well, it wouldn't support anything con-
cerning black defendants being systematically more aggravated
and less mitigated than white defendants.
So, what happens is those blacks kill black cases
really dilute the bad blacks killing whites.
Q Okay. That's work, you say, that's underway now.
Are you continuing with those anaysis of the data in this
direction?
A Well, I don't know if I will get an opportunity to
go into this area indepth, but I've started it.
Q Have you generated any other runs that assist you
in this kind of analysis?
A Yes. I've looked at, just very briefly, comparing
high status and low status victims. And essentially, the
explanation seems to be high status victims are in the proces; Ul
©
x
[
of
3
Zz
ul
z
Z
o
>
<
o
P
E
N
G
A
D
C
O
.
,
172
of being robbed when they're murdered or some contemporaneous
offense. Low status victims are not in the process of being
robbed when they're murdered. It's a result of some precipitat
event where there's 1 of mitigaging factors.
But then, I haven't done much analysis on that.
That's. very sketchy; I haven't pursued that angle at all.
Q Okay. Let me ask you just a couple of technical
questions about your work if I can. Let's look at Table X,
which is Professor Baldus's table that's found on Page 31 of
Petitioner's Exhibit 2. Professor Baldus's own table did not
include Z values; did it?
A I don't believe it did. Excuse me, Table X?
0 Table X.
A On Page 317?
0) That's right.
A I do not believe so. I think he starred significan]
differences.
Q How did you perform a Z test for these aggravating
factors?
A I tested, I just computed the proportion difference
which is .50 minus .07. .Then I divided by the square root
of the average proportion, which would be 16 divided by 41,
because that's the best estimate of what the true proportion
is, given that we're assuming that there's no difference
times
between those cases, f£ime 1 minus that, times 1 over 26 plus
ing
4
3
z
ui
z
Zz
o
>
<
m
P
E
N
G
A
D
C
O
.
,
173
1 over 15, which is the standard formula used to compare
the differences in proportions, under the assumption that
the proportions are the same.
Q Would you mind repeating the very last portion?
A l over 26 plus 1 over 15.
PROFESSOR BALDUS: Times ?
THE WITNESS: Plus.
PROFESSOR BALDUS: thank you.
BY MR. BOGER:
Q Now, let's look at Table XXVIII on Page 47. Excuse
me; that's Table XVIII on Page 47. That's also, I think, a
table of Professor Baldus. Did you compute the Z values the
same way in that table?
A Yes.
Q And Table 37; which is another table of Dr. Baldus's
I'll have to search to find exactly where that is. It's on
Page 100.
A Yess
|®
) Did you compute the Z scores the same way?
A Yess
0 All ‘right. « You mentioned a little earlier inform-
ation that you received from Nicholas Dumick --
A Dumich.
0 Dumich; that's D-u-m-i-c-h -- on individual cases --
you have and written information from him on cases, or was thi
did
174
“ 1 | all orally transmitted?
2 A The usual procedure was just to transmit this
3 | information orally.
4 0 And do you have written record of the information
5 | that you received from him?
6 A No, I don't; except for one document in which
7 | cases in which more than one penalty trial had occurred were
g | tabulated by members of the office.
9 Q Is that readily available?
10 A I think that's in my office.
1 | 0 Could we get it, perhaps, next Tuesday if not this
12 | afternoon, or next Wednesday, I'm sorry, in Syracuse? I
13 | just would like to have your count on that --
14 MS. WESTMORELAND: Is that in some of the notes you
15 | made yesterday?
16 THE WITNESS: No, I don't think so.
17 MR. BOGER: Would that be a problem to produce?
: 18 THE WITNESS: I don't believe so.
19 | BY MR. BOGER:
20 Q All right; thank you. Do you have knowledge of any
errors that were discovered in any of the cases or observations
B
A
Y
O
N
N
E
,
N.
J.
#
7
8
3
2
-
fe
R
No
—
—
p
22 | that Professor Baldus had in either the Procedural Reform
23 | Study or the Charging and Sentencing Study that you located
P
E
N
G
A
D
Co
O.
,
24 | based on this oral information from Nicholas Dumich?
25 A Yes. Well, 'I have information but .I can't be sure
B
A
Y
O
N
N
E
,
N.
J.
2.
00
2
No
m
d
P
E
N
G
A
D
C
O
.
,
24
25
175
that there are errors.
Q In other words, information that there are mis-
matches between the oral information of Mr. Dumich and the
data of Professor Baldus?
A Yes; but I haven't read the file and I don't know
whether in fact Professor Baldus is correct or Mr. Dumich
is correct. l-went --
Q Did you make any list of those places or instances
in which there was possible mismatch or error?
A You mean resulting from my discussions with Mr.
Dumich?
0 That's correct; or otherwise.
A Well, one case that I can't be certain about, but
that might be what you're looking for, is we investigated
Professor Baldus's data on the McCleskey case to see what
was coded there. And Mr. Dumich felt that it wasn't coded
properly, it wasn't indicating all the aggravating circumstance
I think the case we were considering or the study that we were
considering was the Procedural Reform Study. And he had indi;
cated to me that McCleskey had ambushed the policeman, that he
had hid behind some furniture, and as the policeman walked
down the store, he just jumped up and shot him twice. And he
felt that what should have been included for llcCleskey in ‘that,
in the aggravating circumstances should be ambush, unnecessary
and multiple shots.
176
pA 1 Q This is based on his knowledge of the case?
2 A This is what he told me. So, I can't vouch for
3 | any of that making any sense or being true. That's simply
4 the information that he transmitted to me.
5 Q Do you have any other instances in which he gave
6 | you similar kinds of information or suggested erroneous
7 | coding or mismatching?
8 A There was another instance where the Isaacs, I
believe it was Coleman, where he indicated that there were
10 | many more aggravating circumstances than were indicated in
11 | the Procedural Reform Study.
12 Q Did any other members of the Attorney General's
13 | Office give you oral information or written information to
14 | suggest other errors in the coding process?
15 A No, they have not provided them to me; no.
16 Q Have they provided them to anyone connected with
17 | this study?
18 MS. WESTMORELAND: I don't think he knows the
: 19 | answer to that.
20 MR. BOGER: I'm only asking you for information
to which you know the answer.
BA
YO
NN
E,
N.
J.
27
¢
No
—
24 THE WITNESS: I believe that is pretty much the
P
E
N
G
A
D
C
O
.
,
23 | extent to what I've been privileged to in that area in
24 | noticing that there were, according to Mr. Dumich, some
25 | differences in what he would have coded for the aggravating
177
% 1 circumstances compared to what was actually coded.
2 BY MR. BOGER:
3 0 When you talk about being privileged to, I'm not
4 | exactly sure what you mean.
5 A I don't have access to any analysis that's been
6 | done in that area.
7 Q To your knowledge, there may have been such analysis
8 | done?
9 A To my knowledge, I believe there has been some
10 | analysis done.
11 Q Does it employ any of the data that you've had
12 | access to, Professor Baldus's data?
13 A Yes; I've provided them with the data.
* 14 Q To whom have you made that available?
15 A To Mr. Dumich and the others.
16 Q And he is in the Attorney General's Office?
17 A Yes.
18 Q Okay. You indicated that you had done some race
19 | of defendant analysis since receiving Professor Baldus's
20 | working draft dated June 15, I think, 1983. What's the nature
of those race of defendant ‘analyses?
BA
YO
NN
E,
N.
J.
N
—
22 A Pretty much what I described to you in terms of
P
E
N
G
A
D
C
O
.
,
defendant victim facial combination. Plus, I also ran a
PE
No
Ww
24 | comparison, a computer printout that you have in your
26 | possession, comparing white defendant cases and black defendant
178
1 cases.
2 0 Was the sentencing outcome a dependent variable
3 | at this point?
4 A No, this is similar to the information that I used
5 | to gather, to construct the tables.
6 0 I see.
7 A So, I ran them for all the variables defined in
8 | the earlier table for the Charging and Sentencing Study.
9 Q Do you have, have you formulated any additional
10 | research plans with respect to Professor Baldus's working
11 | draft to prepare you for the evidentiary hearing in August?
12 | A Well, I first had to be present at Professor
13 | Baldus's deposition to find out precisely what he's done
14 | so that I can then proceed to analyze it. I'm waiting for
15 | the deposition to analyze it.
16 Q So, until after you know what he says in his depo-
17 | sition, you're not in a position to say what you are going
18 | to do or what analysis you may undertake to prepare for the
19 | hearing?
5 20 A Correct.
} 21 0 Even with ‘receipt of the working draft?
22 A Well, I would like the opportunity to find out
P
E
N
G
A
D
C
O
.
,
23 | what all that means, and we will do that when we take his
24 | deposition. And that will make it easier for me to formulate
25 | possible experiments to test his conclusions.
179
® 1 Q Have you formulated any experiments that you
2 | intend to carry out to this point?
3 A I don't know if they're going to -- I think it all
4 | depends on how the deposition goes and what happens and
5 | whether any of these kinds of experiments might be necessary.
6 MR. BOGER: Let me note for the record, Ms.
7 | Westmoreland, that the reasons the depositions were initially
8 | scheduled in the order they were or initially suggested in
9 | the order that they were was precisely in order to give
10 | Professor Baldus an opportunity to hear what criticisms
11 | in this study may be developed by an expert. It's for that
12 | reason that when Mr. Dumich was given the opportunity to
13 | depose Professor Baldus last November, which was about eight
14 | months ago, that we told the court we would not depose any
15 | experts the State came up with until after Professsor Baldus
16 | had first been deposed; which we anticipated now what this
17 | witness has told us, that he has to hear Professor Baldus's
18 | deposition before he can fully prepare his criticisms.
: 19 With that in mind, and with the observation that
20 | we first knew who your experts were on June 15th, whereas
you knew our expert as early as early as November 15th, we
B
A
Y
O
N
N
E
,
N.
J.
87
0
No
m
h
22 | may well apply to the Court for a further deposition of
P
E
N
G
A
D
C
O
.
,
23 | Dr. Katz under the rationale which we think is reasonable,
24 | that both sides should be prepared for a hearing of this
25 | sort, and that the witness who is propounding a theory and
u
f
a
-
=
2
ui
z
z
Oo
>
Ed
o
P
E
N
G
A
D
C
O
.
,
180
propounding a report needs to hear that criticism in order
to take account of it in his preparations for the hearing.
I just wanted to alert you to that. I know that
you were not the Attorney General on the case back in
November or December or January and so forth. I think there
were some reasons, I suspect, why there was some delay in
taking Professor Baldus's deposition, but you know we're
facing a July llth cut-off at this point of the discovery
period, and I simply want to advise you that we may seek a
deposition after the end of that period based upon what
Professor Katz has told us. I still have some additional
questions, but --
MS. WESTMORELAND: Just in relation to that,
if I could just make one comment. Part of, I think, the
problem is the latest working draft; and I don't believe we
received it until June 17th -- it was either the 15th or
17th. And I think Dr. Katz is contemplating reviewing it
somewhat more thoroughly before our final determination has
been made. I think that is my understanding of the remaining
work that needs: to be done, at least as far as that. I'll
also note that I have not been Counsel of Record for that
greater length of time, and I also do believe there was
a lapse of time, at which point we weren't sure whether
discovery was available, while the initial discovery period
had expired and we were waiting to hear if the discovery
a
9
bs
4
z
ul
z=
z
o
>
<
a
P
E
N
G
A
D
C
O
.
,
181
period was being extended.
MR. BOGER:: That's ‘all correct.
MS. WESTMORELAND: That's just for the record.
MR. BOGER: Again, as we indicated, I think in a
letter to you, Professor Baldus's preliminary report was
sent to you in effect gratis because it really had never
been requested by Mr. Dumich in his discovery; you had asked
only for a final report. And as I'm sure Dr. Katz knows,
one doesn't develop a final report until the last moment
because there are so many changes and corrections and
additions. And yet we thought that it was appropriate when
we deposed your expert, Dr. Katz, to have Professor Baldus's
latest thoughts before, and which is why we didn't forward
the material at that time.
Now, let me ask just a few additional questions,
Dr. Katz, and this long deposition may come to an end. What
works would you cite as authorities, reputable authorities,
in the field of statistics; what texts?
THE WITNESS: Concerning what aspects?
BY MR. BOGER:
Q Well, let's ask multivariate analysis?
A Neter = and Wasserman.
Q What's the title of that? If you would mind
spelling those, actually, for the Court Reporter, we'd all
be, probably, better off.
182
: A N-e-t-e-r, and Wasserman, W-a-s-s-e-r-m-a-n; I
y believe it's "lLinerar Models" or something of that form.
3 Q Are there other authorities you'd recognize in
4 | that field?
5 A Yes.
6 Q Can you give me some names?
y A Not at the moment.
5 Q The names don't occur to you at the moment?
° A Right.
0 Q All right. Do some text names. occur to. you even
% if the authors' names don't?
oy A Well, I don't believe this is a point that I'm
03 going to raise objections to if an author cites some valid
Ft oF formula and conclusion.
16 Q I'm just trying to find out who 'it is you go to
6 when you go to sources?
: 37 A I go to Neter and Wasserman.
18 Q All right. Any texts in the. area of research
10 design or methodology for empirical data gathering?
20 A I would go to my.Kish.book on Sampling.
: > Q Would you spell that for us?
5 A K~-1-s~h.
: 53 Q And the title of that is "Sampling"?
be A I believe so.
25 Q Any other texts?
A
-
z
[]
z
z
Oo
>
Ld
o
P
E
N
G
A
D
C
O
.
,
183
A No.
0 Kish covers data collection methods as well. What
about questionnaire design?
A I don't know right offhand.
Q Okay. And finally, something that always gets
asked to every expert in every case that I've ever seen,
are you being compensated by the State of Georgia for your
work?
A Yes.
Q At what rate?
MS. WESTMORELAND: I don't really think that's
relevant at this proceeding at this time. We will stipulate
that Dr. Katz is under contract to the State of Georgia at
this time and has been.
MR. BOGER: Okay, the objection as to relevance
is noted, and if you will answer the question?
MS. WESTMORELAND: We still direct Dr. Katz not
to answer the question as to the rate of his compensation. I
think that that's something that you're entitled to at this s:
MR. BOGER: We would disagree on that. I suspect
we'll not go to the Judge over it, but of course he will be
asked this question on cross examination. And it does go to
possible interest or bias, and I suspect he'll probably have |
answer it. But as a statistician who remembers every number
that has ever come up, I suspect you'll still be able to do if
don't
tage.
4
z
wl
z
z
Qo
>
<<
om
P
E
N
G
A
D
C
O
.
,
184
questions
note that
Thank you for your patience, Dr. Katz. Will there
on cross-examination?
MS. WESTMORELAND: No, I don't believe so.
MR. BOGER: This deposition is concluded, then.
MS. WESTMORELAND: And we do want the record to
Dr. Katz does not waive reading and signing.
FURTHER DEPONENT SAITH NOT.
-
F
O
R
M
7
4
0
0
7
0
0
2
B
A
Y
O
N
N
E
,
N.
J.
P
E
N
G
A
D
C
O
.
,
24
25
|
|
i
|
i
|
|
|
1
|
i
185
VOLUME II
SU BM I:S 'S.10.N
I, DR. JOSEPH L. KATZ, certify that I have read or have
had read to me the foregoing deposition, and that I have sub-
mitted 23 changes in form and substance to the officer
before whom this deposition was taken, to be included by her
in the deposition.
F 5 JA
/4 2 4
THIS [3 day of _ 19k , 1983.
PET
NZ
P A / | ICTR / Z
1) [JOIN XK
DR. JOSEPH L. KATZ we
Signed, sealed and delivered
in the presence of:
~ /
Sy” ; ~~
“ ol Zz pd ’ SR 4 pa £ FB
<s 5
Notary Publik
wie, (Georgia Steip af * »7=
WARREN MCCLESKEY V. WALTER ZANT
Volume II
Page Line
9 15
9 17
9 23
16 15
28 10
36 16
36 17
37 16
37 20
45 2
66 1
93 2
122 22
123 3
127 14
128 15
No. C81-2434A
Corrections for Deposition
of Dr. Joseph L. Katz
As 1n
Deposition
"Pulasky"
"Provess"
hdi you”
"race V-D"
nol"
"Leech"
"Alfa"
"For consideration"
"Defender"
"tourture"
"mode of"
"aren't analyzed"
"minus of predicting"
"deconfess"
" +O"
Corrected
Pulaski
Process
did you
RACEVD
0, 1
Lerch
Leach
Alpha
From
consideration
Offender
torture
motive
are analyzed
minus the
predicted
DECONFESS
0,1
Line
25
As 1n
Deposition
"life victim"
"srongly"
"THELMUR"
"V witness"
"loss of"
“iime:1l"
Corrected
white victim
strongly
FELMUR
VWITNESS
lots of
times 1
186
1 CoE. RTI FP.I CATE
® 2 || STATE OF GEORGIA
3 COUNTY OF DEKALB
4 I, Earlene P. Stewart, a Certified Court Reporter and
5 || Notary Public in and for the State of Georgia at large, do
6 || hereby certify as follows:
7 THAT I correctly reported by the stenomask verbatim
g || method the foregoing deposition of DR. JOSEPH L. KATZ (Volume IT)
g |i at the date and place stated in the caption hereof;
10 THAT the witness was duly sworn and examined by counsel;
THAT my tape recorded notes were later reduced to type-
writing under my personal supervision and control, and that
12
3 the doregoing transcript contains the full, true and correct
& 14 testimony of said witness on said occasion;
75 THAT I am neither of kin nor counsel to any parties in-
18 volved in the matter, nor in any manner interested in the
iy result thereof;
- THAT the reading and signing of the deposition was not
1 waived by counsel and the witness;
20 THAT all changes submitted by the witness have been
o included in the deposition;
23 THIS «18th day of July, 1983.
23
24 y
EARLENE P. STEWART
» CERTIFIED COURT REPORTER
XXX = O
Ra CR JP
yee Bla = | Lh 14
cells
Ww a= Blac 4
|
|
|
|
—
—
—
—
—
pi
jad Eatl: :
T
r
e
| FE SELTEICE